Transcriptomic and bioinformatics analysis of the mechanism by which erythropoietin promotes recovery from traumatic brain injury in mice

Abstract

Recent studies have found that erythropoietin promotes the recovery of neurological function after traumatic brain injury. However, the precise mechanism of action remains unclear. In this study, we induced moderate traumatic brain injury in mice by intraperitoneal injection of erythropoietin for 3 consecutive days. RNA sequencing detected a total of 4065 differentially expressed RNAs, including 1059 mRNAs, 92 microRNAs, 799 long non-coding RNAs, and 2115 circular RNAs. Kyoto Encyclopedia of Genes and Genomes and Gene Ontology analyses revealed that the coding and non-coding RNAs that were differentially expressed after traumatic brain injury and treatment with erythropoietin play roles in the axon guidance pathway, Wnt pathway, and MAPK pathway. Constructing competing endogenous RNA networks showed that regulatory relationship between the differentially expressed non-coding RNAs and mRNAs. Because the axon guidance pathway was repeatedly enriched, the expression of Wnt5a and Ephb6, key factors in the axonal guidance pathway, was assessed. Ephb6 expression decreased and Wnt5a expression increased after traumatic brain injury, and these effects were reversed by treatment with erythropoietin. These findings suggest that erythropoietin can promote recovery of nerve function after traumatic brain injury through the axon guidance pathway.

Introduction

Traumatic brain injury (TBI) is one of the leading cause of death and disability worldwide (Jamjoom et al., 2021), a major cause of morbidity and mortality, and a public health issue (Capizzi et al., 2020). TBI treatment is based on the severity of the injury. Although mild TBI usually requires no treatment other than rest, moderate and severe TBIs require emergency care. In addition, patients with moderate and severe TBIs usually need additional treatment to reduce inflammation, decrease bleeding, or increase oxygen supply to the brain. Over the years, numerous neuroprotective agents have been developed to treat TBI, including the calcium channel inhibitor nimodipine (Hassan et al., 1999), anti-inflammatory drugs, and anti-apoptotic drugs. Some drugs have been reported to exert a neuroprotective effect by decreasing inflammation or apoptosis after TBI, such as edaravone (Wang et al., 2011), ganglioside (Li et al., 2021b), and tranexamic acid (Nelson Yap et al., 2020). However, there are few reports on drug usage and adverse side effects (Lapchak, 2010; Wu et al., 2014; Xiong et al., 2015).

Erythropoietin (EPO) is a hematopoietic regulatory hormone that confers neuroprotection when directly injected into the ischemic rodent brain (Rangarajan and Juul, 2014) through its anti-apoptotic, anti-inflammatory, anti-oxidant, angiogenic, and neurotrophic actions (Cotena et al., 2008; Velly et al., 2010). A retrospective study of patients with severe TBI showed that the use of rhEPO (a recombinant form of EPO) markedly reduced in-hospital mortality compared with matched case-control groups (Talving et al., 2010). In addition, Zhiyuan et al. (2016) reported changes in the expression of protein-coding genes, such as Claudin-5, Occludin, and ZO-1, after TBI treatment with EPO; yet, the role of non-coding RNAs (ncRNAs) in TBI treated with EPO remains unclear.

ncRNAs are essential to a wide range of biological processes (Li et al., 2017b, 2022; Feng et al., 2022). Long non-coding RNAs (lncRNAs) are involved in regulating gene expression through various transcriptional or post-transcriptional mechanisms. Abnormal lncRNA expression might promote disease development or affect disease progression (Mattick, 2011). Zhang et al. (2019) found that lncRNAMalat1 overexpression inhibited aquaporin 4 (AQP4) expression, as well as the nuclear factor kappa-B and interleukin-6 (NF-κB/IL-6) pathways, to reduce brain edema caused by TBI. Furthermore, circular RNAs (circRNAs) have been found to regulate the recovery of neurological function after TBI. Our team found that circlphn3, which acts as a molecular sponge for mir-185-5p, regulates the expression of tight junction protein after TBI, thereby increasing the permeability of the blood-brain barrier (Cheng et al., 2022). Moreover, melatonin can reduce lipid peroxidation through circptpn14/mir-351-5p/5-lox signal transduction, thus reducing ferroptosis and endoplasmic reticulum (ER) stress in TBI (Wu et al., 2022). lncRNAs and circRNAs can regulate each other by competing for shared microRNA response elements in a regulatory model known as competing endogenous RNAs (ce-RNAs) (Salmena et al., 2011).

In this study, we used RNA sequencing (RNA-seq) to investigate differentially expressed protein-coding genes and ncRNAs in a mouse model of TBI treated with EPO to explore the mechanism by which EPO confers neuroprotection in the context of TBI.

Methods

Animals

Eighty-four C57BL/6J mice (male, 18–24 g, 6–8 weeks old, specific pathogen-free level) were purchased from the Laboratory Animal Center of Chongqing Medical University (license No. SCXK (Yu) 2018-0003). All animals were housed in an environment with a temperature of 22 ± 1°C, a relative humidity of 50 ± 1%, and a light/dark cycle of 12/12 hours (n = 5/cage). All animal procedures (including the mouse euthanasia procedure) were performed in compliance with the regulations and guidelines of Chongqing Medical University Institutional Animal Care (No. 2021-177, March 22, 2021) and conducted according to the Association for Assessment and Accreditation of Laboratory Animal Care and the Institutional Animal Care and Use Committee guidelines for the Care and Use of Laboratory Animals published by the National Institutes of Health.

Mice were randomly divided into three groups: sham, TBI, and TBI + EPO (n = 28/group). Samples were taken from three mice in each group for RNA sequencing, fifteen mice per group were subjected to the Morris water maze test, and samples from five mice per group were used for western blotting, real-time polymerase chain reaction, and immunofluorescence.

rhEPO treatment

rhEPO was purchased from 3SBio Inc (Cat# S19980074, Shenyang, China). rhEPO was administered as described in a previous study (Millet et al., 2016). Briefly, rhEPO was administered as a single dose of 5000 U/kg once daily (intraperitoneal injection) after TBI for 3 days. The first dose of EPO was administered 6 hours after TBI.

Controlled cortical impact model

Anesthesia was initiated with 5% isoflurane (Macklin, Shanghai, China), and spontaneous ventilation was maintained with 1.5% isoflurane in oxygen-enriched air (20% oxygen/80% air). A moderate controlled cortical impact (CCI) damage model was established using an electro-pneumatic impact device (Precision Systems and Instrumentation, Fairfax, VA, USA). Mice were placed on a cranial stereotaxic apparatus (David Kopf Instruments, Tujunga, CA, USA) with the head fixed to the anterior end, and body temperature was maintained at 36.8–37°C by the built-in heating bed. Under strict sterile conditions, a longitudinal incision was made along the midline above the mouse skull, and a craniotomy was performed over the left parietal cortex using a portable trephine (position of the center of the craniotomy coordinates relative to the bregma: posterior 2 mm, lateral 2 mm) (Xie et al., 2022), after which the bone flap was removed, keeping the dura intact. The impact parameters were as follows: the impact diameter was 3 mm, the impact velocity was 5.0 m/s, the impact depth was 1.5 mm, and the impact duration was 100 ms. Finally, cyanoacrylate tissue glue was applied to close the scalp. In the sham operation group, mice underwent the same surgical procedure in the absence of CCI. When the mice regained locomotor activity, they were returned to their cages.

A contralateral limb grasping test was performed 2 hours after surgery to determine whether TBI was successfully induced. The grasping test was performed as described in a previous study (Bertelli and Mira, 1995). In brief, mice were allowed to grasp the grid covering the cage. When the first signs of active finger flexion were noted, the mice were lifted by the tail with increasing firmness until they loosened their grip. If the mouse could not maintain its hold on the grid, indicating low grasping strength, it was considered to have TBI.

RNA-seq and bioinformatics analysis

Brain samples were obtained 3 days after surgery and snap frozen in liquid nitrogen for RNA sequencing (BGI, Shenzhen, China). RNA-sequencing was performed on nine mouse brain tissue samples (three samples per group). Library construction and sequencing were completed by BGI. Total RNA was extracted from tissues using Trizol reagent (Invitrogen, Carlsbad, CA, USA). Approximately 60 mg of brain tissue was added to a 2 mL tube, ground to a powder with liquid nitrogen, homogenized for 2 minutes, and finally rested horizontally for 5 minutes. Samples were then centrifuged at 12,000 × g for 5 minutes at 4°C, after which the pellet was discarded, and the supernatant was transferred to a new Eppendorf tube containing 0.3 mL of chloroform/isoamyl alcohol (24:1). The mixture was then vigorously shaken for 15 seconds and centrifuged at 12,000 × g for 10 minutes at 4°C. After centrifugation, the remaining RNA and upper aqueous phase were moved to another new tube containing an equal amount of isopropanol supernatant and centrifuged at 12,000 × g for 20 minutes at 4°C. The supernatant was discarded, and the RNA pellet was washed twice with 1 mL of 75% ethanol. To collect residual ethanol, the mixture was centrifuged at 12,000 × g for 3 minutes at 4°C; the RNA pellet was air dried for 5–10 minutes and mixed with 25–100 μL of diethyl pyrocarbonate to dissolve the RNA. Finally, the total RNA was quantified and its quality was assessed using a NanoDrop and an Agilent 2100 Bioanalyzer (Thermo Fisher Scientific, Waltham, MA, USA).

mRNA was purified from the total RNA using Oligo (dT)-linked magnetic beads (Thermo Fisher Scientific). Purified mRNA was fragmented in fragmentation buffer at 4°C. First-strand complementary DNA (cDNA) was generated using random hexamer reverse transcription, followed by synthesis of second-strand cDNA. Synthesis was terminated by adding ATailing Mix and RNA Index Adapters (Thermo Fisher Scientific) and incubating the mixture. Polymerase chain reaction (PCR) was performed to amplify the cDNA fragments obtained in the previous step, and the products were purified with Ampure XP Beads (Beckman Biotechnology, Beijing, China) and dissolved in ethidium bromide solution. The quality of the PCR products was assessed using an Agilent Technologies 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). The double-stranded PCR product obtained in the previous step was then heat-denatured and circularized using a splint oligonucleotide sequence to obtain the final library, which contained single-stranded circular DNA. The final library was amplified with phi29 (Thermo Fisher Scientific) to generate DNA nanospheres containing more than 300 single-molecule copies; the DNA nanospheres were loaded into patterned nanoarrays to generate 100 pair-ends on the BGIseq-500 platform (BGI) base reads. We then used SOAPnuke (v1.5.2, BGI, Shenzhen, China) (Li et al., 2008) to filter the sequencing data by 1) removing reads containing sequencing adapters; 2) removing low-quality bases (base quality ≤ 5) exceeding 20% of reads; (3) removing reads with a proportion of unknown bases (‘N’ bases) exceeding 5% to obtain clean reads; and 4) converting the remaining reads to FASTQ format. HISAT2 (v2.0.4, McKusick-Nathans Institute of Genetic Medicine, Johns Hopkins University School of Medicine, Baltimore, MD, USA) (Kim et al., 2015) was used to map clean reads to the reference genome. The clean reads were aligned to the reference coding gene set using Bowtie2 (v2.2.5, Center for Bioinformatics and Computational Biology, Institute for Advanced Computer Studies, University of Maryland, College Park, MD, USA) (Langmead and Salzberg, 2012), after which the expression levels of the genes were calculated using RSEM (v1.2.12, Department of Computer Sciences, University of Wisconsin-Madison, Madison, WI, USA) (Li and Dewey, 2011). Heatmap and Circos plots of the chromosome were generated using bioinformatics (http://www.bioinformatics.com.cn), a free online platform for data analysis and visualization.

Kyoto Encyclopedia of Genes and Genomes and Gene Ontology analysis

Eight differentially expressed lncRNAs (DElncRNAs) and eight differentially expressed circRNAs (DEcircRNAs) with the small inter-group difference and stable expression between TBI and TBI + EPO groups were selected. The RNAhybrid algorithm (https://bibiserv.cebitec.uni-bielefeld.de/rnahybrid) (Krüger and Rehmsmeier, 2006) was used to predict the target miRNAs of DEcircRNAs and DElncRNAs (minimum free energy < –25 kcal/mol, max bulge loop length and max internal loop length < 4). All differentially expressed miRNAs (DEmiRNAs) were uploaded to RNAhybrid, and the possible binding miRNAs were predicted based on the ce-RNA mechanism, after which overlapping interactions were identified. The online database miRDB (http://mirdb.org/mirdb/index.html) (Chen and Wang, 2020) was used to predict the target mRNAs of the DEmiRNAs in the previous step (the predicted target score > 80).

Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) analyses were performed on differentially expressed mRNAs (DEmRNAs), predicted target mRNAs of DElncRNAs, and DEcircRNAs. KEGG analysis (http://www.genome.jp/kegg/) was performed to investigate potentially essential pathways; the GO (www.geneontology.org) terms used in this study included biological process (BP), molecular function (MF), and cellular component (CC). The analysis results were visualized using http://www.bioinformatics.com.cn, a free online platform for data analysis and visualization, and the OmicShare tools (https://www.omicshare.com/tools), a free online platform for data analysis. All P-values were corrected for false discovery rate (FDR).

Real-time polymerase chain reaction

Three days after TBI, anesthesia was initiated with 5% isoflurane, spontaneous ventilation was maintained with 1.5% isoflurane in oxygen-enriched air (20% oxygen/80% air), and the mice were sacrificed by cutting their necks. Total RNA was extracted from the mouse brain tissues using an RNA extraction kit (BioTeck, Beijing, China) according to the manufacturer’s instructions. Samples were first treated with 3 U/mg RNase R (Epicenter, Madison, WI, USA) for 15 minutes at 37°C. The mRNAs, lncRNAs, and circRNAs were then reverse transcribed into cDNA using an RT Master Mix for real-time polymerase chain reaction (qPCR) Kit (MCE, Monmouth Junction, NJ, USA) following the manufacturer’s instructions. Next, the expression levels of selected DEmRNAs, DElncRNAs, and DEcircRNAs were determined by qPCR using SYBR® Green qPCR Master Mix (MCE). GAPDH was used as an endogenous reference transcript for mRNAs/lncRNAs/circRNAs. The 2^–ΔΔCt method (Poorghobadi et al., 2023) was used when measuring relative gene expression. The sequences of the primers used for qRT-PCR are shown in Table 1.

Table 1.

The primer sequences of qRT-PCR

Gene	Sequence (5’–3’)
Nos1	F: CGC TGC TAC AAC CTC GCT AC
	R: GGG TAT GGT AGG ACA CGA TGG
Npy	F: ATG CTA GGT AAC AAG CGA ATG G
	R: TAG TGT CGC AGA GCG GAG TAG
Rgs4	F: AGT CAG ATT CCT GCG AAC ACA
	R: AAA GCT GCC AGT CCA CAT TC
lncGm35161	F: GTT GGA GCA AGG AAT GAG AAT C
	R: TGA GGT CAC AAC CAC GAA ATG
lncGm46491	F: CAG GAC ACC AGA AGG GAA CG
	R: TTG CCA CCA GAG TCA TTT GC
lncAI662270	F: GAA TGT GGA AAT GTC GGT TGT AG
	R: TGG GTT GCC ATT GTT CTT GTC
mmu_circ_0005511	F: TGA AAC GCA ACA GCA CCG
	R: CAA TCA GGG CGA AGA CAC TC
mmu_circ_0005839	F: GTG CCT GAG TCT GTT TCT TGT TAC
	R: GGC TGG TCT ACA ACT CAC TAT GC
mmu_circ_0012957	F: AAG AAG GCT GGA CAG ATG GTG
	R: GCT ATT GGA GGA TGG GAA GTG
GAPDH (mouse)	F: AGG TCG GTG TGA ACG GAT TTG
	R: TGT AGA CCA TGT AGT TGA GGT CA

F: Forward; GAPDH: glyceraldehyde-3-phosphate dehydrogenase; ncAI662270: long non-coding AI662270; lncGm35161: long non-coding Gm35161; lncGm46491: long non-coding Gm46491; mmu_circ_0005511: Mus musculus circular 0005511; mmu_circ_0005839: Mus musculus circular 0005839; mmu_circ_0012957: Mus musculus circular 0012957; Nos1: nitric oxide synthase 1; Npy: neuropeptide Y; qRT-PCR: quantitative real-time polymerase chain reaction; R: reverse; Rgs4: regulator of G protein signaling 4.

Construction of ce-RNA networks

Stably expressed DElncRNAs and DEcircRNAs were identified. The downstream target DEmiRNAs of the DElncRNAs and target DEmiRNAs of the DEcircRNAs were predicted using the RNAhybrid algorithm (Krüger and Rehmsmeier, 2006). The top three DEmiRNAs with the smallest minimum free energy values were selected. RNAhybrid was used to predict the target DEmRNAs of the DEmiRNAs identified in the previous step. The top ten mRNAs with the smallest free energy values were selected. The identified DElncRNAs, DEcircRNAs, DEmiRNAs, and DEmRNAs were used to construct regulatory networks. Sequencing data were analyzed by DEGseq (Wang et al., 2010) for differential expression of non-coding genes and protein-coding genes, with Q value ≤ 0.01 and |log2foldchange| ≥ 0.5 set as the screening thresholds.

Morris water maze test

The Morris water maze (MWM) test was performed to assess the learning ability and spatial memory of the mice. The mice were trained on days 25–29 after TBI. The maze was a circular pool with a diameter of 120 cm and a depth of 50 cm (Zhenghua Bio-Instruments, Huaibei, China). An invisible platform was placed in the northeast quadrant of the pool, 10 cm in diameter and 2 cm below the water surface. In each experiment, mice were randomly placed on the pool wall in one of four orientations (northeast, southeast, northwest, and southwest) and released. The experiment was terminated when the mouse found and remained on the platform for at least 5 seconds, or if the mouse did not find the platform within 90 seconds. Mice that could not find the platform were directed to the platform and kept there for 15 seconds. After each experiment, mice were placed in dry cages. Each mouse performed four tests per day for 5 consecutive days, with a 5-minute interval between tests, and the platform position was kept constant. On the last day, the cognitive function of each mouse was assessed by placing it on the southwest wall of the pool, which no longer contained the platform. A video camera was placed above the maze to record the movements of each mouse, and trajectory analysis software (Zhenghua Bio-Instruments) was used to calculate the results, including swimming trajectory, dwell time, and path length.

Modified neurological severity score

The modified neurological severity score (mNSS) test, which covers motor (2 points), sensory (3 points), balance (6 points), and reflex and abnormal movement (4 points) dimensions, was used to evaluate the neurological function of the mice at 3 days after TBI. The maximum deficit score was 18 points, with a score of 1–6 points representing mild damage, 7–12 points representing moderate damage, and 13–18 points representing severe damage (Yu et al., 2019).

Western blot assay

Proteins were extracted from the ipsilateral cortex (3 days after TBI), resolved by sodium dodecyl sulphate-polyacrylamide gel electrophoresis, transferred to polyvinylidene fluoride membranes, and then incubated with primary antibodies (1:1000) overnight at 4°C. After 1 hour of incubation with horseradish peroxidase (HRP)-conjugated secondary antibodies (1:8000) at room temperature (25°C), bands were detected using an enhanced chemiluminescent substrate (Sigma-Aldrich, St. Louis, MO, USA, Cat#WBKLS0100). ImageJ software (1.8.0_112; National Institutes of Health, Bethesda, MD, USA) (Schneider et al., 2012) was used to analyze gray scale values. The ratio of the gray value of the target band to that of the internal control band was calculated to determine the relative protein expression level. The primary antibodies used were Wnt5a (mouse, ZEN BIO, Chengdu, China, Cat# 619919, RRID: AB_2927750) and Ephb6 (mouse, ZEN BIO, Cat# BES20492, RRID: AB_2927752). β-Actin was used as an internal reference (mouse, GeneTex, Alton Pkwy Irvine, CA, USA, Cat# GTX109639, RRID: AB_1949572). The secondary antibodies used were goat anti-rabbit IgG H&L (HRP; ZEN BIO, Cat# 511203, RRID: AB_2927753) and goat anti-mouse IgG H&L (HRP; ZEN BIO, Cat# 511103, RRID: AB_2893489).

Immunofluorescence

For immunofluorescence, the ipsilateral cortex was obtained 3 days after TBI and snap frozen in liquid nitrogen. The brain was treated with formalin for 24 hours and then sectioned. The resulting brain slices were incubated overnight at 4°C with appropriate primary antibodies (1:100), then incubated with the corresponding fluorochrome-conjugated secondary antibodies (1:400) for 1 hour at room temperature. Nuclei were visualized using a mounting medium containing 4′,6-diamidino-2-phenylindole (Beyotime, Shanghai, China). Brain slices were washed in phosphate-buffered saline (PBS) and then observed under a fluorescence microscope (Zeiss LSM800, Braunschweig, Germany). ImageJ software was used for further analysis. The antibodies used included an NeuN antibody (mouse, 1:200, Protein, Beijing, China, Cat# 66836-1-Ig, RRID: AB_1949572) and the other antibodies mentioned above. The secondary antibodies used were goat anti-rabbit IgG H&L (Alexa Fluor 488; ZEN BIO, Cat# 550037, RRID: AB_2927754) and goat anti-mouse IgG H&L (Alexa Fluor 594; ZEN BIO, Cat# 550042, RRID: AB_2927755). The gray scale values of the intensity and bands were analyzed using ImageJ software.

Statistical analysis

Statistical analyses were performed using GraphPad Prism software (version 9.1.1, GraphPad Software, San Diego, CA, USA, www.graphpad.com). A normality test was performed. The datasets were analyzed using one-way or two-way analysis of variance followed by Tukey’s post hoc test. A P-value < 0.05 was considered statistically significant. The evaluators were blinded to the group assignments.

Results

Transcriptomic analysis of the cerebral cortex in a mouse model of TBI treated with EPO

For the expression profiles of RNA alternations and to investigate the effect of EPO treatment on RNA expression after TBI, we performed RNA-sequencing analysis of total RNA isolated from cerebral cortex tissue from the sham, TBI, and TBI + EPO groups. The experimental design is shown in Figure 1A. The RNA-seq data were uploaded to NCBI (Accession: PRJNA862128). Mapping ratios for clean read counts and sequencing data are shown in Table 2. Compared with the TBI group, the TBI + EPO group exhibited 4065 differentially expressed RNAs in the cerebral cortex, including 1059 DEmRNAs (379 up-regulated and 680 down-regulated), 92 DEmiRNAs, 799 DElncRNAs (281 up-regulated and 518 down-regulated), and 2115 DEcircRNAs (825 up-regulated and 1290 down-regulated) (Figure 1B). Ninety RNAs (including 30 DEmRNAs, 30 DElncRNAs, and 30 DEcircRNAs) whose expression was relatively stable within each group were used to generate a heatmap (Figure 1C). All of the DEmRNAs, eight stably expressed DElncRNAs, and eight stably expressed DEcircRNAs were used in the subsequent analyses.

Figure 1.

Experimental flow chart and the RNA-seq results.

(A) Experimental design. (B) Histogram showing the number of up- and downregulated mRNAs, lncRNAs, and circRNAs. (C) Heatmap plotted by selecting 90 differentially expressed genes (including 30 DEmRNAs, 30 DElncRNAs, and 30 DEcircRNAs) with relatively stable expression within each group. TBI-1, TBI-2, and TBI-3 represent the three mice that sustained traumatic brain injury (TBI) with no treatment. TBI+EPO-1, TBI+EPO-2, and TBI+EPO-3 represent the three mice that sustained TBI and were treated with EPO. circRNA: Circular RNA; EPO: erythropoietin; GO: Gene Ontology; KEGG: Kyoto Encyclopedia of Genes and Genomes; lncRNA: long non-coding RNA; mNSS: modified neurological severity score; qRT-PCR: quantitative real-time polymerase chain reaction; TBI: traumatic brain injury.

Table 2.

Quality control of RNA sequencing

Sample	Total raw reads (M)	Total clean reads (M)	Total clean bases (Gb)	Clean reads Q20 (%)	Clean reads Q30 (%)	Clean reads ratio (%)
TBI-1	114.94	113.84	11.38	98.43	95.25	99.04
TBI-2	112.44	111.37	11.14	98.18	94.78	99.04
TBI-3	112.44	111.43	11.14	98.43	95.23	99.1
TBI+EPO-1	103.84	103.15	10.32	98.39	95.1	99.32
TBI+EPO-2	114.94	114.16	11.42	98.29	94.82	99.32
TBI+EPO-3	114.94	113.96	11.4	98.35	95	99.15

EPO: Erythropoietin; TBI: traumatic brain injury.

DEmRNA expression profile and functional prediction in a mouse model of TBI treated with EPO

For the results of DEmRNA, a volcano plot of the DEmRNAs is shown in Figure 2A, and the chromosomal locations of all of the DEmRNAs is shown in Figure 2B. We tested the expression of three randomly selected mRNAs by qRT-PCR and found that the expression patterns were consistent with our sequencing data (Figure 2C).

Figure 2.

DEmRNA expression profile and functional prediction.

(A) Volcano plot of differentially expressed mRNAs (DEmRNAs) between the TBI group and the TBI + EPO group. (B) Circos plot showing DEmRNA locations on mouse chromosomes. Track 1 (outside track) is the chromosome map of the mouse genome, and the black bars are chromosome cytobands. Track 2 (inner track) shows DEmRNA expression: blue dots represent down-regulated DEmRNAs, and red dots represent upregulated DEmRNAs. (C) Relative expression levels of the DEmRNAs Npy, Rgs4, and Nos-1 were detected by qRT-PCR. The results are presented as mean ± SEM (n = 5/group). ***P < 0.001, ****P < 0.0001, TBI vs. sham; #P < 0.05, ##P < 0.01, TBI + EPO vs. TBI (one-way analysis of variance followed by Tukey’s post hoc test). (D) GO annotation of the upregulated DEmRNAs was performed, covering the domains of biological process (BP), cellular components (CC), and molecular functions (MF). The red points on the circular track show the distribution of upregulated DEmRNAs. (E) KEGG analysis of the upregulated DEmRNAs. The yellow line represents the gene count, and the red bar represents the –log10 (P-value); the P-value was corrected by FDR. (F) GO annotation of downregulated DEmRNAs was performed, covering the domains of BP, CC, and MF. The blue points on the circular track showing the distribution of the downregulated DEmRNAs. (G) KEGG analysis of the downregulated DEmRNAs. The yellow line represents the gene count, the blue bar represents the –log10 (P-value), and the P-value was corrected by FDR. EPO: Erythropoietin; FDR: false discovery rate; GO: Gene Ontology; GRCm38: Genome Reference Consorium mouse genes (build 38); KEGG: Kyoto Encyclopedia of Genes and Genomes; TBI: traumatic brain injury.

To determine the functions of the above DEmRNAs, GO and KEGG analyses were performed. For the upregulated DEmRNAs, regulation of amine transport, learning, and memory were the most abundant terms in the GO BP category. Axon terminus and neuropeptide hormone activity were the most abundant terms in the GO CC and GO MF categories, respectively (Figure 2D). KEGG enrichment analysis further showed that the significantly enriched pathways were mainly neuroactive ligand-receptor interaction and ECM-receptor interaction (Figure 2E).

GO BP analysis of the downregulated DEmRNAs showed that the most abundant terms were neurotransmitter transport and negative regulation of neuron death. In the GO CC and GO MF categories, synaptic membrane and neurotransmitter receptor activity were the most abundant terms, respectively (Figure 2F). As for KEGG analysis, downregulated DEmRNAs neuroactive were highly enriched in ligand-receptor interaction, the calcium signaling pathway, the Wnt signaling pathway, and the Hippo signaling pathway (Figure 2G).

DElncRNA expression profile and functional prediction in a mouse model of TBI treated with EPO

For the results of DE lncRNA, lncRNAs act in the cytoplasm or nucleus to control neuronal cell fate and function (Ang et al., 2020). All lncRNAs with differential expression and eight stably expressed DElncRNAs are shown in a volcano plot in Figure 3A, and the qRT-PCR results confirmed the pattern of expression changes (Figure 3B). The chromosomal locations of the eight selected lncRNAs are shown in Figure 3C. Next, we searched for DEmRNAs adjacent to the genomic locations of the chosen DElncRNAs. We found the AI662270 and Slfn14, Gm46491 and Gm2956, and Gm39178 and Tenm3 were located close to each other (Figure 3D), suggesting that they may interact via cis-regulatory mechanisms (Yan et al., 2017). Notably, Gm46491 is located upstream of Gm2956 and oriented in the opposite direction. A previous study has reported that many lncRNA-mRNA pairs with this kind of relationship have correlated expression patterns and related functions (Guo et al., 2015). Therefore, we assessed the expression levels of these neighboring mRNAs and found a strong correlation between the expression levels of the selected lncRNA-mRNA pairs (Figure 3E). Thus, these lncRNA-mRNA pairs may be coregulated by the same signaling pathway.

Figure 3.

DElncRNA expression profile and functional prediction.

(A) Volcano plot of differentially expressed lncRNAs (DElncRNAs) between the TBI and TBI + EPO groups; the selected DElncRNAs with stable expression are labeled. (B) The relative expression levels of the DElncRNAs lncGm36161, lncAI662270, and lncGm46491, were detected by qRT-PCR. The results are presented as mean ± SEM (n = 5). ****P < 0.0001, TBI vs. sham; #P < 0.05, ##P < 0.01, ####P < 0.0001, TBI + EPO vs. TBI (one-way analysis of variance followed by Tukey’s post hoc test). (C) Circos plot showing the location of the DElncRNAs on mouse chromosomes. Track 1 (the outermost track) is the chromosome map of the mouse genome, and the black bars are chromosome cytobands. Track 2 (the larger inner track) labels the selected DElncRNAs. Track 3 (the smaller inner track) shows the expression of the selected DElncRNAs: blue dots represent downregulation, and red dots represent upregulation. (D) The genomic location of AI662270 and Slfn14, Gm46491 and Gm2956, and Gm39178 and Tenm3, the DElncRNA-DEmRNA pairs, as indicated by data retrieved from NCBI. Thick boxes indicated open reading frames (ORFs), and narrow boxes represent untranslated regions. The solid lines show intronic sequences, and arrows represent the transcriptional direction. (E) Scatter plot showing the correlation of the DElncRNA-DEmRNA pairs listed above. Each point shows the expression level in a cortex sample from a TBI or TBI + EPO mouse. (F, G) GO annotation and KEGG analysis of up- and downregulated DElncRNAs. The red and blue bars show the enrichment scores. BP: Biological process; CC: cellular component; chr: chromosome; DElncRNAs: differentially expressed long non-coding RNAs; EPO: erythropoietin; GO: Gene Ontology; GRCm38: Genome Reference Consorium mouse genes (build 38); KEGG: Kyoto Encyclopedia of Genes and Genomes; MF: molecular function; TBI: traumatic brain injury.

Next, we performed KEGG and GO analyses of the DElncRNAs. For the upregulated DElncRNAs, the most significant GO BP terms were axonogenesis and synapse organization. Synaptic membrane and cell adhesion molecule binding were the most abundant terms in the GO CC and GO MF categories, respectively. The KEGG results showed that axon guidance was the significantly enriched pathway (Figure 3F).

Interestingly, the most significant GO BP term for the downregulated DElncRNAs was also axonogenesis. Regulation of neurogenesis and synapse organization were also detected. Neuron to neuron synapse was the most abundant term in the GO CC category, while GTPase regulator activity was the most abundant term in the GO MF category. KEGG analysis showed that, as for the upregulated DElncRNAs, axon guidance was significantly enriched for the downregulated DElncRNAs (Figure 3G).

DEcircRNA expression profile and functional prediction in a mouse model of TBI treated with EPO

For the results of DEcircRNA, many studies have shown that circRNAs are important regulators of different biological processes, especially in the brain (Rybak-Wolf et al., 2015; You et al., 2015; Shao and Chen, 2016). Among the DEcircRNAs identified in our study, most (79.68%) were exonic, 7.46% were antisense, a few were intronic and sense-overlapping, and a small portion were intergenic (Figure 4A). All of the DEcircRNAs and eight selected DEcircRNAs are shown in a volcano plot in Figure 4B. We assessed the expression of three randomly selected DEcircRNAs by qPCR and confirmed the same pattern of expression changes as that shown in the RNA-sequencing data (Figure 4C). The chromosomal locations of the selected DEcircRNAs in the TBI and TBI + EPO groups are shown in Figure 4D.

Figure 4.

DEcircRNA expression profile and functional prediction.

(A) Annotations of all differentially expressed circRNAs (DEcircRNAs). (B) Volcano plot of DEcircRNAs between the TBI and TBI + EPO groups. The selected DEcircRNAs with stable expression are labeled. (C) Relative expression levels of the DEcircRNAs circ_0005839, circ_0005511, and circ_0012957, as detected by qRT-PCR. The results are presented as mean ± SEM (n = 5). ****P < 0.0001, TBI vs. sham; #P < 0.05, ###P < 0.001, TBI + EPO vs. TBI (one-way ANOVA followed by Tukey’s post hoc test). (D) Circos plot showing the DEcircRNAs locations on mouse chromosomes. Track 1 (the outermost track) is the chromosome map of the mice genome, and the black bars are chromosome cytobands. Track 2 (the larger inner track) shows the selected DEcircRNAs. Track 3 (the smaller inner track) shows the expression of the selected DEcircRNAs: blue dots represent downregulation, and red dots represent upregulation. (E) KEGG analysis of up- and down-regulated DEcircRNAs. The red bars show the enrichment scores of the upregulated DEcircRNAs, and the blue bars show the enrichment scores of the downregulated DEcircRNAs. (F, G) GO annotation of up- and downregulated DEcircRNAs was performed. Track 1 (the outermost track) shows the classification; outside the circle is a gene number ruler. Different colors represent different categories. Track 2 (the largest inner track) shows the p-values and background gene numbers of this classification; the longer the bar the more genes, and the smaller the p-values the redder the color. Track 3 (the larger inner track) shows the number of enriched genes. Track 4 (the smallest inner track) shows the RichFactor value of each classification, and each small bar on the auxiliary line represents 0.1. chr: Chromosome; DEmRNAs: differentially expressed mRNAs; EPO: erythropoietin; GO: Gene Ontology; GRCm38: Genome Reference Consorium mouse genes (build 38); KEGG: Kyoto Encyclopedia of Genes and Genomes; TBI: traumatic brain injury.

KEGG analysis of the upregulated DEcircRNAs showed that the mTOR signaling pathway and cAMP signaling pathway were enriched; for the downregulated DEcircRNAs, the most significantly enriched pathways were axon guidance and the MAPK signaling pathway (Figure 4E). The GO results showed that the most significant GO BP terms for the upregulated DEcircRNAs were dendrite development (GO:0016358) and vesicle-mediated transport in synapse (GO:0099003). Synaptic membrane (GO:0097060) and protein kinase regulator activity (GO:0019887) were the most abundant terms in the GO CC and GO MF categories, respectively (Figure 4F). The most significant GO BP terms for the downregulated DEcircRNAs were small GTPase-mediated signal transduction (GO:0007264) and axonogenesis (GO:0007409), and neuron to neuron synapse (GO:0098984) and GTPase regulator activity (GO:0030695) in the GO CC and GO MF categories, respectively (Figure 4G).

Constructing competing endogenous networks for the DElncRNAs and DEcircRNAs

Endogenous competition among ncRNAs has been gaining increasing attention in recent years. In this study, we constructed ce-RNA networks for the identified DElncRNAs and DEcircRNAs to explore the regulatory relationships among the differentially expressed ncRNAs in a mouse model of TBI treated with EPO. The ce-RNA networks of the upregulated and downregulated DElncRNAs are shown in Figure 5, and the DEcircRNA ce-RNA network is shown in Figure 6. Our ce-RNA networks demonstrated the potential regulatory axis that is activated by EPO treatment.

Figure 5.

Competing endogenous RNA networks of upregulated (A) and downregulated (B) DElncRNAs.

Red: lncRNAs; Green: miRNAs; Black: mRNAs. EPO: Erythropoietin; FC: fold change; mfe: minimum free energy; TBI: traumatic brain injury.

Figure 6.

Competing endogenous RNA networks of upregulated (A) and downregulated (B) DEcircRNAs.

Red: circRNAs; Green: miRNAs; Black: mRNAs. EPO: Erythropoietin; FC: fold change; mfe: minimum free energy; TBI: traumatic brain injury.

EPO promotes recovery of neurological function after TBI via axon-guided isopathic pathways

Impairment of spatial recognition memory is an important complication after TBI (Zeng et al., 2020). To test spatial recognition in our mouse model, we performed the MWM test. The representative paths of the mice are shown in Figure 7A. No significant changes in the location-navigation and spatial exploration tests were observed on the first day after TBI, with or without EPO treatment, whereas on days 25–29 after TBI, EPO treatment significantly reduced the escape latency (Figure 7B). In addition, the mNSS score was decreased after EPO administration (Figure 7C). These findings suggest that treatment with EPO may be improve spatial memory and neurological function in a mouse model of TBI.

Figure 7.

EPO promotes neurological recovery after TBI via axon-guided isopathic pathways.

(A) Representative paths of mice in MWM test on days 1–5 of training. Green line: the path of the mouse; black circle: invisible platform. (B) Line graph showing the escape time in the MWM test from day 1 to day 5 of training (n = 15 mice/group). (C) mNSS score. (D) Ephb6 and Wnt5a localization were detected by double immunofluorescence staining. The distance–gray value curves demonstrate the spatial consistency between NeuN and Ephb6 and Wnt5a. Scale bars: 50 μm, 10 μm (enlarged images). (F) Ephb6 and Wnt5a expression levels in brain tissue were detected by western blotting (n = 5/group). The results are presented as mean ± SEM. ****P < 0.0001, TBI vs. sham; #P < 0.05, ##P < 0.01, ###P < 0.001, ####P < 0.0001, TBI + EPO vs. TBI; (B: two-way analysis of variance followed by Tukey’s post hoc test; C, F: one-way analysis of variance followed by Tukey’s post hoc test). EPO: Erythropoietin; mNSS: modified neurological severity score; MWM: Morris water maze; TBI: traumatic brain injury.

Given that the term axon guidance was highly enriched in the KEGG results, and that improvement in neural function after TBI could be partly attributable to brain plasticity, we speculated that EPO treatment may promote axonal regeneration (Zhong et al., 2017). To explore whether EPO promotes the recovery of neurological function after TBI through the axon guidance pathway, we assessed the expression of Ephb6 and Wnt5a, two key components of the axon guidance pathway, by co-immunofluorescence and western blotting. Ephrins are axon guidance proteins, and Wnt proteins are also factors in axon development, and both protein families have been studied in the context of CNS injury (Biervert et al., 2001; Miyashita et al., 2009). Our results showed that Ephb6 and Wnt5a colocalized with NeuN in brain tissues, and that the increase in Ephb6 expression and decrease in Wnt5a expression induced by TBI were both reversed by treatment with EPO (Figure 7D and E). These changes in Ephb6 and Wnt5a expression before and after EPO treatment were confirmed by western blotting (Figure 7F). Thus, we conclude that EPO might promote recovery of neurological function after TBI through the axon guidance pathway.

Discussion

To the best of our knowledge, this is the first study to report transcriptomic profiling of a mouse model of TBI treated with EPO. The role of ncRNAs in the cerebral cortex in an EPO-treated mouse model of TBI has not been reported before. In the present study, 4065 differentially expressed RNAs were detected. KEGG and GO analysis showed that the axon guidance pathway, Wnt pathway, and MAPK pathway might be crucial to the beneficial effects of EPO seen in a mouse model of TBI. Additionally, ce-RNAs networks constructed for the identified DElncRNAs and DEcircRNAs suggested a possible regulatory axis among ncRNAs and protein-coding genes. Moreover, the positive effect of EPO on axon guidance was demonstrated by detecting the expression of Ephb6 and Wnt5a expression in response to treatment.

There were statistically significant changes in the expression of protein-coding genes in a mouse model of TBI after treatment with EPO. Amine transport was the most abundant GO BP term among the mRNAs that were upregulated after treatment. Among the mRNAs enriched in this term, we found that Rgs4 expression decreased after TBI and increased after EPO treatment. As a member of the Rgs family, Rgs4 has a crucial role in modulating synaptic signaling and brain plasticity (Gerber et al., 2016), which suggests that it may be a potential therapeutic target for TBI. Furthermore, KEGG analysis identified neuroactive ligand-receptor interaction as the main pathway enriched in differentially expressed mRNAs after EPO treatment. Neuroactive ligand-receptor interaction is an aspect of various disease processes. In this pathway, the Npy gene attracted our attention. In a rat model of epilepsy, exogenous EPO can upregulate Npy expression and exerts antiepileptic and neuroprotective effects (Kondo et al., 2009). Npy was also found to be downregulated after brain injury, while its expression increased after EPO treatment; therefore, EPO may promote the recovery of spatial memory in rats, enhance neurotrophy, and protect nerves by upregulating the expression of Npy and other genes.

The most abundant GO terms associated with the downregulated mRNAs were “negative regulation of neurotransmitter receptors” and “neuronal death.” Nos-1 is a component of both of these pathways. Multiple studies have reported that nitric oxide synthase expression is upregulated in neurons in a rat model of TBI (Miszczuk et al., 2016), where it mediates neuronal excitotoxicity by driving NO signaling, leading to cell death, suggesting that inhibiting Nos-1 would provide neuroprotection and aid in neurological function recovery (Mesenge et al., 1996); this is consistent with our sequencing results. In addition, the calcium signaling pathway was significantly enriched after EPO treatment. Calcium (Ca²⁺) acts as a universal second messenger, regulating important activities in all eukaryotic cells, participating in the transmission of depolarizing signals in the nervous system, and contributing to synaptic activity (Liraz-Zaltsman et al., 2021). Thus, we speculate that EPO may provide neuroprotection after TBI and aid functional recovery through the aforementioned pathways.

Although lncRNAs are involved in different aspects of cellular processes, no study to date has described changes in lncRNA expression in response to EPO treatment after TBI. Cis-regulation means refers to genes on the same chromosome that directly regulate the expression of neighboring genes (Yan et al., 2017). In this study, we found three DElncRNA-DEmRNA pairs, AI662270 and Slfn14, Gm46491 and Gm2956, and Gm39178 and Tenm3, that were located close together on chromosomes, which means that they might be cis-regulated DElncRNA-DEmRNA pairs that are vital to mediating the beneficial effects of EPO treatment. We predicted the downstream mRNAs of the identified DElncRNAs based on ce-RNA theory and performed GO and KEGG analyses using the target mRNAs of DElncRNAs. Interestingly, for both up- and downregulated DElncRNAs, the most abundant GO terms were axonogenesis and synapse organization. Axonogenesis involves the emergence of axons from the cell body following an initial polarization phase, elongation, and navigation to target regions guided by a series of environmental cues. This series of biological processes involves the GTPase family (Hall and Lalli, 2010), which is consistent with the results from our GO analysis. Synaptic organization and biogenesis were significantly affected, and related genes were upregulated, in a TBI model treated with mild hypothermia (Feng et al., 2010). Therefore, we speculated that EPO could improve neurological function after TBI by promoting synaptic and axonal growth. Furthermore, KEGG analysis indicated that differentially expressed lncRNAs in the cAMP signaling and mTOR signaling pathways were relatively abundant. The cAMP signaling pathway is a key regulator of neuronal regeneration, neuroplasticity, learning, and memory (Li et al., 2017a). After TBI, cAMP pathway signaling is reduced, which can lead to learning and memory impairment in patients (Titus et al., 2016). The mTOR pathway has an important role in various physiological functions of the nervous system, including regulation of neuronal cell growth, survival, axonal and dendritic development during differentiation, and synaptic plasticity (Jaworski and Sheng, 2006). The mTOR pathway may play different roles in cell death and neuroprotection after TBI: some studies have shown that mTOR inhibition prevents neuronal damage and death after TBI (Huang et al., 2018), while others have shown that increased mTOR signaling after injury promotes regeneration and functional recovery (Gao et al., 2020). Thus, there may be a delicate balance between inhibition and activation of the mTOR pathway to maintain neuronal survival while promoting axonal regeneration. Undoubtedly, this paradoxical effect requires further investigation and clarification in preclinical and clinical studies.

Our data suggest that DEcircRNAs may be a novel role by which EPO confers neuroprotection after TBI. Studies have shown that circRNAs have a key role in neuropsychiatric and neurodegenerative diseases (Wu et al., 2019; Wei et al., 2021). Most differentially expressed circRNAs identified by our RNA-seq analysis were exonic, which is consistent with a previous report (Li et al., 2021a). GO analysis showed that small GTPase-mediated signal transduction and axonogenesis were significantly enriched. The small GTPase is a strictly regulated molecular switch, which makes binary on/off decisions by controlling the loading (activation) of GTP and the hydrolysis of GTP into GDP (inactivation) (Reiner and Lundquist, 2018). In recent years, studies have begun to reveal the complexity of dendritic spine structure regulation, and the small GTPase protein family has come to be seen as a key regulator of structural plasticity, linking extracellular signals to the regulation of dendritic spines, which may affect cognitive performance (Woolfrey and Srivastava, 2016). KEGG analysis of both DElncRNAs and DEcircRNAs indicated that axon guidance and MAPK signaling pathways were significantly enriched. The axon guidance pathway primarily involves a series of guidance molecules that trigger axonal growth and termination, ultimately contributing to the development and maintenance of neural circuits (Curinga and Smith, 2008). Moreover, any defect or change in neural circuits caused by injury can lead to impaired function, so axon guidance pathway molecules have a key role throughout life. Zou et al. (2004, 2021) reported that guidance molecules and protrusion molecules are abundantly expressed during the repair process after spinal cord injury, and that axon guidance molecules have a key role in guiding the proper growth of axons. Several studies have also shown that the expression of many of these axon guidance molecules, especially semaphorins, ephrins, and Wnts, is increased after SCI (Giger et al., 2010), which is consistent with our analysis. In recent years, the MAPK pathway has been reported to be involved in regulating various cellular events mediated by growth factors, such as proliferation, senescence, differentiation, and apoptosis (Dent et al., 2003). Studies have also reported that the MAPK pathway acts as a downstream effector of the EPO signaling pathway for neuronal migration and localization during development (Constanthin et al., 2020). The closely associated PI3K-Akt pathway was also significantly enriched. Moreover, previous literature has reported that the PI3K-Akt signaling pathway can alleviate TBI-induced neuronal apoptosis (He et al., 2018). Digicaylioglu et al. (2004) found that EPO exerts neuroprotective effects by activating the PI3K/AKT pathway after TBI. Tóthová et al. (2021) reported that EPO-mediated induction of MAPK and PI3K pathway signaling has important effects on improving neuronal organization, reducing oxidative stress, and promoting neuronal stem cell differentiation into astrocytes. Therefore, EPO may exert neuroprotective and apoptosis-reducing effects through the aforementioned pathways after TBI.

Axon guidance was repeatedly enriched not only in the KEGG analysis but also in the GO analysis. As mentioned earlier, axon guidance is known to play an important role in the nervous system. Therefore, we analyzed the expression and localization of Wnt5a and Ephb6, key components in the axon guidance pathway. Ephb6 is an adrenergic tyrosine-protein kinase receptor that has an important role in axon guidance and synaptic plasticity (Fox and Kandpal, 2011). In contrast, Wnt5a induces a repulsive axonal response after spinal cord injury, limiting the capacity for axonal regeneration (Hollis and Zou, 2012). The exact role of Ephb6 and Wnt5a in brain injury after treatment with EPO is unclear. Our results showed that Ephb6 and Wnt5a expression are significantly altered by treatment with EPO after TBI, suggesting that EPO might promote the recovery of spatial memory after TBI through the axon guidance pathway.

Several limitations should be noticed in the present study. Firstly, only male mice were used in this study to avoid interference from the estrous cycle of female animals. There are clear monthly changes in estrogen and progesterone levels in female animals, whereas high testosterone levels in male animals remains stable over time. Therefore, we expected to obtain more consistent data by only using male animals. However, a recent study reported the availability of female mouse models (Prendergast et al., 2014), so our findings should be replicated in female mice. Moreover, recent studies have indicated that EPO treatment does not result in significant improvement in patients with brain trauma (Nichol et al., 2015; Liu et al., 2020). However, our study showed that EPO ameliorates deficits in neurological function, and a previous study also suggested that EPO has a positive effect on TBI (Gantner et al., 2018). There are several potential explanations for this apparent discrepancy: 1) there are species differences between rodents and humans; 2) the drug dose and the delivery route used in our experiments differed from those used in clinical trials; 3) patients with brain trauma are more complex than a simple mouse model, and the mechanism and extent of brain injury are also different. Therefore, it is worth studying the effects of EPO more extensively in the future to translate the result from animals to patients.

In summary, this study offers a comprehensive transcriptomic map and integrative analysis of the in potential pathways that mediate the neuroprotective effects of EPO treatment after TBI. Our findings provide new insights into and potential therapeutic targets for treating TBI, while the deeper exact mechanism is worth studying further.