Unveiling the procoagulant state in Alzheimer's disease: A novel PET imaging strategy

Abstract

INTRODUCTION

A subset of Alzheimer's disease (AD) patients exhibits a procoagulant state that remains undiagnosed despite available treatments. We developed positron emission tomography (PET) strategies to in vivo detect cerebral microthrombi in TgCRND8 mice.

METHODS

PET was performed with a fibrin‐binding probe (FBP) in TgCRND8 and wild‐type mice using a zirconium‐89–radiolabeled and a click‐chemistry pre‐targeted gallium‐68 radiotracer. Platelet content was assessed in post mortem brain tissue from AD patients and TgCRND8 mice, and evaluated by PET using a CD41 pre‐targeted gallium‐68 nanoradiotracer.

RESULTS

Cerebral FBP uptake differentiated TgCRND8 from wild‐type mice, particularly ex vivo and with advancing age. Cerebral fibrin burden correlated with platelet content in AD patients, and elevated cerebral platelet burden in TgCRND8 mice was confirmed in post mortem tissue and in vivo by PET.

DISCUSSION

Our findings support PET‐based detection of AD‐associated cerebral microthrombi and the development of imaging biomarkers to stratify AD patients by procoagulant status and inform personalized therapies.

 

Highlights

• Half of patients with Alzheimer's disease (AD) exhibit a procoagulant state.

• AD procoagulant state is characterized by abnormal fibrin deposition in the brain.

• Fibrin burden remains undiagnosed despite available counteracting therapies.

• Fibrin burden correlates with platelet load in the brains of AD patients.

• Novel positron emission tomography (PET) biomarkers enable in vivo detection of cerebral microthrombi in AD mice.

1. BACKGROUND

Over 55 million people are currently living with dementia worldwide ¹ and, with increasing life expectancy, this number is projected to reach 153 million by 2050. ² The term “dementia” encompasses several conditions characterized by memory and cognitive impairment, among which Alzheimer's disease (AD) is the most common. The World Health Organization estimates that AD accounts for 60–70% of all dementia cases. ¹ AD is a complex, multifactorial neurodegenerative disorder characterized by aggregation of amyloid‐β (Aβ) and phosphorylated tau in the brain parenchyma, creating senile plaques and neurofibrillary tangles, respectively. ³ Beyond these classic hallmarks, many other neuropathogenic processes contribute to AD, including neuroinflammation, synaptic dysfunction and a strong vascular component. Together, these mechanisms create a heterogeneous scenario where single‐target treatments are failing partly due to their inability to address the complex set of factors involved. ⁴ Multifocal diagnostic and therapeutic strategies are urgently needed to handle the multifaceted nature of AD and enable personalized interventions.

In 1907, Alois Alzheimer first described the neuropathology of AD and, in his original report, also noted vascular alterations (i.e., arteriosclerotic changes) in Auguste Deter's brain. ⁵ Today, mounting evidence implicates vascular dysfunction in AD, including intracranial atherosclerosis, cerebral amyloid angiopathy (CAA) and neurovascular unit alteration leading to blood–brain barrier (BBB) disruption. ⁶ Furthermore, AD also features an altered hemostatic response ⁷ that affects ∼half of AD patients, ⁶ , ⁸ promoting abnormal formation of blood clots in the brain. ⁸ , ⁹ , ¹⁰ Aβ interacts with fibrin(ogen) — the main protein component of blood clots — hindering access of fibrinolytic enzymes to the fibrin clot mesh and thereby stabilizing microthrombi. ¹⁰ , ¹¹ , ¹² Aβ accumulation within cerebral blood clots has been attributed to platelets, ¹³ the predominant cellular element of thrombi and the major source of circulating Aβ. ¹⁴ , ¹⁵ Contact with Aβ promotes platelet activation, whose exacerbation has been repeatedly linked to AD, ¹⁶ , ¹⁷ , ¹⁸ leading to further increased Aβ release, ¹⁹ platelet adhesion and aggregation, ¹⁹ , ²⁰ and ultimately contributing to CAA, ¹⁹ senile plaques ²¹ and thrombus formation. ²² This prothrombotic milieu in AD ²³ drives abnormal cerebral fibrin(ogen) accumulation that lines or occludes cerebral blood vessels and extravasates into the brain parenchyma, ⁸ exacerbating CAA, ²⁴ cerebral hypoperfusion, ²⁵ BBB disruption, ²⁵ synaptic dysfunction, ⁸ , ²⁶ , ²⁷ neurodegeneration, ⁸ neuroinflammation, ²⁷ , ²⁸ , ²⁹ amyloid deposition ⁹ , ²⁵ and cognitive deficits, ⁹ , ²⁵ all strongly associated with AD. ²³ , ³⁰

Despite the availability of novel anticoagulant therapies ⁷ , ³¹ and fibrin‐targeted immunotherapy ³² , ³³ that show promise in delaying AD progression in animal models, fibrin(ogen) accumulation and fibrin clot formation in the AD brain remain undiagnosed due to the lack of in vivo diagnostic strategies. This limitation is critical, as not all AD patients exhibit abnormal fibrin deposition in their brains, ⁸ stressing the need for appropriate diagnostic tools to identify and stratify patients and guide therapy. Neuroimaging, especially positron emission tomography (PET), is well suited for this purpose, as it enables in vivo, non‐invasive visualization of brain functional processes with exceptional sensitivity and specificity. For example, Aβ and tau PET imaging have transformed our understanding of AD progression, guiding both clinical management and therapeutic decisions. ³⁴ However, while several PET strategies have been developed to target components of the coagulation cascade (e.g., fibrin, factor XIIIa, activated platelets ³⁵ , ³⁶ ), these primarily detect large, well‐formed clots rather than the diffuse cerebral microthrombi that characterize the AD procoagulant state.

RESEARCH IN CONTEXT

1. Systematic review: A thorough literature review using official scientific databases (e.g., PubMed) was performed to frame this work. Half of Alzheimer's disease (AD) patients exhibit a procoagulant state characterized by fibrin and fibrinogen deposition in the brain and microvascular dysfunction. Although counteracting therapies exist, this condition remains largely neglected due to the lack of in vivo diagnostic strategies. All relevant citations are appropriately included.

2. Interpretation: We have developed highly specific positron emission tomography (PET) biomarkers targeting fibrin and platelets, providing the first in vivo neuroimaging approach to detect AD's procoagulant environment in TgCRND8 mice. Our study also offers the first quantitative evidence of increased platelet accumulation in the procoagulant human AD brain and its correlation with cerebral fibrin deposition.

3. Future directions: This manuscript expands current understanding of cerebrovascular contributions to AD and opens the door to non‐invasive diagnostic approaches with tangible translational potential for patient stratification and therapeutic guidance.

Therefore, we aim to develop a novel neuroimaging strategy to diagnose this AD‐associated prothrombotic state (i.e., cerebral fibrin deposition). Using TgCRND8 AD mice, which model this condition, ⁸ , ⁹ , ²⁵ we developed two pre‐targeting approaches based on fibrin‐ and platelet‐binding probes with clear translational potential. Our data provide a direct, non‐invasive in vivo measure of cerebral microthrombi burden in an AD mouse model, laying the groundwork for diagnostic tools that could support patient selection for anticoagulant, antiplatelet and/or fibrin‐targeted interventions aimed at slowing AD progression.

2. METHODS

2.1. Animals

TgCRND8 AD mice, kindly provided from The Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto (Canada), overexpress the Swedish (KM670/671NL) and Indiana (V717F) familial AD mutations in the amyloid precursor protein under the control of the PrP gene promoter. ³⁷ Mice were kept on a mixed background (C57xC3H/C57), genotyped, and TgCRND8 AD heterozygous and non‐transgenic wild‐type (WT) littermates male and female mice were randomly selected throughout the experiments, as detailed below. Sample sizes were calculated empirically according to pilot imaging studies for an estimated power of 95%.

Mice were housed up to five mice per cage at the National Center for Cardiovascular Research (CNIC) Comparative Medicine Unit in a dedicated vivarium, under controlled temperature and humidity conditions, with a 12/12 hour light‐dark cycle. All experimental procedures were conducted by operators who were unaware of the experimental groups, according to European Communities Council Directive 2010/63/EU, following the ARRIVE guidelines, ³⁸ and approved by the Ethics Committee in Animal Experimentation and corresponding authority (PROEX 147.7_2).

Carotid Crush Injury Model – One day before PET acquisition, WT mice underwent surgical induction of a carotid crush injury to serve as positive controls (C+) for the pre‐targeting imaging approaches, as previously described. ³⁹ Briefly, 70‐week‐old WT littermates were anesthetized with sevoflurane (5% induction, 3% maintenance in medical gas). The left common carotid artery was exposed, and blood flow was occluded for 15 minutes, after which the clamp was released to restore perfusion. Immediately after surgery, animals were injected with either the corresponding pre‐targeting probe or saline and processed according to the same experimental protocol used for each radiotracer, as detailed below.

2.2. Radiotracer synthesis

All reagents were purchased from Merck Life Science S.L.U. (Spain) unless otherwise indicated.

2.2.1. Fibrin imaging (direct labeling)

This strategy utilizes zirconium‐89 (⁸⁹Zr) to radiolabel a commercial Fibrin Binding Probe (FBP) ⁴⁰ modified with desferrioxamine B (FBP‐DFO, Pepscan, The Netherlands) at the CNIC Advanced Imaging Unit radiochemistry facilities, adapting a protocol previously published. ⁴¹ First, 92.5 MBq of ⁸⁹Zr (PerkinElmer, USA) were placed in a 2 mL Eppendorf® tube. Then, 200 µL solution of 1 M oxalic acid and 90 µL of 2 M Na₂CO₃ were added to the solution, which was incubated for 3 minutes at room temperature (RT). Later, 300 µL of 0.5 M HEPES, 710 µL of FBP‐DFO, and 700 µL of 0.5 M HEPES were sequentially pipetted into the reaction tube, and the solution was incubated for 60 minutes at RT and 500 rpm in the ThermoMixer® (Eppendorf Iberica S.L.U., Spain). The reaction product was transferred to a C18 SEP‐PAK® light cartridge (Waters Corporation, USA), previously conditioned with 5 mL of EtOH and 10 mL of H₂O. The cartridge was then washed with 0.5 mL of H₂O to remove the free isotope, and finally the desired product was eluted and collected with 0.5 mL of EtOH. For the in vivo studies, the product was partially evaporated and reformulated in saline, maintaining 20% EtOH in the final formulation. The quality control of the final product was performed by HPLC with a C18 column and a flow rate of 0.8 mL/min (Phase A: Water + TFA 0.1%; Phase B: CH₃CN; Gradient: 0–18 minutes from 0 to 100% of phase B, 18–20 minutes 100% of phase B).

Once [⁸⁹Zr]Zr‐DFO‐FBP was purified, animals were intravenously (IV) injected with 7.54 ± 2.98 MBq/mice (mean ± SD) through the tail vein, followed by imaging acquisition. [⁸⁹Zr]Zr‐DFO‐FBP studies included five sequential PET acquisitions immediately after the radiotracer injection (0 hour), 24, 48, 72, and 96 hours post injection. TgCRND8 AD mice and WT littermate males were used (70.77 ± 3.11‐week‐old, mean ± SD. 0 hour: N_AD = 5, N_WT = 5; 24–96 hours: N_AD = 8, N_WT = 8).

2.2.2. Fibrin imaging (pre‐targeting)

This strategy utilizes a biorthogonal reaction based on click chemistry between FBP coupled to transcyclooctene (TCO) and a tetrazine (TZ) derivative radiolabeled with gallium‐68 (⁶⁸Ga), which allows an ultra‐fast, covalent interaction under physiological conditions between these two chemical species (i.e., TZ and TCO). The FBP peptide (FHCHypY(3‐Cl)DLCHIL‐PXD (FBP‐PXD) described by Oliveira et al. ⁴⁰ ) was in‐house synthesized at the Complutense University of Madrid (UCM) by microwave‐supported solid‐phase peptide synthesis to enable temperature control, thereby reducing reaction times and improving product purity an d yield. Coupling reactions were performed with 3 equivalents (eq) of the corresponding amino acid and 3 eq of O‐(Benzotriazol‐1‐yl)‐N,N,N′,N′‐tetramethyluronium hexafluorophosphate, both dissolved in amine‐free dimethylformamide (AF DMF), together with 3.5 eq of N,N‐diisopropylethylamine. On the other hand, Fmoc deprotection steps were carried out using 20% Piperidine in AF DMF. Standard coupling and deprotection reactions were conducted at 75°C for 15 minutes, except for Cysteine (Cys) and Histidine units, which required milder conditions (50°C for 20 minutes). The synthesis was initiated by coupling para‐xylenediamine (PXD) to a 2‐chlorotrityl resin (1.55 mmol/g) at RT. This step was performed using 3.5 eq of PXD in AF DMF with 4 eq of N,N‐diisopropylethylamine, under controlled agitation for 2 hours. After each coupling/deprotection steps, washing procedure was carried out with 5 mL DCM, 5 mL DMF, and 5 mL AF DMF. Following chain assembly, the N‐terminal amine was acetylated using 10% acetic anhydride in AF DMF, under agitation at RT for 1.5 hours. The peptide was cleaved from the resin with a TFA‐based cleavage cocktail (86.5% TFA, 4.5% Triisopropylsilane, 4.5% Dithiothreitol, 4.5% Methanesulfonic acid) for 2 hours with vigorous agitation. The crude peptide was then precipitated in cold diethyl ether. After purification and lyophilization, 234 mg of peptide was obtained, corresponding to a 48% overall yield. Additionally, a non‐functional FBP isomer that differs in the chirality of one cysteine (i.e., DCys‐FBP), preventing the peptide cyclization crucial for fibrin binding, ⁴² was identically synthesized and used as negative control of the fibrin pre‐targeting probe (C‐DCys).

Once synthesized, FBP and DCys‐FBP were coupled to TCO‐PEG4‐NHS (TCI Europe N.V., Belgium) under an argon atmosphere using a 1:1 stoichiometric ratio. Both components were dissolved in anhydrous DMF in argon‐purged vials. To the peptide solution, 20 eq of N,N‐diisopropylethylamine were added to deprotonate the PXD amine, and the mixture was stirred for 1 h RT. TCO‐PEG4‐NHS was then added dropwise, and the reaction was allowed to proceed overnight at RT under continuous stirring. The crude product was precipitated with cold diethyl ether, purified, and lyophilized.

Similarly, TZ was coupled with 1,4,7,10‐tetraazacyclododecane‐1,4,7,10‐tetraacetic acid (DOTA) DOTA, 1,4,7,10‐tetraazacyclododecane‐1,4,7,10‐tetraacetic acid; following a protocol developed at UCM, and radiolabeled at CNIC radiochemistry using ⁶⁸Ga obtained from a germanium‐68/⁶⁸Ga generator (IRE Elit Radiopharma, Belgium). First, we placed 400 MBq of [⁶⁸Ga]GaCl₃ (0.05 – 0.1 M) in a 1.5 mL Eppendorf® tube. Then, 20 µL of TZ‐DOTA (5 mg/mL) were added to the tube, followed by 200 µL of 1 M HEPES. The solution was incubated during 20 minutes at 95°C and 500 rpm in a ThermoMixer®. The reaction product was transferred to a C18 SEP‐PAK® light cartridge (Waters Corporation), previously conditioned with 5 mL of EtOH and 5 mL of H₂O. The cartridge was then washed with 5 mL of H₂O to remove the free isotope, and finally the desired product was eluted and collected with 0.3 mL EtOH. For the in vivo studies, the product was partially evaporated and reformulated in saline, maintaining 20% EtOH in the final formulation. Quality control of the final product was performed by HPLC with a C18 column and a flow rate of 0.8 mL/min (Phase A: Water + TFA 0.1%; Phase B: CH₃CN; Gradient: 0–18 minutes from 0 to 100% of phase B, 18–20 minutes 100% of phase B).

The day before imaging acquisition, animals were IV injected through the tail vein with FBP‐TCO (100 µg/mouse). The next day, animals were IV injected with 12.08 ± 0.64 MBq/mice (mean ± SD) of [⁶⁸Ga]Ga‐DOTA‐TZ through the tail vein and left for 2 hours before PET imaging. Experimental groups of 77.43 ± 13.28‐week‐old (mean ± SD) mice included a total of 14 AD (nine females, five males), 15 WT (eight females, seven males) and 14 AD animals injected with the non‐functional DCys‐FBP‐TCO (C‐DCys; six females, eight males) as negative controls.

2.2.3. Platelet imaging (pre‐targeting)

The platelet probe consisted of Thrombo‐tag, ⁴³ a PET radiotracer based on in vivo click chemistry between a specific anti‐platelet antibody (i.e., anti‐CD41) and new‐generation ⁶⁸Ga‐labeled nanoradiotracers: iron oxide nanoparticles (IONP) doped with ⁶⁸Ga. For functionalization of the anti‐CD41 antibody (#14‐0411‐85, eBioscience™, Thermo Fisher Scientific, USA) with TCO, 0.4 mg/mL of the anti‐CD41 antibody were mixed with 0.18 mg/mL 0.1 M Na₂CO₃ in phosphate buffered saline (PBS) 1× (pH ∼ 9) and 0.02 mg/mL of TCO‐NHS ester. The reaction was stirred at RT for 60 minutes, purified by ultrafiltration (100 kDa filters), and the retentate diluted with PBS to a concentration of 400 µg/mL. This protocol was also used to functionalize an immunoglobulin (Ig) G1 isotype control (#14‐4301‐85, eBioscience™, Thermo Fisher Scientific, USA), used as negative control of the platelet pre‐targeting probe (C‐IgG).

The radiosynthesis of [⁶⁸Ga]Ga‐IONP‐TZ was carried out before each imaging study, at the CNIC and Centre for Energy, Environmental and Technological Research (CIEMAT) radiochemistry facilities. Briefly, IONP‐TZ synthesis involved mixing the nanoparticle precursors (FeCl₃, sodium citrate and [⁶⁸Ga]GaCl₃) with hydrazine, and the resulting mixture was stirred under microwave at 120°C for 10 minutes. After that, the [⁶⁸Ga]Ga‐IONP solution was purified by ultrafiltration, and EDC and sulfoNHS (Thermo Fisher Scientific) were added to the purified product. This solution was incubated at RT and 650 rpm for 30 minutes, and later purified using 30 kDa Amicon filters. The nanoparticles were resuspended in HEPES buffer (pH 8) and 0.2 mg/mL of benzylamino TZ dissolved in dimethylsulfoxide (DMSO) was added. The solution was incubated at RT and 650 rpm for 45 minutes to form the [⁶⁸Ga]‐IONP‐TZ, ultrafiltrated and resuspended in PBS.

The day before the PET study, animals were IV injected through the tail vein with the antibody anti–CD41‐TCO (70 mg/kg). The next day, animals were IV injected with 9.23 ± 1.51 MBq/mice (mean ± SD) of [⁶⁸Ga]Ga‐IONP‐TZ through the tail vein, and PET acquisition was performed 2 hours post‐injection. Experimental groups of 79.10 ± 16.63‐week‐old (mean ± SD) mice included a total of 15 AD (11 females, four males), 18 WT (11 females, seven males), and 11 AD animals injected with the non‐specific IgG1‐TCO probe (C‐IgG; six females, five males) as negative controls.

2.3. Imaging acquisition and processing

Imaging studies were performed under isoflurane anesthesia (5% induction, 3% maintenance in medical air) in the facilities of the Distributed Biomedical Imaging Network (ReDIB, Spain) at CNIC. Temperature control and respiration (through a respiratory pad) were registered along the imaging session. In each imaging study, the whole body of the animal was scanned for 40 minutes: first acquiring a 10‐minute computerized tomography (CT) scan, and then a 30‐minute PET image. Decay and dead time corrections were applied.

Fibrin imaging (direct labelling) using [⁸⁹Zr]Zr‐DFO‐FBP and platelet imaging (pre‐targeting) using Thrombo‐tag (i.e., anti–CD41‐TCO + [⁶⁸Ga]Ga‐IONP‐TZ) were acquired with a dedicated small animal nanoScan® PET/CT scanner (Mediso, Hungary). CT studies were reconstructed with an isotropic voxel size of 0.14 mm. PET images were reconstructed using a 3D‐OSEM algorithm with four iterations and six subsets, and isotropic voxel size of 0.4 mm³ and were spatially normalized to a reference mouse brain atlas ⁴⁴ using a rigid registration algorithm based on mutual information available at the Multimodal Workstation software (BIIG, Spain). ⁴⁵

Fibrin imaging (pre‐targeting) acquisitions using FBP‐TCO + [⁶⁸Ga]Ga‐DOTA‐TZ were obtained with the X‐CUBE® micro‐CT and ɣ‐CUBE® micro‐PET scanners (Molecubes NV, Belgium). CT studies were reconstructed with an isotropic voxel size of 0.1 mm³. PET images were reconstructed using a 3D‐OSEM algorithm with 10 iterations and isotropic voxel size of 0.4 mm³ and were registered to the same brain atlas ⁴⁴ using a rigid registration algorithm based on mutual information in FSL (Oxford University, UK). ⁴⁶

Each registered PET image was smoothed using an isotropic Gaussian filter with Full Width at a Half Maximum of 2.5 times the voxel size. For intensity normalization, [⁸⁹Zr]Zr‐DFO‐FBP images were initially subjected to an iterative histogram‐based normalization ⁴⁷ to provide an objective basis for comparing tracer uptake across different time points (0, 24, 48, 72, and 96 hours post‐injection). After that, the 24‐hour time‐point was chosen for subsequent analyses since it revealed the largest differences between AD and WT groups. All scans were then analyzed using a reference cluster method, ⁴⁸ specifically designed for situations where a baseline bias in the data is expected (e.g., higher uptake in the AD group), correctly accounting for this bias and preventing underestimation of true group differences. Smoothing and normalization steps were performed by means of in‐house scripts made in MATLAB R2023a (MathWorksInc., USA).

PET data were analyzed using hypothesis‐driven and exploratory strategies. On the one hand, classical regions‐of‐interest (ROIs) analyses were performed. ROIs were defined by masks segmented from the reference mouse brain atlas ⁴⁴ for (coordinates from Bregma): whole brain (4.28 mm to ‐8.24 mm), cerebellum (‐5.34 mm to ‐8.24 mm), cortex (2.96 mm to ‐4.84 mm), hippocampus (‐1.58 mm to ‐2.92 mm), striatum (1.4 mm to ‐1.58 mm), subcortex (1.34 mm to ‐2.92 mm), and thalamus (‐0.82 mm to ‐2.92 mm); and directly applied on the previously preprocessed PET images to obtain their mean uptake value. On the other hand, a voxel wise exploratory approach was conducted to identified differences in brain uptake between AD and WT groups without a predefined hypothesis. See the 2.8. Statistical analyses section for details.

2.4. Biodistribution studies

Necropsies were performed immediately after each pre‐targeting PET acquisition. The activity in each brain sample was measured on a dedicated gamma‐counter (PerkinElmer, USA) and expressed as percentage of the injected dose per gram of tissue (%ID/g). The sample size of the biodistribution data does not strictly coincide with that of imaging studies because, on the rare occasions an error occurred during PET acquisition and images could not be recovered, brain samples were still collected for gamma‐counter measuring, maintaining the same uptake periods. Therefore, fibrin pre‐targeting experimental groups included 17 AD (eight females, nine males), 16 WT (nine females, seven males) and 16 C‐DCys (seven females, nine males), while platelet pre‐targeted groups comprised 16 AD (12 females, four males), 18 WT (13 females, five males), and 11 C‐IgG (six females, five males).

2.5. Brain samples

Post mortem frontal cortex samples and data from AD patients included in this study were provided by the Biobank Banco de Tejidos CIEN – BT‐CIEN (PT17/0015/0014), with the appropriate approval of the Ethics and Scientific Committees (CEI PI 91_2019 & CEI PI 14_2023). BT‐CIEN is integrated in the Spanish National Biobanks Network and samples were processed following standard operating procedures. The cohort (N = 39) comprised females (85%) and males (15%) with a mean ± SD age of 87.4 ± 5.7 years and an average post mortem interval of 4.4 hours (Table S1 of supporting information). Blocks of frozen frontal cortex samples were sliced by cryostat and collected in tubes for Western blot (WB) analysis, as previously described. ⁸ An additional subset of these blocks was fixed in 0.5% paraformaldehyde (PFA) for fibrin immunostaining. Additionally, slides containing paraffin‐embedded frontal cortex tissue from a subset of these patients (N = 11) were obtained from BT‐CIEN for subsequent immunohistochemistry assays.

Mice were anesthetized and transcardially perfused with a heparin/saline solution. Brains were removed and one hemisphere was dissected into cortex and hippocampus and directly frozen at −80°C for WB analysis, and the other hemisphere was fixed in PFA for subsequent staining procedures.

2.6. WB

Protein extraction for WB analyses was performed with slight modifications from previously described protocols. ⁸ Human frozen tissue collected in tubes was homogenized in 1:5 (g:ml) of PBS containing 1% Nonidet P‐40, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate and protease inhibitor cocktail (Roche‐11836170001) using the TissueRuptor II (Qiagen, USA) and minimum speed for 5‐7 seconds. The homogenate was centrifuged at 4°C at 10,000 × g for 10 minutes, and the supernatant (soluble fraction) was transferred to a different tube. The pellet was homogenized in 3 M urea, for 2 hours at 37°C and centrifuged at 14,000 × g for 15 minutes. Supernatant was collected in a different tube, and the insoluble (fibrin containing) pellet was resuspended in reducing sodium dodecyl sulfate loading buffer at 65°C for 15 minutes.

The amount of protein in the soluble fraction was quantified with a Bicinchoninic Acid kit for protein determination, following manufacturer's instructions (Thermo Fisher Scientific). Equal amounts of soluble (40 µg) or insoluble (10 µL) protein from each sample were run on 4%‐20% gradient polyacrylamide Criterion gels (Bio‐Rad, USA) under reducing conditions and then transferred to polyvinylidene fluoride membranes. Membranes were blocked with 3% bovine serum albumin or 5% non‐fat milk. Blots containing soluble protein extracts were cut by the 75 kDa marker and incubated overnight at 4°C in the same blocking solution with the following primary antibodies: rat monoclonal anti‐CD41 antibody (upper half; MWReg30, 1:1000, BD Pharmingen, USA) and rat monoclonal anti‐tubulin antibody (lower half; YOL1/34, 1:5000, Abcam, UK). Blots containing insoluble protein extracts were incubated with mouse monoclonal anti‐fibrin antibody (59D8, 1:1000, Thermo Fisher Scientific) under the same conditions. The following day, membranes were incubated with the corresponding secondary antibodies coupled with horseradish peroxidase (HRP; goat polyclonal anti‐mouse antibody, 1:1000, Agilent Dako, Denmark or goat polyclonal anti‐rat antibody, 1:5000, Thermo Fisher Scientific). Incubation was followed by development with either ECL reagent Immobilon Classico Western HRP substrate or Immobilon Forte Western HRP Substrate, as needed. Direct digital images were acquired with the Image Quant LAS 4000 mini (GE Healthcare, USA). After development, membranes containing insoluble protein extracts were stripped with Restore PLUS WB Stripping Buffer (Thermo Fisher Scientific) and incubated with rat monoclonal anti‐tubulin antibody (YOL1/34, 1:5000, Abcam), followed by a secondary goat polyclonal anti‐rat (1:5000, Thermo Fisher Scientific).

Cortical and hippocampal frozen samples from 60‐ to 95‐week‐old mice six AD (four females, two males) and six WT (six females) followed the same protocol as described above for soluble protein extracts from human frozen brain blocks, except that the following primary antibodies were used: rabbit polyclonal anti‐CD41 antibody (1:250, Proteintech, USA), and rat monoclonal anti‐tubulin antibody (YOL1/34, 1:5000, Abcam). Corresponding secondary antibodies included goat polyclonal anti‐rabbit antibody (1:1000, Agilent Dako) and goat polyclonal anti‐rat (1:5000, Thermo Fisher Scientific).

As appropriate, human or mouse platelets lysates and in vitro human fibrin clots ⁸ were run in parallel as positive controls. Tubulin served as loading control given its ubiquitous expression. Densitometry quantification analysis of three different WB experiments was carried out using the Fiji tool “Gel Analysis”.

2.7. Brain staining

Human brain staining—Human frontal cortex 5‐µm paraffin‐embedded sections were subjected to deparaffination and heat induce epitope retrieval with Tris‐ethylenediaminetetraacetic acid (EDTA) buffer (pH 9) and staining were performed by the CNIC Histopathology Core using an automated processing system. Endogenous peroxidase was quenched by incubation with 3% H₂O₂ and endogenous antigens blocked with 5% bovine serum albumin and 10% fetal bovine serum. A rabbit polyclonal anti‐CD41 antibody (1:200, Novus Biologicals, USA) was used and processed following the EnVision+ System HRP Labelled Polymer Anti‐Rabbit (Agilent Dako) protocol. Diaminobenzidine was used for visualization. Sections were counterstained with hematoxylin and scanned with the Axio Scan Z1® (Carl Zeiss, Germany). The total number of CD41‐positive platelets was quantified using the Fiji macro “Analyze Particles” (N = 11; mean area quantified per section = 286 mm²).

For fibrin/CD41 co‐staining, fibrin immunohistochemistry was performed as previously described, ⁸ with the key modification that tissue was fixed in 0.5% PFA (instead of 4% PFA plus proteinase K treatment), because subsequent CD41 was not compatible with proteinase K. Briefly, a subset of frozen human post mortem AD tissue was fixed in 0.5% PFA, cryoprotected in 30% sucrose/PFA and sectioned on a cryostat. Free‐floating 20‐µm sections were incubated in methanol/H₂O₂ to quench endogenous peroxidase activity, blocked in Tris buffer with 2% donkey:horse serum (1:1), and incubated overnight with a monoclonal anti‐fibrin antibody (59D8, 1:1000, Thermo Fisher Scientific). Sections were then incubated with ready‐to‐use EnVision+ System HRP Labelled Polymer Anti‐Mouse (Agilent Dako) and developed with ImmPACT DAB Peroxidase Substrate (Vector Labs). For subsequent Vector Blue labeling, sections were re‐blocked and incubated overnight at 4°C with a rabbit polyclonal anti‐CD41 (1:500, Novus Biologicals, USA). The following day, sections were incubated with alkaline phosphatase‐conjugated ImmPRESS‐AP Polymer Anti‐Rabbit IgG reagent and developed with Vector Blue substrate (Vector Laboratories) for 15 minutes, dehydrated, cleared in histoclear, coverslipped with fluoromount and scanned with the Axio Scan Z1® (Carl Zeiss).

Mouse brain staining—The contralateral cerebral hemisphere from the animals used for the WB experiments was fixed in 4% or 0.5% PFA. After 24 hours of fixation at 4°C, samples were cryoprotected in a solution of PFA with 30% sucrose, and frozen at ‐80°C until sliced. Free‐floating 20‐µm cryostat brain sections were obtained and kept frozen in cryoprotectant buffer until use. The 4% PFA‐fixed floating sections (three sections/mouse) from seven AD (six females, one male) and six WT (five females, one male) 95‐week‐old mice were subjected to immunohistochemistry, as previously described. ²⁵ Briefly, endogenous peroxidase activity was quenched by incubating the sections for 15 minutes in PBS/3% H₂O₂. After rinsing, non‐specific binding was blocked by incubating the tissues in PBS containing 0.1% Triton X‐100, 3% bovine serum albumin, 0.1 M lysine and 5% goat serum (Vector Laboratories, USA). Sections were incubated overnight at 4°C with a rat monoclonal anti‐CD41 (MWReg30, 1:500, BD Pharmingen, USA) followed by several washes and incubation for 1 h with a biotinylated goat anti‐rat secondary antibody (1:1000, Vector Laboratories). Sections were incubated for 1 h with the ImmunoPure Ultra‐Sensitive ABC Peroxidase Staining kit (Thermo Fisher Scientific), developed using the diaminobenzidine method, rinsed, mounted, dehydrated, cleared in xylene and coverslipped using DePex mounting medium. Counterstaining with 0.5% Thioflavin S was performed to identify Aβ. Slides were scanned with the Axio Scan Z1® (Carl Zeiss) and CD41‐positive cells were visually identified and counted using the NanoZoomer Digital Pathology software (Hamamatsu, Japan) across the entire cortex and hippocampus of AD and WT mice (N = 7 AD and N = 6 WT mice; mean area quantified per section = 6.26 mm²).

For fibrin/CD41 co‐staining, we followed the same protocol described above for CD41, except that mouse brain hemispheres were fixed in 0.5% PFA (fibrin(ogen) immunolabeling is suboptimal in 4% PFA‐fixed tissue). 0.5% PFA‐fixed floating sections (three sections/mouse) from three AD and three WT 70‐week‐old males were incubated with a rabbit polyclonal anti‐fibrin(ogen) antibody (1:500, gift from Dr. J. L. Degen, Cincinnati, OH, USA) followed by a biotinylated goat anti‐rabbit secondary antibody (1:1000, Thermo Fisher Scientific). For subsequent Vector Blue labeling, the same procedure described above for the human tissue was applied, using a rat monoclonal anti‐CD41 antibody (MWReg30, 1:500, BD Pharmingen, USA) and the ImmPRESS‐AP Polymer Anti‐Rat IgG reagent. Slides were scanned with the Axio Scan Z1® (Carl Zeiss).

For triple immunofluorescence (6E10/fibrin(ogen)/CD41), 0.5% PFA‐fixed‐free‐floating sections (three sections/mouse) from four AD (two females, two males) and four WT (two females, two males) 70‐week‐old mice were blocked in PBS containing 0.25% Triton X‐100 and 3% normal goat serum and incubated overnight at 4°C with rat monoclonal anti‐CD41 antibody (MWReg30, 1:500, BD Pharmingen, USA), rabbit polyclonal anti‐fibrin(ogen) antibody (1:500, gift from Dr. J. L. Degen, Cincinnati, OH, USA), and biotinylated anti‐Aβ 1‐16 (6E10, 1:1000, BioLegend, San Diego, CA, USA). Sections were then incubated for 1 hour with the following secondary antibodies; Cy3 goat anti‐rat (1:500, Jackson Laboratories, Cambridge, UK), Alexa Fluor 594 donkey anti‐rabbit (1:500, Thermo Fisher Scientific), and Alexa Fluor 647‐conjugated streptavidin (1:250, Thermo Fisher Scientific), counterstained with DAPI, mounted and coverslipped with Fluoromount‐G (SouthernBiotech, Birmingham, AL, USA). Images were acquired using a HC PL Apo CS2 63×/1.4 oil objective on a Leica TCS SP8 confocal microscope (Leica Microsystems, Mannheim, Germany) at the CNIC Microscopy Unit.

All procedures were carried out at RT unless otherwise specified, with thorough washes after steps.

2.8. Statistical analyses

Biodistribution and ROI data were analyzed using RStudio© (2025.05.1.513 Version, Posit Software, PBC, USA). Outliers were checked based on their Z‐scores (values exceeding ± 3), although no data were excluded from the analyses as no outliers were found. Non‐parametric statistics were applied when assumptions of normality and homoscedasticity were not accomplished after testing the data with Shapiro–Wilk's and Levene's tests, respectively. Therefore, to detect differences in the radiotracer uptake between groups, two‐tailed two‐sample T‐tests or Mann–Whitney U analyses (p < 0.05) were performed. The magnitude of the differences between groups was evaluated using Cohen's d as an effect size parameter. Furthermore, Pearson's or Spearman's correlation analyses were carried out to test the association between brain uptake and age. Data were corrected for injected radioactivity dose (ROI) and sex (ROI and biodistribution). Specifically, for [⁸⁹Zr]‐DFO‐FBP whole brain ROI data obtained with the iterative histogram normalization method, Z‐scores normalized to the respective WT group at each time point were calculated. This allowed us to compare the discriminatory power of the probe on the groups across time points.

Additionally, PET images were evaluated by voxel‐wise analysis using SPM12. ⁴⁹ Two‐sample T‐tests, with dose and sex as covariates, were performed to detect differences between groups. To account for multiple comparisons and enhance sensitivity to subtle effects, threshold‐free cluster enhancement ⁵⁰ was applied, estimated using 10,000 permutations. Results were considered statistically significant at p < 0.05 corrected for False discovery rate (FDR). Only significant clusters larger than 50 adjacent voxels were considered and represented as Cohen's d maps overlaid on an MRI brain template. Additionally, the mean ± SD values of Cohen's d maps for the statistically significant clusters are also provided to contextualize the mean effect size of the voxel‐wise differences.

The association between human CD41 and fibrin in WB densitometry measurements was performed by Spearman's correlation using RStudio (p < 0.05). We first corrected the densitometric values of CD41 and fibrin normalized to tubulin levels (from three independent WB experiments) by sex, post mortem interval, and experimental batch. The residual values for each subject were averaged to ensure independent observations. Additionally, using these adjusted fibrin and CD41 values from the human post mortem cases, Spearman's correlation analyses were performed with all available clinical variables shown in Table S1 in supporting information, while Mann–Whitney U tests were used to assess differences according to the presence of “Vascular Pathology” (as defined by BT‐CIEN neuropathologists) ⁵¹ , ⁵² in RStudio (p < 0.05).

Differences in CD41‐positive platelet counts between human AD frontal cortex sections from cases with and without cerebral fibrin were assessed by Mann–Whitney U test.

WB statistical analysis of mouse data was performed using RStudio following the same data screening protocol as for ROIs data. To ensure data independence, densitometric values of CD41 normalized to tubulin were averaged for each subject and analyzed using a two‐tailed t‐test comparing experimental (i.e., AD vs. WT) groups.

Differences in CD41‐positive platelet counts in AD versus WT mice were evaluated by two‐tailed, two‐sample t‐test, with values of p < 0.05 considered statistically significant. All values are presented as mean ± SD, and the effect size of group differences was quantified using Cohen's d.

3. RESULTS

3.1. Fibrin imaging in AD mice

To evaluate the dynamics of cerebral fibrin imaging in TgCRND8 AD mice, our initial approach assessed the specific brain uptake of FBP directly labeled with ⁸⁹Zr (i.e., [⁸⁹Zr]Zr‐DFO‐FBP). The radioisotope's 78.4‐hour half‐life allowed imaging the same animal during 5 consecutive days (Figure 1A). This design enabled us to identify the uptake interval that provided the greatest discriminatory power between AD and WT groups based on brain radiotracer signal. Figure 1B shows [⁸⁹Zr]Zr‐DFO‐FBP brain uptake across all time points, expressed as the z‐score of the uptake ratio. Although no statistically significant differences were observed between groups at any timepoint when values were normalized to a common, valid reference region that allows comparison across all timepoints, the greatest difference between WT and AD occurred at 24 hours post‐injection (AD mean normalized z‐scores: baseline = ‐0.25, 24‐hour = 0.49, 48‐hour = 0.23, 72‐hour = ‐0.38, 96‐hour = ‐0.17), which led us to select this timepoint for further analysis. At 24 hours post‐[⁸⁹Zr]Zr‐DFO‐FBP injection, using data normalized by a data‐driven method optimal for these cross‐sectional comparisons, all examined ROIs showed significant AD‐WT differences (Figure 1C and Table S2 in supporting information). The voxel‐wise analysis identified a single statistically significant cluster indicating a widespread increase in radiotracer uptake across the brain (Figure 1D, Cohen's d = 1.392 ± 0.343). In this sense, while preclinical PET enables detection of widespread pathological processes, its spatial resolution and partial volume effects constrain the ability to resolve highly focal microthrombi. Therefore, observed large clusters may represent the aggregation of multiple affected regions rather than uniform diffuse pathology, and caution should be taken when interpreting regional specificity.

FIGURE 1.

Fibrin PET imaging in TgCRND8 AD mice by a direct‐labeling approach. The [⁸⁹Zr]Zr‐DFO‐FBP probe enables in vivo detection of cerebral fibrin accumulation in TgCRND8 AD mice 24 hours after injection. (A) Study design. (B) Z‐scores of [⁸⁹Zr]Zr‐DFO‐FBP brain uptake (normalized to mean WT levels) in WT and AD mice over 5 consecutive days. The 24‐hour time point shows maximal discrimination between groups. (C) PET ROI analyses across brain regions 24 hours post‐injection comparing AD with WT mice (t‐test results: ***p < 0.005, ****p < 0.001; Mann–Whitney: ^#### p < 0.001). (D) Voxel‐based results shown as Cohen's d‐maps overlaid on a T2‐weighted MRI, indicating significant increase in cerebral [⁸⁹Zr]‐DFO‐FBP uptake in AD versus WT mice 24 hours post‐injection. Color bar represents Cohen's d‐values for increased brain [⁸⁹Zr]Zr‐DFO‐FBP uptake. Hemisphere labels: Right (R) and Left (L), k > 50 voxels, p < 0.05 (FDR‐corrected). AD, Alzheimer's disease; FBP, fibrin binding probe; FDR, false discovery rate; MRI, magnetic resonance imaging; PET, positron emission tomography; ROI, region of interest; WT, wild‐type; [⁸⁹Zr]Zr‐DFO‐FBP, zirconium‐89 radiolabeled fibrin binding probe modified with desferrioxamine B.

To enhance the clinical translatability of our approach, and since the 24‐hour PET acquisition provided the best discrimination between AD and WT groups after [⁸⁹Zr]Zr‐DFO‐FBP injection (Figure 1B), we developed a pre‐targeting strategy based on ⁶⁸Ga (half‐life 67.8 minutes) to reduce radiation dose while improving signal‐to‐noise ratio. We administered the FBP‐TCO 1 day prior to imaging, followed by injection of [⁶⁸Ga]Ga‐DOTA‐TZ 2 hours prior to PET acquisition (Figure 2A). Administering FBP‐TCO in advance allowed clearance of unbound probe, leaving only molecules specifically bound to their target (i.e., fibrin) when the radioligand ([⁶⁸Ga]Ga‐DOTA‐TZ) was administered closer to imaging acquisition, thereby improving PET signal. We observed a clear [⁶⁸Ga]Ga‐DOTA‐TZ signal emitted from the thrombus region in the carotid crush injury model used as C+, which was not present in animals operated but injected with saline (Figure 2B), confirming the specificity of the pre‐targeting FBP‐TCO + [⁶⁸Ga]Ga‐DOTA‐TZ strategy. We then evaluated differential brain uptake of the radiotracer between AD and WT mice, and also included a C‐ group based on the injection of DCys‐FBP‐TCO in AD mice to assess nonspecific tracer uptake. ROI analyses revealed a marked, widespread increase in tracer uptake across all examined brain regions in AD versus WT, while no significant differences were observed for AD versus C‐DCys or WT versus C‐DCys (Figure 2C and Table S3 in supporting information). Voxel‐wise analysis identified a single, statistically significant cluster spanning the entire brain, reflecting elevated tracer uptake in AD versus WT mice (Figure 2D, Cohen's d = 0.741 ± 0.368), but no differences were found between AD and C‐DCys mice (data not shown). However, ex vivo biodistribution, a much more sensitive measurement, showed significantly higher uptake in AD versus WT (Cohen's d = 1.510, p < 0.001), in AD versus C‐DCys (Cohen's d = 0.688, p = 0.049), and in C‐DCys versus WT (Cohen's d = 0.838, p = 0.015) animals (Figure 2E and Table S3 in supporting information), that remained markedly higher in the AD group at all ages compared to WT and C‐DCys (Spearman's ρ = 0.55, p = 0.023; Figure 2F and Table S4 in supporting information). These differences indicate specific fibrin‑related binding in AD mice that PET imaging could not fully resolve in vivo probably due to resolution power.

FIGURE 2.

Fibrin PET imaging in TgCRND8 AD mice by a pre‐targeting approach. The FBP‐TCO + [⁶⁸Ga]Ga‐DOTA‐TZ strategy enables in vivo detection of cerebral fibrin accumulation in TgCRND8 AD mice. (A) Study design. (B) Representative PET images of the carotid crush injury model (C+), showing radiotracer uptake in the injured carotid region of WT mice injected with FBP‐TCO + [⁶⁸Ga]Ga‐DOTA‐TZ. (C) PET ROI analyses across brain regions comparing the different experimental groups (t‐test: **p < 0.01; Mann–Whitney: ^# p < 0.05, ^## p < 0.01). (D) Voxel‐based results shown as Cohen's d‐maps overlaid on a T2‐weighted MRI, indicating significant increased cerebral [⁶⁸Ga]Ga‐DOTA‐TZ uptake in AD versus WT mice. Color bar represents Cohen's d‐values for increased brain [⁶⁸Ga]Ga‐DOTA‐TZ uptake. Hemisphere labels: Right (R) and Left (L), k > 50 voxels, p < 0.05 (FDR‐corrected). (E) Ex vivo biodistribution studies showing post mortem brain retention of [⁶⁸Ga]‐DOTA‐TZ measured by gamma‐counting after PET acquisition (Mann–Whitney: ^# p < 0.05, ^#### p < 0.001) and (F) plotted as a function of age, showing a positive association in AD mice (Spearman's ρ = 0.55, p = 0.023). AD, Alzheimer's disease; C‐DCys, negative control; FBP, fibrin binding probe; FDR, false discovery rate; MRI, magnetic resonance imaging; PET, positron emission tomography; ROI, region of interest; TCO, transcyclooctene; TZ, tetrazine; WT, wild‐type; [⁶⁸Ga]Ga‐DOTA‐TZ: gallium‐68 radiolabeled 1,4,7,10‐tetraazacyclododecane‐1,4,7,10‐tetraacetic acid conjugated to tetrazine.

3.2. Platelets in the AD procoagulant brain

FBP preferentially binds newly formed thrombi rather than established clots, ⁵³ potentially underestimating the total cerebral clot burden in AD. We therefore sought an alternative and complementary thrombotic target to enhance PET signal and evaluated platelets as potential diagnostic markers for this cerebrovascular dysfunction. Since the specific contribution of platelets to AD remains underdefined, ¹⁴ , ²³ we first quantified platelet content in brain tissue from AD patients and TgCRND8 AD mice. To this end, we evaluated platelet content by CD41 expression in post mortem frontal cortex tissue from AD patients and assessed its correlation with cerebral fibrin levels. WB analysis showed heterogeneity in the amount of CD41 across AD patients (Figure 3A and B) but higher CD41 levels were detected in AD samples with greater cerebral fibrin content, revealing a significant positive correlation (Figure 3C; Spearman's ρ = 0.343, p = 0.033). This finding further underscores the complexity of AD pathology and indicates that platelet content may serve as an effective in vivo target of the AD‐associated procoagulant state. Notably, half of AD patients exhibited abnormal insoluble cerebral fibrin levels by qualitative assessment of the WB shown in Figure 3A (i.e., 48.72%), consistent with our previous findings in AD samples from an independent brain bank. ⁸ We also performed exploratory analyses assessing whether the clinical variables from these human post mortem cases (Table S1 in supporting information) were significantly associated with the AD‐procoagulant phenotype. We did not find any significant association with bearing the Apolipoprotein E4 allele, with medication or with comorbidities. However, the levels of cerebral fibrin burden were significantly increased in those patients diagnosed with “Vascular Pathology” (Cohen's d = 0.437, p < 0.001). These exploratory analyses should be interpreted with caution, given the limited sample size (N = 39). Future studies in larger and specifically powered cohorts will be required to define the relationship between demographic factors, cerebral fibrin/CD41 burden, and cerebrovascular pathology in AD.

FIGURE 3.

Platelet deposition in the human AD brain and correlation with the AD‐associated cerebral fibrin burden. WB and immunohistochemistry analyses illustrate the accumulation of fibrin and platelet marker CD41 in the frontal cortex of post mortem samples from AD patients. (A) Representative WB of soluble (CD41) and insoluble (fibrin) protein extracts from the frontal cortex of AD patients (arrows point to the corresponding quantified band). Blots were re‐probed with anti‐tubulin antibody as loading control. Human platelet lysate and an in vitro‐generated human fibrin clot were run in parallel as positive controls, respectively (C+). Each lane corresponds to an individual AD patient. (B) Combined violin and box plots showing the distributions of fibrin/tubulin and CD41/tubulin ratios from the WB quantification in panel A. Each dot represents an individual patient. The fibrin/tubulin ratio of one patient exceeded the y‐axis scale (ratio = 68.11) and is not displayed in the violin plot for better visualization. (C) Scatter plot showing the relationship between CD41 and fibrin WB levels in each AD patient, indicating a positive significant correlation (Spearman's ρ = 0.343, p = 0.033). CD41/tubulin and fibrin/tubulin ratios from three independent WB experiments are represented as averaged residuals for each subject. N = 39 AD patients. (D) Representative CD41 immunohistochemistry images of paraffin‐embedded frontal cortex sections from a subset of the AD patients used in (A) (CD41, brown; hematoxylin counterstaining nuclei, blue). Intravascular platelets are found in AD patients with cerebral fibrin deposition in WB analysis (A), partially occluding large vessels or blocking small capillaries. In contrast, CD41 staining shows only sporadic single platelets or punctate CD41 immunoreactivity in AD patients whose frontal cortex lacked detectable fibrin by WB (A). Scale bar, 25 µm. (E) Representative double immunostaining images of PFA‐fixed frontal cortex sections showing CD41‐positive cells (blue) co‐localizing with fibrin (brown) in AD patients with cerebral fibrin. Scale bars, 25 µm and 50 µm. AD, Alzheimer's disease; PFA, paraformaldehyde; WB, Western blot.

To further characterize platelets in the human AD brain, we performed immunohistochemistry on frontal cortex post mortem human AD samples in a subset of AD cases with high levels of fibrin and compared them with cases lacking cerebral fibrin as determined by WB (Figure 3A). In AD patients with cerebral fibrin burden, CD41 staining was robust, with extensive immunoreactivity primarily intravascular, either occluding small capillaries or partially blocking larger vessels, whereas AD samples without cerebral fibrin showed minimal CD41 signal, with only a few scattered platelets across the brain section (Figure 3D; 21.38 ± 4.73 platelets/mm² versus 10.66 ± 4.83 platelets/mm² in AD cases with high or none cerebral fibrin burden, respectively; Cohen's d = 2.24, p = 0.009). Interestingly, CD41 staining co‐localized with fibrin‐positive areas in the frontal cortex of AD patients exhibiting a procoagulant state (Figure 3E). These results indicate that platelets are present in the subset of AD patients with a procoagulant state, underscoring their potential as biomarkers and therapeutic targets.

Having identified increased platelet content in the subset of AD patients with cerebral fibrin deposition, we next examined whether this feature was also present in TgCRND8 mice, the AD model used for our in vivo PET imaging (Figures 1 and 2). WB analysis showed elevated CD41 levels in cortical and hippocampal TgCRND8 samples relative to age‐matched WT littermates (Figure 4A) and quantification confirmed a significant increase in AD mice (Figure 4B; Cohen's d = 1.79, p = 0.015). To further characterize cerebral platelet accumulation in the AD mouse model, we performed CD41 immunohistochemistry on brain sections from WT and TgCRND8 AD mice. AD mice displayed abundant intravascular platelets, frequently stalling small capillaries in the cortex and hippocampus and occasionally near amyloid plaques, whereas WT mice exhibited only sparse platelets (Figure 4C; 5.69 ± 1.83 platelets/mm² versus 3.30 ± 1.50 platelets/mm² in TgCRND8 and WT mice, respectively; Cohen's d = 1.36, p = 0.029). Interestingly, CD41 staining was found with fibrin(ogen)‐positive thrombotic deposits across the cortex and hippocampus of TgCRND8 mice (Figure 4D), near amyloid plaques (Figure 4E).

FIGURE 4.

Platelet deposition in the mouse AD brain. WB and immunohistochemistry analyses show platelet accumulation in the cortex and hippocampus of AD mice. (A) Representative CD41 WB of soluble cortical and hippocampal protein extracts from WT and TgCRND8 AD mice. Blots were re‐probed with anti‐tubulin antibody as loading control. Mouse platelet lysate was included as positive control (C+). (B) Three different WB experiments were quantified and averaged values for each mouse plotted. CD41 levels were significantly increased in TgCRND8 AD brains compared to age‐matched WT littermate controls (N_AD = 6, N_WT = 6; T‐test: *p < 0.05). (C) Representative immunohistochemistry images showing numerous CD41‐positive platelets (brown) obstructing small vessels or co‐localizing with Aβ deposits (Thioflavin S counterstain, green) in the cortex and hippocampus of TgCRND8 AD mice. In contrast, corresponding regions from age‐matched WT littermates display little to none CD41 staining. (D) Representative double immunostaining images showing CD41‐positive cells (blue) co‐localizing with fibrin(ogen) (brown) in TgCRND8 AD mice, whereas WT mice show sparse fibrin(ogen) and CD41 staining. Scale bar, 25 µm. (E) Representative confocal images of triple immunofluorescence in the cortex of TgCRND8 mice show fibrin(ogen)‐positive structures (red) associated with platelets (CD41, white) in the presence of Aβ (6E10, cyan). Scale bar, 10 µm. Aβ, amyloid‐β; AD, Alzheimer's disease; WB, Western blot; WT, wild‐type.

3.3. Platelet imaging

The increased platelet presence in the brains of AD patients with a procoagulant state (Figure 3) and TgCRND8 mice (Figure 4) supported the development of a suitable biomarker and prompted us to investigate the in vivo detection of platelet accumulation by PET imaging in this mouse model. Building on our previous imaging results, we directly employed a pre‐targeting strategy using Thrombo‐tag, consisting of [⁶⁸Ga]Ga‐IONPs and a CD41‐specific antiplatelet antibody bound in vivo by click chemistry. ⁴³ The anti–CD41‐TCO was injected the day before imaging to allow distribution and clearance of unbound antibody, thereby maximizing the signal from target‐bound platelets when [⁶⁸Ga]Ga‐IONP‐TZ was administered 2 hours prior to PET acquisition (Figure 5A).

FIGURE 5.

Platelet PET imaging in the brain of TgCRND8 AD mice by a pre‐targeting approach. Thrombo‐tag (i.e., anti–CD41‐TCO + [⁶⁸Ga]Ga‐IONP‐TZ) enables in vivo cerebral detection of platelet accumulation in TgCRND8 AD mice. (A) Study design. (B) Representative PET images of the carotid crush injury model (C+) showing high radiotracer uptake in the injured carotid region of WT mice injected with the anti–CD41‐TCO + [⁶⁸Ga]Ga‐IONP‐TZ. (C) PET ROI analyses across brain regions comparing AD versus WT mice (t‐test: *p < 0.05, ****p < 0.001; Mann–Whitney: ^# p < 0.05, ^## p < 0.01, ^#### p < 0.001). Voxel‐based results shown as Cohen's d‐maps overlaid on a T2‐weighted MRI, indicating significant increases in [⁶⁸Ga]Ga‐IONP‐TZ uptake in AD versus WT (D) and AD versus C‐IgG (E) mice. Color bar represents Cohen's d‐values for increased cerebral Thrombo‐tag uptake. Hemisphere labels: Right (R) and Left (L), k > 50 voxels, p < 0.05 (FDR‐corrected). AD, Alzheimer's disease; C‐IgG, negative control; FDR, false discovery rate; MRI, magnetic resonance imaging; PET, positron emission tomography; ROI, region of interest; TCO, transcyclooctene; TZ, tetrazine; WT, wild‐type; [⁶⁸Ga]Ga‐IONP‐TZ: tetrazine functionalized iron oxide nanoparticles doped with 68‐gallium.

As expected, PET images of the carotid crush injury model revealed a positive signal at the thrombotic site in animals injected with the full functional probe due to successful biorthogonal reaction, while control animals injected with saline showed no specific signal in the lesioned area (Figure 5B), hence demonstrating the performance of our pre‐targeting approach. Consistent with the FBP results (Figure 2), ROI analyses showed a robust, widespread increase in Thrombo‐tag uptake across all examined brain regions in AD mice relative to WT controls (Figure 5C and Table S5 in supporting information). Furthermore, the AD group showed statistically significant increases compared to C‐IgG mice in all ROIs but cerebellum and thalamus, with the effect being most pronounced in the cortex and hippocampus, while no differences were observed between WT and C‐IgG in any of the ROIs analyzed (Figure 5C and Table S5 in supporting information). Voxel‐wise analysis identified a single, brain‐wide, statistically significant cluster, indicating consistently higher activity in AD versus WT mice (Figure 5C, Cohen's d = 0.967 ± 0.162). When comparing AD versus C‐IgG mice, voxel‐wise analysis showed again a single cluster along the brain but more restricted to cortical and hippocampal areas (Figure 5D, Cohen's d = 1.051 ± 0.246), supporting again the relevance of these regions in the AD procoagulant pathology. These in vivo data were further corroborated by ex vivo biodistribution measurements (Figure 6A and Table S5 in supporting information), which not only showed a significant increase in brain Thrombo‐tag uptake in AD versus WT (Cohen's d = 0.925, p = 0.001) or C‐IgG (Cohen's d = 0.858, p = 0.013) mice, but also displayed a positive association between age and Thrombo‐tag uptake in AD mice (Spearman's ρ = 0.823, p < 0.001, Figure 6B and Table S6 in supporting information). To further examine age‐related patterns of Thrombo‐tag uptake in vivo within each experimental group, we analyzed radiotracer uptake measured by PET in the whole brain, cortex and hippocampus ROIs as a function of age (Figure 6C and Table S6 in supporting information). In the AD group, Thrombo‐tag uptake increased significantly with advancing age in the cortex (Spearman's ρ = 0.551, p = 0.034) and hippocampus (Pearson's R = 0.727, p = 0.002), while no similar tendencies were found neither in WT nor in C‐IgG animals. Together, these in vivo data demonstrate elevated Thrombo‐tag specific cerebral uptake in TgCRND8 mice with a strong pathology‐burden dependence.

FIGURE 6.

Age‐related platelet radiotracer brain uptake in AD mice. (A) Ex vivo biodistribution studies showing post mortem brain retention of Thrombo‐tag measured by gamma‐counting after PET acquisition (Mann–Whitney results: ^# p < 0.05, ^## p < 0.01) (B) Ex vivo measurements of the radiotracer uptake in the brain as a function of age, showing a strong positive correlation in AD mice (Spearman's ρ = 0.823, p < 0.001). (C) Correlation analyses between age and in vivo Thrombo‐tag PET signal in the whole brain (left), cortex (middle) and hippocampus (right) showing a significant positive association in AD mice (cortex: Spearman's ρ = 0.551, p = 0.034; hippocampus: Pearson's R = 0.727, p = 0.002). AD, Alzheimer's disease; C‐IgG, negative control; PET, positron emission tomography.

4. DISCUSSION

The primary contribution of our study lies in the development of specific PET imaging probes for detecting the procoagulant state in AD. We show that fibrin‐ and platelet‐targeted probes present increased brain uptake in AD animals with a procoagulant state compared to controls, establishing the first in vivo detection of the AD‐associated procoagulant environment (Graphical Abstract). We also provide evidence of significantly increased platelet accumulation in the AD brain concomitant with the AD's procoagulant state, both in patients and in TgCRND8 mice. Together, these findings represent a significant step toward non‐invasive diagnostic strategies for the AD‐procoagulant state, with clear translational potential for clinical practice.

In 2016, the AT(N) system was proposed as a neuropathological classification of AD based on Aβ deposition (A), tau pathology (T), and neurodegeneration (N). ⁵⁴ This framework led to a redefinition of AD from a clinical syndrome to a neuropathological construct, ⁵⁵ emphasizing the pivotal role of imaging and fluid biomarkers in AD research, diagnosis and therapy. ⁵⁵ Specifically, PET neuroimaging of Aβ and tau has been critical for AD diagnosis and staging, mapping temporospatial progression, and enabling disease‐specific trials. ⁵⁶ , ⁵⁷ More recently, the Alzheimer's Association expanded the classification system to include inflammatory/immune mechanisms (I), vascular brain injury (V), and α‐synucleinopathy (S), ⁵⁸ reflecting the aim to integrate AD's multifactoriality in its conceptualization. Beyond Aβ and tau radiotracers, other PET probes are being developed for proteinopathies (e.g., α‐synuclein, TDP‐43⁵⁷), for neuroinflammation targeting active microglia (e.g., TSPO, cyclooxygenase, CSF‐1) or astrogliosis (e.g., MAO‐B, imidazol‐2 binding site), ⁵⁹ and for synaptic disfunction/degeneration (e.g., SV2A, ⁶⁰ VAChT ⁶¹ ). However, all these PET probes remain largely in preclinical phases, ⁵⁷ , ⁶² and the only neuroimaging approved AD PET tracers are restricted to Aβ and tau, limiting current diagnostic capacity.

Hence, deploy imaging strategies sensitive enough to detect widespread, potentially chronic cerebral microthrombi is essential. As reviewed by Stendahl et al, ³⁶ multiple radiotracers have been developed for thrombus detection by nuclear imaging. Probes targeting fibrin, ⁶³ , ⁶⁴ , ⁶⁵ activated platelets, ⁶⁶ , ⁶⁷ , ⁶⁸ , ⁶⁹ and factor XIIIa ⁷⁰ show promise for detecting focal, large‐vessel thrombotic events. However, most of these probes are biased toward sites of active thrombosis, ³⁶ , ⁶⁹ , ⁷⁰ limiting detection of chronic or mature thrombi, and none has been comprehensively evaluated in the diffuse prothrombotic milieu expected in AD. Direct comparison with our imaging approaches is not yet feasible, and head‐to‐head studies will be required to benchmark these PET probes in the AD‐procoagulant context. Here, using fibrin‐ and platelet‐targeted PET probes, we detected higher cerebral uptake in vivo in TgCRND8 mice in comparison with their WT littermates. Cortical and key subcortical regions (e.g., hippocampus) exhibited pronounced increases in radiotracer uptake, indicating substantial involvement of procoagulant mechanisms in areas critical for cognition and affected early in AD. ⁷¹ , ⁷² Consistent with reports of elevated fibrin(ogen) in these regions in AD human and murine tissue, ⁸ our results demonstrate in vivo detection of this cerebrovascular dysfunction. Importantly, probe uptake positively correlated with age only in AD mice, supporting that the PET signal tracks pathology progression rather than nonspecific aging. The consistent increase with two mechanistically distinct tracers supports the specificity of our approach in detecting the AD procoagulant state.

For clinical translation, platelet imaging may be preferable as platelets are predominantly intravascular in both human and mouse AD brain (Figures 3 and 4), whereas fibrin also shows strong intraparenchymal staining ⁸ that our FBP‐PET approach likely misses. Sensitivity for fibrin imaging could be improved by BBB‐penetrating fibrin probes, such as the recently reported proof‐of‐concept bispecific tracers ⁷³ . Also, FBP uptake decrease as clots mature, ⁵³ whereas Thrombo‐tag targets the constitutive platelet antigen CD41 and captures both forming and chronic clots. Combined with FBP's micromolar affinity, ⁴² these features likely contribute to the lower in vivo sensitivity of the FBP‐based pre‐targeting PET approach and the lack of significant AD versus C‐DCys brain differences by PET. In contrast, Thrombo‐tag leverages a nanoparticle platform ⁴³ that enhances click‐chemistry target engagement, likely consistent with its larger in vivo effect size. Cohen's d > 0.8 are considered “large” effects, ⁷⁴ aligned with Thrombo‐tag's performance. While Aβ/tau PET typically yields very large effects (d≈2‐3), ⁷⁵ , ⁷⁶ , ⁷⁷ , ⁷⁸ , ⁷⁹ , neuroinflammation ⁸⁰ , ⁸¹ and synaptic dysfunction ⁶⁰ , ⁶¹ PET tracers often show moderate yet meaningful effects (d≈0.7–1.3, calculated from reported data). These comparisons underscore the strong discriminatory power of our probes and position Thrombo‐tag as the leading biomarker candidate to detect a generalized cerebral procoagulant state in AD. Importantly, preclinical Aβ PET experience indicates that moderate‐to‐large regional differences in mice provide a robust in vivo readout of disease‐relevant pathology and progression, ⁸² , ⁸³ and clinical Aβ PET demonstrates consistent regional retention across tracers/cohorts, supporting translational relevance for diagnosis and longitudinal follow‐up. ⁸⁴ , ⁸⁵ Accordingly, although our effect sizes are smaller than those typically reported for Aβ/tau tracers, they are comparable to established pathophysiology‐targeted tracers and should be detectable in humans, especially in longitudinal designs.

In this context, our study provides the first quantitative evidence of increased platelet accumulation in the human and mouse AD brain and its correlation with cerebral fibrin deposition. The literature on platelets in AD is controversial, likely reflecting their multifaceted nature and stage‐dependent roles across the AD continuum. ²³ Most studies have focused on circulating platelets, whereas evidence of platelet presence within the AD brain, although reported in patients ⁸⁶ and mice, ¹⁹ , ⁸⁷ remains limited. Notably, no study has demonstrated a correlation between brain platelets and the increased cerebral fibrin observed in AD, an association that would be expected a priori, given the well‐established interplay between platelets and coagulation. In AD, however, platelets recruitment may further exacerbate the procoagulant milieu by promoting aggregation and fibrinogen binding, triggering coagulation pathways and further increasing local platelet levels, which may constitute an integral part of the thrombi formed in AD brains. ⁸ , ⁹ , ⁸⁸ Moreover, platelets may contribute to thrombus persistence in AD ⁹ through local Aβ release, facilitating Aβ–fibrin interactions that stiffen clots and exacerbate CAA‐related vascular injury. ¹⁹ , ⁸⁶ Furthermore, elevated platelet activation associates with worse cognitive outcomes ¹⁸ and fibrin has a well‐established causal role in AD pathophysiology, ²³ , ³⁰ including effects on cerebral hypoperfusion. ²⁵ Collectively, these data support a fibrin‐platelet axis of high diagnostic relevance in AD. Here, we introduce a novel non‐invasive neuroimaging strategy designed to detect this pathological process and ultimately enable personalized therapeutic interventions.

Our strategy may help distinguish AD from other neurodegenerative conditions because fibrinogen's involvement is particularly compelling: Aβ binds fibrinogen, ¹⁰ promotes fibrinogen oligomerization, ¹⁰ impairs fibrinolysis, ¹¹ , ¹² and disruption of this interaction rescues thrombosis and cognitive decline in AD mice. ⁸⁹ The procoagulant phenotype is also linked to vascular Aβ deposition, as lowering fibrinogen attenuates CAA ⁹ and CAA‐linked Aβ mutations show increased fibrinogen affinity. ²⁴ Thus, CAA may create focal vascular niches that favor Aβ–fibrinogen–platelet interactions, which our fibrin‐ and platelet‐PET signals may partially reflect.

We acknowledge that these tracers could detect thrombotic processes beyond AD; however, our readout is brain‐restricted, reducing confounding by systemic thrombosis. Clinically, these PET measures should be interpreted alongside established AD biomarkers (Aβ/tau/neurodegeneration PET and/or fluid‐based biomarkers) to capture comprehensive profiling, patient stratification, and mechanism‐guided therapeutic targeting, including anticoagulant, antiplatelet and/or fibrin immunotherapy when appropriate.

This diagnostic strategy is particularly relevant because, although coagulation can be targeted therapeutically, cerebral fibrin burden varies widely across AD patients (Figure 3 and Cortes‐Canteli et al ⁸ ). Identifying those most likely to benefit is critical to maximize efficacy and minimize bleeding risk. Several anticoagulants and emerging antiplatelet therapies have shown potential to slow disease progression in patients and AD mouse models. ⁷ , ²³ In parallel, fibrin‐targeted immunotherapy is promising: a monoclonal antibody against fibrin's pro‐inflammatory epitope (without affecting coagulation) reduced neuroinflammation and provided neuroprotection in AD mice, ³² and a higher‐affinity humanized version is now in Phase 1 trials. ³³ Regardless of the therapeutic modality, AD patients differ in cerebral fibrin burden, highlighting the importance of the diagnostic tool described here.

Finally, we must acknowledge several limitations of our study. First, FBP is a small peptide with high clearance rate via the urinary system, which may have resulted in partial loss of specific signal during imaging. Second, IONPs exhibit a long half‐life in the body, particularly in the liver as the main organ of elimination. However, these nanoparticles have been shown to be safe and non‐toxic, ⁹⁰ and have received approval by the United States Food and Drug Administration for various medical applications. ⁹¹ , ⁹² Third, amyloid overexpression in TgCRND8 mice might amplify fibrin–platelet interactions, consistent with evidence that Aβ binds fibrinogen ¹⁰ and that platelets are the main source of Aβ in blood. ¹⁵ However, we expect similar fibrin and platelet PET signals in tau‐based models given that fibrinogen interacts with tau enhancing cytotoxicity, tau phosphorylation/aggregation and neuronal dysfunction, ⁹³ and tau hyperphosphorylation/deposition interacts with components of the AD‐disrupted BBB. ⁹⁴ Finally, we acknowledge that disease‐related radiotracer uptake thresholds must be established in subsequent studies to enable the clinical translation of this approach as a diagnostic biomarker.

In conclusion, this study provides the first in vivo neuroimaging approach to identify the AD‐associated procoagulant environment through the development of highly specific PET imaging biomarkers. These results not only advance our understanding of the cerebrovascular contributions to AD but also open the door to non‐invasive diagnostic approaches with tangible translational potential for patient stratification and therapeutic guidance.

CONFLICT OF INTEREST STATEMENT

Author disclosures are available in the Supporting Information.

DISCLOSURES

The authors declare no competing financial interests.

CONSENT STATEMENT

All post mortem human samples were provided by the Biobank Banco de Tejidos CIEN (BT‐CIEN), integrated in the Spanish National Biobanks Network. All samples were obtained under informed consent from the donor or, when this was not possible, from a legal representative or next of kin and they were processed following standard operating procedures with the appropriate approval of the Ethics and Scientific Committees.

Supporting information

Supporting information: alz71652‐Sup‐0001‐ICMJE forms.pdf

Supporting information: alz71652‐Supp‐0002‐SuppMatt.pdf

Ceron C, Casquero‐Veiga M, Lamanna‐Rama N, et al. Unveiling the procoagulant state in Alzheimer's disease: A novel PET imaging strategy. Alzheimer's Dement. 2026;22:e71652. 10.1002/alz.71652