犬脑阿尔茨海默病的神经病理学特征及其与生物体液生物标志物和认知功能的关系。 机翻

Introduction

Alzheimer’s disease (AD), a progressive neurodegenerative biological construct characterized by progressive cognitive decline and its defining neuropathologic features, collectively termed Alzheimer’s disease neuropathologic change (ADNC), which includes amyloid-β (Aβ) plaques and tau neurofibrillary tangles (NFTs). Although anti-amyloid immunotherapies are now approved for early-stage AD, their clinical use remains limited to selected patients and they do not stop or reverse disease progression, underscoring the urgent need for innovative research approaches1. Progress in the field has been hindered by limitations in transgenic rodent models, which, despite providing mechanistic insights, often rely on genetic modifications and fail to fully recapitulate the spontaneous, age-related, and multifactorial nature of human AD pathology2. These limitations highlight the need for alternative animal models that better capture the complex interplay of genetic, environmental, and aging factors in AD.

The companion dog represents a distinctive and naturally occurring model of AD, exhibiting age-dependent cognitive decline and neuropathological features that overlap with key elements of human ADNC, including age-associated brain atrophy and the accumulation of Aβ plaques and phosphorylated tau deposits that partially recapitulate early tau changes in human AD3–6. In contrast to traditional rodent models, dogs share human environments, lifestyles, and daily exposures, providing a natural and translationally relevant context for studying the development and progression of AD-like pathology, but on an accelerated timescale. On average, a domestic dog lives approximately 10–12 years compared to an average human lifespan of 77 years in some populations7,8.

Importantly, dogs experience age-associated cognitive decline, termed canine cognitive dysfunction (CCD) syndrome, which typically affects dogs over eight years old and is associated with signs of dementia that are similar to those that occur in AD, including disorientation, altered social interactions, and functional decline9. Recent studies from the Dog Aging Project (dogagingproject.org) have advanced our understanding of CCD epidemiology by assessing signs of dementia in over 15,000 dogs, revealing strong associations between CCD and factors such as age10 and physical activity11. Pathologically, Aβ accumulates progressively with age in dogs, both as extracellular plaques and soluble oligomers, mirroring AD-like pathology with changes in amyloid precursor protein (APP) processing that favors amyloidogenic cleavage12. Tau pathology has generally been reported to be limited and less frequent in canines, particularly when assessed using the AT8 marker. Recently, however, studies leveraging improved phospho-tau–specific reagents demonstrate regionally restricted tau pathology, particularly within limbic structures3. Together, these observations suggest that dogs develop both clinical and neuropathological features relevant to AD. Despite this convergence, a critical gap remains: no rigorous studies have directly linked age-associated cognitive decline in dogs to the extent and regional distribution of AD-like neuropathology. This disconnect is due in part to the absence of a systematic, standardized framework for staging canine Aβ and tau pathology that is comparable to frameworks established for human AD. However, human AD staging criteria should be applied to dogs with caution, because the extent to which canine brain regions, tau spread, and Aβ/tau processing match the human disease remains incompletely understood. Therefore, these frameworks should be interpreted as comparative tools rather than direct human-equivalent staging systems.

To assess clinical behavioral changes in dogs, validated behavioral assessment tools such as the Canine Dementia Scale (CADES) and Cognitive Dysfunction Rating (CCDR) scale are used to measure cognitive deficits in aged dogs and show striking similarities to human AD scoring systems like the Clinical Dementia Rating (CDR)11,13–15. Through evaluation of domains such as disorientation, social interaction, and daily functioning, these tools have revealed progressive cognitive decline in aging dogs, positioning them as a translational bridge between rodent models and human clinical studies16,17. However, without corresponding standardized neuropathologic staging, the relationship between these clinical measures and underlying brain pathology remains poorly defined.

Although behavioral changes in aging dogs have been well described, comprehensive neuropathological mapping aligned with human AD frameworks18 is lacking, hindering full validation of the canine model for AD. To address this gap, we applied a comparative neuropathologic assessment framework adapted from the NIA-AA guidelines18, incorporating regional evaluation of Aβ deposition, pTau181 pathology, and neuritic plaque pathology to characterize AD-like changes in client-owned aging dogs. We further integrated plasma biomarkers used in human AD diagnosis that are also altered in CCD to examine their relationships with neuropathology and cognitive function19–21. This study provides a comparative clinico-pathologic assessment of AD-like changes in aged dogs and evaluates the extent to which CCD neuropathology parallels, and diverges from, human AD.

Methods

Recruitment and clinical assessment

Dogs were selectively recruited through multiple channels: social media and local news platforms, referrals from a network of primary care veterinarians identifying cases of suspected CCD, and from the existing population of senior dogs evaluated by the Neurology Clinical Trials Service at the Colorado State University Veterinary Teaching Hospital, and from the Dog Aging Project’s Brain Health Study22. Informed consent was obtained from all pet owners prior to study enrollment. Clinical assessments included the CCDR Scale, a validated survey instrument that measures signs of CCD15. The inclusion criteria required absence of brain tumors or other neurological conditions that could affect cognition.

A cohort of the dogs was under the University of Washington IRB deemed that recruitment of dog owners for the Dog Aging Project, and the administration and content of the DAP Health and Life Experience Survey (HLES), are human subjects research that qualifies for Category 2 exempt status (IRB ID no. 5988, effective 10/30/2018). All study-related procedures involving privately owned dogs were approved by the Texas A&M University IACUC, under AUP 2021-0316 CAM (effective 12/14/2021). Details regarding age and cognitive scores of participating dogs are listed in Table 1. In this study, we used a modified classification threshold from refs. 1523, dogs with a CCDR score >40 were classified as CCD-positive (CCD+), and those with a score ≤40 were classified as CCD-negative (CCD−).

Collection of plasma

Whole blood was collected, placed into EDTA blood collection tubes and centrifuged to separate plasma. Samples were stored at −80 °C until used for biomarker assays (see below). Plasma biomarker measurements were only available for a subset (n = 14) of the neuropathology cohort due to sample availability.

Biofluid biomarker assay

The Quanterix SR-X platform was used to simultaneously measure Abeta 40 (Aβ40), Abeta 42 (Aβ42), Glial Fibrillary Acidic Protein (GFAP), and Neurofilament light chain (NF-L) in antemortem canine plasma samples (Simoa® Neurology 4-Plex E Advantage Kit). The Simoa® Human Neurology 4-Plex E Advantage Kit was developed for human EDTA plasma and CSF and, to our knowledge, has not been fully validated as a canine-specific assay. However, Simoa-based assays have been used previously to measure NfL and Aβ40/Aβ42 in canine plasma in aging and neurodegenerative disease studies. Therefore, we interpreted these biomarker data as translational measures rather than clinically validated canine diagnostic values20,21. Plasma samples were thawed, centrifuged, and diluted 4X in plasma diluent provided by the kit. Standards (in triplicate), and diluted samples (in duplicate) were loaded onto the plate and the assay performed according to the Simoa® Neuro 4-Plex Advantage Plus kit for SRX instructions. Briefly, antibody conjugated beads and secondary detection antibodies were added to the plate and incubated overnight. After plate washing, activated RGP reagent was added, and the plate was incubated for 10 min. After washes and open-air drying the plates were loaded onto the SRX. Quantification was performed using Quanterix SRX software version 1.2.0 using a 5-parameter sigmoidal curve generated from the standards. All sample aliquots were thawed only on the day of the assay, used once and discarded. Samples with responses above the dynamic range of the calibration curve (n = 14) were diluted 10X, rerun and reanalyzed.

Tissue collection and processing

Twenty-four canine brains were obtained postmortem after euthanasia and necropsy from privately owned dogs with owner consent. The brains were collected at the Veterinary Teaching Hospital at Colorado State University or the University of Washington and fixed in 10% neutral buffer formalin for at least 72 h. Tissue blocks containing pre-frontal cortex, visual cortex, striatum, hippocampus, midbrain and cerebellum were processed using a Leica TP1020 Automatic Benchtop Tissue Processor, embedded in paraffin wax (Cancer Diagnostics, Cat #: EEPAR56), and sectioned on a Thermo Scientific HM 325-2 Manual Microtome at 5 μm thickness and mounted on positively charged glass slides (Superfrost Plus, Cancer 232 Diagnostics, Cat #: 4951) for staining and analysis.

Immunohistochemistry

Immunohistochemical staining was performed to evaluate the pathologic hallmarks of ADNC. Aβ plaques were detected using Anti-β-Amyloid (1-16, 1:5000) on a Leica BondMax automated stainer with citrate antigen retrieval (PH 6.0). Tau hyperphosphorylation was assessed using phospho-Tau (Thr181) Monoclonal Antibody (Thermo Fisher MN1050, 1:200) with manual staining. Briefly, paraffin-embedded sections were dewaxed in xylene and rehydrated through graded ethanols before antigen retrieval in 0.01 M sodium citrate for 20 min at 95 °C. Endoperoxides were removed by incubating sections in a 0.30% hydrogen peroxide and water solution for 30 min. To block non-specific labeling, sections were treated with 10% horse serum in Tris-A with 2% BSA (bovine serum albumin) and 2% Triton-X in Tris buffered saline (TBS). Sections were then incubated overnight at 4 °C with primary antibodies diluted in Tris-A with 2% BSA. After washing with Tris-A with 2% BSA, sections were incubated for 1 h with their corresponding biotinylated secondary antibodies (horse anti-mouse, Vector Labs) diluted in 10% horse serum in Tris-A with 2% BSA at 1:250. Slides were counterstained with hematoxylin, rinsed in bluing reagent, dehydrated through graded ethanols followed by xylene, and mounted with mounting media. Coverslips were applied using #1 cover glass.

To further evaluate classical AD neuropathological features, modified Bielschowsky silver stain was performed manually24. Briefly, formalin fixed paraffin-embedded blocks were dewaxed in xylene, rehydrated through graded ethanols and incubated in fresh 20% silver nitrate for 20 min. Sections were then incubated in titrated ammoniacal silver solution and developed in a formalin-based developer. The reaction was monitored microscopically to optimize contrast. Slides were rinsed in distilled water extensively, treated with 2% sodium thiosulfate to remove unreduced silver, dehydrated through graded ethanols followed by xylene, and mounted with mounting media.

Histology assessment

Neuropathological evaluation was performed by adapting National Institute on Aging-Alzheimer’s Association guidelines for AD neuropathology18 to aging canines. All assessments were conducted blinded to clinical data, including animal age and cognitive scores. Histological assessment of Aβ deposits, pTau181-positive pathology, and neuritic plaques density was carried out on all dogs included in the study. An AD-like neuropathologic change (AD-like NC) classification framework for canine brains was based on: (A) regional Aβ deposition, (B) pTau181-positive pathology, and (C) neuritic plaque pathology. Brain sections analyzed for each assessment are shown in Supplementary Fig. 1.

Thal-inspired regional assessment

Regional Aβ deposition was assessed using brain regions selected on the basis of their relevance to the hierarchical distribution of amyloid pathology described in the Thal framework25. Regions evaluated included the middle frontal gyrus, primary visual cortex, transentorhinal/entorhinal cortex, hippocampus, neostriatum, midbrain, and cerebellum. For each region, tissue sections were sampled from comparable anatomical levels across all dogs. A brain region was considered positive for Aβ pathology if one or more Aβ-positive plaques were identified in the tissue examined. These assessments were interpreted as a comparative evaluation of regional Aβ distribution rather than a validated canine Thal staging system.

Semi-quantitative amyloid burden: Pathology assessment was performed using four categories: 0 = no amyloid pathology, 1 = amyloid pathology covers less than half the tissue, 2 = amyloid pathology covers approximately half the tissue, 3 = amyloid pathology covers the entire tissue.

Phosphorylated Tau Pathology: Braak-Inspired Regional Assessment. We assessed tau pathology for NFT-like lesions based on Braak staging criteria26 in transentorhinal/ entorhinal cortex, hippocampal formation (including CA1 and subiculum), middle frontal gyrus, and primary visual cortex. An area was considered positive for pathology if pTau181-immunoreactive neuronal cells showing perikaryal tau accumulation consistent with pretangle or early tau pathology were identified using an immunohistochemical stain for phospho-tau (Thr181). Morphological criteria included brown DAB-positive cytoplasmic/perikaryal staining within neuronal cell bodies, with extension into proximal neuronal processes, and a staining pattern distinguishable from diffuse background, nonspecific neuropil labeling, or isolated punctate debris. Because Braak staging was developed for human Alzheimer’s disease, this framework was applied comparatively and not as a fully validated canine staging system.

Neuritic plaque density (CERAD Criteria)

We assessed neuritic plaque density using the CERAD criteria27 by evaluating the prefrontal cortex with the Bielschowsky silver stain. A neuritic plaque is characterized by the presence of dystrophic neurites that are typically highlighted on silver stain. The density is assessed as the plaques per mm² per Mirra et al.

Digital Image Analysis

To quantify the burden of amyloid and phosphorylated tau (p-tau) pathology in the canine brain, histologic slides were digitized using an Aperio AT2 whole-slide scanner (Leica Biosystems; software version 102.0.7.5) equipped with a 20×/0.75 NA Plan Apo objective, producing images at 0.50 µm/pixel. Slides were scanned using standardized scanner settings to minimize image-to-image variability. Scanner calibration and routine quality control were performed according to the manufacturer’s procedures and the imaging core facility’s established operating protocol. Scan quality was reviewed for focus, tissue coverage, and staining artifacts before image analysis. Regions of interest (ROIs) were manually delineated and included cortical gray matter. ROIs were defined using anatomical landmarks by an investigator blinded to clinical status, biomarker values, and downstream quantitative results. ROI annotation was performed before quantitative analysis and applied to both Aβ and pTau181 measurements.

For Aβ pathology, area-based quantification was performed using the Area Quantification module in HALO image analysis software (v3.4.2986; Indica Labs). The percentage area occupied by positive 6E10 immunostaining was calculated within each ROI. Thresholding and masking parameters were optimized to identify Aβ-positive plaques while excluding nonspecific background staining and tissue artifacts using negative annotations. After optimization, the same analysis algorithm was applied uniformly to all scans.

For pTau181 pathology, cell-based quantitative analysis was performed using QuPath software version 0.6.028,29. Whole-slide images were imported into QuPath, and ROIs corresponding to the same anatomical regions analyzed for Aβ were selected for analysis. Color deconvolution was applied to separate hematoxylin and DAB signals. Cells were detected using optical density sum as the detection image, and detection parameters were optimized to identify nuclei/cells within the selected ROIs. A supervised machine-learning object classifier was trained to classify detected cells as pTau181-positive or pTau181-negative. The classifier was trained using representative ROIs from six dogs, with four ROIs included per training case. Approximately 1000 pTau181-positive cells and 1000 pTau181-negative cells, defined as cells lacking visible DAB-positive pTau181 immunoreactivity, were manually annotated for classifier training. A random trees object classifier was trained using morphometric and stain-intensity features, including cell/nuclear area, perimeter, circularity, hematoxylin optical density measurements, and DAB optical density measurements such as mean intensity and standard deviation.

After classifier training, the finalized classifier was applied to all selected ROIs using batch analysis. Positively classified pTau181 cells were exported from QuPath, and cell counts were normalized to the analyzed tissue area and reported as pTau181-positive cells/mm². Representative segmentation outputs for HALO-based Aβ area quantification and QuPath-based pTau181 cell classification are provided in Supplementary Fig. 5.

Statistical analysis

Prior to conducting group comparisons, normality was assessed using the Shapiro-Wilk test, and Levene’s test was performed to check for homogeneity of variance. For group comparisons, two-sample t-tests were used where assumptions of normality and homogeneity were met and Mann-Whitney test when normality assumption was violated. Spearman’s correlation was computed to assess the association between CCDR scores, dog age in years, quantitive brain tissue Aβ and blood biomarker concentrations. All statistical analyses were performed using R software (version 4.3.1) and GraphPad Prism (version 10.1.2), with statistical significance set at p < 0.05 for all analyses.

Results

The extent of neuropathological changes in our cohort of aged dogs (eight years or older) was evaluated using a comparative neuropathologic assessment framework adapted from theNIA–AA guidelines for the neuropathological assessment of AD-related pathology. Figure 1 illustrates the study design and neuropathologic evaluation workflow. Figure 2A–F demonstrates the regional distribution of Aβ pathology across the canine brain. All dogs (100%) exhibited Aβ deposition in the neocortex. Aβ pathology was also observed in the hippocampus in 87.5% of dogs, the striatum in 37.5% of dogs, and the midbrain in 8% of dogs, while no Aβ deposits were detected in the cerebellum (Fig. 3B). This regional pattern is consistent with progressive involvement of brain regions affected early in the hierarchical distribution of amyloid pathology described in the Thal framework.

[图：Study design and methodological overview. Schematic representation of the study workflow. Antemortem assessments include]

[图：Senior canines show neuropathological characteristics similar to humans. Representative immunohistochemistry images show]

[图：Regional distribution of β-amyloid and tau pathology in dogs. A Coronal brain sections illustrating the relative density]

Neuropathologic examination of tau pathology was modeled after Braak regional staging principles using immunohistochemistry for phosphorylated tau (pTau181) (Fig. 2I–K). Tau pathology was less pervasive than amyloid, with 71% of animals exhibiting pre-tangle pathology. The hippocampus, particularly CA1, was most frequently involved (62.5%), with fewer animals showing extension into the frontal or primary visual cortices (46%) (Fig. 3B). Among animals with cortical involvement, most exhibited a hippocampus-to-neocortex progression consistent with Braak staging; however, this pattern was not uniform across the cohort. A subset of animals (17%) showed atypical, non-hierarchical tau distribution, indicating that Braak-like progression does not fully capture tau pathology in all dogs. These atypical cases included animals with neocortical pTau181 immunoreactivity in the absence of hippocampal involvement, as well as animals with non-contiguous regional involvement. Specifically, dog CD3 showed visual cortical involvement without detectable hippocampal tau pathology, while dogs 17, 18, and 21 showed involvement of both the hippocampus and visual cortex without the expected involvement of the middle frontal gyrus (Fig. 3B). These findings suggest that although human Braak staging provides a useful comparative framework, tau pathology in dogs may follow species-specific or non-hierarchical patterns.

Using the criteria published by CERAD, we evaluated the brains for neuritic plaques. Despite widespread Aβ deposition detected by immunohistochemistry, no plaques meeting the morphologic criteria for neuritic plaques were identified by Bielschowsky silver staining. Accordingly, all animals received a CERAD score of absent (Fig. 2L and Supplemental Fig. 2).

To generate an overall assessment of AD-like neuropathologic change, we integrated regional Aβ, pTau181, and neuritic plaque findings using a comparative framework informed by the 2012 NIA-AA guidelines and adapted for canine neuropathology18. Based on this framework, the majority of dogs (80%) were classified as having LOW AD-like NC, while 16% were classified as having INTERMEDIATE AD-like NC; no animals met criteria for HIGH AD-like NC (Table 2). Because all canine brains exhibited a CERAD score of C0, the maximum achievable AD-like NC classification within this framework was INTERMEDIATE, as HIGH AD-like NC would require moderate-to-frequent neuritic plaque pathology (CERAD C2–C3). One animal exhibited advanced regional Aβ deposition together with pTau181-positive pathology restricted to the primary visual cortex. Because this pattern was not consistent with the hierarchical regional distribution of tau pathology described by the Braak framework, this case was excluded from AD-like NC classification.

Cerebral amyloid angiopathy (CAA) was observed in 9 dogs. Vascular-associated Aβ deposition was most prominent in posterior regions, particularly the occipital (visual) cortex, with a smaller subset showing involvement of both visual and frontal cortices. Because CAA was assessed descriptively, formal severity grading, vessel caliber measurements, and CAA subtype classification were not performed.

Regional differences in amyloid burden were first assessed using a semi-quantitative scoring approach. The prefrontal cortex exhibited the highest plaque burden, followed by the hippocampus and visual cortex, while the striatum and midbrain showed minimal pathology (Fig. 3A). Across individual dogs, β-amyloid deposition was moderate to high in cortical and hippocampal regions, with considerable inter-individual variability (Fig. 3B). Compared to amyloid pathology, tau pathology was less widespread and appeared more variable between cases (Fig. 3B).

Quantitative analysis revealed a significantly higher Aβ plaque burden in the prefrontal cortex of CCD+ dogs compared with CCD− dogs (Fig. 4A). No statistically significant difference was observed in the visual cortex between groups (Fig. 4C). In contrast, tau burden was significantly higher in the visual cortex of CCD+ dogs compared with CCD− dogs (Fig. 4D), and no significant differences were detected in the prefrontal cortex (Fig. 4B).

[图：Region-specific increases in β-amyloid and P-tau181 pathology in cognitively impaired dogs. Quantitative assessment of β]

Correlation analyses demonstrated a moderate association between age and Aβ plaque burden in the visual cortex (r = 0.57, p-value = 0.003), but no significant association in the prefrontal cortex (r = 0.35, p-value = 0.08). CCDR scores showed no significant correlation with amyloid burden in either regions (Supplementary Fig. 3). Similarly, correlations between tau burden and age or CCDR scores were non-significant across regions (Supplementary Fig. 4).

To determine if the underlying neuropathologic changes in the aged dog brains are associated with antemortem plasma biomarkers, we measured neurofilament light chain (Nf-L), Ab1-40, Ab1-42 and glial fibrillary acidic protein (GFAP). These biomarkers have previously been shown by our group and others to be associated with age, cognitive status, and neurodegeneration in dogs19–21. None of the plasma biomarker levels differed significantly between cognitively impaired (CCD+) and cognitively intact (CCD−) dogs for any of the analytes measured, including NfL, GFAP, Aβ₁–₄₀, Aβ₁–₄₂, or the Aβ₁–₄₂/Aβ₁–₄₀ ratio (Fig. 5). However, despite the absence of group-level differences, exploratory correlation analyses revealed a moderate positive association between plasma GFAP levels and regional β-amyloid plaque burden in both the prefrontal cortex and visual cortex (Supplemental Fig. 5B, E; r = 0.5, p-value = 0.04, for both regions). In contrast, plasma NfL levels and the Aβ₁–₄₂/Aβ₁–₄₀ ratio had minimal correlations with amyloid plaque burden across regions (Supplemental Fig. 4A, B, E, F) and plasma biomarkers demonstrated weak associations with tau burden (Supplementary Fig. 6).

[图：Plasma biomarker levels in aged dogs with and without cognitive dysfunction. Plasma concentrations of NfL ( A ), GFAP ( ]

Discussion

To determine whether and how the aging dog recapitulates key neuropathologic features of Alzheimer’s disease (AD) we applied a comparative neuropathologic assessment framework informed by the National Institute on Aging–Alzheimer’s Association (NIA–AA) neuropathological criteria to dog brains. Using tools and approaches adapted from those used in human studies, we found, consistent with prior reports, that senior dogs develop β-amyloid (Aβ) plaques in an age-dependent manner that partially parallels early amyloid pathology in humans as described in Fig. 1. This finding is consistent with previous studies showing that aged dogs commonly develop diffuse cortical Aβ deposition, particularly in frontal and temporal/entorhinal regions6,30,31, and that plaque burden tends to increase with age and cognitive impairment12,32. Our study extends these findings by applying a structured regional assessment framework informed by NIA–AA and Thal concepts to characterize canine Aβ distribution in relation to patterns observed in human AD. Notably, the neocortex appeared to be the earliest site of plaque deposition, as all dogs in our cohort exhibited at least some cortical diffuse Aβ pathology. This pattern is similar to findings from human autopsy studies, in which approximately 86% of individuals exhibit cortical amyloid deposition25. Consistent with patterns described in humans, amyloid pathology in dogs followed a hierarchical regional distribution with involvement of the hippocampus after the neocortex, followed by the neostriatum, and, in rare cases, the midbrain. Importantly, amyloid deposition in most dogs remained restricted to brain regions associated with early amyloid involvement, indicating that most aged dogs exhibit a pattern of Aβ deposition that overlaps primarily with regions affected early in the Thal framework (Fig. 2). Consistent with this observation, no neuritic plaques, which represent a more mature and pathogenic form of amyloid deposition, were identified in any canine brain examined, suggesting that canine Aβ pathology differs from human AD in plaque maturation and associated neuritic degeneration.

Pre-tangle pathology was also identified in limbic and, less frequently, neocortical regions, indicating that tau pathology is present in CCD syndrome; however tau pathology was less prevalent and less widespread than amyloid deposition and did not fully recapitulate the distribution observed in human AD. Across the assessed brain regions, Aβ deposits were descriptievely diffuse in morphology. Compact or cored plaques were only rarely observed, and Bielschowsky staining did not reveal neuritic plaques. Thus, although Aβ deposition was common in this cohort of aged dogs, the plaque pattern was characterized predominantly by diffuse rather than neuritic/compact plaque pathology. A notable finding was the divergence between regional amyloid and tau pathology in the canine brain. Although some dogs showed relatively extensive regional Aβ deposition, pTau181 pathology remained limited and did not exhibit the widespread hierarchical progression typical of human AD. This discordance may suggest species-specific resistance to tau propagation or maturation, highlighting aged dogs as a comparative model for studying mechanisms that limit tau spreading despite amyloid accumulation. In this study, tau pathology was assessed using an antibody targeting phosphorylated tau at threonine 181 (pTau181) because prior canine studies have reported sparse or infrequent AT8-positive neurofibrillary-like pathology, despite broader phosphorylated tau abnormalities detected with other phospho-tau markers. Phosphorylation at Thr181 is considered an early tau modification in AD pathogenesis and recent studies demonstrate that different phospho-tau epitopes capture distinct aspects of tau pathology across tissue- and biomarker-based assessments33,34. Together with the early Thal-like regional Aβ involvement and the absence of neuritic plaques, these findings indicate that the dog brain exhibits AD-like neuropathologic features that overlap with pathological changes observed early in human AD, while not constituting a validated equivalent of any defined stage of human disease.

A major unresolved question in AD research concerns whether and how amyloid and tau pathology interact to drive disease progression. Our data provide some insight into this relationship in a natural aging model. All dogs with pTau181 pathology also exhibited at least some degree of amyloid deposition, whereas not all dogs with amyloid pathology showed pTau181-positive cells. Moreover, dogs with more severe amyloid pathology in the prefrontal cortex frequently exhibited greater pTau181 burden in the hippocampus, and animals with pTau181 pathology extending from hippocampal regions into neocortical regions typically exhibited more extensive regional Aβ involvement (Fig. 3). This pattern is consistent with the hypothesis that amyloid pathology facilitates, but is not sufficient for, the development and spread of tau pathology35.

Quantitative analyses demonstrated that cognitively impaired dogs (CCD+) exhibited significantly greater Aβ plaque burden in the prefrontal cortex compared with cognitively intact (CCD−) dogs. In addition, CCD+ dogs showed significantly higher pTau181-positive cells in the visual cortex, indicating that both Aβ and pTau181 cortical pathologies are associated with cognitive impairment but may follow a region-specific pattern in the aging canine brain. However, some CCD− dogs displayed Aβ pathology involving both neocortical and hippocampal regions, together with moderate to severe plaque burden. This may indicate that owner-reported behavioral assessments and CCDR scores may not fully capture subtle or preclinical cognitive changes, particularly in the earliest stages of disease. Alternatively, and consistent with observations in humans, some dogs may exhibit resilience to underlying neuropathology, resulting in a dissociation between cognitive performance and pathological burden36–38.

Correlation analyses revealed that age was significantly associated with increased Aβ burden in the visual cortex (r = 0.57, p-value = 0.003), reinforcing age as a strong risk factor for pathology (Supplement Fig. 3). In contrast, CCDR scores were not significantly associated with neocortical Aβ burden (Supplement Fig. 3). Astrocytic GFAP plasma measurement showed significant positive correlation with Aβ burden in both regions (Supplemental Fig. 5B, E, r = 0.5, p-value = 0.04, for both regions). This finding parallels human studies demonstrating that increase in plasma GFAP levels reflect astrocytic activation secondary to Aβ pathology39. However, given the small sample size in our plasma cohort, this finding should be further investigated in a larger cohort of senior dogs. Although this study measured plasma levels of Aβ42/40 and GFAP, plasma p-tau217 has emerged as one of the most promising blood-based biomarkers for AD pathology in humans, showing strong performance for detecting amyloid and tau pathology. Because canine tau pathology in this study was assessed histologically using pTau181, future studies should evaluate whether plasma p-tau217 can be analytically validated in dogs and whether it reflects tissue-level tau burden, amyloid stage, or cognitive decline in CCD. Tau pathology was not significantly associated with age, CCDR scores, or plasma biomarkers in any region (Supplementary Figs. 5 and 6). This observation is consistent with previous findings indicated that tau pathology in senior dogs is limited with variable distribution3,40–42. In contrast, studies in AD have demonstrated strong associations between tau pathology and cognitive decline43,44. These results indicate that tau accumulation and distribution in senior dogs are more variable than amyloid pathology. This variability support the presence of pathological heterogeneity in canine brain, where amyloid deposition appears to follow consistent regional pattern, while pre-tangle pathology shows greater variability in regional involvement and distribution.

Within this cohort, dogs with more extensive regional Aβ involvement and higher pTau181pathology frequently exhibited neocortical involvement; however, marked heterogeneity was observed. Some animals showed substantial amyloid deposition with minimal tau pathology, whereas others exhibited tau pathology in the presence of limited amyloid burden. This variability suggests that additional pathogenic processes, including synucleinopathies, TDP-43 pathology, and vascular disease, may contribute to cognitive impairment and neuropathologic diversity in senior dogs, as found in human neurodegenerative disease38,45.

A major strength of this study is the systematic application of comparative neuropathologic assessment frameworks informed by human AD neuropathology to a naturally aging canine cohort. By adapting concepts from the Thal, Braak, and CERAD frameworks to assess regional Aβ deposition, pTau181 pathology, and neuritic plaque pathology, this work provides the first integrated clinico-pathologic evaluation of AD-like changes in companion dogs using criteria adapted from human neuropathological frameworks, with modifications necessitated by species-specific differences. Importantly, these comparative efforts were conducted in collabortion with human neuropathologists to ensure that pathological interpretations were alligned with similar standards used in AD research. The use of client-owned dogs further enhances translational relevance, as these animals experience heterogeneous genetic backgrounds, shared human environments, and natural aging trajectories that are not captured in traditional laboratory models. Integration of standardized neuropathologic assessment with validated owner-reported cognitive measures and antemortem plasma biomarkers represents an additional strength, enabling evaluation of relationships among pathology, cognition, and peripheral markers within the same individuals.

Several limitations should also be considered. First, the cross-sectional, postmortem design precludes inference regarding the temporal sequence of amyloid and tau pathology and their relationship to cognitive decline. Longitudinal studies incorporating serial cognitive assessments, biomarker measurements, and eventual neuropathologic evaluation will be essential to define disease trajectories in this model. Second, cognitive status was assessed using owner-reported instruments, which, while validated, may lack sensitivity to detect subtle or domain-specific impairments, particularly in preclinical disease stages46. Although the validated CCDR threshold for CCD diagnosis is generally reported as ≥50, this study used a broader CCDR cutoff of >40 to include dogs with clinically apparent CCD as well as dogs in the at-risk or MCI range. Therefore, dogs with CCDR scores between 40 and 49 were interpreted as cognitively impaired or at risk rather than definitively CCD-positive by the original validated diagnostic threshold.Third, tau pathology was assessed using a pTau181-specific antibody. Although previous canine studies have demonstrated AT8 immunoreactivity, AT8-positive pathology has generally been reported as limited and regionally restricted relative to other phospho-tau epitopes. Because we did not perform side-by-side AT8 and pTau181 immunohistochemistry, the full spectrum of tau pathology in dogs may not have been captured. Similarly, different antibodies and methodologies can be used to assess amyloid pathology. Here we utilized 6E10 immunohistochemistry; future studies incorporating Aβ species-specific antibodies (e.g., Aβ42) and alternative antigen-retrieval approaches such as formic acid pretreatment may provide additional information regarding plaque composition, maturation, and regional distribution. Additionally, this study was not powered to assess the influence of sex, breed, body size, or lifespan variability, all of which may contribute to heterogeneity in aging trajectories and neuropathologic burden.

Fourth, this study did not investigate other age-related or neurodegenerative co-pathologies, including TDP-43, α-synuclein, or other comorbid lesions that may contribute to cognitive decline in aged dogs. Fifth, digital image analysis may be influenced by staining variability, tissue processing differences, tissue folds or section artifacts, scan quality, ROI delineation, threshold selection, and classifier performance. Although the same parameters were applied uniformly across all cases, the image analysis outputs were not validated against manual expert count. Sixth, this study did not systematically evaluate the layer specific distribution of cortical Aβ deposits or pretangle-like pTau181 pathology. Seventh, the Human Neurology 4-Plex E Kit was developed for human samples and has not, to our knowledge, been fully validated for canine plasma; thus, these results should be interpreted as translational biomarker measures rather than clinically validated canine values. Eighth, this study focused on Aβ, p-Tau181, and plasma biomarkers and did not assess tissue-level markers of neuroinflammation, synaptic loss, neuronal loss, oxidative injury, or other neurodegenerative processes.

Finally, several neuropathologic features commonly characterized in human AD studies were not systematically evaluated in the present analysis. Although Aβ deposits were observed to be predominantly diffuse and pTau181-positive cells demonstrated morphologic features consistent with tau hyperphosphorylation and pretangle-like pathology, plaque morphology (diffuse versus compact/cored), cortical laminar distribution, vascular Aβ subtype classification, and detailed morphologic subtyping of pTau181-positive cells were not quantitatively scored. Similarly, tissue-level measures of neuroinflammation, neuronal loss, synaptic integrity, and other age-related co-pathologies were beyond the scope of the current study. Consequently, the neuropathologic assessments presented here should be interpreted as a comparative characterization of AD-like pathology rather than a comprehensive evaluation of all neuropathologic processes contributing to cognitive aging in dogs.

Despite these limitations, the strengths of this approach, particularly the use of comparative neuropathologic assessment frameworks informed by human AD pathology in a naturally aging, translationally relevant, support the value of the companion dog as a model for investigating early AD-related pathological processes.

Together, these findings support the validity of companion dogs as a naturally occurring model for AD-like pathology, particularly for investigating the earliest stages of disease. By integrating standardized human neuropathological frameworks with behavioral assessments and plasma biomarkers, our study bridges a critical gap between canine and human aging research. Future studies should expand the characterization of mixed neuropathological markers in dogs and further evaluate biomarker-pathology relationships using longitudinal and postmortem approaches. Such investigations will provide novel insights into the mechanisms that underlie CCD and may inform translational strategies relevant to both veterinary and human neurodegenerative disease.

Supplementary information

[内容请查看原文]