PDE5 inhibition restores mitochondrial function and improves neurobehavioral outcomes after repeated mild blast TBI

Abstract

Repeated mild traumatic brain injury (rmbTBI) is associated with persistent neurological deficits in military personnel arising from secondary cerebrovascular and metabolic impairments. Critically, there are no FDA-approved therapeutics for rmbTBI. Phosphodiesterase 5 (PDE5) inhibition has been explored in preclinical and clinical TBI studies; however, its exact therapeutic mechanism remains unclear. Recent data demonstrate that PDE5 inhibition can rescue mitochondrial dysfunction. As such, we hypothesized that PDE5 inhibition with sildenafil restores mitochondrial health following rmbTBI, conferring neurobehavioral improvement after rmbTBI. To model rmbTBI, male rats were exposed to two 11 psi blast waves 24 h apart, and daily sildenafil was initiated either 15 min or 8 weeks post-blast to assess subacute and delayed treatment effects. At 7 d post-injury, sildenafil restored rmbTBI-induced reductions in hippocampal expression of PGC-1α, a master regulator of mitochondrial biogenesis. Transcriptomic analysis revealed enrichment of oxidative phosphorylation pathways in sildenafil-treated animals, and metabolomic profiling indicated that rmbTBI results in cortical accumulation of glycolytic intermediates and altered NAD⁺ redox balance, both attenuated by sildenafil. To assess the mitochondrial signature of brain capillaries after PDE5 inhibition, we show that sildenafil restored mitochondrial bioenergetics in brain capillaries following AAPH-induced oxidative stress and rmbTBI. rmbTBI increased anxiety-like behavior at 3 d and 7 d post-injury, which was partially normalized by sildenafil. At 20 weeks, rmbTBI animals exhibited persistent fear-conditioned behavior, whereas sildenafil facilitated fear extinction. These findings demonstrate that rmbTBI disrupts mitochondrial function and behavior, and that PDE5 inhibition with sildenafil restores these alterations, supporting its potential as a therapeutic intervention for blast-induced TBI.

Introduction

Repeated exposure to low-level blast (LLB) overpressure is a common occupational hazard for military personnel, particularly during training and operational activities involving weapon firing and breaching. In addition, service members may experience blast-induced traumatic brain injury (bTBI) from higher-intensity blast exposures encountered during enemy firing in military conflict. While severe, combat-related bTBI frequently results in overt brain injury, repeated LLB exposures are more likely to produce subclinical outcomes. Emerging evidence suggests that repeated blast insults may contribute to long-term health risks for service members [1,2]. Further, cumulative LLB exposure has been linked to chronic neurological sequelae, including post-concussive syndrome, tinnitus, headaches, cognitive dysfunction, post-traumatic stress disorder (PTSD), anxiety, and disturbances in sleep and mood [[3], [4], [5], [6]].

Although the pathological outcomes of repeated mild blast TBI (rmbTBI) remain incompletely understood, it has been shown that rmbTBI or LLB exposure can disrupt cerebrovascular function and damage the neurovascular unit (NVU) [7], an integrated network of endothelial cells, astrocytes, pericytes, and neurons that regulate cerebral blood flow (CBF) [2,5,8]. Clinically, these disruptions manifest as chronic alterations in cerebral hemodynamics and neurovascular health, which are increasingly recognized as drivers of persistent neurological dysfunction [5]. For instance, career breachers repeatedly exposed to LLB show elevated serum markers of blood-brain barrier (BBB) disruption and vascular endothelial injury [2]. Preclinical studies have demonstrated that targeted pharmacological intervention can restore NVU and BBB deficits after TBI and improve cognitive function at both acute and chronic timepoints post-injury [9,10].

In addition to neurovascular dysfunction, growing evidence suggests that metabolic dysfunction may be a key mechanism underlying the secondary pathophysiology of rmbTBI. Veterans with a history of blast exposure exhibit chronically reduced cerebral glucose metabolism, indicating that blast-related metabolic disruptions can persist long after initial injury [11,12]. Preclinical studies reinforce these observations, implicating mitochondrial dysfunction as a central driver of these metabolic alterations. Rodent models of blast-induced TBI demonstrate impairments in mitochondrial respiration, adenosine triphosphate (ATP) production, redox homeostasis, and mitochondrial biogenesis [[13], [14], [15], [16], [17]]. Furthermore, these disruptions in mitochondrial and oxidative stress pathways following blast-TBI are likely time-dependent. For example, mitochondrial apoptotic signaling peaks within 24 hours of injury [18], whereas dysregulation of fission-fusion dynamics and oxidative stress markers can persist up to 30 days post-injury [14]. Notably, early correction of TBI-induced aberrant mitochondrial fission has been shown to prevent chronic mitochondrial bioenergetic failure, oxidative damage, and cognitive impairment [19].

Importantly, oxidative stress within the cerebrovascular compartment has emerged as a key mechanistic link between blast exposure and mitochondrial dysfunction. Our previous work demonstrated that brain capillaries are particularly vulnerable to oxidative stress, exhibiting reduced mitochondrial abundance and respiration following exposure to 2,2-azobis (2-amidinopropane) dihydrochloride (AAPH) [20], a free radical generator that spontaneously decomposes to generate carbon-centered radicals that promote lipid peroxidation [21]. Additionally, blast-induced TBI has been shown to increase reactive oxygen species (ROS) and NADPH oxidase signaling within microvessels, contributing to BBB disruption and loss of tight junction integrity [[22], [23], [24]]. Elevated ROS levels have also been detected in brain endothelial cells following blast injury, supporting a role of endothelial oxidative stress in vascular dysfunction and injury progression [25].

However, despite this known pathology, there are currently no FDA-approved therapeutics to treat TBI. Sildenafil, a BBB-permeable vasodilator, is FDA-approved for the treatment of primary pulmonary hypertension and erectile dysfunction. It has also advanced to Phase II clinical trials for the treatment of traumatic microvascular injury in patients with a history of moderate-to-severe TBI [26,27]. While clinical reports so far indicate that sildenafil improves cerebrovascular health following TBI, there are no studies that evaluate sildenafil following bTBI or explore whether sildenafil can produce cellular effects beyond its known vasodilation mechanism. Sildenafil has been shown to improve mitochondrial health in endothelial cells by increasing mitochondrial DNA, enhancing respiration under oxidative stress, and elevating levels of PCG-1α, a transcriptional coactivator and master regulator of mitochondrial biogenesis [28]. Further, recent study by Zink et al. identified sildenafil as the top candidate to treat Leigh syndrome, a primary mitochondrial disease, and demonstrated its ability to regulate calcium homeostasis and mitochondrial membrane potential to improve neurobehavioral outcomes clinically [29]. Based on these recent studies, we conducted a series of experiments to interrogate the mitochondrial properties of sildenafil in a model of rmbTBI and whether sildenafil could improve neurobehavioral outcomes.

Materials and Methods

Animals

All experiments were conducted using male Sprague Dawley rats (Charles River) aged 7–8 weeks. Animal studies were approved by the United States Veterans Affairs Animal Component of Research Protocol (ACORP) and adhered to the guidelines of the Association for the Assessment and Accreditation of Laboratory Animal Care International (AAALAC, International). All animal procedures complied with ARRIVE guidelines and followed the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH Publication No. 8023, revised 1978). Animals were maintained on a 12-h light/dark cycle with ad libitum access to food and water in individually ventilated cages (IVC) (NexGen™ Rat 1800, Allentown Inc.). Random number generators were used to assign animals to groups, and every measure was taken to minimize confounding factors between groups. Additionally, transgenic mice expressing a global mitochondria-targeted fluorescent Dendra 2 reporter (mtD2g; homozygous) were obtained from Jackson Laboratories (B6; 129 S-Gt (ROSA)26Sortm1.1 (CAG-COX8A/Den-dra 2) Dcc/J; strain #108397) for specific experiments. This strain was generated by crossing mtD2g floxed mice with germline-deleting Mesox2-Cre mice, resulting in global mitochondrial expression across cell types. Mouse colonies were maintained at the Division of Laboratory Animal Resources at the University of Kentucky. Experiments were conducted using male mtD2g mice aged 10–12 weeks, with an average body weight of 25 g.

Experimental design and drug administration

Rats from all of the following experiments were subjected to the same rmbTBI paradigm. Sildenafil (Sil) or vehicle (Veh) was administered via oral gavage according to the cohort-specific designs. Experimental designs of subacute (7–8 days post-rmbTBI) studies are represented in Fig. 1A. For most studies, animals were randomized into the following groups: Sham, rmbTBI with vehicle treatment (rmbTBI-Veh), and rmbTBI with sildenafil treatment (rmbTBI-Sil). For select experimental cohorts, an additional Sham-Sil group was included. Because no significant differences were observed between Sham-Veh and Sham-Sil animals for the measured outcomes, sham groups were combined for subsequent analyses.

Fig. 1.

High dose sildenafil upregulates mitochondrial biogenesis regulator PGC-1α in the hippocampus at 7d post-rmbTBI. (A) Experimental timeline of subacute studies. Separate cohorts are indicated by different colors. (B) Bar graphs represent densitometric quantification of PGC-1α expression in the cortex and hippocampus using ImageJ (n = 5 rats/group) (top row). Representative Western blot image shows the expression of PGC-1α in the cortex and hippocampus of rats in the sham, rmbTBI-Veh, rmbTBI-5 mg/kg Sil, and rmbTBI-20 mg/kg Sil groups (bottom row). Data represented as mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 by one-way ANOVA followed by Tukey's post hoc multiple comparisons test.

The first subacute experiment examined the effects of two sildenafil doses (5 and 20 mg/kg) on cortical and hippocampal PGC-1α expression after rmbTBI. Rats received sildenafil or vehicle treatment via oral gavage 15 min after the first blast exposure and then once daily thereafter. Based on the efficacy of the 20 mg/kg dose in this study, subsequent experiments utilized the 20 mg/kg dose.

A second set of subacute experiments was conducted using three parallel cohorts, all of which underwent the same rmbTBI paradigm. All animals received vehicle or sildenafil (20 mg/kg) beginning 15 min after the first blast exposure and continuing once daily thereafter. These cohorts were dedicated to distinct outcome measures, including bulk RNA sequencing, metabolomics, behavior testing, and vascular and mitochondrial analyses. Behavioral assessments consisted of open field testing at 3 days pre- and post-injury and elevated plus maze at 7 days post-injury, with tissue collection performed at 7–8 days post-injury.

An additional experiment examined the effects of rmbTBI and sildenafil treatment on mitochondrial respiration in isolated brain capillaries using Seahorse analysis. This cohort underwent the same rmbTBI paradigm as the subacute cohorts and received vehicle or sildenafil (20 mg/kg) beginning 15 min after the first blast exposure and continuing once daily thereafter. Animals were euthanized at 1 day post-rmbTBI.

A final experiment examined the effects of delayed sildenafil treatment on vascular, mitochondrial, and behavioral outcomes. Animals underwent the same rmbTBI paradigm as the previous studies. Sildenafil treatment (20 mg/kg) was initiated at 8 weeks post-injury and continued once daily for 12 weeks. At 20 weeks post-injury, fear-conditioning behavior was assessed, and glial reactivity, capillary mitochondrial volume, and capillary tight junction expression were quantified.

Repeated blast injury model

For rmbTBI, Sprague Dawley rats were transported to the blast site and injured as previously described using the McMillan Blast Device [30]. Animals were briefly anesthetized with isoflurane using the SomnoSuite Low-Flow Anesthesia System (Kent Scientific Corporation, Torrington, CT) just before injury. Following physiological recordings, the rats were placed in a mesh netting support (Industrial Netting, Minneapolis, MN) and secured in the blast device. The mesh netting prevented head rotation, ensuring animals were only exposed to primary blast injury. During the rmbTBI protocol, sham animals were present at the blast site but were not exposed to injury. To model rmbTBI, animals were exposed to two mild blasts at 11 psi peak static overpressure 24 hours apart. Animal weights were recorded for all groups before and after rmbTBI to ensure consistency in weight range across groups.

Open field test

Open field testing was conducted in the morning hours as previously described [30]. Both rmbTBI and sham groups were assessed for anxiety-like behavior 3 days before and after the first blast exposure. During testing, rats were placed in a 32″ × 32″ × 12″ dimly lit box for 10 min, and their exploration was recorded using EthoVision software. The box was digitally divided into two zones, with an inner zone centered within the larger outer zone and set at half its size. The software tracked the center point of each animal, logging the number of entries and time spent in each zone. Between each test, the box was cleaned with 70 % ethanol (EtOH).

Elevated plus maze

Elevated plus maze testing was conducted in the morning hours as previously described [30]. Both rmbTBI and groups were assessed for anxiety-like behavior 7 days after the first blast exposure. Rats were placed in the Elevated Plus Maze (EPM; Med Associates Inc., Fairfax, VT, USA), which consists of two open arms (20″ × 4″) and two closed arms (20″ × 4″ × 15.94″), arranged in a cross with an open junction in the center (4″ × 4″). The maze, which is roofless, is elevated 29.31″ above the ground. Each rat was placed at the junction of the open and closed arms and allowed to explore freely for 5 min. Entries into each arm and the time spent in both open and closed arms were recorded using Ethovision software. To minimize stress, the lighting was dimmed slightly during the test. The maze was thoroughly cleaned with 70 % EtOH between trials to eliminate olfactory cues.

Fear conditioning

Assessment of fear-conditioned behavior began at 140 days following the initial blast exposure and continued for 4 consecutive days. Freezing, defined as complete immobility except for minor respiratory movements, serves as a widely accepted indicator for conditioned fear [31]. Rats underwent testing in a standardized fear conditioning system (Med Associates Inc., Fairfax, VT). On the first day, rats were placed in a fear conditioning chamber and allowed to explore. A 2-s 0.75 mA foot shock was delivered through the chamber floor at 120 s and then at 60-s intervals, for a total of 6 foot shocks. After the final shock, rats remained in the chamber for an additional 60 s without further shocks before removal. On days 2–4, rats were returned to the same chamber for 480 s without shocks to assess conditioned fear learning and extinction. Freezing behavior was quantified from video recordings using Med Associates Inc. VideoFreezeTM Video Fear Conditioning Software.

Tissue processing

Cohorts of rats were euthanized at 1 day, 7–8 days, or 20 weeks following rmbTBI. For cohorts designated for combined capillary isolation and tissue staining, animals were anesthetized with isoflurane prior to transcardial perfusion with cold, sterile phosphate buffered saline (PBS) (2.7 mM KCl, 1.46 mM KH₂PO4, 136.9 mM NaCl, 8.1 mM Na₂HPO₄, 5 mM d-glucose, 1 mM sodium pyruvate, pH 7.4). Brains were removed, and the hemispheres were separated, with one hemisphere used for capillary isolation and the other hemisphere designated for tissue staining. For cohorts designated solely for capillary isolation, Western blot, and RNA-sequencing, animals were euthanized using CO₂ before decapitation and rapid brain removal. For Western blot analysis, brains were dissected to isolate the cortex and hippocampus, which were immediately frozen on dry ice. For metabolomic analysis, animals were decapitated without anesthesia, and brain tissue was rapidly collected and frozen in liquid nitrogen. Hemispheres designated for tissue staining were fixed with 10% neutral buffered formalin (NBF). Following fixation, tissue was placed into 30% sucrose in PBS buffer solution for at least 48 h for cryoprotection. Brains were then flash frozen in −25 to −35 °C isopentane before being cut into 40 μm thick sagittal brain sections using a sliding microtome (Microm HM 450, Thermo Fisher). Tissue sections were stored at −20 °C in cryoprotectant (30 % glycerol, 30 % ethylene glycol in 1 × Tris buffered saline (TBS)) until used for immunostaining.

Capillary isolation

Mouse and rat brain capillary isolation was performed as described previously [20,32] with slight modification. Briefly, after euthanasia, brains were collected in sterile ice-cold capillary isolation buffer (DPBS; Cat No: 14,080,055, Fisher Scientific with 5.5 mM glucose, 1 mM sodium pyruvate, and 1% BSA, pH 7.4). The cerebral hemispheres were isolated, minced, and transferred into a sterile 2 ml screw-cap tube (Sarstedt Inc Screw Cap Microtube, type H) preloaded with stainless beads (3.2 mm, 1.8 g, fisher scientific, NC0778455). 1 ml of ice-cold capillary isolation buffer was then added into each tube before samples were homogenized in bead homogenizer (Biospec products) for 15 s. The homogenate was transferred into a 5 ml tube and topped with sterile isolation buffer to 5 ml. Followed by a brief vortex, the homogenate was centrifuged at 1000×g for 10 min at 4 °C using a swinging bucket rotor. The pellet was harvested, resuspended in 5 ml of sterile lymphocyte separation medium (25-072-CV; Corning Life Sciences), and then centrifuged at 4500×g for 20 min at 4 °C using a swinging bucket rotor. After removing the top layer insoluble matter and supernatant, the small pellet was resuspended in 1 ml of sterile ice-cold isolation buffer, and the suspension was filtered through a 70 μm membrane filter (15–1070; Tisch Scientific) to remove larger vessels. The membrane was then washed with at least 10 ml of isolation buffer. The filtrate was then filtered through an 18 μm membrane (25 mm diameter) filter (ME17233; Tisch Scientific). At least 10 ml of wash buffer (DPBS; Cat No: 14,080,055, Fisher Scientific with 5.5 mM glucose, 1 mM sodium pyruvate with pH 7.4) was used to wash the membrane to remove RBCs, suspended brain cells and debris. After the final wash with 2 ml of endothelial cell culture medium (ECM) (C-22011; PromoCell), the capillaries from 18 μm membrane filter were released into one 1.5 ml tube filled with 1 ml of ECM. A drop of capillary elute was used to assess the quality and concentration of the capillaries under a bright field microscope. Final centrifugation was performed at 1000×g, 10 min, 4 °C (swinging bucket) to pellet the isolated brain capillaries. The final capillary suspension in ECM from five mice was pooled for ex vivo AAPH and sildenafil treatment.

Immunofluorescence

Two small hydrophobic barriers were drawn on charged microscope slides using a hydrophobic pen. Approximately 150 μL of isolated capillary suspension was loaded within each barrier and allowed to settle onto the slide surface for 1.5 h at 25 °C. The mounted capillaries were then fixed with 10% neutral buffered formalin for 10 min at room temperature. Following fixation, samples were permeabilized and blocked using a solution containing 0.2 % Triton X-100 in PBST, 1 % BSA, and 10 % normal horse serum (NHS) for 1 h at room temperature. Samples were then incubated overnight at 4 °C with various primary antibodies. All the primary antibodies were used in 1:250 dilution: Zona occludens-1 (ZO-1) (302,200, Invitrogen) and OXPHOS (ab110413, Abcam). All secondary antibodies were used in 1:500 dilution for 1 h at room temperature: Alexa flour 594 donkey anti-rabbit IgG (A2120, Invitrogen), Alexa flour 594 donkey anti-mouse IgG (A21203, Invitrogen), Alexa flour 488 donkey anti-rabbit IgG (A21206, Invitrogen), and Alexa flour 488 goat anti-mouse IgG (A11001, Invitrogen). Samples were mounted on glass slides using a Vectashield HardSet Antifade Mounting Medium with DAPI (H-1500-10, Vector laboratories). Fluorescently stained capillaries were imaged using confocal microscopy (100X oil) with Z-stack acquisition. Max intensity projections were generated from Z-stack images, and mitochondria and tight junction protein expression were quantified using Imaris (X64 9.6.1) and NIS-Elements version 4.40.03, respectively. For each animal, 3–5 capillaries were imaged and analyzed using NIS-Elements software. Quantitative measurements from individual capillaries were averaged to generate a single value per animal for statistical analysis.

Western blot

Frozen cerebral cortex or hippocampus samples isolated from left rat brain hemisphere were thawed on ice and homogenized for 10 s using a bead homogenizer (Biospec products). The homogenate (90 μL) was lysed with 10 μL of 10x RIPA buffer (0.5 M Tris-HCl, pH 7.4, 1.5 M NaCl, 2.5% deoxycholic acid, 10% NP-40, 10 mM EDTA) containing 1x protease inhibitor cocktail (78,430, Thermofisher). The samples were snap-frozen in liquid nitrogen and stored at −80 °C. For Western blot analysis, lysates were cleared by centrifugation at 10,000×g for 20 min to remove insoluble fraction. After determination of the total protein concentration with BCA (23,250, ThermoScientific), the sample was prepared with XT sample buffer (1,610,791, Bio-Rad) containing 2 mM β-mercaptoethanol (BME), boiled at 95 °C for 10 min, loaded 20 μg per well, separated on 4–12 % BIS-TRIS SDS-PAGE gels (3,450,125, Bio-Rad), and electrophoretically transferred to Immobilon-P membranes (PVDF membrane, Millipore). The PVDF membrane was blocked with 5 % non-fat dry milk in TBS containing 0.05% Tween-20, followed by overnight incubation at 4 °C with primary antibody against PGC-1α (1:1000; 66,369-1-1 g, Proteintech). Fluorescent secondary antibody IRDye 680RD goat anti-mouse (1:10,000; 926–68070, LI-COR Biotechnology) was incubated with the membrane for 2 h at room temperature. Blots were imaged using the LI- COR Odyssey DLx imager (LI-COR Biotechnology, Nebraska, USA) and protein levels were quantified by densitometric analysis using ImageJ software and normalized with β-actin.

RNA sequencing

RNA-seq data and experimental design

Bulk RNA sequencing data were obtained from rat traumatic brain injury (rmbTBI) experiments comprising three experimental groups: Sham, rmbTBI-Veh, and rmbTBI-Sil. The primary contrasts analyzed were rmbTBI-Sil vs rmbTBI-Veh and Sham vs rmbTBI-Veh. Sample sizes for each comparison are indicated in the corresponding figure legends. Raw sequencing reads and aligned gene-level count matrices were provided by the sequencing facility and used for downstream bioinformatic analysis.

Gene expression quantification and preprocessing

Gene-level read counts were analyzed in R (version 4.4. x). Genes with very low expression were filtered prior to differential expression analysis by requiring a minimum total count across all samples (row sum ≥10). Count data were normalized using DESeq2 size-factor normalization to account for library size and compositional differences across samples.

Differential expression analysis

Differential gene expression analysis was performed using DESeq2, fitting a negative binomial generalized linear model with experimental condition as the design factor. Wald tests were used to assess statistical significance. P-values were adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) method. Unless otherwise stated, differentially expressed genes (DEGs) were defined using a prespecified threshold of FDR <0.05 and |log₂ fold change| > 1. Log₂ fold changes were further stabilized using apeglm shrinkage for visualization and interpretation.

Variance stabilization and quality control

For visualization and quality control, normalized counts were transformed using the variance-stabilizing transformation (VST) implemented in DESeq2. Principal component analysis (PCA) was performed on VST-transformed counts using the most variable genes (DESeq2 default). Sample–sample similarity was assessed using Euclidean distance and Pearson correlation heatmaps based on VST expression values. Additional quality control metrics included dispersion estimates, p-value distributions, and size-factor inspection to ensure appropriate model fit and normalization.

Visualization of differential expression

Results were visualized using MA plots, volcano plots, heatmaps, and single-gene boxplots. MA plots display log₂ fold change as a function of mean normalized expression, with biological effect-size thresholds (|log₂FC| = 1) indicated. Volcano plots summarize statistical significance (−log₁₀ p-value) versus effect size, highlighting genes meeting DEG criteria. Heatmaps show VST-transformed expression values for top DEGs, z-scored per gene to emphasize relative expression differences across samples. Single-gene boxplots were generated when only one gene met DEG thresholds.

Gene set enrichment analysis (GSEA)

Gene set enrichment analysis was performed using a rank-based approach without prefiltering for DEGs. Genes were ranked by the DESeq2 Wald statistic, preserving both direction and magnitude of differential expression. GSEA was conducted against MSigDB Hallmark (H) and C2 curated gene sets using clusterProfiler/fgsea-based methods. Enrichment significance was assessed using permutation testing, and p-values were adjusted for multiple testing. Normalized enrichment scores (NES), adjusted p-values, and leading-edge subsets were extracted for interpretation.

Over-representation analysis (ORA)

Complementary over-representation analysis was performed on statistically significant DEGs using Gene Ontology (GO) Biological Process annotations. Enrichment was tested using a hypergeometric framework with Benjamini–Hochberg correction. Dot plots summarize enriched terms, where dot size represents the number of contributing genes and color indicates adjusted p-value.

Sensitivity and exploratory analyses

To assess robustness, sensitivity analyses were conducted using alternative DEG thresholds and by excluding predefined sample subsets where indicated. In comparisons where no genes met DEG criteria under standard cutoffs, pathway-level analyses (GSEA) were emphasized to capture coordinated but subtle transcriptional changes not detectable at the single-gene level.

Software and reproducibility

All analyses were performed in R using DESeq2, apeglm, clusterProfiler, fgsea, enrichplot, pheatmap, and tidyverse packages. Figures and result tables were generated directly from the analysis scripts, and session information was recorded to ensure reproducibility.

LCMS metabolomics workflow

Sample preparation

Tissue was collected from sham, rmbTBI-Veh, and rmbTBI-Sil groups receiving vehicle or sildenafil treatment (20 mg/kg). All tissues were handled on liquid nitrogen when removed from cryostorage at −80 °C. Brain were pulverized in liquid N₂ using a Freezer/Mill Cryogenic Grinder (model 6875D, SPEX SamplePrep, Metuchen, NJ, USA). Approximately 25 mg of powdered tissue was removed with a micro spatula to a new tube and extracted with 50% MeOH in water. The sample mixtures were vortexed for 10 min at 4 °C, then centrifuged at 24,000×g for 10 min at 4 °C. The aqueous phase containing polar metabolites was filtered to remove lipids using Agilent Captiva EMR-Lipid filter cartridges under a positive pressure manifold. The resulting filtrate was fully dried at 10⁻³ mBar using a SpeedVac (Thermo Fisher Scientific, Waltham, MA, USA) without heat. Samples were then reconstituted in 70% acetonitrile/20% water and 10% MeOH, normalized to mass of tissue used for extraction.

HPLC parameters

Metabolite separation was achieved using the Agilent 1290 Infinity II Stainless HPLC system equipped with an InfinityLab Poroshell 120 HILIC-Z column (2.1 mm × 150 mm; p/n 673,775,924). Mobile phase A consisted of 10 mM ammonium acetate in water supplemented with 0.1% medronic acid and B was pure acetonitrile. The flow rate set at 0.400 ml/min and gradient elution program was as follows.

Time	%B
0	90
1.00	90
8.00	78
12.00	60
15.00	10
18.00	10
19.00	90
23.00	90

MS/MS parameters

Analysis performed on Agilent 6495D LC/TQ with the following parameters: ion mode set at both positive and negative, gas temperature 200 °C, drying gas flow 14 L/min, nebulizer gas at 50 psi, sheath gas temperature at 375 °C, sheath gas flow at 12 L/min, capillary voltage set to (+)3000/(−)2500 V and nozzle voltage set at 0 V.

Data analysis

After data acquisition, initial data analysis performed on Agilent MassHunter Quantitative Analysis for peak identification and quantification. Relative abundance of each metabolite was then exported into Excel and statistical analysis performed using Metaboanalyst (https://www.metaboanalyst.ca). Graphs were made using GraphPad Prism 9.

AAPH and sildenafil treated capillaries

AAPH and sildenafil treatment

The mixed capillary suspension in ECM was aliquoted into five 1.5 ml tubes (500 μl each) for five groups. AAPH (2,2′-Azobis (2-methylpropionamidine) dihydrochloride, Sigma, #440914) or vehicle was added to a final concentration of 0.5 mM, followed by incubation in a 37 °C CO2 cell culture incubator for 3 h. Sildenafil (Sildenafil Citrate, Sigma, #PHR1807) or vehicle was then added to final concentrations of 0 μM, 0.1 μM, 1 μM, and 10 μM, and the samples were incubated overnight. A hole was made in the microcentrifuge cap using a soldering iron gun for CO2 exchange. Following incubation, a small aliquot of capillary suspension (50 μL) was used for mitochondrial imaging analysis, and the rest of the capillary suspension was centrifuged and resuspended in XF assay medium for mitochondrial respiration analysis.

Mitochondrial quantification in imaris

The mitochondrial shape, size, and count were quantified as described previously [20]. Briefly, mtD2g green fluorescence from Z-stack images was quantified using Imaris (X64 9.6.1) using surface creation tool. A smoothed surface with background subtraction (local contrast) was used to separate the mitochondria from the background. A split-touching object was enabled to separate individual adjacent mitochondria. This template file was used to bulk-process all the images from different groups. Mitochondrial volume (sum of total mitochondrial volume detected by Imaris surface detection tool) and count (count of mitochondrial segments detected by Imaris surface detection tool) were normalized to the total capillary area and expressed in 1000 μm².

Mitochondrial respiration in capillaries

To evaluate the effects of rmbTBI on capillary bioenergetics, capillaries were isolated from the cortex of sham and rmbTBI animals (male Sprague Dawley rats (n = 8-11/group)) at 1 day post-injury. For in vivo sildenafil treatment, vehicle or sildenafil (20 mg/kg) was initiated via oral gavage 15 min after both blast exposures. For ex vivo sildenafil treatment, isolated capillaries from rmbTBI were incubated in either control media (Veh) or sildenafil citrate (100 μg/ml) (Sil) for 2 hours prior to analysis.

Brain capillaries were isolated as previously described [20,32]. Purified capillaries were washed and used for Seahorse respirometry experiments. Capillary mitochondrial respiration was assessed using a Seahorse XFe96 Flux Analyzer (Agilent Technologies, Palo Alto, CA, USA). Isolated capillaries were pelleted and resuspended in XF assay medium with substrates (10 mM glucose, 1 mM pyruvate, and 2 mM l-glutamine). Equal amounts of capillary protein (5–25 μg) were plated in triplicate wells of XF96 microplates and centrifuged at 800×g for 10 min to settle capillaries at the bottom of the well. After a 1 h incubation at 37 °C, oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured. After three measurements of baseline OCR (basal respiration), respiratory chain inhibitors/uncouplers were sequentially injected into each well (1 μM Oligomycin, 4 μM FCCP, 0.5 μM Rotenone/Antimycin A, and 5 mM 2-deoxyglucose (2DG)), and OCR was measured three times after each injection. Maximal respiration was determined from OCR following FCCP injection, and basal glycolysis was determined from Extracellular Acidification Rate (ECAR) following 2-DG injection. At the conclusion of the assay, the XF microplate was centrifuged at 1000 g for 10 min, after removing the media, 10 μL RIPA buffer was added to the capillaries in each well, and freeze-thaw cycles were performed 3 times in dry ice before protein was quantified using a BCA kit (23,225; Thermo Scientific). Using Wave software version 2.6.1 (Agilent Technologies, Santa Clara, CA, USA), OCR from various respiration states was calculated and normalized to protein concentration.

Statistics

Power analysis was conducted (using G × Power statistical software, v.3.0.10) based on previous published literature from our group [30]. A priori analysis was performed and effect size was calculated based on expected mean ± SD within each group, which varied based on outcome measure. For behavioral experiments, sample size was determined using the following parameters: α = 0.05, 1 − β = 0.8, and an estimated standard deviation of 20% of the mean (effect size of 1.12).

Statistical analysis was performed using Graph Pad Prism (GraphPad Software, CA, USA). Statistical significance was defined as α = 0.05. Data are graphed as mean ± SEM, with individual data points shown in the figures. Exact p values reported in the Results, and the biological replicates (n) are indicated the figure legends. Statistical significance is noted in the figures as ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Normality was assessed using the Shapiro–Wilk test, and homogeneity of variance was assessed using Brown–Forsythe and Bartlett's tests. As these assumptions were met, parametric statistical tests were applied. For comparisons involving a single independent variable, data was analyzed using one-way ANONA followed by Tukey's post hoc test for all pairwise comparisons or Dunnett's post hoc test for comparisons against a control group, as appropriate. For experiments involving two independent variables (e.g., time and treatment), data were analyzed using two-way ANOVA to assess main effects and interactions, followed by Šidák's post hoc multiple comparisons test. Pearson correlation analysis was also used to assess relationships between mitochondrial and vascular markers.

Data exclusion criteria were applied uniformly across experiments. For the in vivo sildenafil treatment study, at total of n = 3/group animals were excluded due to failed oligomycin injection during Seahorse analysis. For bulk RNAseq experiments, n = 3 rmbTBI-Veh and n = 1 rmbTBI-Sil were excluded based on sample processing.

Results

Sildenafil dose-dependently mitigates mitochondrial alterations following rmbTBI

To identify an optimal therapeutic dose of sildenafil for intervention following rmbTBI, male rats received vehicle (Sham, rmbTBI-Veh), 5 mg/kg sildenafil (rmbTBI-5 mg/kg Sil) or 20 mg/kg sildenafil (rmbTBI-20 mg/kg Sil) via oral gavage beginning 15 min after the first blast and continuing once daily thereafter. To evaluate mitochondrial biogenesis signaling, expression of the transcriptional coactivator PGC-1α was quantified in cortical and hippocampal tissue. A representative Western blot of sham, rmbTBI-Veh, rmbTBI-5 mg/kg Sil, and rmbTBI-20 mg/kg Sil groups is shown in Fig. 1B, bottom row. In the hippocampus, PGC-1α expression differed significantly among groups (one-way ANOVA; p < 0.0001; Fig. 1B, top row). Post hoc analysis revealed reduced PGC-1α expression in vehicle- and 5 mg/kg-treated rmbTBI animals relative to sham (p = 0.0001 and p < 0.0001, respectively), whereas 20 mg/kg sildenafil treatment increased PGC-1α expression compared to both rmbTBI-Veh and rmbTBI-5 mg/kg sildenafil groups (p = 0.0134 and p = 0.0058, respectively). In contrast, PGC-1α expression in the cortex did not differ among groups.

Together, these findings indicate that rmbTBI induces region-specific alterations in mitochondrial markers at 7 days post-injury. In particular, the hippocampus appears especially vulnerable to injury, as evidenced by reduced PGC-1α expression, and treatment with 20 mg/kg sildenafil restored this rmbTBI-induced deficit.

Sildenafil upregulates mitochondrial and metabolic transcriptional pathways following rmbTBI

Although cortical PGC-1α expression remained unchanged following rmbTBI, our previous work has demonstrated mitochondrial and metabolic dysregulation within the cortex after injury [15]. Therefore, to determine whether sildenafil influences broader transcriptional programs associated with these processes, RNA-sequencing was performed on isolated cortical tissue from sham, rmbTBI-Veh, and rmbTBI-Sil (20 mg/kg) groups at 7 days post-injury. Principal component analysis of VST-transformed counts from rmbTBI-Veh (n = 4 rats/group) and rmbTBI-Sil (n = 5 rats/group) samples revealed clear separation between treatment groups (Fig. 2A). Separation between groups primary occurred along principal component 1 (PC1), which accounted for 81% of the total variance, whereas principal component 2 (PC2) accounted for 5%. These data suggest that sildenafil treatment was associated with a marked shift in the cortical transcriptomic profile following rmbTBI.

Fig. 2.

Sildenafil enhances metabolic transcriptional pathways following rmbTBI. (A) Principal components analysis (PCA) on VST-transformed counts for the same samples (rmbTBI-Veh n = 4; rmbTBI-Sil n = 5). Colors indicate treatment group. Axis labels show variance explained (PC1 = 81%, PC2 = 5%). Separation along PC1 reflects the dominant treatment effect. (B) GSEA (MSigDB Hallmark), dot plot Gene Set Enrichment Analysis using an all-gene ranked list (DESeq2 Wald statistic, unshrunken) against MSigDB Hallmark (H) gene sets. The most significant terms (e.g., Oxidative Phosphorylation, Mitotic Spindle, MYC Targets V1) show strong enrichment (q < 0.05) with direction defined by the Wald-statistic rank (positive = Sildenafil-higher). (C) GSEA running-score plot (Hallmark, top pathway). Running enrichment score for the top Hallmark set (HALLMARK_OXIDATIVE_PHOSPHORYLATION). Early accumulation of hits and a positive peak indicate higher expression in the rmbTBI-Sil group. Reported p and q values are from GSEA with BH correction.

Differential expression analysis identified 2306 significantly differentially expressed genes (DEGs’ FDR <0.05, |log₂FC| > 1), including 1036 upregulated and 1270 downregulated genes in rmbTBI sildenafil-treated animals relative to vehicle-treated controls (Fig. S1A). Hierarchical clustering of the top DEGs further demonstrated clear segregation of treatment groups, consistent with a robust transcriptional response to sildenafil treatment (Fig. S1B).

To identify pathway-level transcriptional changes between vehicle and sildenafil treatment in rmbTBI animals, gene set enrichment analysis was performed using the MSigDBHallmark collection on an all-gene ranked list derived from the unshrunken DESeq2 Wald statistic comparing cortical tissue from rmbTBI-Veh and rmbTBI-Sil groups. The dot plot in Fig. 2B summarizes the top Hallmark gene sets meeting significance criteria (FDR <0.05 and |log₂FC| > 1), with dot size reflecting the number of contributing genes and color indicating adjusted p-value. Sildenafil treatment was associated with strong enrichment of oxidative phosphorylation, mitotic spindle, and MYC targets gene sets (q < 0.05).

To visualize enrichment of the top Hallmark pathway, GSEA running enrichment score was plotted for Hallmark oxidative phosphorylation gene set comparing rmbTBI-Veh and rmbTBI-Sil groups (Fig. 2C). The running-score curve showed an early accumulation of oxidative phosphorylation genes at the top of the ranked list and a positive peak in the enrichment score, indicating overall higher expression of this pathway in the sildenafil group. This enrichment was highly significant (p = 1 × 10⁻¹⁰, FDA-adjusted p = 1.67 × 10⁻⁹, Benjamini-Hochberg correction), supporting sildenafil-associated upregulation of oxidative phosphorylation after rmbTBI.

These results suggest that sildenafil treatment is a dominant driver of transcriptional variability in the cortex following rmbTBI, primarily by upregulating mitochondrial oxidative phosphorylation pathways and metabolic regulatory proteins. This claim is supported by the co-enrichment of oxidative phosphorylation gene sets and MYC target pathways, as MYC target pathways have been shown to promote mitochondrial biogenesis and respiration [33,34]. Together, these findings support that sildenafil enhances mitochondrial bioenergetic capacity and MYC-driven transcriptional activity following rmbTBI.

Although our analysis focused primarily on metabolic pathways, over-representation analysis of significantly differentially expressed genes (FDR <0.05, |log₂FC| > 1) using Gene Ontology (GO) Biological Process terms revealed enrichment of neuronal pathways related to synaptic and junctional regulation, including regulation of synapse organization, post-synapse organization, synaptic vesicle cycle, and cell junction assembly (Fig. S1C). These findings are consistent with the DEG-level changes identified by differential expression analysis and suggest that sildenafil-associated transcriptional changes may extend to neuronal structural and signaling processes along with enhancement of metabolic function.

Sildenafil restores cortical metabolic homeostasis following rmbTBI

Cortical and hippocampal metabolites were quantified 7 days after rmbTBI using LC-MS to assess injury- and treatment-associated metabolic changes. Sparse partial lease squares-discrimination analysis (sPLS-DA) was performed on sham, rmbTBI-Veh, and rmbTBI-Sil groups following data filtering, missing value estimation, and normalization in Metaboanalyst (Fig. 3A). Missing values were minimal and did not differ across groups. In the cortex, sPLS-DA revealed separation of experimental groups, with Component 1 and Component 2 accounting for 18.2% and 9% of the variance, respectively. rmbTBI-Veh samples were most distinct from sham, whereas rmbTBI-Sil samples showed greater overlap with shams, consistent with partial normalization of rmbTBI-induced alterations in the cortical metabolomic profile. In contrast, hippocampal metabolites showed less distinct group separation, with Component 1 and Component 2 accounting for 9.4% and 11% of the variance respectively, suggesting that metabolic alterations were more pronounced in the cortex than in the hippocampus.

Fig. 3.

Sildenafil rescues rmbTBI-induced alterations in glycolytic metabolites and NAD redox balance in the cortex at 7d post-rmbTBI. (A) sPLS-DA score plots of metabolite profiles from the cortex and hippocampus. Each point represents an individual animal. Ellipses indicate 95% confidence interval for each group. (B) Volcano plots of cortical metabolites in sham vs. rmbTBI-Veh (left) and rmbTBI-Veh vs. rmbTBI-Sil (right). Each point represents a metabolite. Blue points indicate metabolites significantly altered between groups (p < 0.05). Selected metabolites are annotated. (C) Bar graphs represent the relative abundance of cortical metabolites measured by LC-MS metabolomics in sham, rmbTBI-Veh, and rmbTBI-Sil groups. Data represented as mean ± SEM (n = 5 rats/group). ∗p < 0.05, ∗∗p < 0.01 by one-way ANOVA followed by Tukey's post hoc multiple comparisons test.

Subsequent differential metabolite analyses were performed in the cortex and hippocampus comparing sham versus rmbTBI-Veh and rmbTBI-Veh versus rmbTBI-Sil groups. Results were visualized using volcano plots depicting log₂ fold change versus -log₁₀(p-value) (Fig. 3B). In the cortex, rmbTBI-Veh animals exhibited significantly increased levels of guanosine-5-monophosphate (GMP), fructose-6-phosphate, fructose-1,6-bisphosphate, and glucose 6-phosphate relative to shams (p < 0.05), whereas adenosine diphosphate (ADP) was significantly reduced. Comparison of rmbTBI-Veh and rmbTBI-Sil groups revealed reduced nicotinamide and GMP levels in sildenafil-treated animals, alongside significant increases in NADH and NAD⁺ (p < 0.05). In contrast, analysis of hippocampal metabolites identified fewer significant alterations between groups and did not reveal changes in glycolytic or NAD redox-related metabolites (Fig. S2A).

Additional analysis of cortical metabolites revealed significant group differences in the relative abundance of glucose-6-phosphate, fructose-6-phosphate, nicotinamide, and GMP (one-way ANOVA; p = 0.0183, 0.0176, 0.0461, and 0.0049, respectively; Fig. 3C). Post hoc analysis revealed that nicotinamide is significantly reduced in rmbTBI-Sil animals compared to rmbTBI-Veh (p = 0.0422) and that glucose-6-phosphate,fructose-6-phosphate, and GMP are increased in rmbTBI-Veh animals compared to sham (p = 0.0142, 0.0137, and 0.0036, respectively). However, one-way ANOVA analysis did not reveal any group differences in cortical ADP and fructose-1,6-bisphosphate relative abundance. Analyses of the effects of rmbTBI and sildenafil treatment on additional cortical metabolites are presented in Fig. S2B.

These results demonstrate that rmbTBI disrupts cellular energy metabolism. The accumulation of early glycolytic intermediates in rmbTBI-Veh animals, relative to shams, is suggestive of a potential metabolic bottleneck in downstream glycolytic or mitochondrial oxidative processes. Sildenafil treatment attenuated these injury-induced metabolic disturbances, shifting metabolite levels toward those observed in shams. Additionally, because nicotinamide is a breakdown product of NAD⁺ metabolism, the elevation observed in rmbTBI-Veh animals suggests increased NAD⁺ consumption and metabolic stress following injury. In contrast, sildenafil-treated animals exhibited increased NAD⁺ and NADH levels, indicating restoration of NAD redox homeostasis, which is necessary for mitochondrial respiration.

Sildenafil partially restores oxidative stress-induced alterations in mitochondrial abundance and respiration in capillaries

Our previous studies have determined that AAPH induces oxidative stress and changes mitochondrial function [20]. To test ex vivo capillary mitochondrial function in response to an oxidative insult, we treated isolated brain capillaries from male mice with 0.5 μM of AAPH and either 0, 0.1, 1, or 10 μM of sildenafil. We then analyzed mitochondrial content using Imaris and bioenergetics using Seahorse XFe96 Bioanalyzer. Representative micrographs of isolated cortical capillaries from mtD2 mice are shown in Fig. 4A.

Fig. 4.

Oxidative stress impairs capillary mitochondrial bioenergetics with partial restoration by sildenafil. (A) Representative confocal micrographs of isolated capillaries (mtD2-green) from control and AAPH plus sildenafil treatment groups. (B) Bar graphs represent mitochondrial volume and count in 1000 μm² capillary surface area in isolated cortical capillaries treated with AAPH (0.5 μM) and increasing concentrations of sildenafil (0–10 μM) (n = 9 mice/group). (C) Representative Seahorse mitochondrial stress test traces showing oxygen consumption rate (OCR) and extracellular acidification rate (ECAR). Oligomycin (Oligo), FCCP, rotenone/antimycin A (Rtn/AA), and 2-deoxyglucose (2-DG) were injected at indicated time points. (D) Bar graphs represent basal respiration, maximal respiration, spare respiratory capacity, and glycolytic capacity measured by Seahorse analysis in isolated cortical capillaries under the same treatment conditions as in Fig. 4B (n = 5–6 mice/group). OCR and ECAR values were normalized to protein content. Data represented as mean ± SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 by one-way ANOVA with Tukey's and Dunnett's post hoc multiple comparisons test.

Significant differences in capillary mitochondrial volume and count were observed between experimental groups (one-way ANOVA; p < 0.0001 and p < 0.0003, respectively; Fig. 4B). Post hoc analysis revealed that treatment with AAPH alone resulted in a significant reduction in both mitochondrial volume and count (p < 0.0001 and p = 0.0063, respectively). In comparison to untreated controls, capillaries exposed to 0.1 or 1 μM sildenafil in addition to AAPH exhibited decreased mitochondrial volume (p < 0.0001 and p = 0.0008, respectively) and count (p = 0.0002 and p = 0.0123, respectively). However, capillaries treated with AAPH and 10 μM sildenafil demonstrated greater mitochondrial volume than those treated with AAPH plus 0 μM or 0.1 μM sildenafil (p = 0.0397 and p = 0.0153, respectively) and did not differ in mitochondrial count or volume in comparison to controls.

Furthermore, AAPH-induced oxidative stress decreased mitochondrial respiration in cortical capillaries compared to controls. Capillaries treated with AAPH and varying concentrations of sildenafil exhibited significant differences in basal respiration, maximal respiration, spare respiration capacity, and glycolytic capacity (one-way ANOVA; p = 0.0043, 0.0006, 0.0098, and 0.0026, respectively; Fig. 4D). Post hoc analysis revealed that capillaries treated with AAPH and 0, 0.1, or 1 μM of sildenafil exhibited significantly decreased basal respiration (p = 0.0046, 0.0064, and 0.0038, respectively), whereas capillaries treated with AAPH and 10 μM sildenafil exhibited basal respiration similar to that of control levels. Additionally, in comparison to controls, capillaries treated with AAPH and 0, 0.1, 1, or 10 μM of sildenafil exhibited significantly decreased maximal respiration (p = 0.0040, 0.0008, 0.0004, and 0.0095, respectively), spare respiratory capacity (p = 0.0313, 0.0116, 0.0057, and 0.0287, respectively), and glycolytic capacity (p = 0.0021, 0.0022, 0.0071, and 0.0353, respectively).

Together these data show that oxidative stress robustly reduces mitochondrial abundance and impairs mitochondrial respiration in cortical capillaries ex vivo. While the highest sildenafil concentration tested (10 μM) was supraphysiological, this proof-of-concept data demonstrates the ability of sildenafil to partially restore mitochondrial outcomes in brain capillaries in a dose-dependent manner.

In vivo and ex vivo sildenafil treatment partially rescues rmbTBI-induced mitochondrial respiratory deficits in capillaries

To assess whether in vivo sildenafil administration functionally impacts capillary mitochondrial bioenergetics, sham and rmbTBI animals received vehicle or sildenafil (20 mg/kg) following each blast exposure. Brain capillaries were subsequently isolated at 1 day post-rmbTBI and underwent Seahorse analysis. Significant group differences were observed in basal, maximal, and ATP-linked respiration (one-way ANOVA; p = 0.0042, 0.0096, and 0.0108, respectively; Fig. 5A). Post hoc analyses revealed that basal, maximal, and ATP-linked respiration are significantly decreased in rmbTBI-Veh animals relative to shams (p = 0.0034, 0.0069, and 0.0105, respectively). No group differences were observed in proton leak or glycolytic capacity.

Fig. 5.

In vivo and ex vivo sildenafil treatment partially restores rmbTBI-induced disruption of mitochondrial respiration at 1d post-rmbTBI. (A) Bar graphs represent basal respiration, maximal respiration, ATP-linked respiration, proton leak, and glycolytic capacity measured by Seahorse analysis of cortical capillaries isolated from sham, rmbTBI-Veh, and rmbTBI-Sil groups treated in vivo. Data are represented as mean ± SEM (n = 8–11 rats/group). ∗p < 0.05, ∗∗p < 0.01 b one-way ANOVA followed by Tukey's post hoc multiple comparisons test. (B) Representative Seahorse mitochondrial stress test traces showing oxygen consumption rate (OCR) and extracellular acidification rate (ECAR). Oligomycin (Oligo), FCCP, rotenone/antimycin A (Rtn/AA), and 2-deoxyglucose (2-DG) were injected at indicated time points. (C) Bar graphs represent maximal and basal respiration measured by Seahorse analysis in isolated cortical capillaries from sham and rmbTBI animals incubated with vehicle (Veh) or sildenafil (Sil). Data are represented as mean ± SEM (n = 4 rats/group). ∗p < 0.05 by two-way ANOVA followed by Šidák's post hoc multiple comparisons test. OCR and ECAR values were normalized to protein content.

To assess whether sildenafil directly modulates capillary mitochondrial bioenergetics independent of systemic influences, we isolated brain capillaries from sham and rmbTBI animals at 1 day post-injury and incubated capillaries in either control media (Veh) or sildenafil citrate (100 μg/ml) (Sil) for 2 h prior to Seahorse analysis. Basal and maximal respiration significantly differed between sham and rmbTBI groups (two-way ANOVA; group effect, p = 0.0284 and p = 0.0126, respectively; Fig. 5C), indicating impaired capillary mitochondrial respiration following rmbTBI. Post hoc analysis showed a significant decrease in maximal respiration (p = 0.0174), as well as a trend toward reduced basal respiration in vehicle-treated capillaries from rmbTBI animals compared to shams (p = 0.0524).

Together these findings indicate that rmbTBI impairs capillary mitochondrial oxidative phosphorylation and capacity for ATP production, and that sildenafil is able to restore these bioenergetic deficiencies in vivo and ex vivo.

Sildenafil partially mitigates subacute anxiety-like behavior following rmbTBI

Open field testing was performed at 3 days pre- and 3 days post-rmbTBI, followed by elevated plus maze assessment at 7 days post-injury, to assess subacute anxiety-like behavior. No baseline group differences were observed for all open field outcomes prior to injury (one-way ANOVA; all p > 0.05). Total distance traveled differed significantly between pre- and post-injury assessments (two-way ANOVA; time effect, p = 0.0010; Fig. 6A). Within-group post-hoc analysis revealed that rmbTBI-Veh animals traveled a shorter distance post-injury compared to pre-injury (p = 0.0155), whereas sham and rmbTBI-Sil groups did not exhibit pre-to post-injury differences.

Fig. 6.

Sildenafil partially mitigates anxiety-like behavior following rmbTBI. (A) Bar graphs represent total distance moved (cm), freezing duration (s), time spent in the center zone (s), and latency to first enter the center zone (s) measured in the open field test at 3 d pre- and post-rmbTBI. (B) Representative movement tracking traces generated by EthoVision illustrating exploratory behavior in the open field from one animal per group before and after rmbTBI. (C) Representative elevated plus maze heatmaps generated by EthoVision from one animal per group. (D) Bar graphs represent the number of explorations and time spent (s) in open versus closed arms measured in the elevated plus maze at 7 d post-rmbTBI. Data represented as mean ± SEM (n = 8 rats/group). ∗p < 0.05, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 by two-way ANOVA followed by Šidák's post hoc multiple comparisons test.

Additionally, time spent not moving, or freezing duration, significantly differed between pre-to post-injury assessments (two-way ANOVA; time effect, p = 0.0006). Within-group post hoc analyses revealed increased freezing behavior in rmbTBI-Veh animals post-injury compared to pre-injury (p = 0.0130), whereas sham and rmbTBI-Sil groups did not demonstrate significant changes across time.

Furthermore, cumulative time spent in the center zone differed significantly between pre- and post-injury assessments (two-way ANOVA; time effect, p = 0.0002). Both rmbTBI-Veh and rmbTBI-Sil groups spent less time in the center zone post-injury compared to pre-injury (p = 0.0139 and p = 0.0196, respectively). No significant group differences were observed for latency to enter the center zone. Across all open field measures, no significant group effects or time × group interactions were detected. Representative movement tracings of sham, rmbTBI-Veh, and rmbTBI-Sil groups are shown in Fig. 6B.

Elevated plus maze testing was performed at 7 days post-injury. Representative heat maps of sham, rmbTBI-Veh, and rmbTBI-Sil groups are shown in Fig. 6C. Overall, animals spent less time in open arms compared to closed arms (two-way ANOVA; arm effect, p < 0.0001; Fig. 6D). Post hoc analyses revealed that this difference was significant within both rmbTBI-Veh and rmbTBI-Sil groups (p < 0.0001 and p = 0.0003, respectively). Additionally, arm exploration frequency differed by group (two-way ANOVA; group effect, p = 0.0497), with rmbTBI-Veh animals entering the open arms fewer times than shams (p = 0.0323).

Additionally at 7 days post-injury, mitochondrial outcomes, using OXPHOS staining, and expression of the tight junction protein ZO-1 were measured in isolated cortical capillaries (Fig. S3A). No significant group differences were observed among sham, rmbTBI-Veh, and rmbTBI-Sil groups.

These findings indicate that rmbTBI induces subacute anxiety-like behavior, characterized by increased freezing behavior in the open field and reduced exploration of EPM open arms, which sildenafil partially mitigated. However, sildenafil did not fully preserve the development of anxiety-like behaviors, as both vehicle- and sildenafil-treated rmbTBI animals spent less time spent in the center of the open field and in the open arms of the EPM.

Delayed sildenafil treatment restores chronic fear learning and extinction deficits following rmbTBI

Fear-related behavior was evaluated at 20 weeks post-rmbTBI using a cued fear-conditioning paradigm with a mild foot shock as the unconditioned stimulus. On the training day (Day 1), animals received a single foot shock every 60 seconds starting at 120 seconds. Cumulative freezing episodes increased across minutes on Day 1 for all groups (two-way ANOVA; time effect, p < 0.0001; Fig. 7A, left). The trajectory of freezing acquisition differed between groups (two-way ANOVA; time × group interaction, p = 0.0461), despite no overall group effect (p = 0.1825). Post hoc analyses indicated that these differences emerged in the late acquisition phase, with no between-group differences observed during minutes 1–8. At minute 9, cumulative freezing was lower in rmbTBI-Veh animals compared with shams (p = 0.0192), whereas rmbTBI-Sil animals did not differ significantly from shams.

Fig. 7.

Delayed sildenafil treatment improves fear-associated memory and glial pathology following rmbTBI. (A) Line graphs represent cumulative freeze episodes during fear acquisition on Day 1 and freeze episodes across extinction testing Days 1–4. Bar graphs represent freeze episodes measured during each day of extinction testing. Data for extinction testing were normalized to the corresponding experimental group mean on Day 1. (B) Representative confocal micrographs of isolated capillaries (ZO-1-green; OXPHOS-red; DAPI-blue) from a randomly selected animal (top row). Bar graphs represent capillary mitochondrial volume and count in 1000 μm² surface area and ZO-1 expression (middle row). Scatterplots represent Pearson correlation analyses examining the relationship between capillary mitochondrial volume/1000 μm² and ZO-1 expression in sham, rmbTBI-Veh, and rmbTBI-Sil groups. Solid lines represent simple linear regression line. Pearson correlation coefficient (r) and corresponding p values are shown on each graph (bottom row). (C) Bar graphs represent percent GFAP and IBA-1 positive area in the corpus callosum. Representative confocal micrographs (GFAP-green; IBA-1-red; DAPI-blue) of corpus callosum from sham, rmbTBI-Veh, and rmbTBI-Sil groups. Data represented as mean ± SEM (n = 11–12 rats/group). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 by two-way ANOVA followed by Šidák's post hoc multiple comparisons test.

Across the training day and three days of the fear extinction paradigm (Days 1–4), freeze episodes differed across days (two-way ANOVA; day effect, p < 0.0001) and across treatment groups (two-way ANOVA; group effect, p = 0.0009), with no significant day × group interaction (p = 0.175) (Fig. 7A, middle and right). Post hoc analyses revealed that sham animals exhibited fewer freezing episodes on Day 2 compared with Day 1 (p = 0.0031) and a further reduction by Day 4 compared with Day 1 (p < 0.0001) and Day 2 (p = 0.0205). In the rmbTBI-Sil group, freezing episodes were also reduced by Day 4 compared with Day 1 (p = 0.0063) and Day 2 (p = 0.0048). No significant day-to-day differences were detected within the rmbTBI-Veh group; however, rmbTBI-Veh animals exhibited a greater number of freeze episodes on Day 2 compared to shams (p = 0.0401).

Together, these findings indicate that rmbTBI disrupts fear learning and impairs extinction of conditioned fear, whereas sildenafil treatment following rmbTBI restores both acquisition and extinction of fear responses toward levels observed in sham animals.

Additionally, EPM testing was conducted at 1 and 3.5 months post-rmbTBI in this cohort. No group, time, or group × time interactions were observed for exploration frequency or time spent in the open or closed arms (two-way ANOVA; all p > 0.05; Fig. S4A), in contrast to the persistent fear-associated memory deficits described above.

At five months post-injury, mitochondrial outcomes, using OXPHOS staining, and expression of the tight junction protein ZO-1 were measured in isolated cortical capillaries. Representative micrographs are shown in Fig. 7B (top row). No significant differences were observed among groups in mitochondrial count, mitochondrial volume per area, or ZO-1 expression (Fig. 7B, middle row). However, Pearson correlation analysis identified a significant correlation between mitochondrial volume and ZO-1 expression in the sham (r = 0.6656, p = 0.0182) and rmbTBI-Sil (r = 0.7123, p = 0.0093) groups (Fig. 7B, bottom row). In contrast, no significant correlations were found between mitochondrial volume and ZO-1 expression in the rmbTBI-Veh group.

Given the established interplay between neurovascular dysfunction, glial responses, and behavioral outcomes following TBI, we also investigated whether rmbTBI and delayed sildenafil treatment were associated with alterations in glial immunoreactivity across brain regions. Glial fibrillary acidic protein (GFAP) and ionized calcium-binding adapter molecule 1 (IBA-1) were quantified as markers of astrocytes and microglia, respectively, in both vascular and parenchymal compartments. In the corpus callosum, no significant differences in percent GFAP-positive area were observed between groups (Fig. 7C, left). In contrast, percent IBA-1-positive area differed significantly across groups (two-way ANOVA; group effect, p = 0.0077; Fig. 7C, right), with post hoc analyses indicating increased IBA-1 expression in rmbTBI animals relative to shams across vascular and parenchymal regions (p = 0.0054). Representative micrographs illustrating GFAP and IBA-1 staining are shown in Fig. 7C. No significant group differences in GFAP and IBA-1 expression were observed in the cortex, hippocampus, and amygdala. However, across all examined brain regions, percent GFAP positive area differed significantly between vascular and parenchymal compartments, whereas IBA-1 positive area did not, indicating compartment-specific differences in astrocytic, but not microglial, distribution. Specific comparisons are shown in Fig. S5A.

Together, these findings indicate that while rmbTBI does not produce overt changes in capillary mitochondrial content or tight junction expression chronically, it disrupts the association between mitochondrial and vascular integrity, which is preserved with sildenafil treatment. Additionally, glial analyses identify the corpus callosum as a region exhibiting persistent microglial alterations following rmbTBI.

Discussion

rmbTBI is associated with persistent neurological dysfunction, and the present findings identify microvascular and metabolic dysfunction as key contributors to these outcomes. Sildenafil, a PDE5 inhibitor currently under investigation for vasculopathy following TBI, has demonstrated improved vascular outcomes in clinical TBI studies [26,27]. However, the cellular mechanisms underlying these effects remain incompletely understood, particularly in the context of bTBI. In the present study, we demonstrate that PDE5 inhibition with sildenafil produces coordinated effects across multiple levels of mitochondrial regulation following rmbTBI. Specifically, sildenafil restores hippocampal PGC-1α expression, cortical metabolic homeostasis, and vascular mitochondrial bioenergetic capacity, while also improving behavioral outcomes.

Mitochondrial dysfunction is a hallmark pathology of bTBI [[13], [14], [15],35], positioning mitochondrial recovery as a critical therapeutic target. Previous studies have shown that PDE5 inhibition with sildenafil enhances mitochondrial biogenesis by increasing mitochondrial DNA content and upregulating PGC-1α, a key regulator of mitochondrial biogenesis [28]. Activation of PGC-1α pathways has also been shown to improve mitochondrial function and behavioral outcomes following TBI [36]. In the current study, PGC-1α expression was selectively reduced in the hippocampus, but not the cortex, at 7 days following rmbTBI, and administration of 20 mg/kg of sildenafil restored this blast-induced deficit. Although no previous studies have examined PGC-1α following bTBI, our findings align with reports of reduced hippocampal PGC-1α after impact TBI [37], suggesting that hippocampal mitochondrial biogenesis pathways may represent a key area of injury and recovery. The mechanisms underlying the selectivity of reduced PGC-1α in the hippocampus compared to the cortex have not been specifically investigated; however, this selectivity may reflect differences in the timing of injury-induced mitochondrial dysfunction between regions. For example, Vekaria et al. reported that cortical PGC-1α was significantly reduced within 24–48 hours following controlled cortical impact (CCI) TBI, whereas hippocampal PGC-1α levels remained unchanged, indicating that hippocampal deficits may develop on a delayed timeline relative to the cortex [38]. This discrepancy may also reflect differences in injury biomechanics. In CCI, hippocampal injury is believed to occur secondarily to the cortical injury cascade and therefore may not be apparent within 48 hours. In contrast, blast overpressure is transmitted diffusely through the skull and brain rather than originating from a single focal point [17]. This raises the possibility that, in bTBI, the hippocampus may be directly susceptible to wave transmission rather than being secondarily affected via cortical spread. These observations align with our findings of unaffected cortex but significantly reduced hippocampal PGC-1α at 7 days post-rmbTBI, as well as previous studies demonstrating that hippocampal mitochondrial ultrastructural abnormalities after blast are most pronounced at this same time point [35].

Given the restoration of PGC-1α expression following sildenafil treatment, we next investigated whether sildenafil also affects broader mitochondrial transcriptional programs. Bulk RNA-sequencing revealed oxidative phosphorylation as one of the most strongly enriched pathways in the cortex following sildenafil treatment, suggesting that sildenafil promotes transcriptional programs associated with mitochondrial metabolism and energy production.

Interestingly, RNA-sequencing also revealed an enrichment of MYC target genes in sildenafil-treated rmbTBI animals. MYC is an oncoprotein that contributes to mitochondrial integrity by promoting oxidative phosphorylation, glycolysis, electron transport chain architecture, and mitochondrial fission and fusion. MYC also regulates glycolytic enzymes that regenerate NAD⁺ [34], suggesting a role in NAD redox homeostasis. These findings raise the possibility that sildenafil may promote mitochondrial and metabolic recovery through MYC target pathways following rmbTBI.

Consistent with gene-level alterations in metabolism, this study also investigated whether rmbTBI and sildenafil treatment affect cortical and hippocampal metabolites. While hippocampal metabolite profiles exhibited minimal alterations, more pronounced metabolic changes were observed in the cortex, a brain region our previous studies have shown is vulnerable to metabolic dysregulation following rmbTBI [15]. rmbTBI produced disruption of glycolysis, characterized by accumulation of early glycolytic intermediates (glucose-6-phosphate, fructose-6-phosphate, and fructose-1,6-bisphosphate) in the cortex. These results suggest that blast injury may hinder progression into lower glycolysis and disrupt glycolytic processes downstream of initial phosphorylation steps. These findings offer a potential mechanistic framework for clinical observations of altered brain glucose metabolism in military personnel with a history of blast exposure [11,12].

Alterations in NAD metabolism further support the effect of rmbTBI on cortical metabolism. Vehicle-treated rmbTBI animals exhibited elevated nicotinamide, a breakdown product of NAD⁺, while sildenafil treatment increased NAD⁺ and NADH levels. These metabolite alterations suggest that rmbTBI may drive NAD⁺ consumption, which could, in turn, lead to decreased cytosolic NAD⁺ and disruption of downstream NAD-dependent glycolytic enzymes such as glyceraldehyde-3-phosphate dehydrogenase [39]. These findings raise the possibility that sildenafil rescues glycolytic bottlenecking by enhancing NAD redox buffering. These findings are consistent with other studies showing that sildenafil restores NAD⁺/NADH balance in cells with mitochondrial disease [29]. Whether this treatment effect is driven by enhanced oxidative phosphorylation, a primary driver of NAD⁺ regeneration from NADH [39], or through modulation of other NAD⁺-consuming pathways remains an important question for future investigation.

In addition to glycolytic and redox alterations, rmbTBI also disrupted purine metabolism, as evidenced by elevation of the purine nucleotides xanthosine-5-monophosphate (XMP) and guanosine-5-monophosphate (GMP). Sildenafil normalized these metabolite levels, suggesting a broader role in restoring metabolic homeostasis beyond energy metabolism alone. Although the mechanisms underlying these changes remain unclear, especially as no studies examining purine metabolism following bTBI currently exist, they may reflect PDE5 inhibition-mediated effects on cyclic GMP signaling, secondary mitochondrial rescue, or purine nucleotide metabolism.

Given the known vulnerability of the cerebrovascular compartment following low-level blast exposure [8,40], we next examined whether mitochondrial dysfunction extends to the microvasculature. Previous research has demonstrated that bTBI increases oxidative stress and disrupts mitochondrial function within brain microvessels [[22], [23], [24]], and our prior work has shown that brain capillaries are particularly vulnerable to oxidative stress [32]. Consistent with these findings, oxidative stress reduced mitochondrial content and impaired mitochondrial respiration in brain capillaries, which sildenafil partially restored. Notably, in this ex vivo model, sildenafil restored mitochondrial volume, count, and basal respiration, but maximal respiration, spare respiratory capacity, and glycolytic capacity remained significantly reduced relative to controls, indicating that restoration of mitochondrial abundance does not necessarily confer full recovery of mitochondrial bioenergetic capacity. Furthermore, we demonstrated that both in vivo and ex vivo treatment with sildenafil restores rmbTBI-induced alterations in vascular mitochondrial bioenergetics. Collectively, our data suggest that brain microvasculature is particularly susceptible to injury, and that sildenafil treatment can mitigate these alterations at a systems level and through direct effects on the microvasculature. Our findings complement previous studies showing that sildenafil also improves mitochondrial respiratory capacity in vitro [28]. Overall, microvascular mitochondrial dysfunction may represent an underexplored contribution to subacute and chronic outcomes following rmbTBI.

To further examine vascular outcomes following rmbTBI, this study also investigated whether rmbTBI disrupts the blood-brain barrier (BBB). However, neither rmbTBI nor sildenafil affected brain capillary mitochondrial content and tight junction expression at subacute and chronic time points. BBB disruption is likely resolved by these stages, as previous studies have demonstrated that compromised BBB integrity is most evident at acute time points, with reductions in tight junction expression occurring within the first 24 hours post-injury [30,32,41]. Our previous work has also established that vascular alterations correspond with acute anxiety-like behavior [30,32]. Despite apparent structural recovery of the BBB, behavioral deficits such as anxiety- and fear-like responses persist long after injury [42], suggesting that neurovascular dysfunction following rmbTBI may extend beyond overt BBB breakdown. The present study demonstrates that rmbTBI induces increased anxiety-like behavior and impaired fear learning and extinction, both of which are attenuated by sildenafil at subacute and chronic time points, respectively. These results align with clinical reports indicating that military personnel with a history of cumulative occupational blast exposure experience anxiety disorders, cognitive dysfunction including memory impairment, and other long-term behavioral health conditions [[3], [4], [5]]. Furthermore, the current study identifies sildenafil as a potential therapeutic strategy to target both the pathological and neurobehavioral outcomes of rmbTBI.

Although overt vascular changes were not detected at a chronic timepoint, rmbTBI was associated with region-specific changes in glial immunoreactivity. Specifically, sildenafil normalized microglial elevation in the corpus callosum 20 weeks post-rmbTBI, consistent with previous reports of reduced neuroinflammatory markers following PDE5 inhibition in other models of brain injury [43].

While the primary focus of this study was on mitochondrial and microvascular outcomes, the transcriptomic data also provides evidence for broader neuronal effects of sildenafil treatment. RNA-sequencing revealed that genes differentially expressed between untreated and sildenafil-treated rmbTBI animals were enriched for processes related to regulation of synapse structure, postsynaptic organization, and regulation of synapse organization. Similarly, transcriptomic analyses in an Alzheimer's disease model have shown that sildenafil alters expression of genes involved in synaptic and neurodevelopmental signaling pathways [44], suggesting this may be a broader transcriptional signature of sildenafil treatment across disease contexts. These findings align with prior research showing that sildenafil preserves and promotes synaptic function in models of stroke, Alzheimer's disease, and vascular dementia [44,45]. Furthermore, emerging evidence suggests that sildenafil may facilitate neuronal recovery through mitochondrial mechanisms. For instance, multiple studies have demonstrated that inhibition of ATP-sensitive potassium channels reverses the protective effects of sildenafil, including attenuation of 3,4-methylenedioxy-methamphetamin administration (MDMA)-induced neurotoxicity [46] and beta-amyloid peptide-induced neuronal mitochondrial calcium overload, neuronal death, decreased ATP production, and increased ROS production [47]. Additionally, a recent study by Zink et al. using patient-derived models of Leigh syndrome, a mitochondrial disorder, found that sildenafil improves disrupted neuronal calcium dynamics, development, and survival [29].

While our study highlights the impact of rmbTBI and sildenafil treatment on microvascular and behavioral outcomes in males, our previous work has identified sex differences in vascular integrity and anxiety-like behavior between males and females acutely following rmbTBI [30,32]. Future investigations should include female animals to determine whether females exhibit similar alterations in metabolism and behavior after rmbTBI at subacute and chronic time points. In addition, although we demonstrate that PDE5 inhibition modulates mitochondrial and metabolic function, we did not directly assess its established signaling mechanism: inhibition of cyclic GMP breakdown and subsequent cellular signaling leading to vasodilation. Therefore, our ability to determine whether these mitochondrial changes are mediated through vascular structure and blood flow changes or by direct cellular mechanisms is limited.

Overall, our findings suggest that sildenafil promotes mitochondrial recovery across multiple biological levels, spanning mitochondrial regulation, transcriptional reprogramming, cellular metabolism, and mitochondrial bioenergetic function. Taken together, these findings position metabolic and microvascular dysfunction as central drivers of injury progression following rmbTBI. By restoring mitochondrial function, metabolic disruptions, and microvascular health, sildenafil produces coordinated effects that span from molecular pathways to behavioral outcomes. Collectively, these results highlight sildenafil as a potential therapeutic strategy for targeting the underlying mechanisms of rmbTBI.

Author contributions

Kilgore MO: performed research, analyzed data, wrote first draft of the paper, edited the paper, wrote the paper; Ye S: performed research, analyzed data, edited the paper; Oyelami F: analyzed data, edited the paper; Alareidi A: performed research; Gard J: performed research; Johnson LA: designed research, edited the paper; Velmurugan GV: designed research, performed research, analyzed data, edited the paper; Hubbard WB: designed research, analyzed data, edited the paper.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

W. Brad Hubbard reports financial support was provided by US Department of Veterans Affairs. W. Brad Hubbard reports financial support was provided by National Institute of General Medical Sciences. W. Brad Hubbard reports financial support was provided by US Department of Defense. W. Brad Hubbard reports financial support was provided by Kentucky Spinal Cord and Head Injury Research Board. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article.