C57BL/6N mice (aged 8–10 weeks) were obtained from the Beijing Vital River Laboratory Animal Technology Co., Ltd. and were housed at the Laboratory Animal Center of the Institute of Zoology under standard conditions: 12/12 h light/dark cycle, free access to food and water, noise below 60 dB, a constant comfortable temperature (20–26°C), and humidity (40–70%) Supplementary Video. We ensured that animals did not experience unnecessary pain or fear. All animal experiments were approved by the Animal Care and Use Committees of the Institute of Zoology, Chinese Academy of Sciences (IOZ-IACUC-2021-105).

To delete microglia in mice, PLX3397 (Selleckchem), a CSF1R inhibitor, was formulated into AIN-76A standard diet at 290 and 600 mg/kg. In order to pre-delete microglia, mice were fed with medicated feed for 21 days (or mentioned in the article) before receiving an SCI-inducing surgery. To delete microglia in mature scars, the medicated feed was administered 10 weeks after the surgery (when scar formation was complete and remained stable) and continued for 10 weeks. An AIN-76A standard diet was used as the corresponding vehicle.

The procedure of C57BL/6N mouse T9 spinal cord crush is similar to what was described previously (Li et al., 2020) with modifications. This type of injury is similar to the SCI caused by spinal fracture or dislocation, which is clinically common (Chiu et al., 2017; Bradbury and Burnside, 2019). Briefly, mice were anesthetized via an intraperitoneal injection of pentobarbital (75 mg/kg). A longitudinal incision was made on the shaved backs of the mice to cut open the skin and muscles to expose the thoracic vertebrae. Next, laminectomy was performed at T9 to expose the spinal cord, and the ninth thoracic level was crushed using forceps (JZ, WA3010) with a tip width less than 1 mm. The tip was closed for 10 s. The dura was kept intact throughout the injury process to avoid invasion of the spinal cord by muscle tissues during subsequent healing. Laminectomy was performed in the sham group, but without the crush-induced injury. Next, the open muscles and skins were sutured. Finally, mice were placed at 37°C to recover from anesthesia until their thermoregulation was re-established. Urine was emptied by manual abdominal pressure daily throughout the experiment.

Mice were perfused with normal saline under deep anesthesia to replace blood, followed by 4% paraformaldehyde (PFA) for 10 min. The 20 mm spinal cord segments, including the injured portion, were excised and placed in 4% PFA overnight and in 30% sucrose solution for 2 days to dehydrate. Frozen and dehydrated segments were cut into 20 μm thick sections and blocked with 0.5% Triton-X 100 and 5% bovine serum albumin for 1 h. The samples were incubated with primary antibodies [rat anti-glial fibrillary acidic protein (GFAP), 1:400, Invitrogen, 13-0300; rabbit anti-GFAP, 1:400, Abcam, ab7260; mouse anti- GFAP, 1:500, Millipore, MAB360; rabbit anti-β-Tubulin III (Tuj1), 1:1,000, Sigma, T2200; mouse anti-Tuj1, 1:1,000, T8660; mouse anti-Tuj1, 1:1,000, BioLegend, 801201; goat anti-Iba1, 1: 400, Abcam, ab5076; mouse anti-Iba1, 1: 400, Abcam, ab15690; rabbit anti-collagen I, 1: 400, Abcam, Ab21286] at 4°C overnight. After rinsing with phosphate-buffered saline (3 rinses, 5 min each), sections were incubated with secondary antibodies (Alexa Fluor^® 488-labeled donkey anti-rabbit IgG antibody, 1:1,000, Thermo Fisher, Waltham, MA, United States; Alexa Fluor^® 568-labeled donkey anti-mouse IgG, 1:1,000, Thermo Fisher; Alexa Fluor^® 647-labeled donkey anti-goat IgG antibody, 1:1,000, Thermo Fisher) at 25°C for 2 h in the dark. Finally, sections were rinsed with phosphate-buffered saline and mounted with a DAPI-containing sealant. The samples were observed under a laser confocal microscope (880 Ariyscan, Leica, Wetzlar, Germany). No primary antibodies were used in the controls.

Spinal cord segments (2 mm long, including the injured portion) were homogenized in an ice-cold RIPA Lysis and Extraction Buffer (89901; Thermo, CA, United States) supplemented with a protease inhibitor cocktail (78439; Thermo, CA, United States). The total protein concentration was determined by the bicinchoninic acid method (23250; Thermos, CA, United States) and was separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis at 100 V for 2 h. Separated proteins were electro-transferred onto polyvinylidene difluoride membranes at 300 mA for 90 min. The membranes were blocked with Tris-buffered saline with Tween 20 (TBST) containing 5% skimmed milk powder for 2 h and incubated with the primary antibodies (rabbit monoclonal anti-collagen I, 1:250, Abcam, ab21286; mouse monoclonal anti-GAPDH, 1:4,000; AF0006, Beyotime) at 4°C overnight. After rinsing with TBST (three rinses, 15 min each), the membranes were incubated with secondary antibodies (horseradish peroxidase-conjugated goat anti-rabbit or anti-mouse IgG, 1:1,000, Abcam) at room temperature for 2 h. Immunoreactive bands were visualized with an enhanced chemiluminescence reagent (ECL, Pierce) and quantified using ImageJ software (version 1.53c).

The virus AAV 2/9 rAAV-hSyn-EGFP-WPRE-SV40 polyA (Titer: 2.71E+12 ug/ml Cat# PT-0905, BrainVTA) was used to label the descending spinal tract. A stereotaxis apparatus was placed on both hemispheres of mice brains to construct the coordinates, with the bregma being the coordinate origin. 1 mm beneath the meninges, 0.5 μL of the virus was injected at six selected points, (1.5, 0), (3.0, 0), (1.5, 1.5), (−1.5, 0), (−3.0, 0), and (−1.5, 1.5). Samples were collected 2 weeks after the virus injection surgery.

The hindlimb motor function in mice was evaluated with the locomotor open-field Basso Mouse Scale (BMS), following the method developed by Basso et al. (Basso et al., 2006). Behavioral records were taken at the same time every week since the date of surgery. Two experienced examiners evaluated each mouse for 4 min and assigned an operationally defined score for each hindlimb.

The statistical significance of data was assessed by Student’s two-tailed paired/unpaired t-test between two groups and one-way ANOVA between three groups using Prism v8.0 (GraphPad Software, Inc.). Data distribution was assumed to be normal. Data are presented as the mean ± standard error of the mean. The statistical significance (P) was set at 0.05.

To detect the effect of microglia depletion on astrocyte scar after SCI, we adopted the T9 crush injury of the spinal cord in C57BL/6N mice (Figure 1A, Supplementary Figure S1A). To delete microglia, we fed the mice with PLX3397, a CSF1R inhibitor. This approach can specifically delete microglia from the central nervous system without significantly affecting astrocytes (Supplementary Figure S2A) (Elmore et al., 2014; Ma et al., 2020). We observed that the subject mice had a sharp decrease in food intake within 4 days after SCI surgery (Figure 1B and Supplementary Figures S1B,C; a decrease by over 80% on the first day after surgery and over 50% on the first 4 days), which may have resulted in insufficient PLX3397 intake. Meanwhile, when CSF1R inhibition was removed, microglia in subject mice repopulated at ∼20% per day (Huang et al., 2018; Ma et al., 2020). These may have together caused subsequent incomplete microglia deletion. Considering the potential effects of gavage on mice after non-healing wounds and damaged vertebrae (Bellver-Landete et al., 2019), a PLX3397 concentration of 600 mg/kg was used to ensure minimal participation of microglia in astrocyte scar formation (Elmore et al., 2014) (Figures 1C,D).

Microglia were deleted by administering PLX3397 at 600 mg/kg 3 weeks prior to the SCI surgery (Figure 2A). Astrocyte scar formation defects were observed in the microglia deletion group (Figure 2B), and astrocytes were present in the non-scar-forming state [a low activation state, Figure 2B(b1-b6)].

To detect whether the disruption of astrocyte scar during the operation caused the spread of the damaged core area, we compared the damaged core area of the two groups and found no significant expansion of the injury area when microglia were deleted (Figure 2C). Therefore, the loose astrocyte scar caused by the deletion of microglia retained its ability to isolate damage factors.

The most harmful effects of compact astrocyte scar on the central nervous system are the deletions of neurons and axons in injured core areas and inhibition of axon regeneration (Liddelow et al., 2017; Sofroniew and Vinters, 2010; Chung et al., 2013). To observe the effect of loose astrocyte scar on axon regeneration, we detected the distribution of axons in the damaged area of the loose scar group (PLX3397 group) mice at 10 weeks after SCI. The results showed that the distribution range of neurons in the loose astrocyte scar group was closer to the astrocyte scar boundary (Figure 3A). At the same time, in the PLX3397 group, the damaged spinal tract had a higher density of axons than the vehicle group in the injury core (Figures 3A,B). A similar result was observed when using a fluorescent virus to track the spinal tract’s downward projection: axons were present closer to the scar boundary in the loose group than in the vehicle group (Figures 3C–E). The weight of the mice was recorded as an index to evaluate the overall health status (Sulzbacher et al., 2022). We found that the body weight of the PLX3397 group was slightly higher than that of the wild type after surgery (Figure 3F and Supplementary Figure S1D). These results showed that the health status of the PLX3397 group was better than that of the wild type. These results suggested that the incompact astrocyte scar induced by microglia deletion had a positive effect on axon regeneration and the health status of mice.

Astrocyte activation requires the participation of ECM (Ridet et al., 1997; Hara et al., 2017) and multiple cytokines (Liddelow et al., 2017; Schiweck et al., 2021). Collagen I is an ECM secreted by multiple cells within a few days after injury (Shintani et al., 2006; Yonezawa et al., 2010; Dias et al., 2018) and is required for astrocyte activation. Additionally, we observed that microglia are an important source of collagen I (Jones et al., 2002; Li et al., 2020). Therefore, we speculated that microglia deletion would lead to reduced collagen I expression and thus prevent astrocytes from forming compact astrocyte scars. Immunofluorescence staining showed that collagen I remained concentrated in the damaged core area at all time points in the vehicle group (Figure 4A) and showed a strong co-localization association with microglia/macrophages (as activated microglia were difficult to distinguish from macrophages in morphology) (Figure 4C). In contrast, collagen I in the loose astrocyte scar group dispersed 1 week post injury (Figures 4A,B), and further observation detected no co-localization between collagen I and microglia/macrophages in the damaged core area (Figure 4C). Furthermore, we measured collagen I expression in loose astrocyte scar using western blotting and found a downregulated collagen I expression in the subacute phase (1–4 weeks) of SCI, in response to microglia deletion (Figures 4D,E). These results indicated that microglia deletion would decrease collagen I expression and change the distribution pattern in the damaged area such that activated astrocytes would exit the scar-forming state and no longer generate compact scars.

In the absence of microglia, astrocyte scars will not completely form in the pathological process of SCI, leading to an unclear scar boundary (Bellver-Landete et al., 2019). However, it is not clear whether microglia are required to maintain mature astrocyte scars. To verify that the same, we deleted microglia in mice 10 weeks (Parvin et al., 2021) post SCI and collected their spinal cord samples after another 10 weeks (Figure 5A). The results showed that the deletion of microglia in chronic scars led to the collapse of the compact scar boundaries (Figure 5B). Through further observation, we found a morphological restoration of astrocytes from the scar-forming state (having reduced dendrites and hypertrophy of the cell body) to the non-scar-forming state (having increased dendrites and a thin cell body, Figure 5B, b1-b4). Cell counting showed that the number of astrocytes in astrocyte scar decreased after deletion of microglia, whereas the area of scar did not increase (Figures 5C,D). In summary, microglia deletion in the chronic stage of astrocyte scars disrupted the compact scar arrangement by inactivating astrocytes. Therefore, microglia were required for both the formation and maintenance of compact astrocyte scars.