CaMKIIα inhibition facilitates functional recovery after preterm hypoxic-ischemic brain injury by disrupting microglial GPNMB signaling

Introduction

Preterm birth and its complications are major contributors to mortality in children younger than 5 years old (1,2). While advances in perinatal medicine have improved survival rates, the incidence of brain injury among preterm survivors continues to rise (3). As a prevalent complication of preterm birth, preterm brain injury is closely linked to negative long-term neurodevelopmental effects (4-6).

Preterm brain injury is not confined to white matter but also involves gray matter damage, including neuronal and axonal lesions in the cerebral cortex, basal ganglia, hippocampus, thalamus, and cerebellum (5). These pathologies are characterized by neuronal degeneration and migration abnormalities, axon loss, and synaptic dysfunction, which result in abnormal neural circuit signaling (7,8). Neuroimaging studies have demonstrated substantial cortical and subcortical volume atrophy in preterm children after brain injury. Pathologically, this volume deficit is closely associated with the defective maturation of neuronal dendritic architecture, manifested in decreased complexity of dendritic branching and shortened length (9). In agreement with this, McClendon et al. (10) found that brain injury results in an impairment of caudate nucleus neuron maturation in a fetal sheep hypoxia-ischemia (HI) model, as demonstrated by abnormalities in dendritic spine number and morphology. Thus, our studies and other reports support the idea that the impaired development of dendritic spines is an important pathological step that leads to subsequent synaptic dysfunction and neuronal deficits after preterm brain injury.

Recent studies demonstrated that HI brain injury in immature mice resulted in an early, non-selective increase in the number of both excitatory and inhibitory synapses and upregulation of synaptic proteins synaptophysin (SYN) and postsynaptic density protein 95 (PSD95) in the cortex and hippocampal region at 14 days after HI, which indicated an impairment in the dynamic balance between synapse formation and pruning (11). More importantly, these newly formed dendritic spines were developed in a stagnated state and failed to undergo normal differentiation and maturation, which could lead to aberrant neural circuit function and long-term neurobehavioral deficits (12).

The fine-tuned regulation of dendritic spine formation, maturation, and pruning is important for maintaining synaptic homeostasis and neural circuit function (13,14). Abnormalities in synaptic plasticity lead to spatial memory deficits, a phenomenon most concretely explained by the dysregulation of dendritic spine dynamics. The impairment of long-term potentiation and other forms of plasticity often reflects a failure in the structural-functional coupling at spines, where experience-dependent signals fail to convert labile, immature spines into stable, mushroom-shaped structures. This breakdown in spine maturation and stabilization directly undermines the circuit-level refinement required for accurate spatial memory (15,16). Mouse models of autism exhibit reduced dendritic arborization and synaptic transmission, leading to social deficits (17). In schizophrenia, genetic risk factors may lead to the disruption of microglia-dependent synaptic pruning, which may contribute to neurobehavioral abnormalities (18). Abnormalities in synaptic number and function are involved in the pathogenesis of many neurological disorders that lead to cognitive deficits, learning disabilities, and motor deficits (19).

A conserved regulator of synaptic function is Ca²⁺/calmodulin-dependent protein kinase II (CaMKII), a serine/threonine kinase highly enriched in the brain. The α-isoform of CaMKII (CaMKIIα) is abundant in the hippocampus where it is highly enriched and plays a central role in synaptic signaling, spine morphology and plasticity, as well as learning and memory (20-22). Once activated and autophosphorylated, CaMKIIα can modulate a multitude of downstream targets. CaMKIIα induces dendritic spine growth and stabilization, recruits and locally stabilizes postsynaptic scaffold proteins such as neuroligin-1 (23), and promotes the insertion of the ɑ-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subunit glutamate receptor A1 (GluA1) into the postsynaptic membrane. CaMKIIα also stabilizes the PSD95/GluA1 complex (24,25), a mechanism critical for synaptic strength and long-term potentiation (26).

Glycoprotein non-metastatic melanoma protein B (GPNMB) is a type I transmembrane glycoprotein involved in cell differentiation and neuroinflammation. GPNMB is expressed in the nervous system by microglia, astrocytes and neurons (27,28). Proteolytic cleavage by ADAM proteins releases the soluble extracellular domain of GPNMB (sGPNMB) (29). Recent research has highlighted its role in neurology. GPNMB interacts with α-Synuclein, and reduced GPNMB expression is linked to synaptic loss in Parkinson’s disease (30). Our previous observations indicated a microglia-specific overexpression of GPNMB following HI injury in mice triggers aberrant synapse formation and long-term neurobehavioral dysfunction. We further found that microglia release GPNMB in an activity-dependent manner, correlating with neuronal activity intensity.

However, it remains unknown whether microglial GPNMB influences neuronal synapse development and long-term function by regulating CaMKIIα autophosphorylation and, consequently, the expression and localization of GluA1 in the postsynaptic density. Furthermore, it is unclear whether inhibition of CaMKIIα can disrupt this signaling axis and ameliorate the synaptic and behavioral deficits induced by HI injury.

Based on this, we hypothesize that the GPNMB-CaMKIIα-GluA1 signaling pathway is a key mechanism mediating synaptic deficits and long-term neurobehavioral abnormalities after HI injury in the immature brain. To test this, we employed a postnatal day 3 (P3) mouse model of HI injury, transgenic mice with microglia-specific overexpression of GPNMB, and pharmacological inhibition of CaMKIIα using the selective inhibitor KN93. We present this article in accordance with the ARRIVE reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2025-1-922/rc) (31).

Methods

Experimental animals

Specific-pathogen-free (SPF) C57BL/6J mice were employed throughout all experiments. The mice were given a regular diet and unlimited access to water while being kept in a 12-hour light/dark cycle at 21–25 ℃ and 40–60% humidity. All experimental procedures were approved by the Animal Care and Use Committee of Children’s Hospital of Fudan University (approval No. 2020239) and were conducted in accordance with the National Institutes of Health (NIH) Guidelines for the Care and Use of Laboratory Animals. All effort was made to minimize animal suffering, and the minimal number of animals required to achieve statistical significance was used. The experimental design and timeline are summarized in Figure 1. A protocol was prepared before the study without registration.

Figure 1 Experimental timelines for assessing the role of microglial GPNMB and its effect on CaMKIIα inhibition after HI injury. (A) Experiment 1: timeline for the induction of microglia-specific GPNMB overexpression and HI injury. (B) Experiment 2: timeline for the induction of microglia-specific GPNMB overexpression and administration of the CaMKIIα inhibitor KN93 under normal (non-HI) conditions. (C) Experiment 3: timeline for the induction of microglia-specific GPNMB overexpression, HI injury, and KN93 administration under HI conditions. KN93 (10 mg/kg) or DMSO was administered intraperitoneally 0.5 hours before tissue collection on P17 and 0.5 hours before each behavioral test in adulthood. CaMKIIα, calcium/calmodulin-dependent protein kinase II alpha; DMSO, dimethyl sulfoxide; GPNMB, glycoprotein non-metastatic melanoma protein B; HI, hypoxia-ischemia; KN93, C₂₆H₂₉ClN₂O₄S·H₃O₄P; MG^GPNMB, microglia-specific GPNMB-overexpression mice; MG^WT, microglia-specific wild-type mice; OFT, open-field test; P, postnatal day.

Preterm HI brain injury

A modified Rice-Vannucci model was used to induce HI brain injury in C57BL/6 mice on P3 as previously described (32). In brief, isoflurane was used to anesthetize pups of both sexes with comparable mean body weights. The common carotid artery on the right was dissected and irreversibly ligated for the HI group. After a one-hour recovery period with the dam, the pups were placed in a chamber perfused with a humidified hypoxic gas mixture (10% O₂/90% N₂) for 90 min at 36 ℃. Sham-operated controls underwent identical anesthesia and surgical procedures, including exposure of the artery, but without ligation or hypoxia. All pups were returned to their dams after the procedure and were monitored closely.

Microglia-specific GPNMB overexpression mouse model

To achieve microglia-specific overexpression of GPNMB, we generated a conditional transgenic mouse model using the Cre-loxP system. All transgenic mice used in this study are on a C57BL/6J background. Briefly, R26-CAG-LSL-GPNMB mice {C57BL/6JSmoc-Gt(ROSA)26Sorem1[CAG-Loxp-Stop-Loxp (SLS)-GPNMB-WPRE-polyA]Smoc, customized from Shanghai Southern Model Biotechnology Development} were crossed with Cx3cr1-CreERT2 mice [C57BL/6JSmoc-Cx3cr1em1-(CreER2-WPRE-polyA)Smoc, Cat# NM-KI-200157, purchased from Shanghai Model Organisms Center] to obtain offspring with the genotype R26-CAG-LSL-GPNMB; Cx3cr1-CreERT2. The Cx3cr1-CreERT2 line has been validated for tamoxifen-inducible, microglia-specific Cre activity (validation data available at https://www.modelorg.com/portal/article/index/id/9554/post_type). To induce GPNMB overexpression, tamoxifen was administered to mice to activate Cre recombinase specifically in microglia. This resulted in the excision of the LSL cassette, allowing the expression of GPNMB under the control of the ubiquitous CAG promoter. These mice were hereafter referred to as MG^GPNMB (Microglia-specific GPNMB overexpression), with the genotype Rosa26^GPNMB/GPNMB × Cx3cr1^CreERT2/WT. Control mice (littermates carrying the R26-CAG-LSL-GPNMB allele but lacking the Cx3cr1-CreERT2 transgene) received the same tamoxifen administration. Since they lacked Cre recombinase, the Stop cassette remained intact, and GPNMB expression was blocked. These mice are designated as MG^WT (Microglia-specific wild-type), with the genotype Rosa26^GPNMB/GPNMB × Cx3cr1^WT/WT.

Experimental grouping and drug administration

Experiment 1: four groups of mice were allocated at random: MG^WT + Sham, MG^WT + HI, MG^GPNMB + Sham, and MG^GPNMB + HI. To induce microglial-specific gene manipulation, neonatal pups received a single subcutaneous injection of tamoxifen (20 mg/kg; MedChemExpress, USA) or an equal volume of corn oil vehicle. Three days after injection, the mice underwent either HI or Sham surgery. Tissue samples were collected on postnatal day 17 (P17). Another cohort of mice was raised to adulthood for assessment of long-term neurobehavioral function using the open field test, rotarod test, and Morris water maze test. For each group, half of the animals were used for behavioral tests and the other half for biochemical and molecular analyses (Figure 1A).

Experiment 2: four groups of mice were allocated at random: MG^WT + DMSO, MG^GPNMB + DMSO, MG^WT + KN93, and MG^GPNMB + KN93. Neonatal mice received tamoxifen or corn oil as described in Experiment 1. An intraperitoneal injection of KN93 (10 mg/kg; MedChemExpress, USA) or an equivalent volume of DMSO vehicle was given 0.5 hours before tissue was collected on P17. For behavioral testing in adulthood, KN93 or DMSO was similarly injected 0.5 hours prior to each test session. As in Experiment 1, separate cohorts of mice were used for behavioral testing and biochemical/molecular analyses (Figure 1B).

Experiment 3: four groups of mice were allocated at random under HI conditions: HI + MG^WT + DMSO, HI + MG^GPNMB + DMSO, HI + MG^WT + KN93, and HI + MG^GPNMB + KN93. Neonatal mice received tamoxifen or corn oil followed by HI surgery as described in Experiment 1. KN93 or DMSO was administered as described in Experiment 2. As in Experiments 1 and 2, separate cohorts of mice were used for behavioral testing and biochemical/molecular analyses (Figure 1C).

Genotype identification

Toe tissue samples were collected and lysed in 50 µL of 50 mM NaOH at 95 ℃ for 30 min. After cooling to room temperature, the lysate was neutralized with 5 µL of 1 M Tris-HCl (pH 8.0), vigorously vortexed, and centrifuged at 12,000× g for 5 min. The supernatant (3 µL) was used as the PCR template. PCR was performed in a 25 µL reaction mixture containing 12.5 µL of 2× Taq Master Mix, 1 µL each of forward and reverse primers (10 µM), 3 µL of template DNA, and 7.5 µL of sterile ultrapure water. The amplification process included a 5 min initial denaturation at 95 ℃, 35 cycles of denaturation at 95 ℃ for 30 s, annealing at 60 ℃ for 30 s, extension at 72 ℃ for 1 min, and a final extension at 72 ℃ for 5 min. A Bio-Rad gel imaging equipment was used to observe the PCR results after they had been separated by electrophoresis on a 1.5% agarose gel at 120 V for 30 min.

Tamoxifen administration

Tamoxifen (MedChemExpress, USA) was dissolved in corn oil to a final concentration of 20 mg/mL. The working solution was stored at 4 ℃ protected from light for up to one week. To induce gene recombination, neonatal pups (P0) received a single subcutaneous injection of tamoxifen at a dose of 200 µg per pup (in a volume of 10 µL). Injections were performed using a micro-syringe inserted horizontally into the subcutaneous space along the dorsal midline.

KN93 administration

KN93 (MedChemExpress, USA) was dissolved in 10% DMSO to a final concentration of 1 mg/mL. The solution was stored at 4 ℃ protected from light for up to one week. On P17, mice received a single intraperitoneal injection of KN93 at a dose of 200 µg per pup (in a volume of 200 µL) according to a previously described method (33). Injections were administered 30 min prior to experimental procedures. Proper needle placement within the peritoneal cavity was confirmed by the absence of blood or fluid aspiration before injection.

Western blot

Brain tissues were homogenized in ice-cold RIPA lysis buffer (Thermo Fisher, USA) supplemented with protease and phosphatase inhibitors (Thermo Fisher, USA) and incubated on ice for 30 min. Lysates were centrifuged at 15,000× g for 25 min at 4 ℃, and the supernatant was collected. Protein concentration was determined using a BCA protein assay kit (Beyotime, China). Equal amounts of protein were separated by SDS-PAGE on 10% gels and subsequently transferred to nitrocellulose (NC) membranes. The membranes were blocked with 5% bovine serum albumin (BSA) in TBST for 2 hours at room temperature with gentle shaking. Membranes were blocked and then treated for the entire night at 4 ℃ with primary antibodies diluted in blocking buffer. The main antibodies listed below were employed: Rabbit anti-CaMKII alpha (phospho T286) (Abcam, Cambridge, UK), Rabbit anti-CaMKII alpha (Abcam, Cambridge, UK), Rabbit anti-brain-derived neurotrophic factor (BDNF) (Proteintech, China), Rabbit anti-cAMP response element-binding protein (CREB) (Proteintech, China), Rabbit anti-PSD95 (Abcam, Cambridge, UK), Rabbit anti-SYN (Proteintech, China), Rabbit anti-Glutamate Receptor 1 (Absin, China), Rabbit anti-GAPDH (Absin, China), Rabbit anti-β-actin (Absin, China). Membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies (Goat anti-mouse or Goat anti-rabbit IgG, Absin, China) for one hour at room temperature after being washed five times with TBST. After another round of TBST washes, bands of proteins were observed with a Bio-Rad gel imaging system and an enhanced chemiluminescence (ECL) substrate (Thermo Fisher, USA). Total protein lysates were used; membrane fractionation or surface biotinylation was not performed, so GluA1 signals represent total protein levels. The uncropped full-length blots for all figures are provided in Figure S1.

Transmission electron microscopy (TEM)

Synaptic ultrastructure was examined by TEM using a previously described method (34) with modifications. Briefly, mice were given ice-cold 2.5% glutaraldehyde and PBS transcardially while under profound isoflurane anesthesia. After being quickly extracted, the brains were coronally sectioned into 1-mm-thick slices. The hippocampal CA1 region was microdissected from tissue slices located 1.5–2.5 mm posterior to bregma according to a mouse brain atlas. After collection, the samples were immersed in 2.5% glutaraldehyde overnight at 4 ℃. Following fixation, the samples were rinsed in 0.1 M phosphate buffer and post-fixed with 1% osmium tetroxide for 3 hours. Following a series of graded ethanol dehydration, acetone transition, and epoxy resin embedding, tissues were prepared. A Hitachi TEM system running at 80 kV was used to examine the ultrathin sections (90 nm) that had been cut using a Leica ultramicrotome and stained with uranyl acetate and lead citrate. The synaptic ultrastructure of the hippocampal CA1 region was examined. ImageJ software was used to quantitatively evaluate synaptic parameters.

Open field test

Locomotion and spontaneous activity of mice were assessed using open-field test, with video tracking performed via the system (Shanghai JiLiang Software Technology Co.). The experimental procedure was conducted according to a published protocol (35). Mice were placed in a square open-field (40 cm × 40 cm × 30 cm) and allowed to explore for 5 min. Mice were allowed 30 min to acclimate to the behavioral room prior to testing. All testing was conducted under quiet conditions. Time spent in the central zone (20 cm × 20 cm) and the total distance moved in the center were measured and analyzed.

Rotarod test

A rotarod device (Shanghai JiLiang Software Technology Co.) was used to evaluate motor coordination and endurance following a previously published protocol (36). Mice were acclimated to the testing room for at least 15 min before testing. The rotarod accelerated from 5 rpm to 40 rpm over 300 s. The latency to fall was recorded as the primary endpoint. Mice that gripped the rod and completed a full passive rotation without walking were considered to have fallen, and latency was recorded at the onset of passive rotation. Each mouse underwent three trials with a 15 min interval between trials, and the mean of the three trials was calculated. Mice that jumped off the rod intentionally were excluded. The equipment was cleaned with 75% ethanol and wiped with a dry cloth to remove waste residues.

Morris water maze test

A Morris water maze video analysis system (Shanghai JiLiang Software Technology Co.) was used to evaluate spatial learning and memory. The maze consisted of a circular pool (120 cm in diameter, 45 cm in height) filled with water maintained at 22±1 ℃. A hidden platform (10 cm in diameter) was submerged 1 cm below the water surface. Distinctive visual cues were placed on the walls of the testing room to assist spatial navigation. The navigation phase and the probe trial were the two stages of the experiment, which was carried out in accordance with Vorhees et al. (37). During the navigation phase, mice underwent four trials per day from four different starting quadrants. Each trial lasted up to 60 s, during which escape latency and swimming distance were recorded. If a mouse failed to find the platform within 60 s, it was gently guided to the platform and allowed to remain for 10 s. The inter-trial interval was 15–20 min. Twenty-four hours after the final navigation session, a probe trial was conducted, during which the platform was removed and mice were allowed to swim freely for 60 s. The first crossing latency, number of platform crossings, and time spent in the target quadrant were recorded.

Co-immunoprecipitation (Co-IP)

The interaction between CaMKIIα and GluA1 was analyzed by Co-IP. Cortical and hippocampal tissues from mice were lysed, and the lysates were precleared with 1 µg of non-specific IgG and 20 µL of Protein G agarose beads (Sigma, USA). After that, the supernatants were incubated with 2 µg of either mouse anti-CaMKIIα antibody (Invitrogen, USA) or control IgG overnight at 4 ℃. Subsequently, 40 µL of Protein G agarose beads were added to each sample and incubated for 4 h at 4 ℃. The immunoprecipitated complexes were washed three times with Buffer A (150 mM NaCl, 50 mM Tris-HCl, 0.1% Triton X-100, 1 mM EDTA) followed by three washes with Buffer B (300 mM NaCl, 50 mM Tris-HCl, 0.1% Triton X-100, 1 mM EDTA). The complexes were then eluted and separated by SDS-PAGE, and Western blot was performed to detect GluA1 and CaMKIIα. Input samples and an IgG isotype control were included to ensure experimental specificity. We used total protein lysates for this assay (no membrane fractionation).

Immunofluorescence staining

Immunofluorescence staining was performed as follows. Mice were anesthetized with isoflurane and transcardially perfused with saline followed by 4% paraformaldehyde (PFA) via the left ventricle. Whole brains were post-fixed in 4% PFA for four to six hours at 4 ℃. Following fixation, the brains were successively submerged in 20% and 30% sucrose solutions at 4 ℃ to cryoprotect them. After saturation, the brains were embedded in OCT compound. Coronal sections (40 µm) containing the hippocampal region (1.5–2.5 mm posterior to bregma) were cut on a cryostat at −25 ℃. Sections were rinsed in PBS to remove residual OCT and to ensure anatomical consistency across groups. After blocking the sections for an hour at room temperature with immunofluorescence blocking solution, the sections were incubated overnight at 4 ℃ with primary antibodies diluted in the same blocking solution. Primary antibodies were mouse anti-GluA1 (Proteintech, China) and rabbit anti-PSD95 (Proteintech, China). Sections were reheated to room temperature for half an hour the next day, then rinsed with PBS and left to incubate for two hours in the dark with matching fluorescent secondary antibodies. Secondary antibodies included donkey anti-mouse Cy3 and donkey anti-rabbit Alexa Fluor 488 (Jackson ImmunoResearch, USA). After secondary antibody incubation, sections were washed extensively with PBS −0.1% Triton X-100 and subsequently counterstained with DAPI (Servicebio, China) for 15 min at room temperature in the dark.

Statistical analysis

The data are presented as mean ± standard error of the mean (SEM). GraphPad Prism 9 (GraphPad Software, USA) and SPSS 26.0 (IBM, USA) were used for statistical analyses. Normality and homogeneity of variance were verified prior to analysis of variance (ANOVA). An unpaired Student’s t-test was applied to assess comparisons between two groups. For multiple group comparisons, one-way or two-way ANOVA was used as appropriate, with repeated-measures ANOVA applied for the Morris water maze navigation test. Bonferroni’s post hoc test was used for all ANOVAs. For rotarod fall latency and water maze probe trial first crossing latency, Kaplan-Meier survival curves with log-rank tests were used. A P value <0.05 was established as the threshold for statistical significance.

Results

Microglial GPNMB overexpression induces aberrant CaMKIIα activation and synaptic impairments after HI brain injury

To examine the role of microglial GPNMB in CaMKIIα activation following HI injury, we used a microglia-specific GPNMB-overexpressing mouse model (MG^GPNMB) along with their wild-type littermates (MG^WT). On P3, the pups underwent either HI injury via the modified Rice-Vannucci procedure or sham surgery (Sham), which served as the procedural control. HE staining results indicated that the injured hemisphere (right side) of HI-group mice displayed atrophy in the hippocampus and corpus callosum, accompanied by enlargement of the lateral ventricle. In contrast, the contralateral hemisphere (left side) and the sham-operated group showed no structural alterations (Figure 2A). Western blot analysis showed that GPNMB protein was highly expressed in the brains of MG^GPNMB mice compared with MG^WT mice. HI injury further enhanced the expression of GPNMB in MG^GPNMB mice (Figure 2B,2C). Therefore, we successfully constructed the conditional microglial GPNMB-overexpressing transgenic model.

Figure 2 Microglial GPNMB overexpression potentiates CaMKIIα activation and synaptic abnormalities after HI injury. (A) Representative images of HE-stained brain sections at 14 days post-injury. Scale bar =1 mm. (B) Representative Western blots of GPNMB. (C) Relative GPNMB expression. (D) Representative Western blots of p-CaMKIIα and t-CaMKIIα. (E) Relative p-CaMKII expression. (F) Relative t-CaMKII expression. (G) Representative Western blots of PSD95. (H) Relative PSD95 expression. (I) Representative Western blots of SYN. (J) Relative SYN expression. (K) Representative electron micrographs of hippocampal CA1 synapses. Key features are indicated: synaptic vesicles (yellow arrows), synaptic cleft (red arrows), and postsynaptic density (blue arrows). Scale bar: 1 µm. (L) Number of synaptic vesicles. (M) Synaptic cleft width. (N) PSD thickness. Data are presented as mean ± SEM (n=6/group). *, P<0.05; **,P<0.01; ***, P<0.001. CA1, cornu ammonis 1; CaMKIIα, calcium/calmodulin-dependent protein kinase II alpha; GPNMB, glycoprotein non-metastatic melanoma protein B; HE, hematoxylin-eosin; HI, hypoxia-ischemia; p-CaMKIIα, phosphorylated CAMKIIα; PSD, postsynaptic density; PSD95, postsynaptic density protein 95; SYN, synaptophysin; SEM, standard error of the mean; t-CaMKIIα, total CAMKIIα.

Western blot analysis showed that HI injury increased the expression of phosphorylated CaMKIIα (p-CaMKIIα) and total CaMKIIα (t-CaMKIIα) in the ipsilateral hemisphere of HI + MG^WT group compared with Sham + MG^WT group. Interestingly, microglial GPNMB overexpression further enhanced the expression of CaMKIIα phosphorylation. HI + MG^GPNMB group showed significantly higher expression of p-CaMKIIα and t-CaMKIIα than HI + MG^WT group (Figure 2D-2F). These results indicate that microglial GPNMB overexpression enhanced the CaMKIIα phosphorylation after perinatal HI brain injury. Our results demonstrate that microglia-derived GPNMB regulated CaMKIIα activation under a pathological conditions. To investigate whether microglial GPNMB overexpression induced defective synaptogenesis, we examined the expression of synaptic protein and synaptic ultrastructure. Western blot analysis showed that the expression of postsynaptic protein PSD95 and presynaptic protein SYN were significantly upregulated in HI + MG^WT mice compared with Sham + MG^WT controls. Importantly, GPNMB overexpression further enhanced the expression of PSD95 and SYN in MG^GPNMB group than MG^WT group (Figure 2G-2J). These results indicated that GPNMB overexpression induced defective synaptogenesis by enhancing the accumulation of postsynaptic and presynaptic protein.

Electron microscopy analysis revealed significant synaptic ultrastructural alterations in the hippocampal CA1 region after HI injury (Figure 2K). The HI + MG^WT group showed an increased number of synaptic vesicles, a relatively wider synaptic cleft and a thicker PSD compared with Sham + MG^WT controls. HI + MG^GPNMB group showed a further increase in the number of synaptic vesicles, a wider synaptic cleft and a thicker PSD, and exhibited conspicuous irregular PSD profiles and electron-dense deposits (Figure 2L-2N), suggesting that HI injury induced severe deficits in synaptic integrity, and microglial GPNMB overexpression aggravated these pathological changes.

Microglial GPNMB overexpression exacerbates long-term neurobehavioral deficits after HI brain injury

Consistent with synaptic impairments, microglial GPNMB overexpression significantly exacerbated neurobehavioral deficits induced by HI injury. Behavioral tests were performed to investigate anxiety-like behavior, motor coordination, and long-term memory in Sham + MG^WT, Sham + MG^GPNMB, HI + MG^WT, and HI + MG^GPNMB mice.

The total distance traveled by the four groups in the open field test was not significantly different. Compared with the Sham + MG^WT group, HI + MG^WT mice spent less time in the center and traveled a shorter distance in the center. Moreover, the HI + MG^GPNMB group exhibited even fewer entries into the center, decreased central time, and traveled a shorter explore distance compared with the HI + MG^WT group (Figure 3A-3C). These results demonstrated that HI injury could induce anxiety-like behavior and reduce exploratory behavior, and that microglial GPNMB overexpression further aggravated these behaviors.

Figure 3 Microglial GPNMB overexpression exacerbates long-term neurobehavioral deficits after HI brain injury. A-C) Performance in the open field test. (A) Representative movement trajectories. (B) Central times. (C) Explore distance. (D) Rotarod test: Percentage of mice remaining on rotarod. (E-G) Performance in the Morris water maze navigation test (days 1-4). Quantitative analysis of (F) escape latency and (G) explore distance to the platform. (H-J) Morris water maze probe test (day 5). (H) Platform crossing-free survival, (I) Crossing (frequency), and (J) Percentage of time in the target quadrant. Data are presented as mean ± SEM (n=15/group). *, P<0.05; **,P<0.01; ***, P<0.001 HI + MG^WTvs. Sham + MG^WT group; ^#, P<0.05; ^###, P<0.001 HI + MG^GPNMBvs. HI + MG^WT group. GPNMB, glycoprotein non-metastatic melanoma protein B; HI, hypoxia-ischemia; MG^GPNMB, microglia-specific GPNMB-overexpression mice; MG^WT, microglia-specific wild-type mice; SEM, standard error of the mean.

In the rotarod test, HI + MG^WT mice exhibited a significant decrease in latency to fall compared with the Sham + MG^WT group. HI + MG^GPNMB mice exhibited decreased latency to fall compared with the HI + MG^WT group (Figure 3D). These results demonstrate that HI injury could induce long-term motor coordination deficits, and that microglial GPNMB overexpression further aggravated these deficits.

During the 4-day navigation phase, escape latency and swimming distance to locate the hidden platform progressively decreased across training days in all groups, indicating successful spatial learning. Compared with the Sham + MG^WT group, the HI + MG^WT group exhibited prolonged escape latency and increased swimming distance on all training days. The HI + MG^GPNMB group showed even greater impairments in both parameters compared with the HI + MG^WT group on days 2–4, but not on day 1 (Figure 3E-3G). In the probe test on day 5, the HI + MG^WT group showed longer first crossing latency, fewer platform crossings and less time spent in the target quadrant compared with Sham controls. The HI + MG^GPNMB group exhibited further deterioration in these indices relative to the HI + MG^WT group (Figure 3H-3J).

Together, these results demonstrate that microglial GPNMB overexpression aggravates synaptic surplus and significantly exacerbates long-term neurobehavioral deficits following HI brain injury.

GPNMB-induced synaptic and neurobehavioral deficits are mediated by aberrant CaMKIIα activation

To examine how CaMKIIα activation contributes to deficits caused by GPNMB, we first measured the expression of the GluA1 protein. Western blot analysis of hippocampal lysates showed that microglial GPNMB overexpression significantly increased total GluA1 protein levels in the MG^GPNMB + DMSO group compared to the MG^WT + DMSO controls. This upregulation was suppressed by CaMKIIα inhibition with KN93, as evidenced by the significantly reduced total GluA1 expression in the MG^GPNMB + KN93 group (Figure 4A,4B). These results indicate that CaMKIIα activation is necessary for GPNMB-induced GluA1 upregulation.

Figure 4 Microglial GPNMB drives synaptic disruption through aberrant CaMKIIα activation and direct CaMKIIα-GluA1 binding. (A) Representative Western blots of GluA1. (B) Relative GluA1 expression. (C) Predicted molecular docking model between CaMKIIα (kinase domain, cyan) and GluA1 (C-terminal domain, purple) by AlphaFold3. (D) Co-immunoprecipitation assays using hippocampal lysates. (E) Representative immunofluorescence images showing co‑staining of GluA1 (red) and PSD95 (green) in the hippocampal CA1, CA3 and DG regions. (F, G) Colocalization analysis: (F) Colocalization, (G) Colocalization cells density. (H, I) Fluorescence intensity in colocalized regions: (H) GluA1 intensity in colocalized regions, (I) PSD95 intensity in colocalized regions. Data are normalized and presented as mean ± SEM (n=6/group). ***, P<0.001. CA1, cornu ammonis 1; CA3, cornu ammonis 3; CaMKIIα, calcium/calmodulin-dependent protein kinase II alpha; DG, dentate gyrus; GluA1, glutamate receptor a1; GPNMB, glycoprotein non-metastatic melanoma protein B; PSD95, postsynaptic density protein 95; SEM, standard error of the mean.

To explore the underlying mechanism, we used AlphaFold3-based molecular docking to predict a direct interaction between CaMKIIα and GluA1. The model suggested that the C-terminal domain of GluA1 binds to the kinase domain of CaMKIIα via multimodal interactions, including electrostatic (e.g., GLU105-ARG606) and hydrogen-bonding (e.g., ASN416-THR110) networks (Figure 4C). Co-IP assays experimentally validated this interaction, with anti-CaMKIIα antibodies pulling down GluA1 (from total protein lysates), while control IgG did not (Figure 4D). These data provide structural and biochemical evidence for a CaMKIIα-GluA1 complex that may mediate aberrant signaling in response to GPNMB overexpression.

We further evaluated the synaptic distribution of GluA1 and PSD95 via immunofluorescence staining across hippocampal CA1, CA3, and DG subregions. Quantitative analysis showed that compared with the MG^WT + DMSO group, the positive area of GluA1 and PSD95 (Figure 4E) as well as the percentage of colocalization-positive cells (Figure 4F), the colocalization-positive cell density (Figure 4G), and the mean fluorescence intensity of protein of GluA1 (Figure 4H) and PSD95 (Figure 4I) were up-regulated in the MG^GPNMB + DMSO group. All the above-mentioned up-regulations were inhibited in the MG^GPNMB + KN93 group, indicating that KN93 could partially reverse the effects of GPNMB overexpression.

In summary, we propose that microglial GPNMB alters the localization and expression of synaptic proteins in a CaMKIIα-dependent manner by enhancing the interaction of GPNMB with scaffolding proteins, such as PSD95, or by impairing the AMPAR (GluA1) trafficking. Our results demonstrate that GPNMB induced GluA1 overexpression and synaptic mislocalization via the aberrant CaMKIIα activation, providing evidence for the underlying mechanism by which microglial GPNMB contributes to synaptic and behavioral deficits.

Inhibition of CaMKIIα rescues GPNMB-induced synaptic defects

To investigate whether inhibition of CaMKIIα could rescue GPNMB-induced synaptic defects, we first evaluated the effect of CaMKIIα inhibitor KN93 (10 mg/kg, i.p.) on MG^WT + DMSO, MG^GPNMB + DMSO, MG^WT + KN93, and MG^GPNMB + KN93 mice. The results of western blot analysis showed that KN93 markedly decreased the expression of p-CaMKIIα and t-CaMKIIα in MG^WT + KN93 group and MG^GPNMB + KN93 group compared with their corresponding DMSO group (Figure 5A-5C).

Figure 5 Inhibition of CaMKIIα rescues synaptic defects induced by microglial GPNMB overexpression. (A) Representative Western blots of p-CaMKIIα and t-CaMKIIα. (B) Relative p‑CaMKII expression. (C) Relative t‑CaMKII expression. (D) Representative Western blots of PSD95. (E) Relative PSD95 expression. (F) Representative Western blots of SYN. (G) Relative SYN expression. (H) Representative electron micrographs of hippocampal CA1 synapses. Key features are indicated: synaptic vesicles (yellow arrows), synaptic cleft (red arrows), and postsynaptic density (blue arrows). Scale bar: 1 µm. (I) Number of synaptic vesicles. (J) Synaptic cleft width. (K) PSD thickness. Data are presented as mean ± SEM (n=6/group). *, P<0.05; **,P<0.01; ***, P<0.001. CA1, cornu ammonis 1; CaMKIIα, calcium/calmodulin-dependent protein kinase II alpha; GPNMB, glycoprotein non-metastatic melanoma protein B; p-CaMKIIα, phosphorylated CaMKIIα; PSD, postsynaptic density; PSD95, postsynaptic density protein 95; SEM, standard error of the mean; SYN, synaptophysin; t-CaMKIIα, total CaMKIIα.

Then we investigated the expression of synaptic markers in the cortex and hippocampus. Compared with MG^WT + DMSO group, the expression of postsynaptic protein PSD95 and presynaptic protein SYN were significantly upregulated in MG^GPNMB + DMSO group. Furthermore, KN93 markedly rescued GPNMB-induced upregulation of PSD95 and SYN in MG^GPNMB + KN93 group, and the expression of PSD95 and SYN were decreased to the similar level of MG^WT + DMSO group (Figure 5D-5G).

Electron microscopy analysis of hippocampal CA1 region also revealed that inhibition of CaMKIIα could rescue GPNMB-induced synaptic pathology (Figure 5H). Compared with MG^WT + DMSO controls, the MG^GPNMB + DMSO group showed obvious synaptic pathology, including increased number of synaptic vesicle, widened synaptic cleft and thickened PSD. Inhibition of CaMKIIα by KN93 substantially rescued GPNMB-induced synaptic pathology, as revealed by restored number of synaptic vesicle, narrowed synaptic cleft, reduced PSD thickness and improved synaptic ultrastructure (Figure 5I-5K).

All these results demonstrate that inhibition of CaMKIIα activity could rescue biochemical and ultrastructural synaptic defects induced by microglial GPNMB overexpression, and further supported that CaMKIIα activation mediated GPNMB-induced synaptopathy.

Inhibition of CaMKIIα rescues GPNMB-induced neurobehavioral deficits in mice

We next sought to determine whether inhibition of CaMKIIα could rescue GPNMB-induced neurobehavioral deficits in MG^WT + DMSO, MG^GPNMB + DMSO, MG^WT + KN93, and MG^GPNMB + KN93 mice. Western blot analysis showed that expression of CREB1 and BDNF was significantly enhanced in the GPNMB-overexpressing microglia (MG^GPNMB + DMSO) group compared with the MG^WT + DMSO group. KN93, as CaMKIIα inhibitor, significantly inhibited GPNMB-induced upregulation of CREB1 and BDNF in MG^GPNMB + KN93 group compared with MG^GPNMB + DMSO group (Figure 6A-6D). These results suggest that CaMKIIα inhibition could rescue GPNMB-induced dysregulation of neuroprotective signaling. Behavioral tests further demonstrated that GPNMB overexpression induced anxiety-like behavior, motor deficits, and cognitive impairment in mice.

Figure 6 Inhibition of CaMKIIα ameliorates neurobehavioral deficits induced by microglial GPNMB overexpression. (A-D) CREB1 and BDNF expression: (A, B) Representative Western blots, (C) Relative CREB1 expression, (D) Relative BDNF expression. (E, F) Open field test: (E) Central times, (F) Explore distance. (G) Rotarod test: Percentage of mice remaining on rotarod. (H, I) Morris water maze navigation test (days 1–4): (H) Escape latency, (I) Explore distance. (J-L) Morris water maze probe test (day 5): (J) Platform crossing-free survival, (K) Crossing (frequency), (L) Percentage of time in the target quadrant. Data are presented as mean ± SEM (n=6/group for Western blot; n=15/group for behavioral tests). *, P<0.05; **, P<0.01; ***, P<0.001. MG^GPNMB + DMSO vs. MG^WT + DMSO group; ^###, P<0.001 MG^GPNMB + KN93 vs. MG^GPNMB + DMSO group. BDNF, brain-derived neurotrophic factor; CaMKIIα, calcium/calmodulin-dependent protein kinase II alpha; CREB1, camp response element-binding protein 1; DMSO, dimethyl sulfoxide; GPNMB, glycoprotein non-metastatic melanoma protein B; MG^GPNMB, microglia-specific GPNMB-overexpression mice; MG^WT, microglia-specific wild-type mice; SEM, standard error of the mean.

In the open field test, the time spent and the distance traveled in the central zone were significantly reduced in the MG^GPNMB + DMSO group compared with MG^WT + DMSO group. KN93 administration could partially reverse the deficits in the MG^GPNMB + KN93 group compared with the MG^GPNMB + DMSO group (Figure 6E,6F).

In the rotarod test, the MG^GPNMB + DMSO group exhibited significant motor coordination deficits, as shown by the decreased latency to fall, when compared with the MG^WT + DMSO group. KN93 administration significantly improved motor coordination in the MG^GPNMB + KN93 group compared with the MG^GPNMB + DMSO group (Figure 6G).

Morris water maze was used to test spatial learning and memory. In the navigation test, all groups showed learning progression. Compared with the MG^WT + DMSO group, the MG^GPNMB + DMSO group exhibited prolonged escape latencies. KN93 administration reduced escape latencies and swimming distance in the MG^GPNMB + KN93 group relative to the MG^GPNMB + DMSO group (Figure 6H,6I). In the probe test, the MG^GPNMB + DMSO group displayed longer first crossing latency, fewer platform crossings, and less time in the target quadrant compared with the MG^WT + DMSO group. These deficits were significantly ameliorated by KN93 administration, with the MG^GPNMB + KN93 group showing shorter first crossing latency, increased platform crossings, and increased time in the target quadrant relative to the MG^GPNMB + DMSO group (Figure 6J-6L).

These results demonstrate that inhibition of CaMKIIα activity ameliorates neurobehavioral abnormalities induced by microglial GPNMB overexpression, rescuing anxiety-like behavior, motor coordination, and spatial memory deficits through modulation of CREB1 and BDNF functional neural circuits.

CaMKIIα inhibition disrupts microglial GPNMB signaling and rescues functional deficits after HI brain injury

To determine whether CaMKIIα inhibition acts through the GPNMB signaling pathway to ameliorate HI injury, we examined four groups under HI conditions: HI + MG^WT + DMSO, HI + MG^WT + KN93, HI + MG^GPNMB + DMSO, and HI + MG^GPNMB + KN93.

Consistent with our earlier findings, HI + MG^GPNMB + DMSO mice showed elevated expression of p-CaMKIIα, t-CaMKIIα, PSD95, and SYN compared with HI + MG^WT + DMSO mice. Notably, KN93 administration reduced these proteins in both HI + MG^WT + KN93 and HI + MG^GPNMB + KN93 mice, with a more pronounced effect in the HI + MG^GPNMB + KN93 group (Figure 7A-7E). These results indicate that CaMKIIα inhibition effectively disrupts GPNMB-mediated CaMKIIα activation and downstream synaptic abnormalities.

Figure 7 CaMKIIα inhibition disrupts microglial GPNMB signaling and rescues functional deficits after HI brain injury. (A-E) Western blot analysis of p-CaMKIIα, t-CaMKIIα, PSD95, and SYN: (A) Representative Western blots, (B) Relative p-CaMKII expression, (C) Relative t-CaMKII expression, (D) Relative PSD95 expression, (E) Relative SYN expression. (F, G) Open field test: (F) Central times, (G) Explore distance. (H) Rotarod test: (H) Percentage of time remaining on rotarod. (I,J) Morris water maze navigation test (days 1–4): (I) Explore distance, (J) Escape latency. (K-M) Morris water maze probe test (day 5): (K) Platform crossing-free survival, (L) Crossing (frequency), (M) Percentage of time in the target quadrant. Data are presented as mean ± SEM (n=6/group for Western blot; n=10/group for behavioral tests). *, P<0.05; **, P<0.01; ***, P<0.001 HI + MG_WT + KN93 vs. HI + MG_WT + DMSO; ^###, P<0.001 HI + MG_GPNMB + KN93 vs. HI + MG_GPNMB + DMSO. CaMKIIα, calcium/calmodulin-dependent protein kinase II alpha; GPNMB, glycoprotein non-metastatic melanoma protein B; HI, hypoxia-ischemia; p-CaMKIIα, phosphorylated CaMKIIα; PSD95, postsynaptic density protein 95; SEM, standard error of the mean; SYN, synaptophysin; t-CaMKIIα, total CaMKIIα.

In the open field test, KN93 restored center time and distance in HI + MG^WT + KN93 mice compared with HI + MG^WT + DMSO controls, and partially rescued the more severe deficits in HI + MG^GPNMB + KN93 mice compared with HI + MG^GPNMB + DMSO mice (Figure 7F,7G). In the rotarod test, KN93 improved fall latency in both genotypes, with a more pronounced effect in HI + MG^GPNMB + KN93 mice (Figure 7H). In the Morris water maze test, KN93 reduced escape latency and swimming distance during the 4-day navigation phase in HI + MG^WT + KN93 mice compared with HI + MG^WT + DMSO controls. The more severe learning deficits in HI + MG^GPNMB + DMSO mice were similarly ameliorated by KN93 in HI + MG^GPNMB + KN93 mice (Figure 7I,7J). In the probe trial on day 5, KN93 reduced the first crossing latency, increased platform crossings, and increased time spent in the target quadrant in HI + MG^WT + KN93 mice compared with HI + MG^WT + DMSO controls. The more severe memory deficits in HI + MG^GPNMB + DMSO mice were similarly reversed by KN93 administration in HI + MG^GPNMB + KN93 mice (Figure 7K-7M).

Together, these results demonstrate that CaMKIIα inhibition with KN93 promotes functional recovery after perinatal HI injury and counteracts microglial GPNMB-CaMKIIα signaling. The efficacy of KN93 in both HI + MG^WT + KN93 and HI + MG^GPNMB + KN93 supports the conclusion that this signaling axis is a key pathogenic pathway in the endogenous response to perinatal HI injury.

Discussion

HI brain injury in preterm infants is a leading cause of irreversible neurological impairment, significantly elevating the risks of neurodevelopmental disorders and cognitive deficits (32,38). Effective treatments for the resulting sequelae remain scarce. This study systematically delineates how GPNMB, released by a distinct microglial subset, governs synaptic development and lifelong behavioral outcomes after perinatal HI brain injury. Our results not only confirm our previous finding that microglial GPNMB is specifically upregulated and induces an abnormal increase in synapse number after HI in immature mice, but also reveal the central role of the CaMKIIα-GluA1 signaling pathway in GPNMB-mediated synaptic developmental disorders and neurobehavioral deficits. Specifically, in the perinatal HI model, a discrete population of disease-associated microglia upregulates GPNMB, which in turn potentiates CaMKIIα phosphorylation. Consequently, the aberrant trafficking and anchoring of postsynaptic GluA1 receptors derail synaptic maturation, culminating in lasting neurobehavioral deficits. Moreover, selective CaMKIIα inhibition in mice overexpressing microglial GPNMB fully restored GluA1 abundance and synaptic localization, reversed synaptic developmental defects and erased the associated neurobehavioral impairments. Importantly, KN93 administration also significantly attenuated molecular and behavioral deficits in wild-type HI mice, demonstrating that the GPNMB-CaMKIIα axis is functionally engaged in the endogenous response to perinatal HI injury. Accordingly, we present a unified model (Figure 8) in which HI-triggered, microglia-derived GPNMB hijacks the CaMKIIα-GluA1 signaling axis to derail synaptogenesis and imprint enduring neurobehavioral dysfunction.

Figure 8 Schematic diagram of the proposed mechanism. HI injury upregulates microglial GPNMB, which activates neuronal CaMKIIα and increases GluA1 expression, leading to synaptic and behavioral deficits. Inhibition of CaMKIIα disrupts this pathway and promotes functional recovery after HI injury. CaMKIIα, calcium/calmodulin-dependent protein kinase II alpha; GluA1, glutamate receptor A1; GPNMB, glycoprotein non-metastatic melanoma protein B; TARP, transmembrane AMPA receptor regulatory protein; HI, hypoxia-ischemia.

Encephalopathy of prematurity is typically characterized by cerebral white matter injury (WMI); however, it is often accompanied by gray matter injury (5,39). Most studies have concentrated on WMI pathophysiology, such as oligodendrocyte maturation impairment and dysmyelination. In recent years, however, accumulating evidence has suggested that the effects of WMI are not limited to the cerebral white matter. Beyond pure WMI, WMI may involve structural abnormalities in gray matter, including neuronal migration disturbance, synaptic development abnormalities, and neural network function dysregulation in cortical and hippocampal areas (4,40). These changes might represent a fundamental pathological mechanism for long-term clinical impairments. In the immature brain, HI injury might impair the coordinated development of white matter and gray matter and lead to dysfunctional neural circuit remodeling (41,42). Clinical neuroimaging studies have shown abnormal cortical thickness and altered functional connectivity in the preterm encephalopathy (40). Thus, it is likely that synaptic dysgenesis substantially contribute to long-term neurological sequelae after preterm brain injury.

Our study shows that perinatal HI injury in the immature mouse brain induces an abnormal early increase in synaptic development, characterized by a synapse number surplus, a structural phenomenon of excessive synaptic protein accumulation that is distinct from synaptic dysfunction, which refers to functional impairment in synaptic transmission and plasticity. This early synaptic surplus may appear contradictory to many previous reports from neonatal models of HI, which show synaptic loss (43). We hypothesize that this divergence from some previous reports may stem from methodological variations, such as differences in the HI models used, the post-injury time points examined, or the specific brain regions assessed. In addition, it more likely reflects previously unrecognized pathophysiological complexity. We propose that the net effect of synaptic loss vs. surplus following HI may depend on the pattern of microglial activation and resulting predominant underlying molecular mechanism recruited by the insult. Specifically, when a pro-inflammatory microglial response dominates, robust phagocytic pruning may outcompete other processes and result in net elimination of synapses. In contrast, our results show that a distinct subset of activated microglia, identified by high GPNMB expression, orchestrates a distinct pathologic program. GPNMB-positive microglia perturb the balance of synapse formation and elimination by simultaneously impairing phagocytosis-dependent pruning and directly promoting excess synaptogenesis. Thus, the increase in number at an early stage, rather than representing a compensatory gain, reflects a maladaptive gain that drives subsequent abnormal and inefficient neural networks. This sequence of events provides an explanation for why the resulting early developmental disruption results in permanent neurobehavioral deficits despite a normalization of synaptic density by adulthood. Therefore, our findings redefine a key aspect of synaptic pathology after preterm HI injury, highlighting that the core problem is not solely synaptic loss, but rather synaptic misdirection initiated by a GPNMB-dependent mechanism.

Previous studies have shown that microglia are involved in synaptic remodeling in several ways, such as complement-dependent synaptic pruning at early stages of development and the secretion of neurotrophic factors like BDNF to regulate synaptic homeostasis in mature circuits (44,45). In the case of WMI, obvious activation of microglia is displayed, and the degree of microglial activation is associated with neurodevelopmental outcomes. Exuberantly activated microglia promote neuroinflammation and interfere with repair probably by releasing pro-inflammatory cytokines (e.g., IL-1β, TNF-α), excessive amounts of synaptic pruning and oligodendrocyte damage, and then cause motor and cognitive deficits (46). We previously found that GPNMB was highly expressed in immature mice after HI brain injury and labeled typical DAM, a subtype of activated microglia. In the present study, using a microglia-specific GPNMB overexpression model, we found that upregulated GPNMB aggravated synaptic developmental abnormalities and induced long-term neurobehavioral deficits after perinatal HI injury.

GPNMB has been reported to interact with α-synuclein and regulate synaptic integrity after α-synuclein overexpression in rats (30). However, the molecular mechanisms underlying the regulation of GPNMB on synaptic development and its region-specific effects on WMI remain unclear. CaMKIIα is a key regulator of neuronal signaling and synaptic plasticity. It can induce changes in neuronal morphogenesis during dendritic spine morphogenesis and CREB-BDNF pathway activation (47,48) and has been reported to be involved in cognitive diseases such as schizophrenia and ADHD. CaMKIIα also has specific functions in certain brain regions, such as the mPFC, which is related to working memory (49,50). Building upon these established roles, we found that microglia-derived GPNMB can greatly enhance CaMKIIα activation in the hippocampus and cortex after HI damage. Consistent with this, inhibition of CaMKIIα using KN93 significantly reversed HI-induced synaptic abnormalities and behavioral deficits in wild-type mice, further supporting that CaMKIIα is a critical downstream effector in the endogenous HI response. In neuroinflammatory and depressive models, our results are consistent with recent reports that CaMKIIα inhibition can correct synaptic dysfunction and cognitive deficits (51). Interestingly, CaMKIIα heterozygous knockout mice also exhibit structural abnormalities in the dentate gyrus and schizophrenia-like behaviors (48) correlated with decreased phosphorylation of CaMKIIα. As an essential subunit of AMPA receptors, GluA1 participates in excitotoxic signaling due to its calcium permeability (52). Under HI conditions, GluA1 is overactivated at the synaptic level, leading to enhanced calcium influx, and induction of CaMKIIα phosphorylation at Thr286. Then, activated CaMKIIα can further phosphorylate GluA1 at Ser831 and enhance channel activity in a positive feedback loop, which further aggravates calcium overload, mitochondrial dysfunction, and neuronal apoptosis (25,53). Our present study demonstrates that p-CaMKIIα interacts with GluA1 via multiple protein-protein interaction pockets including electrostatic and hydrogen-bonding interactions. Furthermore, we found that microglial GPNMB upregulated GluA1 expression in cortical and hippocampal areas, which was abolished by CaMKIIα inhibition. Interestingly, GPNMB overexpression increased the interaction between postsynaptic GluA1 and PSD95 in the hippocampal CA1, CA3 and DG subregions, an effect that was specifically reversed via KN93. Recently, it has been reported that CaMKIIα phosphorylation of AMPARs is involved in AMPAR internalization and dendritic spine retraction (54). Therefore, we propose that the CaMKIIα-GluA1 axis mediates the effects of microglial GPNMB on synaptic dysgenesis. This finding provides a new theoretical basis for understanding long-term neurological deficits after perinatal HI injury and suggests potential targets for intervention.

An important finding from our study is that KN93 administration significantly ameliorated synaptic and behavioral deficits in both wild-type HI mice and GPNMB-overexpressing HI mice. The effect was more pronounced in GPNMB-overexpressing mice, consistent with a dose-response relationship in which elevated GPNMB drives greater CaMKIIα activation. These results collectively demonstrate that the GPNMB-CaMKIIα signaling axis is functionally engaged in the endogenous response to perinatal HI injury, and that pharmacological inhibition of CaMKIIα can disrupt this pathway to promote functional recovery. The enhanced efficacy in GPNMB-overexpressing mice further supports the specificity of this mechanism by showing that pathway hyperactivation increases susceptibility to KN93-mediated CaMKIIα inhibition.

Although our study demonstrated an important role for microglia-derived GPNMB in regulating synaptic development via CaMKIIα-GluA1 axis and in inducing long-term neurobehavioral deficits, several limitations should be acknowledged. First, neonatal HI injury in mice may cause visual impairment, which could confound the interpretation of vision-dependent behavioral tests such as the Morris water maze. Future work incorporating vision-independent behavioral assays would help address this concern. Second, in our transgenic mouse model, MG^WT control mice (Cx3cr1^WT/WT) and MG^GPNMB mice (Cx3cr1^CreERT2/WT) differ in Cx3cr1 copy number. While independent studies have shown that Cx3cr1 knockout does not affect synaptic plasticity in the primary visual cortex under basal conditions (55,56), we cannot exclude the possibility that Cx3cr1 haploinsufficiency may influence some phenotypes under HI injury conditions. Future studies using Cre-negative, Cx3cr1-heterozygous littermates as controls are warranted. Third, manufacturer validation data indicate detectable Cre activity in the Cx3cr1-CreERT2 line (Cat# NM-KI-200157) in the absence of tamoxifen. Importantly, MG^WT control mice lack the Cre transgene and are therefore unaffected by leakage. Any background recombination in MG^GPNMB mice would only underestimate group differences, making our findings more conservative. This line has been successfully used in published microglial studies (57). Finally, whether microglia-specific knockdown of GPNMB rescues synaptic development through modulation of the CaMKIIα pathway remains to be investigated, which would provide more definitive evidence for the cell-autonomous role of microglial GPNMB.

Conclusions

In summary, we demonstrate that microglial GPNMB is a critical driver of developmental delay and persistent neurobehavioral deficits after perinatal HI injury. In particular, we identify the CaMKIIα-GluA1 axis as the key conduit through which GPNMB disrupts synaptic AMPAR trafficking and function. Inhibition of CaMKIIα using KN93 ameliorates synaptic and behavioral deficits in both wild-type and GPNMB-overexpressing HI mice, with a more pronounced effect in the latter group. These findings identify the GPNMB-CaMKIIα-GluA1 axis as a pathogenic pathway and a potential target for intervention in preterm brain injury. To confirm these mechanisms across models and investigate the effects of GPNMB-mediated synaptic modulation in particular regions, more research is required.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the ARRIVE reporting checklist. Available at https://tp.amegroups.com/article/view/10.21037/tp-2025-1-922/rc

Data Sharing Statement: Available at https://tp.amegroups.com/article/view/10.21037/tp-2025-1-922/dss

Peer Review File: Available at https://tp.amegroups.com/article/view/10.21037/tp-2025-1-922/prf

Funding: This work was supported by National Natural Science Foundation of China (grant number: 82001596).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tp.amegroups.com/article/view/10.21037/tp-2025-1-922/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All experimental procedures were approved by the Animal Care and Use Committee of Children’s Hospital of Fudan University (approval No. 2020239) and were conducted in accordance with the National Institutes of Health (NIH) Guidelines for the Care and Use of Laboratory Animals.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Perin J, Mulick A, Yeung D, et al. Global, regional, and national causes of under-5 mortality in 2000-19: an updated systematic analysis with implications for the Sustainable Development Goals. Lancet Child Adolesc Health 2022;6:106-15. [Crossref] [PubMed]
2. Cheong JLY, Mainzer RM, Doyle LW, et al. Neurodevelopment at Age 9 Years Among Children Born at 32 to 36 Weeks' Gestation. JAMA Netw Open 2024;7:e2445629. [Crossref] [PubMed]
3. Charkaluk ML, Kana GD, Benhammou V, et al. Neurodevelopment at age 5.5 years according to Ages & Stages Questionnaire at 2 years' corrected age in children born preterm: the EPIPAGE-2 cohort study. Arch Dis Child Fetal Neonatal Ed 2024;109:519-26. [Crossref] [PubMed]
4. Buser JR, Maire J, Riddle A, et al. Arrested preoligodendrocyte maturation contributes to myelination failure in premature infants. Ann Neurol 2012;71:93-109. [Crossref] [PubMed]
5. Back SA. White matter injury in the preterm infant: pathology and mechanisms. Acta Neuropathol 2017;134:331-49. [Crossref] [PubMed]
6. Motavaf M, Piao X. Oligodendrocyte Development and Implication in Perinatal White Matter Injury. Front Cell Neurosci 2021;15:764486. [Crossref] [PubMed]
7. Lear BA, Lear CA, Dhillon SK, et al. Evolution of grey matter injury over 21 days after hypoxia-ischaemia in preterm fetal sheep. Exp Neurol 2023;363:114376. [Crossref] [PubMed]
8. Schneider J, Miller SP. Preterm brain Injury: White matter injury. Handb Clin Neurol 2019;162:155-72. [Crossref] [PubMed]
9. Prasad JD, van de Looij Y, Gunn KC, et al. Long-term coordinated microstructural disruptions of the developing neocortex and subcortical white matter after early postnatal systemic inflammation. Brain Behav Immun 2021;94:338-56. [Crossref] [PubMed]
10. McClendon E, Chen K, Gong X, et al. Prenatal cerebral ischemia triggers dysmaturation of caudate projection neurons. Ann Neurol 2014;75:508-24. [Crossref] [PubMed]
11. Ziabska K, Gewartowska M, Frontczak-Baniewicz M, et al. The Impact of the Histone Deacetylase Inhibitor-Sodium Butyrate on Complement-Mediated Synapse Loss in a Rat Model of Neonatal Hypoxia-Ischemia. Mol Neurobiol 2025;62:5216-33. [Crossref] [PubMed]
12. Vallés AS, Barrantes FJ. Dendritic spine membrane proteome and its alterations in autistic spectrum disorder. Adv Protein Chem Struct Biol 2022;128:435-74. [Crossref] [PubMed]
13. Yang Y, Liu JJ. Structural LTP: Signal transduction, actin cytoskeleton reorganization, and membrane remodeling of dendritic spines. Curr Opin Neurobiol 2022;74:102534. [Crossref] [PubMed]
14. Guo H, Ali T, Que J, et al. Dendritic spine dynamics in associative memory: A comprehensive review. FASEB J 2023;37:e22896. [Crossref] [PubMed]
15. Yang G, Pan F, Gan WB. Stably maintained dendritic spines are associated with lifelong memories. Nature 2009;462:920-4. [Crossref] [PubMed]
16. Pelucchi S, Da Dalt L, De Cesare G, et al. Neuronal PCSK9 regulates cognitive performances via the modulation of ApoER2 synaptic localization. Pharmacol Res 2025;213:107652. [Crossref] [PubMed]
17. Yang F, You H, Mizui T, et al. Inhibiting proBDNF to mature BDNF conversion leads to ASD-like phenotypes in vivo. Mol Psychiatry 2024;29:3462-74. [Crossref] [PubMed]
18. Howes OD, Onwordi EC. The synaptic hypothesis of schizophrenia version III: a master mechanism. Mol Psychiatry 2023;28:1843-56. [Crossref] [PubMed]
19. Penzes P, Cahill ME, Jones KA, et al. Dendritic spine pathology in neuropsychiatric disorders. Nat Neurosci 2011;14:285-93. [Crossref] [PubMed]
20. Yasuda R, Hayashi Y, Hell JW. CaMKII: a central molecular organizer of synaptic plasticity, learning and memory. Nat Rev Neurosci 2022;23:666-82. [Crossref] [PubMed]
21. Lisman J, Schulman H, Cline H. The molecular basis of CaMKII function in synaptic and behavioural memory. Nat Rev Neurosci 2002;3:175-90. [Crossref] [PubMed]
22. Bemben MA, Shipman SL, Hirai T, et al. CaMKII phosphorylation of neuroligin-1 regulates excitatory synapses. Nat Neurosci 2014;17:56-64. [Crossref] [PubMed]
23. Soleimanpour E, Bergado Acosta JR, Landgraf P, et al. Regulation of CREB Phosphorylation in Nucleus Accumbens after Relief Conditioning. Cells 2021;10:238. [Crossref] [PubMed]
24. Gillett DA, Wallings RL, Uriarte Huarte O, et al. Progranulin and GPNMB: interactions in endo-lysosome function and inflammation in neurodegenerative disease. J Neuroinflammation 2023;20:286. [Crossref] [PubMed]
25. Bessières B, Jia M, Travaglia A, et al. Developmental changes in plasticity, synaptic, glia, and connectivity protein levels in rat basolateral amygdala. Learn Mem 2019;26:436-48. [Crossref] [PubMed]
26. Cong L, Ding S, Guo Y, et al. Acupuncture alleviates CSDS-induced depressive-like behaviors by modulating synaptic plasticity in vCA1. Theranostics 2025;15:4808-22. [Crossref] [PubMed]
27. Saade M, Araujo de Souza G, Scavone C, et al. The Role of GPNMB in Inflammation. Front Immunol 2021;12:674739. [Crossref] [PubMed]
28. Han B, Bao MY, Sun QQ, et al. Nuclear receptor PPARγ targets GPNMB to promote oligodendrocyte development and remyelination. Brain 2025;148:1801-16. [Crossref] [PubMed]
29. Rose AA, Annis MG, Dong Z, et al. ADAM10 releases a soluble form of the GPNMB/Osteoactivin extracellular domain with angiogenic properties. PLoS One 2010;5:e12093. [Crossref] [PubMed]
30. Diaz-Ortiz ME, Seo Y, Posavi M, et al. GPNMB confers risk for Parkinson's disease through interaction with α-synuclein. Science 2022;377:eabk0637. [Crossref] [PubMed]
31. Percie du Sert N, Hurst V, Ahluwalia A, et al. The ARRIVE guidelines 2.0: Updated guidelines for reporting animal research. PLoS Biol 2020;18:e3000410. [Crossref] [PubMed]
32. Chen Q, Zhang K, Wang M, et al. A translational mouse model for investigation of the mechanism of preterm diffuse white matter injury. Transl Pediatr 2022;11:1074-84. [Crossref] [PubMed]
33. Tang XH, Zhang GF, Xu N, et al. Extrasynaptic CaMKIIα is involved in the antidepressant effects of ketamine by downregulating GluN2B receptors in an LPS-induced depression model. J Neuroinflammation 2020;17:181. [Crossref] [PubMed]
34. Wang Y, Sun B, Shibata B, et al. Transmission electron microscopic analysis of myelination in the murine central nervous system. STAR Protoc 2022;3:101304. [Crossref] [PubMed]
35. Crawley JN. Behavioral phenotyping strategies for mutant mice. Neuron 2008;57:809-18. [Crossref] [PubMed]
36. Deacon RM. Measuring motor coordination in mice. J Vis Exp 2013;e2609.
37. Vorhees CV, Williams MT. Morris water maze: procedures for assessing spatial and related forms of learning and memory. Nat Protoc 2006;1:848-58. [Crossref] [PubMed]
38. Volpe JJ. Dysmaturation of Premature Brain: Importance, Cellular Mechanisms, and Potential Interventions. Pediatr Neurol 2019;95:42-66. [Crossref] [PubMed]
39. Volpe JJ. Iron and zinc: Nutrients with potential for neurorestoration in premature infants with cerebral white matter injury. J Neonatal Perinatal Med 2019;12:365-8. [Crossref] [PubMed]
40. Dean JM, McClendon E, Hansen K, et al. Prenatal cerebral ischemia disrupts MRI-defined cortical microstructure through disturbances in neuronal arborization. Sci Transl Med 2013;5:168ra7. [Crossref] [PubMed]
41. Ball G, Pazderova L, Chew A, et al. Thalamocortical Connectivity Predicts Cognition in Children Born Preterm. Cereb Cortex 2015;25:4310-8. [Crossref] [PubMed]
42. Diedrichsen J, Yokoi A, Arbuckle SA. Pattern component modeling: A flexible approach for understanding the representational structure of brain activity patterns. Neuroimage 2018;180:119-33. [Crossref] [PubMed]
43. Liu N, Tong X, Huang W, et al. Synaptic Injury in the Thalamus Accompanies White Matter Injury in Hypoxia/Ischemia-Mediated Brain Injury in Neonatal Rats. Biomed Res Int 2019;2019:5249675. [Crossref] [PubMed]
44. Schafer DP, Lehrman EK, Kautzman AG, et al. Microglia sculpt postnatal neural circuits in an activity and complement-dependent manner. Neuron 2012;74:691-705. [Crossref] [PubMed]
45. Miyamoto A, Wake H, Ishikawa AW, et al. Microglia contact induces synapse formation in developing somatosensory cortex. Nat Commun 2016;7:12540. [Crossref] [PubMed]
46. Verney C, Pogledic I, Biran V, et al. Microglial reaction in axonal crossroads is a hallmark of noncystic periventricular white matter injury in very preterm infants. J Neuropathol Exp Neurol 2012;71:251-64. [Crossref] [PubMed]
47. Claiborne N, Anisimova M, Zito K. Activity-Dependent Stabilization of Nascent Dendritic Spines Requires Nonenzymatic CaMKIIα Function. J Neurosci 2024;44:e1393222023. [Crossref] [PubMed]
48. Robison AJ. Emerging role of CaMKII in neuropsychiatric disease. Trends Neurosci 2014;37:653-62. [Crossref] [PubMed]
49. Liu XB, Murray KD. Neuronal excitability and calcium/calmodulin-dependent protein kinase type II: location, location, location. Epilepsia 2012;53:45-52. [Crossref] [PubMed]
50. Ge C, Chen W, Zhang L, et al. Chemogenetic activation of the HPC-mPFC pathway improves cognitive dysfunction in lipopolysaccharide -induced brain injury. Theranostics 2023;13:2946-61. [Crossref] [PubMed]
51. Niu Y, Dai Z, Liu W, et al. Ablation of SNX6 leads to defects in synaptic function of CA1 pyramidal neurons and spatial memory. Elife 2017;6:e20991. [Crossref] [PubMed]
52. Tran L, Keele NB. CaMKIIα knockdown decreases anxiety in the open field and low serotonin-induced upregulation of GluA1 in the basolateral amygdala. Behav Brain Res 2016;303:152-9. [Crossref] [PubMed]
53. Jalan-Sakrikar N, Bartlett RK, Baucum AJ 2nd, et al. Substrate-selective and calcium-independent activation of CaMKII by α-actinin. J Biol Chem 2012;287:15275-83. [Crossref] [PubMed]
54. Kim S, Sohn S, Choe ES. Phosphorylation of GluA1-Ser831 by CaMKII Activation in the Caudate and Putamen Is Required for Behavioral Sensitization After Challenge Nicotine in Rats. Int J Neuropsychopharmacol 2022;25:678-87. [Crossref] [PubMed]
55. Schecter RW, Maher EE, Welsh CA, et al. Experience-Dependent Synaptic Plasticity in V1 Occurs without Microglial CX3CR1. J Neurosci 2017;37:10541-53. [Crossref] [PubMed]
56. Lowery RL, Tremblay ME, Hopkins BE, et al. The microglial fractalkine receptor is not required for activity-dependent plasticity in the mouse visual system. Glia 2017;65:1744-61. [Crossref] [PubMed]
57. Guan X, Zhu S, Song J, et al. Microglial CMPK2 promotes neuroinflammation and brain injury after ischemic stroke. Cell Rep Med 2024;5:101522. [Crossref] [PubMed]