Coding and long non-coding gene expression changes in the CNS traumatic injuries

Abstract

Traumatic brain injury (TBI) and spinal cord injury (SCI) are two main central nervous system (CNS) traumas, caused by external physical insults. Both injuries have devastating effects on the quality of life, and there is no effective therapy at present. Notably, gene expression profiling using bulk RNA sequencing (RNA-Seq) and single-cell RNA-Seq (scRNA-Seq) have revealed significant changes in many coding and non-coding genes, as well as important pathways in SCI and TBI. Particularly, recent studies have revealed that long non-coding RNAs (lncRNAs) with lengths greater than 200 nucleotides and without protein-coding potential have tissue- and cell type-specific expression pattern and play critical roles in CNS injury by gain- and loss-of-function approaches. LncRNAs have been shown to regulate protein-coding genes or microRNAs (miRNAs) directly or indirectly, participating in processes including inflammation, glial activation, cell apoptosis, and vasculature events. Therefore, lncRNAs could serve as potential targets for the diagnosis, treatment, and prognosis of SCI and TBI. In this review, we highlight the recent progress in transcriptome studies of SCI and TBI and insights into molecular mechanisms.

Introduction

Traumatic injury to the central nervous system (CNS) is one of the leading causes of disability and death in modern society, particularly among young adults [1]. In this review, we will focus on the major categories of CNS injury, spinal cord injury (SCI) and traumatic brain injury (TBI), which have both common and distinctive features in pathophysiology. In general, a series of neurological events that occur after CNS traumatic injuries lead to drastic expression changes in both coding and non-coding genes. This involves primary insult, including mechanical damage, such as focal contusions, axonal injury, and hematomas [2]. The primary insult is followed by detrimental secondary injury, causing blood–brain barrier (BBB) disruption, inflammation, hypoxia cascades, cellular toxicity, neuronal cell death, and functional impairment [3–5]. Minimum regeneration occurs spontaneously after severe traumatic CNS injuries. Thereby, a better understanding of the gene expression changes and underlying molecular regulatory mechanisms holds the promise of developing therapeutic strategies to improve recovery.

With the development of next-generation sequencing technologies, complex alterations of gene expression have been characterized after injury. The expression levels of thousands of coding and non-coding genes changed in the injured CNS [5–12]. Notably, 70–90% of the mammalian genome is transcribed into RNA at some point during development, but only < 2% of the genome represents protein-coding genes [13–15]. Therefore, we will focus on a major category of non-coding RNA: the long non-coding RNAs (lncRNAs) in this review.

Regulatory functions of lncRNAs

LncRNAs are defined as gene transcripts with no protein-coding potential and with lengths greater than 200 nucleotides. They are a type of regulatory RNAs with important functions in the CNS. The categories of lncRNA mechanisms have been characterized as decoys, scaffolds, guides, enhancer RNAs and signals, etc. in transcriptional and translational processes [16–18]. For example, decoy lncRNAs sequester the regulatory effectors such as transcription factors, chromatin and miRNAs to negatively regulate these effectors [19]. Scaffold lncRNAs can serve as structural platforms to assemble protein and form functional complexes, such as ribonucleoprotein (RNP) complexes [20]. Guide lncRNAs can interact with the target molecules and direct them to the specific gene positions in cis (neighboring genes) or in trans (distal genes) [21]. In general, cis lncRNAs overlap introns, exons, promoter regions of neighboring protein-coding genes or enhancers (eRNAs). eRNAs are transcribed from enhancer DNAs and enhance the expression of target gens upon TF binding by forming chromatin looping [22]. For instance, in mouse cortical neurons, eRNAs can be transcribed bi-directionally from enhancer regions and positively correlated with its enhancer activity [23]. On the other hand, lncRNAs that overlap protein-coding genes antisense are commonly transcribed in opposite directions from neighboring coding genes. The expression of antisense lncRNAs may regulate the expression of their target protein-coding gene neighbors. Furthermore, lncRNA may also exert its function by acting as a signal molecule in response to different stimuli. For example, lncRNAs have been found to be associated with cell–cell communication through extracellular vesicles. LncRNAs may serve as scaffolds for proteins, mRNAs, or microRNAs (miRNAs), facilitating their exosomal loading into extracellular vesicles (EVs) [22].

Studies including ours have revealed that lncRNAs are transcribed in tissue-, cell type-, and developmental processes- specific patterns to exert their functions [14, 24–28]. They play various roles in cellular processes such as proliferation, apoptosis (programmed cell death), and migration [16–18]. Dysfunction of lncRNAs is closely related to a wide range of diseases, including neurological disorders [5, 16–18, 27, 29, 30]. Importantly, lncRNAs have been reported to play critical roles in CNS injury-induced secondary damage by gain- and loss-of-function approaches [5, 31, 32]. Nowadays, almost every drug is designed to target proteins. Innovative targeting of non-coding RNAs greatly enlarges the number of druggable targets and facilitates new classes of mechanistic discoveries [33, 34]. The negligible intrinsic nuclease activity within the cerebral spinal fluid (CSF) makes RNA-based agents an attractive therapy to treat CNS injury [35]. The above has highlighted the importance of investigating lncRNAs’ functions in CNS trauma, and their potential role as therapeutic targets in the future. Although the functions of lncRNAs in CNS traumatic injury are yet to be better elucidated, several studies have been conducted to unveil lncRNAs’ biological significance in this field. In this review, we summarize related investigations in coding and long non-coding gene expression changes associated with SCI and TBI, to gain deeper insights into both basic and clinical research.

Coding and long non-coding gene expression changes in SCI

SCI is one of the most detrimental CNS injuries [36]. The stages of SCI can be divided into acute, subacute, sub-chronic, and chronic phases [37]. By performing transcriptome analysis with RNA sequencing (RNA-Seq) using SCI models, researchers have characterized global gene profiles over a time course as well as gene transcription in specific cell types [5–10]. Gene expression profile analyses demonstrated aberrant changes in both coding and long non-coding RNA expression in rats and mice after SCI [6, 38–40]. For example, we systematically analyzed differentially expressed genes (DEGs) in sub-chronic and chronic phases of contusive SCI in rats [6]. In total, 4633 protein-coding genes and 277 lncRNAs were differentially expressed at 1, 3, and 6 months post-injury (mpi). The potential regulatory roles of lncRNAs were imputed from the functions of co-expressed protein-coding genes with a “guilt-by-association” approach [5]. The selected differentially expressed (DE) lncRNAs had high normalized enrichment scores (NES) in gene sets from the Molecular Signatures Database (MSigDB), and were categorized into five main function categories: signaling pathways (S), immune response (IR), epigenetic modification (EM), nervous system (N), and extracellular matrix (ECM). Motif analysis in the promoter regions of these DE lncRNAs showed that they contained binding motifs of DE transcription factors (TFs), such as STAT3 [41] and SPI1. The presence of these motifs suggested that these TFs might be involved in regulating the transcription of DE lncRNAs in SCI. In a separate study, we compared DEGs at 2 days post-injury (dpi), 7 dpi, 1 mpi, and 3 mpi after SCI with sham control in mice. 5675 protein-coding genes and 778 lncRNAs were differentially expressed in at least one comparison. 297 DE lncRNAs were upregulated in at least one-time point after SCI. Then, we used slncky software to derive a set of DE lncRNAs that are likely to be functionally important based on evolutionary conservation. Next, we applied the “guilt-by-association” analysis to predict the potential functions of these lncRNAs. We found that the potential functions were highly correlated with immune response, cell differentiation, cell proliferation, and cytokine biosynthesis [5, 7].

In addition, the enriched signaling pathways and functions are altered at different stages of SCI. For example, DEGs were investigated during the acute and subacute phases of SCI in rats [42]. This study obtained 384, 585, and 649 DEGs in SCI at 1, 4, and 7 dpi from rats compared with sham, respectively. Pathway analysis revealed the transition from inflammatory responses to multiple forms of cell death processes from acute to subacute stages. In our study, we compared transcriptome datasets from acute (2 and 7 dpi) to chronic (1 and 3 mpi) SCI stages in mice [5]. The gene set enrichment results showed that genes with functions in cell proliferation and differentiation were enriched during the acute stages, whereas genes encoding cell–cell adhesion and extracellular matrix proteins were enriched during chronic stages.

In addition to bulk RNA-seq, recent development of single-cell RNA-seq (scRNA-Seq) technologies allows for the characterization of heterogeneity and interactions of cell types. For instance, scRNA-Seq analysis of uninjured and injured mouse spinal cord at 1, 3 and 7 dpi identified 15 distinct clusters comprising the major cell types [43]. The top DEGs showed a unique molecular signature for each cell type. Based on annotated marker and Gene Ontology (GO) biological processes terms, four microglial subtypes (steady-state subtype as the homeostatic microglia, and inflammatory, dividing, and migrating subtypes as non-homeostatic microglia) and two major macrophage subtypes (chemotaxis-inducing macrophages and inflammatory macrophages) were identified. These multiple cellular states of both microglia and macrophages displayed unique transcriptomic signatures, functional pathway enrichment, and dynamic changes during the progression of injury. For instance, inflammatory microglia was mostly related to cell death and cytokine production, and these microglia were identified by low expression of P2ry12 and high expression of Igf1. Dividing microglia were mostly associated with the cell cycle and had low expression of P2ry12, increased expression of Msr1, and high expression of Cdk1. In addition, analysis of cell–cell interaction via the specified ligand-receptor pathway suggested potential angiopoietin (Angpt)-Tie2 interactions between vascular and endothelial cells, as well as potential Vegfa–Vegfr interactions of astrocytes with endothelial cells during angiogenesis. This dataset provides a valuable resource for SCI research community [43]. Although lncRNA expression was not a focus of this study, future re-analysis of the dataset and characterization of lncRNAs can lead to additional mechanistic insights.

Furthermore, there are other single-cell studies focusing on the specific cell types involved in SCI. For example, by integrating scRNA-Seq and single-cell assay for transposase-accessible chromatin sequencing (scATAC-seq), one study investigated the ependymal-derived cells from uninjured and injured spinal cord in Foxj1-tdTomato (tdT) and Foxj1-Olig2-tdT mice (conditional and simultaneous expression of OLIG2 and tdT in ependymal cells) at 2 and 4 weeks post-injury (wpi) [44]. Six clusters were identified including one ependymal, three astroependymal (AE1-3), and two ependymal-derived oligodendrocyte lineage (epOL1, 2) subtypes. Pseudotime analysis of Foxj1-Olig2-tdT group demonstrated AE 3 and epOL 2 as the terminal differentiation states of the divergent branches of ependymal cell differentiation. The epOL cells displayed self-amplifying oligodendrocyte progenitor cell-like states and may generate mature myelinating oligodendrocytes without disturbing astrocyte scarring. The ependymal-derived oligodendrocytes remyelinated the axons and contributed to axon conduction in long term by migrating to the demyelinated region after SCI [44].

As above mentioned, after SCI, the primary injury could cause cell damages and initiate a complex secondary injury leading to inflammatory response, astrocyte activation, astrogliosis, cell apoptosis, matrix remodeling, axon degeneration/demyelination, etc. [45, 46]. In the following sections, we will focus on some of these aspects.

Inflammation

Inflammatory response plays a key role in the pathogenesis of SCI [47]. After SCI, the blood–spinal cord barrier (BSCB) is compromised, and the damaged cells release microglial chemotaxis, which activates microglia. The innate immune response triggered by microglia is enhanced by peripheral myeloid cells, including neutrophils and monocytes, which induces the proliferation of astrocytes, microglia, and oligodendrocyte precursor cells (OPCs), thus initiating gliosis. A number of involved transcription regulators have been characterized during inflammatory response. For instance, nuclear factor-κB (NF-κB) is a family of transcription factors that regulates a list of genes involved in the process of immune and inflammatory responses. Previous study showed NF-κB signaling pathway was activated after SCI in rat [48]. This study reported 151 upregulated lncRNAs and 186 downregulated lncRNAs in the 3 dpi group compared with control group. Five lncRNAs (Loc685699-ot1 (Airsci), Aabr07015057.1-ot11, Linc5298, Aabr07000398.1-ot18, and Aabr07055878.1-ot1) were found to be possibly involved in the NF-κB signaling pathway. Further quantitative real-time PCR results showed that lncRNA Airsci increased significantly after SCI. The inhibition of Airsci not only reduced the protein expression levels of NF-κB (p65) and p-IκBα, but also decreased the protein and mRNA expression levels of inflammatory factors (e.g., IL-1β, IL-6, and TNF-α). Histopathological alteration and functional test showed suppressing Airsci repressed lesion size as well as neuronal loss, and promoted the recovery of motor function in SCI rats. These results suggested lncRNA Airsci might affect inflammatory response in the spinal cord after injury through the NF-κB signaling pathway, thereby regulating functional recovery. In addition, toll-like receptor 4 (TLR4) has been shown to trigger innate immune responses by interacting with infectious agents or endogenous ligands in the spinal cord [49]. In rat spinal cord ischemia/reperfusion (I/R) model, the inhibition of TLR4 decreased the release of inflammatory cytokines (such as IL-1β and IL-6) after SCI [50], and the downregulation of lncRNA Tug1 inhibited the expression of TRIL (TLR4-interactor with leucine-rich repeats) and TLR4, as well as the TLR4-mediated inflammatory response of the NF-κB pathway [51]. Moreover, the Janus kinase/signal transduction and activator of transcription (JAK-STAT) signaling pathway is implicated in the inflammatory response. LncRNA Znf667-as1 was downregulated in mouse cervical SCI over a time course, and the overexpression of Znf667-as1 inhibited the mRNA expression and protein levels of JAK2 and STAT3, promoting grip strength recovery of forelimbs. The results suggested Znf667-as1 might inhibit inflammatory response and improve functional recovery after SCI through the JAK-STAT pathway [52].

As mentioned above, microglia are the resident cells and act as early responders. They may have limited phagocytic capacity as compared to blood-borne macrophages but can produce growth factors, cytokines, and chemokines to recruit blood-borne inflammatory cells to assist with debris clearance after SCI [53, 54]. Recent RNA-Seq studies have revealed microglia molecular signatures in the spinal cord. Using scRNA-Seq and functional analyses in FACS (fluorescence-activated cell sorting)-purified microglia from neonatal mice, it was reported that microglia in neonatal mice could express several extracellular and intracellular peptidase inhibitors, as well as other inflammatory molecules (e.g., CD68) [55]. After transplanting neonatal microglia, or adult microglia treated with combination of peptidase inhibitor (E64, a membrane-permeable irreversible inhibitor of a wide range of cysteine peptidases; and SerpinA3N, a serine protease inhibitor) into spinal cord lesion sites of adult mice, the deposition of collagen I and CSPG (chondroitin sulfate proteoglycan) was significantly reduced, and more serotonergic axons crossed the lesion [55]. In addition, lncRNAs could also regulate microglia cell inflammation after SCI. For example, the expression of lncRNA Malat1 dramatically increased in microglia after SCI [56]. The knockdown (KD) of Malat1 not only attenuated lipopolysaccharide (LPS)-induced activation of microglia and proinflammatory cytokine production (TNF-α and IL-1β), but also improved the hindlimb locomotor activity of SCI rats.

Macrophages are differentiated from monocytes and can phagocytose cellular debris and induce fibrosis [43]. They also play critical roles after CNS injury. In addition to clearing debris, activated macrophages can induce retraction and dieback of axons after SCI, as well as interact with stromal non-neural cells in the lesion core [54]. Using the RiboTag method (purifying cell-specific ribosome-associated mRNA (ramRNA) via a hemagglutinin tag (RiboTag)), the macrophage-specific mRNA was extracted directly from mouse injured spinal cord. RNA-seq was conducted to explore the macrophage mRNA transcriptional profile [57]. A total of 15,565 genes were identified with 5597 DEGs between 3 and 7 dpi macrophages. Among the top 20% of the highest expressed genes, both time points had 1,725 genes in common, whereas 367 genes were unique to the macrophages at 3 dpi and 324 genes were unique to 7 dpi. From 3 to 7 dpi, the macrophage-specific transcriptional profile after SCI shifts from cell migration and cytokine signaling to lipid catabolism. Two canonical lipid catabolism pathways which were activated in macrophages at 7 dpi were identified: liver X receptor/retinoid X receptor (LXR/RXR) and peroxisome proliferator-activated receptor alpha (PPARα/RXRα). Knockout of the classic lipoprotein scavenger receptor (Cd36) led to decreased macrophage lipid content, reduced lesion size, and improved locomotor recovery [57]. These observations are consistent with our earlier publication in 2013 on acute SCI, in which we found that the LXR/RXR activation pathway encompassing Cd36 was significantly upregulated at both 2 and 7 dpi. We also established that Cd36 was a “hub” gene by network interaction analysis and was concluded as a gene of interest in SCI [7]. Moreover, as above-mentioned, the recent scRNA-Seq study identified two subtypes of macrophages in SCI: chemotaxis-inducing macrophages and inflammatory macrophages based on GO terms [43]. The chemotaxis-inducing macrophages were related to the chemotaxis of other leukocytes such as neutrophils while the inflammatory macrophages were related to glial and macrophage activation and inflammatory response. They both expressed Cd63 and can be identified by preferential expression of heme oxygenase Hmox1 in chemotaxis-inducing macrophages and Apoe in inflammatory macrophages. Inflammatory macrophages (APOE^hi) were found to be the predominant macrophage subtype at 7 dpi [43].

The functions of macrophage lncRNAs in SCI remain elusive. One recent study reported that lncGBP9 regulated macrophages’ functions in SCI via miR-34a. By competing with SOCS3 (suppressor of cytokine signaling 3) for miR-34a binding, lncGBP9 counteracted miR-34a-mediated suppression on SOCS3 and then regulated STAT1/STAT6 signaling in macrophages [58]. Knocking down lncGBP9 in SCI mouse downregulated the expression of macrophage marker Arg1, along with important signaling molecules STAT1 and SOCS3. This downregulation improved functional recovery after SCI.

Astrogliosis

After SCI, local astrocytes become reactive and undergo a process named astrogliosis. Reactive astrocytes proliferate, surrounding the margins of lost neural tissue and forming an astrocyte limitans border to separate the spared neural tissue from inflammatory cells and stromal cell scar tissue. During this process, reactive astrocytes interact with other cell types, playing a vital role in axon regeneration [59]. To unveil the transcriptomic profile of reactive astrocytes, Anderson et al. purified astrocyte-specific ramRNA via RiboTag from wild type (WT) and STAT3-CKO (STAT3 conditional knock out in astrocytes) mice after SCI [60]. They performed RNA-Seq at 14 dpi. The result showed that 63% of genes significantly regulated in WT SCI were not significantly changed in STAT3-CKO astrocytes after injury. The transcriptomic profiling of STAT3-CKO SCI astrocytes is more similar to the profile of uninjured astrocytes than of WT SCI astrocytes. Moreover, the researchers analyzed the expression of both axon-inhibitory and -permissive molecules in the experimental groups. Compared to WT uninjured group, in WT SCI groups, reactive astrocytes expressed many inhibitory and permissive molecules. Compared to WT SCI, attenuating astrocyte scar formation after SCI with STAT3-CKO reduced the gene expression related to astrogliosis, astrocyte border formation, and axon-permissive molecules, and increased the genes of axon-inhibitory molecules. The failure of astrocyte border formation and lack of axon growth-supporting molecules led to the significant reduction of axon regrowth in in STAT3-CKO SCI animals [60]. In a recent study [61], the same group of researchers found a combination of three mechanisms essential for facilitating axon regrowth in both SCI mice and rats: (1) reactivation of neuronal intrinsic growth capacity by AAV (adeno associated virus) delivery of osteopontin (coded by Spp1), IGF1 and CNTF (AAV-OIC), (2) induction of axon growth-supportive substrates by delivering hydrogel depot containing fibroblast growth factor (FGF) 2 and epidermal growth factor (EGF), and (3) chemoattraction by delivering glial-derived growth factor (GDNF) hydrogel depot. They delivered growth factors FGF2 and EGF to the lesion center and found that multiple genes and networks associated with proliferation and development in astrocytes, and in non-astrocyte cells associated with inflammatory responses, were significantly changed. Through administration of FGF + EGF + GDNF followed by a second depot of GDNF in AAV-OIC rodents, histological analysis showed robust axon regrowth across lesion cores, and synapse markers colocalized in the regrowing axon contacts.

RiboTag precipitation allows for the purification of protein-coding RNAs. Therefore, to investigate both coding and non-coding genes involved in the molecular basis of reactive astrogliosis, we purified astrocytes from the epicenter tissue of GFAP-Cre:R26-tdT mice at 7 dpi, 1 mpi, and 3 mpi [5]. Compared with sham controls, thousands of protein-coding genes and hundreds of lncRNAs were significantly changed. The common DEGs from all stages were enriched in not only the immune system and cytokine production but also in secretory vesicle and complement system functions. Homologous genes of lncRNAs between mouse and human were also identified. We tested a number of DE lncRNAs and found a gene of interest, Zeb2os, which was highly conserved in humans (ZEB2-AS1). The expression of Zeb2os had a high correlation of expression with its antisense protein-coding gene Zeb2 at different SCI stages, as well with an essential TF gene in astrogliosis, Stat3 [41, 62]. Guilt-by-association analysis of Zeb2os in purified astrocytes revealed that Zeb2os had a highly significant positive correlation with astrocyte functions, STAT3 pathway, and integrin pathway. Zeb2os KD in primary astrocytes affected several downstream genes, such as Gfap and Zeb2 (Fig. 1a). Astrocyte proliferation significantly decreased in Zeb2os KD astrocytes, as shown by BrdU staining (Fig. 1b). In addition, our data demonstrate that Zeb2os-Zeb2 colocalized in GFAP + cells by RNAscope in primary astrocyte culture and in SCI tissue sections in vivo (Fig. 1c). The KD of Zeb2os in vivo by AAV injection reduced GFAP (Fig. 1d) and pSTAT3 (Fig. 1e) expression, as well as lesion volume (Fig. 1f) at 17 dpi. The ChIP-Seq data revealed that STAT3 bound to the Zeb2 promoter region; thus, Zeb2os may regulate Zeb2 and Stat3 directly or indirectly, and STAT3 may regulate Zeb2 expression. These results indicate that Zeb2os could regulate astrogliosis through a Zeb2os/Zeb2/Stat3 axis (Fig. 1g).

Fig. 1.

a qPCR analysis of Zeb2os KD in primary astrocytes using lentivirus. Data are presented as means ± SEM; n = 3 independent experiments; ∗p < 0.05 compared with Luci (independent t test); Luci, luciferase shRNA control; Zeb2os-1 and Zeb2os-2, two different Zeb2os shRNA KD constructs. b DAPI (blue)/BrdU (green) double labeling in primary astrocytes from control (Luci ctrl) and Zeb2os KD groups (scale bar, 100 mm). The percentage of BrdU-expressing cells is significantly decreased in the Zeb2os KD group. *p < 0.05 compared with Luci (independent t test). c Combination of RNAscope and immunohistochemistry showing Zeb2os mRNA (red) colocalized with Zeb2 mRNA (green) (indicated by triangles) in scratched astrocyte (left), and insets show a magnification of the boxed area (scale bar, 20 mm); and in injured spinal cord 17 dpi (indicated by arrows) (scale bar, 3 mm). d Immunohistochemistry of GFAP (red) expression in Zeb2os KD in astrocytic scar compared with control (scale bar, 20 mm). e Immunohistochemistry of pSTAT3 (red) expression in Zeb2os KD astrocytes compared with control (scale bar, 20 mm). f Quantification of the lesion volume using GFAP immunostaining. g Working model of a Zeb2os/Zeb2/Stat3 axis in reactive astrogliosis.

Adapted from Wei and others (mouse study) [5] with permission and modification

Another study reported that lncSCIR1 was downregulated at acute/subacute stages in rat contusive SCI model [63]. EdU cell proliferation and transwell assays showed that the migration of astrocytes significantly increased in vitro after lncSCIR1 KD. On the contrary, lncRNA Snhg5 was upregulated in astrocytes in rat SCI [32]. The proliferation of astrocytes decreased after Snhg5 KD [32]. Moreover, overexpressing Snhg5 with the administration of lenti-Snhg5 in rat SCI attenuated Basso–Beattie–Bresnahan scale and Basso Mouse Scale (BMS) scores, while enhancing the protein expression of KLF4, eNOS, GFAP and IBA-1. All these results suggest that lncRNAs could potentially affect astrocyte migration or proliferation after injury.

Cell apoptosis

Cell apoptosis follows the inflammatory response after SCI [64]. Recent studies have revealed that lncRNAs are involved in the regulation of neuronal apoptosis after SCI. In some cases, lncRNAs promote neuronal apoptosis. For instance, it was found that knocking down lncRNA Bdnf-as could downregulate PRDM5 (a family member of PR (PRDI-BF1 and RIZ) domain proteins that regulates cell differentiation, growth, apoptosis, etc.) in acute SCI and suppress neuronal cell apoptosis [65]. Further investigations indicated that PRDM5 is a target of miR-130b-5p. By sponging miR-130b-5p, Bdnf-as acted as a competing endogenous RNA (ceRNA) in neuronal cells. In another example, lncRNA Xist was significantly upregulated after SCI [66]. The shRNA KD of Xist inhibited cell apoptosis by reactivating the PI3K/AKT signaling pathway in rat SCI. In some other cases, lncRNAs can also repress neuronal apoptosis. For example, it was reported that Dgcr5 could directly bind to PRDM5 protein and negatively regulate it, thereby suppressing neuronal cell apoptosis in acute SCI [67]. In addition, Tctn2 repressed apoptosis via the miR-216b-Beclin-1 pathway in SCI [68]. In addition, lncRNA CasC7 could reduce neuronal cell apoptosis by regulating miR-30c expression in spinal cord I/R injury [69].

Coding and long non-coding gene expression changes in TBI

TBI is defined as the brain damage resulting from a rapid movement of the brain within the skull caused by a traumatic event. The complexity of TBI pathology is a limiting factor for the development of diagnostic and prognostic strategies. Gene expression study can help to delineate better the molecular mechanisms underlying the pathophysiology. For instance, transcriptomic and epigenomic profiling analysis of hippocampus and blood leukocytes from TBI rats identified 240 and 1052 DEGs in hippocampus and leukocytes, respectively [11]. Among 18 common genes identified in the two tissues, such as Abca4, Adipor1 and Aqp1, 16 of them showed opposite expression alteration, and 3 showed the same directional change of expression. Functional characterization revealed that both the hippocampal and leukocyte samples had enriched pathways of axon guidance, ECM (extracellular matrix), and lysosome; and numerous signaling pathways for insulin, IGF, FGFR, neurotrophin, PDGF, etc. The alteration of genomic signatures from blood and brain may be used to develop potential biomarkers of TBI. Moreover, the intersection between DEGs in rat TBI with human genome-wide association studies (GWAS) indicated the association of TBI with neurological and psychiatric disorders in humans.

Compared to the bulk RNA-Seq of heterogenous cell mixtures, scRNA-Seq of TBI samples unveils transcription signals from different cell subtypes or states. A recent scRNA-Seq study identified 15 clusters containing the classical hippocampal cell types and 2 unknown clusters in TBI mice using Drop-Seq [12]. The gene expression signatures indicated Unknown1 as migrating endothelial cells with markers of endothelial, and Unknown2 as likely progenitor cells differentiating into a variety of cell lineages. Cell–cell gene co-expression analysis found cell–cell communication of astrocytes and ependymal cells with neurons, as well as interactions between microglia with oligodendrocytes increased. This study also found various pathways related to TBI, such as energy and metabolism in astrocytes and neuron clusters, inflammation and immune response in microglia and OPCs, and myelination in oligodendrocytes. Besides co-expressed genes between cell types, this study reported cell type-specific DEGs as potential biomarkers or therapeutic targets related to TBI pathology. For instance, seizure-related gene Id2 was specifically upregulated in dentate gyrus granule cells after TBI, which makes it a potential target for modulating post-traumatic epilepsy. In addition, gene Ttr was strongly upregulated in most of the neuronal clusters and ependymal cells. Ttr encodes transthyretin to carry the thyroid hormone thyroxine T4 across BBB, and its upregulation may indicate the compensatory requirement of T4 to modulate cellular metabolism after TBI. The protective effect of T4 was verified by treating TBI mice with T4 transporter. RNA-Seq identified 951 affected DEGs, and 93 TBI-altered genes were reversed by T4 treatment. Downstream behavioral tests showed that T4 treatment improved cognitive learning and memory.

In addition to coding genes, increasing evidence shows that lncRNAs are involved in the pathological process of TBI. Using a controlled cortical impact model, one RNA-Seq study in mouse TBI showed that 667 lncRNA were upregulated and 156 lncRNA were downregulated [70]. Furthermore, the expression profile of lncRNAs in human TBI revealed a positive correlation between the total number of altered lncRNA and the injury severity [71]. By investigating the whole blood sample from TBI and healthy subjects using Agilent human lncRNA microarray, the researchers identified 3035 DE lncRNAs (1685 upregulated and 1349 downregulated) in the severe TBI group, 2362 DE lncRNAs (1874 upregulated and 488 downregulated) in the moderate group, and 433 DE lncRNAs (183 upregulated and 250 downregulated) in the mild group. These investigations suggested lncRNAs are involved in TBI pathogenesis, but their roles still need to be further elucidated. Among the studies, pathway analysis of the potential target genes of the DE lncRNAs and functional tests indicate that many lncRNAs are involved in various pathways of TBI. To date, four main biological functions regulated by lncRNAs have been discovered: astrocytic function, neuroinflammation, neuronal death, and cerebral vasculature alterations. Here, we highlight the functional roles of lncRNAs in TBI, which are well documented through high-throughput assays and classical biochemical methods.

Neuroinflammation

Neuroinflammation is a critical contributor to the pathological process of TBI. After TBI, circulating neutrophils, monocytes, and lymphocytes infiltrate into cerebral parenchyma through damaged BBB [3, 15, 72]. These leukocytes produce complement factors and proinflammatory molecules, such as IL-6 and TNF-α, which aggravate BBB permeability, brain edema, and neurological dysfunction [3, 73]. More microglia are recruited to the injury site. The prolonged inflammatory response can cause demyelination, axonal damage, and neurodegeneration [3].

Increasing evidence elucidated that lncRNAs regulate neuroinflammation process through microglia activation. For example, the expression of lncRNA Hotair increased in the mouse TBI model [74]. It was shown to mediate microglia activation and cytokine production by regulating the ubiquitination of MYD88 in LPS-treated microglia. Similarly, lncRNA Kncql antisense transcript 1 (Kcnq1ot1) was also increasingly expressed in TBI mice, along with elevated levels of cytokines (such as IL-1β, IL-6, and TNF-α). Knocking down Kcnq1ot1 repressed these cytokines’ expression and associated neuronal damage [75]. It was further verified that suppressing Kcnq1ot1 attenuated microglial activation and inflammation, and had neuroprotective effects on TBI mice by regulating the miR-873-5p-TRAF6-p38/NF-κB axis. In addition, lncRNA Malat1 has been shown to regulate TBI recovery in a study of the secretome of human adipose-derived stem cells (hASCs), which was administered in TBI rats [76]. Malat1 worked as a vital component of the hASC secretome. Brain RNA-Seq showed that TBI groups had upregulation of inflammatory regulators such as IL1β, IFNG, TLR2, IL27, and CD40 compared to control, whereas the upregulation of these inflammation genes was diminished in the exosome-treated TBI group. Ingenuity Pathway Analysis showed that the potential upstream regulators include PTX3, MMP9, POMC, and CD80. In the TBI group treated with Malat1-depleted exosomes, the gene network and predicted upstream regulators demonstrated a very similar expression pattern as the TBI group. Functional tests showed the treatment of Malat1-depleted exosomes had limited regenerative effects in TBI rats, which is consistent with the pattern of transcriptome analysis mentioned above.

Astrocytic functions

Similar to reactive astrocytes in SCI, astrocytes are also involved in the pathophysiological process and have dual roles (both beneficial and inhibitory effects) in TBI [77–79]. Astrocytic functions in TBI can be regulated by lncRNAs as well. For instance, lncRNA Gm4419 has been shown to promote trauma-induced astrocyte apoptosis by upregulating the expression of inflammatory cytokine TNF-α [80]. Mechanistically, miRNA miR-466l targets TNF-α 3’ UTR, leading to TNF-α degradation and translation inhibition. An RNA precipitation and dual-luciferase reporter assay demonstrated that Gm4419 acts as a sponge of miR-466 l and upregulates TNF-α expression, thereby inducing inflammatory damage as well as astrocytic apoptosis in the traumatic injury model of astrocytes. Another study has validated that astrocyte-derived EVs carrying lncRNA Nkila (NF-κB-interacting lncRNA) promote brain recovery after TBI in vivo [81]. Upregulation of lncRNA Nkila increased neuron proliferation and decreased neuron apoptosis in vitro by competitively binding to miR-195, as well as by upregulating NLRX1 (NLR family member X1). Another study found that Malat1 was downregulated in both the rat TBI model and the astrocyte fluid percussion injury (FPI) model [82]. Overexpression of Malat1 significantly reduced rat brain edema, as well as astrocyte swelling, which may be mediated by inhibiting the production of IL-6, NF-κB, and aquaporin 4 (AQP4, a predominant water channel expressed by astrocytes) after TBI.

Neuronal death

Following TBI, neuronal death can be identified within contusions in the acute stage, and the impact on the remote regions occurs in days and weeks after trauma [83]. There are two types of neuronal death: necrosis and apoptosis. Neuron necrosis is a passive process associated with failure of membrane integrity, irreversible metabolic disturbances, and organelle and/or excitotoxicity after mechanical trauma [84]. Regarding neuronal apoptosis, many signaling pathways are involved, including caspases and proapoptotic members of the B-cell lymphoma-2 (Bcl-2) family, c-Jun N-terminal kinase (JNK), and autophagy signaling pathway (ATG) [83, 84]. Some lncRNAs have been found to affect neuronal repair and death in TBI through the regulation of these key signaling pathways, such as Gas5, Crnde, and Neat1. Transcriptome profiling indicated that lncRNA Gas5 was upregulated in both rat hippocampus and mouse cortex at the early stage of TBI [70, 85]. Further in vitro study confirmed that the overexpression of Gas5 attenuated neuronal cell viability, promoted proapoptotic protein Bax, and lowered protein levels of anti-apoptotic protein Bcl-2 [86]. Gas5 may function as a competing endogenous RNA (ceRNA) to sponge miR-335, which upregulated Rasa1 expression via miR-335/Rasa1 axis in mouse neuronal cells. In addition, another study model showed that Gas5 acted as proinflammatory mediator by mediating microglia functions [87].

Another example is lncRNA CRNDE (colorectal neoplasia differentially expressed), which was highly expressed in the serum of TBI patients [88]. Knocking down Crnde in TBI rats improved neurobehavioral function, inhibited expression of neuroinflammatory factors, enhanced the number of Nissl bodies, and restricted neuronal apoptosis and autophagy. Furthermore, this study also found that downregulation of Crnde increased expression of NGF, Nestin and NeuN proteins, implying the promotion of neuron differentiation and nerve regeneration.

In addition, lncRNA Neat1 was found to be an important mediator in neuroprotective effects of bexarotene on traumatic brain injury in mice [89]. Neat1 was upregulated in mouse TBI compared to sham, and the agonist of RXR-bexarotene can increase Neat1 expression compared to the vehicle group. Using neuronal cell line, ChIP-Seq and luciferase assay verified RXR bound to Neat1 promotor, and RNA-Seq of Neat1 KD cells showed that multiple biological processes were regulated, including synapse formation and axon guidance. Overexpressing Neat1 improved axonal extension. RNA pulldown combined with mass spectrometry and downstream functional assays demonstrated that the overexpression of Neat1 restricted neuron apoptosis and microglia’s inflammatory effect in vitro by targeting Pidd1 protein. Upregulated Neat1 facilitated a better recovery of motor and cognitive function after TBI in mice.

Cerebral vasculature

Like SCI, TBI causes BBB deformation and disruption, leading to hypoperfusion; altered delivery of metabolic substrates; and I/R damage, vasospasm, and coagulopathy [90, 91]. LncRNA also plays a vital role in this process. For instance, using oxygen–glucose deprivation/reoxygenation (OGD/R) as an in vitro I/R injury model for studying blood–brain barrier dysfunction, research showed that lncRNA Malat1 can exert a protective role by inducing brain microvascular endothelial cell (BMEC) autophagy [92]. In addition, lncRNA Snhg1 protected BMEC against OGD-induced injury via sponging miR-338, and upregulated HIF-1α/VEGF-A in BMEC [93].

Perspectives and conclusions

In summary, this review has described the recent progress in transcriptome analysis in CNS traumatic injury. The expression of both coding genes and lncRNA is significantly altered in both rodents and humans after SCI and TBI (Table. 1). At the cellular level, lncRNAs have been shown to regulate the expression of protein-coding RNAs involved in astrogliosis, inflammation, neuronal death, and vasculature events. Therefore, lncRNAs could be promising biomarkers and targets for the diagnosis, treatment, and prognosis of CNS traumatic injuries.

Table 1.

List of lncRNA examples in SCI and TBI (in the order in which they appear in the main text)

lncRNA	Species/cells	Model	Expression after injury	Potential functions	Targets/pathways involved	References
Airsci	Rat	SCI	Upregulated	Inflammatory response	NF-κB, p-IκBα	[48]
Tug1	Rat	SCI	Upregulated	Inflammatory response	TRIL/TLR4, NF-κB	[51]
Znf667-as1	Mouse	SCI	Downregulated	Inflammatory response	JAK-STAT	[52]
Malat1	Rat	SCI	Upregulated	Microglia inflammatory response	miR-199b, IKKβ/NF-κB	[56]
LncGBP9	Mouse	SCI	Upregulated	Microglia inflammatory response	miR-34a/SOC3, STAT1/STAT6	[58]
Zeb2os	Mouse	SCI	Upregulated	Reactive astrogliosis	Zeb2 and Stat3	[5]
LncSCIR1	Rat	SCI	Downregulated	Astrocyte proliferation and migration		[63]
Snhg5	Rat	SCI	Upregulated	Astrocyte proliferation	KLF4	[32]
Bdnf-as	Rat	SCI	Upregulated	Neuronal apoptosis	mir-130b-5p/PRDM5	[65]
Xist	Rat	SCI	Upregulated	Neuronal apoptosis	miR-494/PTEN/PI3K/AKT	[66]
Dgcr5	Rat	SCI	Downregulated	Neuronal apoptosis	PRDM5	[67]
Tctn2	Rat	SCI	Downregulated	Neuronal apoptosis	miR-216b/Beclin-1	[68]
CasC7	Rat	SCI	Downregulated	Neuronal apoptosis	miR-30c	[69]
Hotair	Mouse	TBI	Upregulated	Microglia activation	MYD88	[74]
Kcnq1ot1	Mouse	TBI	Upregulated	Microglial activation and inflammation	miR-873-5p-TRAF6-p38/NF-κB	[75]
Malat1	Rat	TBI	Downregulated	Reduces brain edema and astrocyte swelling Modulates hASC-derived exosomes’ function	NF-κB/IL-6	[76, 82]
Gm4419	Mouse primary astrocytes	TBI	Upregulated	Astrocyte apoptosis	miR-466l, TNF-α	[80]
Nkila	Rat	TBI	Downregulated	Transferred by astrocyte-derived extracellular vesicles and protects against neuronal injury	mir-195, NLRX1	[81]
Gas5	Mouse	TBI	Upregulated	Neuronal apoptosis; regulates microglia’s function	miR-335/Rasa1	[86, 87]
CRNDE	Human, rat	TBI	Upregulated	Inflammation and neuronal apoptosis, differentiation and regeneration		[88]
Neat1	Mouse	TBI	Upregulated	Axon elongation and neuronal apoptosis and microglial inflammation	Pidd1	[89]
Malat1	BMEC	OGD/R	Upregulated	Autophagy inducer of endothelial cell	miR-26b/ULK2	[92]
Snhg1	BMEC	OGD	Upregulated	Protect BMEC from hypoxic/ischemic injury	miR-338/HIF-1α/VEGF-A	[93]

Nowadays, RNA-based therapy has emerged as a promising therapeutic tool in many diseases. For instance, oligonucleotide-based drugs nusinersen (Spinraza), mipomersen (Kynamro) and eteplirsen (Exondys 51) [35] have been recently approved by the FDA for the treatment of spinal muscular atrophy, homozygous familial hypercholesterolemia and Duchenne muscular dystrophy, respectively (http://www.accessdata.fda.gov/scripts/cder/daf/index.cfm). The absence of nucleases within the CSF has made intrathecal injection of RNA-based agents an attractive approach to treat CNS traumatic injuries. Nevertheless, there is still a significant knowledge gap in determining lncRNAs’ mechanisms, as well as the functional outcomes of lncRNA manipulation. Continued investigation will be crucial for revealing the roles of lncRNAs in CNS injuries and could lead to the development of RNA-based therapeutic approaches.

Author contributions

XW, HW, and JQW wrote the manuscript. All the authors read and approved the final manuscript.

Funding

This work was supported by Grants from the NIH, United States (R01 NS088353, R21 NS113068-01, and R21 EY028647-01), The Staman Ogilvie Fund-Memorial Hermann Foundation, and Mission Connect, a program of The Institute for Rehabilitation and Research (TIRR) Foundation.

Availability of data and material

Not applicable.

Declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical approval and consent to participate

No animal or human study was performed.

Consent for publication

All the authors gave their consent for publication.