ECM remodeling in the mPFC exacerbates cocaine-induced hyperactivity and impairs threat vigilance

Abstract

The extracellular matrix (ECM) plays a crucial role in addiction-related neuroplasticity. While chondroitinase ABC (chABC)-mediated ECM degradation has been shown to impair drug-conditioned place preference, its effects on cocaine-induced locomotor activity remain unknown. Using 3D motion capture combined with unsupervised behavioral clustering in a chronic cocaine exposure mouse model, we demonstrate that cocaine treatment significantly increases locomotor activity while impairing environmental threat perception. Mechanistically, we observed elevated Wisteria floribunda agglutinin (WFA) intensity in the prelimbic (PrL), with pronounced effects in the interstitial matrix and perineuronal nets (PNNs) surrounding parvalbumin-positive (PV⁺) neurons in layers 2/3 and 5. Surprisingly, despite successfully reducing WFA staining in both PNNs and interstitial matrix, chABC treatment unexpectedly exacerbated cocaine-induced hyperlocomotion and impaired threat detection behaviors. Small interfering RNA (siRNA)-mediated knock-down of Lox, but not Spp1 partially rescued chABC-aggravated threat perception deficits in cocaine-exposed mice. These findings demonstrate that chronic cocaine exposure induces significant ECM remodeling in the PrL, and that ECM degradation paradoxically worsens behavioral outcomes through specific molecular alterations. Our results highlight the complex, context-dependent role of ECM in addiction-related behaviors and suggest the need for more targeted approaches to ECM-based interventions for substance use disorders.

Introduction

Drug addiction remains a persistent global health challenge, characterized by compulsive drug use despite adverse consequences. The escalating prevalence of stimulant abuse, particularly cocaine, has led to a concerning rise in overdose-related deaths [1, 2]. This chronic relapsing disorder arises from maladaptive neuroplasticity across molecular, cellular, and circuit levels following repeated drug exposure [3–5]. While extensive research has focused on neuronal adaptations in addiction, the extracellular matrix (ECM) - a dynamic structural scaffold that regulates synaptic plasticity - has received comparatively less attention.

The brain ECM exists in two primary forms: a diffuse network throughout the extracellular space and dense perineuronal nets that enwrap specific neuronal populations [6]. Beyond providing structural support, the brain ECM actively participates in neurodevelopment, synaptic stabilization, and experience-dependent plasticity [6–8]. Emerging evidence implicates ECM alterations in various neuropsychiatric disorders, including addiction [9]. Chronic drug exposure modifies PNN density and composition in reward-related brain regions, with excessive PNN deposition associated with impaired neuronal plasticity, and PNN removal has been proven to be an effective therapeutic strategy [10]. Notably, region-specific degradation of chondroitin sulfate proteoglycans (CSPGs) by chondroitinase ABC (chABC) has shown therapeutic promise in addiction models. chABC administration in the prelimbic cortex (PrL) disrupts cocaine conditioned place preference [11], while intra-amygdala injections prevent drug memory reinstatement [12]. However, these studies have primarily examined PNNs and reward-related behaviors, leaving the role of interstitial ECM and its contribution to drug-induced locomotor effects largely unexplored.

The medial prefrontal cortex (mPFC), particularly its prelimbic and infralimbic subdivisions, serves as a critical hub for executive control over drug-seeking behaviors [13, 14]. Human and rodent studies consistently demonstrate mPFC involvement in addiction phases from initial use to relapse [3, 15, 16]. Given its established role in addiction pathophysiology and rich ECM composition, the mPFC represents an ideal region to investigate ECM-drug interactions.

In this study, we employed a multidisciplinary approach to examine how chronic cocaine exposure alters ECM composition in the mPFC and how these changes influence behavioral outcomes. Using advanced 3D behavioral analysis combined with molecular and histological techniques, we provide novel insights into the complex relationship between ECM remodeling and cocaine-induced behaviors. Our findings challenge conventional assumptions about ECM-targeted interventions and reveal unexpected consequences of CSPG degradation in addiction models.

Materials and methods

Animals

All animal procedures were conducted in accordance with the ethical guidance and regulation approved by Institutional Animal Care and Use Committee of Peking University Health Science Center. Male C57BL/6 J mice aged 8 ~ 10 weeks were raised in Department of Laboratory of Animal Science in Peking University. The housing room was maintained in a barrier condition with well-controlled room temperature (22 ± 1 °C), room humidity (50 ± 5%) and reversed 12 h light/dark cycle (light on at 19:00). Standard food chow and pure filtered water were provided ad libitum. Bedding material was renewed twice weekly to maintain hygiene.

Drug administration

Mice underwent chronic drug exposure via intraperitoneal injections as previously described [17, 18]. The cocaine group received daily cocaine hydrochloride (20 mg/kg; Qinghai Pharmaceutical, Qinghai, China) for 7 days. Controls received saline following the same schedule.

Stereotaxic surgery

Three days before cocaine or saline treatment, mice were anesthetized with isoflurane and secured in a stereotaxic apparatus (RWD Life Science, Shenzhen, China). Chondroitinase ABC (chABC, 83.33 U/ml, 200 nl/site; Merck, Shanghai, China) or penicillinase (PEN, 83.33 U/ml, 200 nl/site; Rhawn, Shanghai, China) was bilaterally infused into the PrL (AP: +1.85 mm, ML: ±0.30 mm, DV: -2.15 mm) using nanoinjectors (KD Scientific, Holliston, MA, USA). The syringe was left in place for 5 min post-injection to prevent reflux before slow withdrawal. Following stereotaxic surgery, mice were randomly allocated to either the saline group or the cocaine group.

The siRNAs (synthesized by GenePharma, Shanghai, China) were chemically modified with 2’-O-methyl groups to increase resistance to nuclease degradation and conjugated with cholesterol at the 5’ end to enhance transfection efficiency. The siRNA was dissolved in RNase-free water, mixed with the in vivo transfection reagent Entranster™-in vivo-RNA (Engreen, Beijing, China), and prepared as a transfection complex with a final volume of 350 nl. This complex was microinfused into the PrL using the same stereotaxic coordinates and surgical procedures described above, with each site receiving 0.5 μg of siRNA. Behavioral tests were conducted approximately 72 h after siRNA delivery. Mice were sacrificed 2 h after the final cocaine injection in day 7 for subsequent gene expression analysis. siRNA sequences are provided in Supplementary Table 1.

Real-time quantitative PCR (RT-PCR)

Mice were sacrificed 2 h post-injection, and mPFC tissue punches were collected using 12-gauge needles. RNA was extracted with Trizol reagent (Vayzyme, Nanjing, China), and cDNA was synthesized using the HiScript II 1st Strand cDNA Synthesis Kit (Vayzyme). qPCR was performed on a 7500 Real-Time System (Applied Biosystems, Foster city, USA) with chamQ SYBR master mix (Vazyme) under the following conditions: 95 °C for 30 s, then 40 cycles of 95 °C for 10 s and 60 °C for 30 s. Gene expression was normalized to Gapdh and analyzed using the 2^-ΔΔCt method. Primer sequences are listed in Supplementary Table 1.

Immunofluorescence

All mice used for immunofluorescence staining experiments were sacrificed 2 h after the last injection of cocaine or saline on day 7. Mice were anesthetized and transcardially perfused with saline followed by 4% PFA. Brains were post-fixed in 4% PFA overnight at 4 °C, dehydrated in 20% and 30% sucrose, and sectioned at 30 μm using a cryostat. For immunofluorescence, free-floating sections were washed in PBS, permeabilized with 0.3% Triton X-100/PBS for 30 min, and blocked with 5% donkey serum/0.3% Triton X-100/PBS for 1 h. Sections were incubated with primary antibodies at 4 °C overnight, washed in 0.2% Tween-20/0.3% Triton X-100/PBS, and incubated with secondary antibodies at room temperature for 1.5 h in the dark. For neuronal staining, sections were incubated with NeuroTrace (1:100, Molecular Probes, Eugene, OR, USA, N21479) for 20 min, washed, and mounted with coverslips for imaging.

The primary antibodies included biotin-conjugated lectin Wisteria floribunda (1:500, Sigma-Aldrich, St. Louis, MO, USA, L1516), mouse anti-parvalbumin (1:500, Sigma-Aldrich, SAB4200545), and rabbit anti-Neurocan (1:500, Affinity Biosciences, Cincinnati, OH, USA, DF4236). The secondary antibodies included FITC-conjugated streptavidin (1:500, Sigma-Aldrich, S3762), Alexa Fluor 647 Donkey anti-mouse IgG and Alexa Fluor 647 Donkey anti-rabbit IgG (1:500, Invitrogen, Carlsbad, CA, USA, A32787, A31573).

Images were acquired using a confocal microscope (Olympus FluoView 1000, Center Valley, PA, USA) and were analyzed by Image J. The calculation of WFA intensity was referred to previous study [19]. In brief, image background was subtracted with thresholds set at 2 standard deviations above the mean in brain midline areas (pia mater). Using ImageJ’s ROI tool, PNN structures were manually outlined, and their intensities were measured. PNN density was also quantified accordingly. Intensity values were normalized to the saline control group mean. The top 25% and the lowest 25% intensity of PNN in saline group were used to define low and high intensity PNN separately. Interstitial ECM in this study was defined as the total WFA staining area minus the PNN area. During quantification of Neurocan puncta density, ROIs of consistent size were placed in comparable regions of the PrL or IL across all samples. Subsequently, manual counting was performed using the Cell Counter plugin in Image J. During quantification, the analyst was blinded to group assignment.

Behavioral experiments and analysis

Mice were placed in a cylindrical open field apparatus (50 cm diameter × 50 cm height) for 20 min sessions on four occasions: saline administration, and on days 1, 4, and 7 of cocaine treatment. All experimental mice were acclimated to the testing apparatus for at least 20 min one day prior to the experiment. And the saline control group underwent behavioral recording on Day 7 following a single saline injection. During the behavioral tests, the experimenter was blinded to the animal group allocation.

For the spontaneous behavior collection and analysis, we employed the Behavior Atlas (Bayone, Shenzhen, China). Behavior was recorded using overhead 4 cameras for multi-view 3D reconstruction. The spontaneous behavior analysis in this study employs a hierarchical unsupervised machine learning framework. A pre-trained DeepLabCut (DLC) model tracked 16 body parts to generate 3D skeletal time-series data. It first uses the Dynamic Time Alignment Kernel (DTAK) to measure similarity between behavioral segments while aligning them temporally, handling variability in duration and speed. The high-dimensional behavioral features are then projected into a lower-dimensional space using UMAP (Uniform Manifold Approximation and Projection) with parameters set to balance local and global structural information of the data. Next, hierarchical clustering is adopted to aggregate the low-dimensional behavioral feature data into clusters, forming distinct behavioral phenotype groups. Finally, the Bayesian Information Criterion (BIC) is used to verify the rationality of the number of clusters, ensuring that the determined number of clusters (e.g., 40 for group experiments) is both data-driven and biologically meaningful. Further details are described in previous studies [20, 21]. Animals were excluded if their tracking data did not align with the physical test arena. Sample size was determined based on previous study [21] using this Behavior Atlas system.

The analyzer system quantified 39 kinematic parameters, including velocity, movement energy, body posture, and event-based behaviors (e.g., freezing, escaping). The unsupervised clustering categorized movements into 40 types, which were manually annotated into 7 defined behavioral categories (Supplementary Table 2, Supplementary Fig. 1C-D). Among these 7 behavior phenotypes, upright postures including “Up stretching” and “Head rising” (termed ‘rearing’ in most studies) are considered as active vigilance/risk‑assessment postures that reflect heightened environmental monitoring in rodents [22–24]. “Hunching” is recognized as a passive defensive posture, indicating a protective, withdrawn state [25, 26]. “Sniffing” can be interpreted as horizontal exploratory behavior [27]. “Rotating” and “Turning” might reflect mice’s motor flexibility or coordination [28].

Statistical analysis

All statistical analyses were perform using GraphPad Prism version 8. Data were represented as mean ± SEM. The normality of data distribution and the homogeneity of variances between groups were assessed prior to conducting parametric tests. Statistical significance was calculated by One-way ANOVA with Dunnett’s multiple comparisons test, unpaired t-test and unpaired t-test with Welch’s correction when appropriate. P < 0.05 was considered statistically significant.

Results

Chronic cocaine exposure disrupts locomotor patterns and threat perception

Following 7 days of chronic cocaine treatment, we analyzed spontaneous behaviors using a 3D motion capture system during 20-minute sessions on days 1, 4, and 7 post-injections. Our multi-camera setup combined with machine learning algorithms precisely tracked 16 body landmarks (Fig. 1A), enabling comprehensive behavioral quantification. Principal component analysis revealed distinct clustering between cocaine- and saline-treated groups in both kinematic and movement feature spaces (Fig. 1B). Cocaine administration significantly increased frame distance (Fig. 1C) and elevated movement velocity across all body regions (Supplementary Fig. 1A). These effects were most pronounced on day 4 of treatment, as visualized in kinematic feature maps (Fig. 1B, Supplementary Fig. 1A). Cocaine exposure also induced characteristic postural changes, increasing average body angle and body length (Fig. 1D-E) while body height remained unchanged (Supplementary Fig. 1B), suggesting a shift toward rigid, hyperextended postures.

Fig. 1. Chronic cocaine exposure produced locomotor hyperactivity.

A Mice received 7-day continuous cocaine treatment and the spontaneous behavior tests were conducted 20 min after cocaine treatment on day 1, day 4 and day 7. The Behavior Atlas Analyzer system collects videos from four cameras to estimate 3D coordinates of 16 body parts of mouse and thereby obtaining their kinematic parameters. Deep learning algorithm was applied to cluster mice behaviors into 40 movement types. B Principal component analysis of kinematic parameters (left) and movement features (right) of saline-treated mice and cocaine-treated mice. C The frame distance traveled during the tests. It is the frame-by-frame movement distance of the animal’s body center point. The body center point is defined as the mean of the coordinates (x- and y-axes) of all tracked body points. (One-way ANOVA with Dunnett’s multiple comparisons test: F_3,24 = 19.54, P < 0.0001; Saline vs. Cocaine_d1, P = 0.0273; Saline vs. Cocaine_d4, P < 0.0001; Saline vs. Cocaine_d7, P = 0.0004). D, E As the diagram showed, body angle was the angle of mouse’s neck, back and the root tail. The body length was the 3D distance between mice’s nose and root tail. The example traces of mouse’s body angle (D, left) and body length (E, left). And the comparison of average body angle (D, right) and body length (E, right). (One-way ANOVA with Dunnett’s multiple comparisons test: Body angle: F_3,24 = 12.33, P < 0.0001; Saline vs. Cocaine_d1, P = 0.0010; Saline vs. Cocaine_d4, P < 0.0001; Saline vs. Cocaine_d7, P = 0.0016. Body length: F_3,24 = 26.61, P < 0.0001; Saline vs. Cocaine_d1, P = 0.0001; Saline vs. Cocaine_d4, P < 0.0001; Saline vs. Cocaine_d7, P < 0.0001). All error bars represent mean ± SEM. * P < 0.05, *** P < 0.001, **** P < 0.0001, saline vs. cocaine_d1 or cocaine_d4 or cocaine_d7. N = 7 mice.

To precisely characterize cocaine’s behavioral effects, we systematically classified 40 distinct movement patterns into 7 defined phenotypes: running, sniffing, up stretching, head rising, turning, rotating, and hunching (Fig. 2A, Supplementary Fig. 1C-D, Supplementary Table 2). Integrated analysis of behavioral ethograms, movement skeleton diagrams, and behavioral distribution patterns revealed significant alterations in cocaine-treated mice (Fig. 2A-B). In detail, cocaine exposure induced increased locomotion and stereotypic ground-focused exploration, manifested as increased repetitive running and excessive walking and sniffing. Furthermore, cocaine exposure impaired threat assessment and defensive responses, evidenced by decreased up stretching, head rising, and defensive hunching behavior. Additionally, cocaine exposure impaired mice’s motor flexibility or coordination reflected as reduced proportions of complex rotating, up stretching and hunching movement. These behavioral alterations, most pronounced on day 4, demonstrate that chronic cocaine exposure not only enhances locomotion but also impairs environmental threat perception and motor coordination. The comprehensive behavioral profiling reveals a transition from natural exploratory behaviors to stereotypic, hyperactive patterns characteristic of psychostimulant exposure.

Fig. 2. Chronic cocaine exposure impaired motor coordination and threat perception.

A The ethogram of seven annotated behaviors during the test (left), and the top and side view of average skeleton diagram (right) of mice under four conditions. Movement intensity (MI) values, normalized to a 0 ~ 1 scale (arbitrary units, a.u.), are visualized as heatmaps overlaid on skeleton representations, showing both spatial distribution and relative intensity of movements. B Overall illustrations (upper) and statistical comparisons (lower) of annotated behaviors in mice. (One-way ANOVA with Dunnett’s multiple comparisons test: Running: F_3,24 = 26.52, P < 0.0001; Saline vs. Cocaine_d1, P = 0.0143; Saline vs. Cocaine_d4, P < 0.0001; Saline vs. Cocaine_d7, P = 0.0002. Sniffing: F_3,24 = 4.618, P = 0.0109; Saline vs. Cocaine_d1, P = 0.0090; Saline vs. Cocaine_d4, P = 0.0266; Saline vs. Cocaine_d7, P = 0.5932. Up stretching: F_3,24 = 6.426, P = 0.0024; Saline vs. Cocaine_d1, P = 0.0719; Saline vs. Cocaine_d4, P = 0.0006; Saline vs. Cocaine_d7, P = 0.0466. Head rising: F_3,24 = 4.455, P = 0.0127; Saline vs. Cocaine_d1, P = 0.0758; Saline vs. Cocaine_d4, P = 0.0066; Saline vs. Cocaine_d7, P = 0.6623. Hunching: F_3,24 = 3.979, P = 0.0196; Saline vs. Cocaine_d1, P = 0.1890; Saline vs. Cocaine_d4, P = 0.0075; Saline vs. Cocaine_d7, P = 0.0580. Rotating: F_3,24 = 3.282, P = 0.0382; Saline vs. Cocaine_d1, P = 0.7812; Saline vs. Cocaine_d4, P = 0.0211; Saline vs. Cocaine_d7, P = 0.1461.). All error bars represent mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, saline vs. cocaine_d1 or cocaine_d4 or cocaine_d7. N = 7 mice.

Chronic cocaine exposure remodels ECM structure in the mPFC

To investigate cocaine-induced ECM modifications, we performed WFA staining on the mouse prefrontal cortex following chronic cocaine treatment. WFA specifically labels both PNN and interstitial ECM through its recognition of chondroitin sulfate glycosaminoglycan (CS-GAG) chains on CSPGs [8]. Our analysis revealed clear WFA labeling of PNN structures and diffuse ECM throughout the mPFC (Fig. 3A). Quantitative assessment demonstrated that cocaine exposure significantly increased interstitial WFA signal intensity in both the prelimbic (PrL) and infralimbic (IL) cortices, while PNN density remained unchanged (Fig. 3B-E). Layer-specific analysis identified distinct spatial patterns of ECM remodeling. While both groups showed a characteristic WFA intensity peak approximately 160 μm from the pia mater (corresponding to layer 2/3), cocaine-induced WFA enhancement was most pronounced in deeper cortical layers (layer 5; Fig. 3F-G). Consistent with previous reports [8, 29], our observations confirmed that PNN in the cortex is highly associated with PV⁺ interneurons that are mostly found in layers II/III and V. Notably, cocaine exposure significantly increased the proportion of PV⁺ neurons enveloped by PNNs in the PrL (Fig. 3H), without altering the overall density of PV⁺ neurons or total PNN-enwrapped neurons (Supplementary Fig. 2A-B). Further classification of PNNs based on intensity revealed that cocaine-exposed animals exhibited increased percentage of high- and decreased medium-intensity PNNs in the PrL compared to controls (Fig. 3I). As PNN intensity reflects the density of the surrounding matrix, these findings suggest cocaine exposure induces a general strengthening of PNNs in the mPFC. Together, these results demonstrate that chronic cocaine exposure drives extensive ECM reorganization, characterized by: (1) increased interstitial matrix CSPG deposition, (2) enhanced PNN intensity enwrapping PV⁺ interneurons, and (3) layer-specific patterns of remodeling. These changes point to a selective strengthening of the ECM network, particularly around inhibitory neurons in specific cortical laminae.

Fig. 3. Chronic cocaine exposure remodels ECM structure in the mPFC.

A Experiment scheme of mice received seven days continuous cocaine injections and then were sacrificed for later WFA staining (left). The higher-resolution magnified images of mPFC ECM structure. Scale bar, 20 μm. The PNNs outlined by dashed lines exhibited lattice-like assemblies that envelop neuronal soma, proximal dendrites, and axon initial segments. The areas outside the dashed lines are considered to be the interstitial ECM, which is diffusely distributed as the neural interstitial matrix between cells of the CNS parenchyma. Although often considered as background fluorescence, the interstitial ECM also contains proteoglycans and can thus be labeled by WFA. B Immunofluorescence images of WFA and NeuroTrace in the coronal slice of mPFC from saline and cocaine-treated mice. Scale bar, 200 μm. C, E Quantification analysis of normalized total WFA intensity (C, Unpaired t-test: Saline vs. Cocaine: PrL, t = 3.324, P = 0.0040; IL, t = 2.121, P = 0.0499), average interstitial WFA intensity (D, Unpaired t-test: Saline vs. Cocaine: PrL, t = 4.090, P = 0.0008; IL, t = 2.464, P = 0.0254) and PNN density (E) in the PrL and IL. F Enlarged confocal images of WFA and PV staining in the mPFC from two groups. The dotted lines delineate different layers of PrL and IL from Layer I to Layer VI according to Allen’s brain atlas. PV staining is in Magenta, and WFA in green. Scale bar, 200 μm. G The intensity distribution of WFA from midline to deeper layer of PrL (Two-way ANOVA: Saline vs. Cocaine: interaction: F_{549, 9350} = 1.509, P < 0.0001; Distance: F_{549, 9350} = 6.938, P < 0.0001; Treatment: F_{1, 9350} = 2195, P < 0.0001) and IL (Two-way ANOVA: Saline vs. Cocaine: interaction: F_{546, 9299} = 0.7614, P > 0.9999; Distance: F_{546, 9299} = 5.081, P < 0.0001; Treatment: F_{1, 9299} = 652.5, P < 0.0001) in two groups. H Quantification analysis of percentage of PV⁺ neuron in total PNN enwrapped neurons in the PrL (Unpaired t-test. Saline vs. Cocaine: t = 2.288, P = 0.0382) and the IL. I The percentage of low, medium and high intensity PNNs in the PrL (Unpaired t-test: Saline vs. Cocaine: Medium, t = 2.174, P = 0.0433; High, t = 2.727, P = 0.0138) and IL. The top 25% and the lowest 25% intensity of PNN in saline group were used to define low and high intensity PNN separately. All error bars represent mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, saline vs. cocaine. N = 3 ~ 4 mice.

Transcriptional changes in ECM components following cocaine exposure

The brain ECM comprises a complex network of fibrous proteins (collagens, elastin), glycoproteins (laminin, fibronectin, proteoglycans), and associated molecules (hyaluronan, matricellular proteins, and regulatory enzymes) that collectively form both PNN and the interstitial ECM [30, 31]. This structural scaffold interacts with neurons through specific receptors including integrins (Fig. 4A). To investigate cocaine-induced ECM alterations, we analyzed mRNA levels of key ECM components in the mPFC following chronic cocaine exposure (Fig. 4B). Our results revealed significant transcriptional changes, with a marked upregulation of ncan (encoding the CSPG Neurocan) and downregulation of col1a2 (encoding collagen type I α2 chain). Given that Neurocan is a major component of CSPGs recognized by WFA, the increased ncan expression may underlie the enhanced WFA signaling observed in cocaine-treated mice. Immunohistochemical validation demonstrated significantly elevated Neurocan protein expression in both the PrL and IL cortices of cocaine-exposed animals (Fig. 4C, Supplementary Fig. 2C). These findings demonstrate that chronic cocaine exposure induces substantial remodeling of the extracellular microenvironment through coordinated regulation of both structural (collagen) and regulatory (CSPG) ECM components, with particularly pronounced effects on Neurocan-containing matrix elements.

Fig. 4. ECM compositional changes after chronic cocaine exposure and chABC-mediated degradation of PNNs and interstitial ECM.

A Graphical depiction of major composition of extracellular matrix in the brain. B The mRNA level of different ECM-related genes in the mPFC after chronic cocaine treatment. (Unpaired t-test with Welch’s correction: Saline vs. Cocaine: ncan, t = 2.841, P = 0.0449; col1a2, t = 3.295, P = 0.0179). N = 5 ~ 6 mice. C The density analysis of Neurocan in the PrL (Unpaired t-test: Saline vs. Cocaine: t = 3.856, P = 0.0004) and IL (Unpaired t-test: Saline vs. Cocaine: t = 3.474, P = 0.0011) of mice treated with saline or cocaine. N = 4 mice. D Schematic representation of WFA-mediated CSPG detection and chABC-induced ECM remodeling. WFA bound to CSPGs in the ECM, enabling visualization of PNNs and interstitial ECM. chABC cleaved GAG chains on CSPGs, disrupting ECM integrity and promoting neural plasticity. E WFA Immunostaining showing PNN and interstitial ECM disruption effects at different timepoints after single chABC injection in the prelimbic cortex. Scale bar, 200 μm. F Quantification analysis (One-way ANOVA with Dunnett’s multiple comparisons test) of normalized total WFA intensity (F_{3, 49} = 6.911, P = 0.0006; Control vs. chABC_d3, P = 0.0001; Control vs. chABC_d7, P = 0.0388; Control vs. chABC_d14, P = 0.0332), PNN intensity (F_{3, 49} = 26.27, P < 0.0001; Control vs. chABC_d3, P < 0.0001; Control vs. chABC_d7, P < 0.0001; Control vs. chABC_d14, P = 0.0004) and average interstitial WFA intensity (F_{3, 49} = 16.35, P < 0.0001; Control vs. chABC_d3, P < 0.0001; Control vs. chABC_d7, P < 0.0001; Control vs. chABC_d14, P < 0.0001) in the PrL after chABC treatment. N = 3 mice. G, H Mice received intra-PrL micro-infusion of chABC or control drugs (Penicillinase, Pen) before chronic cocaine treatment (G). After last cocaine treatment, mice were sacrificed for WFA staining (H). Scale bar, 200 μm. I, J Normalized total WFA intensity (I, Unpaired t-test: Pen vs. chABC: t = 15.43, P < 0.0001), PNN density (I, Unpaired t-test: Pen vs. chABC: t = 12.34, P < 0.0001), average interstitial WFA intensity (I, Unpaired t-test: Pen vs. chABC: t = 7.015, P < 0.0001) and Neurocan puncta density (J, Unpaired t-test: Pen vs. chABC: t = 4.974, P < 0.0001) were quantified in the PrL of chABC-treated or Pen-treated cocaine mice. N = 7 mice. All error bars represent mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001, saline vs. cocaine OR control vs. chABC_d3 or chABC_d7 or chABC_d14. ^#### P < 0.0001, Pen vs. chABC.

chABC-mediated degradation of PNNs and interstitial ECM

Previous studies have established that chABC-mediated degradation of PNN can disrupt addiction-related behaviors [32]. In this study, we characterized the temporal and spatial effects of chABC administration on the ECM, focusing on the dynamic plasticity of this structural network (Fig. 4D-F). Following a single microinfusion into the PrL, chABC effectively eliminated PNNs and interstitial CSPG within 3 days, with this effect persisting for at least two weeks (Fig. 4F). As an enzyme that specifically cleaves CSPGs (Fig. 4D), chABC treatment was particularly relevant given our observation of upregulated Neurocan (a key CSPG protein) in cocaine-treated mice. By comparing chABC-treated animals with penicillinase (Pen)-treated controls, we confirmed successful degradation of CSPGs, as evidenced by significantly reduced total WFA intensity, decreased PNN density, diminished interstitial WFA intensity, and lower Neurocan expression (Fig. 4G-J, Supplementary Fig. 2D). These findings demonstrate that chABC administration effectively disrupts both PNNs and interstitial ECM components in the PrL of cocaine-exposed mice, providing a valuable tool for investigating CSPG-dependent mechanisms in addiction-related plasticity.

chABC treatment unexpectedly exacerbates cocaine effects

To assess the behavioral consequences of CSPG degradation, we examined cocaine-induced behaviors following intra-PrL chABC microinfusion using our established 3D motion capture system (Fig. 5A). Principal component analysis of kinematic features revealed distinct behavioral profiles, with the Saline+Pen group clearly separating from both cocaine-treated groups, and the Cocaine+chABC group showing the most pronounced divergence from controls (Fig. 5B). While cocaine administration in Pen-treated mice produced the expected increase in frame distance, chABC pretreatment unexpectedly amplified this hyperlocomotor response, with enhanced movement velocity across multiple body regions (Fig. 5C, Supplementary Fig. 3A). Furthermore, chABC treatment exacerbated cocaine-induced postural abnormalities, including greater increases in body angle and length compared to Pen controls (Fig. 5D-E; Supplementary Fig. 3B). Detailed behavioral analysis demonstrated that chABC potentiated characteristic cocaine-induced phenotypes: mice showed increased repetitive running (by 121%) and ground-sniffing behaviors (by 142%), while displaying greater reductions in threat-assessment postures (up stretching decreased by 48% and head rising decreased by 28%) and coordinated movements (rotating reduced by 64%) compared to Cocaine+Pen animals (Fig. 5F-G; Supplementary Fig. 3C-D). These findings reveal that rather than attenuating cocaine’s effects, chABC-mediated ECM degradation unexpectedly exacerbates the triad of cocaine-induced behavioral alterations: hyperlocomotion, impaired motor coordination, and deficient environmental threat assessment.

Fig. 5. chABC treatment unexpectedly exacerbates cocaine-induced locomotion and threat perception deficit.

A The spontaneous behavior tests were conducted after 7 consecutive days of cocaine treatment in mice receiving intra-PrL chABC microinfusions. B Principal component analysis of kinematic parameters (left) and movement features (right) of Saline+Pen group, Cocaine+Pen group and Cocaine+chABC group. C The frame distance of mouse during the tests. (Unpaired t-test: Saline+Pen vs. Cocaine+Pen: t = 5.055, P = 0.0003; Cocaine+Pen vs. Cocaine+chABC: t = 2.332, P = 0.0341). N = 7 ~ 10 mice. D, E The example traces (D) and quantitative comparison of average body angle (E, Unpaired t-test: Saline+Pen vs. Cocaine+Pen: t = 6.140, P < 0.0001; Cocaine + Pen vs. Cocaine+chABC: t = 2.039, P = 0.0595) and body length (E, Unpaired t-test: Saline + Pen vs. Cocaine+Pen: t = 3.776, P = 0.0026; Cocaine+Pen vs. Cocaine+chABC: t = 3.194, P = 0.0060) among three groups. F The ethogram of seven annotated behaviors (left), and the top and side view of average skeleton diagram (right) of mouse. G Overall illustrations (upper left) and statistical comparisons (upper right and lower) of annotated behaviors among three groups. Unpaired t-test: Running: Saline+Pen vs. Cocaine+Pen: t = 6.859, P < 0.0001; Cocaine+Pen vs. Cocaine+chABC: t = 2.298, P = 0.0363. Sniffing: Cocaine+Pen vs. Cocaine+chABC: t = 2.340, P = 0.0335. Up stretching: Saline+Pen vs. Cocaine+Pen: t = 2.865, P = 0.0142; Cocaine+Pen vs. Cocaine+chABC: t = 2.772, P = 0.0142. Head rising: Cocaine+Pen vs. Cocaine+chABC: t = 2.141, P = 0.0491. Rotating: Cocaine+Pen vs. Cocaine+chABC: t = 3.659, P = 0.0023.). H mRNA levels of ECM-related genes were detected in the mPFC of chABC-treated cocaine mice. (Saline+Pen vs. Cocaine+Pen: Unpaired t-test: ncan: t = 2.545, P = 0.0315. col1a2: t = 2.611, P = 0.0260. Cocaine+Pen vs. Cocaine+chABC: Unpaired t-test with Welch’s correction: spp1: t = 3.038, P = 0.0384. lox: t = 4.407, P = 0.0111. ncan: t = 3.047, P = 0.0315. Unpaired t-test: col1a2: t = 7.301, P < 0.0001.fn1: t = 3.408, P = 0.0078). N = 5 ~ 6 mice. All error bars represent mean ± SEM. * P < 0.05. ** P < 0.01, *** P < 0.001, **** P < 0.0001, Saline+Pen vs. Cocaine+Pen. ^# P < 0.05, ^## P < 0.01, ^#### P < 0.0001, Cocaine+Pen vs. Cocaine+chABC.

chABC induced reorganization of ECM components in the PrL

To investigate the molecular mechanisms underlying chABC’s paradoxical effects on cocaine-induced behaviors, we analyzed ECM-related gene expression profiles in the mPFC. Our results revealed that while chronic cocaine exposure significantly increased ncan (Neurocan) mRNA levels, chABC treatment effectively reversed this upregulation (Fig. 5H). Similarly, cocaine-induced downregulation of col1a2 (Collagen type I α2 chain) was completely rescued by chABC administration. Notably, although cocaine alone did not significantly alter expression of fn1 (Fibronectin), spp1 (Osteopontin), or lox (Lysyl oxidase), chABC treatment induced dramatic upregulation of these genes - particularly spp1 (35-fold increase) and lox (11-fold increase) compared to baseline levels (Fig. 5H). These findings demonstrate that chABC triggers comprehensive ECM remodeling characterized by: (1) normalization of cocaine-altered CSPG (ncan) and collagen (col1a2) expression, and (2) massive induction of matricellular (spp1) and collagen-modifying (lox) factors. The particularly striking upregulation of Osteopontin, a known modulator of neuroimmune interactions [33], and lysyl oxidase, a key enzyme in ECM cross-linking and stiffening [34, 35], suggests these molecules may play central roles in mediating chABC’s unexpected behavioral effects. This transcriptional reprogramming likely creates an altered extracellular environment that potentiates cocaine’s locomotor-activating effects while impairing normal threat assessment behaviors, possibly through modified matrix stiffness and enhanced neuroimmune signaling.

Knock-down of Lox partially rescues chABC-induced threat perception deficits in cocaine-exposed mice

To determine whether the compensatory upregulation of specific extracellular matrix genes mediates the unexpected behavioral effects of chABC, we examined whether targeted knock-down of Lox or Spp1 could reverse the exacerbated cocaine-induced phenotypes. Following intra-PrL chABC microinfusion and subsequent chronic cocaine administration (Fig. 6A), siRNA-mediated gene silencing was performed. Quantitative PCR confirmed successful and selective knockdown of Lox and Spp1 mRNA expression (Fig. 6B). Principal component analysis revealed that in the kinematic feature spaces analysis (Fig. 6C, left), data points from the both Lox-siRNA group and Spp1-siRNA group failed to show clear separation from those of the scramble control group. In contrast, in the movement features analysis (Fig. 6C, right), the Lox-siRNA group but not Spp1-siRNA exhibited a modest but observable separation trend from the scramble controls. Neither Lox nor Spp1 knock-down altered the chABC-potentiated hyperlocomotion, as total frame distance and speed of different body parts showed no statistical significance across all three groups (Fig. 6D, Supplementary Fig. 4A). Similarly, general postural abnormalities including body angle and body length induced by chABC were unaffected (Supplementary Fig. 4B-C). However, Lox knock-down specifically increased average body height compared to the scramble control, suggesting vertical postural alteration (Fig. 6E). Ethographic analysis and behavioral quantification demonstrated that while neither siRNA affected the proportion of repetitive running behavior, Lox silencing elicited a significant behavioral shift. It reduced sniffing behaviors and markedly increased threat-assessment postures, including up stretching and head rising behaviors (Fig. 6F-G, Supplementary Fig. 4D-E). Collectively, these results indicate that Lox, but not Spp1, critically contributes to the chABC-induced deficit in environmental threat perception in cocaine mice. Importantly, Lox knock-down selectively rescued this behavioral impairment without normalizing the hyperlocomotor or general postural effects of chABC, thereby dissociating the molecular mechanism mediating threat vigilance from those governing locomotor activation.

Fig. 6. Knock-down of Lox partially rescues chABC-induced threat perception deficit in cocaine mice.

A The behavioral testing schedule was as follows: (1) Intra-PrL microinfusion of chABC; (2) 7 consecutive days of cocaine treatment, beginning 3 days after chABC infusion; (3) Intra-PrL microinfusion of siRNA 3 days before the (4) final spontaneous behavior test. Mice were sacrificed 2 h after last cocaine injection. N = 7 mice. B mRNA levels of Lox (Scramble vs. Lox-siRNA: Unpaired t-test: t = 3.166, P = 0.0081) and Spp1 (Scramble vs. Spp1-siRNA: Unpaired t-test: t = 2.881, P = 0.0138) were detected in the prelimbic cortex of cocaine mice after chABC and siRNA mutual microinfusion. N = 7 mice. C Principal component analysis of kinematic parameters (left) and movement features (right) of Scramble group, Lox-siRNA group and Spp1-siRNA group. D The frame distance of mouse during the tests. E The example traces (left) and quantitative comparison of average body height (right, Unpaired t-test: Scramble vs. Lox-siRNA: t = 2.257, P = 0.0435) among three groups. F The ethogram of seven annotated behaviors (left), and the top and side view of average skeleton diagram (right) of mouse. G Overall illustrations and statistical comparisons of annotated behaviors among three groups. (Unpaired t-test: Scramble vs. Lox-siRNA: Sniffing: t = 2.086, P = 0.0590; Up stretching: t = 2.316, P = 0.0390; Head rising: t = 2.710, P = 0.0190). All error bars represent mean ± SEM. * P < 0.05, ** P < 0.01, Scramble vs. Lox-siRNA or Spp1-siRNA.

Discussion

Using a 3D motion-capture system, we characterized the behavioral consequences of chronic cocaine exposure, revealing two distinct but interrelated effects: pronounced locomotor hyperactivity and disrupted threat-assessment behaviors (particularly reduced up stretching). Detailed kinematic analysis further demonstrated impaired motor coordination, evidenced by decreased complex movements like rotating. To investigate the neurobiological basis of these behavioral changes, we examined cocaine-induced modifications of the ECM in the mPFC. Histological and molecular analyses identified three key alterations: (1) elevated WFA intensity in layer 5 PNNs and interstitial matrix, (2) upregulated ncan and downregulated col1a2 mRNA expression, and (3) increased PNN ensheathment of PV⁺ interneurons. Surprisingly, while chABC successfully degraded CSPGs in the PrL, it aggravated cocaine’s behavioral effects, intensifying hyperactivity and further impairing threat assessment. This unexpected outcome is probably associated with dramatic transcriptional changes in ECM components - most notably a 35-fold increase in spp1 mRNA and 11-fold increase in lox mRNA. However, subsequent siRNA experiment demonstrated that specific knockdown of Lox in the PrL, but not Spp1, could partially rescue the chABC-mediated exacerbation of threat perception deficits in cocaine mice. Together, these findings reveal a previously unrecognized role for ECM remodeling in modulating cocaine’s behavioral effects, where CSPG degradation triggers compensatory changes that potentiate rather than attenuate cocaine-induced phenotypes.

Our study reveals that chronic cocaine exposure enhanced interstitial ECM components and increased PNN ensheathment of PV⁺ interneurons in the mPFC. These findings contribute to growing evidence that addictive substances dynamically modulate PNNs in a spatiotemporal- and region-specific manner [10]. While previous reports have shown variable PNN alterations depending on administration paradigms - including acute decreases [19], abstinence-related reductions [36], and conditioned place preference-independent changes [37], our work demonstrates that chronic continuous exposure preferentially strengthens PNNs surrounding PV⁺ interneurons in layers 2/3 and 5 of the prelimbic cortices. The laminar specificity of these effects is particularly noteworthy. The observed PNN intensification in layer 5, where PV⁺ interneurons are densely distributed [8], may have particularly significant functional consequences. As fast-spiking GABAergic interneurons, PV⁺ cells play crucial roles in regulating cortical microcircuits through their connections with pyramidal neurons and other interneuron subtypes [29]. The PNNs surrounding these neurons, composed primarily of CSPGs [7], are well-positioned to modulate their synaptic integration and firing properties. Previous electrophysiological evidence showed that chronic cocaine exposure reduces PV⁺ interneurons excitability by altering their synaptic input patterns - reducing the amplitude and frequency of mEPSCs while increasing the amplitude and frequency of mIPSCs into them [19]. The cocaine-induced PNN enhancement we observed may contribute to these functional changes through several potential mechanisms: First, by restricting structural plasticity of PV⁺ neuron dendrites, the enhanced PNNs may limit their capacity for synaptic reorganization. Second, these ECM modifications could modify perisomatic inhibition of pyramidal cells, altering network dynamics. Third, the changes may affect the balance between excitatory and inhibitory inputs to PV⁺ neurons themselves. This PNN-mediated modulation of inhibitory microcircuits may ultimately lead to pyramidal neuron disinhibition, potentially contributing to the hyperexcitability observed in addiction models. The layer-specific nature of these changes suggests they may preferentially affect particular cortical microcircuits. For instance, PFC pyramidal neurons projecting to subcortical areas like nucleus accumbens [38], dorsal medial striatum [39] or lateral habenula [40], might contribute to cocaine-induced hyperlocomotion. Conversely, pyramidal neurons involved in intra-cortical, and corticothalamic circuits [41–43] might more directly underlie the deficits in threat perception. However, direct electrophysiological recordings, projection-specific viral tracing and cell-type-specific manipulations will be essential to verify these hypotheses.

Beyond perineuronal net alterations, our study reveals significant cocaine-induced modifications in the diffuse ECM of the mPFC, as evidenced by elevated WFA signaling intensity. This finding is particularly significant given that the diffuse matrix represents approximately 98% of total CNS CSPGs, while PNNs account for only 2% [44]. The complex molecular composition of the ECM creates a dynamic microenvironment where neurons and glia continuously interact through reciprocal signaling [45]. Our transcriptional profiling identified two particularly noteworthy ECM alterations: upregulation of ncan mRNA and downregulation of col1a2 mRNA. While these molecules have been relatively understudied in addiction contexts, emerging evidence suggests their potential roles in neural plasticity. Neurocan downregulation in the PrL increases stress vulnerability in rats [46], while collagen I reduction inhibits tumor progression in glioblastoma models [47], highlighting their functional significance in CNS pathophysiology.

The ECM transcriptome signature following chABC treatment reveals even more profound reorganization, characterized by coordinated downregulation of CSPGs (ncan) and upregulation of structural (col1a2, fn1), matricellular (spp1), and cross-linking (lox) components. This pattern suggests a fundamental shift in matrix composition that likely alters the biomechanical and biochemical properties of the extracellular microenvironment. Two molecules stand out for their dramatic changes and established roles in neural function: LOX and SPP1. Our subsequent siRNA experiments further reveal a divergent, gene-specific role for these ECM components in cocaine-associated behaviors. Specifically, we conducted intra-PrL injection of Lox-siRNA, Spp1-siRNA or scramble siRNA in chABC-pretreated cocaine mice and assessed their locomotor activity and threat vigilance-related behaviors. Results showed that knockdown of Lox, but not Spp1, partially rescued the chABC-induced exacerbation in threat-perception behaviors without affecting locomotor activity. The protein product of Lox is Lysyl oxidase, a secreted enzyme that increases the abundance of its catalytic substrates, including collagen and elastin [34, 48]. This activity facilitates ECM remodeling and cross-linking, ultimately enhancing tissue tensile strength and maintaining ECM structural integrity [48]. Currently, research on LOX in the brain remains limited. Nevertheless, emerging evidence indicates its potential involvement in various neurological disorders, including Alzheimer’s disease [49], bipolar disorder [50], hypoxic-ischemic injury [35], and amyotrophic lateral sclerosis (ALS) [51]. Microarray analysis and ex vivo LOX inhibition experiment revealed that LOX is a profound regulator of astrocyte morphology and proliferation [50]. Immunostaining results demonstrated that LOX displayed significant enrichment in spinal and brain neurons in ALS mouse models [51]; similarly, it colocalized with amyloid-beta pathology in human Alzheimer’s disease samples [49]. Notably, recent work showed that elevated LOX increased brain stiffness through enhancing ECM components at the hypoxic-ischemic injury site, which triggered mechanosensitive Piezo1 channel activation, and further induced neuronal ferroptosis via a GPX4-dependent pathway. Pharmacological suppression of LOX attenuated brain neuronal ferroptosis and mitigated learning and memory deficits [35]. Hence, our study suggests that Lox-mediated collagen cross-linking might increase ECM stiffness in the PrL thereby potentiating the threat assessment impairment in cocaine mice.

Even though Spp1 knockdown did not reverse chABC-exacerbated hyperlocomotion and threat-vigilance behavior in our study, its drastic upregulation in the PrL requires more discussion. SPP1 is generally considered as a pro-inflammatory mediator, known to regulate microglial activation and function [52, 53]. However, the role of SPP1 is complex and context-dependent, as its expression marked transcriptionally distinct microglial subsets with both anti- or pro-inflammatory phenotypes [54, 55]. Notably, De Schepper et al. [33] demonstrated that SPP1 tunes microglial phagocytosis of dendrite spines through paracrine pathway. Several studies have implicated SPP1 in cognitive processes [56, 57], and prior work showed that intra-PrL chABC infusion impaired cocaine-conditioned memory [11]. Thereby, we speculate that SPP1 may contribute to such cognitive deficits by modulating microglial phagocytic activity and synaptic remodeling within prefrontal circuits. Given that SPP1 also acts as a critical neuroimmune mediator, it may contribute to broader network dysregulation and mediate behavioral phenotypes beyond the specific behaviors measured in our study. Collectively, increased LOX activity could stiffen the matrix, restrict structural plasticity while SPP1 promotes aberrant synaptic stabilization. These changes may be particularly impactful in the mPFC, where precise balance between stability and plasticity is crucial for executive function. Our findings suggest that chronic cocaine exposure induces a maladaptive ECM state that combines reduced CSPG-dependent plasticity with enhanced tissue stiffness and immune signaling- a combination that may underlie both the cocaine-related behaviors and the unexpected effects of chABC treatment.

Critically, our qPCR result and siRNA experiment collaborate to demonstrate that chABC-mediated CSPGs degradation specifically results in an active, targeted ECM remodeling process rather than a chaotic non-specific inflammatory cascade. Moreover, previous study demonstrated that microinfusion of chABC in the PrL inhibited the acquisition and reconsolidation of cocaine-conditioned memory [11], while our study discovered that chABC exacerbates hyperlocomotion and threat-assessment deficits in cocaine mice. Therefore, these observations suggest that chABC’s behavioral consequences are mediated by targeted ECM-neural circuit interactions (e.g., altered PV⁺ interneuron function, ECM stiffness changes) rather than non-specific neural dysfunction. We acknowledge we cannot fully exclude the possibility that mild, localized neuroinflammation or circuit instability contributes to the behavioral exacerbation. Future studies will address this gap by quantifying microglial polarization, cytokine levels, and neuronal firing dynamics following chABC treatment to clarify whether these secondary effects play a role.

Our study employed an advanced 3D motion capture system combined with machine learning algorithms to provide unprecedented resolution in characterizing cocaine-induced behavioral changes. While clinical observations document cocaine’s psychomotor effects in humans, including tachycardia, stereotypic movements, euphoria, and altered perception [58], rodent studies have traditionally relied on simplified locomotor metrics [59]. Our machine learning-based analysis revealed three distinct behavioral alterations during peak drug exposure (20-40 min post-injection), coinciding with maximal dopamine levels in reward-related circuits [60, 61]. First, we observed a dramatic behavioral shift characterized by compulsive running bouts and reduced static postures, consistent with cocaine’s known psychomotor stimulant properties. More significantly, our system detected a 31% reduction in vertical exploration (up stretching behavior), which is considered a reliable marker of environmental threat assessment in rodents [21]. These findings parallel clinical observations of environmental disengagement during cocaine intoxication [58]. Additionally, the decreased frequency of complex motor patterns like rotating and hunching suggests cocaine may transiently impair motor coordination, though this requires further experimental verification. The temporal precision of our 3D tracking approach allowed us to directly correlate these behavioral changes with the known neurochemical timeline of cocaine action. The observed behavioral triad—hyperlocomotion, threat assessment deficits, and motor coordination changes—likely reflects dopamine-driven disruption of cortico-striatal circuits during the drug’s euphoric phase. Notably, the threat perception deficits we quantified may represent a preclinical analog to the impaired judgment and risk assessment seen in human cocaine users [58]. Furthermore, our ability to detect subtle motor coordination changes suggests this approach could reveal previously overlooked dimensions of drug effects that may relate to the loss of behavioral control characteristic of addiction progression. These findings establish 3D behavioral phenotyping as a powerful tool for investigating the multidimensional effects of psychostimulants with translational relevance to human drug responses.

We observed that cocaine-induced locomotion was most pronounced on day 4 instead of day 7, which diverges from the progressive pattern of cocaine-induced sensitization [62]. However, our study focused on cocaine’s direct pharmacological effects on locomotion, rather than establishing context-dependent locomotor sensitization. Standard sensitization requires repeated drug-context pairing to induce gradual locomotion augmentation [62–64]. Our protocol deviated from this paradigm: cocaine was administered primarily in the home cage, and activity was measured only intermittently in the open field (Days 1, 4, 7) within a focused post-injection window. While this approach allowed us to capture stable acute hyperlocomotion [65], it was not optimized to study cocaine-induced sensitization.

Our study still leaves much room for future exploration. Firstly, it is crucial to identify the exact neuron types with enhanced PNN intensity and explore their impact on the synaptic plasticity in mPFC microcircuit. Furthermore, future work should include mass spectrometry-based proteomic analysis in the mPFC to comprehensively verify alterations in ECM composition in protein level. Notably, given that sex differences in ECM biology may contribute to addiction-related behaviors in a sex-dependent manner [66], it will be necessary to replicate these experiments in female mice to confirm the generalizability of our findings. At last, up stretching (rearing) behavior has been reported to increase significantly following cocaine administration in prior literature [65, 67–69]. The observed inconsistencies may be attributed to: use of scoring systems, intermittent time sampling, divergent statistical time windows and paradigm-specific effects [59]. Thus, we need to fully leverage the 3D behavioral analysis technology employed in this study to conduct more comprehensive and prolonged observations of cocaine-induced behavioral alterations in mice.

Collectively, our findings demonstrate that chronic cocaine exposure induces significant ECM reorganization in the mPFC, characterized by enhanced PNNs and interstitial matrix remodeling. Surprisingly, rather than ameliorating cocaine’s behavioral effects, chABC-mediated degradation of CSPGs exacerbated three core behavioral manifestations: potentiated locomotor hyperactivity, impaired motor coordination, and amplified deficits in environmental threat assessment. These paradoxical effects are mechanistically linked to a compensatory upregulation of ECM-stiffening components, most notably LOX, as evidenced by our rescue experiments where Lox knockdown selectively restored threat-assessment behaviors. Our study thus reveals a maladaptive, LOX-driven shift in prefrontal ECM homeostasis as a novel mechanism sustaining addiction-related behaviors, offering a novel perspective for targeted interventions.

Author contributions

LS conceived the project, designed experiments, supervised the project, and revised the manuscript. NW offered expert critique and recommendations to the project. XL carried out the behavioral and molecular experiments, performed the statistical analysis and wrote the manuscript. YH, XW and JY provided critical revision of the manuscript for intellectual contents. YH and XW aided in 3D spontaneous behavior experiment including data analysis. ZL, and XF assisted with immunofluorescence experiment. SG, XY and XZ contributed in intellectual inputs for the project. YL provided critical comments and feedback to the project. All authors agreed on the final manuscript before submission.

Funding

This work was supported by National Natural Science Foundation of China grant 82101231 (L.S.), Beijing Natural Science Foundation of China grant 7252082 (L.S.), Peking University Talent Fund 68263Y1230 (L.S.) and National Natural Science Foundation of China grant 81901350 (N.W.).

Data availability

Data are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41398-026-04014-5.