Repeated mild spinal cord contusions exacerbate tauopathy development in PS19 mice

Abstract

Growing evidence indicates that traumatic brain injury constitutes a significant risk factor for the development of age-related tauopathies, yet the consequences of repeated mild spinal cord injuries (rmSCI) remain insufficiently explored. In this context, we developed a mouse model undergoing two successive mild cervical contusions during young adulthood, which do not produce immediate functional motor deficits, thereby mimicking real-life conditions of occupational- or sports-related trauma. rmSCI were induced in PS19 (hTau^P301S) mice developing late-onset tauopathy in the brain and in the spinal cord. The consequences of rmSCI were assessed on motor outcomes and tauopathy signature in injured PS19 mice during ageing. rmSCI exacerbated age-dependent motor deficits and significantly increased tau hyperphosphorylation on the pSer422 and pSer202/Thr205 epitopes. Pathological changes, initially confined to the epicenter of the contusions, extended along a caudo-rostral axis reaching the thalamus. While rmSCI did not modify intraspinal tau aggregation, spinal cord-derived protein extracts displayed enhanced seeding activity in vitro. Although causality was not established with tau hyperphosphorylation status, early response towards mild spinal contusions included p38 MAPK activation, glial activation and upregulation of interferon-stimulated genes. Together, these data identify rmSCI as a previously underappreciated modifier of tau pathology and disease progression. Moreover, this study provides a novel experimental link between mild central nervous system injuries and exacerbation of tauopathy and suggests the involvement of type I interferon signalling or sustained glial activation. Ultimately, these findings underscore the importance of preventing even mild spinal injuries in individuals at risk of tauopathy.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s40478-026-02378-9.

Introduction

As life expectancy continues to rise with modern lifestyles, populations are increasingly exposed to the risk of age-related neurodegenerative disorders, such as tauopathies. In this family of disorders, Tau protein undergoes abnormal post-translational modifications, including pathognomonic hyperphosphorylations that trigger conformational changes and its destabilization from the microtubule network. As a result, Tau becomes misfolded, relocates to the perikaryon, in which it progressively accumulates into neurotoxic oligomers, paired helical filaments (PHFs), and higher-order aggregates known as neurofibrillary tangles (NFTs), leading to neuronal dysfunction and death [1, 26, 35].

Among tauopathies, chronic traumatic encephalopathy (CTE) was brought to light thanks to the seminal work of Dr. Bennet Omalu, who described the condition in post-mortem brains of National Football League players, with clinical symptoms such as mood disorders, cognitive impairment or dementia [38–41]. The development of CTE has been associated with single, episodic, or repetitive mild or severe head trauma that transmits acceleration–deceleration forces to the brain [38]. This syndrome was further characterized using experimental models. In 2013, a first report demonstrated that repetitive mild traumatic brain injury (TBI) exacerbated Tau pathology in hTau mouse model expressing wild-type human Tau on a murine Tau knockout background [37]. Next, it was reported that moderate to severe TBI enhanced Tau pathology progression and spreading, and exacerbated long-term memory impairments, in the PS19 transgenic mouse model [12]. The association between TBI and an increased risk of dementia has been confirmed in several human cohort studies [13, 19], although findings across studies have been somewhat inconsistent, likely due to various confounding factors such as genetic mutations, drugs, age heterogeneity, environmental factors, or post-mortem delay [10, 14, 36]. CTE is now defined as a progressive post-traumatic neurodegenerative syndrome driven by abnormal Tau phosphorylations and deposits in the brain, and more recently associated with TDP-43 proteinopathy [31].

The spinal cord, as an integral component of the central nervous system, shares substantial structural and molecular similarities with the brain and is likewise vulnerable to tauopathies [27, 38, 56]. Notably, patients with Alzheimer’s disease exhibit deposits of hyperphosphorylated Tau and NFTs within the spinal cord [56]. In this context, spinal cord has been largely underexplored; only a few studies have investigated a potential link between spinal cord injury (SCI) and the development of tauopathies [4, 22, 23, 33, 44, 47]. Among these, experimental spinal hemisection in rats has been shown to induce early Tau phosphorylation [22, 23], that was associated with the activation of calpain and cyclin-dependent kinase 5 [22]. In a rat model of spinal cord compression, Tau hyperphosphorylation was observed to extend up to 5 mm around the lesion epicenter [4]. Similarly, spinal compression in Balb/c mice triggered abnormal Tau phosphorylation, and the highly pathological cis-pTau conformation [33]. Interestingly, this characteristic tauopathy-related signature was also detected in distant brain regions, as further demonstrated in a dorsal column injury model [47]. More recently, these findings were corroborated by demonstrating acute pTau upregulation in a ferret model of blast-induced SCI [44]. Overall, these studies predominantly relied on severe injury models [22, 23, 33, 44], or focused on the identification of tauopathy-related biomarkers, rather than comprehensively addressing their contribution to the progression of tauopathies [4].

Given the limited body of evidence, we aimed to investigate the impact of repeated mild spinal cord injury (rmSCI) on the progression of Tau pathology. To this end, we developed a mouse model involving two mild spinal contusions that do not produce immediate functional deficits. This paradigm was designed to mimic real-life conditions (e.g., occupational or sports-related settings) in which individuals may experience recurrent low-grade spinal trauma without overt clinical symptoms yet remain at risk for delayed neurological sequelae. Among the spontaneously-developing tauopathy models, we selected the PS19 (B6;C3-Tg(Prnp–MAPT*P301S)PS19Vle/J) transgenic mouse model of frontotemporal dementia, which overexpresses a mutant form of human Tau (P301S) in addition to endogenous murine Tau. Mice were euthanized at multiple time points to assess the long-term effects of rmSCI on the onset and progression of Tau pathology.

Material and methods

Animals and experimental design

All mice used in this study were heterozygous PS19 (B6;C3-Tg(Prnp–MAPT*P301S)PS19Vle/J) transgenic animals obtained from The Jackson Laboratory and backcrossed with C57BL/6 J mice from an in-house breeding colony. All experimental procedures were approved by the University of Namur Ethics Committee (Ethics Projects UN 20/362 and 23/403) and conducted in accordance with European directives (2010/63/EU, 86/609/EEC, and 87/848/EEC), as well as the ARRIVE Essential 10 guidelines. Mice were housed under controlled temperature conditions with a 12 h light/12 h dark cycle and had ad libitum access to food and water.

At two months of age (designated as day 0 of the experiment), mice underwent either a mild midline contusion at the C5 spinal level or a sham procedure consisting of laminectomy only (Fig. 1A). Three weeks later, animals received a second contusion or sham surgery. Mice were subsequently monitored for periods ranging from 2 to 224 days following the initial procedure. Animals were euthanized by an overdose of ketamine (120 mg/kg; Nimatek) and xylazine (8 mg/kg; Sedaxylan). Under deep anesthesia, mice were transcardially perfused with cold phosphate-buffered saline (PBS). For tissues intended for histological analyses, perfusion was followed by 4% paraformaldehyde (PFA), and harvested organs were post-fixed in the same fixative for 48 h. For all other applications, tissues were rapidly dissected, snap-frozen in liquid nitrogen, and stored at − 80 °C until further processing.

Fig. 1.

Motor and clinical outcomes in PS19 mice undergoing sham surgeries or rmSCI. A Timeline expressed in days of experiment (from day 0 to day 224) or expressed in months for the age of mice. Sham surgeries or rmSCI at a force of 30 kDyn are scheduled on day 0 and day 21 (vertical black arrows). B rmSCI in a first cohort of PS19 mice produces no immediate motor deficits, and no further difference in forelimb grip strength up to day 126. n = 18 or 7 mice/group. C Same rmSCI protocol in a second cohort of PS19 mice produces a shift and an exacerbation of motor decline between day 140 and day 224, compared to sham mice. n = 12–8 mice for sham and n = 10–6 mice for rmSCI (n decreasing over time due to animals reaching the humane endpoint euthanasia). D Evolution of clinical scoring in both groups of mice. n = 16 mice for sham and n = 15 mice for rmSCI. See Table 1 for detailed clinical criteria included in the scoring system spanning from 0 (normal) to 6 (moribund, humane endpoint). E Humane endpoint-free survival represented as a Kaplan–Meier curve for each group of mice. PS19 mice reaching the humane endpoint of 6 in the scoring grid are euthanized according to ethic rules. n = 16 mice for sham and n = 15 for rmSCI. Data are presented as mean ± SD. *p < 0.05, mixed-effects model followed by Tukey’s multiple comparison test

Surgery

Preoperative analgesia was administered via subcutaneous injections of buprenorphine (0.05 mg/kg; Vetergesic) and carprofen (5 mg/kg; Rycarfa). Anesthesia was induced by intraperitoneal administration of ketamine (120 mg/kg; Nimatek) and xylazine (8 mg/kg; Sedaxylan), which maintained mice sedated for approximately 90 min. Spinal cord contusions at the C5 level were performed with minor adaptations from previously described protocols [29, 48]. Briefly, the surgical area was shaved and disinfected with povidone-iodine. A midline dorsal incision was made through the skin and underlying musculature, and the dorsal and interspinous muscles were retracted between the C3 and C6 vertebrae. A bilateral laminectomy was performed at the C5 vertebral level to expose the spinal cord. A 30 kDynes midline contusion was then delivered using an IH-0400 Infinite Horizon Impactor (Precision Systems and Instrumentation) equipped with a 1 mm diameter tip. A biocompatible membrane (Biobrane; Smith & Nephew, 66,800,174) was placed over the exposed spinal cord. Muscle layers were sutured using 5–0 absorbable thread, and the skin was closed with 9 mm wound clips (Bioseb, 205016). Postoperative analgesia included a second injection of buprenorphine at 6 and 24 h post-surgery, as well as additional carprofen administrations at 24 and 48 h. Sham-operated animals underwent identical procedures, except for the contusion step. Three weeks later, a second surgery was performed following the same protocol. Animals were monitored daily after each surgical procedure.

Behavioral tests

Grip strength test. Forelimb muscle strength was assessed using a Grip Strength Meter (Bio-GS3, Bioseb, France). Measurements were obtained for each forelimb individually by allowing the animal to grasp a horizontal bar connected to a force transducer. The mouse was then gently pulled away from the device until the grip was released, yielding a measure of maximal grip force. Each animal underwent three consecutive trials per forelimb, and the mean value was used for subsequent analyses.

Clinical scoring. All mice were monitored weekly using a custom-made clinical scoring scale (Table 1), incorporating multiple behavioral and clinical parameters, such as body weight, tremor, gait abnormalities, and posture. A predefined humane endpoint was established at a score of 6.

Table 1.

Clinical scoring for longitudinal assessment of well-being in PS19 mice. Four clinical parameters (posture, general activity in the cage, locomotion, weight loss) are followed up on a daily basis for 4 days after each surgery and every week up to reaching humane endpoint. Humane endpoint is set at a score of 6 or above, upon addition of the four parameters

Clinical parameters		Score
Posture	Arched back	3
	Lateral or ventral recumbency	4
Activity	Stereotypy	2
	Trembling/spasms	3
Locomotion	Gait abnormalities	2
	Limb paralysis	4
Weight loss	5–10%	1
	11–15%	2
	16–20%	4
	More than 20%	6

Neurofilament-light chain assay

Blood samples were collected either from the submandibular vein or via intracardiac puncture at the time of euthanasia and transferred into heparin-coated tubes. Samples were centrifuged at 12,000 × g for 10 min, and the resulting plasma was stored at − 80 °C until analysis. Neurofilament light chain (Nf-L) concentrations were quantified in plasma samples diluted 1:4 using the Nf-light® Advantage V2 SR-X Kit (Quanterix). Measurements were performed on an automated SR-X™ platform (Quanterix) in accordance with the manufacturer’s instructions. Twenty-four animals were included for Nf-L assay (n = 12 for sham, n = 12 for rmSCI).

Histology

Following fixation in 4% PFA for 48 h, tissues were rinsed in tap water for 30 min, then dehydrated in 70% isopropanol and processed for paraffin embedding. Spinal cord or brain sections (10 µm thickness) were cut on a microtome and mounted on Superfrost™ Plus slides (Epredia, J1800AMNZ). Prior to histological procedures, slides were dewaxed and rehydrated.

Eriochrome C—Cresyl Violet staining. Slides were incubated in Eriochrome C staining solution for 30 min, briefly rinsed (10 s) in tap water, and differentiated in NH₄OH for 4 min before rinsing in distilled water. Sections were then counterstained with 1% cresyl violet for 1 h and differentiated through successive alcohol baths (95% ethanol, twice, followed by a 50:50 mixture of toluene and absolute ethanol for 5 s each). Slides were finally mounted with DPX mounting medium (BDH Chemicals, 36,029) and coverslips.

Gallyas silver staining. Slides were rinsed in distilled water and incubated in potassium permanganate solution for 10 min, followed by rinsing in distilled water. Sections were bleached with 1% oxalic acid until completely decolorized and rinsed again. Slides were then immersed for 1 min in an alkaline silver iodide solution (NaOH 1 M, KI 0.6 M, AgNO₃ 2 mM), neutralized in 0.5% acetic acid for 3 min, and developed in a 1:1 mixture of Developer A (Na₂CO₃, 50 g/L) and Developer B (NH₄NO₃ 2 g/L, AgNO₃ 2 g/L, tungsten silicate 10 g/L) pre-warmed to 20 °C. After rinsing in 0.5% acetic acid (3 min), slides were treated with 1% gold chloride (5 min), rinsed, fixed in 1% sodium thiosulfate (5 min), rinsed again, counterstained with hematoxylin, dehydrated, and mounted with DPX.

Immunohistochemistry. Following dewaxing, antigen retrieval was performed by incubating slides in citrate buffer (pH 6.0) at 100 °C for 10 min (except for Tau phospho-epitopes detection). Slides were rinsed in tap water, treated for 10 min in methanol containing 3% H₂O₂, and then rinsed in TBS (pH 7.4). Spinal cord or brain sections were blocked for 30 min at room temperature (RT) in TBS containing 5% goat or horse serum (VECTASTAIN® ABC Kit, Vector Laboratories PK4001/PK4002). Slides were incubated overnight at 4 °C with primary antibodies diluted in TBS containing 1% serum: NeuN 1:1000 (Cell Signaling D353I, 12943), GFAP 1:10,000 (Sigma-Aldrich G3893), Iba1 1:1000 (Wako 019–19741), Neurofilament-light polypeptide 1:1000 (Biosensis 6H63), AT8 pTau(Ser202/Thr205) 1:500 (Thermo Fisher Scientific MN1020), pTau(Ser422) 1:5000 (4BioDx 1501), and pTau(Thr181) 1:500 (Cell Signaling D9F4G, 12885). After rinsing, slides were incubated for 1 h at RT with an appropriate biotinylated secondary antibody (VECTASTAIN® ABC Kit, Vector Laboratories PK4001/PK4002), rinsed again, and incubated for 45 min at RT in the streptavidin–peroxidase solution. Immunoreactivity was revealed using diaminobenzidine (DAKO K3468). Sections were counterstained with hemalum, dehydrated, and mounted with DPX. For mouse primary antibodies against pTau, Mouse-on-Mouse Kit (Vector Laboratories BMK 2202) was used for blocking, dilution and detection of primary antibodies. PBS—0.1% Triton X-100 was used in the blocking and dilution buffers instead of TBS for Tau phospho-epitopes detection.

Immunofluorescence. Following dewaxing, slides were incubated twice for 3 min in 0.1 M glycine–PBS (pH 7.2). Sections were then permeabilized and blocked for 30 min at RT in TBS containing 0.02% Triton X-100 and 5% goat or horse serum. Slides were incubated overnight at 4 °C with primary antibodies diluted 1:200 in TBS containing 0.02% Triton X-100 and 1% serum. The following day, slides were rinsed and incubated for 1 h at RT with Alexa Fluor®–conjugated secondary antibodies diluted 1:100 in blocking buffer (Alexa Fluor® 594 goat anti-rabbit IgG, Invitrogen, A11012; Alexa Fluor™ 488 goat anti-mouse IgG, Invitrogen, A11001). Nuclei were counterstained with Hoechst (1:500; Sigma-Aldrich, 94403) for 5 min, and slides were mounted with Mowiol.

Image acquisition and analysis. Images were acquired using an Olympus BX63 microscope or Pannoramic Desk II DW Digital Scanner (3DHISTECH). All image analyses were performed using ImageJ software (version 1.54d). For image quantification of astrocytes, microglia and neurons, results are expressed as a cell density, calculated by the number of positive cells in grey matter, normalized by the total surface of grey matter in mm². For pTau IHC and Gallyas staining, results are expressed as pixel density and calculated by the surface of staining in grey matter (µm²) normalized by total surface of grey matter. One slide was analysed per animal. The raw analyses were performed blindly by three investigators (Z.M., H.C. and N.H.), and data were averaged. One-hundred thirty-nine animals were used for histological analysis (n = 70 for sham, n = 69 for rmSCI).

Western blot

For protein extraction, tissues were homogenized in RIPA buffer supplemented with a protease inhibitor cocktail (1:25; Roche Diagnostics, 04693132001) and a phosphatase inhibitor cocktail (1:1000; Sigma-Aldrich, P5726). Homogenization was performed using a Tissue Grinder NG010 (Genetics). Samples were kept on ice for 2 h, with additional homogenization every 30 min. Tubes were then sonicated on ice (16 cycles of 1 s at 20 kHz) using a Sonics Vibra-Cell instrument (Analis), followed by centrifugation at 13,000 × g for 5 min at 4 °C. The supernatants were collected and stored at − 80 °C. Protein concentrations were determined using the Pierce BCA Protein Assay Kit (Thermo Scientific, 23,227). For SDS–PAGE, 25 µg of protein were denatured at 100 °C for 5 min and loaded onto 10% polyacrylamide gels. Proteins were separated by electrophoresis and transferred onto PVDF membranes (Thermo Scientific, 88518) for 40 min at 100 V. Membranes were then blocked in TBS containing 0.1% Tween-20 (Thermo Scientific BP337) and either 5% BSA (Fisher Chemical 205565) for phosphorylated tau epitopes or 5% non-fat dry milk (ChemCruz 1621) for all other proteins. Membranes were incubated overnight at 4 °C with primary antibodies diluted in the appropriate blocking solution as follows: AT8 pTau(Ser202/Thr205) 1:1,000, pTau(Thr181) 1:1,000, pTau(Ser422) 1:5,000, total Tau 1:50,000 (Dako A0024), human Tau 1:20,000 (Br21 antibody, kindly provided by Professor J.-P. Brion and Professor K. Leroy, ULB, Belgium) GAPDH 1:10,000 (Sigma-Aldrich G8795), ERK1/2 1:1,000 (Cell Signaling 9102S), pERK1/2 (Thr202/Tyr204) 1:1,000 (Cell Signaling 9106S), pGSK3α (Tyr276)/pGSK3β (Tyr216) 1:2,000 (Invitrogen G44-0446), p35/p25 1:2,000 (Santa Cruz C-19 SC820), p38 MAPK 1:1,000 (Cell Signaling 9212L), pp38 (Thr180/Tyr182) 1:1,000 (Cell Signaling 9211S). Membranes were washed three times for 5 min in TBS containing 0.01% Tween-20 and incubated for 1 h at RT with horseradish peroxidase (HRP)-conjugated secondary antibodies: anti-mouse IgG (Cell Signaling 7076S) or anti-rabbit IgG (Cell Signaling 7074S), both diluted 1:1000. After additional washes, immunoreactive bands were detected using the BM Chemiluminescence Blotting Substrate (Roche Diagnostics, 11500694001). Images were acquired using either an ImageQuant LAS 4000 mini biomolecular imager (GE Healthcare) or an Amersham ImageQuant 800 system (Cytiva). Band intensities were converted into grayscale pixels using ImageJ software and values of pixel intensity were measured as area-under-the-curve (AUC). All signals were normalized to GAPDH, and phosphorylated proteins were further normalized to their respective total protein levels. Data were expressed relative to the sham condition, which was set to 100%. Thirty-four animals were used for western blot analyses (n = 17 for sham and n = 17 for rmSCI).

Immunogold electron microscopy

Sarkosyl fractionation. Spinal cord tissues were homogenized in 500 µL of buffer H (10 mM Tris–HCl, pH 7.4; 0.8 M NaCl; 1 mM EDTA; 10% sucrose) using a mechanical grinder. Homogenates were centrifuged at 15,000 × g for 20 min at 4 °C. The supernatant was transferred to a new tube, and Sarkosyl was added to a final concentration of 1%. Samples were vortexed and incubated under rotation overnight at 4 °C. Samples were then ultracentrifuged at 45,000 rpm for 30 min at 4 °C using an Optima TLX ultracentrifuge equipped with a TLA 100.3 rotor (Beckman Coulter). The supernatant was discarded, and the pellet was gently rinsed with 500 µL of PBS before being resuspended in 50 µL of PBS.

Immunogold labeling of tau PHF. Nickel grids (300 mesh, carbon/Formvar-coated; EMS, FCF300-NI-50) were glow-discharged using an Elmo Glow Discharge system (Cordouan Technologies). A 15 µL aliquot of the sample was applied to each grid and incubated for 5 min at RT. Grids were then blotted on filter paper and incubated with 30 µL of blocking solution consisting of buffer A (20 mM TBS, pH 8.2) supplemented with 1% bovine serum albumin (BSA; V fraction; Sigma-Aldrich, A-4503), 1% heat-inactivated normal goat serum (NGS; Gibco, 16210-064), 0.1% Tween-20 (Thermo Fisher, 13257418), and an additional 10% NGS for 1 h at RT. Grids were subsequently incubated overnight at 4 °C with the Br21 primary antibody (1:200 in buffer A), which specifically recognizes human tau protein. The following day, grids were rinsed three times (5 min each) in 30 µL of buffer A and incubated for 1 h at RT with a secondary anti-rabbit antibody conjugated to 10 nm gold particles (1:50; Sigma-Aldrich, G7402). After incubation, grids were rinsed with buffer A and ultrapure water, stained for 3 min with 3% silicotungstic acid, rinsed again with ultrapure water, blotted dry, and examined using a Tecnai 10 transmission electron microscope operated at 100 kV (FEI Technologies).

Tau seeding assay

Protein extraction. Spinal cord tissues were incubated in 100 µL of PBS supplemented with a protease inhibitor cocktail (1:25; Roche Diagnostics, 04693132001) and a phosphatase inhibitor cocktail (1:1000; Sigma-Aldrich, P5726) and homogenized using a mechanical grinder. Samples were then vortexed and sonicated 16 times for 1 s at 20 kHz. Homogenates were centrifuged at 1,000 × g for 10 min at 4 °C, and the resulting supernatant was further centrifuged at 21,000 × g for 15 min at 4 °C. The supernatant and pellet fractions were collected separately and assimilated to soluble and insoluble protein fractions, respectively. Protein concentrations were determined using the Pierce BCA Protein Assay Kit prior to storage at − 80 °C.

HEK293T FRET biosensor cell culture. HEK293T FRET biosensor cells expressing the Tau repeat domain (TRD) fused to CFP or YFP were used as previously described [24]. Seed uptake induces TRD recruitment, generating a FRET signal. Cells were plated at a density of 80,000 cells per well in 6-well plates and maintained in DMEM/F12 medium (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS), 1% penicillin–streptomycin, and 1% GlutaMAX. After 24 h, cells were transfected with 30 µg of total protein extracted from spinal cord using a liposome-based transfection reagent (jetPRIME, Polyplus, 101000001). Cells were then cultured for an additional 5 days before being harvested for flow cytometry analysis.

FRET analysis. Cells were detached using trypsin–EDTA for 5 min at 37 °C. The reaction was stopped by adding culture medium, and cells were transferred into 1.5 mL tubes. Following centrifugation at 1,000 × g for 5 min at RT, the supernatant was removed, and the cell pellet was resuspended in 500 µL of PBS containing 2% FBS. Samples were analyzed using a BD LSRFortessa Cell Analyzer (BD Biosciences). Fluorescence was quantified as integrated fluorescence intensity (percentage of positive cells × mean fluorescence intensity). All data were normalized to the mean value of the sham condition. Ten animals were used for Tau seeding assay (n = 5 for sham, n = 5 for rmSCI).

Gene expression analysis

RNA extraction. Spinal cord samples were homogenized in 500 µL of TRIzol reagent (Invitrogen, 15596018) using a grinder device and incubated at RT for 5 min. Homogenates were centrifuged at 15,800 × g for 10 min at 4 °C, and the supernatant was transferred to a new tube. Chloroform (200 µL) was added, and samples were vortexed and incubated for 3 min at RT. Following centrifugation at 15,800 × g for 15 min at 4 °C, 300 µL of the aqueous phase was collected and mixed with an equal volume of 70% ethanol. The mixture was then applied to an RNA extraction column (High Pure RNA Tissue Kit, Roche Diagnostics, 12033674001), and RNA was purified according to the manufacturer’s instructions.

RNA sequencing. RNA extracts were subjected to RNA sequencing (RNA-seq) using Oxford Nanopore Technologies (ONT), outsourced to Namur Molecular Tech sequencing platform (CHU UCL Namur, Yvoir, Belgium). RNA quality was assessed using the TapeStation system (D5000 ScreenTape®; Agilent Technologies). Only samples with an RNA integrity number (RIN) greater than 8.5 and an RNA concentration exceeding 100 ng/µL were retained for library preparation. Complementary DNA (cDNA) libraries were generated using the Direct cDNA Library Prep Kit (SQK-NBD114.96; ONT) and pooled equimolarly for multiplexing. Sequencing was performed on a PromethION flow cell R10.4.1 (FLO-PRO114; ONT) until flow cell failure (~ 68 h).

RNAseq analysis. Raw sequencing reads were basecalled and demultiplexed using Dorado v7.3.11 with the high-accuracy model (400 bps). The resulting FASTQ files were quality filtered with fastp v0.22.0 using the following parameters: --length_required 80, --length_limit 16000, --qualified_quality_phred 9, and --unqualified_percent_limit 20. Reads passing quality control were aligned to the mouse reference genome (GRCm39) using the splice-aware aligner STARlong v2.7.10b with recommended parameters for high-quality long reads. Gene-level read counts were obtained using featureCounts v2.0.6 with the following options: --maxMOp 100 -M --fraction –largest Overlap -g gene_id --minOverlap 50 -L -O. Cut-off to select differentially expressed genes was set up at p-value < 0.05 for significance and ± 1.3 for fold change. Differential gene expression analysis was performed in R using the edgeR package [5]. Low expression genes were filtered using the filterByExpr function and library size were normalized using the trimmed mean of M-values (TMM) method. Gene-wise dispersions were estimated using the edgeR quasi-likelihood framework. A generalized linear model (~ group + batch) was fitted to the normalized count data, and differential expression between experimental groups was assessed using quasi-likelihood F-tests. P-values were adjusted for multiple testing using the Benjamini–Hochberg procedure to control the false discovery rate (FDR).

Gene set enrichment analysis (GSEA). Genes were ranked by Log2FC RNA expression prior to GSEA using R package ClusterProfiler [61] with default settings and Gene Ontology Biological Process (GOBP) [7] as annotation database. Results were filtered on Benjamini–Hochberg calculated adjusted P-values (pAdj < 0.05). To reduce redundancy in the list of significantly enriched terms identified, a Jaccard similarity matrix was calculated for each term based on core leading edge genes, followed by hierarchical clustering from which 12 clusters were extracted. Complete list of enriched results are shown in Supporting Information Table S1. Representative enriched terms of Cluster 1 and 3 were manually selected from the wider clusters for graphical representation.

Real-Time qPCR. Reverse transcription was performed on Thermal Cycler 2720 (Applied Biosystems) with dNTP (Invitrogen 10219012, 10218014, 1216016, 10217018), DTT 0,1 M (Invitrogen Y00147), First strand buffer (Invitrogen 282038), Oligo-dT (Invitrogen 2296086) and SSIII reverse transcriptase (Invitrogen 2853596). cDNA was diluted 1:100 and amplified with primers at 300 nM, listed in Table 2.

Table 2.

Sequences of forward and reverse primers used in RTqPCR assays

Genes	Forward	Reverse
Irf5	5′-TGTTGCCTTTGACGGACCTA-3′	5′-ACATCTAGCAGCTGGTTCGT-3′
Irf7	5′-ACAGGGCGTTTTATCTTGCG-3′	5′-TGGAGCCCAGCATTTTCTCT-3′
Irf8	5′-AGTACACAACAGGGGCAGAA-3′	5′-TTTCCCCAGAGCAACAGTCA-3′
Ifitm3	5′-AGGAAGAATATGAGGTGGCTGA-3′	5′-ATCACCCACCATCTTCCGAT-3′
Ifi27I2a	5′-TAGCCACACTCCAATCAGCA-3′	5′-TCATCTCCTGCGTAGTCTGT-3′
Ifit1	5′-AAGTGGAGCCAGAAAACCCT-3′	5′-GGGCTTCTTCAATGTGTGCT-3′
Bst2	5′-ATGGCGCCCTCTTTCTATCA-3′	5′-TGTGACGGCGAAGTAGATTG-3′
Pycard	5′-TCCTGGACGCTCTTGAAAAC-3′	5′-AGCCCATGTCTCTAAGCACA-3′
Hprt	5′-GGACCTCTCGAAGTGTTGGAT-3′	5′-CCAACAACAAACTTGTCTGGAA-3′

All results were normalized to Hprt as housekeeping gene. All data were expressed as a relative expression compared to delta Cq value of sham condition. Twenty-four animals were used in total for gene expression analysis (n = 12 for sham, n = 12 for rmSCI).

Statistics

Outlier detection was performed on datasets with n > 3 using the GraphPad outlier calculator (α = 0.05). Comparisons between sham and rmSCI groups were conducted using a two-tailed non-parametric Mann–Whitney test for western blot and histological outcomes. For FRET experiments, a one-tailed Mann–Whitney test was used to assess the rise of tau seeding. For analyses involving multiple comparisons between groups and time points, non-parametric mixed-effects models followed by Tukey’s multiple comparison test was applied. All statistical analyses were performed using GraphPad Prism software (version 10.2.3).

Results

rmSCI produce no immediate impairments but worsen motor decline of PS19 mice later in life

Throughout the 10-month experimental timeline (Fig. 1A), motor function and general health of PS19 mice were longitudinally assessed using the grip strength test and a custom clinical scale (Fig. 1B–E). Following the first or second spinal impact, no immediate impairment in forelimb muscle strength was observed, consistent with the mild and clinically asymptomatic nature of the contusions (Fig. 1B). When assessed up to day 126 in a first cohort of PS19 mice, no differences in grip strength were detected between sham and rmSCI groups, and no changes over time were observed within either group (Fig. 1B). These findings suggest that the progression of tau pathology had not yet produced detectable motor deficits at this stage. In contrast, when a second cohort of PS19 mice were followed up to later time points (days 140–224; corresponding to 7–10 months of age), forelimb grip strength began to decline in the sham group, consistent with the onset of symptomatic tauopathy and in line with the literature on the PS19 model [60]. While this functional motor impairment was also observed in rmSCI group (Fig. 1C), mixed-effects analysis revealed a significant exacerbating effect of rmSCI on the progression of grip strength deficits (p < 0.05 for treatment effect in both left and right forelimbs) (Fig. 1C). In parallel, animal well-being was monitored using a clinical scoring system (Table 1) assessing body weight loss, gait coordination, balance, posture, and tremors. Each symptom was scored, with a maximum cumulative score of 6 defining the experimental humane endpoint. This longitudinal assessment enabled tracking of overall health status over ageing (Fig. 1D). Each coloured rectangle represents the proportion of animals reaching the corresponding cumulative score. At day 126, all animals had a score of 0, indicating no detectable impairments at 6.5 months of age. From day 140 onward, clinical scores progressively increased in both groups, reflecting the emergence of weight loss, gait disturbances, and tremors. Notably, symptoms appeared earlier and were more severe in the rmSCI group compared to sham animals. The Kaplan–Meier curve in Fig. 1E depicts humane endpoint–free survival (defined as reaching a clinical score of 6) for each group. At day 224, survival was 62.5% in the sham group (10/16) and 39.99% in the rmSCI group (6/15); however, this difference did not reach statistical significance due to the limited sample size (p = 0.2537; log-rank (Mantel–Cox) test).

rmSCI induce gliosis and axonal damages without overt tissue disruption

Histological analyses were carried out at the lesion epicentre (C5 level) to evaluate the tissue structure and the changes in nervous cell populations (Fig. 2). Eriochrome C/violet cresyl staining revealed no obvious architectural alterations or tissue loss in either grey or white matters at 2 days following first (Fig. 2A) or second (data not shown) contusion. Glial activation was assessed using GFAP and Iba1 immunostainings. Astrocyte density was significantly increased in the contusion group compared to sham up to 7 days after the first contusion (p < 0.05), but not after the second contusion (Fig. 2B). By day 126, no differences in astrocyte density were observed between sham and rmSCI groups (Fig. 2B). Microglial density was significantly elevated in the contusion group 2 days after each injury (p < 0.01), consistent with an acute neuroinflammatory response following mild spinal cord injury (Fig. 2C). Although microglial density was comparable between groups at day 42, it was slightly but significantly higher in the contusion group at day 126 (p < 0.05), suggesting a delayed microglial reactivation, potentially associated with emerging tau pathology. Early glial activation was supported by bulk RNA-sequencing of spinal cord homogenates (Fig. S1, GEO accession GSE336414). Differential expression analysis revealed 65 differentially expressed genes between sham and single mSCI mice (Fig. S1A). GSEA using the GOBP database, followed by clustering of enriched terms based on the similarity of their leading-edge genes, allowed identification of 12 clusters of enriched terms (Table S1). Interestingly, cluster 3 contained multiple terms related to type I interferon (IFN) activation, while cluster 1 grouped a variety of terms related to immune and inflammatory processes, among which we identified microglial activation as interesting candidate process (Fig. S1B). Indeed, by manual inspection of the dataset using STRING analysis, we found 17 astrocyte-related genes (e.g. Aldh1l1, Nes) or microglial activation-related genes (e.g. Aif1/Iba1, Csf1r, Ctsb, C3ar1, Trem2, Grn, Clu) as well as 12 type I IFN pathway-related genes, with significantly increased expression (Fig. S1C and D).

Fig. 2.

Histopathological outcomes in PS19 mice undergoing sham surgeries or rmSCI. A Representative illustrations of spinal tissue at C5-level (epicentre) sampled 2 days after first mSCI. No obvious architectural alterations, tissue loss in the grey matter or demyelination in the white matter tracts are observed. B Representative GFAP immunostaining at C5-level sampled 2 days after first mSCI. Quantification of astrocytes in the grey matter shows a significant increase at day 2 and 7. n = 4–10 spinal cords/group. C Representative Iba1 immunostaining at C5-level sampled 2 days after first mSCI. Quantification of microglia in the grey matter shows a significant increase at day 2, 23 and 126. n = 3–8 spinal cords/group. D Representative NeuN immunostaining at C5-level sampled 2 days after first mSCI. Quantification of neuronal cell bodies shows a significant effect of time on neuronal density but no difference between sham and rmSCI group at any timepoint. n = 4–7 spinal cords/group. E Representative cleaved caspase 3 immunostaining at C5-level sampled 2 days after rmSCI at a force of 30 kDyn, compared to moderate SCI at 40 kDyn. Numerous cell bodies are caspase 3-positive in 40 kDynes condition, used as positive control for apoptosis. F Representative immunostaining for damaged Nf-L epitope at C5-epicentre sampled 2 days after first mSCI. G Longitudinal follow-up of plasma Nf-L concentration from day 1 to day 56. n = 3–5 blood samples/group. H Plasma Nf-L concentration detected at day 126 (6.5-month-old mice). n = 7 blood samples/group. Plasma Nf-L levels are significantly increased in rmSCI group at day 1, 22, 35 and 126. Scale bars represent 200 µm (A, B) and 20 µm (E). Data are presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, mixed-effects model followed by Tukey’s multiple comparison test (B–D and G) or two-tailed Mann–Whitney test (H)

Neuronal density was evaluated using pan-neuronal immunostaining (Fig. 2D). While no difference was detected between sham and rmSCI group at any time point, an overall time effect demonstrated that NeuN density progressively decreased in both groups (p < 0.05, Fig. 2D), consistent with the ongoing neurodegeneration described in the PS19 model.

The lack of immediate neuronal death is further supported by the absence of cleaved caspase-3–positive cells in the mild contusion (30 kDyn) condition, whereas a more severe injury (40 kDyn) used as positive control, produced apoptosis (Fig. 2E). To assess axonal damages, immunostaining for a damaged neurofilament-light chain (Nf-L) epitope was performed 2 days after each contusion.

Damaged Nf-L was scarcely detected in grey matter but was widely distributed across white matter tracts, including the fasciculi cuneatus and gracilis, dorsal corticospinal, rostral reticulospinal, and ventrolateral spinothalamic tracts (Fig. 2F). In parallel, plasma Nf-L levels were measured at multiple time points following the first and second contusions (Fig. 2G–H). One day after each surgery (days 1 and 22), Nf-L concentrations were significantly increased in the contusion group compared to sham, confirming axonal damages observed in spinal tissue (p < 0.001 and p < 0.05 after the first and second contusions, respectively) (Fig. 2G). Longitudinal analysis showed that Nf-L levels remained elevated for up to 3 weeks after second contusion, although with only a significant difference at day 35 (p < 0.05). By day 52, Nf-L levels tend to return toward baseline. Surprisingly, on day 126, plasma Nf-L levels were significantly higher in the rmSCI group compared to sham animals (p < 0.01) (Fig. 2H).

Tau pathology in PS19 mice is exacerbated at the epicentre 126 days following rmSCI

The development of tauopathy in PS19 mice was evaluated at and around the C5 lesion epicentre using immunohistochemistry and antibodies targeting distinct phosphorylated tau epitopes (Ser202/Thr205, Ser422, and Thr181) (Fig. 3A–C). While pTau immunoreactivity was undetectable in young adult mice 2 days after rmSCI (data not shown), a marked signal was observed in 6.5-month-old animals. At day 126 (6.5 month of age), abnormal pTau immunoreactivity was predominantly localized to neuronal cell bodies, and to a lesser extent, along neuronal processes, and affected indistinctly ventral, intermediate, or dorsal horn neurons. No staining was detected in the white matter tracts. Double immunofluorescence was performed to determine the cellular distribution of phosphorylated tau species. For instance, pTau(Ser202/Thr205) was mainly detected in neuronal cell bodies. In some cases, pTau immunoreactivity was also visualized in microglia, corresponding most likely to phagocytosed debris, but never in astrocytes (Fig. S2). Quantitative analysis revealed a significant increase in Ser422 phosphorylation.

Fig. 3.

Tauopathy signature at day 126 following rmSCI. A–C Tau pathology in PS19 spinal cords is detected using antibodies against pTau(Ser422), (Ser202/Thr205) and (Thr181) epitopes. Increased phosphorylation is observed in situ at C5-level for pTau(Ser422), (Ser202/Thr205) in rmSCI group compared to sham, as well at C7-level for (Ser422) epitope. n = 4–14 spinal cords/group. Representative illustrations are taken at C5-level. D Total Tau, human Tau and phosphorylated Tau epitopes are detected by immunoblotting on protein extracts from the C5-epicentre. E–I Among the three phospho-Tau epitopes, only pTau(Ser422) shows a significant increase following rmSCI. AUC, area under the curve (see methods for relative quantification and normalization). n = 5–6 spinal cords/group. Scale bar represents 50 µm. Data are presented as mean ± SD. *p < 0.05, two-tailed Mann–Whitney test

Ser422 phosphorylation at both the C5 epicentre and the C7 level in the rmSCI group compared to sham (p < 0.05) at day 126 (Fig. 3A). Increased phosphorylation at Ser202/Thr205 was also observed at the C5 level (p < 0.05) (Fig. 3B), whereas no differences were detected for Thr181 phosphorylation at any spinal level (Fig. 3C). The abundance of phosphorylated tau species was further assessed by Western blot analysis of spinal cord extracts (C4–C6 levels) using the same panel of antibodies (Fig. 3D–I). Consistent with immunohistochemistry findings, Ser422 phosphorylation was significantly increased in the rmSCI group compared to sham (p < 0.05), whereas no significant differences were observed for Ser202/Thr205 or Thr181 phosphorylation after normalization to total tau levels. Except for the amygdala, most brain regions showed very low immunoreactivity towards phospho-tau epitopes at this time point (cortex, hippocampus, thalamus, and brainstem) (Fig. S3).

Tau pathology remains increased 196 days following rmSCI and extends beyond the lesion epicentre

Mice were maintained until 9 months of age to determine whether the exacerbated tauopathy observed in the rmSCI group persisted with ageing at the lesion epicentre and extended to rostral or caudal spinal levels. Immunohistochemical analyses targeting the same phosphorylated tau epitopes revealed numerous immunoreactive neuronal cell bodies throughout the cervical spinal cord (Fig. 4A–C).

Fig. 4.

Tauopathy signature at day 196 following rmSCI. Tau pathology in PS19 spinal cords is detected using antibodies against pTau(Ser422), (Ser202/Thr205) and (Thr181) epitopes. Increased phosphorylation is observed in situ at C5-epicentre for pTau(Ser422) in rmSCI group compared to sham and extended rostrally for pTau(Ser422) and (Ser202/Thr205) epitope (A–C). n = 6–10 spinal cords/group. Representative illustrations are taken at C2-level. D Total Tau, human Tau and phosphorylated Tau epitopes are detected by immunoblotting on protein extracts from the rostral C2 to C4 segments. E–I The abundance of phospho-Tau epitopes is not different between sham and rmSCI. AUC, area under the curve. J Representative illustration of PHF extracted from PS19 spinal cord and decorated by 10 nm gold particles. K Gallyas staining is performed on the levels C2 to C7 of PS19 spinal cords to highlight Tau aggregates. Representative illustrations are taken at C2-level. L No difference in the surface of Gallyas-positive staining was measured between sham and rmSCI. n = 7–10 spinal cords/group. Scale bars represent 50 µm (A), 100 nm (J) and 100 µm (K). Data are presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, two-tailed Mann–Whitney test

Quantitative analysis showed a significant increase in Ser422 phosphorylation across multiple levels (C2 to C7) within the grey matter of rmSCI mice compared to sham (p < 0.05, p < 0.01, p < 0.001) (Fig. 4A). In contrast, Ser202/Thr205 phosphorylation just showed a tendency to be increased at the lesion epicentre but remained significantly increased at rostral levels, specifically at C2 and C3 (p < 0.01 and p < 0.05, respectively) (Fig. 4B).

A trend toward increased Thr181 phosphorylation was also observed at the C2 level (p = 0.0541) (Fig. 4C). To further assess these changes, protein levels of pTau(Ser422, Ser202/Thr205, and Thr181) were analyzed by Western blot. No significant differences were detected between groups, either at the lesion epicentre (data not shown) or in rostral segments (Fig. 4D–I). This lack of significance likely reflects the inter-animal variability or variable expression of the human Tau protein (Fig. 4D and F).

Beyond tau phosphorylation, we investigated the presence of pathological tau paired helical filaments (PHFs) in spinal cord homogenates from 9-month-old PS19 mice (Fig. 4J). Insoluble protein fractions were isolated by sarkosyl extraction followed by ultracentrifugation. These fractions were then probed with the Br21 antibody, which specifically recognizes human tau, and scrutinized by immunogold electron microscopy. As shown in Fig. 4J, filamentous structures consistent in size and morphology with PHFs were observed and decorated with gold particles. Building on this observation, we next assessed whether the propensity of tau to form higher-order aggregates was altered by rmSCI (Fig. 4K–L). Although strongly stained neuronal cell bodies were occasionally observed in the spinal cord of PS19 mice, quantitative analysis of Gallyas silver staining did not reveal any significant differences between sham and rmSCI groups, at any spinal level (Fig. 4L).

In addition, immunostaining for damaged Nf-L in rostral segments (i.e., C2–C3) revealed a systematic pattern of axonal degeneration localized to the rostral reticulospinal as well as the ventrolateral spinothalamic tracts (Fig. 5A). This observation prompted us to investigate the ascending and descending brain–spinal pathways. The brainstem and thalamus were specifically examined based on their respective neuroanatomical projections via the reticulospinal and spinothalamic tracts. While both the thalamus and the brainstem contained very rare phospho-tau immunoreactive cells at 126 days post-rmSCI (Fig. S3), obvious pTau (Ser422) immunoreactivity was observed at 196 days in the thalamus (Fig. 5B), which was significantly increased in rmSCI group (p < 0.05), along with a trend toward increased pTau(Ser202/Thr205) phosphorylation (p = 0.077) (Fig. 5C). No differences in Ser422 or Ser202/Thr205 phosphorylation were detected in the brainstem (Fig. 5B and C).

Fig. 5.

Tauopathy signature at day 196 in brainstem and thalamus of PS19 mice. Spinal and brain tissues are respectively immunolabelled using antibodies against damaged Nf-L epitope, pTau(Ser422) or pTau(Ser202/Thr205). A Immunoreactivity for damaged Nf-L at C2-level is systematically observed along rostral reticulospinal tract (rrst) and ventral spinothalamic tract (vst) in both groups. B pTau(Ser422) immunoreactivity is significantly increased in the thalamus but not in the brainstem following rmSCI. C A trend towards increased (Ser202/Thr205) phosphorylation is also measured in the thalamus following rmSCI. B and C No difference between groups is seen in the brainstem for any phospho-epitopes. n = 9–10 brains/group. Scale bars represent 200 µm. Data are presented as mean ± SD. *p < 0.05, two-tailed Mann–Whitney test

Extracts from PS19 spinal cords seed Tau recruitment, with a higher potential in rmSCI group

As we and others have previously demonstrated the expression of human mutant tau in PS19 spinal cords [17], we sought to evaluate potential differences in tau seeding capacity using tissue extracts from sham and rmSCI mice euthanized at 8 months of age. The experimental workflow, summarized in Fig. 6A, included a final high-speed centrifugation step yielding two protein-enriched fractions containing soluble and insoluble material, respectively. Both fractions were first characterized for total human tau and pTau(Ser422) content (Fig. 6B–F).

Fig. 6.

Tau seeding abilities from PS19 spinal cord extracts following rmSCI. A Workflow for preparing soluble and insoluble spinal extracts and FRET analysis. B–F Protein extracts are characterized by immunoblotting for pTau(Ser422) and human Tau expression and do not significantly differ between sham and rmSCI. n = 5 spinal cord extracts/group. G–H Soluble and insoluble protein extracts are transfected in HEK 293 T cells expressing Tau RD P301S FRET biosensor system. Representative fluorescent FRET signal is taken 6 days after transfection. While not different for the soluble fraction, a significant rise of FRET signal is measured for the insoluble fraction in rmSCI condition. n = 5 spinal cord extracts/group. Data are presented as mean ± SD. *p < 0.05, one-tailed Mann–Whitney test (C-I). Scale bar represents 50 µm

Due to substantial inter-sample variability, no significant differences in protein abundance were detected between sham and rmSCI groups for either total human tau or pTau(Ser422) in either fraction. To assess seeding activity, 30 µg of protein from each fraction were transfected into HEK293T biosensor cells expressing the tau microtubule-binding domain (MBD) fused to CFP or YFP. Upon recruitment into aggregates by exogenous tau species, these constructs generate a FRET signal. Consistent with this mechanism, FRET signals were detected in all conditions (sham and rmSCI) and in both fractions (Fig. 6G–H). Importantly, higher FRET signals were observed in cells transfected with rmSCI-derived samples, with a significant increase in the insoluble fraction (p < 0.05). Altogether, these findings indicate that the rmSCI paradigm enhances the seeding capacity of human tau, likely through promoting its accumulation in aggregation-prone forms.

Type I IFN response and p38 MAPK are early induced following mild spinal contusion

From a mechanistic perspective, we sought to decipher the kinase(s) and upstream pathways involved in the exacerbation of tau phosphorylation. Using a candidate-driven approach targeting major tau kinases (ERK1/2, GSK3α/β, and CDK5), the phosphorylated form of p38 MAPK (pp38) emerged as the only kinase upregulated 2 days following mild spinal contusion (p < 0.05) (Fig. S4A–D). Interestingly, at day 126, pp38 was found to be colocalized with neuronal pTau(Ser422) and pTau(Ser202/Thr205) deposits (Fig. S4E). At that time point, while the abundance of its phosphorylated form was no longer different, a significant increase of total p38 was evidenced in rmSCI group (p < 0.05) (Fig. S4F–H). Beside activated microglia and astroglia gene signature, gene ontology analysis on bulk RNAseq data (Fig. S1) 2 days after mild spinal contusion revealed that 12 genes differentially expressed (out of 65) were related to Type I IFN pathway (Fig. 7A). To validate these findings, we performed qPCR assay (Fig. 7B–E) on 8 selected genes associated with the type I IFN pathway and measured their expression at both the epicentre and rostral segment. Among these genes, 6 transcripts (Irf7, Irf8, Ifi27l2a, Ifit1, Bst2, Pycard) were confirmed to be significantly upregulated in the contusion group compared to sham in the epicentre (Fig. 7B), while 4 transcripts (Irf5, Irf8, Ifi27l2a, Pycard) showed increased fold changes in the rostral segment (Fig. 7C), although to a lesser extent than those observed at the epicentre. These results suggest that activation of the type I IFN pathway may represent an early event contributing to the initiation or amplification of tau pathology. At a later time point (6.5 months; day 126), analysis of the same genes at the lesion epicentre revealed no significant differences between sham and rmSCI groups (Fig. 7D). However, in the rostral segment, differential expression between groups persisted, reaching statistical significance for Irf7 (p < 0.05) and showing trends for Ifitm3 and Ifit1 (p = 0.07), likely reflecting limited sample size (Fig. 7E).

Fig. 7.

Activation of Type 1 IFN pathway following mild spinal cord injury. A The volcano plot shows the adjusted P-value (-Log10) as a function of the fold change (Log2FC) for all genes in mSCI compared to sham. All differentially expressed genes (thresholds p < 0.05 for adj. p-value and ± 1.3 for FC) are colored, either in blue for downregulation or red for upregulation. Interferon-stimulated genes (ISGs) belonging to Type 1 IFN pathway are overrepresented 2 days following mSCI (see Fig. S1). Dark red color is used to highlight the genes involved in Type I IFN pathway. n = 7 spinal cords/group. B and C RT-qPCR confirms the early induction of 6 ISGs (Irf7, Irf8, Ifi27l2a, Ifit1, Bst2, Pycard) at the C5-epicentre and 4 of them (Irf5, Irf8, Ifi27l2a, Pycard) at rostral levels. D and E At day 126, the analysis on the same set of genes shows an upregulation of Irf7 and a trend for Ifitm3 and Ifit1 at the rostral levels but not at the epicentre. n = 3–5 spinal cords/group. Data are presented as mean ± SD. *p < 0.05, **p < 0.01, two-tailed Mann–Whitney test

Discussion

Over recent years, increasing attention has been devoted to the relationship between repeated mild traumatic injuries and the development of dementia. Recent evidence indicates that pathognomonic lesions of CTE (i.e. hyperphosphorylated forms of tau protein) can be detected in up to 41.4% of young symptomatic athletes engaged in contact sports, even at the amateur level [30]. In parallel, independent retrospective studies have reported that individuals with a history of traumatic SCI exhibit approximately a twofold increased risk of developing late-onset Alzheimer’s disease or related dementias compared to non-SCI individuals, even after adjusting for confounding factors such as TBI [11, 58]. Despite these clinical observations, experimental studies aiming to model this relationship in rodents have primarily relied on single or severe spinal cord injuries, such as transection or high-impact contusion [4, 22, 23, 33, 34, 44, 47]. In contrast, our study employed a novel paradigm based on two repeated mild spinal cord contusions, resulting in minimal tissue damage and no-detectable immediate motor impairment. On purpose, we used the heterozygous PS19 transgenic mice because they are intrinsically prone to develop clinical signs of age-dependant tauopathy as well as pathognomonic tau signature throughout the nervous system. Longitudinal clinical monitoring revealed a progressive decline in general health, with an ethical lifespan of approximately 10 months, consistent with the well-established neurodegenerative phenotype of this model [27, 50, 60]. Importantly, motor decline was accelerated in the rmSCI group, suggesting that repeated mild spinal cord trauma, even in the absence of overt injury, can exacerbate disease progression and negatively impact functional outcomes.

In this model of mild spinal contusion, histological analyses revealed a rapid and transient gliosis, without neuronal loss or overt tissue disruption. Despite their mild severity, these contusions induced axonal damage and compromised blood–brain barrier, as evidenced by the presence of damaged Nf-L in the tissue and their release into the blood. Axonal injury is consistent with the activation of resident phagocytic cells, as reflected by the RNA-seq gene signature (Ccl9, Selplg, Ctsb, Ctsh, Grn, Fcgr1, Fcgr2, Aif1, Trem2, Cd53, Cyba, Bin2, Serpina3n). Aside from complement components (C1qa, C4b, C3ar1) and interferon-stimulated genes (Irf5, Irf7, Irf8, Ifi27l2a, Ifit1, Bst2, Pycard), no significant induction of inflammatory cytokines was observed following mild spinal contusion. This transcriptional signature appears to be, at least partially, shared across various models of mild central nervous system trauma in both rodents and humans [17, 42, 51, 53].

The progression of tau pathology was monitored thereafter at 3.5, 6.5 and 9 months of age. No pathognomonic phospho-tau signature was detected in the spinal cords of 3-month-old PS19 mice, even shortly after rmSCI. This absence can be easily explained either by the mild severity of the traumas or the lack of spinal and brain pathology at such early stage in the PS19 model [60]. In contrast, at 6.5 months of age, immunohistochemical analyses revealed a significant increase in pTau(Ser422) and pTau(Ser202/Thr205) at the C5 epicentre in rmSCI mice. With the exception of pTau(Ser422) at the C7 level, no significant changes in tau phosphorylation were detected in distant cervical segments, indicating that, at day 126, the exacerbation of pathology remained largely confined to the lesion epicentre. Immunoblot analyses of the same epitopes confirmed the increase in pTau(Ser422) at the epicentre but did not detect changes in pTau(Ser202/Thr205). While Western blotting enables the simultaneous assessment of multiple phosphorylation sites, it lacks spatial resolution due to the relatively large tissue samples required (spanning approximately three spinal levels), which may mask localized alterations in tau phosphorylation. At 9 months of age, a sustained and widespread increase in pTau(Ser422) was found both at the C5 epicentre and across rostral and caudal regions. In contrast, elevated pTau(Ser202/Thr205) was restricted to rostral segments (C2 and C3) and was no longer detected at the epicentre. Immunoblot analysis failed to confirm the local differences in tau phosphorylation observed in rostral segments. This discrepancy may be attributed to variability across spinal cord samples, as previously reported in PS19 brainstems and brains [55]. Overall, our in-situ findings indicate that rmSCI accelerates the hyperphosphorylation on two tau epitopes, namely Ser422 and Ser202/Thr205, and that this abnormal phosphorylation pattern progressively extends along the rostro-caudal axis of the spinal cord. Notably, this abnormal pattern appears to reach the thalamic nuclei, most likely via the ventral spinothalamic tract, along which axonal degeneration was concomitantly observed. The increased pTau immunoreactivity observed at such distant regions of the lesion epicentre could be due to a spreading of abnormal tau species, for instance along the spinothalamic tract, between synaptically connected regions [12] or secondary to the deafferentiation of thalamic neurons, such as observed in hippocampal neurons following an entorhinal lesion [32].

Although PHF-like structures and NFTs were observed in PS19 spinal tissues, no difference in the load of aggregated tau proteins was detected between sham and rmSCI groups at late time points. This finding does not contradict the hypothesis of exacerbated tau pathology, as tau oligomers—rather than large aggregates such as NFTs—are considered among the most neurotoxic species [2]. Indeed, tau oligomers are sufficient to induce neuronal toxicity, whereas NFT burden does not consistently correlate with symptom severity or functional impairment [3, 9, 15, 25, 49]. Interestingly, Tau species in rmSCI samples substantially promoted in vitro tau seeding activity and increased its propensity for oligomerization, even though not evidenced in vivo using the Gallyas staining. In the HEK293T FRET biosensor assay, protein extracts from rmSCI tissues triggered increased seeding capacity, while no difference was observed between sham and rmSCI spinal cord tissues regarding the density of pTau(Ser422) and total human Tau in both soluble and insoluble fractions. These findings suggest that the increased seeding activity cannot be explained solely by changes in overall Tau abundance or Ser422 phosphorylation. An informative experiment would be to perform the same seeding assay using early-stage samples, at a time point characterized by active IFN response but preceding any detectable Tau pathology. Such an approach would help determine whether the increased seeding activity is primarily driven by the burden of pathological Tau species associated with disease progression, or rather by IFN-induced changes in the conformational or biochemical properties of Tau proteins.

From a mechanistic perspective, we sought to decipher the upstream pathways involved in the exacerbation of tau phosphorylation. Among all kinases investigated, only p38 MAPK was shortly and transiently upregulated following rmSCI. Interestingly, interferon-stimulated genes were enriched in the pool of transcripts from rmSCI condition, further supporting the growing body of literature implicating the IFN-I response following mild central nervous system (CNS) trauma [51, 52, 54]. These findings provide new insights into the neuroimmune response to spinal cord injury, particularly in the context of mild and repeated injuries. Several lines of evidence also support the relevance of this pathway in the exacerbation of tauopathy. First, IFN-I–related genes are upregulated in glial cells following traumatic brain injury, with a prominent signature in microglia [6, 51]. Second, activation of the IFN-I pathway has been associated with deleterious effects on the progression of neurodegenerative diseases [21]. Third, a recent study highlights a critical role for IFN-I signalling, promoting both tau hyperphosphorylation and seeding in cellular and animal models of tauopathy [46]. By coincidence, p38 MAPK is a downstream effector of IFN-I signalling. Speculatively, one may hypothesize that the release of damage-associated molecular patterns from injured axons activates resident CNS immune cells, thereby inducing IFN-I production. Upon binding to neuronal IFN-I receptors, this signalling cascade could activate downstream effectors, ultimately leading to p38 MAPK–dependent tau phosphorylation. On the one hand, substantial evidence supports a direct role for p38 in Tau phosphorylation. Previous studies have shown that p38 phosphorylates Tau at the Thr181, Ser202/Thr205, and Ser422 epitopes [8, 16, 18]. Furthermore, activated p38 has been reported to colocalize with phosphorylated Tau in Pick bodies, NFTs and PHFs in Alzheimer's disease [8, 20, 62]. Consistent with these findings, our model also revealed the colocalization of phosphorylated p38 with pTau(Ser202/Thr205) and pTau(Ser422) in neurons. Notably, Ser422 emerged as the predominant Tau phosphorylation site in our paradigm. This observation is particularly relevant because several major Tau kinases, including CDK5, GSK3β, and MARK, do not efficiently phosphorylate this residue, whereas p38 does [16, 18, 45]. This activity appears to be mediated preferentially by the p38β, p38γ and p38δ isoforms [16, 59]. In addition, p38 may promote Tau phosphorylation indirectly by activating downstream kinases, such as MK2 [18, 28]. On the other hand, p38 MAPK is a central regulator of inflammatory signalling and is robustly activated by pro-inflammatory cytokines and Toll-like receptor stimulation [8]. Thus, p38 activation may primarily reflect neuroinflammatory processes, particularly in microglial cells, rather than a neuron-specific mechanism of Tau phosphorylation. In support of this interpretation, studies in the PS19 mouse model have demonstrated an age-dependent increase in hippocampal p38 activity that was predominantly localized to microglia, and to a lesser extent in neurons [43]. Collectively, although our findings demonstrate a close spatial association between p38 activation and Tau hyperphosphorylation, they remain correlative and do not establish a causal role for p38 in driving Tau pathology. Additional studies are required to clarify all these mechanisms, for example through the use of PS19 × Ifnar1−/−mice [46] or selective pharmacological inhibitors targeting isoform-specific p38 and downstream effectors.

Several limitations of this study should be acknowledged. First, although we suspect the involvement of kinase-mediated tau hyperphosphorylation in this paradigm, a concomitant downregulation of phosphatases activity e.g. PP2A cannot be ruled out, such as described following experimental TBI [57]. Second, we intentionally applied the rmSCI paradigm to PS19 mice without performing an extensive characterization in wild-type (WT) littermates. This decision was based on observations made during the initial phase of the study, namely the absence of injury-induced tau pathology following mild contusion in C57BL/6 J WT mice, which remained undetectable up to 10 months of age (8 months post-rmSCI). This contrasts with previous reports describing tau hyperphosphorylation at multiple epitopes in WT rodent models of severe spinal cord injury [33]. Next, only male PS19 mice were included, which may obscure potential sex-specific effects and limit the generalizability of our findings. Males were selected due to their greater tolerance to repeated surgical procedures. Then, sham and rmSCI groups were not compared to naïve, non-operated controls. This comparison could have been informative, as surgical procedures themselves —independently of rmSCI— may have induced peripheral nerve or meningeal damages, leading to inflammatory responses. Such peri- or post-surgical effects may have influenced disease progression and potentially reduced the observed differences between sham and rmSCI groups. Finally, given the inherent variability of the PS19 model and the extensive longitudinal characterization, a further limitation is the relatively limited sample size at each time point. Increasing the number of animals would improve the statistical power and provide additional support for several of the observed trends, particularly for the tauopathy signature and tau seeding assays.

Conclusion

We demonstrated that rmSCI can promote the long-term development of Tau pathology. Our histopathological findings on rmSCI parallel, to some extent, those reported in the context of mild TBI [12, 37]. The early transcriptional response to repeated mild spinal contusions was characterized by activation of glial cells, the complement system, and the IFN-I pathway. Overall, this work provides a valuable framework for investigating the relationship between CNS trauma and the risk of neurodegenerative diseases, as well as a relevant model for studying mild spinal cord injury and associated neuroinflammatory processes.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AUC

Area under the curve

CFP

Cyan fluorescent protein

CTE

Chronic traumatic encephalopathy

FBS

Fetal bovine serum

GOBP

Gene ontology biological process

GSEA

Gene set enrichment analysis

IFN-I

Type I interferon

MBD

Microtubule binding domain

NES

Normalized enrichment score

Nf-L

Neurofilament light chain

NFT

Neurofibrillary tangles

ONT

Oxford Nanopore Technologies

PFA

Paraformaldehyde

PHF

Paired helical filaments

rmSCI

Repeated and mild spinal cord injuries

RNA-seq

RNA sequencing

RT

Room temperature

SCI

Spinal cord injury

TBI

Traumatic brain injury

TRD

Tau repeat domain

YFP

Yellow fluorescent protein

Author contributions

Conceptualization: C.N., K.L., A.L., N.H.; Methodology: K.L., N.H.; Software: N.D., M.C., J.D.; Validation: N.D., L.G., H.C., Z.M., K.A., J.D., N.H.; Formal analysis: N.D, H.C., Z.M., M.C., N.H.; Investigation: N.D. (RNA sequencing), L.G. (in vitro), H.C. (histology), Z.M. (histology, western blot), K.A. (histology, western blot), V.S. (electron microscopy), L.C. (electron microscopy), M.M. (animals follow-up, support for surgery and samples collection), L.S. (animals follow-up, support for surgery and samples collection), N.H (all experiments).; Resources: L.C., A.L., J.D., K.L., C.N.; Writing—Original Draft: C.N., N.H.; Writing—Review and Editing: N.D., K.L., J.D., C.N.; Visualization: N.H.; Supervision: C.N., K.L.; Project administration: C.N.

Funding

Margaux Mignolet is supported by a FNRS/FRIA doctoral fellowship, Karelle Leroy is supported by Fonds Génicot ULB.

Data availability

All RNA sequencing full raw data are available upon request (GEO accession GSE336414).

Declarations

Ethics approval and consent to participate

All experimental procedures were approved by the University of Namur Ethics Committee (Ethics Projects UN 20/362 and 23/403) and conducted in accordance with European directives (2010/63/EU, 86/609/EEC, and 87/848/EEC), as well as the ARRIVE Essential 10 guidelines.

Consent for publication

All co-authors approved the latest version of this manuscript and provided their consent for submission.

Competing interests

The authors declare no competing interests.