Astrocytic Netrin-1 modulates stress-induced anxiety susceptibility in mice through ADAM17-mediated TNFα shedding

Abstract

Anxiety disorder is one of the most prevalent psychiatric diseases, with stress being a high-risk factor that functions by inducing neuroinflammation, particularly through the release of tumor necrosis factor α (TNFα). Although stress-induced transcription of TNFα is well documented, whether it also controls the proteolytic maturation of membrane-anchored TNFα precursor (proTNFα) remains unknown. Here we showed that the astrocyte-derived extracellular protein Netrin-1 was highly expressed in the mouse medial prefrontal cortex (mPFC) and down-regulated by restraint stress. Conditional knockout (cKO) of astrocytic Netrin-1 in astrocytes, both in embryo and adult, increased susceptibility to stress-induced anxiety. Mechanistically, Netrin-1 interacted with A disintegrin and metalloprotease 17 (ADAM17) and blocked ADAM17-mediated shedding of TNFα from proTNFα. Consequently, Netrin-1 deficiency elevated soluble TNFα and altered microglial morphology. Pharmacological neutralization of TNFα with Infliximab normalized anxiety-like phenotypes in stressed Netrin-1 cKO mice. Collectively, our data identify Netrin-1 as a critical brake on TNFα release and position the Netrin-1–ADAM17 axis as a tractable therapeutic target for TNFα-driven neuropsychiatric disorders.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-026-03711-8.

Introduction

Anxiety disorder is a prevalent and disabling mental disease, characterized by excessive fear, worry, and disturbances in emotional regulation [1]. The etiology of anxiety disorder involves complex interactions between environmental influences and physiological mechanisms. Among these factors, stress is a well-established risk factor that activates the hypothalamus‒pituitary‒adrenal (HPA) axis and promotes the release of glucocorticoids (GCs). Elevated GCs trigger neuroinflammation and promote the release of proinflammatory cytokines, underlying the pathogenesis of anxiety disorder [2, 3]. Recent studies have revealed that astrocytes express high levels of glucocorticoid receptor (GR), making them highly sensitive to stress and positioning them as key mediators of stress-induced anxiety and depression-like behaviors [4].

Tumor necrosis factor α (TNFα) is a key proinflammatory cytokine involved in anxiety disorder [5, 6]. TNFα is initially synthesized as a transmembrane precursor (proTNFα) and then released via ectodomain shedding catalyzed by A disintegrin and metalloprotease 17 (ADAM17), also known as TNFα converting enzyme (TACE) [7]. Behavioral studies have demonstrated that TNFα transgenic mice exhibit significantly increased anxiety-like behaviors [8], whereas TNFα knockout mice display reduced anxiety levels [9]. Furthermore, direct administration of TNFα, either through intraperitoneal or ventricular injection, induces anxiety-like behaviors in rodents [10]. In contrast, the inhibition of TNFα signaling with Infliximab or XPro1595 reduces acute and chronic mild stress-indued anxiety phenotypes [11–13]. While it is well established that environmental stress enhances TNFa transcription [13–15], the specific mechanisms underlying the regulation of TNFα shedding in response to stress remain largely unexplored.

Netrin-1 is an extracellular guidance protein that plays essential roles in neuronal migration [16], axon projection [17] and synapse formation [18]. Accumulating evidence suggest that Netrin-1 signaling is closely associated with anxiety and depression, two emotional disorders with a high rate of co-occurrence [19]. A genome-wide association study has identified correlations between Netrin-1 polymorphisms and anxiety-related traits [20]. Another genome-wide methylation study of monozygotic twins reveals altered Netrin-1 methylation in individuals with depression [21]. In addition, the serum levels of Netrin-1 are significantly lower in depressed patients who have suffered a stroke, compared to those who are not depressed [22]. Others have also shown that Netrin-1 receptor, deleted in colorectal cancer (DCC) contributes to the risk for anxiety- and depression-like behaviors [23, 24]. Despite these correlative findings, the direct role of Netrin-1 in anxiety and depression and the underlying mechanisms remain unclear.

In this study, we showed that Netrin-1 was highly expressed in astrocytes in the medial prefrontal cortex (mPFC) and was downregulated by restraint stress via GC/GR signaling. Deletion of Netrin-1 in astrocytes significantly increased susceptibility to stress-induced anxiety in mice. Furthermore, mechanistic studies revealed that Netrin-1 interacted with ADAM17 and suppressed ADAM17-mediated shedding of TNFα from proTNFα. Netrin-1 deficiency resulted in increased TNFα release and morphological changes in microglia. Importantly, blocking the TNFα signaling with Infliximab alleviated anxiety-like behaviors in stressed Netrin-1 cKO mice. These results reveal novel functions of Netrin-1 in regulating TNFα release and stress-induced anxiety behaviors, highlighting its potential as a therapeutic target for TNFα-related neuropsychiatric conditions.

Materials and methods

Animals

Netrin-1-LacZ knock in mice (Netrin-1^LacZ/+) and Netrin-1 floxed mice (Netrin-1^f/f) were kindly provided by Yu-Qiang Ding (Fudan University, Shanghai, China) [25]. Aldh1L1-Cre^ER mice were obtained from Jackson Laboratory (Cat. No 031008). GFAP-Cre mice were initially obtained from Jackson Laboratory (Cat. No 024098) [26]. Mice carrying Netrin-1^f/f alleles together with either GFAP-Cre or Aldh1L1-Cre^ER were referred to as Netrin-1 cKO^GFAP or Netrin-1 cKO^Aldh1L1, respectively; their Netrin-1^f/f littermates without Cre served as controls and were designated wild-type (WT). The mice were housed under 12 h light/dark cycle conditions with food and water ad libitum. Both male and female mice were used for histological and biochemical analysis. Eight- to twelve-week-old male mice were used for the behavioral tests. All experiments were approved by the Institutional Animal Care and Use Committee of Northeast Normal University.

Netrin-1^LacZ/+ mice were genotyped with forward primer (5’-TCC TCG CAG TTT GCT TAG TG-3’) and reverse primer (5’-CAC AAC GGG TTC TTC TGT TAG TCC-3’).

Netrin-1^f/f mice were genotyped with forward primer (5’-TCC TCG CAG TTT GCT TAG TG-3’) and reverse primer (5’-CACAACGGGTTCTTCTGTTAGTCC-3’).

GFAP-Cre mice were genotyped with forward primer (5’-ACT CCT TCA TAA AGC CCT-3’) and reverse primer (5’-ATC ACT CGT TGC ATC GAC CG-3’).

Aldh1L1-Cre^ER mice were genotyped with primer 1 (5’-GAG TCC TGG AAT CAG TCT TTT TC-3’), primer 2 (5’-ATC ACT CGT TGC ATC GAC CG-3’) and primer 3 (5’-CCG CAT AAC CAG TGA AAC AG-3’).

Regents

For immunostaining, the following primary antibodies were used: mouse anti-Netrin-1 (1:500, R&D Systems, MAB1109-100), mouse anti-NeuN (1:1000, Millipore, MAB377), rabbit anti-S100β (1:100, Abcam, ab52642), chicken anti-β-galactosidase (1:2000, Abcam, 9361), rabbit anti-Iba1 (1:500, Wako, 019–19741), and mouse anti-Myc (1:200, Abcam, ab32). For immunoblotting, the following primary antibodies were used: mouse anti-Netrin-1 (1:1000, Enzo, ALX-804–838), rabbit anti-GFAP (1:20000, Abcam,7260), mouse anti-β Ⅲ Tubulin (1:1000, Abcam, 7751), mouse anti-GR (1:500, Abcam, ab2768), rabbit anti-Caldesmon (1:200, Santa Cruz Biotechnology, sc-15374), rabbit anti-ADAM17 (1:500, Abcam, ab2051), anti-mouse GAPDH (1:3000, TransGen Biotech, HC301-01), mouse anti-β-Actin (1:3000, Sigma Aldrich, A1978), mouse anti-Myc (1:1000, ZSGB-BIO, TA-01), mouse anti-Flag (1:1000, Sigma Aldrich, F7425). Alexa Fluor 488-, 546-, and 633-coupled secondary antibodies against mouse, rat, chicken, or goat and HRP-conjugated secondary antibodies against mouse or rabbit were purchased from Invitrogen.

Recombinant murine TNFα was purchased from PeproTech (315-01A-5UG). Dexamethasone (DEX) and Tamoxifen were purchased from Sigma‒Aldrich (D4902 and T5648, respectively). Infliximab was purchased from Mabpharm Limited (CMAB008).

Plasmids

The GFP-proTNFα-RFP plasmid was provided by Dr. Baur [27]. The expression vectors encoding Netrin-1 were generated as described previously [28]. The ADAM17 expression vector was obtained from Addgene (#31713, Addgene, MS, USA) [29].

Cell culture

Primary neurons were cultured as described previously [17]. In brief, embryos (E17) were removed from anesthetized pregnant mice. Cerebral cortices were isolated, chopped and then digested in 0.125% trypsin plus 0.05% DNase at 37 °C for 20 min. Single cells were obtained by mechanical pipetting and filtering. Dissociated cells were suspended in DMEM supplemented with 10% FBS and plated on poly-L-lysine-coated dishes. After 4 h, the medium was changed to Neurobasal medium supplemented with 2% B27 supplement and 2 mM GlutaMAX.

Primary astrocytes were prepared from the cerebral cortex of P1‒P2 neonatal mice. Briefly, cerebral cortices were isolated, chopped, and incubated with 0.125% trypsin at 37 °C for 20 min. Single cells were obtained by pipetting, seeded on poly-L-lysine coated dishes and incubated in DMEM containing 10% FBS. After 4 h, the medium was changed to remove unattached oligodendrocyte precursor cells.

HEK293 cells were grown in DMEM supplemented with 10% FBS and 100 U/ml penicillin‒streptomycin. For the co-immunoprecipitation and pull-down assays, HEK293 cells were transfected with polyetherimide (PEI). Netrin-1 was prepared as previously described [54].

Real-time PCR

Total RNA was extracted via TRIzol reagent (Invitrogen). Complementary DNA was synthesized via Tran Script One-Step gDNA Removal and cDNA Synthesis Super Mix (AT-311–03, TransGen Biotech, Beijing, China). Real-time PCR was performed with SYBR Premix Ex Taq™ II (RR820A, Takara) via a 96-well PikoReal Real-Time PCR System (TCR0096, Thermo Fisher). Each sample was analyzed in triplicate, and the mRNA levels were normalized to the GAPDH mRNA levels with the 2^−ΔΔCT method. Melting curves were automatically recorded and analyzed for each reaction. The primers used were as follows:

Gapdh (Mus musculus), forward 5’-ATGACCCCTTCATTGACCTCA-3’ and reverse 5’-GAGATGATGACCCTTTTGGCT-3’; Tnfa (Mus musculus), forward 5’-CCGATGGGTTGTACCTTGTC-3’ and reverse 5’-AGATAGCAAATCGGCTGACG-3’.

Western blotting

Tissue or cell extracts were prepared with RIPA lysis buffer (50 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA (pH 8.0), 1 mM EGTA (pH 8.0), 1% NP40, 2.5 mM sodium pyrophosphate, 1 mM β-phosphoglycerol, 2 mM NaF, and 1 mM Na3VO4), to which 1 mM PMSF and protease inhibitor cocktail (Roche) were added before use. Individual samples were mixed with 5 × SDS loading buffer and then boiled for 5 min at 100 °C. Western blotting was performed according to a conventional method. The blots were visualized with High-sig ECL Western Blotting Substrate (Biotanon) and a Tanon 5200 Automatic Chemiluminescence Image Analysis System (Biotanon).

Immunofluorescence

The mice were anesthetized with sodium pentobarbital and transcardially perfused with PBS (0.1 M), followed by 4% PFA. The brains were cryoprotected in 30% (w/v) sucrose for 72 h, embedded in optimal cutting temperature (OCT) compound and sectioned at a thickness of 40 μm. The slices were rinsed in PBS and subjected to antigen retrieval in 0.01 M sodium citrate buffer (pH 6.0) at 95 °C for 5 min. To block nonspecific antibody binding, the slices were incubated with 5% BSA in PBST. Both primary and secondary antibodies were diluted in the same blocking solution, and the sections were incubated with primary antibodies at 4 °C overnight and with Alexa-conjugated secondary antibodies at room temperature for 1 h. Fluorescence intensity was quantified with ImageJ as follows: RGB images were converted to 8-bit and the channels were split. For each channel, a region of interest (ROI) was drawn around the cells using the free-hand tool; the same ROI was saved and transferred to the target channel. Integrated density (IntDen) was measured after background subtraction and normalized to the mean of the control group.

Microglia morphological analysis

To quantify microglial morphology, brain sections were labeled with Iba1 and imaged as confocal z-stacks. After binarization and skeletonization in ImageJ, morphometric data were extracted with the AnalyzeSkeleton (2D/3D) plugin, which reports the total branch length of all microglial processes within the field [30]. The average process length per cell was then calculated by dividing this total length by the number of microglia in the same field [30].

ELISA

The mPFC regions of stressed WT and stressed Netrin-1 cKO^Aldh1L1 mice were isolated and homogenized. After centrifugation, the supernatant was collected, and the TNFα levels were quantified with a mouse TNFα ELISA kit (JL10484, JONLNBIO, Shanghai, China) following the manufacturer’s instructions. The absorbance was measured via a microplate reader at 450 nm.

RNA sequencing

After 3 days of restraint stress, the mPFC regions of WT and Netrin-1 cKO^Aldh1L1 mice were isolated, and RNA was extracted with TRIzol reagent (Invitrogen). cDNA library construction and sequencing were subsequently performed by BGI Company, China. Three groups of RNA samples (3 for each genotype) were sequenced. The RNA-seq reads were mapped to the mouse genome (GRCm38.p6) with HISAT2 (version 2.1.0). The mapped reads of each gene were counted with RSEM (v1.3.1). Genes differentially expressed in the WT and Netrin-1 cKO^Aldh1L1 groups were obtained via DESeq2 (v1.4.5) with a Q value < 0.05. Gene set enrichment analysis (GSEA) was performed via the BGI Dr. Tom website.

Re-analysis of published scRNA-seq data

Publicly available single-cell RNA-sequencing datasets from the mouse prefrontal cortex (GSE240975 and GSE124952) [31, 32] were integrated and analyzed using Seurat (v5.0.0) in R. Following quality control, normalization, and data integration, cells were clustered in a low-dimensional space. Cell types were annotated based on the expression of established canonical marker genes. The dplyr package was used to evaluate marker gene expression. The following markers were applied for annotation: Astrocytes (Mlc1, Gjb6, Aqp4, Acsbg1), Endothelial cells (Flt1, Cldn5, Tm4sf1, Pglyrp1, Rgs5), Excitatory neurons (Neurod6, Slc17a7, Nrn1), Inhibitory neurons (Gad1, Gad2, Pnoc, Slc32a1), Microglia (C1qa, C1qb, C1qc, Ctss), NF-Oligodendrocytes (Neu4, Pak4, Enpp6, Brca1), and Oligodendrocytes (Ermn, Opalin, Mog, Aspa, Mobp). The expression of Adam17 and Tnfα across the annotated cell types was visualized using DotPlot plots from ggplot2.

Protein array assay

The mPFC regions of stressed WT and stressed Netrin-1 cKO^Aldh1L1 mice were isolated and homogenized. After centrifugation, the lysates were collected and the total protein was adjusted to the same concentration. The levels of inflammatory factors were quantified with a Mouse Inflammation Array C1 kit (AAM-INF-1, RayBiotech, Norcross, USA) following the manufacturer’s instructions. In brief, the nitrocellulose membranes containing 40 different antibodies were incubated with the above lysates at 4 °C overnight, followed by biotinylated antibody incubation and HRP-streptavidin incubation, and visualized with a Tanon 5200 Automatic Chemiluminescence Image Analysis System (Biotanon).

Co-immunoprecipitation

Myc-ADAM17 and Flag-Netrin-1 were co-expressed in HEK293 cells. After 48 h, the cells were collected and lysed with modified RIPA buffer (50 mM Tris–HCl, pH 7.4; 150 mM sodium chloride; 1% NP-40; 0.25% sodium-deoxycholate; and proteinase inhibitors). The cell lysates were incubated with an anti-Flag antibody at 4 °C for 4 h. Then, the cell lysates were incubated with protein A-G-agarose beads at 4 °C for 1 h. The immunoprecipitated complexes were collected by centrifugation and washed. The immune complexes were resolved by SDS‒PAGE and subjected to Western blotting.

Pull-down assay

Myc-ADAM17 and HA-Netrin-1 were overexpressed in HEK293 cells. Cell lysates were incubated with anti-Myc magarose beads (KTSM1336, AlpaLifeBio, Shenzhen, China) or anti-HA magarose beads (KTSM1335, AlpaLifeBio, Shenzhen, China). After washing, HA-Netrin-1 was eluted with 200 mM glycine (pH 2.5), and the solution was subsequently neutralized with 1 M Tris–HCl (pH 10.4). For the pull-down of Myc-ADAM17 and HA-Netrin-1, 10 μg of HA-Netrin-1 was incubated with anti-Myc-beads or anti-Myc-beads-ADAM17, respectively, overnight at 4 °C. The beads were washed three times with PBST, and protein complexes were resolved by SDS‒PAGE and subjected to Western blotting.

Stereotactic surgeries

For virus injection, anesthetized mice were secured in a stereotaxic apparatus. Holes were drilled into the skull and viruses (titer > 10¹² particles/ml, 200 nl, each side) were microinjected into the mPFC using calibrated glass microelectrodes connected to an infusion pump at a rate of 30 nl/min (coordinates: AP 2.10 mm, ML ± 0.25 mm and DV −1.75 mm). After injection, the pipettes were left in place for 5 min to allow for diffusion of the injected virus before being withdrawn slowly. The mice were tested 3 weeks after injection. The injection locations were validated in each mouse after the experiments.

For TNFα or Infliximab injection, the guide cannula (200 μm indiameter, 0.37 NA, KedouBC, Suzhou, China) was implanted into the mPFC region (coordinates: AP 2.10 mm, ML ± 0.25 mm and DV −1.75 mm). Following surgery, mice were allowed to recover for at least 7 days. TNFα (2.5 pg/μl) was bilaterally infused 24 h before the behavioral tests, with an infusion volume of 200 nl for each side. Infliximab (2 μg/μl) was bilaterally infused once a day after restraint stress for 3 consecutive days, with an infusion volume of 250 nl for each side.

Tamoxifen administration

Tamoxifen was dissolved in corn oil at a concentration of 20 mg/ml. Male mice were i.p. injected with tamoxifen solution once a day at a dose of 75 mg/kg body weight for 5 consecutive days. Littermate male WT mice injected with tamoxifen were used as controls. Behavioral tests were conducted 10 days after the last tamoxifen injection.

Restraint stress

Mice were subjected to restraint stress by being placed in well-ventilated plastic conical tubes for 4 h daily (from 12:00 to 16:00) over three consecutive days. The cylinders, with an inner diameter of 3 cm and a length of 10 cm, restricted the animals' movements, rendering them nearly immobile. Samples were collected 2 h after the last restraint stress.

Behavioral tests

Open field test (OFT)

Mice were placed in a chamber (40 * 40 * 40 cm) and monitored for movement using an overhead camera and tracking software (EthoVision, Noldus). The central region (20 × 20 cm) of the arena was defined as the center. The total locomotor distance traveled in the arena and the duration spent in the central region was noted. All Netrin-1 cKO^GFAP mice and their littermate controls were tested for 30 min. Netrin-1 cKO^Aldh1l1 and their littermate controls as well as all microinjected mice were tested for 20 min.

Light and dark box test (LDT)

The LDT apparatus is composed of a chamber (45 × 26 × 26 cm) that is divided into a dark compartment (1/3 of the total area) and a light compartment (2/3 of the total area). Before testing, mice were habituated to low-light conditions (∼5 lx) for 1 h before testing. The test was conducted using a two-chamber apparatus in which one side (the dark chamber) was kept dark and the light chamber was illuminated (∼4000 lx) with a high-intensity light. Each mouse was placed in the dark chamber of the apparatus and allowed to explore for 10 min. The latency for the mouse to move from the dark chamber to the light chamber and the time spent in the illuminated chamber were recorded as indices of anxiety-like behaviors.

Elevated plus-maze (EPM)

The apparatus consists of an elevated (60 cm above the floor) four-armed (30 × 5 cm2) maze with two open arms that cross in the middle and two arms enclosed by walls (h = 10 cm). The mice were placed in the central square of the maze facing the open arms. The time spent in each arm of the EPM was recorded for 10 min.

Sucrose preference test (SPT)

The SPT was carried out in standard mouse cages containing 2 bottles [one filled with water and one filled with sucrose (1% in water)]. After 24 h of water deprivation, the mice were placed in cages with water- and sucrose-containing bottles for 24 h. Each bottle was weighed before and after the test. Depressive-like anhedonia was evaluated by calculating the sucrose preference ratio.

Tail suspension test (TST)

Mice were suspended by their tails with adhesive tape and recorded for a duration of 5 min. The immobility time was subsequently evaluated. Mice were classified as immobile only when they remained in a hanging position, completely motionless, or exhibited slight swaying that was solely attributable to breathing or residual movements from earlier struggling.

Statistical analysis

The data were analyzed using SPSS 19.0 (IBM) and Prism 8.0 (GraphPad Software). For experiments containing two independent variables (genotype and stress exposure) we used two-way ANOVA followed by post-hoc test. When only one factor was manipulated, we applied unpaired two-tailed Student’s t test or one-way ANOVA followed by post-hoc test. The data are presented as the mean ± standard error of the mean (SEM). P values < 0.05 were considered statistically significant.

Results

Astrocytic Netrin-1 in the mPFC is regulated by stress viaGC/GR signaling

To explore the potential role of Netrin-1 in anxiety and depression disorders, we first examined the distribution of Netrin-1 in the mouse brain. Western blotting revealed that Netrin-1 was abundant in the mPFC, which is a hub region for emotion processing (Fig. 1A). Next, we analyzed the cellular expression pattern of Netrin-1 in the mPFC. Since Netrin-1 is a secreted protein, we employed a Netrin-1-LacZ knock-in mouse model to map its expression pattern. Within this model, a cassette containing the IRES, LacZ and neomycin resistance (Neo) gene was inserted into introns of Netrin-1 and thus LacZ could recapitulate Netrin-1 expression under the control of Netrin-1 regulatory sequences (Fig. 1B). Immunostaining of the LacZ revealed that astrocytes (S100β⁺) expressed higher levels of Netrin-1 compared to neurons (NeuN⁺) in the mPFC (Fig. 1B-C). In addition, primary neurons and primary astrocytes were cultured, and Western blotting confirmed that astrocytes expressed more Netrin-1 than neurons in vitro (Fig. 1D).

Fig. 1.

Decreased expression of astrocytic Netrin-1 in response to stress. A Immunoblotting and quantification of Netrin-1 expression in the indicated brain regions of adult C57BL/6 mice. B Schematic diagram of the Netrin-1-LacZ knock-in mouse and representative images of LacZ expression in neurons (NeuN⁺) and astrocytes (S100β⁺) in the mPFC of adult Netrin-1-LacZ knock-in mice. C Quantification of LacZ levels in neurons and astrocytes in the mPFC. Student’s t test, P < 0.0001. D Immunoblotting and quantification of Netrin-1 expression in cultured neurons and astrocytes in culture. Student’s t test, P < 0.0001. E Schematic diagram of restraint stress. F Immunoblotting and quantification of Netrin-1 expression in the mPFC of adult C57BL/6 mice at the indicated time points of stress exposure. One-way ANOVA followed by two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli. F (4, 10) = 6.04, P = 0.0098. 0 days vs. 3 days, P = 0.0007; 0 days vs 10 days, P = 0.0189; 0 days vs 21 days, P = 0.0336; 0 days vs. 30 days, P = 0.0116. G Representative images of Netrin-1 staining in astrocytes (S100β⁺) and neurons (NeuN⁺) in the mPFC of control mice and stressed mice, respectively. The white circle indicated representative cells. H Comparison of Netrin-1 levels in astrocytes and neurons between the control group and the stress group. The level of Netrin-1 in astrocytes and in neurons in the stress group was normalized to that in the control group, respectively. Student's t test, P < 0.0001 for astrocytes; P = 0.6728 for neurons. I Schematic diagram of DEX treatment of cultured astrocytes. J Immunoblotting and quantification of Netrin-1 and GR expression in cultured astrocytes exposed to different DEX concentrations for 24 h. One-way ANOVA followed by two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli. For GR, F (3, 8) = 185, P < 0.0001. DMSO vs. DEX (10^–8 mol/l): P = 0.0003; DEX (10^–8 mol/l) vs. DEX (10^–7 mol/l): P = 0.0462; DEX (10^–8 mol/l) vs. DEX (10^–6 mol/l): P < 0.0001. For Netrin-1, F (3, 8) = 31.1, P < 0.0001. DMSO vs. DEX (10^–8 mol/l): P = 0.0121; DEX (10^–8 mol/l) vs. DEX (10^–7 mol/l): P = 0.0057; DEX (10^–8 mol/l) vs. DEX (10^–6 mol/l): P = 0.0007. For L-Cald, F (3, 8) = 26.6, P = 0.0002. DMSO vs. DEX (10^–8 mol/l): P = 0.0270; DEX (10^–8 mol/l) vs. DEX (10^–7 mol/l): P = 0.0050; DEX (10^–8 mol/l) vs. DEX (10^–6 mol/l): P = 0.0018. The data are from three different brains for each group or from three independent experiments. ns, no significant difference; *P < 0.05; **P < 0.01; ***P < 0.001

To explore the relevance of Netrin-1 to stress, we exposed male C57BL/6 mice to restraint stress for 4 h daily over periods of 3, 10, 21, and 30 days (Fig. 1E). Western blotting revealed that the expression of Netrin-1 was modestly reduced following 3 days, 10 days, 21 days and 30 days of restraint stress. Among these, the most pronounced decrease occurred at the 3-day time point (Fig. 1F). To identify the specific cell types contributing to the stress-induced reduction of Netrin-1, we detected Netrin-1 expression in astrocytes and neurons, two major cell types in the brain. Netrin-1 in astrocytes was significantly downregulated after 3 days of restraint stress, whereas Netrin-1 in neurons was not changed (Fig. 1G-1H).

To confirm the decrease of astrocytic Netrin-1 was caused by stress through the HPA axis and GC, we treated primary astrocytes (DIV 8) with dexamethasone (DEX), an agonist of GC, to mimic the stress conditions in vitro (Fig. 1I). Primary astrocytes were treated with 10^–8, 10^–7 or 10^–6 mol/l DEX for 24 h, covering a concentration range from basal to stress levels [4]. Relative to the DMSO control, baseline DEX (~ 10^–8 mol/l) elevated the expression of GR and Netrin-1 (Fig. 1J). Raising the concentration to an intermediate level (10⁻⁷ mol/L) further increased both proteins, whereas stress-level DEX (~ 10^–6 mol/l) strikingly suppressed it, a typical characteristic of low-dose stimulation and high-dose inhibition (Fig. 1J) [33]. L-Caldesmon (L-Cald), a well-established transcriptional target of the GC/GR pathway [34], followed the same expression trend as GR, supporting the functional engagement of GC/GR pathway. These results demonstrate that astrocytic Netrin-1 is tightly regulated by stress via GC/GR signaling, suggesting the important role of Netrin-1 in response to environmental stress.

Conditional knockout of astrocytic Netrin-1 increases anxiety susceptibility

To evaluate the role of astrocytic Netrin-1 in anxiety and depression disorders, we crossed Netrin-1^f/f mice with GFAP-Cre mice, allowing the selective deletion of Netrin-1 in astrocytes from the embryonic stage (Fig. 2A). The knockout specificity and efficiency were verified by immunofluorescence and Western blotting (Fig. 2B-E). Subsequently, a series of behavioral tests were performed, and there were no significant differences between Netrin-1 cKO^GFAP mice and WT littermates in spontaneous motility (Fig. 2F-G), anxiety-like behaviors (Fig. 2H-L) or depression-like behaviors (Fig. 2M-P), as evaluated by the open field test (OFT), light and dark box test (LDT), elevated plus maze (EPM), tail suspension test (TST), or sucrose preference test (SPT). These results suggest that Netrin-1 deficiency in astrocytes cannot directly result in observable anxiety or depression-like phenotypes.

Fig. 2.

Increased susceptibility to anxiety in Netrin-1 cKO^GFAP mice. A Generation of Netrin-1 cKO^GFAP mice. B Schematic diagram showing the experimental timeline. C Co-staining of Netrin-1 and S100β in the mPFC of WT and cKO^GFAP mice. D Quantification of Netrin-1 levels in astrocytes (S100β⁺) in the mPFC of WT and cKO^GFAP mice. Student’s t test, P < 0.0001. E Western blotting confirming the decrease of Netrin-1 in the mPFC of Netrin-1 cKO^GFAP mice and quantification of Netrin-1 levels in the mPFC of WT and Netrin-1 cKO^GFAP mice. Student’s t test, P = 0.0013. F Schematic diagram of the open field test. G Quantification of the total distance travelled in the open field chamber. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 36) = 0.117, P = 0.7342; main effect of stress: F (1, 36) = 1.22, P = 0.2760; main effect of genotype: F (1, 36) = 0.0041, P = 0.9491. WT vs. cKO^GFAP: P = 0.9916; WT vs. WT + Stress: P = 0.7366; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.9486; WT + Stress vs. cKO^GFAP + Stress: P = 0.9973. H Quantification of center duration. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 36) = 4.05, P = 0.0517; main effect of stress: F (1, 36) = 19.3, P < 0.0001; main effect of genotype: F (1, 36) = 6.26, P = 0.0171. WT vs. cKO^GFAP: P = 0.9856; WT vs. WT + Stress: P = 0.3490; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.0004; WT + Stress vs. cKO^GFAP + Stress: P = 0.0148. I Schematic diagram of the light and dark box test. J Quantification of the time spent in the dark chamber. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 36) = 5.85, P = 0.0208; main effect of stress: F (1, 36) = 10.1, P = 0.0031; main effect of genotype: F (1, 36) = 2.63, P = 0.1134. WT vs. cKO^GFAP: P = 0.9424; WT vs. WT + Stress: P = 0.9501; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.0019; WT + Stress vs. cKO^GFAP + Stress: P = 0.0340. K Schematic diagram of the elevated plus maze test. L Quantification of the time spent in the open arm. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 36) = 2.21, P = 0.1460; main effect of stress: F (1, 36) = 11.0, P = 0.0021; main effect of genotype: F (1, 36) = 9.54, P = 0.0039. WT vs. cKO^GFAP: P = 0.6719; WT vs. WT + Stress: P = 0.5755; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.0088; WT + Stress vs. cKO^GFAP + Stress: P = 0.0138. M Schematic diagram of the tail suspension test. N Quantification of the time spent immobile in the tail suspension test. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 36) = 0.0934, P = 0.7617; main effect of stress: F (1, 36) = 0.940, P = 0.3388; main effect of genotype: F (1, 36) = 1.11, P = 0.2996. WT vs. cKO^GFAP: P = 0.7725; WT vs. WT + Stress: P = 0.9653; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.8040; WT + Stress vs. cKO^GFAP + Stress: P = 0.9517. O Schematic diagram of the sucrose preference test. P Quantification of sucrose preference ratio. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 36) = 0.0324, P = 0.8582; main effect of stress: F (1, 36) = 1.60, P = 0.2136; main effect of genotype: F (1, 36) = 0.0280, P = 0.8681. WT vs. cKO^GFAP: P = 0.9947; WT vs. WT + Stress: P = 0.8683; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.7375; WT + Stress vs. cKO^GFAP + Stress: P > 0.9999. For immunoblotting, the data are from three independent experiments. For the behavioral tests, the number of mice (n) used is indicated in the graphs. ns, no significant difference; *P < 0.05; **P < 0.01; ***P < 0.001

Anxiety disorder is commonly triggered by stress exposure. Therefore, to examine whether astrocytic Netrin-1 modulates stress-induced anxiety susceptibility, we subjected Netrin-1 cKO^GFAP mice and littermate WT mice to restraint stress for 4 h daily over 3 consecutive days. This regimen was chosen because Netrin-1 expression showed its most pronounced drop at this early time point (Fig. 1F), ensuring the induction of gene expression changes while keeping secondary, chronic-stress adaptations to a minimum. Under this stress paradigm, WT mice showed no behavioral alterations relative to non-stressed controls (Fig. 2F-P), confirming that the 3 days restraint stress alone was insufficient to evoke anxiety- or depression-like phenotypes—thus representing a subthreshold stress paradigm. In contrast, stressed Netrin-1 cKO^GFAP mice exhibited significantly increased anxiety-like behaviors. Specifically, stressed Netrin-1 cKO^GFAP mice spent less time exploring the center of the open field (Fig. 2H), exhibited a preference for the dark chamber in the LDT (Fig. 2I-J), and spent less time in the open arms of the EPM (Fig. 2K-L). However, stressed Netrin-1 cKO^GFAP mice exhibited no detectable depression-like behaviors in the TST or SPT (Fig. 2M-P). These results suggest that knockout of Netrin-1 in astrocytes enhances susceptibility to stress-induced anxiety, without affecting depression-like behaviors under the conditions tested.

To assess whether the observed effects of Netrin-1 deletion were due to developmental alterations, we crossed Netrin-1^f/f mice with Aldh1L1-Cre^ER mice, allowing selective Netrin-1 deletion in astrocytes in adulthood via Tamoxifen injection (Fig. 3A). The knockout efficiency was verified by immunofluorescence and Western blotting (Fig. 3B-D). As in the embryonic knockout model, Netrin-1 cKO^Aldh1L1 mice did not display baseline anxiety- or depression-like behaviors. However, following exposure to restraint stress, Netrin-1 cKO^Aldh1L1 mice displayed significant anxiety-like behaviors, including reduced exploration of the center in the OFT, increased preference for the dark chamber in the LDT, and decreased open arm duration in the EPM test (Fig. 3E-H), without detectable depression-like behaviors in the TST and SPT (Fig. 3I- J). These findings confirm that the absence of Netrin-1 in astrocytes increases stress-induced anxiety susceptibility, which is independent of developmental regulation.

Fig. 3.

Increased susceptibility to anxiety in Netrin-1 cKO^Aldh1L1 mice. A Schematic diagram showing the experimental timeline. B Co-staining of Netrin-1 and S100β in the mPFC of WT and cKO^Aldh1L1 mice. C Quantification of Netrin-1 levels in astrocytes (S100β⁺) in the mPFC of WT and cKO^Aldh1L1 mice. Student’s t test, P < 0.0001. D Immunoblotting and quantification of Netrin-1 expression in the mPFC of WT and cKO^Aldh1L1 mice. Student’s t test, P = 0.0007. E Quantification of total distance moved in the open field chamber. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 28) = 0.0010, P = 0.9746; main effect of stress: F (1, 28) = 1.2050, P = 0.2816; main effect of genotype: F (1, 28) = 0.0683, P = 0.7958. WT vs. cKO^Aldh1L1: P = 0.9984; WT vs. WT + Stress: P = 0.8743; cKO^GFAP vs. cKO^GFAP + Stress, P = 0.8542; WT + Stress vs. cKO^Aldh1L1 + Stress: P = 0.9968. F Quantification of center duration. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 28) = 12.9, P = 0.0012; main effect of stress: F (1, 28) = 4.57, P = 0.0414; main effect of genotype: F (1, 28) = 15.0, P = 0.0006. WT vs. cKO^Aldh1L1: P = 0.9971; WT vs. WT + Stress: P = 0.7343; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.0020; WT + Stress vs. cKO^Aldh1L1 + Stress: P < 0.0001. G Quantification of the time spent in the dark chamber. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 28) = 6.81, P = 0.0144; main effect of stress: F (1, 28) = 4.71, P = 0.0386; main effect of genotype: F (1, 28) = 18.3, P = 0.0002. WT vs. cKO^Aldh1L1: P = 0.6470; WT vs. WT + Stress: P = 0.9893; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.0109; WT + Stress vs. cKO^Aldh1L1 + Stress: P = 0.0002. H Quantification of the time spent in the open arm. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 28) = 2.37, P = 0.1348; main effect of stress: F (1, 28) = 10.9, P = 0.0027; main effect of genotype: F (1, 28) = 5.57, P = 0.0255. WT vs. cKO^Aldh1L1: P = 0.9375; WT vs. WT + Stress: P = 0.6057; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.0099; WT + Stress vs. cKO^Aldh1L1 + Stress: P = 0.0471. I Quantification of the time spent immobile in the tail suspension test. Two-way ANOVA followed by Tukey's multiple comparisons test. F (1, 28) = 0.0010, P = 0.9746; main effect of stress: F (1, 28) = 1.21, P = 0.2816; main effect of genotype: F (1, 28) = 0.0682, P = 0.7958. WT vs. cKO^Aldh1L1: P = 0.9984; WT vs. WT + Stress: P = 0.8743; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.8542; WT + Stress vs. cKO^Aldh1L1 + Stress: P = 0.9968. J Quantification of sucrose preference ratio. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 28) = 0.0144, P = 0.9053; main effect of stress: F (1, 28) = 0.251, P = 0.6204; main effect of genotype: F (1, 28) = 0.939, P = 0.3408. WT vs. cKO^Aldh1L1: P = 0.8672; WT vs. WT + Stress: P = 0.9930; cKO^GFAP vs. cKO^GFAP + Stress: P = 0.9712; WT + Stress vs. cKO^Aldh1L1 + Stress: P = 0.9311. For immunofluorescence and immunoblotting, the data are from three independent experiments. For behavioral tests, the number of mice (n) used is indicated in the graphs. ns, no significant difference; *P < 0.05; **P < 0.01; ***P < 0.001

AAV-mediated astrocytic Netrin-1 knockout in the mPFC of adult mice increases anxiety susceptibility

Since Netrin-1 was highly expressed in astrocytes in the mPFC, we asked whether knockout of astrocytic Netrin-1 in the mPFC would induce anxiety-like behaviors. To answer this question, we bilaterally injected AAV-GfaABC1D-SV40 NLS-Cre or AAV-GfaABC1D-EGFP viruses into the mPFC of Netrin-1^f/f mice for behavioral studies (Fig. 4A-B). Confocal imaging revealed the specific expression of viruses in astrocytes (Fig. 4C), and Western blotting confirmed the Netrin-1 knockout efficiency (Fig. 4D). We next evaluated anxiety-like phenotypes and detected no differences between the mice infected with the GFP virus and the mice infected with the Cre virus (Fig. 4E-H). However, following 3 days of restraint stress, the mice infected with the Cre virus displayed obvious anxiety-like behaviors in comparison with the mice infected with the GFP virus (Fig. 4E-H). These results confirm the role of astrocytic Netrin-1 in modulating stress-induced anxiety susceptibility and underscore the importance of the mPFC in this process.

Fig. 4.

Increased anxiety susceptibility caused by Netrin-1 knockout in astrocytes in the mPFC. A Schematic diagram showing the experimental timeline. B Injection of AAV-GfaABC1D-SV40 NLS-Cre or AAV-GfaABC1D-EGFP into the mPFC of Netrin-1^f/f mice. C Immunostaining and quantification of the transduction of virus (GFP⁺) in astrocytes (S100β⁺). Representative image and quantification of virally encoded GFP fluorescence (GFP⁺) co-localized with astrocytes (S100β⁺). D Immunoblotting showing a reduction of Netrin-1 in AAV-GfaABC1D-SV40 NLS-Cre infected mPFC compared to AAV-GfaABC1D-EGFP infected mPFC. Student’s t test, P = 0.0061. E Quantification of the total distance travelled in the open field chamber. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 24) < 0.0001, P = 0.9980; main effect of stress: F (1, 24) = 4.41, P = 0.0465; main effect of genotype: F (1, 24) = 0.634, P = 0.4338. GFP vs. Cre: P = 0.9426; GFP vs. GFP + Stress: P = 0.4630; Cre vs. Cre + Stress: P = 0.4609; GFP + Stress vs. Cre + Stress: P = 0.9415. F Quantification of center duration. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 24) = 5.66, P = 0.0257; main effect of stress: F (1, 24) = 6.30, P = 0.0192; main effect of genotype: F (1, 24) = 5.61, P = 0.0262. GFP vs. Cre: P > 0.9999; GFP vs. GFP + Stress: P = 0.9997; Cre vs. Cre + Stress: P = 0.0103; GFP + Stress vs. Cre + Stress: P = 0.0130. G Quantification of the time spent in the dark box. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 24) = 1.90, P = 0.1810; main effect of stress: F (1, 24) = 11.0, P = 0.0029; main effect of genotype: F (1, 24) = 6.56, P = 0.0171. GFP vs. Cre: P = 0.8361; GFP vs. GFP + Stress: P = 0.5294; Cre vs. Cre + Stress: P = 0.0143; GFP + Stress vs. Cre + Stress: P = 0.0472. H Quantification of the time spent in the open arm. Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 24) = 2.77, P = 0.1088; main effect of stress: F (1, 24) = 7.78, P = 0.0102; main effect of genotype: F (1, 24) = 7.76, P = 0.0103. GFP vs. Cre: P = 0.8575; GFP vs. GFP + Stress: P = 0.8561; Cre vs. Cre + Stress: P = 0.0210; GFP + Stress vs. Cre + Stress: P = 0.0212. For immunofluorescence and immunoblotting, the data are from three independent experiments. For behavioral tests, the number of mice (n) used is indicated in the graphs. ns, no significant difference; *P < 0.05; **P < 0.01

Hippocampus is also a key structure in stress-related anxiety [35]. We examined whether a similar function existed in the hippocampus. Consistently, Netrin-1 expression in the hippocampus was also modestly down-regulated after the same 3 days restraint stress (Supplement Fig. 1A). Then, we bilaterally injected AAV-GfaABC1D-SV40 NLS-Cre or AAV-GfaABC1D-EGFP viruses into the ventral hippocampus (vHPC) of Netrin-1^f/f mice for behavioral studies (Supplement Fig. 1B-C). However, selective deletion of Netrin-1 in the vHPC did not elicit anxiety-like behaviors under either baseline or stress conditions (Supplement Fig. 1D-G). We speculate that these null effects in the vHPC may be attributed to region-specific differences in baseline Netrin-1 expression and functional sensitivity.

Netrin-1 deficiency elevates TNFα levels by promoting TNFα shedding from proTNFα

To investigate the mechanism by which Netrin-1 cKO increases anxiety susceptibility, we performed RNA sequencing to analyze differentially expressed genes (DEGs) in the mPFC of stressed cKO^Aldh1L1 mice compared to stressed WT mice. The analysis identified only 25 DEGs (24 upregulated genes and 1 downregulated gene) in the mPFC of stressed cKO^Aldh1L1 mice (log₂FC > 0, Q value < 0.05), of which 8 genes exhibited more than a twofold change (log₂FC > 1, indicated in red) (Fig. 5A and Supplement Table 1). These findings suggest that Netrin-1 does not play a dominant role in transcriptional regulation under stress conditions. Using gene-set enrichment analysis (GSEA), the TNFα signaling pathway exhibited a significant normalized enrichment score (NES) of 1.48 (FDR = 0.24, P < 0.0001). GSEA revealed an alteration in TNFα signaling, pointing to a potential link between Netrin-1 and TNFα in stress-induced anxiety.

Fig. 5.

Increased TNFα shedding in stressed Netrin-1 cKO^Aldh1L1 mice. A Volcano plots showing significantly differentially expressed genes (Q value < 0.05, log₂FC > 1) between the mPFC of stressed WT mice and stressed cKO^Aldh1L1 mice. B GSEA suggesting the involvement of TNF signaling in anxiety-like phenotypes in cKO^Aldh1L1 mice. C Antibody array assay showing changes in the levels of inflammatory factors in the mPFC of stressed cKO^Aldh1L1 mice compared with stressed WT mice. TNFα is highlighted in red, CCL25 and TNFR1 are highlighted in orange, and CCL25 is highlighted in blue. D Quantification of the fold changes in the levels of inflammatory factors as determined by the antibody array assay. Student’s t test, P = 0.0492 for TNFα, P = 0.0113 for CCL25, P = 0.038 for TNFR1, P < 0.0001 for CCL24. E Schematic illustration of the experimental design. F Real-time PCR showing the increase of TNFα mRNA levels in the stressed groups compared with that in control groups. Two-way ANOVA followed by Two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli. Interaction: F (1, 8) = 0.244, P = 0.6347; main effect of stress: F (1, 8) = 20.2, P = 0.0020; main effect of genotype: F (1, 8) = 0.604, P = 0.4593. WT vs. cKO^Aldh1L1: P = 0.3950; WT vs. WT + Stress: P = 0.0078; cKO vs. cKO^Aldh1L1 + Stress: P = 0.0222; WT vs. cKO^Aldh1L1 + Stress: P = 0.0058; WT + Stress vs. cKO^Aldh1L1 + Stress: P = 0.8461. G ELISA showing the increase of TNFα protein levels in the stressed cKO ^Aldh1L1 group compared with that in the stressed WT group. Two-way ANOVA followed by Two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli. Interaction: F (1, 16) = 9.10, P = 0.0082; main effect of stress: F (1, 16) = 36.5, P < 0.0001; main effect of genotype: F (1, 16) = 13.9, P = 0.0018. WT vs. cKO^Aldh1L1: P = 0.6199; WT vs. WT + Stress: P = 0.0484; cKO vs. cKO^Aldh1L1 + Stress: P < 0.0001; WT vs. cKO^Aldh1L1 + Stress: P < 0.0001; WT + Stress vs. cKO^Aldh1L1 + Stress: P = 0.0002. H Immunoblotting showing the decrease of proTNFα in the stressed cKO group compared with that in the stressed WT group. Two-way ANOVA followed by Two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli. Interaction: F (1, 8) = 15.4, P = 0.0044; main effect of stress: F (1, 8) = 82.8, P < 0.0001; main effect of genotype: F (1, 8) = 21.4, P = 0.0017. WT vs. cKO^Aldh1L1: P = 0.6307; WT vs. WT + Stress: P = 0.0064; cKO vs. cKO^Aldh1L1 + Stress: P < 0.0001; WT vs. cKO^Aldh1L1 + Stress: P < 0.0001; WT + Stress vs. cKO^Aldh1L1 + Stress, P = 0.0003. For immunofluorescence and protein array assays, the data are from three independent experiments. ns, no significant difference; *P < 0.05; **P < 0.01; ***P < 0.001

Given that TNFα is a key mediator of neuroinflammation, we analyzed the levels of TNFα, TNFα receptors, and various other inflammatory factors via an antibody array assay. The mPFC regions were dissected from stressed WT and stressed Netrin-1 cKO^Aldh1L1 mice and the lysates were incubated with antibody-array nitrocellulose membranes. The analysis revealed a significant increase in proinflammatory factors, including TNFα, CCL25, and TNFR1, alongside a decrease in the anti-inflammatory factor CCL24 in stressed cKO^Aldh1L1 mice compared to stressed WT mice (Fig. 5C-D). Remarkably, TNFα exhibited the most pronounced change.

To clarify how TNFα was increased, we analyzed TNFα mRNA, soluble TNFα and proTNFα levels, as illustrated in Fig. 5E. Real-time PCR showed that Netrin-1 deletion alone left TNFα transcription unchanged; only stress produced a modest increase in TNFα mRNA (Fig. 5F). ELISA revealed that soluble TNFα rose after stress in WT mice, and this rise was dramatically amplified in Netrin-1 cKO^Aldh1L1 mice, whose proTNFα pool was simultaneously decreased markedly (Fig. 5G-H). These findings suggest that Netrin-1 deletion in astrocytes dose not drive TNFα transcription; instead, it promotes the proteolytic release of TNFα from its membrane-bound precursor.

Netrin-1 inhibits proTNFα cleavage through suppressing ADAM17 activity

ADAM17 is the principal sheddase of TNFα [7, 36] (Fig. 6A). Single-cell data showed that both Tnfα and Adam17 transcripts were enriched in microglia in the PFC under baseline and social defeat stress conditions (Fig. 6B-C) [31, 32], confirming microglia as the dominant TNFα source [13] and implicating ADAM17 in its release. Surprisingly, Western blotting revealed no differences in ADAM17 protein levels between stressed Netrin-1 cKO^Aldh1L1 and stressed WT mice (Fig. 6D), suggesting that Netrin-1 might not regulate ADAM17 expression but instead modulate ADAM17 activity. Consistent with this idea, Netrin-1 and ADAM17 co-localized in microglia in the mPFC (Fig. 6E). Co-immunoprecipitation assay together with pull-down assays confirmed the interaction between ADAM17 and Netrin-1 (Fig. 6F-G), providing a mechanistic basis for Netrin-1 regulating TNFα shedding.

Fig. 6.

Suppression of ADAM17-mediated TNFα shedding by Netrin-1. A Schematic illustration showing that pro-TNF-α is cleaved by ADAM17 to release the soluble TNFα. B Single cell data showing the expression patterns of Tnfa and Adam17 in the PFC of P60 mice (GSE124952). C Single cell data showing the expression patterns of Tnfa and Adam17 in the PFC of mice following social defeat stress (GSE240975). D Immunoblotting and quantification of ADAM17 expression in the mPFC of stressed WT and stressed cKO^Aldh1L1 mice. Student’s t test, P = 0.6604. E Co-localization of Netrin-1 and ADAM17 in microglia (Iba1⁺) in the mPFC. F Co-immunoprecipitation of exogenous Myc-ADAM17 and Flag-Netrin-1 from HEK293 cell lysates. G Pull-down assay of the interaction between HA-Netin-1 and Myc-ADAM17. Purified HA-Netrin-1 was incubated with Myc beads or Myc beads-ADAM17. Total amounts of HA-Netrin-1 were visualized by Coomassie blue staining. H Representative images of HEK293 cells transfected with GFP-proTNFα-RFP alone or with both GFP-proTNFα-RFP and Myc-ADAM17. As indicated, Netrin-1 was administered after cell transfection at a concentration of 150 ng/ml for 16 h. I Illustration of the GFP-proTNF-RFP fusion protein and its ADAM17 cleavage site. J Quantification of the cleavage of GFP-proTNF-RFP. One-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (2, 51) = 26.26, P < 0.0001. proTNFα vs. proTNFα + ADAM17: P < 0.0001; proTNFα vs. proTNFα + ADAM17 + Netrin-1: P = 0.0464; proTNFα + ADAM17 vs. proTNFα + ADAM17 + Netrin-1: P < 0.0001. K Representative images of Iba1 immunostaining in the mPFC of WT, cKO^Aldh1L1, stressed WT and stressed cKO^Aldh1L1 mice. L Quantification of microglia numbers (Iba1⁺). Two-way ANOVA followed by Tukey's multiple comparisons test. Interaction: F (1, 32) = 0.677, P = 0.4167; main effect of stress: F (1, 32) = 3.01, P = 0.0923; main effect of genotype: F (1, 32) = 2.75, P = 0.1069. WT vs. cKO^Aldh1L1: P = 0.9340; WT vs. WT + Stress: P = 0.2880; cKO vs. cKO^Aldh1L1 + Stress: P = 0.9165; WT + Stress vs. cKO^Aldh1L1 + Stress: P = 0.3133. M Quantification of microglial process lengths. Two-way ANOVA followed by Tukey's multiple comparisons test. F (1, 56) = 4.07, P = 0.0484; main effect of stress: F (1, 56) = 16.4, P = 0.0002; main effect of genotype: F (1, 56) = 13.7, P = 0.0005. WT vs. cKO^Aldh1L1: P = 0.6328; WT vs. WT + Stress: P = 0.4822; cKO vs. cKO^Aldh1L1 + Stress: P = 0.0004; WT + Stress vs. cKO^Aldh1L1 + Stress: P = 0.0009. The data are from three independent experiments. ns, no significant difference; *P < 0.05; ***P < 0.001

To demonstrate the regulatory effect of Netrin-1 on ADAM17-mediated TNFα shedding, we employed a GFP-proTNF-RFP fusion protein (Fig. 6H-J), which emits both GFP and RFP signals in the full-length form [27]. Once the GFP-proTNF-RFP fusion protein is cleaved by ADAM17, TNFα-RFP is released outside of the cells and GFP-pro is retained in the cells. In HEK293 cells, the GFP-proTNF-RFP alone showed equal red and green signals (Fig. 6H and J). When Myc-ADAM17 was coexpressed, the red signals were reduced, indicating increased cleavage of GFP-proTNF-RFP by ADAM17. However, when Netrin-1 was added to the medium simultaneously, the decrease in red signals was counteracted, suggesting that Netrin-1 suppresses ADAM17-mediated TNFα shedding.

In addition, we assessed the microglia numbers and morphology in the mPFC via Iba1 immunostaining (Fig. 6K). Under basal conditions (without stress), Netrin-1 deficiency had little effect on microglia numbers or process lengths (Fig. 6L-M). However, following exposure to restraint stress, Netrin-1 cKO^Aldh1L1 mice exhibited a significant reduction in the total length of microglial processes, without changes of microglia numbers. This morphological change suggests increased microglial activity, consistent with the elevated TNFα levels in stressed Netrin-1 cKO^Aldh1L1 mice.

Blocking TNFα in the mPFC attenuates anxiety-like behaviors in stressed Netrin-1 cKO^Aldh1L1 mice

TNFα is known to be a key mediator in the development of anxiety disorder [8–10]. To confirm that the increase of TNFα in the mPFC contributes to anxiety-like behaviors, we injected TNFα (200 nl, 0.5 pg/side) into the mPFC of C57BL/6 male mice (Fig. 7A). Behavioral tests revealed that as little as 1 pg TNFα was able to induce anxiety-like behaviors, as evidenced by reduced exploration of the center in the OPT, increased preference for the dark chamber in the LDT, and decreased time spent in the open arms of the EPM (Fig. 7B-E).

Fig. 7.

Alleviation of anxiety susceptibility in stressed cKO mice by blocking TNFα signaling. A Schematic diagram showing the experimental timeline and TNFα injection sites. B Quantification of the total distance travelled in the open field chamber. Student’s t test, P = 0.3932. C Quantification of center duration. Student’s t test, P = 0.0376. D Quantification of the time spent in the dark box. Student’s t test, P = 0.0495. E Quantification of the time spent in the open arms. Student’s t test, P = 0.0362. F Schematic diagram showing the experimental timeline. G Quantification of total distance moved in the open field chamber. One-way ANOVA followed by Tukey's multiple comparisons test. F (2, 21) = 0.318, P = 0.7313. WT + PBS vs. cKO^Aldh1L1 + PSB: P = 0.8303; cKO^Aldh1L1 + PBS vs. cKO^Aldh1L1 + Infliximab: P = 0.9826; WT + PBS vs. cKO^Aldh1L1 + Infliximab: P = 0.7299. H Quantification of center duration. One-way ANOVA followed by Tukey's multiple comparisons test. F (2, 21) = 5.42, P = 0.0127. WT + PBS vs. cKO^Aldh1L1 + PSB: P = 0.0161; cKO^Aldh1L1 + PBS vs. cKO^Aldh1L1 + Infliximab: P = 0.0426; WT + PBS vs. cKO^Aldh1L1 + Infliximab: P = 0.8942. I Quantification of the time spent in the dark box. One-way ANOVA followed by Tukey's multiple comparisons test. F (2, 21) = 6.20, P = 0.0077. WT + PBS vs. cKO^Aldh1L1 + PSB: P = 0.0109; cKO^Aldh1L1 + PBS vs. cKO^Aldh1L1 + Infliximab: P = 0.0255; WT + PBS vs. cKO^Aldh1L1 + Infliximab: P = 0.9231. J Quantification of the time spent in the open arm. One-way ANOVA followed by Tukey's multiple comparisons test. F (2, 21) = 6.69, P = 0.0057. WT + PBS vs. cKO^Aldh1L1 + PSB: P = 0.0053; cKO^Aldh1L1 + PBS vs. cKO.^Aldh1L1 + Infliximab: P = 0.0457; WT + PBS vs. cKO + Infliximab: P = 0.599. The number of mice (n) used is indicated in the graphs. ns, no significant difference; *P < 0.05; **P < 0.01

To establish a causal link between TNFα and the anxiety-like behaviors observed in Netrin-1 cKO^Aldh1L1 mice, we conducted rescue experiments using Infliximab, a chimeric monoclonal antibody against TNFα, which prevents the interaction between TNFα and TNFα receptors (Fig. 7F). Behavioral tests revealed that Infliximab alleviated the anxiety-like behaviors of stressed Netrin-1 cKO^Aldh1L1 mice (Fig. 7G-J). Taken together, these findings suggest that Netrin-1 modulates anxiety-like behaviors by suppressing ADAM17-mediated TNFα shedding.

Discussion

Here, we identified a critical role for astrocytic Netrin-1 in regulating stress-induced anxiety susceptibility. Specifically, Netrin-1 interacted with ADAM17 and inhibited ADAM17-mediated TNFα shedding. Netrin-1 deficiency was inferred to relieve this inhibition and promote TNFα release. The elevated TNFα in stressed Netrin-1 cKO mice and the full rescue of their anxiety-like phenotype by TNFα blockade validated this model and positioned the Netrin-1-ADAM17-TNFα axis as a tractable therapeutic target.

Consistent with earlier reports [37], Netrin-1 is expressed by astrocytes in the mouse forebrain. Remarkably, 3 days of restraint stress selectively reduced astrocytic Netrin-1 without altering neuronal levels (Fig. 1G-H), an effect most likely mediated by the GC/GR signaling. Typically, restraint activates the HPA axis, raising circulating GC that readily enters the brain. Once bound to the GR, the GC-GR complex translocates to the nucleus and either directly binds DNA or tethers to other transcription factors, thereby activating or repressing target-gene transcription [2]. We recapitulated this regulation in cultured astrocytes with the synthetic glucocorticoid DEX: a low “baseline” concentration (10⁻⁸ M) elevated Netrin-1 expression, whereas a “stress-level” dose (10⁻⁶ M) suppressed it; peak expression occurred at 10⁻⁷ M (Fig. 1J). L-Caldesmon levels also followed an identical hormetic curve, confirming the transcriptional regulation of the GC/GR signaling [4, 34]. Thus, the direction and magnitude of Netrin-1 change appear to be tuned to stress intensity. Why only astrocytic Netrin-1 is vulnerable remains unclear; we posit that astrocytic end-feet ensheathing cerebral micro-vessels form the first parenchymal barrier to blood-borne GC, conferring stress sensitivity to astrocytic Netrin-1 expression.

To elucidate the role of astrocytic Netrin-1 in anxiety-like behaviors, two astrocytic Netrin-1 conditional knockout mouse lines were generated. One was the Netrin-1 cKO^GFAP mouse line, in which Netrin-1 was deleted in a portion of astrocytes from embryonic stages, and the other was the Netrin-1 cKO^Aldh1L1 mouse line, in which Netrin-1 was deleted specifically in astrocytes after Tamoxifen-induced Cre-loxP recombination. Both mouse lines displayed no anxiety-like or depression-like phenotypes under baseline conditions (Fig. 2F-P and Fig. 3E-J). Upon exposure to 3 days restraint stress, WT mice exhibited unchanged behaviors, suggesting that 3 days restraint stress was a subthreshold paradigm for anxiety-like behaviors. However, both Netrin-1 cKO^GFAP mice and Netrin-1 cKO^Aldh1L1 mice exhibited obvious anxiety-like behaviors (Fig. 2D-O and Fig. 3E-J), suggesting that Netrin-1 deficiency amplified the anxiogenic impact of stress. Furthermore, specific knockout of astrocytic Netrin-1 in the mPFC was sufficient to induce anxiety-like behaviors following 3 days restraint stress (Fig. 4E-H). These results underscore the critical role of astrocytic Netrin-1 in the mPFC in modulating stress-induced anxiety susceptibility.

RNA-seq and protein array profiling revealed that this susceptibility was accompanied by a selective rise in soluble TNFα (Fig. 5A-D). Consistently, a single intra-mPFC injection of 1 pg TNFα was sufficient to reproduce the anxiogenic phenotype (Fig. 7A-E). This low threshold argues against neuronal death or gross structural damage but in favor of functional synaptic adjustment [38, 39]. Previous studies have shown that TNFα can increase AMPA receptor externalization at excitatory synapses, down-regulate GABA currents, and potentiate glutamate release [38, 40]. All these mechanisms are implicated in the hyper-vigilant state characteristic of anxiety disorders [41]. Since the mPFC integrates interoceptive and exteroceptive cues to guide flexible coping, a focal increase in TNFα can be interpreted as a biochemical switch that biases the circuit from adaptive appraisal to maladaptive threat generalization by increasing the excitatory/inhibitory ratio. The causal relevance of TNFα is further underscored by the efficacy of Infliximab, a clinically used anti-TNFα antibody, which completely reversed the anxiety-like phenotype of stressed Netrin-1 cKO^Aldh1L1 mice after intra-mPFC injection (Fig. 7G-J), aligning with previous reports that TNFα blockade ameliorates anxiety evoked by peripheral inflammation [42], chronic pain [41], obesity [43], neuroinflammation [44] as well as acute and chronic mild stress [12, 13]. Future studies should determine whether this low-dose TNFα effect generalizes to other brain regions and whether early anti-cytokine intervention can prevent the transition from normal stress responses to anxiety disorders.

TNFα shedding is catalyzed mainly by ADAM17 [7]. Single-cell data from naïve mice or socially-defeated mice showed that both Tnfα and Adam17 transcripts were enriched in microglia in the mPFC (Fig. 6B-C) [31, 32], consistent with the view that microglia are the dominant source of TNFα [13, 36]. However, it is important to note that although TNFα is predominantly produced by microglia, astrocytes are also capable of synthesizing and releasing TNFα, and this astrocyte-derived TNFα may additionally contribute to anxiety susceptibility [45]. Prompted by the increased shedding of TNFα in stressed Netrin-1 cKO^Aldh1L1 mice, we hypothesized that Netrin-1 might regulate TNFα release by modulating ADAM17 activity. Co-immunoprecipitation, pull-down and colocalization analyses confirmed a physical interaction between Netrin-1 and ADAM17 (Fig. 6E-G). To monitor ADAM17 activity, we employed the GFP-proTNFα-RFP reporter, which emits equal green and red signals as a whole and emits fewer red signals when TNFα-RFP is cleaved and subsequently released [22]. In HEK293T cells, ADAM17 overexpression led to a decrease of red signals, indicative of increased TNFα shedding. Co-application of Netrin-1 suppressed the decrease of red signals caused by ADAM17 overexpression (Fig. 6H-J), demonstrating that Netrin-1 suppresses ADAM17-mediated TNFα shedding. Although our data establish a direct Netrin-1-ADAM17 interface. Netrin-1 may additionally regulate ADAM17 activity through Integrins. Several integrins, such as Integrin α6β4, α3β1 and α6β1 [46–48] are known Netrin-1 receptors. Notably, Integrin α6β1 has been shown to bind with ADAM17 and enhance ADAM17-mediated TNFα shedding [49]. Netrin-1 could therefore inhibit ADAM17 activity by competitively binding with Integrin α6β1. Further studies are needed to elucidate the cross-talk among Netrin-1, ADAM17 and Integrins.

In conclusion, Netrin-1 is highly expressed in astrocytes in the mPFC, and loss of astrocytic Netrin-1 increases ADAM17-mediated TNFα shedding, which contributes to stress-induced anxiety susceptibility.

Authors’ contributions

YY and XL conducted most of the experiments, including animal behavioral tests, immunostaining and protein array assay. JM and ML helped with protein interaction. LY, JN and JY helped with cell culture. MX, MT and JZ helped with immunoblotting. WC, XH and ZH helped with language editing. ZW and XZ provided critical suggestions for the project. HY supervised the project, conceived the experiments, and wrote the manuscript. All authors have read and agreed to the published version of the manuscript. The authors report no biomedical financial interests or potential conflicts of interest.

Funding

This work was supported by the STI2030-Major Project (grant number 2021ZD0204003 to XZ), National Natural Science Foundation of China (grant number 32371039 to HY), Project supported by Jilin Province and Central Government Guides Local Science and Technology Development Fund (grant number YDZJ202201ZYTS620 to HY).

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declarations

Ethics approval and consent to participate

All animal experiments were approved by Institutional Animal Care and Use Committee of Northeast Normal University. National Standards of the People’s Republic of China (GB/T 35892–2018), Laboratory Animal—Guideline for Ethical Review of Animal Welfare, was the guidance for our animal care and protocols.

Consent for publication

All the authors approved the final version for publication.

Competing interests

The authors declare no competing interests.