Integrated analysis of porcine brain microRNA profiles following pediatric diffuse traumatic brain injury

Abstract

Mild traumatic brain injury (mTBI) is underdiagnosed and can lead to long-term symptoms in children. Currently, we lack diagnostic markers and effective therapies for pediatric mTBI. MicroRNAs (miRNAs) show promise for diagnosing and treating pediatric mTBIs due to their role in regulating key biological processes. Cyclosporine A (CsA) has also shown therapeutic effectiveness in modulating neuronal recovery and protection. We aimed at studying miRNA changes in the frontal lobe (FL) and hippocampus + amygdala (H&A) after a sagittal rapid non-impact head rotation (RNR) in 4 week old piglets (N = 50) at 1 day post-injury, 1 week post-injury, and following 1 day of 20 mg/kg/day Cyclosporine A (CsA) treatment compared to anesthesia-only shams. Interestingly, many miRNAs involved in disrupted neuronal and upregulated glial, epithelial, and endothelial functions were differentially expressed (DE-miRNAs) at 1 day post-injury and most returned to baseline by 1 week post-injury. However, miR-20a-3p, miR-10386, and miR-4331-3p were among the Top 10 DE-miRNAs at 1 day post-injury, and remained altered at 1 week. Upregulated neuroprotective miR-17-3p and miR-212 were also part of the Top 10 DE-miRNAs at 1 day post-injury, and their increases were correlated with decreases in axonal injury. Furthermore, to identify miRNAs that could serve as candidate diagnostic biomarkers of injury, we employed LASSO analysis and identified miRNAs miR-363, miR-15b, and miR-450c-3p as the best predictors of mTBI at 1 day post-injury. Lastly, WGCNA revealed the possible neuroprotective effects of CsA treatment in ameliorating neuronal, immune, stress, and vascular functions disrupted at 1 day post-injury. Overall, our data revealed key miRNAs that were differentially expressed at 1 day and 1 week post-injury, modulated by CsA treatment, and suggest that they may serve as diagnostic biomarkers of pediatric mTBI.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40478-026-02260-8.

Introduction

Mild traumatic brain injury (mTBI), also known as concussion, represents about 80% of TBIs [1]. Mild TBIs are mostly known to be caused by falls and motor vehicle accidents and involve rapid rotational forces that act on the brain, leading to diffuse axonal injuries [2, 3]. Due to their mild nature, these injuries are often undiagnosed, because of lack of visible injuries in MRI scans, delayed onset of symptoms and presentation of symptoms that often overlap with characteristic developmental changes in children [4, 5]. Children are known to be more affected by mTBI than adults, and the global incidence of pediatric mTBI continues to increase exponentially [6–8]. About 15–30% of children with mTBI develop long term symptoms that last for months to 25 years following mTBI [9, 10]. Unfortunately, there is still a paucity of diagnostic molecular markers and effective treatments for pediatric mTBI, underscoring the importance of increasing our understanding of molecular mechanisms in pediatric mTBI pathophysiology.

Many studies have already explored how several molecular signatures are changed following pediatric TBI and microRNAs (miRNAs) have become recent biomarker targets [11–15]. miRNAs are small non-coding molecules that play an important role in posttranscriptional gene regulation by modulating mRNA degradation and inhibiting protein translation [16]. They are known to be stable and abundant in brain tissues and biofluids, and they can cross the blood brain barrier (BBB) due to their small size [17]. Based on these characteristics, they have continuously been explored in adult animal and human TBI models [18–24]. These prior studies have shown that miRNAs regulate processes involved in neuronal degeneration and repair in TBI pathophysiology. Additionally, although very valuable in brain research, most of these studies have been done in rodents, which have very different brain sizes and neuroanatomy compared to humans [25]. Only a handful have studied miRNA changes in pediatric TBI and although many of these studies were conducted in children, they have all focused on salivary miRNA changes [12, 26–28]. To truly understand how miRNAs are changing post-mTBI and capture their therapeutic potential, it is important to study how they change specifically in the brain tissue, preferably in pediatric large animal rotational injury models that more closely mirror human neuroanatomy. In summary, our study is the first to explore miRNA alterations in a pediatric rotational TBI animal model.

Understanding how miRNAs are modulated provides a foundation for exploring therapeutic strategies that have shown promise in TBI, like Cyclosporine A (CsA). CsA is an FDA approved therapy that has been studied in mTBI and highlighted for its neuroprotective role [29–34]. We have previously reported that CsA accelerates neuronal recovery genes, improves mitochondria health, improves cerebral blood flow, decreases apoptosis, decreases pro-inflammatory markers, and decreases injured brain volume in 4-week-old piglets who received a sagittal rapid non-impact head rotation (RNR) or controlled cortical impact (CCI) [31, 32, 35, 36]. We showed that a 20 mg/kg/day dose of CsA administered via continuous intravenously infusion starting at 6 h post-injury and sustained for 1 day was the optimal dose for driving neuroprotection in 4 week old piglets experiencing either a focal or diffuse TBI [36]. Hence, using this same optimal dose we previously studied CsA’s role in modulating gene (mRNA) expression and intend to utilize the same samples to study its role in modulating miRNA expression. Although other studies have shown that CsA can induce mRNA and miRNAs expression to alter post-transcriptional regulation of genes indirectly [37–40], there are currently no studies that have examined CsA’s role in modulating miRNA expression following a rotational mTBI.

In this study, we hypothesized that miRNAs would be affected at 1 day and 1 week following a single RNR injury in both the frontal lobe (FL) and in the hippocampus + amygdala (H&A), and that these changes would be ameliorated by intravenous 20 mg/kg/day administration of CsA treatment initiated 6 h post-mTBI. Additionally, we aimed to correlate these changes with axonal injury, identifying subsets of miRNAs that best predict injury and CsA treatment, and connect these changes to differentially expressed genes (DEGs) from our previously published mRNA data to help add to our biological understanding of how miRNAs and genes drive mTBI pathophysiology. We observed miRNA alterations consistent with neuronal degeneration and protection at 1 day post-injury, identified the neuroprotective role of CsA in modulating neuronal, immune, and stress induced miRNAs post-mTBI, and identified key miRNAs that were the best predictors of mTBI at 1 day post-injury.

Materials and methods

Animals and experimental procedures for mTBI

Fifty-three (N = 53), 4 week old female Yorkshire swine (Sus scrofa) were purchased and housed in groups of 2–3 in a housing pen with adequate space to allow socialization and subjected to 12 h light–dark cycle. Animals were exposed to an anesthesia-only sham event (n = 18) or a sagittal rapid non-impact head rotation (RNR) and survived for 1 day post-injury (124.3 ± 1.7 rad/s, n = 10), 1 week post-injury (124.7 ± 3.3 rad/s, n = 15) or treated with 1 day of cyclosporine (CsA) treatment post-injury (124.8 ± 2.5 rad/s, n = 10) (Fig. 1). No differences were observed in Naïve (4 week olds, no additional survival time) and Sham samples with 1 day, 4 days, or 1 week survival, so they were combined into a single Sham group used for comparison to determine influence of injury, survival time and treatment (Supplementary Figure S1). Using an ambulatory wireless programmable infusion pump, the CsA treated animals received 20 mg/kg/day of CsA intravenously. The pump delivered a loading dose initiate at 6 h post-injury, with continuous infusion sustained for 1 day, yielding a total survival time of 30 h postinjury [36]. All experimental groups were randomized to their study group, anesthetized with isoflurane and oxygen via a gas mask, received a toe pinch reflex test, and then intubated and mechanically. Animal snouts were then secured on a padded snout clamp and sagittal RNR injuries were delivered using a pneumatic HYGE device (HYGE, Inc., Kittanning, PA, USA) [41]. At the appropriate post-injury timepoint, brains were harvested, formalin fixed and paraffin embedded (FFPE), and stored in a dry cool cabinet. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at University of Pennsylvania. Experiments were carried out in an AAALAC International (Association for Assessment and Accreditation of Laboratory Animal Care) accredited facility.

Fig. 1.

Schematic of study. miRNAs from all groups and regions were subjected to DESeq2, WGCNA, and LASSO analysis. All DE-miRNAs were correlated with percent axonal injury. Predicted gene targets were identified using miRDB and TargetScan databases. Predicted gene targets that were also identified as DEGs in our previous study were identified. DE-miRNA-mRNA interactions were constructed using miRDB, TargetScan and Cytoscape. GO overrepresentation analysis was conducted on all predicted gene targets. LASSO analysis was coupled with DESeq2 and WGCNA to identify miRNAs that could serve as robust candidate biomarkers. DE-miRNAs, differentially expressed micro-RNAs. GO, gene ontology.

Small RNA sequencing protocol

Frontal lobe (FL) and hippocampus + amygdala (H&A) FFPE blocks were selected and one 10 μm (H&A) and two 5 μm (FL) sections were obtained from regions shown in Supplementary Figure S2. Small RNA was isolated using the Mag-Bind^® FFPE RNA 96 Kit (Omega Bio-tek Inc, Norcross, GA, USA), by following the manufacturer’s protocol. RNA concentration and purity were measured using AccuBlue^® Broad Range RNA Quantitation assay (Biotium, Fremont, CA, USA) and Agilent high sensitivity Screentape (Agilent Technologies, Santa Clara, CA, USA). All small RNA samples passed quality control and were processed for library preparation. cDNA libraries were constructed from 1ug of total RNA using QIAseq miRNA Library Preparation Kit (Qiagen 331502, Hilden, Germany).

Libraries were measured by the Qubit 2.0 fluorometric quantitation (Thermo Fisher Scientific, Waltham, MA, USA), qPCR, and size distributions were determined by the Agilent Tapestation (Agilent Technologies, Santa Clara, CA, USA). Libraries were sequenced on NovaseqX plus platform (Illumina, San Diego, CA, USA) with data yield at 40 M reads in PE150 format.

Samples generated between 8.3 and 20.4 million reads. The raw reads were trimmed using Trimmomatic v0.39 to remove Illumina adapters and any reads with length less than 18 bp per usual miRdeep2 requirements [42, 43]. Then cleaned using cutadapt v.1.9.1 which removed the commonly known miRNA specific adapter sequence, AGATCGGAAG [43, 44]. FastQC v0.12.1 was used to derive quality check information and generate .html summaries both pre- and post- read trimming and alignment [45, 46]. This was to check for completeness, depth and read quality. Reads were mapped using the miRdeep2 mapping script within miRDeep2.pl against MiRBase and mature miRNAs for Sus scrofa (ss11.1) [43]. Resulting miRDeep2 predicted miRNAs and mature known miRNAs were used to map against with BowTie2 for quantification purposes. Alignments were sorted, filtered, and indexed to per-sample miRNA raw counts .bam files using samtools v1.3 [47]. Proprietary BASH scripts were used to combine all samples into a master raw miRNA counts table while denoting known and miRDeep2-predicted putative miRNA nomenclature.

Filtering, data cleaning, and normalization

Raw read counts generated above were analyzed for differential expression using RStudio (v4.2.2). miRNAs with less than 10 counts in 50 samples were filtered out to remove noise. Outliers were detected and removed using mahalanobis outlier detection and multivariate outlier detection package in R [48]. 274 miRNAs remained post filtering and 50 animals remained post outlier detection for data analysis, Sham (n = 15 – FL and 18 – H&A), Injured 1 day post-injury (n = 10 – FL and H&A), Injured 1 week post-injury (n = 15 – FL and 14 – H&A), CsA treated (n = 9 – FL and 10 – H&A) (Fig. 1). FL and H&A samples from the same animals were utilized. All miRNAs were normalized using the DESeq2 package, available on Bioconductor in R [49]. The raw and processed miRNA-seq data for all samples are available under accession number GSE312832 on the NCBI Gene Expression Omnibus (GEO).

Differential expression analysis

DESeq2 package was utilized on all 274 miRNAs to identify differentially expressed miRNAs (DE-miRNAs) that were distinct and common within groups and part of the Top 10 DE-miRNAs (Fig. 1). Wald’s test and Benjamini–Hochberg false discovery rate (FDR) procedures were utilized for hypothesis testing and multiple comparisons adjustments. DE-miRNAs were selected based on a threshold of FDR ≤ 0.05, |log2FC| ≥ 0.5.

Correlation analysis between DE-miRNAs and axonal injury

Spearman correlation was employed to examine the correlation of all DE-miRNAs with axonal injury. DE-miRNAs with significant (p ≤ 0.05) correlation coefficients (R) were reported within the study and plotted using the corrplot package in R.

Identification of predicted gene targets for all miRNAs

Gene targets for all miRNAs were identified using the miRDB v6.0 and TargetScan v8.0 databases [50, 51]. The identified predicted genes target common to the two databases were selected as the target genes for the miRNAs. Those also identified in our previous paper as differentially expressed genes (DEGs) were further analyzed [31]. miRNA-mRNA interactions for the DE-miRNAs and their predicted target DEGs [52] were generated using Cytoscape v3.10.3 software [52].

Co-expression gene network analysis of miRNAs

Weighted gene co-expression analysis (WGCNA) was performed on all 274 miRNAs to identify module eigengenes (MEs) of highly correlated miRNAs using the WGCNA package in R [53, 54] (Fig. 1). To identify our MEs and their expression scores, a network was constructed in a single block using the blockwiseModules() function. Within this function, the soft thresholding power β was 6 based on the network topology analysis (Supplementary Figure S3), the minimum number of miRNAs in the module was 10, the module detection sensitivity was 4 (deepSplit = 4), the cut height for merging of modules was 0.10 (mergeCutHeight = 0.10), the reassignment threshold was 0.05, and a “bicor” correlation type and a “mean” TOM denominator were chosen. miRNAs with the highest expression scores within each ME were selected as the Top 5 hub miRNAs. ME expression scores were compared between all groups using a Wilcox test and Bonferroni adjustment were applied (p ≤ 0.05).

Selection of good predictor miRNAs for the injured and treated groups

Least Absolute Shrinkage and Selection Operator (LASSO) regression analysis was performed using the glmnet package in R [55] with a “multinomial” family argument and 10 fold cross-validation (nfolds = 10) to identify the optimal (best) regularization parameter (lambda.1se, λ) in the FL and H&A (Fig. 1). The best λ value was then applied to construct a final model, from which miRNAs with non-zero coefficients were identified in the 1 day post-injury and CsA treated groups in both regions. To ensure robust model performance, a repeated nested stratified cross validation was implemented. The outer loop contained stratified 10-fold split to ensure each fold had the same group proportions as the original dataset. Within each outer training set, an inner stratified 10-fold split cross validation was conducted to optimize the model hyperparameters (i.e. the LASSO λ). This λ was utilized to fit the model to the train dataset and then probabilities were predicted using the test dataset. This procedure was repeated 5 times to reduce variability and model performance was averaged across all outer folds. A receiver operating characteristic (ROC) analysis was conducted using the pROC package in R [56] and the area under the curve (AUC) was quantified for the predictions. The miRNAs identified by LASSO, identified as DE-miRNAs or Top 10 DE-miRNAs by DESeq2 or Top 5 hub miRNAs by WGCNA were highlighted as the best predictors of injury and CsA treatment.

Overrepresented GO analysis

Gene ontology (GO) overrepresentation analysis was conducted using PANTHER (Protein Analysis Through Evolutionary Relationships, v19.0) to identify biological process (GO terms) overrepresented in each group [57, 58]. The appropriate background was generated using the predicted gene targets of all 274 miRNAs or the default sus scrofa whole-genome list. Fisher’s exact test, with the Benjamini–Hochberg false discovery rate (FDR ≤ 0.05) correction was employed to identify statistically significant overrepresented GO terms. Unique GO terms with the highest fold enrichment scores were highlighted in the results.

Plot generation

All venn diagrams, bar plots, and box plots were generated using the ggplot2 and ggpubr packages in R. Heatmaps were generated using the heatmap3 and ComplexHeatmap packages in R. Volcano plots were generated using the EnhancedVolcano package in R.

Results

MicroRNAs are altered at 1 day post-injury and most return to sham level by 1 week post-injury

To test our hypothesis, we examined microRNA (miRNA) expression at 1 day and 1 week post-injury. Bulk RNA-sequencing was performed to identify a total of 274 miRNAs from the frontal lobe (FL) and hippocampus + amygdala (H&A) regions of porcine brain following mild TBI. Hierarchical clustering of the miRNAs revealed three clusters indicating the effect of injury, region, and/or time on miRNA expression (Fig. 2a). Cluster I consisted of miRNAs that were different between FL and H&A in all 4 groups. Cluster II included miRNAs upregulated at 1 day post-injury and return to baseline at 1 week post-injury in both regions. Cluster III was representative of miRNAs highly upregulated in the FL at 1 day post-injury and going back down in the FL at 1 week post-injury. DESeq2 analysis revealed differentially expressed miRNAs (DE-miRNAs) between our groups. Surprisingly, there were more DE-miRNAs at 1 day post-injury compared to 1 week post-injury or 1 day + CsA treated groups (Fig. 2b, c). By 1 week post-injury, there were only 1 downregulated (miR-10386) and 1 upregulated (miR-20a-3p) DE-miRNAs in the FL, and 2 downregulated (miR-10386, miR-4331-3p) and 0 upregulated DE-miRNAs in the H&A, compared to Sham (Fig. 2d). Additionally, in both regions there were no DE-miRNAs between 1 day post-injury versus CSA treated groups, indicative that CsA does not exert acute effects on miRNAs.

Fig. 2.

MicroRNAs are altered at 1-day post-injury in both regions, and most return to sham level by 1 week post-injury. a Hierarchal clustering of all miRNAs across all injury groups in the FL (light green) and H&A (dark green). Each row represents a miRNA, and each column represents an animal subject (rows were z-scored). b DE-miRNAs in the injured 1 day post-injury group in the FL and c H&A, consecutively. d DE-miRNAs in the injured 1 week post-injury group in the FL and H&A. DE-miRNAs were selected based on threshold FDR ≤ 0.05, |log2FC| ≥ 0.5. e Downregulated DE-miRNAs shared and distinct between FL and H&A and their overrepresented GO terms. f Upregulated DE-miRNAs shared and distinct between FL and H&A and their overrepresented GO terms. DE-miRNAs, differentially expressed miRNAs. FDR, false discovery rate. Log2FC, log2 fold change. FL, frontal lobe. H&A, hippocampus + amygdala. GO, gene ontology

Focusing on DE-miRNAs at 1 day post-injury, many were observed in the FL and H&A (Supplementary Table S1). Surprisingly, only few were distinct between both regions (Fig. 2e, f). We identified predicted gene targets of these DE-miRNAs (Methods), then applied GO over-representation test on these predicted gene targets in distinct to each region and shared (FL + H&A) between each region. Significant GO terms were only identified for predicted gene targets of the FL only and shared DE-miRNAs. Among the downregulated DE-miRNAs (Fig. 2e), in the FL-only group, significant GO terms were involved in cellular stress response processes, glucose metabolic processes, and potassium (K+) ion transport. The shared downregulated DE-miRNAs had GO terms that may mainly be associated with neurogenesis, synaptic signaling, and actin cytoskeletal organization processes. Moving to the upregulated DE-miRNAs (Fig. 2f), the overrepresented functions in the FL only section were involved in neuronal functions like chemical synaptic transmission and neurogenesis, immune and vascular functions, glial cell migration, endothelial cell migration as well as terms related to developmental functions, like epithelial cell migration. Similarly, the shared upregulated DE-miRNAs had predicted gene targets with an overrepresentation of GO terms that may be associated with neuronal and developmental functions like neural crest differentiation, chemical synaptic transmission, axonogenesis, and epithelial cell migration.

These findings indicate that many DE-miRNAs are dysregulated by 1 day post-injury and most return to baseline by 1 week post-injury. Additionally, predicted gene targets of DE-miRNAs in both regions are involved in downregulated neuronal signaling and growth processes, as well as upregulated neuronal signaling and developmental processes. However, in the FL, predicted gene targets were involved in downregulated stress response, metabolic, and signaling process, as well as upregulated neuronal signaling, vascular, immune, and developmental processes.

Top 10 upregulated DE-miRNAs in both regions at 1 day post TBI are mainly involved in neuroprotective functions

Next, because the magnitude of miRNA changes is often related to the most biologically significant regulators of pathological change, we sort to identify our big changes based on Log2FC. We focused on the Top 10 downregulated and upregulated DEGs in both regions at 1 day post-injury. Top 10 downregulated DE-miRNAs in the FL were miR-7173-3p, miR-139-3p, miR-423-5p, miR-193a-5p, miR-885-3p, miR-4331-3p, miR-92b-5p, miR-10391, miR-4332, and miR-10386 (Fig. 3a). Their downregulation have been identified to be involved mostly in degenerative functions (Fig. 3b) [59–64]. Additionally, miR-4331-3p and miR-10386 were also DE-miRNAs at 1 week post-injury in the FL and/or H&A. The Top 10 upregulated DE-miRNAs in the FL were miR-20a-3p, miR-451, miR-130b-5p, miR-148b-3p, miR-212, miR-27a, miR-135, miR-7-3p, miR-363, and miR-17-3p (Fig. 3a). Their upregulation may be mainly associated with neuroprotection [65–73], and some harmful functions [74, 75] (Fig. 3b). miR-20a-3p was also a DE-miRNA at 1 week post-injury in the FL. Most of our Top 10 down- and upregulated DE-miRNAs in the FL may have been associated with degenerative and protective functions, respectively.

Fig. 3.

Top 10 DE-miRNAs in both regions at 1 day post-injury and their corresponding DEGs targets. a Top 10 downregulated (blue) and upregulated (red) DE-miRNAs in the FL. b Functions of Top 10 DE-miRNAs in the FL. c Top 10 downregulated (blue) and upregulated (red) DE-miRNAs in the H&A. d Functions of Top 10 DE-miRNAs in the H&A. e Common Top 10 DE-miRNAs in both regions. f miRNA-mRNA interaction for the Top 10 common downregulated DE-miRNAs in both regions and their predicted DEG targets. g miRNA-mRNA interaction for the Top 10 common upregulated DE-miRNAs in both regions and their predicted DEG targets. DE-miRNAs, differentially expressed miRNAs. DEG, differentially expressed genes. FL, frontal lobe. H&A, hippocampus + amygdala

By focusing on the H&A, the Top 10 downregulated DE-miRNAs were miR-10a-5p, miR-769-3p, miR-193a-5p, miR-92b-5p, miR-574-5p, miR-885-3p, miR-10391, miR-4331-3p, miR-4332, miR-10,386 (Fig. 3c). Many of these DE-miRNAs were common between FL and H&A, they were miR-10,386, miR-4332, miR-4331-3p, miR-10,391, miR-885-3p, miR-92b-5p, and miR-193a-5p (Fig. 3e). The downregulation of those distinct to the H&A are related to distress i.e. elevated inflammation, cell death, mitochondrial metabolic dysregulation, inhibited neurogenesis and TBI (Fig. 3d) [76–78]. The Top 10 upregulated DE-miRNAs were miR-451, miR-130b-5p, miR-212, miR-1271-5p, miR-363, miR-374a-5p, miR-204, miR-27a, miR-17-3p, and miR-9-1 (Fig. 3c). Similarly, a good number of these DE-miRNAs were common in both regions, they were miR-451, miR-130b-5p, miR-212, miR-363, miR-27a, and miR-17-3p (Fig. 3e). The upregulation of those distinct to H&A are related to some protective and neuronal growth functions (Fig. 3d) [79–81]. Similar to the FL, the upregulated Top 10 DE-miRNAs in the H&A may have mostly been associated with protective functions.

We then identified predicted gene targets of the common Top 10 DE-miRNAs in both regions in Fig. 3e and then focused on those that were part of the differentially expressed genes (DEGs) from our previously published paper [31]. For the Top 10 downregulated DE-miRNA, only two had predicted gene targets that were DEGs from our previously published paper, miR-885-3p and miR-193a-5p (Fig. 3f). CLU which was a predicted DEG target for miR-885-3p, was also a Top 10 upregulated DEG in both regions from our previously published paper [31]. It is involved in amyloid β (Aβ) clearance [82]. Within the upregulated DE-miRNAs, miR-212, miR-17-3p, miR-130b-5p, miR-363, and miR-27a had predicted gene targets that were part of the DEGs from our previously published paper (Fig. 3g). Interestingly, four Top 10 upregulated DEGs in the FL and/or H&A were predicted gene target for three of our Top 10 upregulated DE-miRNAs. TMSB10, MICU3, ID4 were predicted gene targets for miR-27a, while PGAM1 was a gene target for miR-363, and miR-17-3p. These DEGs are involved in actin remodeling, mitochondrial maintenance, neurodevelopment, and neuroprotective functions, respectively [83–86]. By focusing on those DEGs with multiple Top 10 DE-miRNA targets, GABBR2, known for its role in synaptic stability, was a predicted DEG target for miR-193a-5p, miR-17-3p, and miR-130b-5p [87]. MAP1B known for its role neuronal morphogenesis was a predicted DEG target for miR-363, miR-17-3p, and miR-27a [88]. Lastly, downregulated ENDOU known to be correlated with increased M2 macrophage infiltration was a predicted DEG target for miR-27a and miR-130b-5p [89]. Overall, the DEGs from our previous study were predicted gene targets of the DE-miRNAs in this study and some were involved in neuronal development and protection.

Collectively, many of the Top 10 DE-miRNAs were shared between both regions. The Top 10 downregulated DE-miRNAs were involved in both protective and harmful functions, while the Top 10 upregulated DE-miRNAs were mainly protective. Additionally, the Top 10 DEGs and DEGs targeted by multiple Top 10 DE-miRNAs were specifically involved in increased neuronal growth, stability, and protection.

Axonal injury correlation with down- and up-regulated DE-miRNAs

Next, to understand more about our data, we sought to identify DE-miRNAs at 1 day post-injury correlated with increases or decreases in the volume of acute axonal injury. We previously reported the volume of axonal injury data in these animals 1 day and 1 week after TBI, computed by dividing the total outlined area of injured white matter (β-amyloid precursor protein, β-APP) in every coronal slice in the brain, expressed as a fraction of total brain area (Supplementary Figure S4, Supplementary Table S2) [31]. Spearman correlation (correlation coefficient & p ≤ 0.05) was utilized to identify DE-miRNAs of interest from both regions and timepoints. We focused on our DE-miRNA distributions from Fig. 2e and f. Among downregulated DE-miRNAs (Fig. 4a), miR-22-3p was the only one specific to the FL that was positively correlated with axonal injury. Downregulation of miR-22-3p is associated with increased Aβ deposition in Alzheimer’s disease (AD) [90]. A preclinical study in AD mice showed that increases in miR-22-3p could improve apoptosis and reduce Aβ deposition by acting on the Sox9 through the NF- κB pathway [90]. miR-1306-3p and miR-490-5p from the H&A were negatively and positively correlated with axonal injury, respectively. The inhibition of miR-1306-3p has been associated with reduced synaptic transmission in sensory neurons and amyotrophic lateral sclerosis (ALS) [91, 92]. miR-490-5p is involved in cell cycle arrest [93]. A study highlighted its potential therapeutic role in osteoarthritis by showing how its inhibition facilitated chondrogenic differentiation and protected the chondrocyte phenotype [94]. miR-206 was downregulated in both regions, but only its expression in the FL was positively correlated with axonal injury. It’s downregulation has been associated with increased brain-derived neurotrophic factor (BDNF), Nissl bodies, Egr-1 and PSD-95 which are crucial for neuron growth, survival and synaptic signaling [95]. Overall, the downregulated DE-miRNAs correlated with axonal injury were involved with a mix of neurodegenerative and neuroprotective functions.

Fig. 4.

DE-miRNAs correlated with axonal injury and miR-20a-3p expression at 1 week post-injury. a Downregulated DE-miRNAs in the FL and H&A correlated with axonal injury. b Upregulated DE-miRNAs in the FL and H&A correlated with axonal injury. Top 10 DE-miRNAs are indicated with purple box. c miR-20a-3p expression in all groups. Significance was achieved using the Wilcox test, with the Bonferroni adjustment (p ≤ 0.05). d miRNA-mRNA interaction for the predicted DEG targets of miR-20a-3p. Asterisks correspond to, *p ≤ 0.0, ** p ≤ 0.01. DE-miRNAs, differentially expressed miRNAs. DEG, differentially expressed genes. FL, frontal lobe. H&A, hippocampus + amygdala

In the upregulated DE-miRNAs, they were all negatively correlated with axonal injury (Fig. 4b). miR-454 is part of the DE-miRNAs specific to the FL region, and its elevation has been shown to alleviate brain damage in ischemic stroke [96]. miR-181a is part of the DE-miRNAs specific to the H&A region, and its overexpression is involved in neuronal degeneration and neuroinflammation [97]. A preclinical study in 3xTg-AD mice, showed that its inhibition alleviated memory deficits and increased GluA2 and GluA1 levels [98]. Of all the DE-miRNAs in both regions, miR-374b-5p was negatively correlated with axonal injury in both regions. It’s increase is protective and have been linked to increases in neuronal viability and attenuated neuroinflammation [99]. While miR-107, miR-137, miR-17-3p, miR-212, and miR-369 expressions in the FL were negatively correlated with axonal injury. miR-17-3p and miR-212 were part of Top 10 upregulated DEGs in both regions and their upregulation are involved in neuroprotection [67, 73]. miR-107, miR-137, and miR-369 are also neuroprotective in AD and ischemic stroke [100–102]. miR-107 has also been implicated in promoting angiogenesis during hypoxia [103]. Lastly, miR-105-1, miR-22-5p, and miR-574-3p expressions in the H&A only were negatively correlated with axonal injury. They have also been involved in neuroprotection [104, 105]. Overall, many of these upregulated DE-miRNAs may be associated with neuroprotection and correlated with a reduced axonal injury.

None of the DE-miRNAs at 1 week post-injury were correlated with axonal injury (miR-20a-3p, miR-10386, miR-4331-3p). Interestingly, they were all part of the Top 10 downregulated or upregulated DE-miRNAs at 1 day post-injury and they remained dysregulated at 1 week post-injury. miR-20a-3p is particularly interesting because it is the only one of the 3 that is homologous to human miRNAs. It has also been identified to be protective in stroke studies [65]. It was up-regulated at 1 day post-injury with or without CsA treatment (Fig. 4c). We further explored the genes targeted by miR-20a-3p, specifically those that were identified as DEGs at 1 day and 1 week post-injury in our previously published paper [31]. Interestingly, more up- and down-regulated predicted DEG targets were present at 1 week compared to 1 day post-injury (Fig. 4d). The DEG targets at 1 day post-injury were involved in protective functions like DNA repair [106, 107], neuronal differentiation [108], neuron survival [109], glutamate homeostasis [110], and degenerative functions like altered development [111] and synaptic disruption [112]. Similarly, the predicted DEG targets appearing and persisting until 1 week postinjury, were also involved in protective functions like decrease in reactive oxygen species (ROS) [113], Aβ clearance [114], microtubule stabilization [115], mitophagy [116], neuron survival [117, 118], stress granule formation [119], neuronal differentiation [120], neuroprotection [121], and degenerative functions like blood brain barrier (BBB) damage [122], endothelial cell injury [123], neuronal apoptosis after neuroinflammation [124], and increased APP expression [125]. Overall, more predicted DEG targets of miR-20a-3p were present at 1 week post-injury, they are implicated in protective and degenerative functions, and the presence of these DEGs implies that they should be further explored in TBI.

WGCNA revealed key miRNAs displaying effect of injury and CsA treatment

We next performed WGCNA on 274 miRNAs to identify co-expression patterns in TBI responses (1 day, 1 week, and CsA treatment), which revealed 6 functional module eigengenes (MEs) as shown in the cluster dendrogram (Fig. 5a). Although DESeq2 revealed DE-miRNAs at 1 day and 1 week post-injury in both regions, there were no individual DE-miRNAs identified between 1 day post-injury versus CSA treated groups in either region. Our subsequent WGCNA revealed networks of miRNAs modulated by CsA treatment, and we decided to explore these integrated networks. To understand how these MEs changed across injury groups, we visualized their module eigengenes (Fig. 5b). Interestingly, in both regions (FL and H&A), the expression scores of ME1, ME2, ME3, ME5, and ME6 were changed at 1 day post-injury, and returned to Sham levels by 1-week post-injury (Supplementary Figure S3). We then focused on the Top 5 hub miRNAs for these MEs and explored the GO terms of their predicted gene targets using the GO overrepresentation analysis (Fig. 5c, Supplementary Table S3). ME1 displayed functions represented by downregulated neuronal development and synaptic signaling. While ME2, ME3, ME5, and ME6 represented an upregulation of immune responses, stress responses, endothelial cell development, growth, signaling and autophagy. Interestingly, some of the Top 5 hub miRNAs that were bold had been previously identified as part of the Top 10 DE-miRNAs or correlated with axonal injury Starting with ME1, miR-423-5p and miR-139-3p had been identified as part of the Top 10 downregulated DE-miRNAs in the FL. Within ME3, ME5, and ME6 upregulated miR-107, miR-105-1, and miR-137 were identified to be negatively correlated with axonal injury in the FL and H&A, such that increases may be associated with decreases in axonal injury. Overall, the MEs changing at 1 day post-injury were involved in downregulated neuronal functions and upregulated immune, stress, protective, signaling, and growth functions and some were correlated with decreased axonal injury and part of the Top 10 DE-miRNAs in the FL.

Fig. 5.

WGCNA analysis reveals co-expression of miRNAs ameliorated by CSA treatment. a Hierarchical clustering tree of 274 miRNAs depicting the 6 module eigengenes identified. Each ME has a distinct color representing co-expressed miRNAs. b Heatmap representing the expression scores all MEs in all groups. Each row represents an animal, and each column represents a ME (rows were z-scored). c The top 5 hub miRNAs and the overrepresented GO terms of their predicted gene targets. Bolded miRNAs are those present in Top 10 DE-miRNAs or correlated with axonal injury. d Expression scores of the MEs displaying an effect of CsA treatment in the H&A (ME1 and ME3). Significance was achieved using the Wilcox test, with the Bonferroni adjustment (p ≤ 0.05). e Overrepresented GO terms of ME1. Red to blue indicates high to low FDR p-values. f Overrepresented GO terms of ME3. Red to blue indicates high to low FDR p-values. FL, frontal lobe. H&A, hippocampus + amygdala. CsA, cyclosporine A. ME, module eigengene

We next focused on MEs affected by CsA treatment, finding effects in the H&A region only for ME1 and ME3 (Fig. 5d). ME1 represented those miRNAs downregulated at 1 day post-injury, but upregulated to baseline by CsA. While ME3 represented those upregulated at 1 day post-injury but downregulated to baseline by CsA. Overrepresented GO terms of the predicted gene targets of Top 5 hub miRNAs, revealed that miRNAs upregulated by CsA in ME1 were involved specifically in neuronal and axon development, as well as fibroblast proliferation and synaptic transmission (Fig. 5e). ME3, which represented those miRNAs downregulated by CsA, revealed a GO term, the calcineurin–NFAT signaling pathway, known to be inhibited by CsA (Fig. 5f) [126]. Additionally, ME3 GO terms involved in BBB maintenance, synaptic plasticity, immune signaling, endothelial cell proliferation, and epithelial cell proliferation were also overrepresented. Overall, CsA treatment resulted in an upregulation of miRNAs involved in neuronal and axon development, as well as a downregulation of immune signaling, BBB maintenance, endothelial proliferation, and calcineurin–NFAT signaling.

LASSO identified key miRNAs that can discriminate injury and CsA treated groups

Because miRNAs are gaining recognition for their diagnostic and therapeutic potential in brain injury [127–129], we sought to identify miRNAs that were potential diagnostic predictors of injury and CsA treatment. To do so, we performed LASSO analysis on all 274 miRNAs. A 10-fold cross validation was performed to determine the best lambda value (λ) for FL and H&A (Fig. 6a). The cross validation produced two special lambda values (λ), lambda.min and lambda.1se which are known to be regularization parameters. The lambda.1se was chosen as the best λ value because it reduces overfitting, thereby making it the best for feature selection. A multinomial LASSO model was then fit using the selected lambda.1se (λ) value for FL and H&A. The LASSO model identified miRNAs with non-zero coefficients at 1 day post-injury and miRNAs with non-zero coefficient following CsA treatment in both regions. These miRNAs were considered the best predictors for the 1 day post injury and CsA groups. To understand the performance of the LASSO model, we measured the prediction performance of the model. To do this, we used a repeated nested stratified cross validation using a 10-fold split for the outer and inner loop, and 5 repeats to generate prediction probabilities [130]. These predictions probabilities were averaged and assessed using a receiver operating characteristic (ROC) analysis (Fig. 6b). The area under the curve (AUC) values for the miRNA predictions for the 1 day post-injury group and for the CsA treated group in both regions were between 0.746 and 0.877 signifying that the LASSO model used to identify the miRNAs had a high predictive power.

Fig. 6.

LASSO predicts key miRNAs that are good predictors of injury at 1 day and CsA treatment. a LASSO cross validation plot depicting the best lambda (λ) values. b ROC curves for LASSO predictions for the Injured 1 day post injury and CsA treated groups. c miRNA selected by LASSO for the Injured 1 day post-injury and CsA treated groups in the FL and H&A. FL, frontal lobe. H&A, hippocampus + amygdala. CsA, cyclosporine A

To identify the miRNA predictions that had biological relevance and could serve as robust candidate biomarkers for injury and CsA treatment, we highlighted the LASSO miRNA predictions that were also part of the DE-miRNAs or Top 10 DE-miRNAs from the DESeq2 analysis or part of the Top 5 hub miRNAs from the WGCNA analysis. These were the miRNAs selected as the best predictors of injury and CsA treatment (Fig. 6c). Beginning with the FL, at 1 day post-injury, let-7d-5p, miR-15b, and miR-363 were identified by LASSO at 1 day post-injury. Interestingly, miR-363 known for its neuroprotective role was also part of the Top 10 upregulated DE-miRNAs in both regions (Fig. 3b, c). Whilst miR-15b known for regulating apoptosis was part of the upregulated DE-miRNAs in both regions [131]. In the CsA treated group, miR-182, miR-335, miR-374a-3p, miR-451 were identified by LASSO in the FL. miR-335 and miR-451 were part of the Top 5 hub miRNAs in the ME6 (Fig. 5c) and Top 10 upregulated DE-miRNAs in both regions, respectively (Fig. 3b, c). They regulate protective functions like antiepileptic activity (miR-335) [132] and neuroprotection in ischemia (miR-451) [66]. In summary, in the FL, miR-363 and miR-15b was identified as the best candidate biomarker for injury at 1 day and miR-335 and miR-451 were identified as the best candidate biomarkers for CsA treatment.

In the H&A, at 1 day post-injury, miR-16 and miR-450c-3p were identified by LASSO (Fig. 6c). miR-450c-3p was part of the upregulated DE-miRNAs in both regions (Fig. 2b, c). Lastly, miR-935 was the only miRNA identified by LASSO in the CsA treated group. It was part of the Top 5 hub miRNAs in ME5 (Fig. 5c). In summary, in the H&A, miR-450c-3p was identified as the best candidate biomarkers for injury at 1 day and miR-935 was identified as the best candidate biomarker for CsA treatment.

Overall, by combining these 3 analysis LASSOS, DESeq2, and WGCNA, we have identified a subset of miRNAs that have biological relevance and can predict injury at 1 day and those that can predict CsA treatment. All these miRNAs have been identified as being protective in neurodegenerative disorders and could potentially serve as robust candidate biomarkers for pediatric mTBI.

Discussion

MicroRNAs have shown significant utility as potential biomarkers and targets of novel therapies in traumatic brain injury (TBI) [127, 133–135]. In this study, we examined how miRNAs changed at 1 day post-injury, 1 week post-injury, and following 1 day of 20 mg/kg/day cyclosporine A (CsA) treatment in the frontal lobe (FL) and hippocampus + amygdala (H&A) of piglets that experienced a sagittal rapid non-impact head rotation (RNR) injury. Our analysis identified many differentially expressed miRNAs (DE-miRNAs) that were changed at 1 day post-injury then returned to baseline by 1 week post-injury, many of which were shared between the FL and H&A regions and involved in dysregulated neuronal and growth functions related to neurodegeneration and neuroprotection. Interestingly, in the FL region, the predicted genes targeted by the DE-miRNAs were involved in upregulated vascular, immune, and developmental processes. Two Top 10 upregulated DE-miRNAs involved in neuroprotection were correlated with decreased axonal injury in the FL. Only 3 DE-miRNAs remained at 1 week post-injury, and they were all part of the Top 10 DE-miRNAs at 1 day post-injury. Additionally, WGCNA identified modules of miRNAs displaying the protective effect of 1 day of CsA in ameliorating downregulated neuronal processes and upregulated immune and vascular processes. Lastly, LASSO identified key miRNAs that were the best predictors of mTBI at 1 day post-injury and CsA treatment. (Fig. 7). Additionally, WGCNA identified co-expression of miRNAs that may contribute to the protective effect of CsA in ameliorating downregulated neuronal processes and upregulated immune and vascular processes. Lastly, LASSO identified key miRNAs that could potentially serve as candidate biomarkers of pediatric mTBI in brain tissue at 1 day post-injury and CsA treatment. (Fig. 7).

Fig. 7.

Findings from DESeq2, WGCNA, and LASSO analysis. We observed key DE-miRNAs involved in damage/neurodegeneration and neuroprotection in both regions. In the FL, GO analysis of target genes revealed upregulated immune response and vascular processes. Only 3 DEGs remained by 1 week post-injury, and miR-20a-3p is known for its role in neuroprotection. WGCNA revealed the role of CsA treatment in ameliorating downregulated neuronal and upregulated immune, stress, and vascular responses. LASSO revealed key miRNAs that serve as candidate biomarkers for mTBI at 1 day and CsA treatment. Illustration from NIAID NIH BioArt Source

As highlighted in our previous paper [31], piglets are very useful models to serve as surrogates for humans to investigate TBI, specifically children due to their similarities in brain anatomy, growth, and postnatal development compared to human infants and children [136–138]. Additionally, they are known to experience similar responses to brain trauma like humans [41]. Thus, they provide great potential in evaluating brain specific molecular changes in response to TBI. Various studies have evaluated molecules like RNAs and proteins following TBI to expand their understanding on cellular and molecular mechanisms in TBI pathophysiology [11–15, 139–144]. Of all these molecules, miRNAs show great promise due to their unique characteristics in regulating multiple mRNAs and biological processes like cell proliferation, differentiation, survival and motility, all of which are vital in TBI molecular response [145]. Additionally, they are even more useful in examining neurological disorders due to their stability, abundance in brain tissue, small size, ability to cross the BBB, and modulate BBB permeability [17, 146]. By combining this useful animal model and molecule, we are the first to utilize piglets to study brain derived miRNAs changes in two fundamental brain regions, at two timepoints post-injury, and following CsA treatment after pediatric mTBI.

We identified DE-miRNAs at 1 day post injury shared in both regions that returned to baseline by 1 week. We interpreted possible effects of the DE-miRNAs by identifying miRNA target genes to define possible biological functions of the miRNA in normal or injury physiological conditions [147, 148]. We then identified predicted gene targets that were observed as differently expressed genes (DEGs) in our prior work that utilized the same samples as this study [31]. GO analysis of our predicted gene targets highlighted overrepresented functions that were mostly involved in neuronal and synaptic functions. Neuronal damage and neurogenesis are very common following TBI and in our study we observed a downregulation of synaptic functions, cytoskeletal organization, and an alteration of neurogenesis associated DE-miRNAs [149, 150]. Additionally in the FL only, we further identified overrepresentation of upregulated glial and vascular cell migration gene targets. Following TBI, studies have observed migration of astrocyte and microglia cells, in addition to a migration of vascular cells in order to mediate repair post-injury [151, 152].

Interestingly, the Top 10 upregulated DE-miRNAs shared in both regions at 1 day post-injury (miR-451, miR-130b-5p, miR-212, miR-363, miR-27a, and miR-17-3p) have all been evaluated for their therapeutic potential in driving neuroprotection, survival, and regeneration in neurological disorders [66, 71, 153–157]. Additionally, some have specifically been observed in previous TBI and ischemia studies. For example, miR-451 has been identified to be upregulated in the serum and CSF of both mTBI and severe TBI (sTBI) human patients and the cortex of CCI mice [18, 158]. miR-363 has been identified to be upregulated in the blood of ischemic stroke patients [159]. Our analysis identified miR-363 as the best predictors of injury at 1 day post-injury, implying that it should be further evaluated as a diagnostic biomarker in TBI. miR-17-3p is a member of the miR-17-92 cluster which has been identified to be upregulated following TBI and neuroprotective [20]. In contrast, two of our Top 10 upregulated DE-miRNAs, miR-27a and miR-212 have been observed as downregulated in young adult TBI models [160, 161]. While miR-130b-5p has been observed to be downregulated after ischemia [162]. Their overexpression have also been observed to drive neuroprotection and we speculate that since they were upregulated in our pediatric data, they may be vital for driving neuroprotection in pediatric mTBI models [71, 154–156]. Some of these Top 10 upregulated DE-miRNAs have also been identified to enhance blood brain barrier integrity and tight junction stabilization, as well as reduce vascular leakage and endothelial inflammation following stroke and intracerebral hemorrhage, miR-212, miR-27a, and miR-17-3p [73, 163, 164]. Additionally, these vascular-associated DE-miRNAs, miR-212 and miR-17-3p, were correlated with decreased axonal injury (Fig. 4). Although there are no established links between axonal injury and cerebrovascular integrity after TBI, these miRNAs may help drive both vascular/BBB repair and axonal injury amelioration. Additionally, some of the predicted DEG targets of our Top 10 upregulated DE-miRNAs in both regions, were also identified as Top 10 DEGs in our previous work, including upregulated CLU and downregulated PGAM1, TMSB10, MICU3, and ID4 [31]. Similar to the DE-miRNAs, their changed expression are involved in protective functions like remodeling, mitochondrial maintenance, development and amyloid β (Aβ) clearance [82–86]. Overall, all our Top 10 upregulated DE-miRNAs and their target genes should be studied further as potential strategic opportunities for development of neuroprotective therapies in pediatric TBI.

There are currently no effective diagnostic biomarkers or therapeutic markers for pediatric TBI. miRNAs offer critical potential to improve diagnosis, prognosis, and therapeutic identification for pediatric TBI. Statistical methods like LASSO are particularly valuable to achieve this as they can efficiently identify subsets of injury predictive miRNAs, thereby highlighting clinically relevant signatures for pediatric TBI. In our study, the DE-miRNAs identified by LASSO, DESeq2, and WGCNA as the best predictors of mTBI at 1 day post-injury have all been implicated in TBI and ischemia. miR-363 is upregulated in the blood of ischemic stroke patients and known to be neuroprotective [159]. miR-15b has been identified in AD and it’s upregulation has been related with decreased APP and Aβ in AD [165]. miR-450c-3p has not been extensively studied in neurodegenerative disorders. Interestingly, a miR-363 and miR-15b have been highlighted for their potential to serve as interventions for ischemic stroke and AD treatment [68, 165]. Overall, these key miRNAs could serve as fundamental diagnostic biomarkers and therapeutic interventions for mTBI, given their prior identification and preclinical analysis in neurological disorder studies.

In our study we observed more acute compared to sustained miRNA responses. By 1 week post-injury, most of the DE-miRNAs had returned to baseline and only 3 were sustained as DE-miRNAs. Similarly, other TBI studies have observed a reduction in DE-miRNAs at 1 week post-injury compared to 1 day post-injury [166, 167]. One of these studies conducted in CCI mice, observed that miRNAs that remained at 1 week post-injury were involved in brain repair mechanisms [166]. In our study, upregulated miR-20a-3p, and downregulated miR-10,386, and miR-4331-3p were the only 3 DE-miRNAs that remained and they were also part of the Top 10 up-and downregulated DE-miRNAs at 1 day post injury in the FL and/or H&A. miR-20a-3p, was the only one homologous to humans and it was observed in the FL at 1 day and 1 week post-injury, is particularly interesting due to its known neuroprotective role in reducing infarction and improving sensory motor function in male and female stroke models [65, 168]. Additionally, its predicted gene targets identified as DEGs in our previous study are involved in protective and degenerative mechanisms like neuron survival, neuroprotection, synaptic disruption, BBB damage, and endothelial cell injury, all of which are involved in TBI pathophysiology. Overall, we speculate that the few miRNAs that remain dysregulated at a later timepoint of 1 week post-injury are crucial in TBI pathophysiology and attention should be placed on them during miRNA analysis in TBI.

Lastly, WGCNA analysis revealed modules of miRNAs ameliorated by 1 day of CsA treatment. We observed that ME1 and ME3 were ameliorated by CsA treatment in the H&A region only. Similarly in our previous study which focused on genes, we identified a higher effect of CsA in the H&A region compared to the FL [31]. The 5 hub miRNAs within ME1, miR-296-3p, miR-423-5p, miR-2366, miR-139-3p, and miR-320 have been observed to be downregulated in neurological and neurodegenerative disorder studies. miR-296-3p has been observed to be downregulated in AD impaired aged mice with amyloidosis and tau pathology [169]. miR-423-5p is downregulated in spinal cord injury (SCI) and lower levels are associated with reduced hippocampal volume, worse cognition, and higher brain Aβ deposition [60, 61]. miR-139-3p is downregulated in TBI induced post-traumatic epilepsy mice and Huntington’s disease (HD) [59, 170]. miR-320 has been identified to be downregulated in depressed patients and Parkinson’s disease (PD) [171, 172]. Furthermore, miR-423-5p and miR-139-3p were also observed as Top 10 DE-miRNAs in the FL and should be further explored. ME3 included co-expression of miRNAs upregulated acutely at 1 day. Similar to ME1, the 5 hub miRNAs within ME3, miR-27b-3p, miR-130b-3p, miR-107, miR-23b, and miR-30b-5p have also been identified in neurodegenerative disorder studies, specifically TBI. miR-27b-3p is involved in aggravated microglia mediated neuroinflammation and accelerated apoptosis is upregulated in PD patients [173, 174]. miR-130b-3p is upregulated in the serum of severe TBI patients [175]. miR-107 is upregulated in concussion (mTBI) and middle cerebral artery occlusion (MCAO) [103, 176, 177]. miR-30b-5p also involved in inflammation has also been identified to be upregulated in CSF of severe TBI patients and PD patients [178, 179]. In contrast, miR-23b although upregulated in our analysis has been observed to be downregulated in TBI patients and rats [180]. Additionally, it was identified as a good CsA predictor miRNA. In total, the 5 hub miRNAs from ME1 and ME3 have been implicated in several neurodegenerative/neurological disorders, specifically AD, SCI, TBI, epilepsy, HD, and PD. In our study we found that they were ameliorated by CsA treatment, further highlighting the crucial role of CsA in targeting key neurological related miRNAs. Studying these CsA mediated miRNAs could open doors to identify miRNA targets with therapeutic potential in TBI.

Limitations

Our study had several limitations. First, we studied a single, narrowly defined severity of rapid sagittal head rotation without impact. Due to mTBI heterogeneity, it is generally important to study a range of magnitudes and even multiple mTBIs. Secondly, we only studied females, which are generally known to have worse TBI clinical outcome [181]. Third, although several databases present validated and predicted miRNA binding sites, some of the databases are not comprehensive and may lack some predicted gene targets [50, 51]. Although we utilized two databases, miRDB and TargetScan, there is a possibility that not all gene targets for each miRNA were included in our study. Lastly, microRNA alterations were profiled in brain tissue. We first evaluated their expression in brain tissue, to help identify a subset involved in TBI pathophysiology. To establish their diagnostic potential, they will need to be examined in serum. Future studies should determine the level of those significant miRNAs identified in this study in biofluids like serum after TBI to assess their potential as clinical diagnostic metrics. Future studies will include comparison between sexes and studying more TBI mechanisms.

Conclusion

By examining microRNA changes in an immature swine model of diffuse mTBI, we identified more acute than sustained responses in miRNAs. Many DE-miRNAs at 1 day were shared between both regions and involved in both neurodegenerative and neuroprotective processes related to TBI pathophysiology, some of which have been studied in previous TBI studies. Two Top 10 vascular associated miRNAs, miR-17-3p and miR-212 were correlated to decreases in axonal injury in the FL. Only 3 DE-miRNAs, miR-20a-3p, miR-10386, miR-4331-3p were present at 1 week post-injury. Key miRNAs miR-363, miR-15b, and miR-450c-3p were the best predictors of injury at 1 day following mTBI and were also identified to be differentially expressed in both regions with one being part of the Top 10 DE-miRNAs in both regions. CsA treatment ameliorated downregulated neuronal and synaptic processes and upregulated immune, vascular, and stress response processes only in the H&A region. Overall, our data reveal key miRNAs regulated at 1 day and 1 week post-injury, and the role of CsA treatment in ameliorating miRNAs that may be associated with neuronal, immune, stress, and vascular processes, thereby highlighting miRNAs with great diagnostic and therapeutic potential in pediatric mTBI.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

mTBI

Mild traumatic brain injury

FFPE

Formalin-fixed paraffin-embedded

FL

Frontal lobe

H&A

Hippocampus + amygdala

RNR

Rapid non-impact head rotation

CsA

Cyclosporine A

miRNAs

MicroRNAs

DE-miRNAs

Differentially expressed microRNAs

DEGs

Differentially expressed genes

Log2FC

Log2 fold change

FDR

False discovery rate

RNA-seq

RNA sequencing

GO

Gene ontology

WGCNA

Weighted gene co-expression network analysis

ME

Module eigengene

LASSO

Least absolute shrinkage and selection operator

Author contributions

Conceptualization, S.S.M, L.B.W., and O.A.; Data Curation, O.A.; Formal Analysis, O.A.; Funding Acquisition, S.S.M.; Investigation, S.S.M, L.B.W., and O.A.; Methodology, S.S.M, L.B.W., and O.A.; Project Administration, S.S.M, L.B.W., and O.A.; Software, O.A.; Resources, S.S.M. and L.B.W.; Supervision, S.S.M. and L.B.W.; Validation, S.S.M, L.B.W., and O.A.; Visualization, O.A.; Writing—original draft, O.A.; Writing—review and editing, S.S.M., L.B.W., and O.A. All authors reviewed and approved the final version of the manuscript.

Funding

This study was funded by the Georgia Research Alliance and the National Institutes of Health (National Institute of Neurological Disorders and Stroke) awards R01NS097549, U01NS069545, R01NS039679, and R01NS115994. The funding institutions had no role in the conceptualization, study design, data collection, analysis, preparation, writing, review, and decision to submit this manuscript for publication.

Data availability

All raw and processed RNA-seq datasets generated and analyzed in this study are available in the NCBI Gene Expression Omnibus (GEO) repository under accession number GSE312832. Source values for all figures are provided with this paper in supplementary tables.

Declarations

Ethics approval and consent to participate

All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Pennsylvania. Experiments were carried out in an Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC) accredited facility.

Competing interests

The authors declare no competing interests.

Consent for publication

Not applicable.