Astrocytic heterogeneous nuclear ribonucleoprotein U is involved in scar formation after spinal cord injury

Abstract

Astrocytes have a beneficial role in tissue repair after central nervous system (CNS) injury. Although astrocyte proliferation is activated in response to injury, the intracellular mechanisms of astrocyte proliferation during acute phase of injury are not fully clarified. In this study, by functionally screening the highly expressed genes in the pathological state of spinal astrocytes, heterogeneous nuclear ribonucleoprotein U (Hnrnpu) is identified as a potential endogenous molecule that regulates astrocyte proliferation and the following scar formation. Inhibition of Hnrnpu in astrocytes impairs the formation of astrocytic glial scar, motor function recovery, and neuronal regeneration after spinal cord injury (SCI) in mice. In human astrocytes, HNRNPU knockdown downregulates the genes related to the astrocyte functions in scar formation and neuronal regeneration. These findings uncover that modulation of endogenous astrocytic function would be a promising therapeutic avenue to restore neurological function after CNS injury.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12974-025-03351-4.

Background

Central nervous system (CNS) injury causes severe neurological impairments, such as motor, sensory, and autonomic dysfunctions. Recovery from the neurological dysfunction requires the regeneration of damaged neuronal network by the injury [1]. Regeneration of damaged neuronal network is thought to be controlled by the mechanisms intrinsic to neurons as well as extracellular environment [1]. In particular, the glial scar is a key cellular and extracellular compartment that forms the compact border around the lesion core in response to CNS injury and crucially contributes to neuronal regeneration [2]. The scar contains multiple cells, of which astrocytes are the major scar-forming cells that are activated to proliferate and migrate in response to injury. Traditionally, the glial scar is considered to be detrimental to axonal growth and functional recovery, because the glial scar forms a physical barrier that inhibits axonal regrowth. However, recent studies have revealed the beneficial and protective roles of glial scar, which is crucial for the proper resolution and recovery after SCI. For example, Anderson et al. found that astrocyte has a beneficial effect on tissue repair after spinal cord injury [3, 4]. Furthermore, promoting astrocyte proliferation through the Yes-associated protein (YAP) and glial scar formation aids functional recovery in mice after SCI [5]. Whereas astrocytes produce inhibitory molecules that impede axonal regeneration and prevent the resolution of inflammation after SCI [6, 7], they also produce axon-growth permissive and promoting molecules and generate reparative microenvironment. For example, astrocytes produce extracellular matrix molecules such as laminins provide a permissive substrate for axon growth [3, 8, 9]. Moreover, while chondroitin sulfate proteoglycans (CSPGs) produced by astrocytes are often associated with inhibitory on axonal growth, some CSPGs, such as CSPG4 and CSPG5, have permissive and promoting effects on axonal growth in certain conditions [3, 4]. As such, the roles of glial scar and astrocytes are multifaceted. Elucidating the mechanisms that regulate astrocyte reactivity and proliferation is crucial not only for better understanding the biological processes of SCI response and recovery but also for developing therapeutic strategies that promote functional recovery.

After SCI, astrocytes become activated, leading to their proliferation, migration, and the production of various molecules involved in glial scar formation and other critical processes. Past decades, much effort has focused on identifying the astrocyte-derived factors that regulates regeneration of injured neuronal network [7, 10]. However, reports on the mechanisms regulating the expression of these factors and associated astrocyte proliferation have been limited to intracellular signalling pathways induced by injury-related inflammatory cytokines such as TGF-β and IFN-γ [11, 12]. In addition, cell types other than astrocytes are also involved in the regulation of neuronal circuit repair, for example, pericytes inhibit axonal regeneration, while microglia exert tissue repair functions [13, 14]. However, the mechanisms that trigger cell-specific responses, such as controlling astrocyte reactivity, are not well understood.

Heterogeneous nuclear ribonucleoprotein (hnRNP) family proteins are involved in a wide range of cellular processes, such as cell proliferation and morphogenesis [15, 16]. In the nervous system, hnRNPs reportedly regulate the survival of neural progenitor cells [17], neurite formation [18], and synaptic plasticity changes [19]. Additionally, hnRNPs are associated with many neurodegenerative disorders such as Alzheimer’s disease, stroke, and amyotrophic lateral sclerosis (ALS) [20, 21]. Hnrnpu, also known as scaffold attachment factor A (SAF-A) and SP120, is a member of the hnRNP DNA- and RNA-binding protein family. Hnrnpu is predominantly located in the nucleus and is considered to be mainly involved in transcription [22] and alternative splicing [23]. HNRNPU gene mutations, such as microdeletions and point mutations in HNRNPU, have been linked to seizure and severe intellectual retardation [17, 24, 25], suggesting that Hnrnpu plays a crucial role in the CNS development and function. Previous studies have mainly focused on the function of hnRNPs in neurons [16, 21, 26], whereas their functions in glial cells, including astrocytes, remain insufficiently understood.

Here we performed an siRNA screening of the highly expressed genes in spinal cord astrocytes and found that Hnrnpu promotes astrocyte reactivity. Inhibiting Hnrnpu suppresses astrocyte migration and alters expression of the genes associated with astrocyte reactivity in vitro. In the spinal cord injury model, suppression of Hnrnpu decreases astrocyte proliferation, impairs astrocytic scar formation, prevents the aid of neuronal circuits, and motor function recovery. Moreover, inhibition of Hnrnpu in human astrocytes shows a similar phenotypic response in vitro as well as downregulation of the genes required for astrocyte reactivity and neuronal repair. These finding suggest that astrocytic Hnrnpu is a potential therapeutic target to promote neuronal regeneration after CNS injury.

Materials and methods

Animals

Six to eight-week-old C57BL/6 J mice were obtained from Japan SLC. Mice were maintained in 22 °C with a 12-h light–dark cycle under specific pathogen free conditions, and were given ad libitum access to food and water. Postnatal day 1 (P1) mice for primary astrocyte culture were purchased from Tokyo Laboratory Animals Science. All animal experiments were approved by the Committee on the Ethics of Animal Experiments of the National Institutes of Neuroscience, National Center of Neurology and Psychiatry (2023041).

Primary astrocyte culture

Culture of primary astrocytes were obtained from P1 mice brain [27]. Briefly, P1 mice brain were dissected and minced with fine scalpels in ice-cold phosphate-buffered saline (PBS), followed by dissociated with 0.25% trypsin/EDTA in Dulbecco’s modified Eagle’s medium (DMEM) at 37 °C for 15 min. Then, add 1/100 vol of DNaseI (10 mg/mL) at 37 °C for 1 min. After neutralization with DMEM containing 10% fetal bovine serum (FBS), cells were centrifuged at 500 × g for 10 min. Cell pellets were fully triturated in DMEM supplemented with 10% FBS and filtrated through a 70 μm nylon cell strainer. Cells were plated on poly-L-lysine (PLL; P2636, Sigma-Aldrich)-coated 75-cm flasks at a density of 2.5 × 10⁷ cells/flask and cultured at 37 °C with 5% CO₂ in DMEM supplemented with 10% FBS. Culture medium was changed every 2–3 days. Ten days after culturing, cells were washed with PBS and the adherent cells were detached by 0.05% trypsin/EDTA-PBS. Cell pellets were suspended in DMEM supplemented with 10% FBS and cultured in a new culture dish. When confluence was attained, repeat the procedure of cell passage, resulting in cultures of astrocyte for experiment.

Human primary astrocyte

Human primary astrocyte was purchased from ScienCell (#1800) with the culture of Astrocyte Medium (AM, ScienCell, #1801) under manufacturer’s instructions. Briefly, human astrocytes were cultured on a poly-L-lysine-coated dishes (2 µg/cm²) with AM culture medium were changed every three days, until the culture reached to 70% confluent. For subculturing, 0.05% trypsin/EDTA in PBS were used for the detachment of astrocytes.

siRNA transfection

Mouse or human astrocytes were plated in 96-well plate at a density of 1 × 10⁴ cells/well. After 24 h, the following siRNAs for target genes were transfected into cultured astrocytes using Lipofectamine RNAiMAX (Thermo Fisher Scientific). ON-TARGETplus Non-targeting siRNA pool (D-001810-10-05), mouse heterogeneous nuclear ribonucleoprotein U (Hnrnpu) siRNA SMARTpool (M-051574–01–0005), and human HNRNPU siRNA SMARTPool (L-013501–00–0005) were purchased from Horizon Discovery Ltd. The total list of siRNAs targeting candidate genes purchased from Horizon Discovery Ltd. was shown in Table S1.

Bromodeoxyuridine (BrdU) incorporation assay

Astrocytes were seeded in 96-well plates at 1 × 10⁴ cells/well in 100 μL of DMEM/FBS. On the next day, cells were transfected with the above-described siRNAs. After 48 h, BrdU was added to the corresponding wells and incubated for 24 h at 37 °C with 5% CO₂. BrdU incorporation was measured according to the manufacturer’s instruction (Roche, #11647229001) and detected by microplate reader (Molecular Devices, SpectraMax 5).

MTT assay

Astrocytes were seeded in 96-well plates at 1 × 10⁴ cells/well in 100 μL of DMEM/FBS. On the next day, cells were transfected with the indicated siRNAs. After 48 h, 10 μL of the MTT labeling reagent (Roche, #11647229001) was added to each well. After incubation at 37 °C with 5% CO₂ incubator for 4 h, 100 μL of the Solubilization solution was added into each well and stand overnight in the incubator. Absorbance of the samples were measured using a microplate reader (Molecular Devices, SpectraMax 5) under 590 nm.

Wound healing assay

Mouse primary astrocytes were seeded at 3.9 × 10⁴ cells in 70 μL of DMEM/FBS medium/well in 2-well silicone culture insert (ibidi GmbH) settled in a 24-well plate, and incubated for 12 h at 37 °C with 5% CO₂. Cell migration was initiated by removing the inserts and adding 1 mL/well of DMEM/FBS. DIC images were acquired immediately after removing the inserts to locate the initial positions of cell-free gaps using an inverted fluorescence microscope (IX71, Olympus). Cells were incubated for 24 h for migration, fixed by 4% paraformaldehyde (PFA), and subjected to image acquisition again. The wound healing closure were measured by using Wound_healing_size_tool plugin for ImageJ software (National Institutes of Health [NIH]).

qPCR

Total RNAs from primary astrocytes and that from mouse spinal cords were extracted using the TRIzol reagent (Thermo Fisher Scientific) and purified by RNeasy Micro Kit (Qiagen) under the manufacturer’s instruction. For reverse transcription, cDNA was synthesized using the PrimeScript™ RT Master Mix (Takara Bio). Quantitative PCR was performed using KAPA SYBR Fast Master Mix (KAPA Biosystems) with the primer pairs listed in the supplementary Table 2. PCR was conducted by CFX Connect Real-Time PCR Detection System (Bio-Rad) with conditions included one cycle at 95 °C for 30 s, followed by 39 cycles of 95 °C for 5 s and 60 °C for 45 s. A melting analysis was carried out following PCR to monitor amplification specificity. Relative mRNA expression was quantified using ∆/∆-Ct method with the Gapdh mRNA as a gene loading control.

Spinal cord injury model construction (dorsal hemisection)

Mice were anesthetized with isoflurane (Pfizer) and a laminectomy at Th10 was performed to expose the spinal cord. The Th10 dorsal hemisection was created by using a micro feather ophthalmic scalpel at a depth of 1 mm. After spinal cord transection, the bladder was manually pressed once a day to assist the mice in urination until spontaneous bladder voiding contractions reappeared. The animals were randomly numbered after surgery and blindly evaluated in the experimental group for the below indicated behavioral tests.

AAV production and intraspinal injection

Adeno-associated viruses (AAV) 2/5 vector with a minimal GFAP promoter (GfaABC1D) was used to target selectively to astrocytes. AAV- flip- excision (FLEX)- enhanced green fluorescent protein (EGFP)-mir30 (Scn9a-scrambled) was a gift from Scott Sternson (Addgene plasmid #79671; http://n2t.net/addgene:79671; RRID: Addgene_79671) [28], and modified by removing FLEX and encoding a short hairpin RNA (shRNA) targeting Hnrnpu (shHnrnpu: TGCCCGTAAGAAGCGAAATTT) [29]. To generate chimeric AAV2/5 vectors, we co-transfected AAV-293 cells (Takara) with pAAV2/5 (Addgene plasmid #104964; http://n2t.net/addgene:104964; RRID: Addgene_104964), AAV-EGFP-mir30 (Scn9a-scrambled, shScramble) or shHnrnpu vector, and the adenovirus helper plasmid pHelper with the 1: 1: 2.5 weight ratio using polyethyleneimine MAX (Polysciences). Cells were harvested 3 days after transfection and AAVs were purified using an AAVpro Purification Kit (Takara). Viral concentrations were titered by qPCR using the following primers: forward ITR primer, 5'-GGAACCCCTAGTGATGGAGTT-3’, reverse ITR primer, 5'-CGGCCTCAGTGAGCGA-3’.

For intraspinal injection, 1 µL of AAV virus were manually injected with 5 × 10¹² vg/mL titer at the Th10 vertebral level. Injections were performed at the rate of 0.3 µL/min at a depth of 1 mm. Following injections, the needle was held at the injection site for another 1 min and then slowly withdrawn from the injection site of spinal cord.

BrdU administration

BrdU (0.4 g/kg body weight, Sigma-Aldrich) was administered intraperitoneally immediately after spinal cord injury, followed by a second intraperitoneal injection 6 h later. Subsequently, BrdU (1 mg/ml, Sigma-Aldrich) was provided in the drinking water supplemented with 1% (wt/vol) sucrose. The drinking water was protected from light and change every 3 days throughout the labeling period (7 days).

FACS

EGFP- labeled astrocytes from AAV-injected spinal cords were isolated by SH800S Cell Sorter (Sony) using 70 μm microfluidic sorting chips (Sony), and data were analyzed with FlowJo software (Tree Star).

MACS

Anti-astrocyte cell surface antigen-2 (ACSA-2) microbeads (#130-097-678, Miltenyi Biotec) were used to label ACSA-2⁺ astrocytes in spinal cord lesions after SCI for 3, 7, and 14 days. Then, cells were subjected to magnetic-activated cell sorting system under the manufacturer’s instruction.

Immunohistochemistry

Mice were anesthetized with a cocktail of medetomidine hydrochloride (0.3 mg/kg, Orion Pharma), midazolam (4 mg/kg, Maruishi Pharmaceutical), and butorphanol (5 mg/kg, Meiji Animal Health), and transcardially perfused with ice-cold phosphate-buffered saline (PBS) and 4% PFA in phosphate buffer (PB). Spinal cords were carefully dissected out and post-fixed overnight with 4% PFA at 4 ºC and cryoprotected with 30% sucrose in PBS overnight at 4 ºC. Sections of the spinal cord spanning the injury sites were collected on a cryostat (Leica) set at 20-µm thickness. Sections were collected every 120 intervals. 6–8 sections were collected from each mouse for quantification analysis.

For immunohistochemistry, in the case of BrdU staining, DNA hydrolysis was performed at 45 °C for 30 min prior to the subsequent steps. Sections were washed with PBS for 10 min twice, and permeabilized with 0.1% TritonX-100 in PBS for 10 min twice. After that, sections were blocked with 3% normal donkey serum (NDS, Sigma Aldrich) for 1 h at room temperature, and incubated with the following primary antibodies: rat anti-Ki67 (1:500, 14-5698-82; eBioscience), rabbit anti-Hnrnpu (1:1000, ab20666; Abcam), goat anti-Sox9 (1:300, AF3075; R&D systems), mouse anti-Nestin (1:500, MAB353; Millipore), chicken anti-Vimentin (1:500, AB5733l; Millipore), rabbit anti-Sox9 (1:1000, ab5535; Sigma), mouse anti-NeuN (1:1000, MAB377; Millipore), rat anti-BrdU (1:500, ab6326; Abcam), rat anti-GFAP (1:1000, 13-0030; Invitrogen), rat anti-CD4 (1:1000, 550278; BD Pharmingen), rabbit anti-Iba1 (1:1000, 019-19741, Wako), rat anti-CD11b (1:1000, MCA74GA, BioRad), rabbit anti-PDGFRβ (1:500, ab32570; Abcam), rabbit anti-5-HT (1:1000, 20080; Immunostar), mouse anti-LAMA5 (1:500, MAB1924, Sigma), mouse anti-LAMB2 (1:500, sc-377379, Santa Cruz Biotechnology), mouse anti-CSPG (1:300, SAB4200696, Sigma) at 4 °C overnight. After 2 times wash, sections were treated with Alexa 488-, 568-, 647-conjugated secondary antibodies (1:500; ThermoFisher Scientific) at room temperature for 1 h. After washing with PBS twice, sections were mounted with fluoromount-G (SouthernBiotech) and imaged with FV3000 confocal laser scanning microscopy (Olympus). Quantification of astrocytic scar formation and fibrotic scar formation was based on the percentage of GFAP⁺ and PDGFRβ⁺ areas in a 1 mm square centered on the lesion site. For the quantification of Iba1⁺, CD11b⁺, and CD4⁺ cells around the lesion site, the average number of positive cells per mm² of tissue section was calculated. Axonal regeneration was assessed by the fluorescencet intensity in rostral and caudal parts around epicenter. Data are represented as 5-HT fluorescence intensity (a.u) of shScramble or shHnrnpu. NeuN⁺ neurons were quantified within a region spanning 1 to 2 mm caudal to the lesion epicenter. Quantification was performed using Fiji/ImageJ software (NIH).

Ladder walk test

Mice were habituated and trained to run on the apparatus before injury [27]. On the day before injury and on selected days after injury, mice were tested. An irregular rung pattern (with uneven spacing between rungs) is used for all the mice based on the time point relative to the injury to prevent the mice from compensating for their limb impairment by learning a particular pattern. A video camera was used to record the mice steps and the video recordings were analyzed by an observer using naked eyes. Deep slips/misses, minor slips and misplacement are considered as fault steps. The percentage of fault steps relative to the total steps were used to represent the results.

Basso mouse scale (BMS)

BMS [30] was measured before or after spinal cord injury for 1 to 7, 14, 21, 28, 35, 42, 49, and 56 days. Mice were allowed to walk freely in an open field and were observed for at least 5 min.

Footprint test

Footprint test was performed 8-week post spinal cord injury to measure the locomotor activity of mice limbs. The fore- and hind-paws of the mice were coated with red or blue nontoxic paints, respectively. Mice were allowed to walk along a (50 cm long, 4.5 cm wide) narrow white paper-covered corridor. To characterize the walking pattern of each mouse, the distance between each stride (stride length) and left and right hind footprints (stance length) were measured at least 4–5 sequential steps to determine the mean values for each parameter. The measurements were analyzed using ImageJ software (NIH).

RNA-seq analysis

Total RNA was isolated from cells using RNeasy mini kits (Qiagen) according to the manufacturer’s instructions. Poly-A mRNA was extracted from total RNA with Oligo-dT beads from a NEBNext Poly(A) RNA Magnetic Isolation Module (New England Biolabs), after which RNA-seq libraries were prepared using a NEBNext Ultra II RNA Library Prep Kit for Illumina (New England Biolabs) according to the manufacturers’ protocols. Libraries were single-end-sequenced or paired-end-sequenced on a NovaSeq sequencer (Illumina). Reads were aligned to the hg38 human genome using STAR [31]. Aligned read files were analyzed using HOMER [32]. Expression analysis of the RNA-seq data was performed using HOMER. Differential gene expression analyses were performed using DESeq2 [33] and Bioconductor (v3.9) in R (v4.2.1). The RNA-seq data presented were from replicate 3 (GSE236433), and all trends were observed in both replicates. Differentially expressed genes were defined by FDR < 0.05 (n = 3). In the heatmap analysis, the values of DEGs were normalized with a z-score on transcripts per million (TPM) values, and heatmap analysis was performed with ‘pheatmap’ in the libraries of Bioconductor. Gene ontology analysis was performed using Metascape [34] and the Database for Annotation, Visualization and Integrated Discovery v6.8 [35]. Gene set enrichment analyses were performed using GSEA [36] with rank files generated from expression data analyzed using DESeq2 [33].

Statistical analysis

Data are presented as mean ± SD. Statistical analyses were performed using Prism 9 (GraphPad Software). Two-tailed Student’s t-test was performed to compare two independent sample sets. One-way AVOVA and Tukey’s post hoc multiple comparisons were applied for group analysis. Two-way ANOVA and Bonferroni’s post hoc multiple comparisons were used to determine two variables on a continuous response. A p value < 0.05 was considered significant. Significant differences are indicated by * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

Results

Hnrnpu is a potential factor regulating astrocyte proliferation

To investigate the potential factors regulating astrocyte reactivity, we selected 1691 potential candidates which are highly expressed (TPM > 10) in spinal cord astrocytes based on previously published RNA-seq datasets [37, 38]. Among these genes, we focused on the top 18 genes, which were enriched in the gene set "cell proliferation" (GO: 0008283) as an initial phenotype of astrocyte reactivity (Fig. 1A) [39], suggesting the potential role of them in contributing to glial scar formation and neuronal regeneration. To assess the impact of these candidate genes on astrocyte proliferation, we utilized siRNA-mediated gene silencing and measured their effects on proliferation in mouse primary astrocytes. Quantifying BrdU incorporation revealed that Hnrnpu knockdown exhibited the most pronounced inhibitory effect on astrocyte proliferation (Fig. 1B, C). We confirmed Hnrnpu knockdown at both the mRNA and protein levels (Fig. S1A, B), without affecting the cell viability (Fig. S1C), indicating that Hnrnpu is a predominant gene promoting astrocyte proliferation. Since reactive astrocytes migrate to the lesion epicenter and facilitate tissue repair in response to CNS injuries [40, 41], we then conducted wound healing assay. In the result, knockdown of Hnrnpu impaired would closure (Fig. 1D), suggesting that Hnrnpu inhibition may affect astrocyte migration in addition to proliferation. We also asked whether inhibition of Hnrupu prevents the expression of well-established markers that are associated with astrocyte reactivity. qPCR analysis showed that Hnrnpu knockdown cells decreased the expression of many genes characteristic to reactive astrocytes (RA) or scar-forming astrocytes (SA) (Fig. 1E, F) [37]. These results suggest that Hnrnpu in astrocytes is involved in phenotypic and transcriptomic changes related to glial scar formation. To determine whether the protein expression of RA-related genes is affected by Hnrnpu knockdown in astrocytes, we performed immunostaining of GFAP, nestin, vimentin on primary astrocytes following Hnrnpu siRNA transfection. Our results showed a reduction in GFAP intensity in Hnrnpu-negative cells, but no changes in nestin or vimentin protein levels (Fig. S1D, E).

Fig. 1.

Hnrnpu is a potential regulator of astrocytes. A Schematic diagram of screening methods for potential genes involved in spinal cord astrocyte proliferation. B siRNA-based loss-of-function genomic screen in primary astrocytes. Astrocyte proliferation was measured by BrdU incorporation assay. C Astrocyte proliferation was assessed by BrdU incorporation assay at 24 h after transfection with control siRNA or Hnrnpu siRNA for 48 h (n = 4). D Wound healing assay using primary astrocytes transfected with control (ctrl) siRNA or Hnrnpu siRNA. Representative images after 24 h (left) and quantification of relative wound closure (right) are shown (n = 4). Dashed lines indicate the initial wound area. Blue lines indicate the migrated area. Scale bar: 200 µm. E, F mRNA levels of reactive astrocyte (RA)-related genes E or scar-forming astrocyte (SA)-related genes F in primary astrocytes transfected with control or Hnrnpu siRNA (n = 3–4). Data are mean ± SD. P values were calculated using two-tailed unpaired t-test. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, ns, not significant. n refers to the number of biological replicates

Hnrnpu contributes to glial scar formation after SCI

To ask the function of Hnrnpu in astrocytes in vivo, we investigated the expression change of Hnrnpu in the mice spinal cords after injury (Fig. 2A). Immunohistochemical analysis revealed that Hnrnpu expression in astrocytes positive for Sox9, an astrocyte-specific marker that is upregulated in reactive astrocytes [42], was increased in the spinal cords after injury (Fig. 2B, C). The upregulation of Hnrnpu mRNA was also observed in astrocyte cell surface antigen (ACSA)−2⁺ astrocytes sorted from spinal cords after injury (Fig. S2A). Consistent with the time change of Hnrnpu expression, Hnrnpu⁺Ki67⁺Sox9⁺ cells (proliferating astrocytes) were also increased in injured spinal cords (Fig. 2C).

Fig. 2.

Hnrnpu contributes to astrocyte proliferation after SCI in vivo. A Schematic diagram of experimental procedure. B Co-immunostaining of Hnrnpu (green), Sox9 (astrocyte marker, magenta), Ki67 (proliferated cell marker, cyan) in spinal cord lesions of C57BL/6J mice at sham, 3-, 7-, and 14-days following SCI. White arrowheads indicate the Hnrupu⁺Ki67⁺Sox9⁺ cells. Scale bars, left, 300 µm; right, 30 µm, 15 µm (insets). C Quantification of Hnrnpu⁺Sox9⁺Ki67⁺ cells around the lesion site (n = 4). Data are mean ± SD. P values are calculated using one-way ANOVA post Tukey’s test. * p < 0.05, *** p < 0.001, **** p < 0.0001. n refers to the number of biological replicates

Because Hnrnpu expression was not limited in astrocytes but also detected in NeuN⁺ neurons and Olig2⁺ oligodendrocytes in the spinal cord (Fig. S2A, B), we constructed astrocyte-specific adeno-associated viruses (AAV) 2/5-mir30 based shRNA knockdown vectors that express the enhanced green fluorescent protein (EGFP) under the direction of the glial fibrillary acidic protein (GFAP) promoter to manipulate the expression of Hnrnpu in astrocytes in vivo (Fig. 3A, Fig. S3A-C). We intraspinally injected AAV expressing shRNA targeting Hnrnpu (shHnrnpu) or scrambled shRNA (shScramble) into the mouse spinal cord at the Th10 level. After 2 weeks post-injection, we conducted spinal cord injury in mice and evaluated whether silencing of Hnrnpu prevents astrocyte proliferation. Immunohistochemical analysis revealed that the inhibition of Hnrnpu expression in astrocytes significantly reduced the number of BrdU⁺Sox9⁺EGFP⁺ in Sox9⁺EGFP⁺ astrocytes around the lesion site compared with the control (Fig. 3B, C), indicating that Hnrnpu expression is associated with astrocyte proliferation after SCI. Additionally, no significant difference in the expression of astrocyte reactive marker proteins was detected 7 days post-injury following Hnrnpu knockdown in vivo (Fig. S3D, E).

Fig. 3.

Hnrnpu inhibition impairs astrocyte proliferation and astrocytic scar formation in vivo. A Schematic of AAV-mir30-based shRNA knockdown vectors (left) and experimental procedure (right). B Co-immunostaining of BrdU (cyan), Sox9 (magenta), and EGFP (green) in the spinal cord lesions 1-week post-injury. Scale bars: 150 µm (left), 40 µm (right). C Quantification of BrdU⁺Sox9⁺EGFP⁺/Sox9⁺EGFP⁺ cells in the spinal cord lesions with AAV2/5-GFAP-shScramble or shHnrnpu treatment (n = 3). White arrowheads indicate the BrdU⁺Sox9⁺EGFP⁺ cells. D Immunostaining of GFAP in the spinal cord lesions of AAV-injected mice after 8 weeks post-injury. Scale bars: 300 µm (left), 100 µm (right). E, F Quantification of GFAP⁺ scar areas within a 1 mm² region centered on the lesion site (E, n = 5) and lesion areas (F, n = 6). Data are mean ± SD. P values are calculated using two-tailed unpaired t-test. * p < 0.05, ** p < 0.01. n refers to the number of biological replicates

As previous studies have indicated that the glial scar border is mainly formed by proliferating astrocytes [43], we examined whether silencing Hnrnpu expression in astrocytes also reduce glial scar formation after spinal cord injury. Immunohistochemical analysis revealed that mice infected with AAV containing shHnrnpu showed a significant decrease in GFAP fluorescence intensity at the lesion borders of spinal cord compared to the control (shScramble) (Fig. 3D-E). Glial scar forms a barrier-like structure that delineates spared tissues from the central region characterized by inflammatory and fibroblast-like cells within the lesion epicenter [43]. The result also showed a larger lesion core area in mice silenced with Hnrnpu compared with that in control (Fig. 3F). Considering the contribution of non-neural cells to scar formation, we assessed microglial/macrophage populations using the cell markers Iba1 and CD11b to evaluate their involvement in the lesion. Our results showed no significant difference in the accumulation of Iba1⁺ cells within the GFAP⁺ scars and CD11b⁺ cells within PDGFRβ⁺ scars between the shScramble and shHnrnpu groups (Fig. S4A-G). In addition, no changes were observed in the CD4⁺ T cells in PDGFRβ⁺ between the groups (Fig. S4H, I), suggesting that astrocytic Hnrnpu does not affect macrophage and infiltrating monocyte accumulation. These data suggest that astrocytic Hnrnpu-mediated glial scar formation mainly depends on astrocyte proliferation.

Hnrnpu inhibition suppresses motor recovery and neural regeneration of injured mice

Glial scar formation is known to have controversial dual effects on axonal regeneration and functional recovery after CNS injury [2, 44]. We thus examined how Hnrnpu deficiency in astrocytes affects the recovery from neurological dysfunction after spinal cord injury. Assessment of open field locomotion by the Basso mouse scale (BMS) revealed that Hnrnpu deficiency in astrocytes significantly impairs the recovery of locomotion in injured mice at later stage of SCI (Fig. 4A, B). Ladder walk test, which detects the misstep during walking on the ladder, showed that Hnrnpu-deficient mice exhibited a higher ratio of missteps compared with the control (Fig. 4C). Additionally, the footprint test also supported that the motor coordination and balance are impaired in Hnrnpu-deficient mice, as evaluated by stride length (Fig. 4D, E) and stance length (Fig. 4F). These results suggest that inhibition of astrocytic Hnrnpu expression causes impairment of neural tissue repair and functional recovery after injury, suggesting that endogenous astrocytic Hnrnpu is involved in functional recovery of mice after SCI.

Fig. 4.

Hnrnpu deficiency in astrocytes impaired motor function recovery and neural regeneration of injured mice. A Schematic diagram of the behavior tests procedure. B, C Locomotion recovery after SCI quantified by the BMS score (B) and fault step ratios in the ladder walk test (C). n = 9 for shScramble, n = 6 for shHnrnpu. D–F Footprint test. Representative images (D). Quantification of stride length (E) and stance length (F). n = 9 for shScramble, n = 6 for shHnrnpu. G Representative images of 5-HT⁺ axons around the lesion site 8 weeks post-injury. Scale bar: 300 µm. Arrowheads indicate the lesion epicenter. H Quantification of 5-HT⁺ axons intensity rostral (negative) and caudal (positive) to the lesion epicenter. Grey and light-green areas indicate the SEM error bands for shScramble and shHnrnpu groups, respectively. I Representative images of NeuN⁺ cells around the lesion site 8 weeks post-injury. Scale bar: 300 µm. J Quantification of the number of NeuN⁺ within zones Z1 (~ 1 mm caudal to the lesion epicenter) and Z2 (1 mm to ~ 2 mm caudal to the lesion epicenter) in I (n = 5). Data are mean ± SD for B, C, E, F, and J. P values are calculated using two-way ANOVA post Bonferroni’s test (B, C, and H), two-tailed unpaired t-test (E, F, and J). * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001

We next asked whether Hnrnpu-mediated motor recovery is sustained by the axon growth and regeneration of neuronal network. To test this, we immunohistochemically detected 5-hydroxytryptamine (HT)⁺ raphespinal tract around the lesion. Notably, we found that Hnrnpu suppression in astrocytes resulted in a significant decrease in the density of 5-HT⁺ fibers surrounding the lesion site at 8 weeks post-SCI (Fig. 4G, H). Additionally, the NeuN⁺ neurons was reduced in the shHnrnpu group compared to the shScramble group at the Z1 position (~ 1 mm caudal to the lesion epicenter) (Fig. 4I, J). These findings support that the impaired motor function caused by Hnrnpu inhibition is likely due to restricted axon growth and regeneration following the injury.

Transcriptomic signature of Hnrnpu suppression in human astrocytes

We finally tested whether our observations in mouse astrocytes were also conserved in human astrocytes. We evaluated the cell proliferation of astrocytes isolated from human brain cortex in vitro. Consistent with our observations in mouse astrocytes, we found that Hnrnpu inhibition suppressed BrdU incorporation into human astrocytes (Fig. 5A) and the percentage of Ki67⁺Sox9⁺ proliferating astrocytes among Sox9⁺ astrocytes (Fig. 5B, C). We confirmed HNRNPU knockdown at both the mRNA and protein levels (Fig. 5D, E; Fig. S5A) There is no significant difference in cell viability measured by MTT between the groups (Fig. S5B). These observations suggest that Hnrnpu is required for human astrocyte proliferation.

Fig. 5.

Inhibition of Hnrnpu suppresses human astrocyte proliferation and affects their transcriptomic profiles. A Hnrnpu inhibition suppresses human astrocyte proliferation as assessed by BrdU assay (n = 4). B Representative images of Ki67 (green) and Sox9 (magenta) in control or HNRNPU knockdown astrocytes. Arrowheads indicate Sox9⁺Ki67⁺ cells. Scale bar: 100 µm. C Quantification of the percentage of Ki67⁺Sox9⁺ cells among Sox9⁺ cells (n = 6). D Representative images of Hnrnpu (green) and Sox9 (magenta) in control or HNRNPU knockdown astrocytes. Scale bar: 100 µm. E Quantification of the percentage of Hnrnpu⁺Sox9⁺ cells among Sox9⁺ cells (n = 6). F Heatmap of z-score normalized RNA-seq of significant differentially expressed genes (DEGs) with control or HNRNPU siRNA-transfected human astrocytes. G MA plot displaying DEGs. Red and blue dots represent up or down regulated genes, respectively. H–K Heatmap showing representative DEGs involved in regulation of cell population proliferation (H, p = 9.1E-2), cell migration (I, p = 5.1E-3), wound healing (J, p = 4.6E-4), and axon guidance (K, p = 2.6E-2). Data are mean ± SD. P values are calculated using two-tailed unpaired t-test (A, C, and E). * p < 0.05, *** p < 0.001, **** p < 0.0001

Our in vivo experiment using SCI mice showed that inhibition of astrocytic Hnrnpu suppresses astrocyte proliferation and subsequent scar formation. We asked if HNRUPU in human astrocytes might also contribute to those processes. We performed RNA-seq analysis on human astrocytes transfected with either control or HNRNPU siRNA and identified hundreds of differentially expressed genes (DEGs) (Fig. 5F, G). Of these DEGs, 308 genes were found to be downregulated and 528 genes were found to be upregulated (FDR < 0.05) (Fig. 5G). Gene ontology (GO) enrichment analysis demonstrated that knockdown of HNRNPU influences GO biological processes (Fig. 5H-K, Fig. S5C-E), including cell population proliferation (GO:0008283, p = 9.1E-2) (Fig. 5H), which supports our observation of astrocyte proliferation regulated by Hnrnpu. Furthermore, Hnrnpu knockdown human astrocytes change the expression of gene sets annotated with cell migration (GO:0016477, p = 5.1E-3), wound healing (GO:0042060, p = 4.6E-4), and axon guidance (GO:0007411, p = 2.6E-2), suggesting that astrocytic Hnrnpu may modulate these potential molecules to promote neuronal regeneration and functional recovery after CNS injury. Additionally, GSEA analysis found the Top 15 terms enriched in the Hallmark gene sets, showing downregulated (Fig. S6A) and upregulated (Fig. S6B) terms in HNRNPU knockdown astrocytes. The results suggest that HNRNPU may regulate cell proliferation primarily through pathways involving MYC targets (V1, V2), E2F targets, mTORC1 signaling, and TNFα signaling via NF-κB (Fig. S6C-G). Together, these results suggest that Hnrnpu controls the genes involved in human astrocyte proliferation, glial scar formation, and axon regeneration following injury.

Hnrnpu inhibition changes the expression of astrocyte-derived factors

Astrocyte is known to express the molecules that regulates axon regeneration [3, 4]. To ask whether astrocytic-Hnrnpu modulate the expression of the factors that regulate axon regeneration, we compared the TPM levels of the well-established factors regulating axon regeneration between the groups based on our RNA-seq results. We found that the facilitating molecule CSPG4, LAMA5 and LAMB2, was significantly decreased in HNRNPU knockdown cells (Fig. 6A). In contrast, inhibition of Hnrnpu did not change the TPM levels of inhibitory molecules including aggrecan (ACAN), brevican (BCAN), and neurocan (NCAN) compared to the control (Fig. 6A). We then asked whether the change of mRNA expression is conserved in the protein level. Immunohistochemical analysis revealed that Hnrnpu suppression downregulates the enrichment of CSPG4 in GFAP⁺ astrocytic scar region, suggesting decreased CSPG4 production in this area (Fig. 6B). Additionally, the expression of LAMA5, and LAMB2 around the spinal cord lesions was also reduced 14 days post-injury (Fig. 6C-D). These results suggest that astrocytic Hnrnpu maintains expression of astrocyte-derived growth-permissive proteins, thereby sustaining axon regeneration and functional recovery (Fig. 6E).

Fig. 6.

Hnrnpu inhibits permissive molecules that promotes axon regeneration. A TPM values of inhibitory molecules (ACAN, BACN, NCAN) and permissive molecules (CSPG4, CSPG5, LAMA5, and LAMB2) in axon regeneration from the RNA-seq analysis of control or HNRNPU-siRNA transfected human astrocytes. (n = 3 each). B Representative images showing the CSPG4 localization (cyan), astrocytic scar (GFAP and EGFP, green), and fibrotic scar (PDGFRβ) around the lesion site of shScramble or shHnrnpu injected spinal cords, 14 days post-injury. Scale bars: 150 µm (enlarged area); 50 µm (insert). Right panel shows the quantification of CSPG4 intensity in the GFAP⁺ scar area. Data are shown as the relative values compared to shScramble group (n = 3). C, D Immunocytochemistry of LAMA5 (C) and LAMB2 (D) around the lesion site (left). Scale bars: 100 µm. Right panels show the quantification of LAMA5 (C) and LAMB2 (D) intensity. Data are shown as the relative values compared to shScramble group (n = 5). P values are calculated using two-tailed unpaired t-test (A–D). * p < 0.05, ** p < 0.01, *** p < 0.001. E Schematic diagram showing that astrocytic Hnrnpu is essential for motor function recovery and neuronal regeneration in injured mice. Hnrnpu is an intrinsic factor expressed in the nucleus of astrocytes and plays a crucial role in astrocyte reactivity. Inhibition of astrocytic Hnrnpu expression suppresses astrocyte proliferation both in vitro and in vivo. In addition, suppressing Hnrnpu expression in astrocytes inhibits astrocytic scar formation, resulting in impaired neuronal regeneration and axon regrowth, and restoration of motor function after spinal cord injury in mice

Discussion

Glial scar has been shown to have both beneficial and inhibitory effects on neuronal network regeneration. Previous studies regarding the inhibitory effects of the glial scar on neural circuit repair have led to the development of therapeutics targeting molecules, such as repulsive guidance molecule (RGM)-a RGM [45] and semaphorins [46], which prevent neuronal regeneration. However, early studies indicated that ablating scar-forming astrocytes increases the lesion size and spread of inflammatory cells, resulting in larger lesion volumes and reduced behavioral recovery after SCI [5, 44]. While this opposing effect is found to be a time-course change after injury [44], and detailed molecular mechanisms of astrocyte impact on the extracellular environment are being elucidated, the endogenous mechanism of the initial mechanism of astrocyte reactivity is not fully understood. In this study, through functional screening of molecules that regulate astrocyte proliferation, we identified Hnrnpu as a potential regulator of astrocytes in tissue repair. We showed that suppression of Hnrnpu impaired neurite extension as well as motor recovery in mice after spinal cord injury, suggesting that Hnrnpu supports neuronal regeneration.

As observed in mouse astrocytes, knocking down HNRNPU also suppressed human astrocyte proliferation and downregulated genes related to cell proliferation, migration, and wound healing, indicating that HNRNPU is important for scar formation in both human and mouse astrocytes. From the RNA-seq atlas and immunohistochemistry analysis, HNRNPU knockdown does not affect the inhibitory molecules such as ACAN, BCAN, and NCAN, whereas the facilitating molecules LAMA5, LAMB2, and CSPG4 were downregulated both at the transcriptional and protein levels. Therefore, we concluded that Hnrnpu inhibition may mainly inhibit the axon regeneration and functional recovery of injured mice by suppressing the astrocyte-derived axon growth-permissive molecules. However, Acan expression was suppressed in mouse cells, but not in human cells, suggesting species differences in HNRNPU signaling. Future studies would need to further elucidate the role of HNRNPU in humans. Although it is known that many factors are targeted by hnRNP proteins [16, 21], further studies would be clarifying the predominant molecules and the regulated molecular mechanisms involved in neural regeneration regulated by astrocytic Hnrnpu. In addition, we should note that interaction between astrocytes and other types of CNS cells is also important to control neural circuit regeneration, because the role of Hnrnpu in astrocytes is significant, modulating its expression results in only partial differences. For example, it is possible that Hnrnpu expression in astrocytes may directly affect neurons or the interactions of neurons with other corresponding cells, such as endothelial cells through the regulation of Sema3C [47]. Understanding the changes in neuronal and immune system cells associated with changes in astrocyte function will also be important for further functional analysis.

Though neural circuits regeneration is known to be affected by immune cells and fibrotic scar formation at the site of injury, histological analysis showed no difference in the accumulation of inflammatory cells and PDGFRβ⁺ cells by astrocytic Hnrnpu knockdown, suggesting that Hnrnpu in astrocytes may not directly affect PDGFRβ⁺ cell activation and inflammation, at least by changes in cell numbers. That said, Hnrnpu knockdown also affected the expression of the genes important for extracellular matrix remodeling, such as matrix metalloproteinases (MMPs) and laminins. The receptors for the ECM (e.g. integrin) are known to be expressed in T cells and microglia that are known to be located around the scar and express factors regulating axon regeneration [48, 49], noting that the remaining possibility that Hnrnpu-dependent astrocytic functional change modulates indirectly regulates neural regeneration and functional recovery.

Our data demonstrate that the upregulation of Hnrnpu after injury is important for astrocyte reactivity. Though the signaling mechanism that controls Hnrnpu expression in astrocytes is yet to be identified, Hnrnpu expression is regulated by TLR4 signaling in the macrophage cell line [50]. Because TLR4 is expressed in astrocytes [51] and injured nervous tissue is rich in TLR4 activating molecules, such as fibronectin [52] and high mobility group box 1 (HMGB1) [53], it is likely that TLR4 signaling is involved in the induction of Hnrnpu in reactivated astrocytes. Importantly, these TLR4 ligands have also been shown to promote functional recovery after spinal cord injury [54]. It will be important to elucidate the network of genes, including Hnrnpu, that respond to damage-associated molecular patterns, such as TLR4 ligands, and activate astrocytes for neural repair. We should also note that characters of reactive astrocyte markers that are changed temporally depending on the stimulus [55]. In general, there can be species differences in the regulation of molecular expression. Additional experiment that identifies the molecular mechanism of astrocyte reactivity by Hnrnpu is important to understand the role of the associated molecules clearly.

The genes affected by Hnrnpu knockdown, particularly those involved in cell proliferation, partially overlap with the Hnrnpu-dependent genes in cancer cells [56], suggesting that Hnrnpu may control cell proliferation across cell types. In this study, we detected expression of Hnrnpu in the injured spinal cord, which is not limited to astrocytes. Hnrnpu, for instance, was found in oligodendrocytes. Oligodendrocytes also contribute to the recovery of nerve function after spinal cord injury by promoting myelin repair [57]. It would be important to address whether Hnrnpu might contribute to neural repair by regulating oligodendrocytes after injury. In addition, since brain repair capacity also decreases in elderly [58] and Hnrnpu expression was decreased during aging in skeletal muscle [59], Hnrnpu may also contribute to aging-related brain diseases.

Abbreviations

CNS

Central nervous system

Hnrnpu

Heterogeneous nuclear ribonucleoprotein U

SCI

Spinal cord injury

SAF-A

Scaffold attachment factor A

hnRNP

Heterogeneous nuclear ribonucleoprotein

ALS

Amyotrophic lateral sclerosis

RA

Reactive astrocytes

SA

Scar-forming astrocytes

ACSA

Astrocyte cell surface antigen

PBS

Phosphate-buffered saline

DMEM

Dulbecco’s modified Eagle’s medium

FBS

Fetal bovine serum

BrdU

Bromodeoxyuridine

NIH

National Institutes of Health

AAV

Adeno-associated viruses

FLEX

Flip-excision

EGFP

Enhanced green fluorescent protein

shRNA

short hairpin RNA

GFAP

Glial fibrillary acidic protein

BMS

Basso mouse scale

5-HT

5-Hydrocytryptamine

DEGs

Differentially expressed genes

GO

Gene ontology

CCAR1

Cell division cycle and apoptosis regulator

STEAP3

Six-transmembrane epithelial antigen of the prostate 3

RGM

Repulsive guidance molecule

MMPs

Matrix metalloproteinases

ECM

Extracellular matrix

HMGB1

High mobility group box 1

PFA

Paraformaldehyde

TPM

Transcripts per million

CSPG

Chondroitin sulfate proteoglycan

ACAN

Aggrecan

BCAN

Brevican

NCAN

Neurocan

YAP

Yes-associated protein

Author contributions

L.Q. performed all the experiments and drafted the manuscript. A.U. supported analysis and writing. I.M. conducted RNA-seq analysis. R.M. designed the studies, wrote the manuscript, and supervised this project.

Funding

This work was supported by a Grand-in-Aid of Scientific Research from the Japan Society for the Promotion of Sciences to L.Q (23K14367) and R.M. (22H02962), AMED under Grant Number (JP22gm6210020, JP24gm1510009) to R.M, and Japan Science and Technology Agency (JST, JPMJMS2023) to I.M.

Availability of data and materials

The RNA-seq data presented were from replicate 3 (GSE236433).

Declarations

Ethics approval and consent to participate

All animal experiments were approved by the Committee on the Ethics of Animal Experiments of the National Institutes of Neuroscience, National Center of Neurology and Psychiatry, Japan (2023041).

Competing interests

The authors declare no competing interests.