The Neuropathologic Basis for Translational Biomarker Development in the Macaque Model of Late-Onset Alzheimer’s Disease

Abstract

Background:

Accurate placement of the macaque within the Alzheimer’s disease (AD) research framework is essential to discover early-stage predictive biomarkers.

Objective:

To assess utility of the aging macaque in advancing translational biomarker development for preclinical AD, we evaluated relative signal strength of comparable neuropathologic phenomena in macaques and patients.

Methods:

We compared pathology in patient and macaque formalin-fixed paraffin embedded (FFPE) tissues using identical criteria. We quantified expression of amyloid-β, pTau, and inflammatory and senescence markers across species. Distribution of AD-relevant markers were compared in FFPE and perfused frozen macaque brain to assess expression of labile proteins that could inform in-life fluid biomarkers.

Results:

Amyloid-β pathology in macaques closely approximated patient pathology. Complex plaque composition in macaques implied significant disruption of synaptic connectivity. In FFPE tissue, pretangle pTau immunoreactivity placed the macaque in Braak Stage 1b. In perfused frozen tissue, soluble pTau distribution approximated Braak Stage III-IV. In macaque, ß-amyloid, pTau and acetylcholinesterase labeling co-localized to AD-vulnerable circuits. Significant association of glial fibrillary acidic protein with amyloid-β occurred in humans only. The senescence marker p16 correlated positively with pTau expression and negatively with amyloid-β in patients only. Macaques lacked neuropathologic co-morbidities.

Conclusions:

AD-relevant neuropathologic signals in the macaque support biomarker discovery in the areas of amyloid-β plaque evolution and associated synaptic disruption as well as early-stage Tau phosphorylation. Relative protection from accumulation of senescence markers, fibrillar tau and neuropathologic co-morbidities in macaque implicate species difference in rates of biological brain aging. We provide over 4000 digital slides for further study.

Introduction

Interest in non-human primates (NHPs) as models of Alzheimer’s disease (AD) has been enduring. ^1–7 This appeal stems from their evolutionary proximity to humans, ⁸ transcriptomic and neuroanatomical similarity of circuits most affected by AD, ^8–10 and development of spontaneous cerebral amyloidosis with age. ¹¹ In macaques, declining age-related cognitive function ¹² is correlated with synaptic loss in prefrontal cortex. ¹³ The principal argument against NHPs as fully accurate models of AD are their failure to spontaneously develop significant neurofibrillary tangle (NFT) pathology. ¹⁴ Additionally, while hippocampal spine density declines in aged animals, ¹⁵ neuronal numbers are preserved. ¹⁶ These disparities are resolved by proposing the aged NHP as a model of preclinical AD. ^{1, 8}

Accurate placement of the macaque within the Alzheimer’s disease (AD) research framework ¹⁷ is essential to discover early-stage predictive biomarkers. Cognitively normal adults experience heterogenous accumulation of amyloid (A), tau (T) and neurodegeneration (N) biomarkers ¹⁸ that may begin in middle age ¹⁹ and allow staging of preclinical AD ^{17, 20} and eventual AD risk. ²¹ Because early intervention has the greatest potential for delaying clinical trajectories of AD, significant effort is directed at developing predictive biomarkers for prodromal AD. Utility and interpretation of comparable biomarkers in macaques remains an underdeveloped area, particularly for imaging biomarkers. ^22–25 In part, this is because it is unclear precisely where to place NHPs in the preclinical AD spectrum, and which aspects of age-related neuropathologic change could be captured to develop predictive biomarkers.

To assess the robustness of neuropathologic signals underpinning AD biomarker discovery in macaques, we compared expression of comparable pathologic phenomena underlying core and non-core AD biomarkers used for diagnosis and staging of Alzheimer’s disease. ¹⁷ Using this framework, we quantified amyloid-β and pTau expression underpinning core amyloid (A) and tau (T) biomarkers in macaque and human brains. We similarly compared expression of the inflammatory (I) marker glial fibrillary acidic protein (GFAP), with additional quantification of the microglial marker ionized calcium-binding adapter molecule (Iba1), and the senescence marker p16. Lastly, we assessed evidence for non-AD specific neuronal injury (N), vascular (V) and synucleinopathy (S) processes. We assessed AD-neuropathologic scores in macaques and humans according to National Institute of Aging-Alzheimers Association (NIA-AA) Consensus Guidelines for the Neuropathologic Assessment of AD. ²⁶ These comparative studies were performed using identical methods on formalin fixed paraffin embedded (FFPE) macaque and human brains. Additionally, we explored whether brainstem pTau expression described in young humans ²⁷ occurred in younger macaques. Lastly, we compared expression of neuropathologic entities underlying AD biomarkers in perfused frozen macaque tissue to assess expression of labile proteins that may be lost due to post-mortem interval, immersion fixation and FFPE processing, ^{3, 28} but would inform in-life fluid biomarker signals. Our raw image data are freely accessible (MacBrain: https://medicine.yale.edu/neuroscience/macbrain/) for further comparative studies.

Materials and Methods

Human and animal subjects

Human tissue was assessed from decedents enrolled through the Yale Alzheimer Disease Research Center. Non-identifiable human biologic specimens are exempt from Institutional Review Board approval. No associated clinical data from patients or animals were included in the study. No experimental procedures were performed in macaques. Tissue for this study were obtained from animals submitted for diagnostic necropsy or for perfusion for unrelated conditions. Macaques were maintained under protocols approved by the Yale University Animal Care and Use Committee and their care conformed to the Guide for the Care and Use of Laboratory Animals of the Office of Laboratory Animal Welfare, National Institutes of Health. Euthanasia was performed in accordance with approved IACUC protocols according to the American Veterinary Medical Association Guidelines for the Euthanasia of Animals (https://olaw.nih.gov/avma-guidelines-2020.htm). Macaque tissue exhibited a female sex bias (sourced from a colony populated in part by retired females from a breeder colony in the same facility), and human tissue a male sex bias as a function of available patients and animals.

Human tissue, immersion fixation and formalin-fixed paraffin embedded (FFPE) preparations.

The brain from five AD patients with increasing severity of AD neuropathologic change (Table 1) was immersion fixed in 10% neutral buffered formalin (NBF) for 3 weeks. The right hemisphere was used for all NHP and patient studies utilizing FFPE material. Brain tissue samples (Supplementary Table 1) were embedded in paraffin blocks, sectioned at 5μm, mounted on charged slides and stained with hematoxylin and eosin. Immunoperoxidase stains were performed at the Yale University Department of Pathology using antibodies listed in Supplementary Table 2, using routine methods (Supplementary Methods).

Table 1:

Animal and patient samples

Macaques
Macaque species (ID #)	ABC score	Sex	Age (years)	PM interval	Fixation & processing
1. M. mulatta (B88)	A0 B0 C0	Male	10.2	0 hr	PFA perfusion, frozen
2. M. mulatta (22–363)	A0 B0 C0	Female	21.7	<1 hr	Immersion, FFPE
3. M. mulatta (20–210)	A0 B0 C0	Female	22.4	<1 hr	Immersion, FFPE
4. M. mulatta (B101)	A0 B0 C0	Male	23.8	0 hr	PFA perfusion, frozen
5. M. mulatta (19–3441)	A0 B0 C0	Female	25.1	<1 hr	Immersion, FFPE
6. M. mulatta (17–1925)	A1 B0 C1	Female	25.4	<1 hr	Immersion, FFPE
7 . M. mulatta (13–14795)	A2 B0 C1	Female	26.6	<1 hr	Immersion, FFPE
8 . M. mulatta (20–283)	A2 B0 C2	Female	28.0	<1 hr	Immersion, FFPE
9. M. mulatta (19–1568)	A2 B0 C1	Female	28.1	<1 hr	Immersion, FFPE
10 . M. mulatta (B95)	A2 B0 C2	Female	32.4	0 hr	PFA perfusion, frozen
11 .M. arctoides(19–1802)	A2 B0 C2	Female	37.7	<1 hr	Immersion, FFPE
Patients
Patient #	ABC score	Sex	Age (years)	PM interval	Fixation protocol
1	A2 B1 C1	Male	>80	7 hrs	Immersion, FFPE
2	A3 B2 C1	Male	70–75	8 hrs	Immersion, FFPE
3	A3 B3 C2	Female	>80	16 hrs	Immersion, FFPE
4	A3 B3 C3	Male	>80	12 hrs	Immersion, FFPE
5	A3 B3 C3	Male	75–80	9 hrs	Immersion, FFPE

FFPE: Formalin-fixed paraffin embedded

PFA: paraformaldehyde

A: A score (derived from Thal phase) describing distribution of amyloid.

B: B score (Braak) describing anatomical progression of neurofibrillary pathology

C: C score (CERAD) for density of neuritic amyloid plaques

Macaque tissue, immersion fixation and FFPE.

Animals used for FFPE tissue evaluation (n=8; Table 1) were submitted to the Diagnostic Pathology Service in the Department of Comparative Medicine following sedation with 10mg/kg intramuscular (IM) ketamine and intravenous (IV) euthanasia (1ml/lb Euthasol Bridgeton, MO) on humane grounds for various unrelated health issues. To assess progression of ageing pathology, including brainstem pTau pathology, ²⁷ we examined a spectrum middle-aged to elderly rhesus macaques (21.7–28.1 years). We also included descriptive pathology of a very old stump-tailed macaque (37.7 years), but did not include this animal in any statistical analyses. Brains were immersed in 10% NBF for 2–4 weeks prior to coronal sectioning of the right hemisphere to yield comparable brain regions (Supplementary Table 1) needed for neuropathologic assessment of AD in patients. ²⁶ Tissues underwent routine FFPE and sectioning at 5μm in the Comparative Medicine Histology Core, followed by mounting on positive-charged slides, staining with hematoxylin and eosin, and Perl’s iron method staining for hemosiderin. Immunoperoxidase stains were performed at the Yale University Department of Pathology using the same antibodies and methods as described for human FFPE tissue (Supplementary Table 2). These macaque brains were used for neuropathologic assessment and quantification of amyloid-ß, AT8, GFAP, Iba1 and p16 load (see Methods below) using identical methods to similar assessment of human brains.

Macaque tissue, perfusion fixation and frozen sections.

Because detection of labile post-translationally modified proteins such as p-Tau may be reduced in immersion fixed FFPE tissue, ^{3, 28} we repeated immunohistochemical analyses in perfused frozen macaque brains. After sedation with 10 mg/kg ketamine IM, followed by IV catheterization and induction of a deep anesthetic state with sodium pentobarbital (40 mg/kg), three rhesus macaques from 10.2–32.4 years (B88, B101 and B95) were transcardially perfused with ~500 ml of phosphate-buffered saline (PBS, 0.9%) followed by ~3,000 ml of 4% PFA in PBS. All animal procedures were approved by the Yale Institutional Animal Care and Use Committee. After overnight postfixation at 4°C in 4% PFA in PBS, the hemispheres were separated at the midsagittal plane and immersed in 20% and then 30% sucrose (in PBS) until they sank. After sucrose cryoprotection, one brain hemisphere was sent to FD NeuroTechnologies (Columbia, MD) via overnight priority shipping at room temperature (RT), wrapped in gauze (non-woven 4” × 4” sponges), and sealed in a 250-ml Nalgene jar filled with cold 20–30% sucrose in PBS. Brain hemispheres were cut in the coronal plane at 50 μm on a freezing microtome and sections stored in freezing solution (25% glycerol and 30% ethylene glycol by volume in PBS, or FD tissue cryoprotection solution^™ until further processing). Sequential sections were stained with Nissl stain, reacted histochemically for acetylcholinesterase ²⁹ and immunohistochemistry performed on 50μm thick free-floating coronal sections (Supplementary Methods) using the antibodies listed in Supplementary Table 2. Validation of antibodies is illustrated in Supplementary Figure 1. Links to galleries illustrating validation procedures are given in Supplementary Table 3. In two animals, we utilized the right hemisphere (B88 and B95). The left hemisphere in B101 was used due to a nick in the right hemisphere sustained during brain removal. B101 and B88 were not used for quantification. B95 was used only for relative quantification of markers within the same brain.

Neuropathologic assessment.

We assessed pathology in FFPE tissue from eight animals and five AD patients according to the NIA/AA Consensus Guidelines for the Neuropathologic Assessment of AD. ²⁶ Neuropathologic assessment in AD patients was performed by a board-certified neuropathologist (AH); similar evaluation of macaque tissue was conducted by a board-certified veterinary neuropathologist (CZ). Data were recorded using the Coding Guidebook for the National Alzheimer’s Coordinating Center (NACC) Neuropathology Data form V11 (Supplementary Dataset 1). In macaques, Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) score ³⁰ was based on density of all plaques as diffuse and neuritic plaques are not clearly distinguishable in this species.

Slide scanning.

A total of 3014 FFPE and frozen sections of macaque and human tissues were scanned at constant illumination with a 20× objective, using an Aperio CS2 HR scanner (Leica Corp). Digitized images were loaded and organized into Leica’s Aperio eSlide Manager. Links to human and macaque galleries in MacBrain (https://medicine.yale.edu/neuroscience/macbrain/) are given in Supplementary Table 3.

Quantification of amyloid-ß, AT8, GFAP, Iba1 and p16 load.

We quantified expression of these markers across nine homologous regions in FFPE sections from eight animals and five AD patients (Supplementary Dataset 2; Supplementary Table 1). Total amyloid-β load (6E10 immunopositive dense and diffuse plaques, as well as vascular amyloid) was quantified across species in FFPE material. Color images were taken at 5X magnification using Leica’s Aperio eSlide Manager and imported into Image J ³¹. The relevant elements were isolated using the color threshold function ³², and after setting the correct scale in Image J using the embedded scale bar. Threshholded elements were quantified using the Analyze Particles function and data for each anatomical region compiled from three independent 500 μm2 areas (Supplementary Dataset 2).

Statistical analyses:

We explored relationships between expression of amyloid-ß, Tau, and neuro-inflammatory and senescence markers in FFPE tissue from macaques and patients. Staining for Iba1, GFAP, and p16 (in %) were used to model staining for amyloid-ß or pTau (AT8) (in %) using a linear mixed effect model. A compound symmetry covariance structure was used to account for intra-subject correlations of repeated observations across brain regions. Human and macaque data was modeled separately. When assessing FFPE macaque tissue, an animal with epilepsy-associated hippocampal sclerosis (21.7 years), and the 37.7 year-old M. arctoides were excluded. To explore anatomical colocalization of amyloid-ß, pTau and AchE, we utilized frozen sections from the perfused 32.4-year-old female rhesus macaque, in which the entire brain had been sectioned coronally. We quantified amyloid-ß, AT8 and AchE expression in 19 coronal sections spanning prefrontal cortex to cerebellum and encompassing 101 anatomical regions as described above (Supplementary Dataset 3). Using the percentage staining of each, we calculated the Pearson’s Correlation Coefficient (N-2) where N = number of data points.

Abbreviations.

Abbreviations according to Paxinos ³³ are given in Supplementary Dataset 4.

Results

Macroscopic features.

Progressive medial temporal atrophy and hippocampal atrophy accompanied increasing AD neuropathologic stage in AD patients. In contrast, very mild hippocampal atrophy with accompanying mild lateral ventricular dilation was evident only in the oldest animal (Supplementary Figure 2).

Amyloid-β deposition.

We assessed pathology in formalin-fixed paraffin embedded (FFPE) tissue from eight macaques and five AD patients (Table 1) according to the NIA/AA Consensus Guidelines for the Neuropathologic Assessment of AD. ²⁶We confirmed the known observation ¹¹ of increasing amyloid-β plaque deposition with age in rhesus macaques. As noted previously ¹¹, plaques were rare in animals under 25 years of age. In the amygdala, amyloid-β load in macaques was comparable to patients (Figure 1). However, in cortical regions macaques experienced lower plaque load (2–3% coverage with C scores of 0–2) than patients (8–9% coverage with C scores of 1–3). Macaques did not experience amyloid-β deposition in midbrain, substantia nigra or cerebellum, thus achieving lower Thal stages (0–2) than patients (Thal stage 1–3). Cerebral amyloid angiopathy (CAA) was evident in the oldest animals, although CAA scores were consistently higher in patients.

Figure 1:

Amyloid-β pathology in AD patients and macaques

a. Amyloid-β plaque load (% area) in FFPE tissue of AD patients (blue) and aging macaques (red). Plaque load in patients increased with increasing ABC score, particularly in cortical regions. Plaque load increased with age in macaques and was negligible under 26 years.

b, c: Plaque morphology in FFPE tissue. A range of diffuse to dense core (arrows, b) plaques were evident in patient tissue. In macaques (c), most plaques were diffuse.

d-m: Amyloid-β plaque content in perfused frozen tissues from a 32.5 year-old rhesus. Constituents comprised phosphorylated Tau AT8 (d), pT217 (e), SMI-32 (f), GFAP (g), Iba1 (h), AchE (i), TH (j), SV2A (k), Nissl (l) and Prussian blue for iron (m). Not all plaques demonstrated all of these constituents.

Patient locations: MFG: middle frontal gyrus; S/MTG: superior and middle temporal gyrus; AC: anterior cingulate gyrus; ERC: entorhinal cortex; O: occipital cortex (BA17, 18). Macaque locations: 46d: dorsolateral prefrontal cortex area 46d; ST2: superior temporal gyrus area ST2; 24: anterior cingulate gyrus; Amyg: amygdala; ERC: entorhinal cortex; P: parietal cortex; O: occipital cortex (BA17, 18).

AchE: acetylcholinesterase, GFAP: glial fibrillary acidic protein, Iba1: ionized calcium-binding adapter molecule, SV2A: synaptic vesicle glycoprotein 2A, SMI-32: non-phosphorylated form of the neurofilament heavy protein, TH: tyrosine hydroxylase.

Bar = 50 μm (b, c); 20 μm (d-m).

As noted in other studies ¹¹, amyloid-β plaques in macaques typically assumed diffuse morphology, distinct from the dense-core appearance noted in patient plaques (Figure 1). However, many amyloid-β plaques contained dystrophic neurites immunoreactive for AT8 and thus meet the definition of neuritic plaques (defined as amyloid plaques with pTau positive dystrophic neurites with or without dense cores). ²⁶ Additionally, local neuronal loss, astro-and microgliosis, iron deposition, and altered expression of a wide variety of other antigens (acetylcholinesterase, tyrosine hydroxylase, SV2A and SMI-32) were evident within subsets of amyloid-β plaques (Figure 1).

pTau expression.

Defined hierarchical progression of AT8 immunoreactive neurofibrillary tangles (NFTs) through limbic and cortical structures using FFPE material is used to assign Braak stage. ^{26, 34, 35} This approach was applied to FFPE tissue from five AD patients and eight macaques (Supplementary Dataset 1). Additionally, AT8 immunoreactivity (aggregate of plaque associated dystrophic neurites, neuropil threads and neurofibrillary tangles) was quantified in FFPE tissue across species (Supplementary Dataset 2). In FFPE macaque tissue, granular cytoplasmic neuronal pTau immunoreactivity occurring within morphologically normal neurons, consistent with pretangles ³⁴ was evident in isolated neurons within locus coeruleus, substantia nigra, hippocampus (CA1 region and subiculum) and isolated prefrontal cortical neurons of the oldest animal (Figure 2 and Supplementary Figure 3). However, the major proportion of pTau immunoreactivity was amyloid-β associated, with staining for AT8, pTauT217 and pTauT181 evident in dystrophic neurites within amyloid-β plaques (Figure 2). This pattern was also evident in perfused frozen tissue from the 32.4 year old animal (Figure 1). Mature neuropil threads (NTs) and neurofibrillary tangles (NFTs), defined as tightly packed fibers in neurons with a dislocated or shrunken nucleus, ³⁴ were absent. Subcortical Braak Stages a-c ²⁷ could be assigned on the basis of pretangle pTau staining in the locus coeruleus and substantia nigra (Figure 2). Braak Stage 1a-b was assigned based on pretangles in hippocampal neurons and processes, however this did not rise to Braak NFT stages 1 and 2 ³⁵ due to absent NFT morphology (Supplementary Dataset 1). While isolated AT8 positive neurons were evident in prefrontal cortex of the oldest macaque (Supplementary Figure 2), this did not accompany the pattern of cortical progression of NFTs that constitutes the basis for higher Braak scoring in AD patients. ³⁵ Therefore, we assessed an overall Braak score of B0 in FFPE material from all macaques (Table 1). We did not observe patterns consistent with a diagnosis of age-related tau astrogliopathy (ARTAG) in macaques.

Figure 2:

Immunostaining against pTau epitopes AT8, pT217 and pT181in FFPE macaque tissue.

a-d: Pretangle pTauT217 and AT8 immunoreactivity within individual neurons (arrow) in locus coeruleus (LC) and substantia nigra (SN) in FFPE (37.7 year old M. arctoides)

e-j: Immunostaining in FFPE tissue from a 37.7 year old M. arctoides against pTau epitopes AT8, pT217 and pT181. Within the subiculum (e, f, g), similar pretangle morphology in neurons was noted with all three antibodies. Within amyloid-β plaques, pTau positive dystrophic neurites were evident with all p-Tau antibodies. Double immunohistochemistry for amyloid-β and pT217 is illustrated in i.

Bar = 50 μm (b, d, e, g, i); 20 μm (a-d, h-j).

FFPE: formalin-fixed paraffin embedded

Quantifiable AT8 was detected only in patients (Supplementary Figure 3). Patients with low Braak scores (B1) exhibited dense NTs and mature NFTs in presubiculum and subiculum. With increasing Braak scores, these progressed through CA1 hippocampal region, entorhinal cortex, and progressive cortical regions in highest scoring individuals (B3). In these patients, masses of AT8 immunoreactive neurons were arrayed upon a dense mat of AT8 positive NTs to occupy 20–25% of the area assessed (Supplementary Figure 3).

As previously noted, ^{3, 28} pTau detection in tissue may be reduced by the combined effects of post-mortem interval, immersion fixation and paraffin processing. We evaluated rostro-caudal distribution of AT8 expression in the brain in frozen sections of three perfused animals (10.2 years, male, 23.8 years, male and 32.4 years, female; Supplementary Figure 4). Diffuse AT8 immunoreactivity was evident in neurons and processes of AD-vulnerable cortical, limbic and subcortical regions of the elder macaque corresponding to the distribution of Tau pathology characterizing Braak III-IV staging in human FFPE material. ³⁶ Intense AT8 staining was also noted in ventral forebrain and ventral pallidum, locus coeruleus and dorsal motor nucleus of the vagus nerve. Both younger animals experienced AT8 immunoreactivity in the dorsal motor nucleus of the vagus nerve, as noted in young humans, ²⁷ with increasing expression in basal forebrain and subiculum in the middle-aged animal. Similar spatial patterns of immunoreactivity were evident for pTauT217, however its staining within animal was relatively weaker than AT8 at comparable primary antibody concentrations (galleries in Supplementary Table 3). Because pTauT181 does not accumulate significantly in CSF of aging cynomolgus and rhesus, ^37–39 it was not extended to perfused whole brain analysis.

Spatial relationship of acetylcholinesterase, amyloid-β and AT8 expression.

Using frozen coronal sections, we quantified amyloid-β, AchE and AT8 expression across 106 regions (Supplementary Dataset 3 and Figure 3) in the 32.4 year old animal (neither amyloid-β or AT8 expression were sufficient to quantify in the 23.8 year old animal). AT8 quantification included amyloid-β plaque-associated immunoreactivity (evident in both FFPE and frozen tissues) and diffuse immunoreactivity evident in neuronal cell bodies and processes (evident in perfused frozen tissue only). Regions with the highest combined expression of all three markers are illustrated in Supplementary Figure 5 and Figure 3. Expression of AT8 and AchE was most highly correlated (r= 0.56; p < .00001), with slightly weaker associations between AchE and amyloid-β (r= 0.45; p < .00001) and AT8 and amyloid-β (r= 0.47; p < .00001; Supplementary Dataset 3), thus implicating a particular vulnerability of cortical and limbic cholinergic circuitry to accumulation of pTau and fibrillar amyloid-β. The majority of the subcortical neuraxis was densely reactive for AchE, and was not utilized for Pearson’s correlation as these regions did not contain amyloid-β.

Figure 3:

Spatial relationship of acetylcholinesterase, amyloid-β and AT8 expression in perfused frozen tissue from a 32.4 year old rhesus.

a-g: AchE histochemistry, anterior to posterior coronal sections. AchE staining was evident in cortical regions, basal forebrain structures, amygdala and hippocampus that were also reactive for AT8. Some subcortical regions (basal ganglia, thalamus, midbrain) were only reactive for AchE.

h-n: AT8 immunohistochemistry, similar sections. Regions of AT8 and AchE co-expression included anterior cingulate cortex (areas 32, 24 and 25:a, h), orbitofrontal cortex (b, i), basal forebrain (c, j), amygdala and temporal cortex with emphasis on perirhinal and parahippocampal regions (c, d, j, k), entorhinal cortex and subiculum (e, f, l, m).

o-p: Higher magnification of AchE, amyloid-βand AT8 staining using area 32 as an example. AchE was evident with processes and within amyloid-βplaques (arrows, o). amyloid-β plaques (arrows, p) contained AT8 positive dystrophic neuritis (arrows, q). Diffuse AT8 immunoreactivity within cortical neuronal processes was evident (asterisk, q).

Bregma locations in mm are illustrated above.

Bar = 1500 μm (a, h); 1600 μm (b, i); 2700 μm (c-g, j-m); 100 μm (o-q), 20 μm (inset, q).

Neuroinflammation and senescence.

We compared grey matter astrocytic (GFAP), microglial (Iba1) and p16 load in FFPE tissue from five AD patients and eight macaques in the same regions used to assess amyloid-β and AT8 load (Supplementary Dataset 2, Figure 4 and Supplementary Figure 6). We used a linear mixed effect model to explore relationships between expression of amyloid-ß, Tau, and neuro-inflammatory and senescence markers in both species. Human and macaque data was modeled separately. Two macaques (the animal with epilepsy-associated hippocampal sclerosis and the 37.7 year-old M. arctoides) were excluded from analysis. Consistent with histologically evident accumulation of astrocytes within amyloid-β plaques, ⁴⁰ the percentage of GFAP staining was significantly associated with the percentage of amyloid-β staining in patients (estimate = 1.16, standard error = 0.18, pval<0.0001). While a similar positive correlation between these two entities was noted in macaques (estimate = 0.15, standard error = 0.16), this relationship did not achieve statistical significance (pval = 0.34). In both species, amyloid-β plaques harbored Iba1 positive microglia and cytoplasmic p16 expression in cells with morphologic features consistent with microglia (Figure 1 and Figure 4). However, Iba1 load was not significantly correlated with amyloid-β load in patients (estimate = 0.48, standard error = 0.44, pval = 0.28) or macaques (estimate = 0.06, standard error = 0.14, pval = 0.77), implying a dominant role of astrocytes in amyloid-β plaque development. In patients, but not macaques, microgliosis and p16 expression was prominent in AT8 immunopositive regions harboring NFTs (Figure 4). In corresponding regions in macaques, p16 expression could be found in cells consistent with glia, but this was rare. In macaques, both AT8 and p16 were so scarce that they were not quantifiable at 5x magnification. In our patient cohort, p16 expression was positively correlated with pTauAT8 load (estimate = 6.61, standard error = 1.99, pval = 0.002), and negatively correlated with amyloid-β load (estimate = −1.26, standard error = 0.46, pval = 0.01).

Figure 4:

p16 immunoreactivity in FFPE tissue from patients and macaques

a: p16 load (% area) in FFPE tissue of AD patients (blue). P16 load in patients was variable with respect to AD neuropathologic score. It was not quantifiable at 5x in macaques.

b-g: Middle frontal gyrus, AD patient (b-d) and Area 46d, 28.1 year old macaque (e-g). Iba1 (arrows, c, f) and P16 (arrows, d, g) labeling is evident within amyloid-β plaques (b, e).

h-k: Subiculum, AD patient (h-j) and 28.1 year old macaque. Abundant AT8, Iba1 and p16 immunoreactivity is evident within patient subiculum. Morphology of p16 positive cells is consistent with microglia (inset, j). In contrast, p16 positive glia were rare in the macaque subiculum.

Bar = 100 μm (h-j); 50 μm (b, e); 20 μm (c, d, f, g, k)

Neuropathologic comorbidities.

Vascular pathology (atherosclerosis, arteriosclerosis, prior infarcts) or white matter rarefaction was evident in all patients (Supplementary Dataset 1) but was rare in macaques. Hippocampal sclerosis was evident in four of five patients. Apart from the youngest macaque with epilepsy in which hippocampal sclerosis was marked, remaining animals did not exhibit this lesion. Iron load in substantia nigra, internal and external globus pallidus, ventral pallidum and dentate nucleus of the cerebellum increased with age in macaques and was evident in both frozen (Supplementary Figure 7) and FFPE tissue. This lesion is well-described in humans ⁴¹ and macaques. ⁴² In macaques, increasing punctate iron with age was evident within entorhinal cortex (Supplementary Figure 7), and, as noted in patients, ⁴³ within amyloid-β plaques (Figure 1). Lewy body pathology with nigral neuronal loss occurred in one patient. Consistent with prior reports in macaques ⁴⁴ aged animals experience modest neuronal loss within substantia nigra (Supplementary Figure 8). We identified increased dendritic diameter of remaining neurons, accompanied by more intense tyrosine hydroxylase immunoreactivity, as noted in the locus coeruleus of aged rats. ⁴⁵

Discussion

Revised criteria for diagnosis and staging of Alzheimer’s disease ¹⁷ define AD as a biological process that can be diagnosed on appearance of an abnormal Core 1 biomarker(s) in asymptomatic individuals. Biomarkers are divided into three broad categories, i) Core AD biomarkers: ß-amyloid (A) and tau (T), ii) non-specific biomarkers important to, but not specific to AD: neuronal injury (N) and inflammation (I), and iii) biomarkers of common AD neuropathologic comorbidities: vascular injury (V) and synucleinopathy (S). In this framework ^{17, 46}, appearance of Core AD biomarkers in cognitively intact individuals can be used to assess risk of progression to cognitive decline accompanied by AD neuropathologic change. This preclinical period is relevant when considering AD-relevant neuropathology and potential for attendant biomarker abnormalities in the aging macaque.

Amyloid-ß (A):

As noted in other NHPs, we identified age-associated accumulation of amyloid-ß plaques in macaques. ^{4, 7, 47–54} In patients, plaque deposition is reflected by declining CSF and blood Aβ42/Aβ40 ratio ⁵⁵ and a positive amyloid PET signal ⁵⁶. In NHPs, amyloid-ß plaque accumulation is also reflected by declining Aβ42/Aβ40 ratios in CSF ^{37, 38, 57, 58} and serum. ⁵⁹ However, to date, attempts at detecting the cardinal Core 1 imaging biomarker (amyloid-ß PET) have yielded inconsistent results in macaques. ^{22, 23, 25, 60} To assess whether plaque load may account for this discrepancy, we quantified amyloid-ß load in macaques and patients. In patients, correlation of 6E10 amyloid-ß immunohistochemical load (% area) at postmortem with in vivo PiB retention (distribution volume ratio; DVR) in the same patient indicate that amyloid-ß loads of 0.5–4% correlate with PiB DVRs of 1.5–1.8. ⁶¹ This extent of PiB uptake exceeds the optimal cutoff (PiB DVR of 1.14) associated with future cognitive decline. ⁶² Using neuropathology, we and others ¹¹ demonstrate that amyloid load sufficient to be detectable by PET (0.5–3%) accumulates after 25 years in macaques (Figure 1). Amyloid PET studies have used animals that range from maximum ages of 19.7 years, ²² 22.6 years, ²⁵ and 20–24 years. ²⁴ Therefore, amyloid-ß loads are likely to have been below the level of detection in these studies. Nevertheless, correlation between 11C-PiB uptake and amyloid-ß immunoreactivity was demonstrated in a study older animals.²⁴ In our study, animals with 0.5–3% amyloid-ß load achieved moderate Thal scores (2) and mild to moderate CERAD scores (1–2). Patients with moderate to high Thal (A2–3) and CERAD (C2–3) scores are reliably distinguished from those with low (A0–1, C0–1) scores using 11C-PiB scans ⁵⁶ or flutemetamol ⁶³. In patients, dense-core plaques generate approximately three times the signal of diffuse plaques, however both dense-core and diffuse plaques (which prevail in macaques) are detected by either 11C-PiB and flutemetamol ^{63, 64}. Therefore, assuming similar tracer binding capacity ⁶⁰ and amyloid plaque composition ⁵⁴ across species, and providing animals are sufficiently aged, it appears feasible that amyloid-ß plaque-bearing animals in the C1–2 range could be distinguished from amyloid-ß negative animals. Very high amyloid plaque burden (7–10% plaque load) reported in cortex of cognitively impaired AD patients ⁶⁴ was similarly noted in our patients with highest Thal and CERAD scores (A3C3), but was not evident in our or other ¹¹ aged macaques.

We identified neuritic plaques ²⁶ quite readily in old macaques. Many larger plaques contained dystrophic neurites immunoreactive for various antigens (AT8, pT217, pT181, AchE and TH). In frozen tissue, we identified localized depletion of some proteins (SV2A, labile cytoplasmic AT8, pT217 staining) and neurons within plaques, as well as accumulation of iron, long considered contributory to AD pathogenesis. ⁶⁵ Our data suggest that there is significant disruption of neural connectivity in regions with high ß-amyloid load. Consequently, therapeutic removal of amyloid-ß in established AD patients is likely to leave an imprint of neuritic injury that may limit cognitive recovery.

Tau (T):

We assigned Braak scores to macaques and AD patients in FFPE tissue using identical criteria for AD neuropathologic change. ²⁶ Pretangle pTau immunoreactivity in locus coeruleus and substantia nigra allowed staging at subcortical pretangle Braak a-c ²⁷ and replicates findings in young humans. ⁶⁶ Pretangle pTau pathology in hippocampus and entorhinal cortex placed the macaque in Braak Stage 1a-b. No animal achieved an NFT Braak score >0 in FFPE tissue. However, the distribution of pTau labeling in perfused frozen tissue approximated Braak Stage III-IV, thus identifying limbic and association cortex destined to develop selective injury as a cumulative phospho-epitope landscape develops with age. ⁶⁷ Further, biochemical and ultrastructural studies demonstrate increasing insoluble, fibrillated tau in the entorhinal cortex of aging macaques ³ supporting the position of rhesus within a translationally relevant trajectory of Tau pathogenesis. We readily identified pTau and AchE within amyloid-β plaques in macaques, consistent with the enduring roles of the cholinergic and Tau hypotheses. The highest correlation occurred between AchE and AT8. Notably, these markers colocalized in the basal forebrain cholinergic system, amygdala, hippocampus and medial temporal regions. These regions demonstrate high connectivity in rhesus studies ⁶⁸ and are particularly susceptible to pretangle Tau phosphorylation in humans. ⁶⁹

Neuropathologic Braak staging can be recapitulated in vivo using tau PET ligands. ⁷⁰ Inability to detect AT8 positive non-argyrophilic pretangles is a consistent feature of current Tau PET tracers. ^{34, 71, 72} Therefore, predominant pretangle Tau pathology in aged NHP brains is consistent with negative Tau PET findings to date. ^{25, 73} Fluid pTau biomarker profiles in patients precede a positive Tau PET signal. ⁷⁴ pTau profiles in cognitively normal and AD brains ^{28, 67, 75} are quite similar, but relatively amplified in AD patients. However, some emerging pTau markers (e.g. pT217, pT231, pT181 and pT205) may have discriminatory potential. ^{46, 76–81} Using FFPE tissue, we identified AT8, pT217 and pT181 immunoreactivity that was most robust within ß-amyloid plaques and manifested as pretangle morphology in rare hippocampal neurons. In cognitively intact patients, elevated fluid biomarkers pT217 and pT231 are more closely associated amyloid-ß-PET positivity than PET-detectable tangles. ^{55, 76–78} Similarly, in simultaneous mass spectrometric analysis of patient CSF, pT217, pT231 and pT205 levels are specific for positive amyloid status in cognitively impaired individuals. ⁷⁹ In later disease stages these also inform on tau tangle accumulation. ^{78, 79} Some Tau phosphorylation sites exhibit age-related fluid biomarker increases in rhesus macaques (pS202/T205, pT217, pT181, pS235, pS356, pS396 and pS214) ^{39, 82–85} and cynomolgus macaque (pS202/T205, pT231),⁸⁶ and in light of human data, may inform on Core 1 biomarker events rather than Core 2 pathology (characterized by positive medial temporal Tau-PET). ¹⁷). As noted in macaques, ^{82, 83, 85} longitudinal studies in middle-aged cognitively intact patients identify increasing pTau with age. ^{87, 88} Because macaques express pretangle Tau pathology, association of the established relationship between fluid pTau and Tau PET in patients ^{46, 74} cannot be assessed in this species. Rather, age-related fluid biomarker changes in macaques reflect earlier stages of amyloid and pretangle pathology noted in older animals and in cognitively normal middle-aged humans, ²⁷ prior to appearance of abnormal PET biomarkers.

Neurodegeneration (N):

Our gross pathology confirms previously noted morphologic ⁸⁹ and MRI ^{90, 91} observations that significant hippocampal atrophy does not occur with age in macaques. PET signal for the synaptic vesicle glycoprotein 2 A (SV2A) tracer [11C]UCB-J in prefrontal cortex and hippocampus of aged macaques correlates with working memory. ²⁵ However, significant synaptic loss similar to that detected by [11C]UCB-J PET in AD patients ⁹² is not evident, consistent with lack of significant neuronal loss in aging macaques. ¹⁶ Elevated neurofilament light (Nfl) levels identify preclinical AD patients. ⁵⁵ No CSF increase in Nfl have been noted associated with age in macaques ³⁷ or in amyloid-β oligomer-injected rhesus brain. ⁹³

Inflammation (I):

Plasma GFAP is increased in preclinical AD. ⁵⁵ Our study identified significant correlation between GFAP and amyloid-ß load in FFPE sections from patients, but not macaques. As both species accumulate astrocytes within amyloid-ß plaques, ^{40, 94} we suspect that this may be due to relatively greater accumulation of amyloid-ß across all of our patients, compared to relatively lower amyloid load evident only in the older subset of our macaques. This may have resulted in insufficient power to detect a significant relationship between GFAP and amyloid-ß in macaques. In our patients, GFAP load was significantly associated with amyloid-ß load but not pTau load. This preferential association is reflected in PET imaging ⁹⁵ and blood biomarker studies. ⁹⁶ The neuroinflammation PET marker (11C-PBR28) is predictive of longitudinal cognitive decline in patients, ⁹⁷ and has been successfully employed in macaques to detect lipopolysaccharide induced neuroinflammation. ⁹⁸ Our data indicate that similar relationships could potentially be captured using imaging ⁹⁸ and blood biomarkers ⁹⁹ in macaques, as long as they are sufficiently aged (>25 years).

The accumulation of senescent cells is a major driver of biological aging underlying the accumulation of age-related morbidities, ¹⁰⁰ including AD.¹⁰¹ Increased glial cytoplasmic and nuclear expression of p16 is evident in brains of AD patients, ¹⁰² and clearance of p16-expressing microglia improves cognitive function and brain immune aging in mice. ¹⁰³ In our patients, p16 expression was significantly positively associated with AT8 load, but negatively associated with amyloid-ß load. Transcriptomic studies in astrocytes ¹⁰⁴ and microglia identify a spectrum of active to senescent proteomic profiles depending on regional pathology and location, ¹⁰⁵ thus complicating extraction of predictive data using microglial activation biomarkers. ^{106, 107} Notably, in macaque FFPE tissue, both AT8 and p16 expression were comparably scarce, suggesting that macaques enjoy relative protection from NFT accumulation and immune senescence by life’s end compared to humans.

Synucleinopathy (S) and Vascular injury (V):

We did not identify marked nigral neuronal loss or synucleinopathy in macaques, although we did demonstrate pretangle Braak stage a-c pTau labeling in nigral neurons similar to those noted in young humans. ⁶⁶ The entanglement between diabetes, hypertension, vascular dementia, and AD is well-recognized in humans ^{108, 109} and may underlie ethnoracial differences in occurrence of neuropathologic comorbidities. ¹¹⁰ Although arteriosclerosis is a commonly histologic lesion, spontaneous hypertension in aging macaques is rare, and more commonly noted in the context of metabolic syndrome. ¹¹¹ Type II diabetes is common comorbidity of macaques, and is associated with increased amyloid-β plaque load and systemic pro-inflammatory biomarkers. ¹¹²

Conclusions:

We, and others, ^{1, 8} have proposed the aging macaque as a model of preclinical AD¹¹³ i.e. the clinically asymptomatic state harboring the underlying AD pathophysiologic disease process. This underlying disease is defined by characteristic AD neuropathologic change, from which arise biomarker abnormalities that permit detection of individuals at risk for developing dementia. ^{17, 113} Our data place the aging macaque within the early preclinical spectrum of AD, specifically within Core 1 biomarker territory. ¹⁷ Positive amyloid PET is sufficient to diagnose biological stage A, whereas progressive detection of Tau by PET is required for diagnosis of subsequent stages (B-D). In humans, Core 1 (amyloid and Tau) fluid biomarkers can establish that an individual is in biological stage A or higher, but cannot currently discriminate between stages A–D.¹⁷ Our comparative amyloid-β data indicate that PET imaging of macaques older than 25 years is indicated for in vivo detection of amyloid-β plaques. Our comparative pTau data indicate that current Tau PET technologies are unlikely to be useful at any age in macaques as these fail to develop significant fibrillar Tau necessary for PET tracer binding. ^{71, 72, 114} Nevertheless, our study in macaque FFPE tissue is the first to assign Subcortical Braak Stages a-c ²⁷ to macaque brain on the basis of pretangle pTau staining in the locus coeruleus and substantia nigra. Additionally we could assign Braak Stage 1a-b on the basis of pretangles in hippocampal neurons and processes. These stages occur in young to middle aged cognitively normal humans ^{27 66} and are considered the earliest manifestations of aged-related Tau evolution. Due to its complex neuritic plaque composition, the macaque may be a robust model to study retention or restoration of synaptic connectivity following removal of ß-amyloid. Accumulation of amyloid-ß plaques in macaques ^{11, 115–117} are accompanied by altered fluid amyloid biomarkers ^{37, 58, 59, 118} and age-related increased fluid levels of pTau ^{59, 83} species. Consequently, the strength of the aged macaque model lies in the area of preclinical fluid biomarker discovery. Notably, subcortical pTau characteristic of pretangle Braak stage a-c is evident, and age-related accumulation of prefibrillar pTau in AD-vulnerable regions in perfused brains recapitulates the pattern of NFT Braak III-IV staging. This finding implies that similar changes may occur in middle-aged humans - these cannot be detected on FFPE pathology, but are likely to inform biomarker studies in early preclinical AD. Our NHP data reveal pTau accumulation in regions that also express AchE and amyloid-ß, allowing study of early amyloid-tau interactions in AD-vulnerable cholinergic circuits. Compared to humans, while macaques can develop NFTs, their extent of fibrillar Tau accumulation is much more restricted, even when very old ⁵⁴. Macaques are also spared from neuropathologic comorbidities, and experience relative protection from expression of the senescence marker p16. The pathogenetic relationship of these findings remains unexplored, but suggests that differences in genetics, lifespan and lifetime exposures ¹¹⁹ may drive differential brain biologic aging ¹²⁰ between macaques and humans.

We recognize several limitations of our study. Although our study is unique in that it directly compares and quantifies various AD-associated neuropathologic markers across patients and aged animals using identical methods, our sample size is low. Our p16 data support a pathogenetic relationship between immune senescence and progression of NFT accumulation and imply that humans and macaques may experience differential rates of biological aging in the brain by life’s end. This hypothesis requires further study. We do not report behavioral data in macaques, or clinical data in humans. However, our intent is to directly compare the strength and distribution of AD-relevant neuropathologic signals across species, and assess expression of these in perfused tissue reflective of in-life processes. Macaques present a singular opportunity to discover early predictive fluid biomarkers and correlate these with in-vivo spatial expression in perfused brain. These findings may be extended to patient tissue to assess predictive biomarker potential within the physiologic to pathologic aging continuum.

Data Availability:

Our raw image data are freely accessible (MacBrain: https://medicine.yale.edu/neuroscience/macbrain/) for further comparative studies. Links to human and macaque galleries in MacBrain (https://medicine.yale.edu/neuroscience/macbrain/) are given in Supplementary Table 3.