Hexaraphane as a potential therapeutic strategy for tauopathies

TL;DR

Hexaraphane, a compound from wasabi, reduces harmful TAU protein changes in Alzheimer's disease models, offering a new treatment possibility.

Contribution

Hexaraphane's novel mechanism of reducing TAU phosphorylation via PP2A activation is identified as a potential therapeutic strategy for tauopathies.

Findings

Hexaraphane reduces pathological TAU phosphorylation in neurons and mouse models.

The effect is mediated through PP2A activation, not via GSK-3β inhibition or NRF2 signaling.

Hexaraphane treatment lowers brain and blood TAU levels and improves cognitive and motor function in mice.

Abstract

Alzheimer's disease (AD) is characterized by pathological hyperphosphorylation of TAU protein, leading to neurofibrillary tangle formation, synaptic dysfunction, neuroinflammation, and neuronal loss. Hexaraphane (6-(methylsulfinyl) hexyl isothiocyanate; HXN), a bioactive compound derived from Wasabia japonica, exhibits neuroprotective and anti-inflammatory properties, yet its potential role in tauopathies remains unknown. Here, we investigated whether HXN modulates pathological TAU phosphorylation and explored the underlying mechanisms in vitro and in vivo. Using primary neurons from APP/TAU transgenic mice with either NRF2 wild-type or knockout backgrounds, combined with complementary genetic and pharmacological approaches, we found that HXN markedly reduced pathological phospho-TAU epitopes (AT8 and PHF1). Notably, this effect occurred independently of NRF2 signaling. Mechanistically, HXN did not suppress GSK-3β activity or alter upstream PI3K/AKT or MAPK pathways. Instead, pharmacological inhibition experiments and phosphatase assays demonstrated that HXN promotes PP2A-dependent TAU dephosphorylation, identifying phosphatase activation as a central mechanism of action. Chronic oral administration of HXN in APP/TAU mice led to significant reductions in brain phospho-TAU levels across multiple regions and decreased blood circulating TAU-pThr217 concentrations. These molecular changes were accompanied by attenuation of neuroinflammatory markers, preservation of neuronal integrity, restoration of synaptic plasticity, and improvements in cognitive and motor performance. Collectively, our findings identify HXN as a potent modulator of pathological TAU phosphorylation. These results support the development of HXN as a promising disease-modifying therapeutic strategy for AD and other TAU-driven neurodegenerative disorders.

Image 1

•Hexaraphane significantly reduces pathological TAU hyperphosphorylation in cellular and in vivo models.•Reduce-pTAU effect by hexaraphane occurs independently of NRF2 activity and is not mediated by GSK-3β inhibition.•PP2A activation is a key mechanism underlying hexaraphane-induced TAU dephosphorylation.•Chronic oral hexaraphane treatment lowers brain and circulating pTAU levels and attenuates neuroinflammation in APP/TAU mice.•Hexaraphane improves cognitive and motor performance, supporting its potential as a disease-modifying therapy for tauopathies.

Hexaraphane significantly reduces pathological TAU hyperphosphorylation in cellular and in vivo models.

Reduce-pTAU effect by hexaraphane occurs independently of NRF2 activity and is not mediated by GSK-3β inhibition.

PP2A activation is a key mechanism underlying hexaraphane-induced TAU dephosphorylation.

Chronic oral hexaraphane treatment lowers brain and circulating pTAU levels and attenuates neuroinflammation in APP/TAU mice.

Hexaraphane improves cognitive and motor performance, supporting its potential as a disease-modifying therapy for tauopathies.

Figures

Full text

Introduction

1

Alzheimer's disease (AD) is the most common neurodegenerative disorder [1], with a global prevalence exceeding 50 million cases in 2015 and projected to rise to 135 million by 2050. Despite over a century of intensive research, the discovery of effective therapies remains a critical objective [2]. Although AD is neuropathologically defined by the accumulation of extracellular amyloid-β (Aβ) plaques and intracellular neurofibrillary tangles (NFTs), mounting evidence indicates that pathological TAU hyperphosphorylation is a central driver of neurodegeneration, synaptic dysfunction, and cognitive decline.

The phosphorylation state of TAU is tightly regulated by the balance between kinases and phosphatases [[3], [4], [5]]. Disruption of this equilibrium leads to TAU hyperphosphorylation, microtubule destabilization, and neuronal dysfunction. Among TAU-directed enzymes, glycogen synthase kinase-3β (GSK-3β) and cyclin-dependent kinase 5 (CDK5) are major kinases promoting pathological phosphorylation, whereas protein phosphatase 2A (PP2A) accounts for most TAU dephosphorylation activity in the brain. Importantly, PP2A activity is reduced in AD, further contributing to TAU pathology [[3], [4], [5]]. Despite extensive efforts to inhibit TAU kinases, therapeutic success has been limited, underscoring the need for alternative strategies aimed at restoring phosphatase activity.

Hexaraphane (HXN; 6-(methylsulfinyl) hexyl isothiocyanate) is a bioactive compound found in Wasabia japonica that exhibits anti-inflammatory, antioxidant, and neuroprotective properties [[6], [7], [8], [9], [10]]. Additionally, HXN has a low toxicity risk, making it an attractive candidate for the prevention or treatment of several diseases. HXN activates the NRF2/ARE pathway [9,[11], [12], [13]], a master regulator of cellular homeostasis [14,15], leading to the activation of a broad range of cytoprotective genes involved in detoxification, biotransformation, antioxidant defense, inflammation, and intermediary metabolism [15]. NRF2 transcriptional activity is controlled by KEAP1-dependent ubiquitination, which maintains very low basal NRF2 levels [16,17]. However, under oxidative or electrophilic stress, KEAP1 undergoes cysteine modifications that impair its ubiquitin ligase activity, leading to NRF2 stabilization and nuclear accumulation [17,18]. Various electrophilic KEAP1 inhibitors, including TBE-31, omaveloxolone, and isothiocyanates such as sulforaphane (SFN) and HXN, have been used to mitigate disease progression in mouse models [13,[19], [20], [21]].

The neuroprotective effects of HXN have been addressed in the APP^NL−G-F/NL−G-F^ knock-in transgenic mice (App^NLGF^) [13] and after intracerebroventricular Aβ_1-42_ peptide injection [9]. In these amyloidopathy models, NRF2 activation by HXN improved cognitive deficits during a passive-avoidance task [13]. Additionally, intraperitoneal administration of HXN alleviated Aβ_1-42_ oligomer-induced memory impairment, oxidative stress, neuroinflammation, and hippocampal neuronal degeneration in Aβ_1-42_-injected mice [9]. However, while these findings suggest a therapeutic effect for HXN in amyloidopathy, its potential benefits in mitigating TAU hyperphosphorylation remain unexplored.

Given the interplay between TAU pathology, oxidative stress, and neuroinflammatory responses, we hypothesized that HXN may exert disease-modifying effects by modulating TAU phosphorylation and restoring cellular homeostasis. Therefore, we investigated the impact of HXN on pathological TAU phosphorylation, neuroinflammatory markers, and functional outcomes in primary neurons and APP/TAU transgenic mice with either NRF2-competent or NRF2-deficient backgrounds, aiming to elucidate both NRF2-dependent and independent mechanisms of action.

Material and method

2

An additional description of some methods is presented in the Supplemental Material.

Cell culture and reagents. Mouse hippocampus-derived HT22 cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS, HyClone, CH30160.03) with 80 μg/ml gentamicin (Normon Laboratories). HXN (6-methylsulfonyl hexyl isothiocyanate or 6-MSITC), provided by Kinjirushi Company Ltd. Other reagents, including tert-butylhydroquinone (tBHQ, Fluka 19986), SB216763 (Sigma-Aldrich S3442), LY294002 (TOCRIS, 934389-88-5), and Okadaic acid (OA, Sigma-Aldrich O7760). Okadaic acid, SB216763, LY294002, and tBHQ were dissolved in dimethyl sulfoxide (DMSO). The final concentration of DMSO in cell culture was less than 0.2%.

Primary neuron cells. Primary neuronal cultures derived from C57BL/6J mice were prepared as previously [22] from the cortices of embryonic day 16 (E16) from APP/TAU-NRF2^WT^ (AT-NRF2^WT^) and APP/TAU-NRF2^KO^ (AT-NRF2^KO^)-NRF2^WT^ mice of either sex (see details in Supplemental Material).

Primary astrocytes and microglia cells. This protocol was performed as previously described [23] (see details in Supplemental Material).

Animals and treatments. All procedures were performed in accordance with the Guide for the Care and Use of Laboratory Animals and were previously approved by the Autonomous Community of Madrid (PROEX 061.5/25). APP/TAU-NRF2^WT^ (AT-NRF2^WT^) and APP/TAU-NRF2^KO^ (AT-NRF2^KO^) mice on the C57/BL6J background were described earlier in Refs. [24,25]. (see details in Supplemental Material). Animals were housed at room temperature under a 12-h light-dark cycle. Food and water were provided ad libitum. Animal care complied with protocols approved by the Ethical Committee for Research of the Autonomous University of Madrid, following institutional, Spanish, and European guidelines (Boletín Oficial del Estado (BOE) of March 18, 1988; and 86/609/EEC, 2003/65/EC European Council Directives) (see details in Supplemental Material).

Pharmacokinetic HXN assay. C57BL/6J mice were randomized and allocated into five experimental groups (n = 4 per group), corresponding to different time points (0 h, 0.5 h, 1 h, 2 h, 4 h, and 8 h). Mice received HXN (100 mg/kg) by oral gavage and were sacrificed at the designated time points for the collection of serum, liver, and brain regions (neoneocortex, hippocampus, and brainstem). All samples were processed as described below and subsequently quantified by the HPLC-MS.

High-performance liquid chromatography (HPLC-MS). The mouse liver and brain samples (∼80 mg) were treated with 0.5 ml of methanol, ground in a potter, and centrifuged. Mouse serum samples were solubilized in methanol and centrifuged. The supernatant was collected and filtered through a 0.45 μm PTFE microfilter (Fisherbrand). The samples were first analyzed by HPLC-UV. The eluted peaks were further analyzed by HPLC-MS. As a standard, HXN was prepared in methanol and analyzed by HPLC-MS.

Electrophysiological recordings. Long-term potentiation (LTP) was induced in the dentate gyrus by theta-burst electrical stimulation at the perforant path. Data were obtained from six urethane-anesthetized (1.6 g/kg i.p.) adult mice per group as described in Ref. [26] (see details in Supplemental Material).

Novel object recognition test (NOR). During the habituation stage, animals were introduced into an empty cage and allowed to explore freely for 10 min. After a 24-h interval, they were presented with two identical objects and permitted to investigate them for 8 min. In the testing session, conducted 24 h later, one of the familiar objects was replaced with a novel item differing in both shape and color. Memory performance was evaluated over 6 min by comparing the duration of interaction with the new object to that with the familiar one. To eliminate potential olfactory influences, all objects were carefully cleaned between trials using 40% ethanol followed by distilled water. Exploratory behavior was defined as direct nose contact, sniffing, or whisker-oriented investigation toward the objects. Behaviors such as climbing on objects were excluded from the analysis. Exploration preference was calculated manually by determining the difference between the time spent investigating the novel object (Tn) and the familiar object (Tf), divided by the total exploration time (Ttot). This value was expressed as the Novel Object Recognition (NOR) discrimination index using the formula: NOR index = (Tn − Tf)/Ttot.

Evaluation of motor alterations (Treadmill). Exercise training was conducted using a four-lane enclosed treadmill system equipped with a compatible controller (LE 8710, Harvard Apparatus, USA). The apparatus featured an adjustable incline ranging from −15° to 25° (including a flat 0° setting). A built-in electrical stimulus system delivered mild shocks (0.2–2.0 mA) whenever an animal contacted the metal grid positioned at the rear of the running belt. The lanes were separated by partitions, and ventilation openings located at the front and back of each lane allowed passive airflow within the enclosure. All training sessions were performed at a consistent time of day (once per week). Each session began with a 5-min warm-up period, which was not counted toward the total training duration, allowing the animals to progressively reach the predetermined target speed. Training progression was structured across three-week intervals (weeks 0, 3, and 5). Motor performance changes were assessed using the previously established A-score system [27], which evaluates body posture and behavioral responses of the mouse during treadmill running or while attempting to avoid running (see details in Supplemental Material).

Mouse blood sample TAU-pThr^217^ analysis. A fully automated Simoa® HD-X Analyser (Quanterix, Billerica, MA, USA) was used for the quantification of mouse AT-NRF2^WT^ serum using the commercially available kit ALZpath p-TAU 217 (#104570) and according to the manufacturer's instructions with modifications (Quanterix). Two quality control samples were run at the same time as the samples for each assay. Calibrators and serum samples, in the case of Simoa, were run in duplicate, and the average of the two measurements was used for statistical analysis. Samples with coefficients of variation higher than 20% were excluded.

Lentiviral vector production and infection. The lentivirus particles used in this study, including pWPXL-GFP and pWPXL-HA/GSK3β^Δ9^, were generated in HEK293T cells. Briefly, a transfection mixture containing 6 μg of envelope plasmid (pMD2G, Addgene #12259), 6 μg of packaging plasmid (psPAX2, Addgene #12260), and 10 μg of the transfer vector was prepared in DMEM (Sigma-Aldrich). This plasmid combination was introduced into HEK293T cells (2 × 10^6^ cells per 100 mm dish) using Lipofectamine (Invitrogen, 18324-012) together with Plus Reagent (Invitrogen, 11514-015). The cells were then maintained in culture for 24 h. After incubation, the viral supernatant was collected and filtered through a 0.45 μm membrane to remove cellular debris. Target cells were subsequently transduced in medium supplemented with 4 μg/mL polybrene (Sigma-Aldrich) to enhance viral infection efficiency.

Immunoblotting. The primary antibodies used in this work are described in Suppl. Table 1. Briefly, cells were washed once with cold PBS and lysed with lysis buffer (TRIS pH 7.6 50 mM, 400 mM NaCl, 1 mM EDTA, 1 mM EGTA, and 1% SDS). The samples were sonicated and precleared by centrifugation, and cell lysates were resolved in SDS-PAGE and transferred to Immobilon-P membranes (Millipore, Billerica, MA). These membranes were analyzed using the primary antibodies indicated above and the appropriate peroxidase-conjugated secondary antibodies. Proteins were detected by enhanced chemiluminescence (GE Healthcare).

PP2Ac assay. This enzymatic activity was determined as reported elsewhere [28] using a PP2A immunoprecipitation phosphatase assay (No. 17-313; Upstate Biotech) that measures free phosphate with a malachite green dye (see details in Supplemental Material).

Analysis of mRNA levels by real-time quantitative PCR. Total RNA was extracted using TRIzol reagent according to the manufacturer's instructions (Invitrogen). Reverse transcription and quantitative PCR were done as detailed elsewhere [29]. Primer sequences are shown in Suppl. Table 2. Data analysis was based on the ΔΔCT method with normalization of the raw data to housekeeping mouse genes Vcl and Gapdh (Applied Biosystems). All PCRs were performed in triplicate.

Immunohistochemistry and immunofluorescence. 30 μm-thick sections from fixed brains were immunostained as indicated in Ref. [30], with appropriate primary antibodies (Suppl. Table 1). For immunohistochemistry, the sections were subsequently incubated in 0.05% 30-30 diaminobenzidine tetrahydrochloride (Sigma-Aldrich) in Tris-HCl buffer, pH 8.0, for 25 min, and then developed in the same buffer containing 0.003% hydrogen peroxide (Sigma-Aldrich). The sections were mounted on gelatin-coated slides, air-dried, and finally dehydrated in graded alcohols, cleared in xylene, and coverslipped. For immunofluorescence, the sections were incubated with secondary antibodies Alexa-Fluor^546^ or Alexa-Fluor^488^. Images were acquired using the LSM710 spectral microscope confocal (Zeiss, Germany) and analyzed using Fiji's plugin ComDetv.0.5.5. The lasers used were Ar 488 nm for green fluorescence and Ar/HeNe 543 nm for red fluorescence.

Immunohistochemistry and immunofluorescence images analysis. Quantitative image analysis was performed using ImageJ/Fiji by an experimenter blinded to genotype and treatment. For all analyses, regions of interest were defined using consistent anatomical landmarks and identical region sizes across animals. Thresholding parameters were established empirically based on signal-to-noise characteristics and then applied uniformly across all experimental groups within each experiment. Background subtraction was performed using the same settings for all images, and no post hoc adjustments were made based on experimental condition.

Western blot image analysis. Densitometric analyses of immunoblots were performed using ImageJ. Band intensities were quantified within identical rectangular regions of interest and normalized to the corresponding loading control. All samples from a given experiment were processed and quantified in parallel using identical settings.

Statistical Analysis. Data were analyzed using parametric tests under the assumption of approximate normal distribution, which is generally considered appropriate for continuous biochemical measurements obtained from independent biological replicates, given the robustness of parametric tests to moderate deviations from normality [31,32]. Unless otherwise indicated, all experiments were performed at least 3 times, and all data presented in the graphs represent the mean of at least 3 independent experiments. Data are presented as mean ± S.D. (standard deviation). Statistical differences between groups were assessed using GraphPad Prism 8 software by the unpaired Student's t-test. One-way and two-way ANOVA analyses of variance followed by Bonferroni's post-hoc test were used for multiple comparisons. A p-value ≤0.05 was considered statistically significant. Statistically significant differences are indicated in the figures (∗∗∗p values < 0.001, ∗∗ < 0.01 and ∗ < 0.05) (###p values < 0.001, ## < 0.01 and # < 0.05).

Results

3

HXN reduces pTAU levels in primary neurons independently of NRF2

3.1

To determine if hexaraphane (HXN) modulates pathological TAU phosphorylation, we used primary cortical neurons derived from APP/TAU transgenic mice (AT) with either wild-type NRF2 expression (AT-NRF2^WT^) or NRF2 deficiency (AT-NRF2^KO^). The APP/TAU model recapitulates key features of amyloid and tau pathology [33,34], whereas NRF2-deficient mice have been previously characterized as a loss-of-function model for redox and inflammatory regulation [14,16]. These genetic backgrounds allowed us to evaluate both NRF2-dependent and independent effects of HXN on TAU phosphorylation. Primary neurons were maintained for 14 DIV to reach maturity [22] and treated with increasing concentrations of HXN (1, 3, and 9 μM) for 16 h. As expected, HXN induced NRF2 stabilization and increased HO1 protein levels in AT-NRF2^WT^ neurons, confirming activation of the NRF2 pathway (Fig. 1A and B). Consistently, transcript levels of canonical NRF2 target genes, including Hmox1, Nqo1, Slc7a11, and Osgin1, were significantly upregulated following HXN treatment (Fig. 1C). No significant induction of these targets was observed in AT-NRF2^KO^ neurons. Importantly, HXN produced a marked dose-dependent decrease in the levels of two pTAU epitopes, namely AT8 (pSer^202^/pThr^205^) and PHF1 (pSer^396^/pSer^404^), both involved in the onset and progression of TAU pathology [35] (Fig. 1A and B) without altering total TAU protein levels.Fig. 1HXN reduces pTAU levels in an NRF2-independent manner. (A) Primary neurons from AT-NRF2^WT^ mice were maintained for 14 DIV and then treated with the indicated concentrations of HXN for 16 h. Representative immunoblots of NRF2 (arrowhead) and HO1, with VCL and GAPDH used as loading controls, and pTAU (AT8) and pTAU (PHF1), with total TAU used as loading control. (B) Densitometric quantification of NRF2, HO1, pTAU (AT8), and pTAU (PHF1) protein levels from the immunoblots shown in (A), expressed as ratios to VCL, GAPDH, or total TAU, respectively. Data are presented as mean ± S.D. (n = 3 independent experiments). ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001 vs. vehicle-treated cells according to one-way ANOVA followed by Bonferroni's post-hoc test. (C) Primary neurons from AT-NRF2^WT^ mice were treated with the indicated concentrations of HXN for 16 h. Transcript levels of Hmox1, Nqo1, Slc7a11, and Osgin1 were determined by quantitative real-time PCR (qRT-PCR) and normalized to the geometric mean of Gapdh, Vcl, Tbp, and Actb. Data are presented as mean ± S.D. (n = 3 independent experiments). ∗∗p < 0.01; ∗∗∗p < 0.001 vs. vehicle-treated cells according to one-way ANOVA followed by Bonferroni's post-hoc test. (D) Primary neurons from AT-NRF2^WT^ and AT-NRF2^KO^ mice were treated with 9 μM HXN for 16 h. Representative immunoblots of NRF2 (arrowhead) and HO1, with VCL and GAPDH as loading controls, and pTAU (AT8) and pTAU (PHF1), with total TAU as loading control. (E) Densitometric quantification of NRF2, HO1, pTAU (AT8), and pTAU (PHF1) protein levels from the immunoblots shown in (D), expressed as ratios to VCL, GAPDH, or total TAU, respectively. Data are presented as mean ± S.D. (n = 3 independent experiments). ∗∗p < 0.01; ∗∗∗p < 0.001 vs. respective vehicle-treated cells; ##p < 0.01; ###p < 0.001 vs. AT-NRF2^WT^ cells according to Student's t-test. (F) Primary neurons from AT-NRF2^WT^ and AT-NRF2^KO^ mice were treated with 9 μM HXN for 16 h. Transcript levels of Hmox1, Nqo1, Slc7a11, and Osgin1 were determined by qRT-PCR and normalized to the geometric mean of Gapdh, Vcl, Tbp, and Actb. Data are presented as mean ± S.D. (n = 3 independent experiments). ∗∗∗p < 0.001 vs. respective vehicle-treated cells; ##p < 0.01 vs. AT-NRF2^WT^ cells according to Student's t-test.Fig. 1

To get insight if HXN was regulating pTAU levels in an NRF2-dependent manner, we compared the effect in primary neurons of AT-NRF2^WT^ and AT-NRF2^KO^ mice. Neurons were incubated with 9 μM of HXN at different time points (2, 4, and 8 h) (Suppl. Fig. S1Α-C) and at the final time point, 16 h (Fig. 1D and E). NRF2 (apparent MW of 105 kDa, over the non-specific lower band) and HO-1 protein levels were increased in AT-NRF2^WT^ neurons and were not detectable in AT-NRF2^KO^ neurons. Consistently, the transcript levels of Hmox1, Slc7a11, Nqo1, and Osgin1 genes were activated by HXN in AT-NRF2^WT^ neurons but absent or markedly attenuated in AT-NRF2^KO^ neurons (Fig. 1D–F). Surprisingly, a significant decrease in pTAU levels was observed in both neuron types as indicated with the AT8 and PHF1 antibodies (Fig. 1D and E). Moreover, we also observed the change in electrophoretic mobility of total TAU protein with the reduction of the upper fosy band and the appearance of the faster migrating bands (Suppl. Fig. S1Α and Fig. S1D). These findings demonstrate that HXN reduces pathological TAU phosphorylation through a mechanism that is largely independent of NRF2 signaling, suggesting the analysis of alternative regulatory pathways.

HXN does not reduce pTAU levels through control GSK-3 activity

3.2

Because GSK-3β is a major kinase responsible for pathological TAU phosphorylation [36], we next examined whether HXN-mediated pTAU reduction could be explained by modulation of GSK-3β activity. Previous studies suggested that HXN may partially affect GSK-3 signaling [6], raising the possibility that suppression of this kinase underlies its effects on TAU phosphorylation.

We first compared the effects of HXN with those of the well-characterized GSK-3 inhibitor SB216763 in primary AT-NRF2^WT^ neurons. As shown in the Suppl. Fig. S2D and S2E, both compounds increased NRF2 and HO-1 protein levels and markedly reduced the AT8 and PHF1 pTAU epitopes, raising the possibility that HXN might reduce pTAU through inhibition of GSK-3. However, pharmacological comparisons alone cannot distinguish between direct GSK-3β inhibition and indirect effects on TAU phosphorylation.

To directly address this question, we designed experimental conditions in which GSK-3β activity was maintained at high levels (Fig. 2). Primary AT-NRF2^WT^ neurons were infected with a lentivirus encoding HA-GSK-3β^Δ9^, a constitutively active mutant lacking the inhibitory Ser^9^ phosphorylation site (Fig. 2A−C) [37], or treated with the PI3K inhibitor LY294002, which suppresses AKT activity and thereby enhances GSK-3β activation (Fig. 2 D−F) [6]. Under both conditions, GSK-3β signaling was robustly activated, as confirmed by decreased GSK-3β-pSer^9^ and AKT-pSer^473^ levels (Fig. 2D and F). Despite this sustained GSK-3β overactivation, HXN consistently reduced AT8 and PHF1 pTAU levels to near-basal values (Fig. 2B, C, 2E, and 2F). These results indicate that HXN-mediated pTAU reduction does not require inhibition of GSK-3β activity.Fig. 2HXN reduces pTAU levels despite GSK-3β overactivity. (A-C) Primary neurons from AT-NRF2^WT^ mice were infected with a lentiviral vector expressing either GFP (control) or constitutively active HA-GSK-3β^Δ9^, lacking its N-terminal pseudosubstrate domain. After infection, cells were treated with 9 μM HXN for 16 h. (A) Representative immunoblots of NRF2 (arrowhead) and HO1, with VCL and GAPDH used as loading controls. (B) Representative immunoblots of GFP and HA (to confirm transduction efficiency), and pTAU (AT8) and pTAU (PHF1), with total TAU used as loading control. (C) Densitometric quantification of NRF2, HO1, pTAU (AT8), and pTAU (PHF1) protein levels from the immunoblots shown in (A) and (B), expressed as ratios to VCL, GAPDH, or total TAU, respectively. Data are presented as mean ± S.D. (n = 3 independent experiments). ∗∗p < 0.01; ∗∗∗p < 0.001 vs. LV-GFP control; ##p < 0.01; ###p < 0.001 vs. LV-HA-GSK-3β^Δ9^ according to Student's t-test. (D–F) Primary neurons from AT-NRF2^WT^ mice were treated with 9 μM HXN, 30 μM LY294002 (PI3K inhibitor), or both treatments for 6 h. (D) Representative immunoblots of NRF2 (arrowhead) and HO1, with VCL and GAPDH as loading controls. (E) Representative immunoblots of pTAU (AT8) and pTAU (PHF1), AKT-pSer^473^ and total AKT, GSK-3β-pSer^9^ and total GSK-3β, with total TAU used as loading control for pTAU. (F) Densitometric quantification of NRF2, HO1, AKT-pSer^473^, GSK-3β-pSer^9^, pTAU (AT8), and pTAU (PHF1) protein levels from the immunoblots shown in (D) and (E), expressed as ratios to VCL, GAPDH, total AKT, total GSK-3β, or total TAU, respectively. Data are presented as mean ± S.D. (n = 3 independent experiments). ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001 vs. vehicle-treated cells; ##p < 0.01; ###p < 0.001 vs. HXN-treated cells according to Student's t-test.Fig. 2

We further examined if HXN modulates upstream signaling pathways controlling GSK-3β activity, comparing the effect of HXN and tert-butylhydroquinone (tBHQ), a known PTEN inhibitor and PI3K/AKT pathway activator that leads to GSK-3 β inhibiton andna activator of MAPK pathways [38]. In contrast to tBHQ, HXN failed to induce significant changes in AKT, GSK-3β, p38, or ERK1/2 phosphorylation (Suppl Fig. S2A−C), despite producing comparable reductions in pTAU levels. Moreover, chronic HXN administration did not significantly alter the phosphorylation status of multiple established GSK-3 substrates in the brain, although these substrates were clearly modulated in peripheral tissues such as the liver (Suppl. Fig. S3). While tBHQ robustly increased AKT-pSer^473^ and GSK-3β-pSer^9^ levels, HXN failed to reproduce these effects at comparable time points (Suppl. Fig. S2A and S2C). Similarly, tBHQ induced phosphorylation of MAPKs, including p38 and ERK1/2, whereas HXN did not elicit significant changes in these pathways (Suppl. Fig. S2A and S2C). Nevertheless, both compounds reduced pTAU levels, albeit with distinct temporal profiles (Suppl. Fig. S2B and S2C), indicating that HXN acts through a mechanism that is largely independent of canonical PI3K/AKT/GSK-3 and MAPK signaling. As an additional control, we examined whether HXN alters the phosphorylation status of other well-established GSK-3 substrates in the brain. Notably, HXN did not significantly modify the protein levels of β-CATENIN, GS-pSer^641^/pSer^645^, GSK-3α-pTyr^279^, GSK-3β-pTyr^216^, or c-JUN-pThr^239^ in the neocortex, hippocampus, or brainstem (Suppl. Fig. S3). In contrast, these GSK-3-related targets were clearly altered in the liver following HXN administration (Suppl. Fig. S4D−G), consistent with the higher concentrations of HXN detected in peripheral tissues compared with the brain (see later).

Taken together, these results indicate that HXN reduces pathological neuronal TAU phosphorylation independently of GSK-3β inhibition or upstream PI3K/AKT and MAPK signaling pathways. The absence of changes in multiple GSK-3 substrates in the brain, despite robust modulation in the liver, strongly supports the conclusion that GSK-3 is not the primary mediator of HXN-induced TAU dephosphorylation in the central nervous system, motivating the exploration of alternative phosphatase-dependent mechanisms.

Pathological pTAU species are reduced by HXN through the regulation of phosphatases

3.3

The coordination between kinases and phosphatases is essential for regulating the phosphorylation status of TAU. Four major serine/threonine phosphatase families, PP1, PP2A, PP2B, and PP2C, account for most of the cellular phosphatase activity [39]. Among them, PP2A has been most strongly implicated in the etiopathogenesis of AD [[40], [41], [42]]. To test this, AT-NRF2^WT^ neurons were treated with HXN alone or in combination with the well-characterized phosphatase inhibitor okadaic acid, which targets several phosphatases but exhibits specificity for PP2A at low concentrations [43]. As shown in Fig. 3A and B, HXN (9 μM, 6 h) increased endogenous NRF2 protein levels and decreased pTAU (AT8 and PHF1 epitopes) as expected. In contrast, okadaic acid (15 nM, 6 h) strongly increased pTAU levels and induced a clear band-shift of TAU protein. Strikingly, co-treatment with okadaic acid completely abolished the effects of HXN, maintaining high pTAU levels, thereby pointing to the requirement of phosphatase activity.Fig. 3Okadaic acid blocks HXN-mediated pTAU reduction through PP2A inhibition. (A) Primary neurons from AT-NRF2^WT^ and AT-NRF2^KO^ mice were treated with 9 μM HXN, 15 nM okadaic acid (OA), or both treatments for 6 h. Representative immunoblots of NRF2 (arrowhead), pTAU (AT8) and pTAU (PHF1), and PP2Ac, with VCL and GAPDH as loading controls and total TAU used as loading control for pTAU. (B) Densitometric quantification of pTAU (AT8) and pTAU (PHF1) protein levels from the immunoblots shown in (A), expressed as ratios to total TAU. Data are presented as mean ± S.D. (n = 3 independent experiments). ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001 vs. vehicle-treated cells; ##p < 0.01; ###p < 0.001 vs. HXN-treated cells according to Student's t-test. (C) Endogenous PP2Ac phosphatase activity assay performed in primary neurons from AT-NRF2^WT^ and AT-NRF2^KO^ mice treated with 9 μM HXN, 15 nM OA, or both for 6 h. Data are presented as mean ± S.D. (n = 3 independent experiments). ∗p < 0.05; ∗∗p < 0.01 vs. vehicle-treated cells according to Student's t-test. (D) Representative immunoblots of immunoprecipitated PP2Ac used as a control for the activity assay shown in (C). (E) In vitro PP2Ac activity assay performed using PP2Ac immunoprecipitated from HT22 cells and incubated with the indicated concentrations of HXN. OA (15 nM) and the absence of primary antibody during immunoprecipitation (Ctrl 1°Ab) were included as controls. Data are presented as mean ± S.D. (n = 3 independent experiments). ∗p < 0.05; ∗∗∗p < 0.001 vs. vehicle according to one-way ANOVA followed by Bonferroni's post-hoc test; ##p < 0.01 vs. vehicle according to Student's t-test. (F) Representative immunoblot of immunoprecipitated PP2Ac used as a control for the activity assay shown in (E). (G) Primary neurons from AT-NRF2^WT^ mice were treated with the indicated concentrations of OA for 6 h, in the presence or absence of 9 μM HXN. Representative immunoblots of pTAU (AT8) and pTAU (PHF1), and PP2Ac, with VCL and GAPDH as loading controls, and total TAU used as a loading control for pTAU. (H) Densitometric quantification of pTAU (AT8) and pTAU (PHF1) protein levels from the immunoblots shown in (G), expressed as ratios to total TAU. Data are presented as mean ± S.D. (n = 3 independent experiments). ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001 vs. vehicle-treated cells according to one-way ANOVA followed by Bonferroni's post-hoc test.Fig. 3

To further investigate whether HXN promotes TAU dephosphorylation, possibly through PP2A activation, we incubated AT-NRF2^WT^ and AT-NRF2^KO^ neurons with 15 nM okadaic acid or 9 μM HXN for 6 h. Then, we immunoprecipitated the catalytic subunit of PP2A (PP2Ac) and performed an in vitro PP2A phosphatase assay using malachite green as substrate. HXN increased PP2A activity, while okadaic acid inhibited this effect (Fig. 3C). Fig. 3D shows the PP2Ac protein levels as a control for the immunoprecipitation step in the enzymatic assay. To further determine if HXN directly modulates PP2A activity rather than acting exclusively through upstream regulatory pathways, we next assessed PP2A enzymatic activity in immunoprecipitated complexes. We immunoprecipitated PP2Ac protein from HT22 hippocampal cells and conducted an in vitro phosphatase assay with several HXN concentrations (0.1−9 μM), using okadaic acid as a control. As shown in Fig. 3E and F, HXN exhibited an inverted U-shaped dose-response curve, producing activating effects at 0.3 μM and inhibitory effects at higher suboptimal concentrations (3–9 μM). These findings indicate that HXN can directly regulate PP2A activity.

Although PP2A is considered the dominant phosphatase regulating pTAU, several others, including PP1, PP2B, PP5, and PTEN, have also been implicated in abnormal TAU phosphorylation and aggregation [42,[44], [45], [46]]. Therefore, we tested if additional phosphatases might contribute to HXN-mediated pTAU reduction by treating neurons with increasing concentrations of okadaic acid (0.5−45 nM) that progressively inhibit multiple phosphatases. As shown in Fig. 3G and H, okadaic acid induced a dose-dependent increase in pTAU levels at the AT8 and PHF1 epitopes. Notably, HXN produced a modest reduction in pTAU levels at low okadaic acid concentrations that selectively inhibit PP2A. These findings suggest that when PP2A activity is inhibited, HXN may still act through additional phosphatases to reduce pTAU. Consistent with this interpretation, the effect of HXN was lost at higher okadaic acid concentrations that inhibit a broader range of phosphatases. Together, our data suggest that HXN upregulates PP2A, the principal TAU phosphatase, and also suggest that additional TAU phosphatases might also participate.

HXN decreases pTAU species in a mouse model of tauopathy

3.4

Having established that HXN reduces pathological TAU phosphorylation through phosphatase-dependent mechanisms in primary neurons, we next evaluated whether these effects translate into vivo models of tauopathy. AT-NRF2^WT^ and AT-NRF2^KO^ mice were administered HXN (100 mg/kg) or vehicle by oral gavage for five consecutive days. Four hours after the final administration, mice were sacrificed, and neocortex and brainstem regions exhibiting high TAU pathology [25] were analyzed. To confirm systemic delivery of HXN, we used liver tissue as a control. In AT-NRF2^WT^ mice, HXN increased NRF2 protein levels and induced its transcriptional signature, whereas no changes were observed in AT-NRF2^KO^ mice (Suppl. Fig. S5A−C). HXN also significantly increased PP2A activity in liver tissue by approximately two-fold relative to controls, in both NRF2 genotypes (Suppl. Fig. S5D). Finally, the presence of HXN in the liver was confirmed by HPLC analysis (Suppl. Fig. S5E).

Neocortex

3.4.1

HPLC analysis confirmed the presence of HXN in the neocortex (Fig. 4D), indicating that HXN crosses the blood-brain barrier. Consequently, HXN significantly decreased pTAU levels at the AT8 and PHF1 epitopes in both AT-NRF2^WT^ and AT-NRF2^KO^ mice (Fig. 4A and B). Although under these conditions, we did not detect a significant increase in NRF2 protein levels, some NRF2 target genes, such as Hmox1 and Slc7a11, were modestly upregulated, whereas Aox1 and Osgin1 remained unchanged (Fig. 4A and C). The lack of detectable NRF2 protein induction may reflect that the amount of HXN reaching the neocortex is sufficient to slightly activate downstream targets but insufficient to increase NRF2 protein levels appreciably. To further assess TAU phosphorylation, we performed immunohistochemistry using the AT8 antibody. No major differences in overall pTAU AT8 staining were observed between HXN-treated and vehicle-treated mice in AT-NRF2^WT^ (Fig. 4E) and AT-NRF2^KO^ (Suppl. Fig. S6A). However, intracellular pTAU aggregates and dystrophic neurites appeared less pronounced and more organized upon HXN treatment. In vehicle-treated mice, pTAU-positive neurites were abundant but disorganized and poorly stained, whereas HXN treatment resulted in more traceable, coherent fibers (Fig. 4E and Suppl. Fig. S6A). Immunofluorescence analysis confirmed a reduction in pTAU AT8 staining upon HXN treatment compared with vehicle in AT-NRF2^WT^ mice (Fig. 4F and G), suggesting that HXN reduces pTAU burden in vivo. Consistent with previous liver and primary neuron data, HXN also significantly increased PP2A activity by approximately 1.5-fold in both NRF2 genotypes (Fig. 4H and I), supporting its role in TAU dephosphorylation.Fig. 4Acute HXN administration reduces pTAU levels and increases PP2A activity in the neocortex. (A) AT-NRF2^WT^ and AT-NRF2^KO^ mice were treated with vehicle or HXN (100 mg/kg) by oral gavage for 5 consecutive days. Representative immunoblots of NRF2 (arrowhead), pTAU (AT8), and pTAU (PHF1), and PP2Ac from neocortical extracts, with VCL and GAPDH used as loading controls, and total TAU used as a loading control for pTAU. (B) Densitometric quantification of pTAU (AT8), pTAU (PHF1), and PP2Ac protein levels from the immunoblots shown in (A), expressed as ratios to total TAU or GAPDH, respectively. Data are presented as mean ± S.D. (n = 5 animals per group). ∗∗p < 0.01 vs. vehicle-treated mice; #p < 0.05; ###p < 0.001 vs. AT-NRF2WT mice according to Student's t-test. (C) Transcript levels of NRF2 target genes (Hmox1, Aox1, Slc7a11, and Osgin1) in neocortical tissue were determined by quantitative real-time PCR (qRT-PCR) and normalized to the geometric mean of Gapdh, Vcl, Tbp, and Actb. Data are presented as mean ± S.D. (n = 5 animals per group). ∗p < 0.05; ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (D) Qualitative HPLC–UV analysis of neocortical extracts confirming the presence of HXN in AT-NRF2^WT^ and AT-NRF2^KO^ mice after oral administration (n = 5 animals per group). (E) Immunohistochemical staining of sagittal brain sections from AT-NRF2^WT^ mice using anti-pTAU (AT8) antibody in the indicated treatment groups. (F) Double immunofluorescence staining of sagittal brain sections from AT-NRF2^WT^ mice. Green: pTAU (AT8). Red: total TAU (SP70). (G) Quantification of pTAU (AT8) fluorescence intensity normalized to total TAU (SP70) from images shown in (F). Data are presented as mean ± S.D. (n = 5 animals per group). ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (H) Endogenous PP2Ac phosphatase activity assay performed in neocortical extracts from AT-NRF2^WT^ and AT-NRF2^KO^ mice treated with vehicle or HXN (100 mg/kg) for 5 days. Data are presented as mean ± S.D. (n = 5 animals per group). ∗p < 0.05; ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (I) Representative immunoblots of immunoprecipitated PP2Ac used as a control for the phosphatase activity assay shown in (H).Fig. 4

Brainstem

3.4.2

In parallel with the neocortex, we assessed HXN's ability to reduce pTAU levels in the brainstem. As shown in Fig. 5A and B, HXN significantly decreased pTAU levels at the AT8 and PHF1 epitopes in both AT-NRF2^WT^ and AT-NRF2^KO^ mice. Similar to the neocortex, NRF2 protein levels were not significantly altered while the NRF2 target genes Osgin1 and Slc7a11 were upregulated, and Aox1 and Hmox1 remained unchanged (Fig. 5A and C). Immunofluorescence analysis confirmed a reduction in pTAU AT8 staining with HXN treatment compared to vehicle in AT-NRF2^WT^ mice (Fig. 5D and E), further supporting HXN's ability to reduce pTAU burden also in the brainstem. Immunohistochemistry using AT8 and PHF1 antibodies revealed that overall patterns of pTAU staining were similar between vehicle- and HXN-treated mice across both AT-NRF2^WT^ (Fig. 5F and G) and AT-NRF2^KO^ (Suppl. Fig. S6B and S6C), with intracellular pTAU aggregates and dystrophic neurites still present. Notably, pTAU-positive neurites in vehicle-treated mice were abundant but disorganized and poorly stained, whereas HXN treatment resulted in fibers that were more traceable and organized, consistent with a reduction in pathological TAU burden (Fig. 5F and G and Suppl. Fig. S6B and S6C).Fig. 5Acute HXN administration reduces pTAU levels in the brainstem. (A) AT-NRF2^WT^ and AT-NRF2^KO^ mice were treated with vehicle or HXN (100 mg/kg) by oral gavage for 5 consecutive days. Representative immunoblots of NRF2 (arrowhead), pTAU (AT8) and pTAU (PHF1), and PP2Ac from brainstem extracts, with VCL and GAPDH used as loading controls, and total TAU used as loading control for pTAU. (B) Densitometric quantification of pTAU (AT8), pTAU (PHF1), and PP2Ac protein levels from the immunoblots shown in (A), expressed as ratios to total TAU or GAPDH, respectively. Data are presented as mean ± S.D. (n = 5 animals per group). ∗p < 0.05; ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (C) Transcript levels of NRF2 target genes (Hmox1, Aox1, Slc7a11, and Osgin1) in brainstem tissue were determined by quantitative real-time PCR (qRT-PCR) and normalized to the geometric mean of Gapdh, Vcl, Tbp, and Actb. Data are presented as mean ± S.D. (n = 5 animals per group). ∗p < 0.05; ∗∗p < 0.01 vs. vehicle-treated mice; ###p < 0.001 vs. AT-NRF2^WT^ mice according to Student's t-test. (D) Double immunofluorescence staining of sagittal brain sections from AT-NRF2^WT^ mice. Green: pTAU (AT8). Red: total TAU (SP70). (E) Quantification of pTAU (AT8) fluorescence intensity normalized to total TAU (SP70) from images shown in (D). Data are presented as mean ± S.D. (n = 5 animals per group). ∗∗∗p < 0.001 vs. vehicle-treated mice according to Student's t-test. (F) Immunohistochemical staining of sagittal brain sections from AT-NRF2^WT^ mice using anti-pTAU (AT8) antibody in the indicated treatment groups. (G) Immunohistochemical staining of sagittal brain sections from AT-NRF2^WT^ mice using anti-pTAU (PHF1) antibody in the indicated treatment groups.Fig. 5

HXN lowers the TAU-pThr217 levels in blood

3.5

Having demonstrated that HXN reduces pathological TAU burden in brain regions, we next asked whether this effect can be detected in blood as a clinically relevant outcome, using the ultra-sensitive Quanterix-Simoa immunoassay kit ALZpath p-TAU 217. We focused on the TAU-pThr^217^ epitope, as it has been identified as a promising biomarker for early AD detection and differential diagnosis [47]. As shown in Fig. 6A, baseline levels of TAU-pThr^217^ were comparable across groups; however, following HXN treatment, AT-NRF2^WT^ mice exhibited a marked reduction in seric TAU-pThr^217^ levels compared with vehicle-treated controls. Longitudinal analysis of individual animals further confirmed these findings, as HXN-treated animals showed a significant decrease over 5 weeks (Fig. 6B). These results show that the levels of pathological pTAU species can be readily used in drug response and disease monitoring in tauopathies.Fig. 6Chronic oral HXN administration reduces circulating pTAU levels in AT-NRF2^WT^ mice. (A) Plasma levels of TAU phosphorylated at Thr^217^ (pThr^217^) were measured in AT-NRF2^WT^ mice treated with vehicle or HXN (100 mg/kg) by oral gavage for 5 weeks. Blood samples were collected before the first administration (baseline) and at the end of the treatment period (final). Data are presented as mean ± S.D. (n = 6 animals per group). ∗p < 0.05 vs. baseline (initial bleeding) according to Student's t-test. (B) Longitudinal representation of individual animals showing changes in circulating pTAU (Thr^217^) levels over the 5-week treatment period corresponding to the data shown in (A).Fig. 6

HXN attenuates inflammatory responses in primary glial cells

3.6

To determine whether the anti-inflammatory effects of hexaraphane (HXN) observed in vivo could arise from a direct action on glial cells. Primary astrocytes and microglia isolated from neonatal mouse brains were stimulated with lipopolysaccharide (LPS) to induce a robust inflammatory response and subsequently treated with HXN. As expected, LPS exposure markedly increased the expression of classical inflammatory mediators at both the protein and transcript levels. In contrast, HXN treatment significantly attenuated the induction of pro-inflammatory markers in both astrocytes and microglia, while concomitantly increasing NRF2 and HO-1 protein levels (Suppl. Fig. S7). Specifically, in primary astrocytes, HXN reduced the LPS-induced upregulation of NOS2, COX2, IL-6, and pre-IL-1β protein levels (Suppl. Fig. S7A and S7B), as well as the transcript levels of Il1b, Ptgs2, Il6, and Tnf (Suppl. Fig. S7C). Similarly, in primary microglia, HXN markedly suppressed the induction of NOS2, COX2, IL-6, and pre-IL-1β proteins (Suppl. Fig. S7D and S7E), together with reduced expression of Il1b, Ptgs2, and Tnf mRNAs (Suppl. Fig. S7F). These results demonstrate that HXN exerts a direct anti-inflammatory effect on glial cells in vitro and provide a rationale for subsequent analyses of HXN in mice.

Pharmacokinetic profile of HXN in plasma, liver, and brain

3.7

To gain insight into the regional and functional effects of chronic HXN administration, we first characterized its pharmacokinetic profile following acute and repeated dosing. These analyses were designed to determine systemic exposure, brain penetration, regional distribution, and the temporal relationship between HXN availability and its biological effects.

Pharmacokinetic profiling following a single oral administration of HXN (100 mg/kg) revealed rapid systemic absorption, with peak serum concentrations of approximately 60 ppb detected at 0.5 h post-gavage (Suppl. Fig. S8A). Plasma levels declined sharply thereafter, reaching approximately 25 ppb at 2 h and approaching baseline values between 4 and 8 h, indicating fast clearance. Liver concentrations followed a similar temporal profile, with peak levels of approximately 40-45 ppb at 0.5-1 h and a rapid decline thereafter, consistent with efficient distribution and elimination (Suppl. Table 4). HXN was also detected in the brain, albeit at lower concentrations than in plasma, with a region-dependent distribution pattern (Suppl. Fig. S8B). Peak brain concentrations were observed between 0.5 and 1 h post-administration, reaching approximately 20 ppb in the neocortex, 12-15 ppb in the brainstem, and 4-6 ppb in the hippocampus. These values correspond to estimated brain-to-plasma ratios of ∼0.3 for the neocortex, ∼0.2 for the brainstem, and ∼0.1 for the hippocampus at peak exposure. Brain concentrations declined rapidly and were near or below the limit of detection by 4-8 h, indicating transient central nervous system exposure (Suppl. Table 4). Based on the plasma profiles, the elimination half-life of HXN was estimated to be approximately 1-2 h. These results indicate that orally administered HXN rapidly reaches the brain, crosses the blood-brain barrier, and displays a preferential accumulation in the neocortex and brainstem compared to the hippocampus.

We next evaluated the pharmacokinetic relevance of HXN under chronic treatment. In these experiments, AT-NRF2^WT^ mice received HXN (100 mg/kg, oral gavage) every three days per week for five weeks. The correct delivery of HXN was assessed by the NRF2 transcriptional signature in the liver. As shown in the Suppl. Fig. S4A and S4B, HXN led to increased NRF2 protein levels and induction of its canonical target gene Gclc. In addition, qRT-PCR analysis revealed a statistically significant upregulation of several NRF2 target genes, including Hmox1, Nqo1, Osgin1, Aox1, and Gclc, consistent with NRF2 pathway activation by HXN (Suppl. Fig. S4C).

Although HXN does not accumulate in the brain over time due to its rapid clearance, repeated dosing ensured recurrent transient exposure of brain tissue to pharmacologically active concentrations. Consistent with the acute pharmacokinetic profile, HXN was readily detectable in the neocortex and brainstem following chronic administration, whereas hippocampal levels remained substantially lower. Importantly, the concentrations of HXN achieved in the brain under chronic dosing conditions fall within the low micromolar or submicromolar range, which is sufficient to promote TAU dephosphorylation and functional recovery without eliciting detectable modulation of classical GSK-3 targets in the central nervous system (Suppl. Fig. S3). These findings support the notion that repeated transient brain exposure, rather than sustained accumulation, is sufficient to drive the therapeutic effects of HXN.

Collectively, the pharmacokinetic data demonstrate that HXN exhibits rapid absorption, efficient brain penetration, and region-specific distribution following oral administration.

In the neocortex, HXN markedly reduced pTAU AT8 and PHF1 epitopes compared with vehicle-treated animals (Fig. 7A and C). Immunohistochemistry and immunofluorescence analyses with the pTAU AT8 antibody revealed abundant intracellular pTAU aggregates, thick dystrophic neurites, and an overall higher pTAU burden in the vehicle group, whereas HXN treatment substantially reduced these pathological features (Fig. 7F and G). To further assess neuronal integrity, we evaluated CALBINDIN-D28k expression, a major calcium-binding and buffering protein essential for neuronal survival, calcium homeostasis, and synaptic plasticity. Immunofluorescence analysis showed that neurons from the vehicle group, which exhibited elevated pTAU AT8 levels, had markedly reduced CALBINDIN-D28k expression (Fig. 7H). Importantly, HXN enhanced CALBINDIN-D28k expression, with a higher number of positive neurons detected (Fig. 7H and I), suggesting a neuroprotective effect. We next assessed the impact of HXN on inflammatory responses in the neocortex. HXN resulted in a downward trend in the protein levels of COX2, pro-IL-1β, and IL-6 (Fig. 7B and D), as well as reduced mRNA expression of inflammatory genes, including Il1b, Il6, Ptgs2, Gfap, Iba1, Cd11b, and Inos (Fig. 7E). Astrogliosis and microgliosis were evaluated using GFAP and IBA1 immunostaining, respectively. In the vehicle group, astrocytes displayed enlarged somas and thickened processes (type B morphology), characteristic of a reactive state, whereas HXN significantly reduced GFAP^+^ astrocytes and maintained astrocytes in a resting morphology (type A), indicating a potent anti-inflammatory effect (Fig. 7J and L). Concerning microgliosis, HXN significantly reduced expression in IBA1^+^, compared with the vehicle group, reinforcing the idea that HXN exhibits an anti-inflammatory effect. Regarding morphology, microglia can switch between a quiescent (type A), an initiating microglial activated non-phagocytic (type B), and an activated and phagocytic (type C) state [48] (Fig. 7K). We observed that microglia morphology was activated and phagocytic state in vehicle-groups, while HXN treatment showed a microglial morphology related to a resting state, corroborating an anti-inflammatory effect (Fig. 7K and M).Fig. 7Chronic HXN administration reduces pTAU levels and attenuates neuroinflammatory markers in the neocortex of AT-NRF2^WT^ mice. AT-NRF2^WT^ mice were treated with vehicle or HXN (100 mg/kg) by oral gavage for 5 weeks (n = 6 animals per group). (A) Representative immunoblots of pTAU (AT8) and pTAU (PHF1) from neocortical extracts, with total TAU used as a loading control. (B) Representative immunoblots of COX2, pre-IL-1β, and IL-6 from neocortical extracts, with VCL and GAPDH used as loading controls. (C) Densitometric quantification of pTAU (AT8) and pTAU (PHF1) protein levels from the immunoblots shown in (A), expressed as ratios to total TAU. Data are presented as mean ± S.D. (n = 6 animals per group). ∗p < 0.05; ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (D) Densitometric quantification of COX2, pre-IL-1β, and IL-6 protein levels from the immunoblots shown in (B), expressed as ratios to VCL or GAPDH. Data are presented as mean ± S.D. (n = 6 animals per group). ∗p < 0.05; ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (E) Transcript levels of pro-inflammatory genes (Il1b, Il6, Ptgs2, Gfap, Iba1, Cd11b, F4*/*80, and Inos) determined by quantitative real-time PCR (qRT-PCR) in neocortical tissue and normalized to the geometric mean of Gapdh, Vcl, Tbp, and Actb. Data are presented as mean ± S.D. (n = 6 animals per group). ∗∗p < 0.01; ∗∗∗p < 0.001 vs. vehicle-treated mice according to Student's t-test. (F) Immunohistochemical staining of sagittal brain sections from AT-NRF2^WT^ mice using anti-pTAU (AT8) antibody. (G) Quantification of pTAU (AT8) fluorescence intensity normalized to total TAU (SP70) from double immunofluorescence images. Data are presented as mean ± S.D. (n = 6 animals per group). ∗p < 0.05 vs. vehicle-treated mice according to Student's t-test. (H) Double immunofluorescence staining of sagittal brain sections. Green: pTAU (AT8). Red: CALBINDIN-D28K. (I) Quantification of the ratio of CALBINDIN-D28K-positive cells relative to pTAU (AT8)-positive cells. Data are presented as mean ± S.D. (n = 6 animals per group). Differences among groups were assessed using Student's t-test. (J) Immunohistochemical staining of sagittal brain sections using anti-GFAP antibody. Representative astrocyte morphologies are shown and classified as resting/basal (type A) or reactive (type B). (K) Immunohistochemical staining of sagittal brain sections using anti-IBA1 antibody. Representative microglial morphologies are shown and classified as resting/basal (type A), activated non-phagocytic (type B), or phagocytic (type C) according to Sánchez-Guajardo et al. [48]. (L) Quantification of GFAP immunoreactivity density. Data are presented as mean ± S.D. (n = 6 animals per group). Differences among groups were assessed using Student's t-test. (M) Quantification of IBA1 immunoreactivity density. Data are presented as mean ± S.D. (n = 6 animals per group). Differences among groups were assessed using Student's t-test.Fig. 7

In the hippocampus, HXN resulted in a reduction of pTAU AT8 and PHF1 epitopes similar to that observed in the neocortex (Fig. 8A and C). Immunohistochemistry and immunofluorescence analyses in the CA1 region confirmed the presence of intracellular pTAU aggregates, thick dystrophic neurites, and higher pTAU burden in both somata and axons of the vehicle group, which were markedly reduced in the HXN-treated group (Fig. 8F and G). Unlike the neocortex, assessment of neuronal integrity based on CALBINDIN-D28k expression revealed only a modest increase in CALBINDIN-D28k levels in the HXN group compared with vehicle-treated mice, which did not reach statistical significance (Fig. 8H and I). These findings suggest that HXN may provide lower neuroprotection in the hippocampus compared to the neocortex. Regarding neuroinflammation, HXN modestly reduced COX2 and pre-IL-1β protein levels, as well as the Ptgs2 and Il1b transcripts (Fig. 8B, D, and 8E). By contrast, the expression of other inflammatory genes, including Il6, Gfap, Iba1, Cd11b, and Inos, remained unchanged (Fig. 8E). In line with these results, immunostaining for GFAP and IBA1 showed no significant differences in astrogliosis or microgliosis between HXN- and vehicle-treated groups. Astrocytes in both groups displayed reactive morphology (type B) (Fig. 8J and L), while microglia in vehicle-treated animals were predominantly activated and phagocytic (type C), HXN treatment continued to display reactive type B morphology, and did not seem to restore microglia to a resting state (type A) (Fig. 8K and M). These data indicate that, in contrast to the neocortex, HXN treatment did not elicit a strong anti-inflammatory effect in the hippocampus. Pharmacokinetic analyses revealed lower concentrations of HXN in the hippocampus compared to the neocortex (Suppl. Fig. S8B). This limited bioavailability may underline the modest reduction in inflammatory hallmarks observed in this region.Fig. 8Chronic HXN administration reduces pTAU levels and modestly attenuates glial activation in the hippocampus of AT-NRF2^WT^ mice. AT-NRF2^WT^ mice were treated with vehicle or HXN (100 mg/kg) by oral gavage for 5 weeks (n = 6 animals per group). (A) Representative immunoblots of pTAU (AT8) and pTAU (PHF1) from hippocampal extracts, with total TAU used as a loading control. (B) Representative immunoblots of COX2, pre-IL-1β, and IL-6 from hippocampal extracts, with VCL and GAPDH used as loading controls. (C) Densitometric quantification of pTAU (AT8) and pTAU (PHF1) protein levels from the immunoblots shown in (A), expressed as ratios to total TAU. Data are presented as mean ± S.D. (n = 6 animals per group). ∗p < 0.05; ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (D) Densitometric quantification of COX2, pre-IL-1β, and IL-6 protein levels from the immunoblots shown in (B), expressed as ratios to VCL or GAPDH. Data are presented as mean ± S.D. (n = 6 animals per group). ∗p < 0.05; ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (E) Transcript levels of pro-inflammatory genes (Il1b, Il6, Ptgs2, Gfap, Iba1, Cd11b, F4/80, and Inos) determined by quantitative real-time PCR (qRT-PCR) in hippocampal tissue and normalized to the geometric mean of Gapdh, Vcl, Tbp, and Actb. Data are presented as mean ± S.D. (n = 6 animals per group). ∗∗p < 0.01; ∗∗∗p < 0.001 vs. vehicle-treated mice according to Student's t-test. (F) Immunohistochemical staining of sagittal brain sections using anti-pTAU (AT8) antibody. (G) Quantification of pTAU (AT8) fluorescence intensity normalized to total TAU (SP70) from double immunofluorescence images. Data are presented as mean ± S.D. (n = 6 animals per group). ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (H) Double immunofluorescence staining. Green: pTAU (AT8). Red: CALBINDIN-D28K. (I) Quantification of the ratio of CALBINDIN-D28K-positive cells relative to pTAU (AT8)-positive cells. Data are presented as mean ± S.D. (n = 6 animals per group). Differences among groups were assessed using Student's t-test. (J) Immunohistochemical staining using anti-GFAP antibody. Representative astrocyte morphologies are shown and classified as resting/basal (type A) or reactive (type B). (K) Immunohistochemical staining using anti-IBA1 antibody. Representative microglial morphologies are shown and classified as resting/basal (type A), activated non-phagocytic (type B), or phagocytic (type C) according to Sánchez-Guajardo et al. [48]. (L) Quantification of GFAP immunoreactivity density. Data are presented as mean ± S.D. (n = 6 animals per group). Differences among groups were assessed using Student's t-test. (M) Quantification of IBA1 immunoreactivity density. Data are presented as mean ± S.D. (n = 6 animals per group). Differences among groups were assessed using Student's t-test.Fig. 8

In the brainstem, HXN also led to a reduction of pTAU AT8 and PHF1 epitopes (Fig. 9A and C). Immunohistochemistry and immunofluorescence analysis of pTAU AT8 revealed intracellular aggregates localized predominantly in neuronal bodies, but not in axons, with a markedly higher burden in vehicle-treated animals compared with the HXN-treated group (Fig. 9F and G). Similar to the neocortex, CALBINDIN-D28k immunofluorescence demonstrated that high pTAU levels in the vehicle group were associated with a reduction of this neuroprotective marker (Fig. 9H). Importantly, HXN preserved CALBINDIN-D28k expression, suggesting enhanced neuronal survival (Fig. 9H and I). Assessment of inflammatory parameters in the brainstem revealed significant decreases in the protein levels of COX2, pre-IL-1β, and IL-6 following HXN treatment (Fig. 9B and D). These effects were accompanied by reduced mRNA expression of Il1b, Il6, Ptgs2, and Gfap, while no significant changes were observed for Iba1, Cd11b, F4/80, or Inos (Fig. 9E). Analysis of glial responses further supported the anti-inflammatory effects of HXN. Astrogliosis, measured by GFAP immunostaining, was significantly reduced in HXN-treated animals compared with the vehicle group (Fig. 9J and L). Astrocytes from vehicle-treated mice exhibited reactive morphology (type B), whereas astrocytes from HXN-treated animals maintained a quiescent, resting type A morphology (Fig. 9J). Similarly, microgliosis was significantly attenuated by HXN treatment, as reflected by reduced IBA1^+^ staining compared with vehicle-treated mice (Fig. 9K and M). Morphologically, microglia in vehicle-treated animals were predominantly activated and phagocytic (type C), whereas HXN treatment restored microglia to a resting state (type A) (Fig. 9K and M). Additionally, the pharmacokinetic profile of HXN revealed that the compound reaches comparable concentrations in the brainstem and neocortex (Suppl. Fig. S8B), indicating good bioavailability and efficient distribution through the neurovascular system, and exerts both neuroprotective and anti-inflammatory effects in the brainstem, closely mirroring those observed in the neocortex.Fig. 9Chronic HXN administration reduces pTAU levels and attenuates neuroinflammatory markers in the brainstem of AT-NRF2^WT^ mice. AT-NRF2^WT^ mice were treated with vehicle or HXN (100 mg/kg) by oral gavage for 5 weeks (n = 6 animals per group). (A) Representative immunoblots of pTAU (AT8) and pTAU (PHF1) from brainstem extracts, with total TAU used as a loading control. (B) Representative immunoblots of COX2, pre-IL-1β, and IL-6 from brainstem extracts, with VCL and GAPDH used as loading controls. (C) Densitometric quantification of pTAU (AT8) and pTAU (PHF1) protein levels from the immunoblots shown in (A), expressed as ratios to total TAU. Data are presented as mean ± S.D. (n = 6 animals per group). ∗p < 0.05; ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (D) Densitometric quantification of COX2, pre-IL-1β, and IL-6 protein levels from the immunoblots shown in (B), expressed as ratios to VCL or GAPDH. Data are presented as mean ± S.D. (n = 6 animals per group). ∗∗p < 0.01 vs. vehicle-treated mice according to Student's t-test. (E) Transcript levels of pro-inflammatory genes (Il1b, Il6, Ptgs2, Gfap, Iba1, Cd11b, F4/80, and Inos) determined by quantitative real-time PCR (qRT-PCR) in brainstem tissue and normalized to the geometric mean of Gapdh, Vcl, Tbp, and Actb. Data are presented as mean ± S.D. (n = 6 animals per group). ∗∗p < 0.01; ∗∗∗p < 0.001 vs. vehicle-treated mice according to Student's t-test. (F) Immunohistochemical staining of sagittal brain sections using anti-pTAU (AT8) antibody. (G) Quantification of pTAU (AT8) fluorescence intensity normalized to total TAU (SP70) from double immunofluorescence images. Data are presented as mean ± S.D. (n = 6 animals per group). ∗p < 0.05 vs. vehicle-treated mice according to Student's t-test. (H) Double immunofluorescence staining. Green: pTAU (AT8). Red: CALBINDIN-D28K. (I) Quantification of the ratio of CALBINDIN-D28K-positive cells relative to pTAU (AT8)-positive cells. Data are presented as mean ± S.D. (n = 6 animals per group). Differences among groups were assessed using Student's t-test. (J) Immunohistochemical staining using anti-GFAP antibody. Representative astrocyte morphologies are shown and classified as resting/basal (type A) or reactive (type B). (K) Immunohistochemical staining using anti-IBA1 antibody. Representative microglial morphologies are shown and classified as resting/basal (type A), activated non-phagocytic (type B), or phagocytic (type C) according to Sánchez-Guajardo et al. [48]. (L) Quantification of GFAP immunoreactivity density. Data are presented as mean ± S.D. (n = 6 animals per group). Differences among groups were assessed using Student's t-test. (M) Quantification of IBA1 immunoreactivity density. Data are presented as mean ± S.D. (n = 6 animals per group). Differences among groups were assessed using Student's t-test.Fig. 9

Collectively, our results demonstrate that HXN exerts beneficial therapeutic effects against tauopathy by modulating neurodegenerative and inflammatory processes in the neocortex and brainstem, whereas its efficacy appears to be less pronounced in the hippocampus.

HXN reduced the Aβ plaque burden in the AT-NRF2WT mouse model

3.8

Since the AT-NRF2^WT^ animal model employed in this study also overexpresses the APP^V717I^ amyloid transgenic human form, we next investigated whether HXN could influence the burden and size of APP/Aβ plaques. To this end, we performed immunohistochemistry using the 4G8 antibody in sections from the neocortex, hippocampus, and brainstem. We detected an increased density of APP/Aβ plaques in the vehicle-treated group compared with the HXN-treated group in the neocortex (Suppl. Fig. S9A and S9D) and hippocampal regions (CA1 and subiculum) (Suppl. Fig. S9B and S9E), accompanied by a significant increase to 50% in plaque number observed in the neocortex, but not in hippocampi. Conversely, in the brainstem, we did not detect APP/Aβ plaques [25]; instead, we observed a higher presence of intracellular vesicles and extracellular aggregates in neurons of vehicle-treated mice, which were markedly reduced in HXN-treated mice (Suppl. Fig. S9C and S9F). Together, these findings suggest that HXN may exert a broad protective effect, attenuating both TAU- and amyloid-associated pathology in this model.

HXN improves synaptic plasticity, memory, and motor performance

3.9

Having demonstrated that HXN reduces pathological TAU phosphorylation, attenuates neuroinflammatory responses, and modulates both central and peripheral biomarkers, we next investigated whether these molecular and cellular effects translated into functional benefits. To this end, we evaluated cognitive performance, motor coordination, and synaptic plasticity in 10-month-old AT-NRF2^WT^ mice subjected to the chronic HXN protocol (Fig. 10A). Before treatment initiation, animals were weighed and subjected to motor and novel object recognition (NOR) tests to establish baseline values, which confirmed the presence of early cognitive and motor impairments associated with tauopathy at this age (Fig. 10B and C). Throughout the treatment period, animals were monitored for body weight and behavioral performance.Fig. 10Chronic HXN administration improves cognitive performance, motor function, and synaptic plasticity in AT-NRF2^WT^ mice. Ten-month-old AT-NRF2^WT^ mice were treated with vehicle or HXN (100 mg/kg) by oral gavage once every two days for 6 weeks (n = 6 animals per group). (A) Experimental timeline showing treatment schedule and behavioral assessments. Motor performance (treadmill test) and novel object recognition (NOR) were evaluated before the first administration (baseline). The treadmill test was repeated during the third and fifth weeks, and NOR was assessed during the first, third, and fifth weeks of treatment. (B) Discrimination index in the NOR task (time spent exploring the novel object divided by total exploration time). Data are presented as mean ± S.D. (n = 6 animals per group). Statistical analysis was performed using two-way ANOVA followed by Bonferroni's post-hoc test. ∗p < 0.05 vs. vehicle-treated mice. (C) Treadmill performance score assessing motor coordination and gait parameters at the indicated time points. Data are presented as mean ± S.D. (n = 6 animals per group). Statistical analysis was performed using two-way ANOVA followed by Bonferroni's post-hoc test. ∗p < 0.05 vs. vehicle-treated mice; #p < 0.05 vs. baseline or genotype control, as indicated. (D) Body weight evolution of AT-NRF2^WT^ mice treated with vehicle or HXN throughout the experimental period is shown in (A). Data are presented as mean ± S.D. (n = 6 animals per group). (E) Long-term potentiation (LTP) was recorded in hippocampal slices from 10-month-old mice after 6 weeks of treatment. The plot represents the mean field excitatory postsynaptic potential (fEPSP) slope over time. Data are presented as mean ± S.D. (n = 6 animals per group). Statistical analysis was performed using two-way ANOVA followed by Bonferroni's post-hoc test. ∗∗∗p < 0.001 vs. vehicle-treated AT-NRF2^WT^ mice.Fig. 10

In the NOR test, vehicle-treated AT-NRF2^WT^ mice exhibited a reduced discrimination index compared with NRF2^WT^ controls, consistent with impaired recognition memory (Fig. 10B). Notably, HXN-treated AT-NRF2^WT^ mice showed a clear tendency toward improved performance, with a 2.0-2.5-fold increase in the discrimination index relative to vehicle-treated animals, indicating a partial restoration of memory function (Fig. 10B).

Motor performance was assessed using a treadmill running assay. At baseline, AT-NRF2^WT^ mice displayed impaired motor coordination compared with NRF2^WT^ controls (Fig. 10C). Chronic HXN administration resulted in a modest but consistent improvement in motor performance, which became apparent by the third week of treatment and was maintained through week five (Fig. 10C). These functional improvements were accompanied by a slight increase in body weight in HXN-treated animals compared with vehicle-treated controls (Fig. 10D), potentially reflecting preservation of feeding behavior, a parameter known to deteriorate with disease progression in this mouse model [24].

To further assess synaptic function, hippocampal long-term potentiation (LTP) was recorded in the dentate gyrus following perforant path stimulation. As expected, high-frequency stimulation induced robust LTP in NRF2^WT^ control mice, whereas this response was markedly reduced in vehicle-treated AT-NRF2^WT^ mice (Fig. 10E). Importantly, HXN-treated AT-NRF2^WT^ mice exhibited a significant recovery of LTP compared with vehicle-treated counterparts, demonstrating that chronic HXN administration restores hippocampal synaptic plasticity (Fig. 10E).

Collectively, these findings indicate that the reduction of pathological pTAU burden and neuroinflammatory markers elicited by HXN translates into measurable improvements in synaptic transmission, cognitive performance, and motor function. Thus, HXN not only modulates key molecular hallmarks of tauopathy but also confers functional benefits at the organismal level, supporting its potential as a disease-modifying therapeutic strategy.

Discussion

4

The present study identifies hexaraphane (HXN) as a potent modulator of pathological TAU phosphorylation in vitro and in vivo. While HXN has been extensively characterized as an NRF2 activator with antioxidant and anti-inflammatory properties [9,[11], [12], [13]], our findings reveal an additional and mechanistically distinct action on TAU homeostasis. Specifically, HXN reduces pathological TAU phosphorylation predominantly through a PP2A-dependent mechanism that occurs independently of NRF2 signaling.

NRF2 has been demonstrated to be a promising target for slowing the progression of neurodegenerative disorders, as it modulates the main hallmarks of these diseases, including proteinopathy, oxidative stress, and chronic inflammation. Indeed, several electrophilic drugs have been used for proof-of-concept studies indicating that activation of NRF2 may provide a benefit against Parkinson's disease [29,49] and amyloidopathy of Alzheimer's disease [50,51], but there is little evidence about the role of NRF2 in tauopathies [52]. Considering HXN as a great NRF2 inductor in the literature [12,53], we demonstrated that HXN is capable of regulating pTAU-specific epitopes involved in the onset and progression, such as AT8 and PHF1, in an NRF2-independent manner, either in primary neurons or our AD-like mouse model (AT-NRF2^WT^ and AT-NRF2^KO^). A similar mechanism has been described for DMF, which disrupts the KEAP1–NRF2 interaction; however, our findings further reveal that this compound also modulates the GSK-3/NRF2 signaling pathway, providing a double mechanism of activation of NRF2 that might be used to treat TAU-related neurodegeneration [54]. Previously published in our group, HXN was described as regulating GSK-3 activity [3], which led to the idea that HXN may prevent hyperphosphorylation of TAU through alternative mechanisms of the NRF2 pathway. Surprisingly, as shown in Fig. 2, HXN can still reduce pTAU levels despite overexpressing GSK-3, raising concern that HXN might be regulating pTAU levels through an alternative, and not yet described mechanism. Additionally, we confirmed that HXN did not alter PTEN/PI3K/AKT/GSK-3 signaling in primary neurons, like the previous report [6]. On the other hand, this study demonstrates robust reductions in pathological TAU phosphorylation and TAU-associated functional deficits, but does not directly assess insoluble TAU species or seeding activity. While our findings support a disease-modifying effect of HXN at the biochemical, cellular, and behavioral levels, future studies will be required to determine its impact on TAU aggregation and prion-like propagation.

So far, an increasing number of strategies have been aimed at inhibiting TAU hyperphosphorylation for the treatment of tauopathies, mainly through inhibition of TAU-protein kinases. However, there are other, less explored therapeutic options in TAU therapies, such as treatment with PPs activators. PPs involved in TAU dephosphorylation include PP1, PP2A, PP2B, PP2C, and PP5 [42]. Physiologically, PP2A accounts for the majority of TAU-related phosphatase activity (∼71%), compared to PP2B (∼7%), PP5 (∼11%), and other phosphatases (predominantly PP1, ∼11%) [42,55]. Thus, PP2A is the most active phosphatase for TAU in the brain: it accounts for around 70% of TAU phosphatase activity in the normal brain, but in AD brains, its activity decreases by about 20% in the gray matter and 40% in the white matter [42]. These proteins represent new avenues that are currently being explored as pharmacological targets. Such is the case of a specific PP2A activator called sodium selenate, which has shown reasonable effectiveness in reducing TAU phosphorylation in animal models [56,57]; although, only moderate benefits have been observed in clinical trials [58]. In this study, we propose an additional PP activator to prevent TAU hyperphosphorylation, such as HXN. Interestingly, HXN has been capable of regulating PP2A activity, as demonstrated by direct PP2A assays in vitro, as well as by the classical PP inhibitor okadaic acid, which blocks HXN's ability to reduce pTAU and pathological species. Although HXN appears capable of increasing PP2A activity, we were not able to conclude whether this effect depends directly on PP2A regulation of pTAU levels. Although okadaic acid was used to functionally implicate PP2A in HXN-mediated TAU dephosphorylation, its limited selectivity at higher concentrations precludes exclusive attribution of the observed effects to a single phosphatase. Nevertheless, the ability of HXN to increase PP2Ac activity in cell-free assays, together with the complete loss of TAU dephosphorylation upon phosphatase inhibition, supports a central role for PP2A, while not excluding contributions from additional serine/threonine phosphatases. Employing shRNA technology might have been one approach, but we have never achieved sufficiently decreased PP2A levels to analyze this. Conversely, we believe that HXN likely modulates various PPs, including PP1, PP2A, PP3, PP4, and PP5, as shown in the experiments conducted with dose-ranging of okadaic acid in combination with HXN, demonstrating its capacity to decrease pTAU from the IC_50_ concentration of okadaic acid, around PP2A and PP4 inhibition, up to higher concentrations (15 nM−45 nM), approaching PP1 inhibition levels (Suppl. Table 3). These findings demonstrate that high concentrations of okadaic acid completely abolish the activity of HXN, likely because okadaic acid broadly inhibits multiple serine/threonine phosphatases across a wide concentration range. In summary, HXN can increase PP2A activity, either in vitro or in vivo assays, but we cannot confirm that HXN regulates hyperphosphorylated TAU solely through PP2A-dependent mechanisms.

Although this study does not answer the follow-up question of how HXN regulates PP2A activity, we can speculate that HXN may work similarly to other well-characterized PP2A activator reporters in the literature. PP2A activation can be controlled by different direct and indirect mechanisms. The direct PP2A activators will bind to the holoenzyme to induce activation, working as protein-protein interaction glue molecules. For instance, a series of molecules known as small molecule activators of PP2A (SMAPs), increasing PP2A activity through boosting the binding of PP2A with the A-scafold subunit to induce the conformational change that is needed for its activation, or by preventing the destabilization of the PP2A-B56α holoenzyme [[59], [60], [61]]. In addition to SMAP, metformin has been shown to result in the PP2A up-regulation, leading to anti-neoplastic activity in lung, prostate, and breast cancers [62,63]. These effects have also been attributed to the interference with the association of the catalytic subunit of PP2Ac to the so-called MID1-α4 protein complex, which regulates the degradation of PP2Ac and thereby influences PP2A activity [62]. Furthermore, sodium selenate can activate PP2A by upregulating its regulatory subunit PP2A/PR55, promoting beneficial effects in desphosphorylation of TAU in AD and in other tauopathies [64], and improving mild traumatic brain injury outcomes [65]. On the other hand, the indirect activation of PP2A relies on the inhibition of the endogenous inhibitors of PP2A, such as SET and CIP2A. SET and CIP2A, which are highly expressed and assist in the reduction of PP2A activity [61]. Thus, these molecules can inhibit these endogenous inhibitors by preventing their binding to PP2A. Probably, HXN may have some of this skill function, since it possesses electrophile properties and a hydrophobic carbon chain capable of interacting with many other macromolecules. Our data demonstrate that HXN functionally activates PP2A and that PP2A activity is required for HXN-mediated TAU dephosphorylation, but the precise molecular mechanism by which HXN modulates the PP2A holoenzyme, either through direct interaction with the catalytic or regulatory subunits, or indirectly via modulation of endogenous PP2A inhibitors such as SET or CIP2A, remains to be elucidated.

To get a preliminary insight into the mechanistic regulation of PP2A by HXN, we have performed molecular docking analyses using available crystal structures of the PP2A holoenzyme (PDB: 2IAE and 6NTS) (data not shown). Using the small-molecule activator DT-061 as a structural reference, our in silico modeling suggests that HXN may bind in proximity to the reported SMAP interaction site located at the interface between the scaffold (A) and the regulatory (B) subunits. Although these findings provide a potential structural rationale for PP2A modulation, they remain exploratory and do not establish whether HXN functions as a bona fide PP2A holoenzyme stabilizer. Further work will be required to elucidate the precise molecular mechanisms underlying PP2A modulation, including potential interactions with PP2A subunits, post-translational modifications, or redox-dependent regulatory pathways. Extended pharmacokinetic characterization and long-term safety studies will also be essential to support clinical translation.

In addition to NRF2-dependent transcriptional responses, HXN may also exert direct redox-modulating effects due to its electrophilic properties [6]. Such direct interactions could influence intracellular thiol redox balance and modify redox-sensitive signaling molecules, including kinases and phosphatases involved in TAU phosphorylation [6]. Redox-dependent modulation of inflammatory pathways may likewise contribute to the observed attenuation of neuroinflammatory markers. Although our data support PP2A activation as a key mechanism underlying TAU dephosphorylation, we cannot exclude the possibility that direct redox effects of HXN synergize with phosphatase activation to shape the overall molecular and cellular response. Future studies specifically dissecting redox-sensitive signaling events will help clarify the relative contribution of these mechanisms.

Several therapeutic strategies have been proposed to restore PP2A activity in tauopathies, including enhancement of selective PP2A activity, inhibition of endogenous PP2A inhibitors such as SET, and the development of small-molecule PP2A activators. For instance, pharmacological promotion of PP2A activation using sodium selenate has been shown to reduce pathological TAU phosphorylation and ameliorate behavioral deficits in tauopathy models [64,66]. Similarly, targeting endogenous PP2A inhibitors such as SET or CIP2A has been explored as a means to restore phosphatase activity and limit TAU pathology [44,67]. On the other hand, the development of small-molecule PP2A activators for neurological diseases remains limited, as most efforts to date have been primarily focused on cancer treatment [60,68,69]. While these approaches share the goal of reducing pathological TAU phosphorylation, they differ substantially in their mechanisms of action, pharmacokinetic properties, and safety considerations. In this context, HXN represents a distinct strategy that promotes PP2A activity at low concentrations, achieves limited and transient brain exposure, and reduces pathological TAU phosphorylation without requiring sustained or global manipulation of the PP2A holoenzyme.

We observed that HXN is not uniformly distributed across different brain regions. Pharmacokinetic profiling revealed that HXN was readily detected in the neocortex and brainstem, whereas its concentration in the hippocampus was markedly lower (Suppl. Fig. 8). This regional difference may reflect variations in tissue accessibility, vascularization, or drug distribution within distinct brain structures, potentially limiting hippocampal exposure to the pharmacological treatment. The inverted U-shaped dose–response observed in vitro suggests that HXN activates PP2A within a defined concentration window, whereas higher concentrations may become suboptimal or inhibitory (Fig. 3E). Pharmacokinetic analyses indicate that brain concentrations might be achieved following oral administration and remain within a low micromolar or submicromolar range, which correlates with PP2A activation, reduced TAU phosphorylation, and neuroprotection in vivo. Consequently, HXN elicited only a modest reduction in pro-inflammatory markers in the hippocampus compared with the neocortex and brainstem (Fig. 7, Fig. 9). Despite this, mice treated with HXN (100 mg/kg) for five weeks exhibited improvements in memory performance and electrophysiological function, indicating that even the low hippocampal concentration achieved by HXN is sufficient to confer protection against tauopathy-related impairments.

Experimental evidence suggests that HXN plays a prominent role in targeting brain proteinopathies, improving synapses, and reducing inflammation, mainly through NRF2 activation [9,22,53]. Oxidative stress is known to promote TAU hyperphosphorylation, protein aggregation, and neuroinflammation through multiple mechanisms, including activation of stress kinases and impairment of phosphatase activity [70]. Therefore, it is plausible that the antioxidant properties of HXN indirectly contribute to its overall neuroprotective profile. For instance, it had already been described that HXN prevents cognitive impairment in the App knock-in AD model (App^NLGF^) mice by citoprotective mechanisms and anti-inflammatory properties by inducing NRF2 [13]. We further corroborate this issue, since HXN treatment also reduces either APP/Aβ plaque size or its burden in the brain in our AD-like mice model.

However, our data indicate that the reduction of pathological TAU phosphorylation induced by HXN occurs independently of its antioxidant/NRF2 activity. Specifically, HXN reduced TAU phosphorylation at AT8 and PHF1 epitopes in both AT-NRF2^WT^ and AT-NRF2^KO^ neurons and mice, demonstrating that NRF2-mediated antioxidant responses are not required for TAU dephosphorylation. Moreover, HXN did not significantly modulate major kinase pathways commonly linked to oxidative stress–driven TAU phosphorylation, such as GSK-3β or MAPKs, further supporting a mechanism distinct from classical redox signaling.

In contrast, the anti-inflammatory effects observed following HXN treatment are likely to be partially mediated by its antioxidant and NRF2-activating properties. Chronic administration of HXN reduced astrogliosis, microgliosis, and pro-inflammatory cytokine expression in vivo, effects that are consistent with NRF2-dependent suppression of neuroinflammatory pathways. Thus, while HXN-induced TAU dephosphorylation is primarily driven by phosphatase activation, particularly PP2A, the compound's antioxidant activity likely contributes to limiting neuroinflammation and promoting neuronal resilience during disease progression.

In correlation, we also detected a improvements in motor performance, dynamic movement, and memory-related behaviors in the brains of 10-month-old AT-NRF2^WT^ mice treated with HXN, likely mediated by a combination of mechanisms involving NRF2 upregulation, phosphatases activation, improved amyloidopathy, and reduction in pathological pTAU species. The cognitive and motor improvements observed following HXN treatment are modest in magnitude but highly consistent across behavioral paradigms, and closely parallel the degree of pathological TAU phosphorylation reduction. Notably, prior studies have demonstrated that partial attenuation of pTAU levels is sufficient to elicit measurable functional benefits in tauopathy models, without requiring complete normalization of TAU pathology [71,72]. In this context, the effect sizes and variability observed here are consistent with those reported for other TAU-targeting disease-modifying strategies.

A limitation of the present study is the absence of long-term behavioral assessments in NRF2-deficient mice. While our data indicate that TAU dephosphorylation and short-term functional improvements occur independently of NRF2, future studies will be required to determine whether sustained behavioral benefits during chronic treatment rely, at least in part, on NRF2-mediated neuroprotective mechanisms.

Additionally, the pTAU levels detected in the brain upon treatment with HXN by oral gavage were also found in blood, since additional analysis discloses a significant decrease of TAU-pThr^217^ in mouse serum, by employing Simoa technology. In this way, HXN is the first molecule in the literature to have been published in correlation with improvement against tauopathies in mouse models, either by decreasing pTAU levels in the brain and blood, thereby conferring significant preclinical value for future translational clinical trials in humans.

Conclusion

5

HXN emerges as a multifunctional compound capable of modulating key pathological features of tauopathies. By promoting PP2A-dependent TAU dephosphorylation independently of NRF2, while simultaneously exerting NRF2-mediated anti-inflammatory and cytoprotective effects, HXN may rebalance kinase–phosphatase dynamics and restore neuronal homeostasis. These findings support further investigation of HXN as a disease-modifying therapeutic candidate for AD and related tauopathies.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

AI-Technologies

AI was only used to improve the clarity and quality of the English language during manuscript preparation. All content was then reviewed and revised by the authors.

Funding

This study was funded by KINJIRUSHI Co., Ltd., and the Spanish National Research Agency (PID2021-122650OB-I00, PID2022-141786OB-I00, PDC2021-121421-I00, PDC2022-133765-I00, PID2022-140236OB-I00, CPP2021-008389), CIBERNED/ISCIII (CB06/05/0010), the Autonomous Community of Madrid (P2022_BMD-7230), and the Basque Government (grant IT1551- 22) and BIOEF (Convocatoria de Ayudas a la Investigación en Alzheimer de la Fundación Vasca de Innovación e Investigación Sanitarias, BIO22/ALZ/014).

CRediT authorship contribution statement

Ángel Juan García-Yagüe: Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Validation, Visualization, Writing – original draft, Writing – review & editing. Daniel Carnicero-Senabre: Formal analysis, Investigation, Validation, Visualization, Writing – review & editing. Ángel Núñez: Formal analysis, Investigation, Validation, Visualization, Writing – review & editing. Raffaela Cipriani: Formal analysis, Investigation, Validation, Visualization, Writing – review & editing. Estibaliz Capetillo-Zarate: Formal analysis, Investigation, Validation, Visualization, Writing – review & editing. Maribel Escoll: Formal analysis, Investigation, Validation, Visualization, Writing – review & editing. Isao Okunishi: Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing. Ana I. Rojo: Formal analysis, Funding acquisition, Investigation, Project administration, Resources, Supervision, Validation, Visualization, Writing – review & editing. Antonio Cuadrado: Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing.

Declaration of competing interest

This study was supported by KINJIRUSHI Co., Ltd. The funding body had no role in the design of the study, collection, analyses, or interpretation of the data, writing of the manuscript, or the decision to publish the results.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this manuscript.