Exosomes originating from neural stem cells undergoing necroptosis participate in cellular communication by inducing TSC2 upregulation of recipient cells following spinal cord injury

Abstract

We previously demonstrated that inhibiting neural stem cells necroptosis enhances functional recovery after spinal cord injury. While exosomes are recognized as playing a pivotal role in neural stem cells exocrine function, their precise function in spinal cord injury remains unclear. To investigate the role of exosomes generated following neural stem cells necroptosis after spinal cord injury, we conducted single-cell RNA sequencing and validated that neural stem cells originate from ependymal cells and undergo necroptosis in response to spinal cord injury. Subsequently, we established an in vitro necroptosis model using neural stem cells isolated from embryonic mice aged 16–17 days and extracted exosomes. The results showed that necroptosis did not significantly impact the fundamental characteristics or number of exosomes. Transcriptome sequencing of exosomes in necroptosis group identified 108 differentially expressed messenger RNAs, 104 long non-coding RNAs, 720 circular RNAs, and 14 microRNAs compared with the control group. Construction of a competing endogenous RNA network identified the following hub genes: tuberous sclerosis 2 (Tsc2), solute carrier family 16 member 3 (Slc16a3), and forkhead box protein P1 (Foxp1). Notably, a significant elevation in TSC2 expression was observed in spinal cord tissues following spinal cord injury. TSC2-positive cells were localized around SRY-box transcription factor 2–positive cells within the injury zone. Furthermore, in vitro analysis revealed increased TSC2 expression in exosomal receptor cells compared with other cells. Further assessment of cellular communication following spinal cord injury showed that Tsc2 was involved in ependymal cellular communication at 1 and 3 days post-injury through the epidermal growth factor and midkine signaling pathways. In addition, Slc16a3 participated in cellular communication in ependymal cells at 7 days post-injury via the vascular endothelial growth factor and macrophage migration inhibitory factor signaling pathways. Collectively, these findings confirm that exosomes derived from neural stem cells undergoing necroptosis play an important role in cellular communication after spinal cord injury and induce TSC2 upregulation in recipient cells.

Introduction

Spinal cord injuries (SCI), primarily resulting from falls and traffic accidents, constitute a substantial portion of the global injury burden. The incidence of SCI is projected to rise due to escalating population density, an aging demographic, and increased motor vehicle usage. The need for expensive, complex medical care and the decreased productivity of individuals with SCI significantly strain the healthcare system. In addition, the irreversibility of primary injury, the complexities of secondary injury mechanisms, and the limited regenerative potential of the spinal cord significantly contribute to the unfavorable prognosis of SCI (Fan et al., 2018; Hellenbrand et al., 2024).

Neural stem cells (NSCs), situated in the periventricular zone encircling the central canal of the spinal cord (Fauser et al., 2021), are capable of self-renewal and differentiation into a diverse array of cell types, as verified by in vitro “neurosphere” assay (Watanabe et al., 2021). Definitive NSCs (dNSCs) predominate in the adult spinal cord, originating along the neuraxis at the embryonic day (E) 16.5 and persisting in the periventricular zone into adulthood (Marqués-Torrejón et al., 2021). dNSCs are characterized by the expression of specific markers, such as the intermediate filament protein nestin (Shu et al., 2022) and vimentin (Morrow et al., 2020), as well as the transcription factor SRY-box transcription factor 2 (SOX2) and the cilia marker forkhead box J1 (FoxJ1) (Gilbert et al., 2022). There is controversy over whether ependymal cells serve as the primary source of dNSCs (Furube et al., 2020).

Necroptosis is a type of programmed cell death distinct from apoptosis, characterized by necrotic morphological features that occur independent of caspase-8 (Yan et al., 2023). Necroptosis is mediated by the necrosome, which consists of mixed lineage kinase domain-like protein (MLKL) (Zhan et al., 2021), receptor-interacting protein kinase 1 (RIPK1) (Yuan et al., 2019), and RIPK3 (Morgan and Kim, 2022; Ye et al., 2023). We previously showed that necroptosis occurs in nestin-positive cells following SCI in mice. Moreover, inhibiting necroptosis is associated with enhanced recovery from SCI (Tong et al., 2023). However, our earlier studies focused on NSC depletion post-necroptosis and did not explore the impact of necroptosis on NSC secretory function.

Exosomes, cellular vesicles that range in size from 30 to 200 nm, are integral to the secretome. Carrying diverse components such as nucleic acids, proteins, and lipids, exosomes play a crucial role in intercellular communication (Harmati et al., 2019; Walbrecq et al., 2020). Recent studies indicate that exosomes from mouse embryonic fibroblasts undergoing necroptosis carry the necroptosis-associated proteins RIPK3 and MLKL. Additionally, these necroptosis-associated exosomes are linked to calcium ion depletion and lysosomal rupture (Gupta et al., 2022). Notably, the impact of necroptosis on exosomal transcriptome has not been reported.

The aims of this study were to examine alterations in the exosome transcriptome following necroptosis in NSCs, to identify hub genes associated with necroptosis in NSCs, and to determine whether these hub genes are involved in cellular communication after SCI.

Methods

Animals

Pregnant female mice (E 16–17) and 7-week-old male C57BL/6J mice weighing 20–25 g were procured from Liaoning Changsheng Biotechnology Company (Benxi, Liaoning, China, license No. SCXK (Liao) 2020-0001). All experiments involving animals were approved by the Institutional Animal Care and Use Committee of Sun Yat-sen University (approval No. SYSU-IACUC-2021-000196) on April 19, 2021. The mice had ad libitum access to food and water and were housed in a controlled environment with a constant temperature of 24 ± 2°C, humidity maintained at 55% ± 5%, and a 12-hour light/dark cycle throughout the study.

Neural stem cell isolation and culture

Pregnant mice were initially anesthetized with 3% isoflurane (RWD, Shenzhen, Guangdong, China) via inhalation and maintained under anesthesia with 1.8% isoflurane. A cesarean section was then performed, and the uterus and placenta were meticulously dissected to extract the fetal mice. Both pregnant and fetal mice were euthanized by an overdose of anesthesia. Fetal mice aged E16 to E17 were immersed in 75% ethanol, a dorsal incision was made through the skin and cartilage, and the spinal cord was acquired. The excised spinal cord was then placed in pre-cooled Dulbecco’s modified Eagle’s medium/Nutrient Mixture F-12 (Gibco, Gardena, CA, USA), rinsed, cut into small pieces, transferred to a 100-μm cell sieve (Falcon, Corning, NY, USA), and rinsed with a generous amount of Dulbecco’s modified Eagle’s medium/Nutrient Mixture F-12. The resulting cell suspension was centrifuged at 300 × g for 5 minutes at 4°C. The pelleted cells were resuspended in Dulbecco’s modified Eagle’s medium/Nutrient Mixture F-12 supplemented with 2% B27 (Gibco), glutamine (0.5 mM, Gibco), penicillin (100 IU/mL), streptomycin (100 mg/mL, Gibco), EGF (20 ng/mL, Peprotech, Rocky Hill, NJ, USA), and bFGF (20 ng/mL, Peprotech) and incubated at 37°C with 5% CO₂ for 48 hours to generate neurospheres. Subsequently, the suspension was centrifuged at 250 × g for 5 minutes, and the supernatant was removed to isolate the neurospheres. These neurospheres were then enzymatically digested into single-cell suspensions for passaging; all subsequent experiments used third-generation NSCs (Ceto et al., 2020). NSCs were characterized by light microscopic observation as well as immunofluorescence staining, as described in more detail below.

Neural stem cell necroptosis induction

NSCs were pre-treated with 20 μM zVAD-fmk (MCE, Monmouth Junction, NJ, USA) for 30 minutes, then treated with recombinant mouse tumor necrosis factor α (TNFα, 40 ng/mL, Peprotech) and 4 μM SMAC mimetic (MCE) for 6 hours to establish a TSZ (TNFα, SMAC mimetic, zVAD-fmk)-induced in vitro model of necroptosis. All of the drugs were dissolved in dimethyl sulfoxide (DMSO), and thus the Control group was treated with an equal volume of DMSO, ensuring that concentrations of DMSO did not exceed 0.1% (Gupta et al., 2022).

Isolation of exosomes

The culture medium was harvested and centrifuged sequentially: first at 300 × g at 37°C for 10 minutes, then at 2000 × g for 10 minutes at 4°C, and finally at 10,000 × g for 60 minutes at 4°C. Following centrifugation, the supernatant was passed through a 0.22-μm sterile filter (Falcon) to eliminate cellular debris. The filtrate was ultracentrifuged at 120,000 × g for 70 minutes, the supernatant was discarded, and the precipitate was resuspended in 200 μL of phosphate buffered saline (PBS, Gibco) to obtain the exosomes (Liu et al., 2021).

Characterization of exosomes

Morphological characterization of the exosomes was performed by directly examining their structure via transmission electron microscopy (HT-7700, Hitachi, Tokyo, Japan). Nanoflow technology (N30E, NanoFCM, Xiamen, Fujian, China) was used to measure the exosome sizes and concentration. The presence of markers CD9, CD63, and CD81 was also assessed by Nanoflow technology (Welsh et al., 2024).

Exosome labeling

To fluorescently label the exosomes, 1,1-dioctadecyl-3,3,3,3-tetramethylindodicarbocyanine (4 mg/mL, AAT Bioquest, Sunnyvale, CA, USA) was added to exosomes at a ratio of 1:200 and incubated according the manufacturer’s instructions. Excess dye exosomes were eliminated by ultracentrifuging three times at 100,000 × g for 1 hour at 4°C, resuspending the pellet in PBS between each ultracentrifugation step. The final pellet was then resuspended in PBS (Wiklander et al., 2015).

Whole-transcriptome sequencing of exosomes

Exosomes isolated from NSCs from both the control and TSZ groups (n = 3 for each group) were subjected to whole-transcriptome sequencing using an Illumina platform (San Diego, CA, USA), carried out by Novogene Co. Ltd. (Beijing, China). Total exosome-derived RNA was purified using an exoRNeasy Maxi Kit (Qiagen, Strasse 1, 40724 Hilden, Germany). RNA quality and purity were assessed on 1% agarose gels. The purity of the isolated RNA was further verified using a NanoPhotometer® spectrophotometer (IMPLEN, Schillerstrasse 3, Germering, Germany, 82110). The RNA concentration was determined using a Qubit® RNA Assay Kit and a Qubit® 2.0 Fluorometer (Life Technologies, Carlsbad, CA, USA). Additionally, RNA integrity was evaluated using a RNA Nano 6000 Assay Kit and an Agilent Bioanalyzer 2100 system (Agilent Technologies, Santa Clara, CA, USA). Next, long non-coding RNA (lncRNA) and microRNA (miRNA) sequencing libraries were constructed using the NEBNext Ultra Directional RNA Library Prep Kit (NEB, Ipswich, MA, USA) for Illumina and NEBNext Multiplex Small RNA Library Prep Set for Illumina, respectively, following the protocols provided by the supplier. To construct the lncRNA sequencing library, the RNA was fragmented, first- and second-strand complementary DNA (cDNA) synthesis was performed, and the double-stranded cDNA products were purified and end-repaired, followed by adaptor ligation. The resulting lncRNA library was then amplified by PCR. For miRNA library construction, the RNA fragments were ligated to 3′ and 5′ adapters, RT-PCR was performed to generate cDNA, the sequencing of the miRNA libraries were carried out using the Hiseq2100 Truseq SBS Kit v3-HS (Illumina) (Ulintz et al., 2019; Zhang et al., 2023). The details of the quality control and analysis of the raw data are shown in Additional file 1 ^{(177.7KB, pdf)} .

Other data sources and general information

The GSE149669 and GSE162610 datasets were obtained from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/gds). The GSE149669 dataset was generated from female C57BL/6 mice subjected to acupuncture to induce moderate SCI at the T9–T10 segment, followed by precise extraction of ependymal cells via laser microscopy. The dataset includes the control group (n = 6) and the SCI group 3 day post-SCI (n = 6). The GSE162610 dataset was generated by inducing moderate SCI at the T8 segment in female C57BL/6 mice using Allen’s method. The dataset comprises the control group (n = 5) and three time points (1, 3, and 7 days post-SCI) from the SCI group (n = 5, 3, and 3, respectively).

Ependymal cell RNA sequencing

Initially, GEO2R (https://www.ncbi.nlm./geo/geo2r/) differential analysis was conducted on the GSE149669 dataset, and the gene expression data were sorted based on the order of the gene sets. Subsequently, the R package “fgsea” was used to perform gene set enrichment analysis (GSEA). GSEA enrichment scores and trends were identified by comparing the gene expression data to the programmed cell death-related gene sets (Additional Table 1) and necroptosis-associated gene set (Additional Table 2) from the Kyoto Encyclopedia of Genes and Genomes (KEGG, https://www.kegg.jp) database (Fan et al., 2022; Li et al., 2022).

Additional Table 1.

Four gene sets for programmed cell death from KEGG database

Necroptosis	Apoptosis	Autophagy	Ferroptosis
H2ab1	Apaf1	Atg5	Alox15
H2al2b	Birc3	Igbp1	Atg5
H2al1a	Birc2	Ppp2ca	Cp
H2al1b	Xiap	Ppp2cb	Cybb
H2al1c	Birc5	Atg9b	Acsl1
H2al1d	Bak1	Pik3c3	Fth1
H2al1f	Bax	Atg4d	Ftl1
H2al1g	Bcl2	Atg4c	Gclc
H2al1h	Bcl2l1	Atg9a	Gclm
H2al1i	Bid	Ulk2	Lpcat3
Gm38574	Bcl2l11	Atg2a	Gss
Chmp7	Birc6	Igbp1b	Hmox1
Sharpin	Casp3	Atg13	Slc3a2
Ticam1	Casp7	Wipi1	Slc11a2
Camk2d	Casp8	Becn1	Pcbp2
Pygb	Casp9	Gabarap	Prnp
Pygl	Cycs	Mlst8	Sat1
Parp1	Cyct	Mtor	Slc39a14
Vps4a	Fadd	Gabarapl1	Acsl6
Alox15	Bbc3	Atg4b	Trf
Slc25a4	Ngfr	Atg4a	Tfrc
Slc25a5	Septin4	Atg10	Trp53
Birc3	Tnfrsf1a	Atg12	Vdac2
Birc2	Mapk10	Atg3	Vdac3
Xiap	Mapk8	Atg101	Pcbp1
Bax	Mapk9	Atg7	Slc7a11
Bcl2	Birc7	Rptor	Ncoa4
Bid	Bok	Wipi2	Acsl5
Camk2a	Becn1	Pik3r4	Acsl4
Camk2b	Pmaip1	Atg2b	Slc40a1
Camk2g	Htra2	Atg16l1	Gpx4
Capn1	Diablo	Gabarapl2	Map1lc3a
Capn2			Map1lc3b
Casp1			Slc39a8
Casp8			Ftmt
Cflar			Steap3
Cybb			Sat2
Fadd			Inhca
Fas			Acsl3
Fasl			Atg7
Tlr3
Fth1
Ftl1
Glul
Glud1
H2ac18
H2ax
Hmgb1
Hsp90ab1
Hsp90aa1
Ifna1
Ifna11
Ifna2
Ifna4
Ifna5
Ifna6
Ifna7
Ifna9
Ifnab
Ifnar1
Ifnar2
Ifnb1
Ifng
Ifngr1
Ifngr2
Il1a
Il1b
Irf9
Jak1
Jak2
Jak3
Sqstml
Pla2g4a
Eif2ak2
Pygm
Ripkl
Vps4b
Smpd1
Chmp6
Stat1
Stat2
Stat3
Stat4
Stat5a
Stat5b
Stat6
Pla2g4b
Nlrp3
Tlr4
Tnf
Tnfaip3
Tnfrsf10b
Tnfrsf1a
Traf2
Traf5
Tnfsf10
Vdac1
Vdac2
Vdac3
Ticam2
Ifna13
Ifna16
H2aj
Pla2g4c
Chmp1a
H2al1j
Rbck1
Ifna15
Ifna12
Spata2
Mapk10
Mapk8
Mapk9
Ppia
Rnf31
Macroh2a1
Aifm1
Pla2g4f
Usp21
H2ac25
H2ac1
H2ac6
H2ac7
H2ac8
H2ac11
H2ac12
H2ac15
H2ac22
H2ac24
H2ac4
H2ac10
H2ac20
H2ac13
H2ac19
Pla2g4e
H2al1o
H2al3
H2alln
Gm5396
Ifna14
Macroh2a2
H2az1
H2al1k
H2al1e
Tyk2
H2al2c
Ripk3
Zbp1
Trpm7
H2ac21
H2ab2
H2ab3
Chmp4c
H2ac23
Chmp3
Pycard
Chmp1b2
Chmp1b
Ppid
H2al2a
Chmp2b
Chmp2a
Tradd
Pgam5
Slc25a31
Dnm1l
Cyld
Mlkl
Chmp4b
H2al1m
Chmp5
Il33
H2az2
Pla2g4d
Spata2l

Additional Table 2.

Marker genes in the necroptosis gene set

Ripkl

Ripk3

Mlkl

Tnf

Tnfrsfla

TnfsflO

TnfrsflOb

Fasl

Fas

Ifnal

Ifnarl

Tlr3

Tlr4

Zbpl

Single-cell RNA sequencing

The GSE162610 dataset was analyzed using a variety of R software packages. The “Seurat,” “ggplot2,” “cowplot,” “Matrix,” “dplyr,” “ggsci,” and “SingleR” packages were used for integration, batch correction, dimensionality reduction, clustering, annotation, and plotting of the single-cell RNA-sequencing data. Additionally, the “Seurat,” “dplyr,” “tidyverse,” “stringr,” “CellChat,” “patchwork,” “mindr,” and “svglite” packages were used to analyze and plot cellular communication via analysis of ligand-receptor pair expression among the single-cell RNA-sequencing data. To enhance the data quality, we only included cells with over 400 genes and 1000 transcripts and excluded genes that were expressed in fewer than 10 cells. Additionally, cells for which mitochondrial genes accounted for more than 5% of the RNA reads were excluded. Dimensionality reduction and clustering were achieved through uniform manifold approximation and projection and 16-point principal component analysis (Rosenberg et al., 2018; Sathyamurthy et al., 2018; Milich et al., 2021; Yang et al., 2022). Cell types were named based on the annotated results from the “SingleR” package analysis, the CellMarker database (http://bio-bigdata.hrbmu.edu.cn/CellMarker/), and relevant literature (Aran et al., 2019). Proportions of cellular subtypes were calculated, and hub genes and receptor-ligand pairs were subjected to interaction network analysis using the STRING database (http://string-db.org). Patterns of communication between the different cellular subtypes were then investigated (Szklarczyk et al., 2023). The Uniform Resource Locators of all R packages used are displayed in Additional Table 3.

Additional Table 3.

The R packages and Uniform Resource Locators (URLs) used in this study

R package	URL
fgsea	https://aur.archlinux.org/r-fgsea.git
Seurat	https://satij alab.org/seurat/
ggplot2	https://ggplot2.tidyverse .org/
Cowplot	https://github.com/wilkelab/cowplot
Matrix	https://cran.r-project.org/web/packages/Matrix/index.html
dplyr	https://dplyr.tidyverse.org/
ggsci	https://github.com/nanxstats/ggsci
SingleR	https://bioconductor.org/packages/release/bioc/html/SingleR.html
tidyverse	https://www.tidyverse.org/
stringr	https://stringr.tidyverse.org/
CellChat	https://stringr.tidyverse.org/
patchwork	https://patchwork.data-imaginist.com/
mindr	https://github.com/pzhaonet/mindr
svglite	https://cran.r-project.org/web/packages/svglite/index.html
DESeq	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
GOseq	https://bioconductor.org/packages/release/bioc/html/goseq.html

Spinal cord injury model establishment

In previous studies, we established an SCI model using male mice, which resulted in better SCI model homogeneity. Consequently, only male mice were used to establish the SCI model used in this study. Mature mice (n = 12) were randomly divided into two groups: the sham group (n = 6), which underwent modified laminectomy only, and the SCI group (n = 6), which was subjected to SCI using Allen’s method, employing a weight-drop contusion mechanism, following modified laminectomy (Wu et al., 2022). Prior to the surgical procedure, all animals were anesthetized with 3% isoflurane inhalation, followed by 1.8% isoflurane inhalation for maintenance of the anesthesia. To fully expose the spinal cord, a laminectomy at the T10 vertebrae was performed via layer-by-layer incision, involving the skin, subcutaneous tissues, and fascia, with separation of the lateral muscular tissues, as previously described. Subsequently, a controlled contusion injury was inflicted using a 10-g hammer in free fall from a height of 4 cm to induce SCI. Mice in the sham group underwent laminectomy only. Post-surgery, urination assistance was provided daily until reestablishment of the micturition reflex.

At 3 days post-SCI, mice were anesthetized with 3% isoflurane and then transcardially perfused with 4% paraformaldehyde (Aladdin, Shanghai, China) or PBS. Subsequently, spinal cord specimens were collected for further analysis.

In vitro cell culture

The neuroblastoma × spinal cord hybridoma cell line (NSC34, RRID: CVCL_D356) was provided by the Xiayu group at the Seventh Affiliated Hospital of Sun Yat-sen University. Cells were cultured and maintained in complete Dulbecco’s modified Eagle’s medium (Gibco, C11995500BT) supplemented with 10% Fetalgro bovine growth serum (Gibco, 10091148) and 1% penicillin-streptomycin (Gibco, 15140-122) (Khosla et al., 2023).

Flow cytometry

A Vybrant FAM Caspase-8 Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA) was used for cell death analysis. NSCs were rinsed with the provided rinse solution and subsequently treated with 150× FLICA (Caspase-8, Thermo Fisher Scientific) for 60 minutes at 37°C. Following another round of washing with the rinse solution, 5 μL of propidium iodide (PI, Thermo Fisher Scientific) was added, and the cells were incubated on ice for 10 minutes. Subsequently, flow cytometry data were acquired using a CytoFLEX LX instrument (Beckman Coulter, Brea, CA, USA). Flowjo software (FlowJo, Ashland, OR, USA, version 8.0.2) was used for data analysis (Fritsch et al., 2019; Gupta et al., 2022).

Immunofluorescence staining

Spinal cords harvested from mice 3 days post-SCI were fixed by immersing in 4% paraformaldehyde for 1 day. Next, each specimen was sequentially transferred to 10%, 20%, and 30% sucrose (Biofroxx, Shanghai, China) solutions for gradient dehydration, over 96 hours. The dehydrated tissues were embedded in optimal cutting temperature compound (SAKURA, Tokyo, Japan) and sectioned using a frozen sectioning machine (Leica, Wetzlar, Germany). Treated NSCs and NSC-34 cells and the frozen tissue sections were fixed in pre-cooled paraformaldehyde (4%, w/v) for 20 minutes, then permeabilized with 0.2% Triton X-100 (Aladdin) for an additional 20 minutes, followed by blocking with 10% normal goat serum. Finally, the cells were incubated overnight at 4°C with the following primary antibodies: anti-Nestin (1:500, mouse IgG, Abcam, Cambridge, UK, Cat# ab6142, RRID: AB_305313), anti-SOX2 (1:500, rabbit IgG, Cell Signaling, Boston, MA, USA, Cat# 23064, RRID: AB_2714146), anti-SOX2 (1:500, mouse IgG, Abcam, Cat# ab171380, RRID: AB_2732072), and anti-tuberous sclerosis 2 (TSC2, 1:500, rabbit IgG, Cell Signaling, Cat# 4308, RRID: AB_10547134). The next day, the cells and frozen tissue sections were incubated at 37°C for 1 hour with the following secondary antibodies: Alexa Fluor 488 (goat anti rabbit IgG, 5 µg/mL, Thermo Fisher Scientific, Cat# A-11034, RRID: AB_2576217) and Alexa Fluor 594 (goat anti mouse IgG, 5 µg/mL, Thermo Fisher Scientific, Cat# A-11005, RRID: AB_2534073). Finally, the nuclei were counterstained with 4′,6-diamidino-2-phenylindole (Abcam) for 10 minutes. The immunoreactivity was visualized using a confocal fluorescence microscope (Carl Zeiss, Oberkochen, Germany) and a pathology slide scanner (Olympus, Tokyo, Japan).

Western blotting

Total protein was extracted from lysed NSC-34 cells or homogenized spinal cord tissues harvested 3 days post-SCI using radioimmune precipitation assay (Thermo Fisher Scientific) lysis buffer containing a mixture of protease and phosphatase inhibitors. Subsequently, the proteins were separated by polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membranes using an iBlot 3 Dry Blotting System (Thermo Fisher Scientific). The membranes were probed with primary antibodies, including anti-solute carrier family 16 member 3 (SLC16A3, 1:1000, rabbit IgG, Thermo Fisher Scientific, Cat# PA587977, RRID: AB_2804558), anti-Forkhead box protein P1 (FOXP1, 1:1000, rabbit IgG, Abcam, Cat# ab134055, RRID: AB_2107103), anti-TSC2 (1:1000, rabbit IgG, Cell Signaling, Cat# 4308, RRID: AB_10547134), and anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH, 1:2000, rabbit IgG, Cell Signaling, Cat# 2118S, RRID: AB_561053) overnight at 4°C. Next, the membranes were incubated with a horseradish peroxidase–conjugated rabbit-specific antibody (1:3000, Cell Signaling, Cat# 7074, RRID: AB_2099233) for 1 hour at 37°C. The resulting bands were visualized using an enhanced chemiluminescence reagent (Thermo Fisher Scientific) and a chemiluminescent imaging system (ChemiDoc, BIO-RAD, Hercules, CA, USA). Relative optical density was evaluated using ImageJ Software (version 1.53, National Institutes of Health, Bethesda, MD, USA; Schneider et al., 2012). TSC2, SLC16A3, and FOXP1 expression levels were normalized to GAPDH.

Statistical analysis

No statistical methods were used to predetermine sample sizes. However, our sample sizes are similar to those reported in a previous publication (Tong et al., 2023). No animals or data points were excluded from the analysis. Data collection and analysis were carried out by investigators blinded to the group assignments. Statistical analysis was performed using GraphPad Prism version 8.0.2 for Windows (GraphPad Software, Boston, MA, USA, www.graphpad.com). The data reporting method varies depending on the sample size and the type of analysis conducted. Western blotting and immunofluorescence data are presented as mean ± standard deviation (SD). Conversely, RNA-sequencing and single-cell RNA-sequencing data are reported as mean ± standard error of mean (SEM). Two-tailed unpaired Student’s t-test was employed when comparing two groups. For multiple groups, one-way analysis of variance was used, followed by Bonferroni’s post hoc test. Statistical significance was defined as P < 0.05.

Results

Identification of neural stem cells mainly deriving from ependymal cells and undergoing necroptosis

To ascertain if NSCs originate from ependymal cells and undergo necroptosis, we analyzed the GSE162610 and GSE149669 datasets. We constructed an SCI single-cell atlas from the GSE162610 single-cell RNA-sequencing data that showed 29 distinct cell clusters (Figure 1A). Subsequent annotation of these clusters resulted in the identification of 15 distinct cell types (Figure 1B and C). Ependymal cells exhibited high levels of Sox2 and Foxj1 expression. Additionally, Vimentin was expressed in a wide range of cells, excluding oligodendrocyte progenitor cells (OPCs) and microglia, and was expressed at particularly high levels in ependymal cells. Nestin expression was detected in both endothelial and ependymal cells. Notably, a subset of ependymal cells that expressed high levels of vimentin also expressed Nestin (Figure 1D). Examination of necroptosis-associated gene expression showed that ependymal cells expressed high levels of the key necroptosis genes Ripk3 and Mlkl at various time points following SCI. Additionally, the tumor necrosis factor receptor superfamily member 1A (Tnfrsf1a) and tumor necrosis factor receptor superfamily member 6 (Tnfsf6/Fas) ligand genes demonstrated elevated expression after SCI, with expression peaking 1 day after SCI (Figure 1E). Furthermore, GSEA revealed enrichment of three gene sets, including necroptosis, apoptosis, and ferroptosis, in isolated ependymal cells 3 days after SCI (Figure 1F). Similarly, expression of the ligand-encoding gene Tnfrsf1a was highest in isolated ependymal cells, consistent with the single-cell RNA-sequencing results (Figure 1G). On this basis, we inferred that ependymal cells primarily undergo necroptosis after SCI through the tumor necrosis factor (TNF) signaling pathway. Therefore, TSZ treatment (TNF receptor–induced initiation) was used to induce necroptosis in our in vitro system.

Figure 1.

Identification of NSCs mainly derived from ependymal cells and undergoing necroptosis.

(A) UMAP plot of single-cell type clusters in SCI. (B) Annotated UMAP plot of single-cell type clusters in SCI. (C) Bubble plot of cell type marker gene expression in SCI. (D) Expression of NSC-related marker genes as determined by single-cell RNA-sequencing. (E) The expression of key genes and ligand receptor–related genes associated with necroptosis in ependymal cells was examined in the sham group and 1, 3, 7 days following SCI. (F) GSEA of PCD in ependymal cells (3 days post-SCI vs. control). (G) Enrichment of key genes and ligand receptor–related genes associated with necroptosis as assessed by GSEA. GSEA: Gene set enrichment analysis; NSC: neural stem cell; PCD: programmed cell death; SCI: spinal cord injury; UMAP: uniform manifold approximation and projection.

Identification of neural stem cells undergoing necroptosis in vitro and characterization of exosomes

To characterize the exosomes released subsequent to necroptosis in NSCs, we isolated NSCs, induced necroptosis, and collected the exosomes released by the NSCs. The NSCs formed neurospheres in vitro that exhibited a typical spherical shape and refractive index (Figure 2A). Both the neurospheres and individual cells express the NSC markers SOX2 and Nestin (Figure 2B). TSZ treatment induced NSC necroptosis that was characterized by a notable rise in the count of PI-positive and caspase-8–negative cells (Figure 2C). Exosomes from both the control and TSZ groups exhibited a classical “Teato-like” (semi-concave on one side) bilayer membrane structure, as observed by transmission electron microscopy. Furthermore, nanoflow analysis showed that these exosomes demonstrated particle sizes ranging from 50 to 150 nm, with a concentration of 4.48 × 10⁹ particles/mL in the control group and 5.1 × 10⁹ particles/mL in the TSZ group (Figure 2D and E). We validated the expression of the three surface markers CD9, CD63, and CD81 on exosomes via nanoflow analysis and observed that CD9 expression was the highest among the three markers in both the control and TSZ groups, followed by CD81 (Figure 2F).

Figure 2.

Validation of the necroptosis model and characterization of NSCs and exosomes.

(A) Characterization of NSC neurospheres. The NSCs displayed a typical spherical neurosphere morphology. Scale bars: 500 μm (upper) and 100 μm (lower). (B) Representative immunofluorescence images of SOX2 (green, Alexa Fluor 488) and Nestin (red, Alexa Fluor 594) expression in NSCs. Both neurospheres and individual NSCs expressed SOX2 and Nestin. Cell nuclei were counterstained with DAPI (blue). Scale bar: 100 μm. (C) NSCs were treated with DMSO or TSZ for 8 hours, after which cell death was analyzed by flow cytometry. (D) Exosome morphology as assessed by transmission electron microscopy. The exosomes in both the NSC-control and NSC-TSZ groups exhibited a typical one-sided semi-concave bilayer membrane structure. Scale bars: 200 nm. (E, F) Nanoflow cytometry detection of particle size distribution, concentration, and exosome surface marker expression. DAPI: 4′,6-Diamidino-2-phenylindole; DMSO: dimethyl sulfoxide; NSC: neural stem cell; SOX2: SRY-box transcription factor 2; TSZ: tumor necrosis factor α, SMAC mimetic, zVAD-fmk.

Identification, KEGG pathway and GO enrichment analysis of differentially expressed mRNAs, lncRNAs, circRNAs, and miRNAs

To investigate the impact of necroptosis on the NSC-derived exosome transcriptome, we conducted RNA sequencing of exosomes obtained from both the control and TSZ groups, performed Venn diagram analysis (Additional Figure 1 ^{(4.3MB, tif)} A), and analyzed the differential expression of messenger RNAs (DEmRNAs), lncRNAs (DElncRNAs), circular RNAs (DEcircRNAs), and miRNAs (DEmiRNAs). We identified 108 DEmRNAs (60 down-regulated and 48 up-regulated), 104 DElncRNAs (55 down- regulated and 49 up-regulated), 720 DEcircRNAs (253 down- regulated and 647 up-regulated), and 14 DEmiRNAs (9 down- regulated and 5 up- regulated) (Additional Figure 2 ^{(6.4MB, tif)} A). Clustering maps of the DEmRNAs, DElncRNAs, DEcircRNAs, and DEmiRNAs indicated clear differentiation between the control and TSZ samples (Additional Figure 1 ^{(4.3MB, tif)} B).

The top 20 KEGG pathways enriched in DEmRNAs, as well as target genes of the DEmRNAs, DElncRNAs, and DEcircRNAs, are presented as bubble plots in Additional Figure 2 ^{(6.4MB, tif)} B–D. Among them, the “Ubiquitin-mediated proteolysis” and “Autophagy” pathways were significantly enriched in DEmRNAs, DElncRNAs, and DEcircRNAs. The “Thermogenesis” and “Spliceosome” pathways were significantly enriched in DEmRNAs and DElncRNAs. The “Neurotrophin signaling pathway” was significantly enriched in DEmRNAs and DEcircRNAs. Additionally, the “Chronic myeloid leukemia” pathway was significantly enriched in DEcircRNAs, while only the “MAPK signaling pathway” was significantly enriched in DEmiRNAs.

Additional Figure 3 ^{(6.3MB, tif)} displays bar charts of the top 20 results of the GO analysis for the three fundamental classifications. For biological processes, “cellular process” was significantly enriched in DEmRNAs, DElncRNAs, DEcircRNAs, and DEmiRNAs. For cellular components, “cell part” and “intracellular part” were significantly enriched in all four groups, which could be because the transcriptome originated from exosomes. For molecular functions, “protein binding” was the most enriched in all four groups, suggesting that DEmRNAs, DElncRNAs, DEcircRNAs, and DEmiRNAs primarily function through protein binding.

Identification of hub genes by whole-transcriptome sequencing analysis of DEmRNAs, DElncRNAs, DEcircRNAs, and DEmiRNAs

To comprehensively analyze the roles of DEmRNAs, DElncRNAs, DEcircRNAs, and DEmiRNAs, we performed whole-transcriptome analysis. According to the competing endogenous RNA (ceRNA) theory, we identified lncRNA/circRNA–target gene pairs sharing the same miRNA binding site. We then established the lncRNA-miRNA-mRNA/circRNA-miRNA-mRNA regulatory relationships with lncRNA/circRNA as the decoy, miRNA as the core, and mRNA as the target and visualized the ceRNA network using Cytoscape software.

Following target gene prediction pairing, we constructed a regulatory network comprising 52 lncRNAs/74 circRNAs, 6 miRNAs, and 14 mRNAs (Additional Tables 4 and Additional Table 5 ^{(186.4KB, pdf)} ). On the basis of the competing ceRNA theory and the whole-transcriptome sequencing network, we generated two ceRNA networks: “down (lncRNA/circRNA)-up (miRNA)-down (mRNA)” and “up (lncRNA/circRNA)-down (miRNA)-up (mRNA).” The “down-up-down” network consisted of 21 down-regulated lncRNAs/37 down-regulated circRNAs, 3 up-regulated miRNAs, and 2 down-regulated mRNAs. The up-down-up network included 14 up-regulated lncRNAs/15 up-regulated circRNAs, 1 down-regulated miRNA, and 1 up-regulated mRNA (Figure 3A and B).

Additional Table 4.

Whole transcriptome sequencing network of LncRNA-miRNA- mRNA

RNA id	RNA name	Type	Up down
TCONS 00006110	ENSMUSG00000026469-AS3	LncRNA	Up
TCONS 00227925	ENSMUSG00000031788-AS3	LncRNA	Down
TCONS 00105206	Prr3-OT1	LncRNA	Down
TCONS 00168233	E130114P18Rik-OT7	LncRNA	Down
TCONS 00172228	Smim1-OT1	LncRNA	Up
TCONS_00168236	E130114P18Rik-OT5	LncRNA	Down
TCONS_00168237	ENSMUSG00000028565-AS1	LncRNA	Up
TCONS_00168235	E130114P18Rik-OT4	LncRNA	Down
TCONS_00060387	Cd180-OT4	LncRNA	Down
TCONS_00123925	ENSMUSG00000058835-AS4	LncRNA	Down
TCONS_00060389	Cd180-OT3	LncRNA	Down
TCONS 00063475	LINC328	LncRNA	Up
TCONS 00063473	LINC330	LncRNA	Down
ENSMUST00000225342	Hdgfl2-208	LncRNA	Up
TCONS 00015967	LINC60	LncRNA	Up
TCONS 00103582	Fam234a-OT1	LncRNA	Down
TCONS 00100465	Crb3-OT2	LncRNA	Down
TCONS 00015968	LINC63	LncRNA	Down
TCONS 00133538	Dnajc1-OT6	LncRNA	Up
TCONS 00079175	Chrac1-OT2	LncRNA	Down
TCONS 00061523	Tbce-OT2	LncRNA	Up
ENSMUSG00000045503	Sys1	mRNA	Down
TCONS 00088884	Ncbp2-OT5	LncRNA	Up
TCONS 00227132	Trmt1-OT6	LncRNA	Up
ENSMUSG00000002496	Tsc2	mRNA	Up
TCONS 00033267	Abhd15-OT2	LncRNA	Down
TCONS 00144648	ENSMUSG00000037111-AS1	LncRNA	Up
mmu-miR-27b-3p	mmu-miR-27b-3p	miRNA	Down
TCONS 00236182	ENSMUSG00000043051-AS1	LncRNA	Down
mmu-miR-378c	mmu-miR-378c	miRNA	Up
TCONS 00059908	Gm20379-OT2	lncRNA	Up
TCONS 00033979	LINC166	lncRNA	Up
mmu-miR-466i-5p	mmu-miR-466i-5p	miRNA	Down
TCONS 00157470	Gm12353-OT2	LncRNA	Down
TCONS 00157471	Gm12353-OT1	LncRNA	Down
TCONS 00128180	Cstf3-OT1	LncRNA	Down
ENSMUST00000231067	Lmf2-208	LncRNA	Up
ENSMUST00000167718	Cct3-207	LncRNA	Up
TCONS 00181761	E130116L18Rik-OT2	LncRNA	Up
TCONS 00072993	Cacna1d-OT1	LncRNA	Up
mmu-miR-320-3p	mmu-miR-320-3p	miRNA	Up
TCONS 00166885	ENSMUSG00000038827-AS2	LncRNA	Up
ENSMUST00000171017	Racgap1-212	LncRNA	Down
ENSMUSG00000027323	Rad51	mRNA	Down
TCONS 00262406	LINC1373	LncRNA	Up
TCONS 00026283	ENSMUSG00000061589-AS1	LncRNA	Down
TCONS 00263509	LINC1399	LncRNA	Down
TCONS 00102440	Dynlt2a2-OT7	LncRNA	Up
ENSMUSG00000108348	Pnma8c	mRNA	Down
TCONS 00143114	ENSMUSG00000027589-AS2	LncRNA	Down
TCONS 00057742	LINC271	LncRNA	Up
ENSMUSG00000030067	Foxp1	mRNA	Down
ENSMUSG00000038729	Pakap	mRNA	Down
ENSMUSG00000027339	Rassf2	mRNA	Up
ENSMUST00000209186	Rtn2-203	LncRNA	Down
ENSMUSG00000035367	Rmi1	mRNA	Down
TCONS 00033271	Abhd15-OT3	LncRNA	Down
TCONS 00033270	Abhd15-OT4	LncRNA	Up
ENSMUSG00000029769	Ccdc136	mRNA	Up
TCONS 00022924	Zbtb2-OT1	LncRNA	Up
mmu-miR-1306-3p	mmu-miR-1306-3p	miRNA	Up
TCONS 00110943	Ldlrad4-OT2	LncRNA	Down
TCONS 00093162	Slc51a-OT1	LncRNA	Down
TCONS_00123934	ENSMUSG00000058835-AS2	LncRNA	Down
TCONS 00244613	Kri1-OT4	LncRNA	Down
ENSMUST00000159935	Ptger1-202	LncRNA	Up
TCONS 00085942	ENSMUSG00000022623-AS1	LncRNA	Down
TCONS 00095752	Dynlt1c-OT4	LncRNA	Down
TCONS 00115178	Zfp236-OT4	LncRNA	Down
TCONS 00115176	Zfp236-OT2	LncRNA	Up
ENSMUSG00000025161	Slc16a3	mRNA	Down
TCONS 00092103	ENSMUSG00000026181-AS2	LncRNA	Up
TCONS 00092104	ENSMUSG00000026181-AS1	LncRNA	Up
TCONS 00123924	ENSMUSG00000058835-AS3	LncRNA	Down
ENSMUSG00000061175	Fnip2	mRNA	Up
TCONS 00127784	1700029115Rik-OT1	LncRNA	Up
TCONS 00085301	Ndufa6-OT3	LncRNA	Up
ENSMUSG00000039000	Ube3c	mRNA	Up
ENSMUSG00000025402	Nab2	mRNA	Down
TCONS 00033272	Abhd15-OT6	LncRNA	Down
TCONS 00095749	Dynlt1c-OT3	LncRNA	Up
TCONS 00010465	ENSMUSG00000026782-AS2	LncRNA	Down
TCONS 00010464	ENSMUSG00000026782-AS1	LncRNA	Up
TCONS 00077020	Gpr183-OT3	LncRNA	Down
TCONS 00147606	St7l-OT2	LncRNA	Down
ENSMUSG00000030512	Snrpa1	mRNA	Up
ENSMUST00000130120	Jmjd6-203	LncRNA	Up
TCONS 00151642	Gm21954-OT3	LncRNA	Up
mmu-miR-378a-3p	mmu-miR-378a-3p	miRNA	Up
TCONS 00087226	Glis2-OT3	LncRNA	Up
TCONS 00181129	LINC943	LncRNA	Up
TCONS 00032852	Ankfy1-OT1	LncRNA	Up
TCONS 00022183	ENSMUSG00000025795-AS2	LncRNA	Up

Figure 3.

WTS network construction and the identification and validation of hub genes.

(A, B) Construction of the WTS regulatory network, which was subsequently divided into two ceRNA networks (down-up-down/up-down-up). In these networks, circles represent miRNAs, squares represent mRNAs, and triangles represent lncRNAs and circRNAs. Red indicates up-regulation, while green indicates down-regulation. (C) Bidirectional hierarchical clustering maps of hub genes between the control and TSZ groups. (D) Expression of hub genes at different time points following SCI, as assessed by single-cell RNA-sequencing. ceRNA: Competing endogenous RNA; circRNA: circular RNA; lncRNA: long non-coding RNA; miRNA: microRNA; SCI: spinal cord injury; WTS: whole-transcriptome sequencing.

To further clarify the role of the ceRNA, hub genes were identified based on the target genes of ceRNA network, including the up-regulated tuberous sclerosis 2 (Tsc2) gene and the down-regulated solute carrier family 16 member 3 (Slc16a3) and Forkhead box protein P1 (Foxp1) genes. The bidirectional clustering heatmap shown in Figure 3C depicts the differential mRNA expression levels of the hub genes between the control and TSZ groups.

Validation of hub gene expression levels after spinal cord injury

To confirm whether the expression of hub genes post-SCI corresponds to their expression in exosomes, we generated bubble plots of hub gene expression after SCI from the single-cell RNA-sequencing data. Tsc2 expression increased significantly after SCI but decreased over time. In contrast, Slc16a3 and Foxp1 expression levels were highest in the control group and decreased with time after SCI (Figure 3D). The expression trends of the hub genes that were differentially expressed after SCI were generally consistent with the expression patterns seen in the ceRNA networks.

Western blot analysis showed a substantial increase in TSC2 expression in spinal cord tissues from the SCI group compared with the sham group (P = 0.0298). Conversely, FOXP1 and SLC16A3 expression levels did not change significantly (Figure 4A and B). Immunofluorescence analysis of spinal cord tissues showed a marked increase in the number of SOX2-positive cells in the SCI group compared with the sham group, and these cells were concentrated in close proximity to the injury site. We observed a concurrent increase in the number of TSC2-positive cells within the injury zone. These findings suggest that exosomes secreted by endogenous NSCs upregulated TSC2 expression in exosome receptor cells (Figure 4C).

Figure 4.

Western blot and immunofluorescence analysis of the expression of proteins encoded by hub genes in the spinal cord 3 days following injury.

(A, B) Western blot and semi-quantitative analyses of the expression of proteins encoded by hub genes in the spinal cord 3 days after SCI. Data are expressed as mean ± SD (n = 3), and were analyzed by two-tailed unpaired Student’s t-test. (C) Immunofluorescence staining for the NSC marker SOX2 (green, Alexa Fluor 488) and the hub gene product TSC2 (red, Alexa Fluor 594). Compared with the sham group, TSC2 and SOX2 expression levels were up-regulated in the SCI group, and the fluorescence signals from both proteins overlapped. Cell nuclei were counterstained with DAPI. Scale bars: 250 μm (left) and 50 μm (middle and right). DAPI: 4′,6-Diamidino-2-phenylindole; dpi: day post-injury; ns: not significant; NSC: neural stem cell; SCI: spinal cord injury; SOX2: SRY-box transcription factor 2; TSC2: tuberous sclerosis 2.

Validation of changes in TSC2 expression induced by exosomes derived from TSZ-induced neural stem cells in vitro

Next, we performed an in vitro uptake assay in which NSC-34 cells were exposed to exosomes secreted by NSCs undergoing necroptosis (1 × 10⁹ particles/mL, equivalent to adding 196 μL of exosomes to 1 mL of exosome-free medium). We then measured TSC2 expression following direct cellular uptake of exosomes derived from TSZ-treated NSCs. Western blot analysis showed a significant increase in TSC2 expression during the first 9 hours post–exosomes uptake in vitro. Conversely, no notable change in TSC2 expression was observed from 9 to 24 hours after exosome uptake (Figure 5A and B). Furthermore, immunofluorescence analysis showed that the TSC2 fluorescence intensity was more pronounced 12 hours after exosome uptake in cells that had taken up a larger quantity of exosomes (red, 1,1-dioctadecyl-3,3,3,3-tetramethylindodicarbocyanine–labeled) (Figure 5C).

Figure 5.

Western blot and immunofluorescence analyses of TSC2 expression after uptake of exosomes derived from TSZ-treated NSCs in vitro.

(A, B) Western blot and semi-quantitative analyses of TSC2 expression at different time points after uptake of exosomes derived from TSZ-treated NSCs in vitro. Data are expressed as mean ± SD (n = 3), and were analyzed by one-way analysis of variance followed by Bonferroni’s post hoc test. (C) Immunofluorescence staining for TSC2 (green, Alexa Fluor 488) and DID-labeled exosomes derived from TSZ-treated NSCs (red) 12 hours after uptake in vitro. When the NSC-34 cells took up the exosomes, there was a significant increase in TSC2 expression. Cell nuclei were counterstained with DAPI. Scale bars: 50 μm (left) and 20 μm (middle and right). DAPI: 4′,6-Diamidino-2-phenylindole; DID: 1,1-dioctadecyl-3,3,3,3-tetramethylindodicarbocyanine; TSC2: tuberous sclerosis 2; TSZ: tumor necrosis factor α, SMAC mimetic, zVAD-fmk.

Ependymal cell communication involving hub genes after spinal cord injury

Next, we conducted ligand-receptor pair analysis to analyze cellular communication among all cells in the control group and at three time points after SCI (Additional Tables 6 ^{(1.6MB, pdf)} –9 ^{(1.7MB, pdf)} ). Cellular communication was observed between ependymal cells and all cell types (including other ependymal cells) at all time points (Additional Figure 4 ^{(5.4MB, tif)} ). Ependymal cell communication with astrocytes and OPCs was stronger in the control group than in the SCI groups. At 1 day post-injury (dpi), almost all cellular communication decreased to similar levels. At 3 dpi, cellular communication was stronger with astrocytes and microglial cells. At 7 dpi, cellular communication in the SCI groups was similar to that seen in the control group, and again stronger with astrocytes and OPCs. Additionally, we elucidated the receptor-ligand pairs involved in intercellular communication between ependymal cells and all cell types in the control group and at three time points after SCI (Figure 6A).

Figure 6.

Ependymal cell communication in SCI and cellular communication involving hub genes.

(A) Ependymal cell communication with other cell types after SCI. (B) Cellular communication through the TNF signaling pathway after SCI. (C) Interaction network of hub genes and receptor–ligand genes after SCI, constructed using data from the STRING database. (D, E) Cellular communication involving the interaction of hub genes with receptors and ligand genes. The line thickness signifies the strength of the intercellular communication; the line color corresponds to the color of the cell from which the ligand originates. dpi: Day(s) post-injury; SCI: spinal cord injury; TNF: tumor necrosis factor.

On the basis of our previous findings indicating that ependymal cells undergo programmed necroptosis after SCI, mainly through the TNF signaling pathway, we further analyzed cellular communication through this pathway. Ependymal cells in the control and 7 dpi groups did not receive any cellular communication related to the TNF signaling pathway. In the 1 and 3 dpi groups, ependymal cells primarily received cellular communication related to the TNF signaling pathway from microglia, along with additional TNF signaling from neutrophils at 1 dpi. From this, we infer that Tsc2 up-regulation is mainly involved in ependymal cell communication at 1 and 3 dpi, while Slc16a3 and Foxp1 down-regulation was mainly involved in the control and 7-dpi groups (Figure 6B).

To determine whether the hub genes are involved in cellular communication after SCI, we analyzed the interactions among hub genes, ligands, and receptors. Slc16a3 and Foxp1 exhibited textmining (extracting a list of protein interactomes from the literature) and co-expression relationships with the vascular endothelial growth factor A (Vegfa) and interferon gamma (Ifnγ) ligand-encoding genes, respectively, whereas Tsc2 exhibited a textmining relationship with the Tnf, midkine (Mdk), transforming growth factor beta-1 (Tgfβ1), and Tgfβ2 ligand-encoding genes. Regarding receptors, textmining and experimentally determined relationships were found with Slc16a3 and Cd44, whereas Foxp1 did not show a relationship with any receptor. In contrast, Tsc2 showed textmining and experimentally determined relationships with transforming growth factor beta receptor type-2 (Tgfβr2), epidermal growth factor receptor (Egfr) and insulin-like growth factor 1 receptor (Igf1r), textmining and experimentally determined co-expression relationships with Tnfrsf1a, and a textmining relationship with Tlr4 (Figure 6C).

We deduced the involvement of hub genes in ependymal cell communication based on the relationship between hub genes and related ligand-receptor pairs. Specifically, Tsc2 participated in ependymal cell communication with OPCs and fibroblasts cells through the EGF signaling pathway at 1 dpi. At 3 dpi, Tsc2 involvement was limited to cellular between ependymal cells with fibroblast cells. Moreover, Tsc2 played a role in ependymal cell communication with all other cells type through the MDK signaling pathway at both 1 and 3 dpi, with greater communication observed with OPCs, fibroblasts, and astrocytes at 3 dpi compared with 1 dpi (Figure 6D). Furthermore, in the control group Slc16a3 played a role in cellular communication between ventricular ependymal cells and endothelial and pericyte cells through the VEGF signaling pathway. Concurrently, Slc16a3 participated in ventricular ependymal cell communication with dendritic cells and monocytes in both the control group and the 7-dpi group via the Cd74–Cd44 receptor-ligand pair, which is part of the macrophage migration inhibitory factor (MIF) signaling pathway, and was involved with ventricular ependymal cell communication with division-like-myeloid cells at 7 dpi (Figure 6E). Additionally, Foxp1 participated in ventricular ependymal cell communication.

Discussion

Following SCI, NSCs originating from the spinal periventricular zone proliferate, migrate to the injury site, and differentiate into mature neural cells (Vieira et al., 2018). Despite the prevailing consensus that NSCs derive from ependymal cells, which is supported by evidence from lineage tracing experiments (Ripoll et al., 2023), there are still dissenting perspectives (Ren et al., 2017). In this study, we examined the expression of NSC markers through single-cell RNA-sequencing data analysis and observed that Sox2 and Foxj1 were predominantly expressed in ependymal cells. While nestin and vimentin were expressed in other cells, ependymal cells expressed high levels of both proteins. From this, we inferred that NSCs originate from this specific subset of ependymal cells.

In our previous study, in which we used Nestin as a marker for NSCs, we demonstrated that nestin-positive cells undergo necroptosis after SCI and that inhibiting programmed necroptosis promotes recovery from SCI (Tong et al., 2023). In this study, we analyzed genes related to necroptosis using single-cell RNA-sequencing data and RNA-sequencing of ependymal cells extracted after SCI. We found that necroptosis occurred in ependymal cells after SCI and that Tnfrsf1a expression increased significantly after SCI. From this, we inferred that NSCs undergo necroptosis after SCI through the classical TNF signaling pathway.

The classical necroptosis pathway is a form of programmed cell death characterized by caspase-8 inhibition (Rodriguez et al., 2022), wherein the death process involves the release of cellular contents, including damage-associated molecular patterns. Previous studies have regarded necroptosis as being detrimental to SCI recovery, mainly due to the reduction in the number of NSCs and the increase in the inflammatory response (Liu et al., 2018; Hu et al., 2022). However, recent research suggests that necroptosis can alter the mode of exosome secretion and the exosome proteome (Gupta et al., 2022). Additionally, exosomes play a crucial role in cellular communication (Walbrecq et al., 2020; Huber and Wang, 2024; Xu et al., 2024; Yu et al., 2024). These findings indicate that necroptosis in SCI not only may damage cells but also may be involved in exosome-mediated ependymal cell communication. dNSCs are present at E16.5 and can persist into adulthood (Redmond et al., 2019). Because extracting NSCs after SCI yields low numbers of both cells and exosomes, we opted to use the spinal cord of E16 to E17 fetal mice as the source of NSCs. Exosome morphology, particle size distribution, concentration, and surface markers remain unchanged after TSZ induction in vitro.

In this study, we performed comprehensive transcriptome sequencing of purified exosomes. Initially, we conducted distinct analyses for mRNAs and non-coding RNAs, resulting in the identification of 108 DEmRNAs, 104 DElncRNAs, 720 DEcircRNAs, and 14 DEmiRNAs. Notably, the “Ubiquitin-mediated proteolysis” and “Autophagy” signaling pathways were significantly enriched in DEmRNAs, DElncRNAs, and DEcircRNAs. The heightened presence of “Autophagy” may be attributable to the transition from exocytosis to a lysosomal-mediated mechanism induced by necroptosis (Gupta et al., 2022). Furthermore, a recent report highlighted the involvement of ubiquitination in necroptosis (Chen et al., 2022), suggesting that the enriched “Ubiquitin-mediated proteolysis” pathway presents a novel avenue for understanding epigenetic modifications in exosomes. However, none of the mRNA classes studied were enriched in “BF” or “CC” in the GO enrichment analysis, whereas only “binding protein” was enriched in the “MF” analysis. This implies that the exosomal transcriptome might exert its influence through protein-mediated proteolysis and protein binding.

Subsequently, we constructed a ceRNA network involving lncRNA/circRNA-miRNA-mRNA interactions, adhering to the ceRNA principle. Through analysis of target genes within this network, we identified hub genes associated with programmed necroptosis in NSCs, specifically noting up-regulation of Tsc2 and down-regulation of Slc16a3 and Foxp1. Furthermore, we validated the expression trends of hub genes, showing that their up- and down-regulation after SCI aligned with the observed changes in hub gene expression following necroptosis in NSCs.

In our in vivo experiments, we observed a substantial increase in TSC2 in spinal cord tissues after SCI. Conversely, there were no significant alterations in SLC16A3 or FOXP1 after SCI. This result was in contrast to our single-cell RNA-sequencing findings. We hypothesize that this discrepancy may arise from our choice to assess protein expression levels 3 days post-SCI. It is also possible that extracting total protein from spinal cord tissues includes parts beyond the targeted region. Building on the in vivo validation results, we proceeded to confirm in vitro that exosomes influence TSC2 expression because TSC2 expression was increased in cells that had taken up exosomes in vitro. Notably, within the first 9 hours after exosome uptake, TSC2 expression increased in a time-dependent manner. Subsequently, from 9 to 24 hours after uptake, TSC2 expression stabilized, suggesting that the mRNAs carried by the exosomes had undergone transcription and translation. TSC2 expression in recipient cells exposed to exosomes was significantly elevated. Thus, our in vivo and in vitro results show that NSCs undergoing necroptosis can elevate TSC2 expression in recipient cells by releasing exosomes. This underscores the involvement of necroptotic NSCs in communicating with other cells via exosomes. The increased expression of TSC2 in recipient cells represents the most immediate outcome of this cellular communication, while subsequent downstream changes remain to be explored.

TSC2 is a pivotal regulator of various biological processes that binds to TSC1 to form the TSC1–TSC2 complex, which is a central integrator of external stress and serves as a core regulator in the mTORC1 and mTORC2 signaling pathways (Karalis et al., 2024). The TSC1–TSC2 complex plays a crucial role in orchestrating an appropriate stress response by modulating the level of mTOR signaling. By regulating Rheb activity, the TSC1–TSC2 complex inhibits mTORC1 signaling, while also activating mTORC2 signaling through direct interaction. Additionally, it plays a vital role in the crosstalk between mTORC1 and mTORC2 signaling through PI3K/Akt feedback regulation. In this context, Akt directly phosphorylates TSC2, thereby inhibiting TSC2’s inhibitory effect on Rheb and mTORC1. This limitation on the inhibitory function of TSC2 allows for the fine-tuning of mTORC1 signaling and underscores the intricate regulatory network involving the TSC1–TSC2 complex in cellular responses (Lin et al., 2023). The inhibitory effect of TSC2 on mTORC1 can result in a reduction in cellular metabolism and biosynthesis, along with the stimulation of autophagy, ultimately culminating in diminished cell growth (Liu and Sabatini, 2020). Previous studies propose that lysosomal function could represent a crucial upstream cellular event in the regulation of various pro-death pathways during secondary SCI. These pathways include inhibition of autophagic flux, activation of endoplasmic reticulum stress-dependent apoptosis, and sensitization to necrosis. In other central nervous system trauma models, such as models of traumatic brain injury, inhibiting the autophagy/lysosomal pathway and activating necrosis are associated with cell loss and tissue damage. Integrating these findings with our results, it is plausible that exosomes secreted after NSC necroptosis may mitigate further damage to spinal cord tissues. This potential protective effect could be achieved by upregulating TSC2 expression, inhibiting mTORC1 expression, and consequently promoting autophagic flux expression. In our previous studies, we investigated the role of Rictor in the mTORC2 complex following SCI and found that Rictor overexpression contributes to SCI recovery by exerting anti-inflammatory effects, as well as supporting neuronal and oligodendrocyte survival (Chen et al., 2020). Integrating these results with the outcomes of the present study, it can be inferred that exosomes released following NSC necroptosis post-SCI may protect spinal cord tissues. This protection is likely achieved by increasing TSC2/mTORC2 signaling in recipient cells, ultimately helping preserve spinal cord tissues (Chen et al., 2024). Questions regarding the downstream effects of TSC2, including modifications in the mTORC1/2 signaling pathway, will be a primary focus of our subsequent studies. In the current study, we delved deeper into the role of hub genes in SCI using single-cell RNA-sequencing to discern their roles in NSC communication with other neural cell types.

Analysis of ependymal cell communication revealed consistent interactions with all cell types, irrespective of the presence of SCI or the time point after SCI. Interestingly, the overall strength of cellular communication diminished at 1 dpi. Furthermore, we observed that communication between ependymal cells and astrocytes was significantly stronger than that between ependymal cells and other cell types. This finding suggests that NSCs, beyond their role as astrocyte precursors (Yeon et al., 2021), may influence glial cells through complex cellular communication pathways. Additionally, we analyzed the role of TNF signaling in cellular communication following SCI and found that ependymal cells received TNF signals from microglia at 1 and 3 dpi, providing further support for our earlier observations that NSC necroptosis post-SCI is mediated by the classical TNF signaling pathway and shedding light on the origin of TNF signals. Notably, TNF-α, a well-known pro-inflammatory factor, has recently been recognized as being crucial for NSC-mediated spinal cord regeneration (Tsarouchas et al., 2018). This suggests that the NSC response to TNF-α, including necroptosis, could have a positive impact on NSC activation.

We delved further into the relationship between hub genes and the interactions among all receptor ligands, focusing on the cellular communication associated with the up-regulated gene Tsc2 in the 1-dpi and 3-dpi groups, as well as the down-regulated genes Slc16a3 and Foxp1 in the control and 7-dpi groups, in light of reception of TNF signaling from ependymal cells. Our analysis showed that Tsc2 was intricately involved in ependymal cell communication through the EGF and MDK signaling pathways. Similarly, Slc16a3 played a role in ependymal cell communication through the VEGF and MIF signaling pathways. Although Foxp1 was not directly involved in cellular communication, the decreased expression of Foxp1, a motor neuron marker (Chiba et al., 2021), in exosomes after NSC necroptosis suggested potential disruption of NSC differentiation into neurons.

This study had several limitations. First, we were unable to directly extract NSCs after SCI for exosomal transcriptome sequencing. Exosomal sequencing demands a large number of cells, and NSC extraction relies on subsequent expansion and sorting in culture medium. It is challenging to extract NSCs from adult mice after SCI, and they exhibit poor survival in vitro. To overcome this limitation, spinal tissue from fetal mice aged E 16 to E17 were used as a source of NSCs to establish the in vitro necroptosis model. Furthermore, we did not perform a detailed analysis of the ubiquitin-mediated proteolysis pathway, which was enriched in differentially expressed RNAs and could be a promising avenue for future research. Finally, we employed CellChat for the analysis of cellular communication, with a specific emphasis on investigating ligand-receptor pairs within various cell types. A more recent analytical tool, NicheNet, has a greater focus on functional alterations of receptor cells and not only confirms the presence of cellular communication but also predicts the subsequent changes induced by such communication (Browaeys et al., 2020). Using NicheNet in a future study could help predict downstream effects resulting from cellular communication.

In summary, we conducted transcriptome sequencing of exosomes derived from NSCs following necroptosis and used whole-transcriptome sequencing analysis to identify hub genes (Tsc2, Slc16a3, and Foxp1) within the ceRNA network. We validated the expression of the hub genes in spinal tissues from a mouse model of SCI and observed a significant elevation in TSC2 expression post-SCI. Additionally, our findings suggest that NSCs may enhance TSC2 expression in cells at the injury site by secreting exosomes to that specific area. Furthermore, our in vitro results showed a time-dependent increase in TSC2 expression within 9 hours of cells taking up exosomes produced by NSCs treated with TSZ. This supports the notion that exosomes produced by NSCs treated with TSZ can effectively upregulate TSC2 expression in recipient cells. We also found that Tsc2 is involved in ependymal cell communication at 1 and 3 dpi through the EGF and MDK signaling pathways. Furthermore, Slc16a3 was found to be involved in cellular communication in the control group and 7 dpi through the VEGF and MIF signaling pathways. These findings provide novel insight into the role of necroptosis in SCI and shed light on the impact of exosomes in the context of SCI.

Additional files:

Additional Figure 1 ^{(4.3MB, tif)} : Differential analysis on mRNAs, LncRNAs, CircRNAs, and miRNAs.

Additional Figure 1.

Differential analysis on mRNAs, LncRNAs, CircRNAs, and miRNAs.

(A) Venn diagram analysis of mRNAs, LncRNAs, CircRNAs and miRNAs. (B) Clustering maps of differentially expressed of mRNAs, LncRNAs, CircRNAs and miRNAs. CircRNA: Circular ribonucleic acid; LncRNA: long non-coding RNA; miRNA: microRNA.

Additional Figure 2 ^{(6.4MB, tif)} : Differential analysis and KEGG enrichment analysis on mRNAs, LncRNAs, CircRNAs, and miRNAs.

Additional Figure 2.

Differential analysis and KEGG enrichment analysis on mRNAs, LncRNAs, CircRNAs, and miRNAs.

(A) Volcano plots show differentially expressed mRNAs, LncRNAs, CircRNAs, and miRNAs in the control and TSZ groups. The criterion of |log2 (fold change)| = 1 and P-value < 0.05 was used in three groups, except miRNAs (|log2 (fold change)| = 0 and P-value < 0.05). (B-E) KEGG analysis of differentially expressed mRNAs, as well as LncRNAs, CircRNAs, and miRNAs (target), in the TSZ group compared to the control group was conducted using KOBAS software. Rich factor represents the ratio of the number of genes located in the entry of the pathway to the total number of annotated genes. CircRNA: Circular ribonucleic acid; KEGG: Kyoto Encyclopedia of Genes and Genomes; LncRNA: long non-coding RNA; miRNA: microRNA; TSZ: tumor necrosis factor α, SMAC mimetic, zVAD-fmk.

Additional Figure 3 ^{(6.3MB, tif)} : GO enrichment analysis of DEmRNAs, DELncRNAs, DECircRNAs, and DEmiRNAs.

Additional Figure 3.

GO enrichment analysis of DEmRNAs, DELncRNAs, DECircRNAs, and DEmiRNAs.

(A-D) Enrichment analysis for GO was conducted on differentially expressed mRNAs, as well as LncRNAs, CircRNAs, and miRNAs (target), using the “GOseq” R package. The three different classifications correspond to the fundamental categories of GO terms: biological process (BP, red), cellular component (CC, blue), and molecular function (MF, green). CircRNA: circular ribonucleic acid; GO: Gene Ontology; LncRNA: long non-coding RNA; miRNA: microRNA.

Additional Figure 4 ^{(5.4MB, tif)} : Receptor-ligand pairs for intercellular communication between ependymal cells and other cells at different time points after SCI.

Additional Figure 4.

Receptor-ligand pairs for intercellular communication between ependymal cells and other cells at different time points after SCI.

The size of the circles represents the magnitude of the P-values, and the color depth represents the strength of communication. SCI: Spinal cord injury.

Additional Table 1: PCD-related gene sets in GSEA.

Additional Table 2: Genes tagged in necroptosis gene set.

Additional Table 3: The R packages and Uniform Resource Locators (URLs) used in this study.

Additional Table 4: Whole transcriptome sequencing network of LncRNA-miRNA-mRNA.

Additional Table 5 ^{(186.4KB, pdf)} : Whole transcriptome sequencing network of CircRNA-miRNA-mRNA.

Additional Table 5

Whole transcriptome sequencing network ofCircRNA-miRNA-mRNA

Additional Table 6 ^{(1.6MB, pdf)} : The communication signal pathways and related ligand receptors in the control group.

Additional Table 6

The communication signal pathways and related ligand receptors in the control group

Additional Table 7 ^{(2.4MB, pdf)} : The communication signal pathways and related ligand receptors in the 1-day post-spinal cord injury group.

Additional Table 7

The communication signal pathways and related ligand receptors in the 1-day post-spinal cord injury group

Additional Table 8: ^{(1.8MB, pdf)} The communication signal pathways and related ligand receptors in the 3-day post-spinal cord injury group.

Additional Table 8

The communication signal pathways and related ligand receptors in the 3-day post-spinal cord injury group

Additional Table 9: ^{(1.7MB, pdf)} The communication signal pathways and related ligand receptors in the 7-day post-spinal cord injury group.

Additional Table 9

The communication signal pathways and related ligand receptors in the 7-day post-spinal cord injury group

Additional file 1 ^{(177.7KB, pdf)} : Methods for whole transcriptome analysis of exosomes.

Additional file 1

Methods for whole transcriptome analysis of exosomes

Additional file 2: Open peer review report 1 ^{(82.1KB, pdf)} .

Funding Statement

Funding: This study was supported by the National Natural Science Foundation of China, No. 81801907 (to NC); Shenzhen Key Laboratory of Bone Tissue Repair and Translational Research, No. ZDSYS20230626091402006 (to NC); Sanming Project of Medicine in Shenzhen, No. SZSM201911002 (to SL); Foundation of Shenzhen Committee for Science and Technology Innovation, Nos. JCYJ20230807110310021 (to NC), JCYJ20230807110259002 (to JL); and Science and Technology Program of Guangzhou, No. 2024A04J4716 (to TL).

Data availability statement:

part of datasets analyzed in this study are available in the GEO database, including the datasets GSE149669 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE149669), and GSE162610 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE162610). The data on the exosome transcriptome came from the researcher’s sequencing data. The data generated or analysed during this study are available from the corresponding author on reasonable request.