PPAR-γ activation by pioglitazone inhibits microglia-mediated neuroinflammation and alleviates disease in neuromyelitis optica spectrum disorders Authors: Leung Wah Yick, Ethel Yin Ying Chan, Wenying Zou, Krystal Xiwing Yau, Jason Shing Cheong Kwan, Chi Yan Lee, Koon Ho Chan Published: 2025 ABSTRACT Microglia contribute to antibody-dependent cell-mediated cytotoxicity (ADCC) in neuromyelitis optica spectrum disorders (NMOSD). This complement-independent process has been implicated in early neuroinflammation following aquaporin-4 immunoglobulin G autoantibodies (AQP4-IgG) binding to astrocytes. We investigated whether pioglitazone, a clinically approved peroxisome proliferator-activated receptor gamma (PPARγ) agonist, could be repurposed to reduce complement-independent neuroinflammation driven by activated microglia and ameliorate disease in mice receiving passive transfers of human AQP4-IgG. Pioglitazone treatment alleviated motor impairments and NMOSD-like pathologies, associating with reduced microglial activation and pro-inflammatory cytokine levels. Antagonizing PPARγ reversed the beneficial effects of pioglitazone, indicating that these effects were PPARγ-dependent. Microglial depletion attenuated AQP4-IgG-induced motor impairments and pathologies, supporting the detrimental role of microglial activation. Serum levels of soluble triggering receptor expressed on myeloid cells 2 (sTREM2), a microglial activation biomarker, were increased in both the mouse model and patients with NMOSD. Pioglitazone reduced sTREM2, which were correlated with attenuated AQP4-IgG-induced microglial activation, neuroinflammation, and pathologies in mice. Elevated sTREM2 levels were associated with cytokines and chemokines related to activated microglia in patients with NMOSD, suggesting the clinical benefits of pioglitazone. Our findings highlight the potential of repurposing PPAR-γ agonists for reducing microglia-mediated neuroinflammation in acute NMOSD attacks.Supplementary InformationThe online version contains supplementary material available at 10.1186/s40478-025-02182-x. Introduction Neuromyelitis optica spectrum disorders (NMOSD) are autoimmune inflammatory disorders of the central nervous system (CNS) with secondary demyelination. They are characterized by recurrent attacks of acute neuroinflammation, which commonly presents as optic neuritis, transverse myelitis, and area postrema syndrome [19]. Most patients with NMOSD are seropositive for pathogenic immunoglobulin G (IgG) autoantibodies against aquaporin-4 (AQP4), known as AQP4-IgG [21, 25]. In the CNS, the AQP4 water channel is predominantly localized at the endfeet of astrocytes. AQP4-IgG binding to astrocytic AQP4 causes astrocytopathy, leading to neuroinflammation, blood–brain barrier (BBB) disruption, and subsequent demyelination and neuronal injury [39]. Multiple types of CNS lesions have been observed in patients with NMOSD and in animal models of the disease, involving both complement-dependent and -independent mechanisms [17, 33]. Extensive studies have been conducted to investigate the role of complement-dependent cytotoxicity (CDC) in NMOSD pathophysiology [11]. Noteworthily, C1q, the initiator of the complement cascade, is absent in the homeostatic CNS. Complement-independent mechanisms, including antibody-dependent cell-mediated cytotoxicity (ADCC), likely contribute to the early stage of lesion development after AQP4-IgG binding to astrocytic AQP4 [18, 54, 55]. However, less is known about whether modulation of ADCC, which is primarily driven by microglia, can alleviate complement-independent neuroinflammation and lesion development in NMOSD. Microglia, which are derived from hematopoietic myeloid progenitor cells, are resident phagocytes of the CNS [41]. They monitor brain activity through their ramified processes to maintain normal brain function and act as primary cells responding to pathogen infection and injury [27, 46, 49]. Prominent microglial activation has been observed in active lesions in patients with NMOSD [33, 40], as well as in various experimental animal models of the disease [2, 3, 14, 54, 55]. The levels of soluble triggering receptor expressed on myeloid cells 2 (sTREM2), a potential biomarker of microglia-mediated neuroinflammation, are increased in patients with NMOSD [42]. A recent study has revealed that the evolving NMOSD pathophysiology relies on an C3a receptor-mediated interaction between astrocytes and microglia [7]. These findings indicate that microglia play a pivotal role in the inflammatory response initiated when AQP4-IgG bind to astrocytes. Therefore, microglia are a potential therapeutic target for ameliorating neuroinflammation and disease severity in NMOSD. Peroxisome proliferator activated receptor-γ (PPAR-γ) is a ligand-binding transcription factor that belongs to the nuclear receptor superfamily [12]. Apart from its role in regulating carbohydrate and lipid metabolism, PPAR-γ inhibits the polarization of microglia into a pro-inflammatory phenotype [4, 24, 45]. Pioglitazone, a highly selective PPAR-γ agonist, is a clinically approved drug for the treatment of type 2 diabetes. In addition to its efficacy on metabolic diseases, pioglitazone has been shown to ameliorate neuroinflammation and disease severity in different experimental models of CNS disorders [16, 24, 28, 37, 38]. However, the potential efficacy of pioglitazone in alleviating disease and reducing neuroinflammation in NMOSD remains unexplored. We have shown that repetitive passive transfers of high-dose IgGs from AQP4-IgG-seropositive NMOSD patients induce complement-independent CNS pathologies in mice, including astrocytopathy, microglial activation, demyelination, and axonal loss [54]. In the current study, we used this mouse model to investigate whether pioglitazone treatment alleviates NMOSD-like motor impairments and pathologies. We then examined if the efficacy of pioglitazone was dependent on PPAR-γ and linked to the reduction of microglia-mediated neuroinflammation. Next, we assessed whether pioglitazone normalized serum sTREM2 levels and their correlations with microglial activation, neuroinflammation, and pathologies in the model. Finally, using serum collected from patients with NMOSD, we investigated the association between sTREM2 levels and markers of microglia-mediated neuroinflammation. Our findings indicate that PPAR-γ agonists, such as pioglitazone, hold promise for repurposing for clinical benefits in acute NMOSD attacks. Methods Human subjects Blood samples were collected from six healthy controls (HC) and eight patients with NMOSD recruited at the Department of Medicine, Queen Mary Hospital, Hong Kong. None of the HC had any history of disease or infection, nor had they received any treatment in the past 2 months. The inclusion criteria for patients with NMOSD were an age of ≥ 18 years, a diagnosis of AQP4-IgG-seropositive NMOSD according to the 2015 International Panel for NMO Diagnosis (IPND) consensus criteria [51], and regardless of the disease course or the disease-modifying treatment. The exclusion criteria were another immune disorder, infection, or cancer. AQP4-IgG serostatus was assessed using cell-based indirect immunofluorescence assay in our hospital as reported previously [5]. Age, gender, disease duration, current treatment, and relapse/remission status were recorded. Degree of disability was assessed by neurological examination with Expanded Disability Status Scale (EDSS) score. Written informed consent was obtained from all participants. All procedures were approved by the Institutional Review Board of The University of Hong Kong/Hospital Authority Hong Kong West Cluster (HKU/HA HKW IRB, UW 09-289). Animals Female wild-type C57BL/6 mice aged 6 to 8 weeks were used in this study. Mice were housed in groups of five per cage at the animal facilities of the Centre for Comparative Medicine Research at The University of Hong Kong. They were maintained under a 12-h dark/light cycle and provided ad libitum access to water and chow. The "Principles of laboratory animal care" (NIH publication No. 86-23, revised 1985) were followed. All experimental procedures were approved by the Committee on the Use of Live Animals in Teaching and Research (CULATR No: 5738-21) of The University of Hong Kong. Human IgG purification IgGs from AQP4-IgG-seropositive NMOSD patients and HCs were isolated from serum or plasma samples using HiTrap Protein G Sepharose columns (GE Healthcare Biosciences). Then they were purified with Slide-A-Lyzer Dialysis Cassettes (Thermo Scientific) and concentrated with Amicon Ultra-15 centrifugal filters (Merck Millipore). Protein concentration was determined using Bradford assay (Bio-Rad). Pooled IgGs isolated from HC were referred to as IgG(ctrl), and pooled IgGs isolated from patients with AQP4-IgG-seropositive NMOSD were designated as IgG(AQP4). BBB disruption Mice were anaesthetized with an intraperitoneal (i.p.) injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). At day 7 prior to IgG transfer, mice received subcutaneous injections of complete Freund’s adjuvant (CFA, BD Biosciences) containing heat-killed H37Ra Mycobacterium tuberculosis (Difco) at four sites (50 µg in 50 µl CFA per site) on the hind flank. Additionally, at days 7 and 3 before IgG transfer, they received i.p. injections of pertussis toxin (PTx, 200 ng in 200 µl PBS, List Biological Laboratories). Passive transfer of human IgGs and drug treatments After the disruption of BBB, mice were administered daily i.p. injections of either IgG(ctrl) or IgG(AQP4) from day 0 to 7, as previously described [54]. Each injection contained 4 mg of IgGs in 500 ul of PBS, resulting in a total of 32 mg of IgGs administered per mouse. Pioglitazone (Sigma-Aldrich) was mixed with corn oil (Sigma-Aldrich) and sonicated at room temperature for 20 min to produce a homogeneous mixture. GW9662 (Sigma-Aldrich), a selective PPAR-γ antagonist, was dissolved in dimethyl sulfoxide (DMSO). Both pioglitazone and GW9662 solutions were freshly prepared before each administration. Mice were randomly assigned to two sets of experiments. For preventive treatment, IgG(AQP4) mice received oral gavage of either vehicle, 20 mg/kg, or 80 mg/kg body weight pioglitazone daily from day 0 to 7. IgG(ctrl) mice that received vehicle were used as a sham control. For therapeutic treatment, mice were randomly divided into five groups: 1) IgG(ctrl) + vehicle, 2) IgG(AQP4) + vehicle, 3) IgG(AQP4) + 80 mg/kg/day pioglitazone, 4) IgG(AQP4) + 3 mg/kg/day GW9662, and 5) IgG(AQP4) + 80 mg/kg/day pioglitazone + 3 mg/kg/day GW9662. Treatments with vehicle, pioglitazone, and GW9662 were initiated at day 4 and continued once daily until day 7. Animals were euthanized at day 8. Depletion of microglia/macrophages Clodronate liposomes were used to deplete microglia/macrophages as previously described [3]. Briefly, 200 ul of PBS or clodronate liposomes (Liposoma) was injected (i.p.) into IgG(AQP4) mice at 2 days before, 2 days after, and 6 days after the initiation of IgG passive transfer at day 0. Mice were euthanized at day 8. In vivo imaging Myelin in mice was labelled using the near-infrared fluorescence dye 3,3-diethylthiatricarbocyanine iodide (DBT) as previously described [50]. In vivo imaging of the cervical spinal cord was performed using the IVIS Spectrum imaging system (Perkin Elmer). Mice were anaesthetized with an i.p. injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). To minimize the effects of autofluorescence and light scattering, all mice underwent shaving before imaging. Each mouse was injected intravenously through the tail vein with 0.3 mg/kg DBT in 100 ml of PBS with 1% DMSO. Images were captured at 0, 2, 4 and 6 min after injection using the following settings: Epi-Fl illumination, excitation at 745 nm and emission at 800 nm, exposure time of 2 s, focus on B 6.5 cm, binning at 8, and f-stop at 2. Near-infrared signals were analysed using LivingImage software (Perkin Elmer). Signal from each region of interest was normalized by subtracting the average background signal. Beam walking test The motor function of the mice was examined using beam walking test [31, 55]. The test assessed the mice’s ability to maintain an upright position and walk across an elevated narrow beam towards a platform. The apparatus consisted of two beams, each 80 cm long, with widths of 1.2 cm and 0.6 cm, respectively. These beams were positioned 50 cm above a tabletop and were supported by two stands. During the test, the time taken by the mouse to cross each beam and the number of paw slips it made while doing so were recorded. Serum collection and tissue harvesting At the experimental endpoint, mice were euthanized with an excessive dose of ketamine and xylazine. Blood samples were harvested from the facial vein of the mice with animal lancets (Medipoint). Sera were obtained by collecting the supernatants after centrifuging the clotted blood samples at 6000 rpm for 15 min at 4 °C. In some animals, the cervical spinal cords were freshly dissected out, immediately frozen in liquid nitrogen, and stored at -80 °C. In other animals, transcardial perfusion was performed using ice-cold PBS, followed by 4% paraformaldehyde. Cervical spinal cords and optic nerves were then harvested, post-fixed with 4% paraformaldehyde, cryoprotected with 30% sucrose, and cut into 10 µm sections using a cryostat. Immunofluorescence Cryosections of spinal cords and optic nerves were incubated with the following primary antibodies at 4 °C overnight: (1) rabbit anti-AQP4 (1:200, catalog A5971, Sigma-Aldrich), (2) mouse anti-glial fibrillary acidic protein (GFAP, 1:200, catalog SC33673, Santa Cruz Biotechnology), (3) goat anti-myelin basic protein (MBP, 1:200, catalog SC13914, Santa Cruz Biotechnology), and (4) rabbit anti-neurofilament heavy polypeptide (NF-H, 1:400, catalog N4142, Sigma-Aldrich), (5) rabbit anti-ionized calcium-binding adapter molecule 1 (Iba-1, 1:200, catalog 019-19741, Wako), (6) mouse anti-cluster of differentiation 68 (CD68, 1:200, catalog YM3161, Immunoway), (7) mouse anti-PPAR-γ (1:200, catalog A3409A, Invitrogen). Sections were then incubated with the appropriate Alexa-Fluor-conjugated secondary antibodies (Thermo Fisher Scientific) at room temperature for 1 h. They were coverslipped using an antifade reagent containing DAPI (Thermo Fisher Scientific). Image processing and quantification Measurement of immunofluorescence intensity was performed on eight alternate cross sections of the ventrolateral white matter of the cervical spinal cord, taken from rostral to caudal in each mice. All images were captured using the same microscope (Eclipse Ni, Nikon). Fluorescence signal intensity was quantified using ImageJ (Version 2.9.0, NIH). For the morphometric analysis of Iba-1+ microglia, z-stacked images with 3-dimensional reconstruction were acquired using the ImageXpress Pico imaging system (Molecular Devices). The analysis of microglial branch number and length was performed using the skeletonize plugin within ImageJ. ELISA The concentrations of interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumour necrosis factor-α (TNF-α) in the spinal cord homogenates of mice were determined using commercially available mouse ELISA kits (IL-1β, catalog ELM-IL-1b; IL-6, catalog ELM-IL6; TNF-α, catalog ELM-TNFa; RayBiotech) following the manufacturer’s protocols. The concentrations of sTREM2 in mice sera samples were measured using TREM2 ELISA kit (ELM-TREM2; RayBiotech). The concentrations of sTREM2, IL-1β, IL-6, TNF-α, CXCL7 in human sera samples were measured using commercially available human ELISA kits (TREM2, catalog ELH-TREM2; IL-1β, catalog ELH-IL-1b; IL-6, catalog ELH-IL6; TNF-α, catalog ELH-TNFa; CXCL7, catalog ELH-NAP2; RayBiotech). Briefly, standards and samples were introduced into 96-well microplates precoated with primary antibodies. Following incubation, the wells were washed to remove unbound proteins. Biotinylated detection antibodies were then added. After a subsequent wash step, HRP-conjugated streptavidin was applied to the wells. The final wash was performed prior to adding the TMB substrate for colorimetric detection. The optical densities of the wells were measured at 450 nm using microplate reader (CLARIOstar, BMG Labtech). The concentrations (pg/ml) of the target proteins were normalized with the total protein content of the samples (pg/mg protein). Statistical analysis All statistical analyses were performed using Prism (Version 10, GraphPad Software Inc). The difference between two groups was determined using two-tailed Student’s t test. Differences between groups were compared uisng one-way ANOVA with post hoc Tukey’s test. Associations between variables were analyzed using Spearman’s two-tailed correlation analysis. P values of less than 0.05 were considered significant. Results Delayed oral pioglitazone treatment ameliorates motor impairments in IgG(AQP4) mice Previously, we have demonstrated that mice exhibit motor impairments and NMOSD-like pathologies after repetitive passive transfers of high-dose human IgGs from AQP4-IgG-seropositive NMOSD patients, along with BBB disruption induced by pre-treatment of CFA and PTx [54, 55]. Here, we used this mouse model of NMOSD to determine the optimal oral pioglitazone dosage for preventing of motor impairments. IgG(AQP4) mice were treated with 20 or 80 mg/kg/day of pioglitazone or vehicle from day 0 to 7 (Fig. 1a). At day 8, beam walking test was performed to assess the time it took for the mouse to cross a beam and the number of paw slips during the process. On a 1.2 × 80 cm (width × length) beam, treatment with pioglitazone at 80 mg/kg/day, but not at 20 mg/kg/day, prevented the onset of motor impairments (Fig. 1b and c). Similar findings were observed in walking test using a narrower 0.6 × 80 cm beam that allowed for the detection of more subtle motor impairments (Fig. 1d, e). Therefore, oral pioglitazone at a dose of 80 mg/kg/day was used in subsequent experiments.Fig. 1Delayed oral pioglitazone treatment ameliorates human AQP4-IgG-induced motor impairments in mice in a PPAR-γ-dependent manner. a Experimental design for evaluating the dose–response effects of pioglitazone treatment on human AQP4-IgG-induced motor impairments in mice. b and c Beam walking test measuring the time taken (b) and the number of paw slips (c) while the mouse walked across a 1.2 × 80 cm (width × length) beam. d and e Time taken to cross (d) and number of paw slips (e) while crossing a 0.6 × 80 cm beam. n = 6–8 mice per group. f Experimental design for evaluating the effects of delayed pioglitazone treatment on human AQP4-IgG-induced motor impairments in mice. g and h Beam walking test measuring the time taken (g) and the number of paw slips (h) while the mouse walked across a 1.2 × 80 cm (width × length) beam. i and j Time taken to cross (i) and number of paw slips (j) while crossing a 0.6 × 80 cm beam. n = 8 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant, *P < 0.05, **P < 0.01, and ***P < 0.001 Delayed oral pioglitazone treatment ameliorates human AQP4-IgG-induced motor impairments in mice in a PPAR-γ-dependent manner. a Experimental design for evaluating the dose–response effects of pioglitazone treatment on human AQP4-IgG-induced motor impairments in mice. b and c Beam walking test measuring the time taken (b) and the number of paw slips (c) while the mouse walked across a 1.2 × 80 cm (width × length) beam. d and e Time taken to cross (d) and number of paw slips (e) while crossing a 0.6 × 80 cm beam. n = 6–8 mice per group. f Experimental design for evaluating the effects of delayed pioglitazone treatment on human AQP4-IgG-induced motor impairments in mice. g and h Beam walking test measuring the time taken (g) and the number of paw slips (h) while the mouse walked across a 1.2 × 80 cm (width × length) beam. i and j Time taken to cross (i) and number of paw slips (j) while crossing a 0.6 × 80 cm beam. n = 8 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant, *P < 0.05, **P < 0.01, and ***P < 0.001 Therapeutic treatment with oral pioglitazone at a dose of 80 mg/kg/day was started at day 4 after the initial passive transfer of IgG(AQP4) (Fig. 1f). The treatment significantly reduced both the time for the mice to cross the 1.2 × 80 cm beam and the number of paw slips during the process (Fig. 1g and h). Similar findings were observed in walking test using a 0.6 × 80 cm beam (Fig. 1i and j). The administration of the PPAR-γ antagonist GW9662 alone did not reduce the time taken by the IgG(AQP4) mice to cross the beams, nor decrease the number of paw slips. However, the effects of pioglitazone were counteracted by the concurrent administration of GW9662 (Fig. 1g–j). These findings indicate that pioglitazone treatment prevents the development of motor impairments caused by human AQP4-IgG in mice, and this effect is dependent on PPAR-γ. Pioglitazone treatment attenuates astrocytopathy in a PPAR-γ-dependent manner To investigate the efficacy of pioglitazone treatment on mitigating astrocytopathy induced by IgG(AQP4), we performed immunofluorescence analysis of AQP4 and GFAP in the spinal cord and optic nerves. The immunofluorescence signals of AQP4 and GFAP in the spinal cords of vehicle-treated IgG(AQP4) mice were lower than that of sham control. Therapeutic treatment with oral pioglitazone significantly reduced the loss of AQP4 and GFAP immunofluorescence signals compared to vehicle-treated IgG(AQP4) mice, while the treatment effect was reversed by GW9662 (Fig. 2a). The evaluation of immunofluorescence intensity confirmed that the levels of AQP4 and GFAP were significantly increased in the pioglitazone-treated IgG(AQP4) mice compared to vehicle, GW9662 only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice (Fig. 2b and c). Similar findings were observed in the optic nerves (Supplemental Fig. 1).Fig. 2Pioglitazone treatment attenuates human AQP4-IgG-induced astrocytopathy through PPAR-γ. a Co-immunostaining for AQP4 and GFAP in the spinal cord of sham, vehicle, pioglitazone, GW9662 only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice. Nuclei were counterstained with DAPI. Images are representative photomicrographs showing cross sections of the ventrolateral white matter of the cervical spinal cord. Scale bar: 50 μm. b and c Quantification of AQP4 (b) and GFAP (c) immunofluorescence intensities. n = 5 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant and ***P < 0.001 Pioglitazone treatment attenuates human AQP4-IgG-induced astrocytopathy through PPAR-γ. a Co-immunostaining for AQP4 and GFAP in the spinal cord of sham, vehicle, pioglitazone, GW9662 only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice. Nuclei were counterstained with DAPI. Images are representative photomicrographs showing cross sections of the ventrolateral white matter of the cervical spinal cord. Scale bar: 50 μm. b and c Quantification of AQP4 (b) and GFAP (c) immunofluorescence intensities. n = 5 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant and ***P < 0.001 Pioglitazone treatment alleviates demyelination and axonal loss in a PPAR-γ-dependent manner To investigate the effects of pioglitazone treatment on the progression of demyelination in IgG(AQP4) mice, a longitudinal study employing in vivo imaging of the cervical spinal cord was conducted. The imaging utilized a near-infrared fluorescent dye DBT, which has a strong affinity for binding to myelin [50]. No differences in DBT fluorescence intensity were observed between different groups at both day 0 and day 3. Starting from day 5, a substantial decrease in the fluorescence intensity was observed in the spinal cord of IgG(AQP4) mice treated with vehicle, GW9662 only, or a combination of pioglitazone and GW9662, compared to both the sham control and mice treated with pioglitazone alone. There was no notable difference in the fluorescence intensity between the IgG(AQP4) mice treated with pioglitazone and the sham control group. This observation persisted until day 7 (Fig. 3a and b).Fig. 3Pioglitazone treatment mitigates human AQP4-IgG-induced demyelination and axonal loss through PPAR-γ. a Representative in vivo near-infrared fluorescence images showing the myelin-binding dye DBT in the cervical spinal cord of sham, vehicle, pioglitazone, GW9662 only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice at day 7. Blue rectangles depict the region of interest. b The time courses of the average radiant efficiency of DBT near-infrared fluorescence signals in different groups of mice at days 0, 3, 5, and 7. c Immunostaining for MBP and NF-H in the spinal cord of different groups of mice. Nuclei were counterstained with DAPI. Images are representative photomicrographs showing cross sections of the ventrolateral white matter of the cervical spinal cord. Scale bar: 50 μm. d and e Quantification of MBP immunofluorescence intensity (d) and the number of NF-H spots (e). n = 5 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant, *P < 0.05, **P < 0.01, and ***P < 0.001 Pioglitazone treatment mitigates human AQP4-IgG-induced demyelination and axonal loss through PPAR-γ. a Representative in vivo near-infrared fluorescence images showing the myelin-binding dye DBT in the cervical spinal cord of sham, vehicle, pioglitazone, GW9662 only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice at day 7. Blue rectangles depict the region of interest. b The time courses of the average radiant efficiency of DBT near-infrared fluorescence signals in different groups of mice at days 0, 3, 5, and 7. c Immunostaining for MBP and NF-H in the spinal cord of different groups of mice. Nuclei were counterstained with DAPI. Images are representative photomicrographs showing cross sections of the ventrolateral white matter of the cervical spinal cord. Scale bar: 50 μm. d and e Quantification of MBP immunofluorescence intensity (d) and the number of NF-H spots (e). n = 5 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant, *P < 0.05, **P < 0.01, and ***P < 0.001 Next, we evaluated whether pioglitazone treatment prevented IgG(AQP4)-induced demyelination and axonal loss using immunofluorescence staining. Vehicle-treated IgG(AQP4) mice displayed prominent and patchy loss of MBP and NF-H immunofluorescence signals in the white matter of the spinal cord compared to sham control. Pioglitazone treatment reversed the loss of MBP and NF-H signals in IgG(AQP4) mice. However, administering GW9662 alone or in combination with pioglitazone did not prevent the loss of MBP and NF-H signals (Fig. 3c). The immunofluorescence intensity evaluation confirmed that the level of MBP was significantly elevated in pioglitazone-treated IgG(AQP4) mice compared to vehicle-treated, GW9662-only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice. (Fig. 3d). Furthermore, the relative number of NF-H spots in pioglitazone-treated IgG(AQP4) mice was similar to that in the sham control (Fig. 3e). Consistently, pioglitazone treatment prevented the loss of MBP and NF-H signals in the optic nerves of IgG(AQP4) mice (Supplemental Fig. 2). Together, these results indicate that pioglitazone prevents AQP4-IgG-induced demyelination and axonal loss, while this protective effect is reversed by antagonizing PPAR-γ. PPAR-γ inhibition reverses the effects of pioglitazone on alleviating microglial activation CNS lesions exhibit prominent microglial activation in animal models and in patients with NMOSD [3, 14, 33, 40, 53, 54]. Pioglitazone has been shown to attenuate microglial activation and reduce neuroinflammation in various experimental models of CNS disorders [10, 16, 28, 37]. To examine whether pioglitazone reduces microgliosis in IgG(AQP4) mice, we performed immunofluorescence staining on spinal cord sections using the microglial marker Iba-1. The immunofluorescence intensity of Iba-1 and the number of Iba-1+ cells were significantly decreased in pioglitazone-treated IgG(AQP4) mice compared to vehicle, GW9662 only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice (Fig. 4a-c). Microglia in pioglitazone-treated IgG(AQP4) mice displayed a homeostatic phenotype characterized by a ramified morphology. In contrast, microglia in vehicle, GW9662 only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice exhibited an activated phenotype characterized by a bushy appearance (Fig. 4d–f). Co-immunofluorescence for Iba-1 and GFAP revealed that activated microglia overlapped with astrocytes in vehicle, GW9662 only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice. Pioglitazone treatment reversed the colocalization of microglia and astrocytes (Supplemental Fig. 3). These findings suggest that pioglitazone can dampen microgliosis and induce the transition of microglia to a homeostatic morphology in IgG(AQP4) mice, a process that is dependent on PPAR-γ.Fig. 4Pioglitazone reverses PPAR-γ-dependent microgliosis. a Immunostaining for Iba-1+ microglial cells in the spinal cord of sham, vehicle, pioglitazone, GW9662 only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice. Nuclei were counterstained with DAPI. Images are representative photomicrographs showing cross sections of the ventrolateral white matter of the cervical spinal cord. Scale bars: upper panel, 50 μm; lower panel, 25 μm. b and c Quantification of Iba-1 immunofluorescence intensity (b) and the number of Iba-1+ cells (c). d Representative images of the morphology of Iba-1+ cells in the spinal cord of different groups of mice (Z-stack: red; skeletonized: white; 20 microglial cells per mouse). Scale bar: 5 μm. e and f Quantification of the branch number (e) and length (f) of Iba-1+ cells. n = 5 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant, *P < 0.05, **P < 0.01, and ***P < 0.001 Pioglitazone reverses PPAR-γ-dependent microgliosis. a Immunostaining for Iba-1+ microglial cells in the spinal cord of sham, vehicle, pioglitazone, GW9662 only, or concurrent pioglitazone and GW9662-treated IgG(AQP4) mice. Nuclei were counterstained with DAPI. Images are representative photomicrographs showing cross sections of the ventrolateral white matter of the cervical spinal cord. Scale bars: upper panel, 50 μm; lower panel, 25 μm. b and c Quantification of Iba-1 immunofluorescence intensity (b) and the number of Iba-1+ cells (c). d Representative images of the morphology of Iba-1+ cells in the spinal cord of different groups of mice (Z-stack: red; skeletonized: white; 20 microglial cells per mouse). Scale bar: 5 μm. e and f Quantification of the branch number (e) and length (f) of Iba-1+ cells. n = 5 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant, *P < 0.05, **P < 0.01, and ***P < 0.001 Pioglitazone promotes microglial PPAR-γ expression and reduces neuroinflammation We then evaluated whether pioglitazone enhanced PPAR-γ expression in microglia in IgG(AQP4) mice. PPAR-γ immunoreactivity was observed exclusively in the nuclei of Iba-1+ cells in pioglitazone-treated IgG(AQP4) mice, while it was absent in vehicle-treated IgG(AQP4) mice and the sham control (Fig. 5a and b). Activated microglia secrete pro-inflammatory cytokines, such as IL-1β, IL-6, and TNF-α, which contribute to neuroinflammation [9]. To investigate the impact of increased PPAR-γ expression on the release of pro-inflammatory cytokines, ELISA was used to measure the levels of IL-1β, IL-6, and TNF-α in spinal cord homogenates. The concentrations of IL-1β, IL-6, and TNF-α in vehicle-treated IgG(AQP4) mice were significantly higher than those in sham control mice. Pioglitazone treatment normalized the concentrations of these cytokines (Fig. 5c–e). These data suggest that the beneficial effects of pioglitazone treatment are associated with the reduction of proinflammatory cytokine levels, potentially via suppressing microglia-mediated neuroinflammation through upregulating PPAR-γ expression.Fig. 5Pioglitazone promotes PPAR-γ expression and reduces pro-inflammatory cytokine levels. a Representative images of co-immunostaining for Iba-1 and PPAR-γ in the spinal cord of sham, vehicle, and pioglitazone-treated IgG(AQP4) mice. Nuclei were counterstained with DAPI. Scale bar: 5 μm. b Quantification of the number of Iba-1+PPAR-γ + cells. n = 5 mice per group. c–e ELISA of IL-1β (c), IL-6 (d), and TNF-α (e) levels in spinal cord homogenates of sham, vehicle, and pioglitazone-treated IgG(AQP4) mice. n = 3 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant, *P < 0.05, **P < 0.01, and ***P < 0.001 Pioglitazone promotes PPAR-γ expression and reduces pro-inflammatory cytokine levels. a Representative images of co-immunostaining for Iba-1 and PPAR-γ in the spinal cord of sham, vehicle, and pioglitazone-treated IgG(AQP4) mice. Nuclei were counterstained with DAPI. Scale bar: 5 μm. b Quantification of the number of Iba-1+PPAR-γ + cells. n = 5 mice per group. c–e ELISA of IL-1β (c), IL-6 (d), and TNF-α (e) levels in spinal cord homogenates of sham, vehicle, and pioglitazone-treated IgG(AQP4) mice. n = 3 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant, *P < 0.05, **P < 0.01, and ***P < 0.001 Depletion of microglia/macrophages attenuates AQP4-IgG-induced pathologies To further confirm the role of microglia/macrophages in the current model of NMOSD, we depleted these cells in IgG(AQP4) mice using clodronate liposomes (Fig. 6a). Iba-1 and CD68 immunofluorescence staining revealed the depletion of both peripheral macrophages and spinal cord microglia/macrophages following the administration of clodronate liposomes in IgG(AQP4) mice (Fig. 6b, Supplemental Fig. 4). In the walking test using 1.2 × 80 cm and 0.6 × 80 cm beams, clodronate liposomes significantly reduced both the crossing time and the number of paw slips in IgG(AQP4) mice compared to the PBS-treated control (Fig. 6c–f). The spinal cords of IgG(AQP4) mice treated with clodronate liposomes displayed increased AQP4 and GFAP immunoreactivities compared to the control (Fig. 6g–i). Similarly, MBP immunoreactivity and the number of NF-H spots in these mice were higher than those in the control (Fig. 6j–l). These findings provide further evidence that microglia/macrophages contribute to the development of astrocytopathy and axonal injury, leading to motor impairments in NMOSD.Fig. 6Depletion of macrophages/microglia alleviates motor impairments and spinal cord pathologies induced by human AQP4-IgG. a Experimental design. b Representative images of immunostaining for Iba-1 and CD68 in the spinal cord and liver of PBS- and clodronate liposome-treated IgG(AQP4) mice. Nuclei were counterstained with DAPI. n = 5 mice per group. Scale bar: 50 μm. c–f Beam walking test measuring the time taken and the number of paw slips while the mouse walked across a 1.2 × 80 cm (width × length) beam (c and d) and a 0.6 × 80 cm beam (e and f). n = 8 mice per group. g Representative images of coimmunostaining for AQP4 and GFAP in the spinal cord. Nuclei were counterstained with DAPI. Scale bar: 50 μm. h and i Quantification of AQP4 (h) and GFAP (i) immunofluorescence intensities. n = 5 mice per group. j Representative images of immunostaining for MBP and NF-H in the spinal cord. Nuclei were counterstained with DAPI. Scale bar: 50 μm. k and l Quantification of MBP immunofluorescence intensity (k) and the number of NF-H spots (l). n = 5 mice per group. Data are mean ± SEM. Student’s 2-tailed t test. *P < 0.05, **P < 0.01, and ***P < 0.001 Depletion of macrophages/microglia alleviates motor impairments and spinal cord pathologies induced by human AQP4-IgG. a Experimental design. b Representative images of immunostaining for Iba-1 and CD68 in the spinal cord and liver of PBS- and clodronate liposome-treated IgG(AQP4) mice. Nuclei were counterstained with DAPI. n = 5 mice per group. Scale bar: 50 μm. c–f Beam walking test measuring the time taken and the number of paw slips while the mouse walked across a 1.2 × 80 cm (width × length) beam (c and d) and a 0.6 × 80 cm beam (e and f). n = 8 mice per group. g Representative images of coimmunostaining for AQP4 and GFAP in the spinal cord. Nuclei were counterstained with DAPI. Scale bar: 50 μm. h and i Quantification of AQP4 (h) and GFAP (i) immunofluorescence intensities. n = 5 mice per group. j Representative images of immunostaining for MBP and NF-H in the spinal cord. Nuclei were counterstained with DAPI. Scale bar: 50 μm. k and l Quantification of MBP immunofluorescence intensity (k) and the number of NF-H spots (l). n = 5 mice per group. Data are mean ± SEM. Student’s 2-tailed t test. *P < 0.05, **P < 0.01, and ***P < 0.001 Pioglitazone treatment decreases serum sTREM2 levels in IgG(AQP4) mice TREM2 is a transmembrane cell surface receptor primarily expressed by microglia, associated with triggering microglial activation and their response to inflammatory stimuli [8]. Upon proteolytic cleavage, TREM2 releases its ectodomain from activated microglia into the CSF and blood in a soluble form (sTREM2) [48]. The CSF and blood levels of sTREM2 are potential biomarkers of microglia-mediated neuroinflammatory and neurodegenerative diseases, including NMOSD [15, 20, 42, 56]. We analyzed the correlation between serum sTREM2 levels and markers of neuroinflammation, as well as NMOSD-like pathologies, following pioglitazone treatment in the current NMOSD model. We found that serum sTREM2 levels in vehicle-treated IgG(AQP4) mice were significantly higher than those in sham control mice. Notably, pioglitazone treatment effectively normalized serum sTREM2 levels (Fig. 7a). Serum sTREM2 levels in sham control, vehicle-treated, and pioglitazone-treated IgG(AQP4) mice displayed a strong correlation with the percentage of Iba-1+ cells (r = 0.763, P = 0.017) (Fig. 7b), MBP intensity (r = − 0.713, P = 0.031) (Fig. 7c), number of NF-H spots (r = − 0.789, P = 0.012) (Fig. 7d). Furthermore, serum sTREM2 levels in sham control, vehicle-treated, and pioglitazone-treated IgG(AQP4) mice were closely related to the levels of pro-inflammatory cytokines in the spinal cord, including IL-1β (r = 0.915, P = 0.001) (Fig. 7e), IL-6 (r = 0.912, P = 0.001) (Fig. 7f), and TNF-α (r = 0.874, P = 0.002) (Fig. 7g).Fig. 7Correlation of serum sTREM2 and CNS pathological markers in pioglitazone-treated IgG(AQP4) mice. a Quantification of serum sTREM2 levels in sham, vehicle-, and pioglitazone-treated IgG(AQP4) mice. n = 3 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant, **P < 0.01. b–g Scatter plots representing the correlations between serum sTREM2 level and Iba-1+ cell number (b), MBP intensity (c), number of NF-H spots (d), IL-1β level (e), IL-6 level (f), and TNF-α level (g). Lines of regression are red. Spearman’s correlation coefficient (r) and P values are shown Correlation of serum sTREM2 and CNS pathological markers in pioglitazone-treated IgG(AQP4) mice. a Quantification of serum sTREM2 levels in sham, vehicle-, and pioglitazone-treated IgG(AQP4) mice. n = 3 mice per group. Data are mean ± SEM. One-way ANOVA with post hoc Tukey’s test. ns not significant, **P < 0.01. b–g Scatter plots representing the correlations between serum sTREM2 level and Iba-1+ cell number (b), MBP intensity (c), number of NF-H spots (d), IL-1β level (e), IL-6 level (f), and TNF-α level (g). Lines of regression are red. Spearman’s correlation coefficient (r) and P values are shown Microglia-mediated neuroinflammation in patients with NMOSD To determine the concentration of sTREM2, as well as cytokines and chemokines known to be secreted by activated microglia in the blood of patients with NMOSD and HC, serum samples were analysed using ELISA [9, 30]. The demographic characteristics of the subjects are shown in Supplemental Table 1. Patients with NMOSD exhibited significantly elevated concentrations of serum sTREM2 compared to HC (Fig. 8a). The serum of patients with NMOSD showed a trend of increased IL-1β concentrations compared to HC, although the difference was not statistically significant (Fig. 8b). The concentrations of IL-6, TNF-α, and CXCL7 in the serum of patients with NMOSD were significantly higher than those of HC (Fig. 8c–e).Fig. 8sTREM2 is positively correlated with markers of microglia-mediated inflammation in patients with NMOSD. a–e ELISA of serum levels of sTREM2 (a), IL-1β (b), IL-6 (c), TNF-α (d), and CXCL7 (e) in healthy controls (n = 5–6 subjects) and patients with NMOSD (n = 8 subjects). Data are mean ± SEM. Student’s 2-tailed t-test. ns, not significant, *P < 0.05, **P < 0.01. f–i Scatter plots representing the correlations between serum levels of sTREM2 and IL-1β (f), IL-6 (g), TNF-α (h), and CXCL7 (i). Lines of regression are red. Spearman’s correlation coefficient (r) and P values are shown sTREM2 is positively correlated with markers of microglia-mediated inflammation in patients with NMOSD. a–e ELISA of serum levels of sTREM2 (a), IL-1β (b), IL-6 (c), TNF-α (d), and CXCL7 (e) in healthy controls (n = 5–6 subjects) and patients with NMOSD (n = 8 subjects). Data are mean ± SEM. Student’s 2-tailed t-test. ns, not significant, *P < 0.05, **P < 0.01. f–i Scatter plots representing the correlations between serum levels of sTREM2 and IL-1β (f), IL-6 (g), TNF-α (h), and CXCL7 (i). Lines of regression are red. Spearman’s correlation coefficient (r) and P values are shown Next, we examined whether serum sTREM2 levels were correlated with activated microglia-related cytokines and chemokines. Significant positive correlations were found between serum sTREM2 and IL-1β (r = 0.732, P = 0.005) (Fig. 8f), IL-6 (r = 0.685, P = 0.010) (Fig. 8g), TNF-α (r = 0.567, P = 0.035) (Fig. 8h), and CXCL7 (r = 0.539, P = 0.047) (Fig. 8i). In conjunction with the data from the current mouse model, these findings suggest that pioglitazone could be translated clinically to mitigate microglia-mediated neuroinflammation in patients with NMOSD. Discussion In this study, we investigated whether pioglitazone reduced microglia-driven ADCC and neuroinflammation in NMOSD using a complement-independent mouse model, which was established by repetitive passive transfers IgGs from AQP4-IgG-seropositive NMOSD patients [54, 55]. We found that delayed pioglitazone treatment effectively ameliorated motor impairments and NMOSD-like pathologies in this model. Furthermore, pioglitazone treatment mitigated microgliosis induced by human AQP4-IgG. Mechanistically, the beneficial effects of pioglitazone were reversed by PPAR-γ inhibition, indicating that these effects depended PPAR-γ. Microglial ablation reduced AQP4-IgG-induced pathologies, providing further evidence for the detrimental impacts of microglial activation in NMOSD. Serum levels of sTREM2 were elevated in both the mouse model and patients with NMOSD. Pioglitazone treatment decreased sTREM2 levels, which were correlated with alleviated human AQP4-IgG-induced microglial activation and pathologies. Elevated sTREM2 levels were positively associated with activated microglia-related cytokines and chemokines in patients with NMOSD. Our findings highlight the potential of repurposing pioglitazone as a novel therapeutic to alleviate microglia-mediated neuroinflammation during acute NMOSD attacks. Complement-independent pathophysiologies, including glutamate excitotoxicity and ADCC, are believed to play a crucial role in the early development of NMOSD lesions after AQP4-IgG binds to astrocytic AQP4 [18, 19, 40, 54]. Given that human IgG does not activate mouse complement [44], we have found that repetitive passive transfers of high-dose IgGs from AQP4-IgG-seropositive NMOSD patients exert complement-independent pathogenic effects in mice [54]. Using this model, we have demonstrated that N-methyl-D-aspartate (NMDA) receptor antagonist alleviates glutamate excitotoxicity and NMOSD-like pathologies [55]. Here, our results suggest that reducing ADCC and neuroinflammation caused by activated microglia is a potential therapeutic strategy for NMOSD. Histopathological studies have consistently revealed prominant microglial activation in the lesions of NMOSD patients, as well as in various animal models of the disease [3, 6, 33, 40, 53, 54]. Microglial activation and parenchymal infiltration of macrophages, granulocytes, and natural killer cells upon Fc receptor binding can lead to astrocyte cytotoxicity via ADCC [13, 32, 43]. Interferon-β (IFN-β) promotes microglial activation and expansion, and exacerbates astrocytopathy in experimental NMOSD [52]. Recently, Chen et al. studied the evolving NMOSD pathophysiology using a mouse model developed by continuously infusing AQP4-IgG, without exogenous complement, into the spinal subarachnoid space. It demonstrates that AQP4-IgG binding to astrocytes leads to their activation and an early-activated CNS-intrinsic complement C3a release. Subsequently, the C3a binds to microglial C3a receptor, triggering microglial activation. The activated microglia upregulates C1q production, which further enhances astrocyte-microglia interactions. This process promotes complement cascades, ultimately leading to enhanced neuroinflammation, neuronal damage, and motor deficits [7]. Here, we investigated the effects of pioglitazone treatment on microglial activation in mice receiving repetitive passive transfers of human AQP4-IgG without exogenous human complement. We found that pioglitazone treatment significantly reduced microglial cell numbers and transformed microglial morphology from a bushy reactive appearance to a ramified homeostatic form in IgG(AQP4) mice. In addition, pioglitazone suppressed microglia-astrocyte interaction via direct cellular contact. These observations were associated with ameliorated motor impairments and NMOSD-like pathologies. Additionally, the depletion of microglia/macrophages alleviated NMOSD-like disease activities in this model. In Chen et al.’s study, the authors observed an increase in GFAP immunoreactivity, while we found GFAP loss after passive transfers of AQP4-IgG. This discrepancy may due to a higher amount and a longer period of AQP4-IgG delivery in our model. Importantly, findings from both Chen et al.’s and our studies suggest that AQP4-IgG binding to astrocytic AQP4 can lead to profound microglial activation, neuroinflammation, and NMOSD lesions without complement. Taken together, these findings support that targeting activated microglia is a potential strategy for preventing lesion development in NMOSD. Further experiments showing microglial engulfment of astrocytic material using confocal 3D reconstruction or ultrastructural imaging would support microglial phagocytosis and degradation of astrocytic structures in this model. Additionally, because clodronate liposome treatment depletes both resident microglia and infiltrating macrophages, a more specific approach, such as using microglia-specific depletion models or flow cytometric analysis, will be needed to further clarify the microglial contribution to NMOSD pathologies. PPAR-γ is a ligand-activated transcription factor in immune cells that attenuates their inflammatory responses [47]. It inhibits pro-inflammatory polarization and signalling cascades in microglia [23]. The anti-inflammatory effects of PPAR-γ have gained significant attention in recent years due to their role in autoimmune diseases such as multiple sclerosis, rheumatoid arthritis, and systemic lupus erythematosus [1, 29, 34, 36]. PPARγ overexpression attenuates microglial activation to a disease-associated microglia-like (DAM-like) phenotype and enhances the activation of tissue repair signalling pathways in a mouse model of repetitive mild traumatic brain injury with chronic neuroinflammation [38]. Here, we found that the beneficial effects of pioglitazone treatment on IgG(AQP4) mice were eliminated when co-administered with the PPAR-γ antagonist GW9662. Moreover, pioglitazone treatment upregulated PPAR-γ expression in the microglia of IgG(AQP4) mice. This observation was accompanied by reduced levels of pro-inflammatory cytokines typically associated with microglial activation, including IL-1β, IL-6, and TNF-α. These results suggest that pioglitazone mitigates pro-inflammatory microglial polarization through the activation of PPAR-γ. Further investigations are needed to elucidate the downstream signalling mechanisms following PPAR-γ activation in microglial cells in this model. TREM2 is predominantly expressed by microglia in the CNS. It regulates microglial activation, phagocytosis, and the production of cytokines, chemokines, and complements. The proteolytic cleavage of TREM2 by disintegrin and proteases releases its ectodomain in a soluble form, sTREM2, from activated microglia into the extracellular space. The increase in sTREM2 levels in the CSF and/or serum has been suggested to be closely associated with enhanced inflammation in the CNS. sTREM2 levels are positively linked to markers of neural damage and disease severity in primary angiitis [15], Alzheimer’s disease [56], amyotrophic lateral sclerosis [20], and multiple sclerosis [35]. Indeed, a recent study has demonstrated that CSF sTREM2 levels are positively correlated with markers of microglial activation, neuroinflammation, and neuronal injury in both patients and a mouse model of NMOSD [42]. Consistently, we found that elevated serum sTREM2 levels were correlated with markers of microglial activation in patients with NMOSD. Using our NMOSD mouse model, we demonstrated that pioglitazone treatment effectively restored serum sTREM2 levels, which were strongly correlated with microglial activation, demyelination, axonal loss, and pro-inflammatory cytokine levels. To further support that the current observation is primarily due to microglial activation, future studies could measure cerebrospinal fluid (CSF) sTREM2 levels or employ CNS-specific TREM2 conditional knockout approaches. Our findings suggest the therapeutic potential of repurposing pioglitazone for NMOSD. Repurposed drugs have established safety profiles from prior human use, increasing their overall safety. Concerns have been raised about a potential link between pioglitazone use and an increased risk of bladder cancer. However, a prospective study of diabetic patients found no evidence of a cancer link with pioglitazone [26], and a double-blind trial stroke patients showed no significant difference in cancer incidence between those treated with pioglitazone and those given a placebo [22]. If pioglitazone is effective in reducing neurological disability during NMOSD attacks, it may be prescribed for short-term use of 3–4 weeks, making long-term bladder cancer risks less concerning. This study has certain limitations. First, while we examined the effects of pioglitazone on microglia, its impact on other inflammatory cells including lymphocytes and dendritic cells remains to be investigated. Second, we did not transfer IgG from NMOSD patients together with human complements to the mice. Complement activation by the Fc portion of human IgG bound to mouse astrocytic AQP4, leading to early terminal complement activation product (C5b9)-mediated astrocytic lysis (lytic lesion), was absent in this model. The loss of GFAP in our model was likely due to ADCC mediated by activated microglia [54]. Moreover, although the reduction of AQP4 and GFAP immunoreactivity implies antibody binding, demonstrating human antibody infiltration into the CNS and co-localization with astrocytic AQP4 using immunostaining would provide further evidence for direct astrocytic targeting by human IgG. Third, due to the rarity of NMOSD, we recruited a small number of patients. Future studies with larger sample sizes will help clarify the relationship between microglial activation and neuroinflammation in NMOSD. Additionally, we cannot rule out the possibility that serum sTREM2 is released by peripheral monocytes and macrophages. In conclusion, our study demonstrates the therapeutic efficacy of pioglitazone and elucidates its mechanism of action in a mouse model of NMOSD. Our findings identify an association between serum sTREM2 and markers of microglia-mediated neuroinflammation in both the mouse model and patients with NMOSD. Importantly, pioglitazone treatment effectively normalizes serum sTREM2 levels in this mouse model, underscoring its potential therapeutic efficacy in acute NMOSD attacks. Further clinical studies are needed to validate the efficacy of PPAR-γ agonists in patients with NMOSD. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1