Targeting microglial NAAA-regulated PEA signaling counters inflammatory damage and symptom progression of post-stroke anxiety

Abstract

Post-stroke anxiety (PSA) manifests as anxiety symptoms after stroke, with unclear mechanisms and limited treatment strategies. Endocannabinoids, reported to mitigate fear, anxiety, and stress, undergo dynamic alterations after stroke linked to prognosis intricately. However, endocannabinoid metabolism in ischemic microenvironment and their associations with post-stroke anxiety-like behavior remain largely uncovered. Our findings indicated that endocannabinoid metabolism was dysregulated after stroke, characterized by elevated N-palmitoylethanolamide (PEA) hydrolase N-acylethanolamine-acid amidase (NAAA) in activated microglia from ischemic area, accompanied by rapid PEA exhaustion. Microglial PEA metabolite exhaustion is directly associated with more severe pathological damage, anxiety symptoms and pain sensitivity. Naaa knockout or pharmacological supplementation to boost PEA pool content can effectively promote stroke recovery and alleviate anxiety-like behaviors. In addition, maintaining PEA pool content in ischemic area reduces overactivated microglia by confronting against mitochondria dysfunction and inflammasome cascade triggered IL-18 release and diffusion to contralateral hemisphere. Meanwhile, maintenance of microglial PEA pool content in ischemic-damaged lesion can preserve contralateral vCA1 synaptic integrity, enhancing anxiolytic pBLA-vCA1^Calb1+ circuit activity by alleviating microglial phagocytosis-mediated synaptic loss. Thus, we conclude that microglial NAAA-regulated lipid signaling in the ischemic focus remodels contralateral anxiolytic circuit to participate in post-stroke anxiety progression. Blocking PEA signaling breakdown promotes stroke recovery and mitigates anxiety-like symptoms.

Graphical Abstract

Microglial NAAA-regulated lipid signaling involves with post-stroke anxiety.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12964-025-02202-2.

Highlights

1. PEA metabolite is exhausted in microglia after ischemic stroke.

2. NAAA signaling activation exacerbates ischemic brain injury and abnormal behavior.

3. PEA metabolism represses IL-18 inflammatory cascade mediated contralateral vCA1 circuit remodeling.

4. Boosting PEA metabolite mitigates anxiety-like behavior and improves function recovery after stroke.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12964-025-02202-2.

Introduction

Stroke has remained the second leading cause of death worldwide. With the advancement and widespread implementation of population-wide primary and secondary stroke prevention efforts, along with acute intervention strategies, the incidence and mortality rates of stroke have significantly decreased [1]. However, post-stroke anxiety (PSA), which is usually ignored, still burdens 1/4 of stroke patients, leading to a serious decline in quality of life [2]. The etiology of PSA is considered multifactorial, including biological factors and psychosocial factors. And effective intervention strategies are limited because of obscure pathophysiological mechanisms. Recently, amino acid metabolism, lipid metabolism, endocannabinoid metabolism are all reported to be associated with post-stroke negative emotions [3–6]. Endocannabinoid signaling involves various classical cannabinoid receptors (e.g. CB1, CB2) and their endogenous ligands such as 2-arachidonoylglycerol (2-AG), arachidonoylethanolamide (AEA), N-oleoylethanolamide (OEA) and N-palmitoylethanolamide (PEA), which are closely correlated with the severity of acute ischemic stroke. Clinical studies indicated that serum endocannabinoid levels undergo dynamic changes post stroke, mainly exhibiting a decreased trend except 2-AG [6]. Moreover, endocannabinoids have been reported to guard against fear, anxiety and stress by activating their respective receptors or modulating glial cell function and neural circuits [7]. So, we firstly speculated that the exhaustion of endocannabinoids may involve the development of PSA, though this assumption lacks substantial evidence.

PEA, AEA, and OEA share some common metabolism pathways, involving N-acyltransferases 1 and 2 (NAT1/2) and N-arachidonyl-phosphatidylethanolamine phospholipase D (NAPE-PLD) in their biosynthesis, N-acylethanolamine-hydrolysing-acid amidase (NAAA) and fatty acid amide hydrolase (FAAH) in their degradation [8, 9]. When the body is challenged by pathological stress including inflammation, neuronal damage, and pain [10], PEA metabolism is notably dysregulated, characterized by an increased expression of endocannabinoid hydrolases, potentially correlating with the severity of disease outcomes [11]. PEA is metabolized by various cell types [12], especially in monocytes and macrophages [10]. Notably, PEA acts as a PPAR-α agonist to inhibit NLRP3 inflammasome activation [13], and attenuate macrophage M1 polarization [14]. Microglia are similar in morphology and function to macrophages, whether endocannabinoids are exhausted in activated microglia from ischemic lesion is not clear. Microglia play a key role in modulating anxiety-like behaviors. Specifically, Yang et al. demonstrated that ischemic damage triggers contralateral microcircuit remodeling, directly driving anxiety-like behaviors by microglia regulating synaptic connectivity in the vCA1-pBLA circuit. Critically, this process is exacerbated by interleukin- 18 (IL- 18), a classical downstream final effector of neuroinflammation [15], promoting synaptic pruning via enhanced microglial phagocytosis in contralateral vCA1 [16]. Ischemic stroke features with hypoxia-induced pathological stress. But endocannabinoid metabolism in ischemic microenvironment is largely uncovered. Neuronal circuits for anxiety involved with multiple brain areas such as hippocampus, medial prefrontal cortex (mPFC) and amygdala [17–19]. However, whether and how endocannabinoid metabolism in damaged lesions is involved in activated microglia-mediated contralateral microcircuit remodeling to influence anxiety-like behavior remains elusive.

Here, we discovered endocannabinoid metabolism is dysregulated mainly in ischemic injury foci after stroke. Moreover, endocannabinoids’ hydrolases were screened correspondingly and found that PEA hydrolase NAAA mainly elevated in activated microglia from ischemic area, accompanied by PEA rapid exhaustion as the course of ischemic stroke progresses. Microglial PEA exhaustion is directly associated with more severe negative emotions such as anxiety and pain sensitivity. Next, boosting PEA metabolism with Naaa knockout or pharmacologic supply effectively promotes stroke recovery and alleviates PSA progression. Maintaining PEA content in the ischemic area brakes overactivated microglia by confronting against mitochondria dysfunction and inflammasome cascade. Based on Yang’s research, we further found elevating PEA content in ischemic damaged lesion can safeguard contralateral vCA1 synaptic integrity, maintaining the activity of pBLA-vCA1 ^Calb1+ circuit by relieving microglial phagocytosis mediated synaptic loss via inhibiting ischemic damaged lesion derived IL- 18. Thus, our study displays mechanistic and functional insights into the unseen effects of microglial PEA metabolite exhaustion during the process of PSA, and therefore provides significant clinical implications for early anxiety assessment and treatment after stroke.

Materials and methods

Animal

In this study, both 6-week-old and 18-month-old wild-type C57BL/6 mice were obtained from GemPharmatech Co., Ltd. (Nanjing, China). Naaa knockout (Naaa^−/−) mice were purchased from Cyagen Biosciences Inc. (Guangzhou, China). All types of mice were randomly grouped and housed five per cage with ad libitum feeding in a specific-pathogen-free (SPF) level animal facility, where a constant temperature (25 ± 1℃) and humidity (50 ± 5%) were maintained, with a regular 7:00 ~ 19:00 light–dark cycle. Possible measures were taken to minimize pain and discomfort in the animals. All experiments complied with standards established by the Institutional Animal Care and Use Committee of USTC and were approved by the Experimental Animal Center of the First Affiliated Hospital of USTC (No. 2022-N(A)− 175).

Modeling

Photothrombotic stroke model (PTS)

After mice were anesthetized, the right skull overlying a small section (AP, − 0.3 ~ − 2.3 mm; ML, 0.5 ~ 3.0 mm) was ground to transparency with a dental drill while leaving the dura intact. Five min after single intraperitoneal injection of Rose Bengal (Sigma Chemical Co., St. Louis, MO, USA) solution (100 mg/kg), the exposure window was treated with a 150 W halogen light source (Nanjing Hecho Technology Co., Ltd., China) for 10 min closely. Mice in sham group underwent the same operation except an extra illumination. Then, the incision was closed with 5–0 silk sutures and mice were allowed to recover in the home cage.

Post-stroke anxiety model (PSA)

On the 1 st to 3rd day after PTS modeling, the marble burying test was conducted once daily (steps are shown in “ Behavioral tests”), and the number of buried marbles was recorded. Mice with results of 10 or more on all three tests were classified as the post-stroke anxiety (PSA) group, while those with < 5 marbles were designated as PTS. PTS and PSA groups were housed separately.

After screening, PSA and PTS mice were randomly selected in a 1:1 ratio using a computerized random number generator, ensuring equal sample sizes for subsequent experiment. This randomization process eliminated systematic differences between groups unrelated to anxiety phenotypes rather than arbitrary exclusion. Moreover, our supplementary data confirm that Naaa^-/- mice do not exhibit basal anxiety (Figure S2A ~ K), allowing us to apply the same MBT screening method in Naaa^−/− mice.

Laser speckle imaging

Laser speckle imaging system (RFLSIIII, RWD Life Technology Co.,China) was used to monitor regional cerebral blood flow (CBF) before ischemia and 10 min after PTS. The CBF decreased by at least 80% relative to the contralateral baseline is indicated the success of artery occlusion.

Magnetic resonance imaging (MRI)

All MRI experiments were performed on a 9.4 Tesla/30-cm-diameter-bore magnet imaging system (uMR 9.4 T, United Imaging Life Science Instrument, P.R. China) with a maximum gradient strength at 1000 mT/m and a slew rate up to 10,000 T/m/s. An 86-mm volume coil was used for transmission and a 15-mm planar surface coil was used for reception. After anesthetized with 3% isoflurane in an induction chamber, mice were positioned on an MRI-compatible cradle equipped with a face mask connected to the ventilator with 1.5% oxygenated isoflurane. The body temperature was kept at 36 ~ 37℃ using a recirculating water bath. Respiratory frequency and heart beat were continuously monitored during scanning. The T2-weighted high-resolution multi-slice fast spin echo (FSE) sequence parameters was set as follows. TR = 3000 ms, TE = 49 ms, ETL = 19, BW = 300 Hz/pixel, field of view: 18 × 20 mm², matrix size = 331 × 368, slice thickness = 0.5 mm, slice gap = 0 mm, 20 slices were scanned to encompass the entire brain. Number of averages = 4, total acquisition time = 15 min. Image analysis and 3D-reconstruction were performed using open-source ITK-SNAP software (http://www.itksnap.org).

Drug administration

Intraperitoneal injection (i.p.)

According to the instruction, Palmitoylethanolamide (TargetMol, China) was dissolved in vehicle contained 10%DMSO, 40%PEG, 5%Tween80,and 45% saline with ultrasonic breaking. PEA (10 mg/kg [20, 21]) or vehicle were administered intraperitoneally once per day for 14 consecutive days post-surgery. The behavioral tests were performed 6 h after injection to avoid potential effects.

Intra-cerebroventricular injection (i.c.v)

After completion of photothrombotic stroke model, a stainless-steel guide cannula (O.D. 0.41 mm, RWD, China) was then implanted above the contralateral vCA1 (AP, − 3.28 mm; ML, − 3.3 mm; DV, − 4.6 mm) through a small hole drilled on the skull and fixed with dental cement firmly. An injection cannula (O.D. 0.21 mm, RWD, China) connected to a 10-μL syringe via a polyethylene (PE) tubing was inserted into the guide cannula. Interleukin- 18 binding protein (IL- 18BP, 1 μg/kg, #HY-P75841, MCE, USA) reconstituted in saline (1000 μg/mL) was injected at a rate of 50 nL/min.

Behavioral tests

Before all experiments, mice were moved to the apparatus in behavioral testing room to habituate in advance at least 3 days, for 1 h each day. All tests were performed under 20 lux brightness condition with constant temperature and humidity during the light cycle.

Marble burying test (MBT)

Irradiated, unscented corn-cob granules were added into polycarbonate cages (40 cm × 25 cm × 15 cm) as the bedding material (5-cm-thick). Fifteen black-glass-marbles (five rows of three, each 4 cm apart) were evenly spaced across the bedding. Mice were gently placed into the cage and allowed to explore undisturbedly for 30 min. Marbles covered by bedding more than two-third were considered as buried.

Open-field test (OFT)

The OFT apparatus consisted of a 50 × 50 × 50 cm³ square box made of opaque white acrylic. Each mouse was gently placed in the central zone (25 × 25 cm) of the open field and allowed to move freely for 5 min. After the activity video recorded by an overhead camera, total distance traveled and time spent in center were tracked and analyzed with Ethovision XT 14 (Noldus, The Netherlands). The apparatus was respectively cleaned by 75% alcohol and clean water to avoid odor cues after each trail.

Elevated plus-maze test (EPM)

The elevated plus maze (60 cm above floor), made of white polypropylene, consisted of two closed arms (35 × 5 × 10 cm³)and two open arms (35 × 5 × 0.5 cm³) connected by a center platform forming a cross. The head and the upper-body of mice were removed in the center of the maze facing an open arm at random. All behaviors of freely moving mice were recorded for 5 min with a video camera. Ethovision XT 14 software was used to automatically track and quantify the time spent in open arms.

Rotarod test

Mice were tested on the rod of rotarod apparatus (Shanghai Xinruan Information Technology Co., Ltd) which was programmed to accelerate uniformly from 4 rpm up to 40 rpm over 5 min. Each mouse received three trials 30 min apart and the latency to fall off was recorded for further analysis as a measurement of motor coordination with a maximum time of 600 s.

Von Frey filament test

Each mouse was individually covered with a plastic ventilated cubicle placed on a wire mesh platform 50 cm above ground. After a 30-min regular acclimatization to minimize stress on the testing day, mechanical allodynia was measured by a series of von Frey filaments (Stoelting Inc., US) in ascending order (0.02, 0.04, 0.07, 0.16, 0.4 and 0.6 g) which were used to vertically stimulate the mid-plantar surface of right hind paw for two seconds. Briefly, the average force that induced positive responses (fast paw withdrawal, flinching, or licking) three times was defined as the paw withdrawal threshold (PWT). In the absence of pain reaction, a filament of nest greater force was applied. The intertrial interval between each measurement was set to be to 5 min at least to avoid eliciting aversive behaviors. Mice with abnormal basic nociceptive thresholds were excluded.

Immunofluorescence (IF)

Mice were deeply anesthetized and transcardially perfused with 4℃ PBS followed by perfusion with 4% paraformaldehyde (PFA). Brains were dissected out and post-fixed overnight in 4% PFA at 4℃. Dehydration with 20% sucrose and 30% sucrose respectively. 35 μm-thick coronal sections were prepared by the freezing microtome (Leica CM 1860). For staining, sections were washed three times with PBS and then incubated with 5% bovine serum albumin (BSA) and 0.2% Triton X-100 dissolved in PBS at room temperature (RT) for 1 h. Next, the sections were incubated in diluted primary antibodies overnight at 4 °C. After washed three times, the sections were then incubated with diluted secondary antibodies for 2 h at RT. Finally, each section was mounted with 10 μL a fluorescence quenching sealing tablet containing with DAPI (BL739B, Biosharp) and immediately mounted with cover slips. Images were captured using ZEISS Axio Imager 2 microscope or LSM 800 confocal microscope for analyses. Slides were scanned using the pannoramic scanner system (PanoBrain, Meca Scientific). Cell counts and colocalization were subjected by ImageJ software (Fiji edition, National Institutes of Health, USA). Antibody details in Supplementary Table 1.

Three-dimensional (3D) reconstruction and morphological analysis

Confocal stacks of Iba1-positive microglia were obtained using Zeiss LSM800 microscope at 2 μm intervals covering 30 μm depth along the z-axis with a 40 × objective and were imported into the IMARIS 9.6.2 software (Bitplane) to 3D-reconstructe. We opened the Z-stacks image file and selected “Add new filaments and Create Calculate Diameter of Filaments” in a region of interest. After setting the diameter between 0.25 μm and 15 μm, we classified dendrite and seed points according to the actual size and selected the diameter of “Sphere Region” as 30 μm. Then, the threshold of dendrites was adjusted under 1.70 and “Detect Spines” was deselected in “Spine Points Diameter” step. Finally, specific values of microglia were available by clicking “statistics” option and further quantified soma size, process length, and number of branch points. Microglia was rendered by “Surface” module. PSD95⁺ puncta were 3D-reconstructed by “Spots” function and colocalized puncta were identified manually. To evaluate microglial engulfment, we used “split into surface objects” to analyze PSD95⁺ particles within IBA1⁺ microglia.

Golgi staining

Mice were anesthetized and decapitated. The brains were removed immediately without any transcardial perfusion. After rinsing with PBS, the brains were prepared using FD Rapid GolgiStain™ kit (FD Neuro Technologies, USA) by the same person according to the manufacture’s instructions. The brains were first immersed in the impregnation solution (A and B) with a replacement of fresh solution after 12 h and then kept in dark at room temperature for 2 weeks. Next, they were transferred into solution C for 24 h which was changed by fresh solution for another 48 h. Following, the tissue was frozen with dry ice and stored at − 80℃ until sliced coronally in 250-μm-thick sections on a VT1200S vibratome (Leica, Germany). The sections were mounted on gelatin-coated microscope slides and dried naturally for 2 days. Before and after a 10-min staining with the developing solution (D and E), the slides were rinsed twice for 4 min each in deionized water. Subsequently, the sections were serially dehydrated in 50%, 70%, 95%, and 100% alcohol for 4 min each, then transparentized in xylene three times for 4 min each and coverslipped with Permount (Fisher Scientific, USA). Finally, images were captured by as 40 × objective and the density of dendritic spines was quantified by ImageJ software.

Transmission electron microscope (TEM)

As described previously, anesthetized mice were successively perfused with pre-warm PBS and 4% PFA. Tissues (volume less than 1 mm³) collected from the ischemic penumbra were immersed in fixative solution (2% paraformaldehyde and 3% glutaraldehyde, pH 7.2) overnight at 4 °C. Next, the samples were post-fixed with 1.5% (w/v) potassium ferrocyanide and 1% (w/v) osmium tetroxide, 1% (w/v) osmium tetroxide, 2% (w/v) uranyl acetate in 0.1 M cacodylate buffer for 1 h each in dark. After washing three times with ddH₂O, the samples were dehydrated with gradient ethanol on a shaker, 10 min each (30% ethanol once, 50% ethanol once, 70% ethanol once, 80% ethanol once, 95% ethanol once, and 100% ethanol twice). Following this, absolute acetone: epon solutions were used to embed with a series of concentrations (3:1, 1:1, 1:3). The samples were replaced with fresh pure epon every 12 h in a 3-day period twice at room temperature and once in an electronic drying cabinet. Then the embedded samples were polymerized at 60℃ in an oven. Ultrathin sections (thickness of 70 nm) were cut with a diamond knife (Diatome, Biel, Switzerland) on a Leica EM UC7 Ultramicrotome and collected on copper single slot grids coated with formvar (Agar Scientific, UK). After freeze-substitution with 1% osmium tetroxide, 0.1% glutaraldehyde, and 2% H₂O, images of mitochondria or synapse were acquired with FEI Tecnai T12 transmission electron microscope with Eagle CCD and CryoBox and quantified by using ImageJ software.

Purified synaptosomes extractation

Mice were anesthetized and perfused with ice-cold PBS, and brains were removed rapidly. The contralateral vCA1 was carefully dissected with blunt puncture needles and homogenized with a Dounce homogenizer on ice. Synaptosomes were then isolated using Synaptic Protein Extraction Reagent (Syn-PER™, Thermo Fisher Scientific) per the manufacturer’s instructions. The homogenate was centrifuged at 1200 g for 10 min at 4 °C, saving 20 μL of the supernatant (S1) for total protein analysis. The remaining supernatant was further centrifuge at 15,000 g for 20 min at 4 °C, producing the pellet (synaptosomal fraction, P2).

Western blot

All samples, including cells or tissues, were lysed and sonicated in ice-cold RIPA (biosharp) buffer containing protease or phosphorylase inhibitors for 10 min, then lysates were centrifuged at 12,000 rpm at 4℃ for 15 min. Protein concentration was measured using BCA assay kit (biosharp). Samples were denatured by SDS protein loading buffer and separated in 8 ~ 12% SDS–polyacrylamide gel. A Tri-Color Prestained Protein Standard (EC2019, Sparkjade) was used as molecular weight marker. The proteins were electrically transferred to PVDF (Millipore) membrane and blocked in TBST containing 5% milk powder at room temperature for 1 h. The membranes were incubated with primary antibody solution at 4 °C overnight. After removing any non-bound primary antibodies, membranes were incubated with corresponding HRP-conjugated secondary antibody at room temperature for 1 h and washed in TBST for another three times. The abundances of proteins were detected with ProteinSimple Fluorchem R system with enhanced chemiluminescence (ECL) reagents (biosharp) and quantified by ImageJ software. Antibody details in Supplementary Material 1.

In vitro electrophysiological recordings

Acute brain slice preparation

Mice were anesthetized and transcardially perfused with 10 mL N-methyl-d-glucamine (NMDG) cutting solution (pH 7.3 ~ 7.4) contained (in mM): 93 NMDG, 93 HCl, 2.5 KCl, 1.2 NaH₂PO₄, 30 NaHCO₃, 20 HEPES, 10 MgSO₄, 0.5 CaCl₂, 25 glucose, 5 Na-ascorbate, 2 thiourea, 3 Na-pyruvate. The brains were rapidly dissected out and placed in the same ice-cold NMDG cutting solution. Coronal brain slices (300 mm) containing the vCA1 were sectioned on a VT1200S vibratome (Leica, Germany). The slices were initially recovered in NMDG cutting solution at 34℃ for 10 ~ 13 min and then incubated in oxygenated artificial cerebrospinal fluid (ACSF) contained (in mM): 126 NaCl, 2.5 KCl, 2 CaCl₂, 2 MgCl₂, 26 NaHCO₃, 1.25 NaH₂PO₄, and 10 glucose at 25℃ for 1 h. The vCA1 slices were gently transferred to a recording chamber (RC- 27, Warner Instruments) at room temperature, where they were continuously perfused with oxygenated ACSF at a flow rate of 2.5 ~ 3 mL/min. The vCA1 neurons were visualized using an upright microscope (BX51 WI, Olympus, Tokyo) equipped with a 40 × water-immersion objective lens, and the target neurons were identified by their morphology and location.

Whole-cell patch-clamp recording

To measure miniature excitatory postsynaptic currents (mEPSCs), whole-cell patch-clamp recordings were performed with patch pipettes (resistance 3.5 ~ 5 MΩ, Sutter Instruments, BF150-86-10) that were pulled from borosilicate glass capillaries (outer diameter 1.5 mm) on a P1000 micropipette puller (Sutter Instruments). The borosilicate pipettes were filled with internal solution contained (in mM): 135 K-gluconate, 5 KCl, 0.5 CaCl₂, 2 Mg-ATP, 0.1 GTP,10 HEPE and 5 EGTA (pH adjusted to 7.3 with KOH and osmolality to 300 mOsm with sucrose). During measurements, tetrodotoxin (TTX, 1 μM) and bicuculline (10 μM) were added to block voltage-dependent Na⁺ channels and GABA_A receptors respectively. mEPSCs were recorded with a holding potential of − 70 mV. The recorded signals were amplified with MultiClamp 700B amplifier (Molecular Devices), sampled at 5 kHz and filtered at 1 kHz with a Digidata 1440 A digitizer (Molecular Devices). The frequency and amplitude of mEPSC events were determined over a 5-min recording and analyzed by MiniAnalysis 6.07 software (Synaptosoft) with an amplitude threshold set at 10 pA. Data within 15% of initial values (25 ~ 35 MΩ) were accepted.

Fiber photometry

After completion of photothrombotic stroke model as previously described, an AAV-CaMKII-GCaMP6 s virus (BrainVTA CO., Ltd, Wuhan, China) was injected targeted to contralateral vCA1 (AP, − 3.28 mm; ML, − 3.3 mm; DV, − 4.6 mm) at a rate of 50 nL/s for 200 nL, the needle stayed for another 10 min before withdrawn. A fiber optic cannula (RWD, R-FOC-BL200 C- 39 NA) was inserted 0.2 mm above vCA1 injection sites secured to the skull with dental cement (Super Bond C&B). After surgery, the mice were permitted to recover for at least 1 week. To decrease laser bleaching, the laser power at the tip of the optical fiber was adjusted to 10 ~ 20 μW. The fluorescence signals were further low-pass filtered (40-Hz cutoff). The analog voltage signals were digitalized at 50 Hz and recorded by the multichannel fiber photometry recording system. The light intensity for 470 nm was set within a range of 20 ~ 40 μW, while the intensity for 410 nm was set within a range of 10 ~ 20 μW. Prior to the stimuli, animals habituated in the open field and freely move for 10 min, during which time baseline fluorescent signals were collected. Each mouse was tested five times and compared responses to irritant gas stimulation. Fluorescence signals were recorded continuously by signal acquisition software (Inper Studio) and further normalized as ΔF/F by analysis software (Inper Datz Process). All injection sites were verified histologically. We only included mice with virus expression limited to the targeted regions.

Quantitative real-time PCR (qRT-PCR)

Total RNA was isolated from tissue samples by TRIzol reagent (15596018 CN, Thermo Fisher Scientific Inc., USA). RNA quality and concentration were determined by NanoDrop One spectrophotometer (Thermo Fisher Scientific). cDNA was reverse‐transcribed using the HiScript III RT SuperMix for qPCR (R323, Vazyme, China) according to the instructions. All primers were synthesized by Genaral Biosystems (Anhui, China). The Real-time PCR was performed with ChamQ Universal SYBR qPCR Master Mix (Vazyme, Q711) on a LightCycler 480II system (Roche, USA). Relative mRNA expression was calculated using the 2^−ΔΔCT method with GADPH as a loading control for normalization. All primer sequences were as follows(5’− 3’):

• mIL- 1β, 5’-GAAATGCCACCTTTTGACAGTG- 3’(forward),

• 5’-TGGATGCTCTCATCAGGACAG- 3’(reverse);

• mIL- 18, 5’-GTGAACCCCAGACCAGACTG- 3’(forward),

• 5’-CCTGGAACACGTTTCTGAAAGA- 3’(reverse);

• mGADPH, 5’- AGGTCGGTGTGAACGGATTTG- 3’(forward),

• 5’-GGGGTCGTTGATGGCAACA- 3’(reverse).

RNA-seq analysis

Total RNA extraction was performed as previously described. Samples were sent to T Tsingke Biotechnology Co., Ltd. for sequencing. Sequencing was performed on Illumina NovaSeq × Plus platform. Raw reads were first processed by FastQ software, the edgeR package (http://www.rproject.org/) was used to identify Differentially expressed genes (DEGs). Gene Ontology (GO), Kyoto Encylopedia of Genes and Genomes (KEGG), and Gene Set Enrichment Analysis (GSEA) were conducted based on DEGs.

Liquid chromatography‐tandem mass spectrometry

The concentration of PEA in brains was analyzed by an ACQUITY premier ultraperformance liquid chromatography (ACQUITY premier UPLC, Waters) coupled with a Xevo TQ-S triple quadrupole mass spectrometer (Xevo TQ-S, Waters). Analytes were separated on an ACQUITY UPLC BEH C18 (1.7 μm; 2.1 × 100 mm; Agilent Technologies) column maintained at 40℃ with a mobile phase consisting of 0.1% formic acid aqueous solution as solvent A and 0.1% acetonitrile solution as solvent B. The linear gradient was set as follows: 40% B in 0 ~ 0.2 min, change to 10% B at 2.0 min continuing to 4.0 min, change to 40% B at 4.10 min, and hold to 6 min. The flow speed was 0.3 mL/min and the autosampler temperature was set at 10℃. The mass spectrometry was operated in the positive electrospray ionization (ESI) mode. Analytes and internal standards were detected by Multiple-reaction monitoring (MRM) of the following transitions: PEA = m/z 300.3 > 62.2, PEA-D4 = m/z 304.4 > 62.2. Capillary voltage was 5000 V. Desolvation temperature was 500℃ with a flow of 1000 L/h. Nebulizer pressure was 7 bar. MassLynx V4.2 SCN1035 was used for data acquisition and processing.

Primary microglia culture

As previously described, postnatal 0–3-day-old mice were sterilized, decapitated, and brains were dissected on ice. After removal of meninges and blood vessels in pre-cooled HBSS (biosharp), the cortices were transferred into complete DMEM medium and gently triturated by pipetting until no large tissue remained. The cell suspension passed through a 70 μm cell strainer was plated onto poly-L-lysine-coated T75 flasks. Medium was changed to complete DMEM medium containing 50 ng/ml mouse GM-CSF (R&D Systems) the next day. Primary microglia were harvested by shaking (37℃, 200 rpm, 30 min) after 10 ~ 12 days in culture and every 3 days thereafter. The obtained cells were seeded into six-well plates at a density of 1 × 10⁶ per well for overnight before further treatment.

Small interfering RNA (siRNA) transfection

For gene knockdown experiments, the Lipofectamine RNAiMAX reagent (Thermo Fisher Scientific, USA) and siRNA were mixed well in optimal-MEM and incubated at room temperature for 5 min. The transfection mixture was added into the media at a final concentration of 50 nM and cultured for at least 6 h. NAAA siRNA (GGUUCCGAGUUGAGACAAATT) was synthesized by Genaral Biosystems (Anhui, China).

Oxygen–glucose deprivation/reoxygenation (OGD/R) model

The OGD model was established as previously described to simulate ischemia–reperfusion in vitro [22]. In brief, microglia were incubated in glucose-free DMEM (Gibco) and transferred into an incubator with a 5% CO₂ and 95% N₂ atmosphere at 37 °C. 6 h later, medium was replaced by fresh complete DMEM under normoxic conditions for an additional 24 h as reperfusion.

Mitochondrial membrane potential (MMP) assay

JC- 1 is a fluorescent probe that is widely used to detect MMP (ΔΨm). JC- 1 monomers emit green fluorescence in damaged mitochondria, whereas JC- 1 aggregates emit red fluorescence in healthy mitochondria. According to manufacturer’s protocol (C2006S, beyotime), cells were incubated with JC- 1 dye solution for 20 min at 37 °C. After washed twice with JC- 1 buffer, cells were then observed under a fluorescence microscope (Leica, Germany) and calculated by ImageJ Fiji. MMP is represented as the mean of red/green fluorescence intensity ratio.

Mitochondrial permeability transition pore (mPTP) assay

The mPTP opening was detected by Mitochondrial Permeability Transition Pore Assay Kit (C2009S, beyotime). Calcein-AM fluoresces diffusely throughout the entire cell, whereas CoCl₂ quenches calcein in the cytoplasm. Briefly, cells were collected and incubated with Calcein-AM staining solution for 30 min at 37℃. After washing by assay buffer twice, the fluorescence intensity was measured by flow cytometer. Loss of green fluorescence indicates opening of mPTP.

Reactive oxygen species (ROS) measurement

Intracellular ROS abundance was measured using Reactive Oxygen Species Assay Kit (S0033M, Beyotime). DCFH-DA is hydrolyzed to DCFH by esterase, which can be further oxidized by ROS to fluorescent DCF. In short, cells were collected and incubated with 10 μM DCFH-DA solution at 37℃ for 20 min and then washed by serum-free culture medium for three times. The sample was determined using a flow cytometer (FACSFortessa, BD Biosciences) and analyzed with FlowJo V.10 software.

Microglial phagocytosis assay

Primary microglia were pre-incubated with IL-18BP dissolved in PBS (40 ng/mL) or vehicle for 24 h, following by incubated for another 12 h under CM or control media as starvation treatment. Then, fluorescent red labeled 0.1-µm latex-beads (Polysciences, Warrington, PA) were added into medium at a ratio of 5:1 (beads: cell). After incubating for 2 h, cells were collected and detected by flow cytometry (FACSFortessa, BD Biosciences) to determine the phagocytic capacity by fluorescent intensity and analyzed with FlowJo V.10 software.

Quantification and statistical analysis

All data were statistically analyzed using GraphPad Prism 8 and are shown as mean ± SD. Unpaired two-tailed Student’s test was used for two group comparisons. One-way or two-way ANOVA, followed by Bonferroni’s correction was used for multiple comparisons. P value < 0.05 was regarded significant. Significance levels are indicated as follows: not significant (n.s.), *p < 0.05, **p < 0.01 and ***p < 0.001.

Results

Microglial PEA metabolite is exhausted after stroke

Several recent studies have revealed that endocannabinoids metabolites are detected in the CNS and their concentration shows significant changes in pathological stress microenvironment [10]. A transient release of PEA at high concentrations in the brain has been identified following stroke in clinic, which may play significant roles in regulating post-ischemic pathophysiological processes by its neuroprotective effects [6]. To investigate the roles of PEA in the courses of stroke, a mouse model of photothrombotic stroke (PTS) was established. We firstly monitored behavior changes after cortex stroke and the follow-up experimental arrangements were shown in Figure S1A. Mice were screened according to the number of buried marbles for three consecutive days. Mice with buried marbles more than 10 were classified as post-stroke anxiety (PSA) and mice with buried marbles less than 10 were identified as non-anxiety individual (called PTS). Surprisingly, we found 20% mice manifested anxiety-like behavior after stroke (Figure S1B and C). To investigate the influence of infarct volume differences to the anxiety sensitivity, we applied the MR Imaging to ensure there was no significant difference between PTS and PSA (Figure S1D and E). Besides, PTS and PSA mice had similar motor coordination assessed by rotarod test (Figure S1G). Open filed test was performed at day 14 after stroke for exploratory behavior and anxiety, total traveled distance had no difference between two groups (Figure S1 H and I). However, PSA mice spent less time in the central area (Figure S1J). Correspondingly, in the elevated plus maze test, PSA mice showed a lower propensity to remain in the open arms (Figure S1K and L). So, the anxiety-like behavior caused by ischemic stroke is not simply affected by the infarct volume itself.

Then we next detected the concentration of PEA in the ischemic penumbra at 0, 6, 12 and 24 h after stoke via mass spectrometry, with mean values of 149.77, 176.99, 115.30 and 112.06 pmol/mL, respectively. The result showed that PEA concentration rapidly increased followed by a continue exhaustion, as the course of the disease prolongs (Fig. 1A), which is consistent with clinical cohort study on stroke [6]. Surprisingly, PEA exhaustion was more significant in mice with negative emotion, declining to half or even lower than physiological level at 3 d after stroke (mean value of Sham group: 134.99 pmol/mL, mean value of PTS group: 109.36 pmol/mL, mean value of PSA group: 77.42 pmol/mL) (Figure S1F). PEA, as one of endocannabinoids, has been reported to guard against fear, anxiety and stress in neuropsychiatric disorders. PEA maintained in dynamic balance by synthesis and degradation enzymes. It is synthesized via NAPE and NAPE-PLD, and rapidly hydrolyzed by NAAA and FAAH, with NAAA being more efficient [23, 24]. So, we detected the PEA metabolizing enzymes in ischemic cortex, hippocampus and amygdala of mice from Sham, PTS and PSA group. NAAA in ischemic cortex elevated dramatically post stroke, especially and obviously in mice from PSA group (Fig. 1B). In addition, there was also a strong negative correlation between the relative expression level of NAAA and time on rod in rotarod test (Fig. 1C) or time in center in OFT (Fig. 1D), respectively. Since stroke primarily occurs in the aged population, we further detected the NAAA expression of ischemic hemisphere in both 6-week-old (6 W) and 18-month-old (18 M) mice. Notably, NAAA protein level is significantly upregulated in PSA group either young or aged mice (Figure S1M), corresponding to the mass spectrometry data of PEA metabolites (Figure S1F). NAAA is primarily expressed in glial cells within the CNS based on single-cell sequencing dataset following the stroke modeling [25]. The RNA expression of NAAA shows significant increase only in microglia, but not in astrocyte (Fig. 1E), which was consistent with immunofluorescence outcomes either aged or young mice (Fig. 1F and G, Figure S1N). Therefore, the exhaustion of PEA metabolites after stroke is mainly mediated by the elevated NAAA in microglia, which is closely related to post-stroke anxiety. This phenomenon is observed in both young and aged mice.

Fig. 1.

PEA metabolism is exhausted in microglia after ischemic stroke. A Trends of PEA concentrations in the ischemic penumbra detected by LC–MS at 0, 6, 12 and 24 h post PTS (n = 3). B Representative western blot images of PEA metabolizing enzymes at 72 h after modeling in three regions of brain as indicated (n = 5). C, D The Pearson’s correlation analysis between the relative protein level of NAAA in hippocampus and the latency fall from the rotarod (C) (n = 8), and time spent in the center in open-field tests (D) (n = 8). E Analysis of NAAA expression level in microglia and astrocytes at 24 h post MCAO (middle cerebral artery occlusion) or Sham operation, the single-cell sequencing data was obtained from GSE174574. F, G Representative immunofluorescence images of NAAA (red) co-labeled with IBA1 (green) and GFAP (cerulean blue) in Sham and PTS mice at 24 h after PTS (scale bars: 500 μm), and the local magnification of ischemic penumbra (scale bars: 100 μm) (F). Statistical analysis of NAAA⁺ microglia (NAAA⁺ IBA1⁺) and astrocyte (NAAA⁺ GFAP⁺) in the ischemic penumbra (G) (n = 20). The data is shown as mean ± SD. Two-way ANOVA with Bonferroni post hoc test (A); unpaired Student’s t test (E and G); ns, no significance; ***p < 0.001

Boosting PEA metabolite contributes to stroke recovery and alleviates PSA progression

To investigate whether boosting PEA metabolite is associated with the stroke recovery and PSA progression alleviation, we used Naaa^−/− mice to prevent the degradation of PEA. Naaa^−/− mice and Wt mice under model of photothrombotic stroke were performed experiments as flowchart shown (Fig. 2A). Western blot was used to verify the Naaa was effectively knockout (Figure S2A). To rule out the impact of NAAA on neural development, Wt and Naaa^−/− mice had similar general observation (Figure S2B) and no obvious differences in rotarod test (Figure S2C), Von Frey filament test (Figure S2D), OFT (Figure S2E ~ G) and EPM tests (Figure S2H and I). So, Naaa deficiency does not affect the psychoneurological function of mice. Laser speckle flowmetry showed that cerebral blood flow was similar in Wt and Naaa^−/− mice 10 min after reperfusion. Both genotypes had comparable ischemic core and penumbra areas, indicating that Naaa deficiency minimally affects cerebral blood flow and both groups experienced similar initial ischemic injury (Fig. 2 B and C). But, Naaa^−/− mice significantly contributed to reduction of infarct volume at day 7 after modeling (Fig. 2D and E). Naaa^−/− groups showed an accelerated functional recovery post stroke assessed by consecutive rotarod tests (Fig. 2 F). In addition, Naaa^−/− mice manifested reduced number of buried marbles (Fig. 2G and H), increased total travel distance and the duration of time in the central area of the open field (Fig. 2I, J and K). The time stay in the open arm was also increased significantly in Naaa^−/− mice (Fig. 2L and M). PSA is frequently comorbid with pain, which exacerbates the duration and severity of anxiety and drives a vicious cycle in between [26]. PSA-Naaa^−/− group performed lower sensitivity to mechanical stimuli compared to PSA-Wt group (Fig. 2N). To exclude gender differences in ischemic stroke, female mice were performed experiments as flowchart shown (Figure S2L). The results of OFT (Figure S2M ~ O) and EPM test (Figure S2P and Q) all manifested that female PSA-Naaa^−/− mice were significantly less anxiety than female PSA-Wt mice. So, based on the above evidence, we confirmed that boosting PEA metabolite contributes to stroke recovery and alleviates PSA progression, with the protective effects being independent of sex.

Fig. 2.

NAAA deficiency alleviates pathology and anxiety behavior in PSA mice. A Schedule of the experimental design. B, C Representative laser speckle contrast imaging of Sham, PTS-Wt and PTS-Naaa^−/− groups pre-operative and 10 min post-reperfusion (B), the red area indicates high blood perfusion. The quantification of cerebral blood flow (% contralateral hemisphere) of three groups (C) (n = 5). D, E Sequential MRI and 3D reconstructed images of the lesion area in Sham, PTS-Wt and PTS-Naaa^−/− groups at 7 d after PTS (D). Statistical analysis of the infract volume E (n = 5). F The latency before fall from the rotarod of three groups at indicated time points from pre-PTS to 14 d after PTS (n = 5). G, H Representative images of marble burying test within three groups at 14 d after PTS (G), and the statistical analysis (H) (n = 8). I ~ K Representative 5 min activity heatmap in OFT at 16 d after PTS (I). Statistical analysis of total travel distance (J) and the percentage of time spent in center (K) within three groups (n = 8). L, M Representative 5 min activity heatmap in EPM test at 18 d after PTS (L) and analysis of time spent in open-arm within three groups (M) (n = 8). N Mechanical pain threshold trends detected by Von Frey filament test, eight detections at indicated time points from pre-PTS to 18 d after PTS (n = 8). The data is shown as mean ± SD. One-way ANOVA with Bonferroni post hoc test (C, E, H, J, K, M); two-way ANOVA with Bonferroni post hoc test (F, N); ns, no significance; *p < 0.05, **p < 0.01, ***p < 0.001

Exogenous supplementation of PEA mitigates pathological progression and anxiety like behavior after stroke

Firstly, we intraperitoneally injected 10 mg/kg PEA to untreated wild-type mice and detected the concentration of PEA in brain at 0, 15, 30, and 60 min post-injection by mass spectrometry, with mean values of 199.8, 1012.7, 854.4, and 252.8 pmol/ml, respectively (Supplementary Material 2). The concentration peaked at 15 min, corresponding to previous studies [10], indicating that PEA can quickly cross the blood–brain barrier. Next, PEA was administrated to the mice with 10 mg/kg daily for consecutive 2 weeks after stroke (Fig. 3A). The brains were processed for MRI to measure infarct volume at 7 days. Strikingly, mice treated with PEA showed a significant reduction in cerebral infarct volume by about 30% compared with vehicle group (Fig. 3B and C). Mice showed increased thresholds to mechanical stimulation at the hind paw in PSA-PEA group over a 2-week recovery period, reversing the high pain susceptibility of PSA-Veh group (Fig. 3D). Moreover, anxiety-related behavioral tests including MBT, OFT and EPM were mainly performed at 14, 16 and 18 days after stroke to assess spontaneous exploratory activity. Mice intraperitoneally injected PEA traveled longer total distances in the arena, although the difference was not statistically significant (Fig. 3H). And they also exhibited less typical anxiety-like performance, evidenced by fewer buried marbles in the MBT tests (Fig. 3E and F) and less spent time in the center zone and the open arm in the OFT (Fig. 3G and I) and EPM tests (Fig. 3J and K), respectively. These results indicated that boosting PEA metabolite with pharmacological supplementation can both alleviate the anxiety behavior as well as pain sensitivity in the course of stroke.

Fig. 3.

Exogenous supplementation of PEA mitigates pathological progression and anxiety behavior after stroke. A Schedule of the experimental design. B, C Sequential MRI images of the lesion area in Sham, PSA-Veh and PSA-PEA groups at 7 d after PTS (B), and the statistical analysis of infarction volume (C) (n = 5). D Trends of mechanical pain thresholds of three groups, eight detections at indicated time points from pre-PTS to 18 d after PTS (n = 8). E, F Representative images of marble burying test at 14 d after PTS (E) and the statistical analysis (F) (n = 8). G ~ I Representative activity heatmap in OFT at 16 d after PTS (G). Statistical analysis of total travel distance (H) and percentage of time spent in center (I) (n = 8). J, K Representative activity heatmap in EPM test at 18 d after PTS (J) and analysis of time in open-arm within indicated three groups (K) (n = 8). The data is shown as mean ± SD. One-way ANOVA with Bonferroni post hoc test (C, F, H, I, K); two-way ANOVA with Bonferroni post hoc test (D); ns, no significance; *p < 0.05, **p < 0.01, ***p < 0.001

To rule out possible age effects, we purchased 18-month-old (18 M) mice, and treated with PEA or vehicle after modeling as flowchart shown (Figure S3A). In 18 M PTS-PEA mice, infarct volume was significantly reduced by day 7 post-modeling (Figure S3B and C). Moreover, these mice exhibited faster functional recovery after stroke, as demonstrated by consecutive rotarod tests (Figure S3D). Behavioral tests revealed that 18 M PSA-PEA mice exhibited attenuated anxiety-like phenotypes: fewer buried marbles in MBT (Figure S3E and F), more total travel distance and longer time spent in the center of open field (Figure S3G ~ I). Furthermore, PEA-treated mice showed willingness to explore the open arms in EPM test compared to vehicle group (Figure S3J and K).Thus, PEA administration can accelerate stroke recovery and improve anxiety symptoms of aged mice, which enhances the possibility of the study for therapeutic applications.

Contralateral vCA1 activity and dendritic complexity are remolded by PEA metabolism after PSA

Studies have shown that the contralateral hemisphere can undergo structure changes and functional reorganization in response to the injury such as stroke, which may be a compensatory mechanism to recover lost functions [27–29]. Our previous study found that contralateral vCA1 can be remodeled by cytokines derived from ischemic-damaged lesion, which has been strongly implicated with negative emotion after stroke such as depression or anxious behavior. Specifically, vCA1 contains a cluster of excitatory pyramidal neurons named “anxiety cells” [30],which are specifically activated by anxiogenic environments and play a crucial role in avoidance behaviors. The vCA1 related circuit serves as a direct route through which the hippocampus can rapidly influence innate anxiety behaviors. We initially assessed neuronal activity by evaluating c-Fos expression in various anxiety-related regions, among which the activity of entire pBLA-vCA1 circuit exhibited the most significant reduction in PSA group (Figure S4A and B). But this decreased activity in pBLA-vCA1 circuit could be alleviated by Naaa knockout (Fig. 4A and C). The anterior basolateral amygdala (aBLA) and posterior basolateral amygdala (pBLA) respectively regulate deep calbindin 1-negative neurons (Calb1⁻) and superficial calbindin 1-positive neurons (Calb1⁺) in the vCA1. Activation of the aBLA-vCA1 Calb1⁻ circuit induces anxiety-like behaviors, whereas activation of the pBLA-vCA1 Calb1⁺ circuit exhibits anxiolytic effects [31, 32]. We further confirmed that Calb1⁺ neurons are responsible for the decrease of neuronal activity of contralateral vCA1 in post-stroke anxiety mice (Figure S4C and D). So, this evidence indicated that inactivation pBLA-vCA1 Calb1⁺ circuit contributes to post-stroke anxiety behavior, which is consistent with previous study [16]. Boosting PEA metabolite with Naaa knockout reactivated pBLA-vCA1 Calb1⁺ circuit to mitigate anxiety.

Fig. 4.

Naaa knockout mitigates microglia activation and synaptic complexity decline in contralateral vCA1. A, C Representative immunofluorescence images of NeuN (green) and c-Fos (violet) co-labeled cells in contralateral vCA1 of Sham, PSA-Wt and PSA-Naaa^−/− mice. White arrows indicate co-labeled cells (A) (scale bars: 50 μm). Quantification of c-Fos and NeuN double-positive cells in vCA1 (C) (n = 6). B, D Representative images of neuronal dendrites by Golgi-Cox staining in the contralateral vCA1 at 18 d after PTS (B) (scale bars: 5 μm). Summarized data for spine numbers per 10 μm (D) (n = 30 synapses from 3 mice). E ~ H Representative TEM images of synapses structure in the contralateral vCA1 at 18 d after PTS (E) (scale bars: 500 nm). Quantification of vesicle number in the PSD area (F), presynaptic compartments (G), and synapse density (H) (n = 30 synapses from 3 mice). I Representative membrane potential of pyramidal neurons in the vCA1 of Sham, PSA-Wt and PSA-Naaa^−/− mice groups at 18 d after PTS. J, K Cumulative curves and histograms for mEPSC frequency (J) and amplitude (K) (n = 6). L, M The average of behavior-triggered ΔF/F signals in Sham, PSA-Wt and PSA-Naaa^−/− mice groups at 18 d after PTS (L) and the peak of curves (M) (n = 5). N Representative heatmap of depicting the Ca²⁺ responses in Sham, PSA-Wt and PSA-Naaa^−/− groups. O Representative western blot images of the indicated components of AMPAR in extracted vCA1 synapses (S) and total (T) vCA1 (n = 5). P Representative immunofluorescence images (left) of IBA1 (green) and PSD95 (red) in contralateral vCA1 (scale bars: 10 μm). Imaris-based 3D reconstruction images (middle) of microglia and the local magnification (right, scale bars: 1 μm), red dots represent engulfed synapses. Q Quantification of IBA1⁺ cell soma size in the contralateral vCA1 from Sham, PSA-Wt and PSA-Naaa^−/− mice (n = 6). R, S Imaris-based semi-automatic quantification of total process length (R) and number of branchpoints (S) of microglia in contralateral vCA1 from Sham, PSA-Wt and PSA-Naaa^−/− groups (n = 6). T Quantification of PSD95⁺ puncta in microglia from indicated groups (n = 20). All samples for sectioning and protein lysates were collected at 90 min after the last behavioral test on the 18 th day post PTS. The data is shown as mean ± SD. One-way ANOVA with Bonferroni post hoc test (C, D, F ~ H, J, K, M, Q ~ T); ns, no significance; *p < 0.05, **p < 0.01, ***p < 0.001

Decreased neuronal activation in response to pathological stress, such as stroke, can be attributed to alterations in synaptic structure and decrease of dendritic spine density [33, 34], which are critical for synaptic transmission and plasticity. So, we next assessed dendritic spines on individual dendrites of pyramidal neurons by Golgi staining. And we found the spines number decreased in PSA-Wt group (Fig. 4B and D, Figure S2J and K). TEM analysis showed a lower postsynaptic density (PSD) area, presynaptic vesicle number and synapse density in the vCA1 of the PSA-Wt group compared to those in the sham group. However, these phenotypes can be effectively reversed in Naaa knockout mice (Fig. 4E ~ H). Similarly, mini excitatory postsynaptic current (mEPSC) records of glutamatergic neurons reflected greater numbers of excitatory inputs into vCA1 by PSA-Naaa^−/− mice, presenting an increase in interevent intervals and an insignificant decrease in amplitude (Fig. 4I ~ K). Additionally, to explore whether vCA1 glutamatergic neurons are sensitive to aversive stimulus with air puff tests, AAV-CaMKII-GCaMP6s, a virus expressed calcium indicators in glutamatergic neurons, was injected into vCA1 and then an optical fiber was implanted in close proximity to this region for fibre photometer recordings of PSA mice. Calcium signals decreased strongly and reliably in PSA-Wt group when mice received a irritant gas stimulation (Fig. 4L ~ N), compared with PSA-Naaa^−/− group. AMPA receptors (AMPARs) in vCA1 region are reported primarily responsible for mediating excitatory synaptic transmission within the central nervous system. The phosphorylation of various subunits influences both the calcium permeability and the stability of these receptors, which can indirectly assess synaptic loss [32, 35]. There is a decrease in phosphorylation at the GluA1 S845 and S831 sites of AMPARs under stress-induced anxiety, resulting in a diminished single-channel open probability and a reduction in single-channel conductance, respectively [36]. We then investigated molecular changes in vCA1 by detecting the expression of AMPARs and PSD95. The expression of GluA1 and PSD95 in Wt mice decreased more apparently after post-stroke anxiety but not in Naaa^−/− mice, compared to sham group. Phosphorylation at S831 but not S845 of GluA1 was significantly decreased in contralateral vCA1, which was reversed by Naaa knockout (Fig. 4O).

Microglia are known to interact with neurons [37, 38], facilitating the reshape of neural circuit patterns through synaptic pruning over extended periods [39]. Pruning of synaptic terminals by microglia is an important mechanism underlying synapse loss in pathological stress [40–42], which is associated with chronic anxiety-related disorders such as post stroke anxiety [43]. Microglial activation was observed in contralateral vCA1 region of PSA mice characterized by distinct morphological alterations, including a reduction in process length and branching points and a growth in soma size. Microglial three-dimensional reconstruction showed that abundant immunoreactive puncta of PSD95 colocalized with microglia in PSA-Wt mice. And these phenotypes were partly reversed in Naaa^−/− mice (Fig. 4P ~ T). Together, these results indicated that PEA metabolism reshapes contralateral vCA1 dendritic complexity, Calb⁺ neuron activity and microglial engulfment after PSA. So, maintenance of microglial PEA pool content in ischemic-damaged lesion can preserve contralateral vCA1 synaptic integrity, enhancing anxiolytic pBLA-vCA1^Calb1+ circuit activity by alleviating microglial phagocytosis-mediated synaptic loss.

Exogenous PEA administration alleviates microglia activation and synaptic decline in contralateral vCA1

Our data supported that hypoactivity of contralateral vCA1 Calb⁺ neurons develops in PSA mice. To investigate whether exogenous supplementation PEA can influence contralateral vCA1 Calb⁺ neurons, immunofluorescent staining was performed with c-Fos, a marker of neuronal activity (Fig. 5A). The number of c-Fos⁺ NeuN⁺ cells in PEA-treated mice increased significantly compared with vehicle group (Fig. 5 D). In line with these results, pretreatment with PEA significantly enhanced the loss of GluA1 and PSD95 levels in contralateral vCA1 region from those PSA-veh mice. Phosphorylation at S831 and S845 of GluA1 was reversed in contralateral vCA1 of PSA-PEA mice (Fig. 5B). Upon exposure to stress, surveillant microglia transform into reactive microglia, subsequently influencing neurocognitive outcomes and emotions through a mechanism termed microglia-dependent synaptic remodeling [39]. Furthermore, CD68⁺ IBA1⁺ cells, associated with reactive microglia status, were found to be increased in proportion in the post-stroke anxiety mice (Fig. 5C). Proportion of CD68⁺ IBA1⁺ cells in PSA-PEA mice approximately restored to sham level (Fig. 5E). In addition, compared with PSA-veh group, the morphological characteristics of IBA- 1-labelled microglia in PSA mice were significantly changed by exogenous supplementation PEA (Fig. 5F). Our results showed that the size of normalized soma was significantly reduced by PEA treatment (Fig. 5G). Boosting PEA reversed the PSA-induced reduction in both the process length and the branch point in microglia (Fig. 5 H and I). Three-dimensional reconstruction further revealed that abundant postsynaptic marker PSD95-labelled immunoreactive puncta (red) and Iba-labelled microglia (green) colocalized in PSA mice. Compared with the vehicle control, the engulfed synapses in microglia from contralateral vCA1 region were significantly reduced after mice treated with PEA (Fig. 5J). These results collectively showed that contralateral vCA1 synaptic loss in PSA mice was due to over microglial phagocytosis, which can be reversed upon exogenous administration of PEA.

Fig. 5.

Exogenous PEA administration alleviates microglia activation and synaptic decline in contralateral vCA1. A, D Representative immunofluorescence images of NeuN (green), c-Fos (violet) co-labeled cells in contralateral vCA1 of Sham, PSA-Veh and PSA-PEA mice (A) (scale bars: 100 μm). The quantification of c-Fos and NeuN double-positive cells (D) (n = 6). B Representative western blot images of the indicated components of AMPAR in extracted vCA1 synapses (S) and total (T) vCA1 of Sham, PSA-Veh and PSA-PEA mice (n = 5). C, E Representative CD68 (red) and IBA1 (green) co-labeled cells in the vCA1 from Sham, PSA-Veh and PSA-PEA mice (C) (scale bars: 10 μm). Statistical analysis of the proportion of CD68⁺ microglia in contralateral vCA1 (E) (n = 6). F Representative immunofluorescence images (left) of IBA1 (green) and PSD95 (red) in contralateral vCA1 (scale bars: 10 μm). Imaris-based 3D reconstruction images (middle) of microglia and the local magnification (right, scale bars: 1 μm), red dots represent engulfed synapses. G Quantification of microglia soma size in the contralateral vCA1 from Sham, PSA-Veh and PSA-PEA mice (n = 6). H, I Imaris-based semi-automatic quantification of total process length (H) and number of branchpoints (I) of microglia in contralateral vCA1 from Sham, PSA-Veh and PSA-PEA mice (n = 6). J Quantification of PSD95⁺ puncta in microglia (n = 20). All samples for sectioning and protein lysates were collected at 90 min after the last behavioral test on the 18 th day post PTS. The data is shown as mean ± SD. One-way ANOVA with Bonferroni post hoc test (D, E, G ~ J); ns, no significance; *p < 0.05, **p < 0.01, ***p < 0.001

Inhibition PEA exhaustion reduces the diffusion of impaired lesion-derived lL- 18 to contralateral brain

PEA acts on multiple molecular targets in the central and peripheral nervous systems, which is known to possess important anti-inflammatory, neuroprotective and analgesic effects [23, 44, 45]. To further investigate whether inhibition PEA exhaustion with Naaa knockout could brake excessive inflammation reaction after stroke, we next performed RNA sequencing (RNA-seq) analyses on ischemic penumbra from PSA-Naaa^−/− and PSA-Wt mice. The volcano plot showed that 464 genes were significantly downregulated with 327 genes upregulated after Naaa knockout (fold change ≥ 2.0, p ≤ 0.05) (Figure S5A). KEGG analysis revealed that the downregulated genes were predominantly enriched in pathways related to cytokine-cytokine receptor interaction, TNF signaling, and NF-κB signaling, which suggests an inhibition of the inflammation response as a result of Naaa knockout (Figure S5B). Further gene set enrichment analysis (GSEA) identified significantly downregulated pathways, including the Gene Ontology Biological Process (GOBP) categories for inflammatory response and response to interleukin- 18 in PSA-Naaa^−/− mice (Figure S5C). Several genes associated with pro-inflammatory cytokines including IL- 18, IL- 1β, were mainly downregulated (Fig. 6A), consistent with qPCR results validated from peri-lesion areas in Wt and Naaa^−/− mice (Fig. 6B).

Fig. 6.

NAAA-induced inflammation response and anxiety-like behavior are dependent on IL- 18 signal activation. A The RNA-seq analysis was performed on ischemic penumbra at 3 d after PTS. Expression heatmap of genes associated with pro-inflammation and microglia phagocytic function between PSA-Wt and PSA-Naaa^−/− mice. B Representative real-time PCR data specific to IL- 18 and IL- 1β in ischemic penumbra of PSA-Wt and PSA-Naaa^−/− mice at 3 d after PTS. C Representative western blot images of the indicated mitochondrial dynamic proteins and NLRP3 inflammasome associated proteins in ischemic penumbra from Sham, PSA-Wt and PSA-Naaa^−/− mice at 7 d after PTS (n = 5). D Representative western blot images of the indicated mitochondrial dynamic proteins and NLRP3 inflammasome associated proteins in ischemic penumbra from Sham, PSA-Veh and PSA-PEA mice at 7 d after PTS (n = 5). E Representative TEM images of mitochondria in ischemic penumbra of Sham, PSA-Wt and PSA-Naaa^−/− mice at 7 d after PTS (scale bars: 500 nm). F ~ H Quantitative analysis of the mitochondrial perimeter (E), area (F) and circularity (G) of Sham, PSA-Wt and PSA-Naaa^−/− mice (n = 30). I Representative western blot images of IL- 18 in ischemic penumbra and contralateral vCA1 from Sham, PSA-Wt and PSA-Naaa^−/− mice at 7 d after PTS. J, K Representative images of marble burying test in Sham, PSA-Veh and PSA-IL- 18BP groups at 14 d after PTS (J), and the statistical analysis (K) (n = 8). L ~ N Representative activity heatmap in OFT 16 d after PTS (L). Statistical analysis of total travel distance (M) and percentage of time spent in center (N) within three groups (n = 8). O, P Representative activity tracking heatmap in EPM test at 18 d after PTS (O). Analysis of time spent in open-arm within three groups (P) (n = 8). Q Trends of mechanical pain thresholds of Sham, PSA-Veh and PSA-IL-18BP groups eight detections at indicated time points from pre-PTS to 18 d after PTS (n = 8). The data is shown as mean ± SD. One-way ANOVA with Bonferroni post hoc test (F ~ H, K, M, N, Q); two-way ANOVA with Bonferroni post hoc test (Q); ns, no significance; *p < 0.05, **p < 0.01, ***p < 0.001.

Inflammation and various cytokines are implicated in the mechanisms underlying negative emotions after stroke [46, 47]. IL- 18, as a novel biomarker, is independently correlated with depressive symptoms subsequent to a stroke [48, 49]. IL- 18 is a member of the interleukin- 1 cytokine superfamily, and its inactive precursor protein is ubiquitously present in nearly all cell types. IL- 18 serves as a pivotal link between mitochondrial fragmentation and the inflammasome cascade, driving the phagocytic response of monocytes and contributing to the overall immune defense mechanism [50]. Upon excessive mitochondrial fragmentation and inflammasome cascade activation, IL- 18 is activated through cleavage by Caspase1 and subsequently secreted into the extracellular space to exert its functions [16]. The processes of Drp1 assembly, oligomerization, and recruitment to the mitochondrial outer membrane initiate the fragmentation [51, 52]. In mice, Drp1 phosphorylation at serine residues 600 and 579 (equivalent to human serine 616 and 637) is crucial for its movement from the cytosol to the mitochondrial membrane [53]. Mitochondrial fusion, involving GTPases mitofusin (Mfn) 1 and 2 on the outer membrane, protects by merging two mitochondria into one tubular structure, strengthening the network [54]. Consequently, we assessed IL- 18 upstream signal about mitochondrial dynamics-related proteins and inflammasome cascade activity in damaged cortex. The expression levels of total Drp1 (t-Drp1) and Mfn1 proteins showed no significant differences among the groups. In contrast, p-Drp1^S579 expression increased and Mfn2 expression decreased after stroke in PSA mice. Both Naaa knockout mice and PEA-treated mice manifest less mitochondrial fission (Fig. 6C and D). The mitochondria perimeter and area became shorter and smaller with circularity increased in PSA mice visualized by electron microscopy, indicating an increase in mitochondrial fission (Fig. 6E ~ H). The NLRP3 inflammasome pathway related proteins significantly increased after PSA, inhibition PEA exhaustion with Naaa^−/− mice reduced the pathway activation (Fig. 6C). Exogenous PEA similarly modulated inflammation by reducing IL- 18 pathway activity, as indicated by the downregulation of NLRP3, Caspase- 1, and IL- 1β (Fig. 6D). The above results suggest that in post-stroke anxiety mice, the IL- 18 pathway activity in ischemic injury foci is significantly enhanced. Inhibition PEA exhaustion with exogenous PEA supplementation or degradation enzyme deficiency reduces IL- 18 pathway activity in impaired lesion. IL- 18 expression was predominantly observed in neurons during the early stress phase and in microglia during the later stress phase [49]. Microglia are capable of responding to IL- 18, which promotes their activation and enhances monocytes phagocytosis, ultimately resulting in the loss of synapses and neurons [10, 16]. The elevated expression of IL- 18 not only promotes the synaptic pruning function of microglia but is also closely related to post-stroke mood disorders, which remodels anxiolytic pBLA-vCA1^Calb1+ circuit in contralateral vCA1 region after stroke [16].

Then, we speculated that diffusion of impaired lesion-derived IL- 18 could also remodel contralateral vCA1. The results confirmed our hypothesis that IL- 18 protein level increased in contralateral vCA1 compared with sham group. And boosting PEA pool content with degradation enzyme deficiency reduced IL- 18 diffusion to contralateral vCA1 (Fig. 6I).

Inhibition of IL- 18 signal reverses the microglia mediated synaptic engulfment and activity in contralateral vCA1.

Furthermore, the pro-inflammatory effects of IL- 18 are counteracted by the constitutively secreted IL- 18 binding protein (IL- 18BP), which serves as an intrinsic inhibitor of IL- 18 [15, 55]. IL- 18BP (1 μg/kg) was administered daily via intra-cerebroventricular injection (ICV) for a duration of two weeks after modeling (Fig. 7A and B). Mice treated with IL- 18BP exhibited greater willingness to spontaneous exploratory activity, including increased total travel distances and more time spent in the center zone and open arm in the OFT (Fig. 6L ~ N) and EPM test (Fig. 6O and P), respectively. Additionally, the buried marbles test and von Frey filament test results also indicated that anxiety-like behavior was relieved with an elevated mechanical pain threshold (Fig. 6J, K and Q). Western blot analysis of contralateral vCA1 tissue in the PSA-Veh group revealed that diminished expression levels of GluA1 and PSD95 accompanied by synaptic instability as evidenced by phosphorylation of the GluA1 S831 site, but not the S845 site. Notably, PSD95 levels were significantly decreased in the contralateral vCA1, with a less pronounced effect observed in the PSA-IL- 18BP group. These findings suggested that IL- 18, serving as the primary inflammatory mediator, regulated the microglia-activation-induced remodeling of the contralateral vCA1 after ischemic stroke (Fig. 7C). Results of immunofluorescence showed that the colocalization of c-Fos and NeuN increased after IL- 18BP treated, which indicated neuron activity was reversed in contralateral vCA1 (Fig. 7D and E). Meanwhile, morphological analysis by confocal images indicated lower soma size with higher process length and branch points in microglia of PSA-IL- 18BP group, suggesting that suppressing IL- 18 may reduce microglial activation (Fig. 7F ~ I). We found that activated microglia engulf more synaptic structures excessively (marked by PSD95) in PSA-Veh group, while IL- 18BP reverse this phenomenon (Fig. 7J). Taken together, boosting PEA could brake inflammasome pathway and reduce the diffusion of impaired lesion-derived lL- 18 to contralateral brain, exerting anti-anxiety roles by preserve contralateral vCA1 synaptic integrity, enhancing anxiolytic pBLA-vCA1^Calb1+ circuit activity by alleviating microglial phagocytosis-mediated synaptic loss.

Fig. 7.

Inhibition of IL- 18 signal reverses the microglia mediated synaptic engulfment and activity in contralateral vCA1. A Schedule of the experimental design. B General observation of intra-cerebroventricular injection in mice (left). Representative coronal Nissl-stained brain section and corresponding brain map (right). C Representative western blot images of the indicated components of AMPAR in extracted vCA1 synapses (S) and total (T) vCA1 of Sham, PSA-Veh and PSA-IL-18BP mice (n = 5). D, E Representative immunofluorescence images of NeuN (green) and c-Fos (violet) co-labeled cells in contralateral vCA1 of Sham, PSA-Veh and PSA-IL- 18BP (D) (scale bars: 100 μm,). The Quantification of c-Fos and NeuN double-positive cells per mm² (E) (n = 6). F Representative immunofluorescence images of IBA1 (green) and PSD95 (red) in contralateral vCA1 (left, scale bars: 10 μm,). Imaris-based 3D reconstruction images (middle) of microglia and the local magnification (right, scale bars: 1 μm), red dots represent engulfed synapses. G Quantification of microglia soma size in the contralateral vCA1 from Sham, PSA-Veh and PSA-IL- 18BP mice (n = 6). H, I Imaris-based semi-automatic quantification of total process length (I) and number of branchpoints (J) of microglia in contralateral vCA1 from Sham, PSA-Veh and PSA-IL- 18BP mice (n = 6). J Quantification of PSD95 + puncta in microglia (n = 20). All samples for sectioning and protein lysates were collected at 90 min after the last behavioral test on the 18 th day post PTS. The data is shown as mean ± SD. One-way ANOVA with Bonferroni post hoc test (E, G ~ J); ns, no significance; *p < 0.05, **p < 0.01, ***p < 0.001

PEA-enriched microglia modulates phagocytosis via IL- 18

Following tissue or cell damage, danger signals initiate a molecular program that lowers PEA production and heightens NAAA-catalyzed PEA degradation [56]. Using isolated primary microglia, we performed oxygen–glucose deprivation (OGD) model to mimic in vivo ischemic injury to further study whether inhibiting PEA exhaustion modulates inflammation activation and microglia phagocytosis in vitro. OGD induced a persistent increase of IL- 18 upstream signal about inflammasome cascade (NLRP3, Caspase1, IL- 18) and mitochondrial dynamics-related proteins (p-Drp1^S579) activity 24 h after reoxygenation, with decrease of mitofusin related protein Mfn2. Inhibition microglial PEA degradation with Naaa knockdown (named PEA-enriched microglia) remarkably attenuated protein levels of NLRP3, Caspase1, IL- 1β, IL- 18 and p-Drp1^S579, with increase expression observed in Mfn2 (Fig. 8A). These findings indicated that the activity of IL- 18 related pathway in microglia was suppressed at elevated concentration of PEA. Furthermore, IL- 18-related upstream pathways, such as mitochondrial fragmentation and excessive inflammation, are intricately connected to the maintenance of mitochondrial membrane homeostasis [57, 58]. Lipid metabolism, particularly the metabolites of palmitoylethanolamide, alters membrane lipid saturation regulating fluidity and permeability, which are crucial for preserving membrane homeostasis, including that of the mitochondrial membrane. Therefore, we examined the effects of palmitoylethanolamide metabolites on the integrity of the mitochondrial membrane [59–61]. We then evaluated the mitochondrial membrane potential (MMP) by JC- 1 and mPTP, which can reflect the mitochondrial membrane integrity as well as mitochondrial function. The degree of mitochondrial membrane fragmentation is negatively correlated with the mitochondrial membrane potential. Throughout the OGD exposure, intracellular JC- 1 monomer of microglia significantly increased, representing a decline in MMP (Fig. 8B). Given that the decrease in MMP is largely attributed to the dysfunction of the mitochondrial permeability transition pore (mPTP), we evaluated mPTP opening using calcein-AM and CoCl₂. After OGD/R modeling, there was a significant increase in mPTP opening in microglia, however, PEA-enriched microglia can reverse this change (Fig. 8C and D). Extramitochondrial ROS serve as a marker of mitochondrial damage and inflammation activation. So, we used DCFH-DA probe to measure ROS levels among groups. Significant release of ROS from mitochondria was induced by OGD/R, while PEA-enriched microglia were characterized by alleviated excessive ROS release (Fig. 8E and F). In vitro experiments demonstrated that sustaining PEA pool content in microglial plays a role in mitigating IL- 18 mediated hyperinflammation, preserving the balance between mitochondrial fusion and division, and reducing mitochondrial fragmentation and damage.

Fig. 8.

Naaa knockdown modulates microglia phagocytosis via IL- 18 in vitro. A Representative western blot images of the mitochondrial dynamic proteins and NLRP3 inflammasome associated proteins in primary microglia under indicated treatments (n = 5). Samples were collected after OGD/R modeling (6 h OGD followed by 24 h reperfusion). B Representative images of mitochondrial membrane potential (MMP) detected by JC- 1 under different treatment after OGD/R modeling. Green for JC- 1 monomer (top) and red for JC- 1 aggregate (middle) (scale bars: 50 μm). C, D mPTP opening measured by Calcein-AM and CoCl₂ in different treatment groups and detected by flow cytometry after OGD/R modeling (C). Quantification of the relative mPTP openness levels (D) (n = 5). E, F ROS level measured by DCFH-DA staining and detected by flow cytometry in different treatment groups after OGD/R modeling (E), and quantification of the relative intensity (F) (n = 5). G, H Primary microglia were pre-treated with 40 ng/mL IL-18BP or vehicle for 24 h, followed by a 12-h incubation in either CM or control media for starvation. Fluorescent latex beads were added into medium and cells were collected after 2 h. Flow cytometric analysis (G) and quantification (H) of microglial phagocytic activity using fluorescent microspheres (n = 5). I, J Representative western blot images of the indicated proteins in primary microglia under different treatments (I), and quantification analysis (J) (n = 5). Data were collected across 5 independent experiments, each including three replicate wells. The data is shown as mean ± SD. One-way ANOVA with Bonferroni post hoc test (D, F, H, J); ns, no significance; *p < 0.05, **p < 0.01, ***p < 0.001

Previous studies have shown that the inflammatory response is not limited to the site of injury but can extend to the contralateral hemisphere, affecting neuronal function and recovery outcomes [62]. To simulate the transfer process of inflammatory factors from the lesion to contralateral hemisphere, we collected the condition medium (CM) derived from PEA-enriched and OGD/R treated microglia, and added it to new dish of microglia pretreated with 40 μg/mL IL- 18BP or vehicle for 24 h. CM from OGD/R treated microglia significantly increased phagocytosis of latex beads by primary microglia (Fig. 8G and H). The spleen tyrosine kinase (SYK) is regarded to instruct signaling and effector functions downstream of TREM2, CD33, and CD22 receptors, which are critical for the phagocytic activity of microglia [63]. Notably, residue Tyr525 is located in the activation loop of SYK and its phosphorylation is important for induced SYK activation [64]. Phosphorylation of SYK at T519 in mouse is equivalent to T525 in human [65], which couples the activated immunoreceptors to downstream signaling events including phagocytosis, proliferation and differentiation [63, 66, 67]. Our outcome indicated that CM from OGD/R treated microglia can significantly induce an increase in SYK and p-SYK^T519, while PEA-enrichment and IL- 18BP pretreatment alleviated the activation of SYK (Fig. 8I and J). So, phagocytosis activation would decrease when using the CM from PEA-enriched microglia, and further suppressed after IL- 18BP treatment. Altogether, PEA non-depletion in microglia contribute to decrease IL- 18 pathway activity, maintain mitochondrial membrane homeostasis, and reduce activated microglial phagocytosis function.

Discussion

It has been found that stroke is associated with aberrant lipid metabolism, but the causes and consequences of these metabolic changes remain unclear. In this study, we found that PEA, as one of lipid derivatives belonged to endocannabinoids, is exhausted in activated microglia because of elevated expression of specific hydrolytic enzyme NAAA. And, we observed that the expression of NAAA in the ischemic penumbra region directly relates to the susceptibility to post-stroke anxiety. Microglial PEA exhaustion leads to more severe pathological damage, anxiety-like behaviors and pain sensitivity. Boosting PEA pool content in ischemic lesion can effectively promote stroke recovery and relief anxiety symptoms. In addition, PEA helps to confront against microglial mitochondria dysfunction, and inflammasome cascade triggered IL- 18 release and diffusion to contralateral hemisphere. Microglial PEA pool content maintenance in ischemic damaged lesion can safeguard contralateral vCA1 synaptic integrity, enhancing anxiolytic pBLA-vCA1 ^Calb1+ circuit activity by relieving microglial phagocytosis mediated synaptic loss. Thus, our studies in this paper demonstrate that microglial NAAA-regulated lipid signaling in the ischemic focus remodels contralateral anxiolytic circuit to participate in post-stroke anxiety progression. And targeting NAAA might provide an opportunity to alleviate stroke-induced excessive inflammation and anxiety-like behaviors.

Post-stroke mood disorders, such as anxiety, depression and posttraumatic stress disorder, are commonly associated with worse functional recovery, poorer return-to-work rates and quality of life. The previous researches focused on post-stroke energy metabolism in the acute phase, such as oxidative stress and glucometabolic [68, 69]. However, post-stroke anxiety is a long-term process, and the abundance of various metabolites undergoes changes [70, 71], particularly adaptive alterations in endocannabinoids metabolism [72, 73]. Endocannabinoids have been reported to guard against fear, anxiety and stress but has not been clearly elucidated in post-stroke mood disorders. Our results manifested that PEA, as a cannabinoid-like compound, is exhausted rapidly after stroke, which is consistent with previous clinical cases research [6]. PEA interacts with various molecular targets within both the central and peripheral nervous systems, and is recognized for its significant anti-inflammatory, neuroprotective, and analgesic properties [74–77]. PEA exhaustion occurs in several inflammatory diseases, accompanied by increased NAAA availability [24, 77, 78]. Sensory neurons, macrophages and other host-defense cells generate PEA in amounts. Following tissue damage, danger signals initiate a molecular program that lowers PEA production and heightens NAAA-catalyzed PEA degradation. Consistent with this scenario, stroke itself is one of a pathological stress disease. Despite employing a distinct stroke model, we demonstrated that NAAA expression dynamically increased in activated microglia of the ischemic penumbra region but not in astrocytes, which aligns with the single-cell RNA-seq observations from Zheng et al.’s study [25]. Pharmacological inhibition or genetic deletion of NAAA is reported to increase PEA and AEA levels in tissue microenvironment, thereby mitigating excessive inflammation and delaying disease progression [76, 79]. In line with these observations, we found that Naaa knockout or exogenous PEA supply both alleviates pathological damage and anxiety-like behaviors after stroke. Cellular functional analysis revealed that microglial activation under hypoxic conditions was suppressed in Naaa-deficient models or with PEA treatment. This aligns with D'Aloia et al.'s findings demonstrating PEA-mediated inhibition of pro-inflammatory microglial phenotype switching and prevention of LPS-induced Ca²⁺ transients [80].

The inflammatory mediators released by activated microglia can remodel remote brain microcircuits and participate in emotion regulation [16, 43]. The hippocampus, easily disturbed during the courses of stroke, contains distinct functional regions along its longitudinal axis, with the dorsal hippocampus mainly involved in memory and cognition. While the ventral hippocampus primarily is associated with stress and anxiety [31, 81]. A population of excitatory pyramidal neurons in the vCA1 termed"anxiety cells"has been identified to encode anxiety-related behaviors. Studies demonstrate that the number of c-Fos-positive neurons in vCA1 decreases in conditions such as chronic inflammatory pain and depression-like behaviors [32, 82], which is also observed in contralateral vCA1 region under ischemic stroke. pBLA-vCA1^Calb1+ circuit can be specifically activated by anxiogenic environments and play a crucial role in avoidance behaviors. Prior Studies revealed that environmental and external stress can lead to synaptic integrity defects and subsequent behavioral alterations correlated with anxiety, depression and pain [16, 31, 32]. Our results elucidated that both dendritic complexity and neuronal excitability decreased in contralateral vCA1 after stroke. Microglia, as key regulators of neural circuit activity and synaptic plasticity, may mediate these changes [43]. Specifically, IL- 18 released by activated microglia in ischemic damaged lesion can promote contralateral vCA1 microglia activation and phagocytosis activity, resulting in dendritic complexity reduction, influencing anxiolytic pBLA-vCA1^Calb1+ circuit activity, leading to post-stroke anxiety [15, 16]. In the current study, we show that in the contralateral vCA1 region, microglia undergo morphological activation and cause excitatory synapse structure disruption mediated by synaptic engulfment after stroke. These phenotypes were partially reversed at some extent under Naaa knockout or PEA pharmacologic supply.

Emerging evidence suggests that PEA modulates cannabinoid receptor type 1 (CB1) and type 2 (CB2) through indirect mechanisms, such as the"entourage effect"by inhibiting FAAH to elevate endogenous anandamide levels [83]. CB2R is abundantly expressed in microglia [84] and mediate PEA's suppression of microglial phagocytosis and neuroinflammation [85], phenocopying with PPAR-α activation [78]. Activation of PPAR-α has been shown to ameliorate stress-induced anxiety behaviors and exhibit a dose-dependent antidepressant effect [86]. While CB1 receptors (CB1R) on GABAergic and glutamatergic neurons regulate anxiety via presynaptic inhibition of neurotransmitter release [87, 88]. In parallel, PEA engages other receptors including transient receptor potential vanilloid 1 (TRPV1) and G-protein-coupled receptor 55 (GPR55) to modulate inflammatory responses in monocytes and microglia, which are implicated in stress and anxiety pathways [89, 90]. So, we cannot rule out direct effects of these receptors during the courses of post-stroke anxiety. Naaa knockout or PEA administration may affect anxiety not only via microglial phagocytosis but also potentially through mechanisms involving other receptors that function independently of microglial activity.

Our study demonstrates PSA arises from dysfunction in anxiety-regulating neural circuits following ischemic stroke. Specifically, IL- 18 can promote phagocytosis of myeloid cells [15, 91, 92]. Naaa knockout or PEA intervention significantly reduced IL- 18 levels in the contralateral vCA1 region, suggesting that ischemic core-derived IL- 18 diffused to this region, triggering microglial overactivation, excessive synaptic pruning, and suppression of the anxiolytic pBLA-vCA1 Calb1⁺ circuit. To preliminarily verify the diffusion of IL- 18, we employed an in vitro model: conditioned medium from OGD-treated microglia activated resting microglia and enhanced phagocytosis, which can be blocked by IL- 18BP. While IL- 18 may spread via extracellular vesicles or BBB disruption post-stroke, there are no established animal models or standardized paradigms to validate these pathways. Definitive characterization of IL- 18’s intracerebral transport will likely require integrated approaches such as fluorescent tracing, BBB permeability assays, and transgenic technologies [93].

In this work, we observed evidence of endocannabinoid dysfunction coexist in microglia after stroke, and identified mechanisms involved. Due to the use of global Naaa knockout mice rather than microglia conditional knockout mice, we cannot exclude the potential impact of Naaa knockout in neurons or astrocytes on the pathological processes of PSA [94]. Research indicates that inhibition of NAAA in astrocytes can alleviate secondary damage following traumatic brain injury [75]. Thus, the role of astrocytic NAAA in the courses of stroke may be critical and cannot be ignored easily. Peripheral myeloid cells can cooperate with microglia to promote pathological changes. For example, NAAA in monocytes has been implicated in the development of hyperalgesia [10]. Similarly, as a myeloid cell, the expression of NAAA in microglia shows a significantly reactive increase following stroke [95]. But whether NAAA has non enzymatic functions involved in the regulation of signaling pathways has not been explored neither. Furthermore, additional evidence is required to investigate whether the concentration of PEA in the serum can serve as an indicator for evaluating prognostic recovery. Although this study has several limitations, what can be confirmed is that Naaa knockout or PEA treatment compromised microglial phagocytosis mediated synaptic loss to counteract post stroke anxiety. This might be attributed to the comprehensive effects of multiple receptors and mechanisms. These findings bridges mechanistic exploration of PSA with translational potential, offering both conceptual innovation and preclinical evidence for targeted drug development.

Authors’ contributions

SW and HRZ conceived and designed the project. TYY, SJS, LP, MMY, MYL, XLY and FYY conducted all experiments. TYY and SJS performed a statistical analysis. TYY wrote the manuscript. HRZ drew revised the manuscript. SW acquired funding and reviewed the manuscript. All authors have read and approved the final version of the manuscript.

Funding

National Natural Science Foundation of China, 82272225. National Natural Science Foundation of China, 81860249. USTC Research Funds of the Double First-Class Initiative YD9110002084.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

All experiments complied with standards established by the Institutional Animal Care and Use Committee of USTC and were approved by the Experimental Animal Center of the First Affiliated Hospital of USTC (No. 2022-N(A)- 175).

Competing interests

The authors declare no competing interests.