Role of peptidylarginine deiminase 2 in a murine model of traumatic brain injury

Abstract

INTRODUCTION:

Traumatic brain injury (TBI) is among the leading causes of death and disability worldwide. Studies have linked peptidylarginine deiminases (PADs) with TBI outcomes. However, nonspecific PAD inhibition makes it difficult to decipher the exact role of specific PAD enzymes in neurotrauma. Since both PAD2 and PAD4 have been linked with neurodegeneration, we sought to clearly establish their roles in TBI.

METHODS:

Male mice (11–14 weeks) were subjected to controlled cortical impact TBI (n = 5/group). Experimental groups included wild-type plus TBI (WT-TBI), PAD2 knockout plus TBI (PAD2-KO–TBI), PAD4 knockout plus TBI (PAD4-KO–TBI), and PAD2/4 double knockout plus TBI (PAD2/4-DKO–TBI). Twenty-four hours post-TBI, frozen brain sections were stained (Nissl and immunofluorescence) to determine lesion size and expression of PAD2 and PAD4. We also assessed the impact of PAD2-KO on neurologic severity scores (1–8 days post-TBI) and visuospatial learning using the Morris water maze test (21–30 days post-TBI).

RESULTS:

Overall, PAD2-KO–TBI and PAD2/4-DKO–TBI animals displayed significantly smaller brain lesion sizes than WT-TBI (p = 0.005 and 0.005, respectively) and PAD4-KO–TBI (p = 0.005 and 0.004, respectively). However, there was no significant difference in lesion size between the PAD4-KO–TBI and WT-TBI (p = 0.880) groups. Analysis of archived snRNA-seq data and immunofluorescence staining 24 hours post-TBI showed upregulation of PAD2 (primarily in astrocytes) in WT-TBI compared with sham (p = 0.048), whereas PAD4 was undetectable. Overall, PAD2-KO–TBI had a significantly lower neurologic severity score on postinjury days 1 to 6 compared with WT-TBI (all p < 0.05) group. Moreover, Morris water maze test demonstrated that the cumulative cued and noncued spatial learning was worse in WT-TBI compared with PAD2-KO–TBI (p < 0.05) animals.

CONCLUSION:

Our results suggest that PAD2, but not PAD4, blockade can improve outcomes following TBI, which justifies its exploration as a potential target for novel neuroprotective therapies.

Traumatic brain injury (TBI) is among the leading causes of death and disability worldwide. It is responsible for approximately 3 million emergency department visits, hospitalizations, and deaths, annually.^1–3 Despite extensive research efforts over the past 100 years, few effective therapies for TBI exist.⁴ The etiology of TBI is twofold: a primary insult and a secondary cascade of brain injuries that are multifactorial and may last for years after impact.^5,6 Mechanistically, several factors contribute to the secondary injuries, including oxidative stress, lipid peroxidation, neuroinflammation, axon degeneration, and apoptotic cell death.⁶ These secondary insults should be explored as targets for therapeutic interventions.⁷

Peptidylarginine deiminases (PADs) are a family of calcium-dependent posttranslational modification enzymes that convert arginine to citrulline in proteins. Of five known isozymes (PAD1–4 and PAD6), each exhibits distinct tissue and cell-type specificity. PAD1 is absent from the brain, while PAD3 and PAD6 are present in the CNS but have unclear roles in neurodegeneration.^8,9 In contrast, PAD2 and PAD4 are strongly linked to neurodegenerative processes, with PAD2 abundantly expressed in CNS cells including oligodendrocytes, microglia, and astrocytes.¹⁰ Recent studies suggest a prominent role of PAD2 in catalyzing neurodegeneration in various neuroinflammatory disorders, such as multiple sclerosis (MS), amyotrophic lateral sclerosis, and Alzheimer’s disease,^10–12 through protein citrullination. Nonetheless, its implication in TBI remains unexplored.

Despite its abundance in the CNS, minimal research has been conducted to understand the impact of PAD2 in neurorecovery post-TBI. Previous studies have implicated the PAD family in TBI, but the exact contribution of specific PAD isoforms (particularly PAD2) remains elusive. Studies have shown that post-TBI, pan-PAD inhibition reduces neuronal apoptosis and improves short-term neurological functions.^13–15 These studies attributed these benefits to the PAD4 isoform, primarily because of its important role in neutrophil extracellular trap formation and close implication in brain edema. However, non-specific inhibitors such as Cl-Amidine cannot rule out contributions from other isoforms, notably the CNS-enriched PAD2. Because of the widespread distribution of PAD2 in the brain and its prominent neurodegenerative role in inflammatory disorders, it is likely that PAD2 is associated with accentuation of lesion size and reduction of cognitive recovery post-TBI. In the current study, using transgenic mouse strains and a well-established model of TBI, we sought to investigate the role of PAD2, hypothesizing that its aberrant upregulation exacerbates lesion development and impairs functional and neurocognitive recovery post-TBI.

MATERIALS AND METHODS

Animals and Ethics

Adult male mice (C57BL/6j background), 11 to 14 weeks at the time of surgery were used. C57BL/6j mice were used as wild type (WT; Jackson Labs, Bar Harbor, ME). PAD2 knockout (PAD2-KO), PAD4 knockout (PAD4-KO), and PAD2PAD4 knockout (PAD2/4-DKO) mice were provided by an external collaborator and bred in-house (Supplemental Digital Content, Supplementary Fig. 1, http://links.lww.com/TA/E908).^16,17 All genetically modified mice underwent polymerase chain reaction and genetic testing (Supplemental Digital Content, Supplementary Fig. 2, http://links.lww.com/TA/E909).¹⁸ All mice were individually housed in an enclosure with controlled temperature and humidity and a standard 12-hour light/dark cycle. The mice had access to food and water ad libitum. The experimental protocol was approved by the Institutional Animal Care and Use Committee. The study adhered to the Animal Research; Reporting InVivo Experiments guidelines (Supplemental Digital Content, Supplementary Data 1, http://links.lww.com/TA/E910).

TBI Model

Controlled cortical impact (CCI) has been widely used as a reliable model of TBI.^13,19 All WT and genetically modified mice underwent CCI TBI. We included four sham groups: wild-type sham (WTsham), PAD2-knockout sham (PAD2-KOsham), PAD4-knockout sham (PAD4-KOsham), and PAD2/4 double knockout sham (PAD2/4-DKOsham), which did not undergo any injury. Briefly, all mice (including sham) were anesthetized with isoflurane and provided with adequate analgesics. Traumatic brain injury mice underwent a 4-mm craniectomy located in the right parietal bone midway between the bregma and lambda, centered 2 mm lateral from the sagittal suture with the dura matter intact. The bregma levels were held constant for all groups (levels ranging between −0.35 mm and −4 mm). Traumatic brain injury was induced by a digital electromagnetic CCI device (David Kopf Instruments, Tujunga, CA) at a depth of 1 mm and a velocity of 4.5 m/s over a 100 m/s dwell time. The incision was closed, and mice were placed in heated cages until recovery post-TBI. Twenty-four hours after TBI, mice were euthanized per protocol, organs were perfused with normal saline, and brains were embedded in optimal cutting temperature medium (Sakura, Oakland, CA). Coronal sections (10 μm thickness) were cut using a cryostat at −20°C and mounted onto poly-l-lysine-coated slides. A total of 10 sections were taken 300 μm apart to ensure that the entire lesion side had been obtained. Mice were euthanized before the endpoint if they could not tolerate TBI.

Nissl Staining

Nissl staining was used to estimate neuronal damage and delineate the lesion size. Frozen brain sections were rehydrated in a graded series of alcohol and treated with Cresyl violet for 10 minutes.²⁰ Lesion size was measured by two independent team members. The area of damaged neurons was estimated using ImageJ software (Wayne Rasband, version1.46r; Bethesda, MD) for 10 consecutive brain sections, and the contralateral cerebral area acted as an internal control. We obtained the total lesion volume by summing up the area of each slice and then multiplying it to the slice thickness (10 μm). Since each section was 300 μm apart, we multiplied the lesion volume at each section with 300.

Immunofluorescence Staining

Brain cryosections were washed with phosphate buffered solution, postfixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100 (Sigma Aldrich, St. Louis, MO). Following 1 hour incubation with blocking buffer (3% bovine serum albumin) at room temperature, the tissue sections were then incubated overnight at 4°C with primary antibodies against proteins including citrullinated histone 3 (CitH3; 1:200, designed in house, as described previously²¹), PAD2 (3:200, Protein Tech, 66386-1-Ig), PAD4 (1:200, Protein Tech, 17373-1-AP), caspase-11 (1:200, Abcam, Ab 180673), and glial fibrillatory acidic protein (GFAP; 1:100, Thermo Fisher, 13-0300). Thereafter, the sections were incubated with the appropriate secondary antibodies (1:1000, Goat anti-Mouse IgG2b, A21141; 1:1000, Donkey anti-rabbit Jackson ImmunoResearch, 711-545-152; 1:500, Goat anti-human Jackson ImmunoResearch, 109-545-088; and 1:1000, Goat anti-rat Jackson ImmunoResearch, 112-605-003) for 1 hour at room temperature in the dark. Following counter-staining with 4′,6-diamidino-2-phenylindole (DAPI; Thermo Fisher, 815-968-0747), images were acquired with an inverted fluorescence microscope (Olympus, Tokyo, Japan). The data were analyzed from randomly selected microscopic fields using the ImageJ program (version1.46r).

CitH3 Enzyme-Linked Immunosorbent Assay

A “sandwich” enzyme-linked immunosorbent assay, which has been described previously, was used.²² The specific methodology of building the enzyme-linked immunosorbent assay plate can be found on paragraph A of Supplemental Digital Content (Supplementary Data 2, http://links.lww.com/TA/E911).

Analysis of snRNA-Seq Data

To investigate the particular cell type wherein PAD2 and PAD4 might exert significant influence, we analyzed two publicly available data sets (GSE230253 and GSE198074) from the Gene Expression Omnibus database, encompassing single-cell sequencing data from mouse control and TBI samples.^23–25 In both data sets, 7-week-old male mice underwent blast-induced TBI, and cells in the brain were assessed 48 hours post-TBI for PAD2 and PAD4 expression. One data set assessed cells in the hippocampus, while the other assessed the cell population of the subventricular zone. These data sets differed from our study in TBI mechanism. We elected to use them because the blast model TBI induces more diffuse injury affecting areas other than the cortex like the hippocampus that contains abundance of cells and is an important site for secondary brain injury.

The preprocessing, normalization, data scaling, and cell clustering of the aforementioned single-cell sequencing data sets were carried out using Seurat (version 5.0.3, Wilmington, MA). For cluster identification, we used Single R package (version 2.0.0, BMW Motorrad, Berlin, Germany) to identify specific cell populations and then assessed the expression of PAD2 and PAD4 across different cell types in these two data sets.

Long-term Neurological Outcomes

Desired mice experimental groups were survived up to 31 days post-TBI, and their motor, sensory, and cognitive recovery were assessed.

Neurologic Severity Scores

To evaluate the neurological motor function deficit in mice after TBI, we used a previously established neurologic severity score (NSS) model (Supplemental Digital Content, Supplementary Table 1, http://links.lww.com/TA/E912).²⁶ This score is a composite of motor, sensory, reflex, and balance tests. The NSSs ranged from 0 (no deficit) to 15 (comatose) and was monitored on each postinjury day by one blinded and one unblinded reviewer. Baseline scores were obtained from uninjured mice (WTsham and PAD2-KOsham).

Morris Water Maze Exercises

After a recovery period of 20 days, learning and memory were evaluated in a Morris water maze test using previously published protocols.²⁷ Briefly, experiments consisted of three types of swimming trials to assess differential learning and memory: (1) cued learning trials (days 21–23 post-TBI), (2) spatial learning trials (days 26–30 post-TBI), and (3) probe memory trials (days 29–30 post-TBI). All swimming trials were conducted and scored by the same operator, at the same time of day on all experiment days incorporating a computerized video-tracking and recording system over the pool to facilitate analysis (Actimetrics WaterMaze software; Actimetrics, Wilmette, IL). The video-tracking system collected data on the time taken to reach the platform (Latency to given zone) and average swimming velocity for the cued and spatial trials. For the probe trial, the duration spent in the desired northwest zone (area where the hidden platform was previously located) was recorded. Specific experimental details for each water maze experiment can be found on paragraph B of Supplemental Digital Content (Supplementary Data 2, http://links.lww.com/TA/E911).

Statistical Analyses

The data are expressed as the mean ± SE and were analyzed with GraphPad Prism software (version 8.1.2; GraphPad Software, San Diego, CA). Data normality of every group was tested by the Shapiro-Wilk test. Multiple comparisons were analyzed by one-way analysis of variance followed by Tukey’s multiple comparison test. When comparing two groups, unpaired Student’s t test or Welch’s t test were performed. A p value of <0.05 was considered significant. For immunofluorescence, % cells stained positive was calculated using the following formula: (cells stained positive for protein/cells stained positive for DAPI) × 100.

Sample size for the primary outcome-lesion size was calculated using a power of 90%, large, expected effect size (Cohen’s d > 0.8), and p value <0.05, with the goal of using the lowest numbers to achieve our objective (Stata v.14.2, StataCorp LLC, TX, USA). We found that a minimum of five mice per group would be needed to detect a significant p value for lesion size and eight mice per group for the long-term neurobehavioral outcomes. The larger sample size for behavioral studies is due to the fact that outcomes are dependent on a range of factors including circadian rhythm, environment, pheromones, and genetic background.^28,29 Accounting for the inherent variability in mouse behavior often justifies the use of larger sample sizes.^28,29

RESULTS

Hypothesis Confirmation

PAD2 Expression Is Aberrantly Upregulated in CNS, Primarily in Astrocytes, Post-TBI

To investigate the dynamic changes in PAD2 expression following TBI, we analyzed the snRNA-seq data from publicly archived data sets of murine TBI. Following cell type clustering, we observed abundant PAD2 mRNA level and significant upregulation in multiple glial cell populations post-TBI, particularly in astrocytes (Supplemental Digital Content, Supplementary Fig. 3, http://links.lww.com/TA/E913).^23,24 In contrast, PAD4 mRNA was barely detectable (Supplemental Digital Content, Supplementary Fig. 4, http://links.lww.com/TA/E914). This is consistent with its myeloid specificity and existing data under healthy state from the Human Protein Atlas.³⁰ Since the TBI model used in the snRNA-seq was different from our experimental model, we validated these results using immunofluorescence staining. Consistent with the bioinformatics finding, immunofluorescence staining also showed a significant increase in PAD2 post-TBI (p = 0.02) (Fig. 1), with 64.8% of PAD2 positive cells costained with the astrocytic marker protein GFAP (Supplemental Digital Content, Supplementary Fig. 3, http://links.lww.com/TA/E913). In contrast, no discernable PAD4 staining could be observed in neither sham nor post-TBI samples (Supplemental Digital Content, Supplementary Fig. 4, http://links.lww.com/TA/E914, and Supplementary Fig. 5, http://links.lww.com/TA/E915).

Figure 1.

Immunofluorescence shows a significant increase in PAD2 staining 24 hours following TBI. (A) Representative images of immunofluorescence staining in mouse brain sections. Images were taken with Nikon confocal microscope (Nikon Instruments Inc, Tokyo, Japan); (B) bar graph to represent staining. *Significant increase in PAD2 staining in WT compared with sham. DAPI, 4′,6-diamidino-2-phenylindole.

Comparison of Lesion Size 24 Hours After TBI

PAD2-KO and PAD2/4-DKO Are Associated With a Significantly Smaller Lesion Size

Next, we sought to dissect the individual roles of PAD2 and PAD4 in the progression of brain lesions. Following TBI, only PAD2-KO and PAD2/4-DKO but not PAD4-KO mice displayed a significantly smaller lesion size than WT (Fig. 2). The lesion sizes between PAD2-KO and PAD2/4-DKO were comparable. Considering the lack of phenotype in the PAD4-KO mice, we contend that PAD2 was the primary contributor to the post-TBI lesion development. Lesion size of mice brains span from the cortex to the hippocampus, and the pattern observed was similar (higher lesion size in WT and PAD4-KO and lower in PAD2-KO and PAD2/4-DKO).

Figure 2.

Nissl staining of brain sections 24 hours after TBI. (A) Nissl-stained images. (B) Representative bar graphs showing lesion size. Images taken using EVOS M7000 (Life Technologies Corporation, Washington, USA). Wild-type–TBI had a significantly higher lesion size than PAD2-KO–TBI and PAD2/4DKO–TBI; PAD4-KO–TBI had a significantly higher lesion size than PAD2-KO–TBI and PAD2/4DKO–TBI; no difference in lesion size between WT-TBI and PAD4-KO–TBI groups and PAD2-KO and PAD2/4-DKO groups respectively. ***p < 0.001. L, lesion.

Downstream Mechanism of PAD2 in the CNS

PAD2-KO Has No Impact on Perilesional and Plasma CitH3 Levels

To better understand the downstream mechanistic pathways of PAD2, we first examined the impact of PAD2 and/or PAD4 loss-of-function on CitH3 levels in the brain and circulation post-TBI, given their reported roles in driving CitH3 and extracellular trap formation. Consistent with prior reports,¹³ TBI led to significant increase in perilesional CitH3 levels in WT and perilesional and plasma CitH3 in PAD2-KO. This was not observed in PAD4-KO or PAD2/4-DKO groups (Supplemental Digital Content, Supplementary Fig. 6, http://links.lww.com/TA/E916). This suggests that the observed neuroprotection in PAD2 loss-of-function animals was likely not mediated through modulation of CitH3 and extracellular trap formation, systemically or locally.

Impact of PAD2-KO on Secondary Brain Injury

To investigate the noncanonical functions and the role of PAD2 in neuroinflammation in TBI, we conducted a caspase-11 immunofluorescence staining. Caspase-11 plays a role in the inflammatory response following TBI by contributing to the maturation of inflammatory cytokines. Post-TBI, caspase-11 was significantly reduced in PAD2-KO compared with WT (p = 0.012). Moreover, PAD4-KO had a significantly higher caspase-11 staining than all three TBI experimental groups (Supplemental Digital Content, Supplementary Fig. 7, http://links.lww.com/TA/E917). Thus, the downstream mechanisms of PAD2 are likely related to its impact on noncanonical functions in the brain.

Long-term Neurologic and Functional Outcomes in PAD2-KO TBI Mice

Since PAD2-KO but not PAD4-KO exhibited a protective phenotype in comparison with WT control acutely (24 hours) post-TBI, we decided to investigate the impact of PAD2-KO on long-term neurological and neurobehavioral recovery. Wild-type and PAD2-KO mice were survived for 31 days post-TBI.

As demonstrated in Figure 3, our experimental TBI model resulted in significant motor and sensory deficits in both WT and PAD2-KO mice post-TBI, as evidenced by marked impairment in the beam walk performance observed on the first postinjury day. As animals gradually recovered from the acute injury, PAD2-KO mice exhibited significantly improved NSSs in comparison with WT from postinjury days 1 to 6. Since none of the sham groups underwent injury, all had an NSS of 0 on all days (not shown in graph)

Figure 3.

Comparison of NSSs between WT-TBI and PAD2-KO–TBI mice post-TBI. Significantly higher NSS in WT-TBI compared with PAD2-KO–TBI (*p< 0.05, **p< 0.01, ***p < 0.001). PID, postinjury day.

During the cued learning trial, WT and PAD2-KO mice in the sham control groups demonstrated comparable performances across all 3 days. In contrast, TBI significantly lowered the cumulative speed and prolonged the time to reach the platform only in the WT group but not in PAD2-KO mice on trial day 1. This suggests that inhibition of PAD2 results in enhanced recovery of visual cognitive memory post-TBI (Supplemental Digital Content, Supplementary Fig. 8, http://links.lww.com/TA/E918).

During the spatial learning phase with hidden platforms, WT mice demonstrated significantly poor learning compared with the rest of the experimental subgroups following TBI. In contrast, PAD2-KO mice demonstrated a very similar learning pattern with both WTsham and PAD2-KOsham (Fig. 4). By the end of spatial learning trials, wild type plus TBI (WT-TBI) mice had the worst overall cumulative spatial learning (20.8 ± 9.1 seconds) compared with PAD2 knockout plus TBI (PAD2-KO–TBI; 8.3 ± 2.6 seconds, p = 0.002), WTsham (10.0 ± 2.5 seconds, p = 0.019), and PAD2-KOsham (9.7 ±4.1 seconds, p = 0.01), without significant differences between the latter three groups. Of note, during the spatial learning trial, WT-TBI mice also had a significantly lower cumulative speed on trial days 1 to 3 compared with shamPAD2-KO (p < 0.05 for all comparisons). However, there was no significant difference in speed between PAD2-KO and any of the sham groups (not shown in figure).

Figure 4.

(A) Spatial trials of Morris water maze test conducted 26 to 30 days post-TBI. (B) Representative images of paths taken by WT-TBI and PAD2-KO–TBI on the last day of the spatial trial. Wild-type–TBI mice took a significantly longer duration to find hidden platform compared with PAD2-KO–TBI mice (**p < 0.01); #WT-TBI mice took a significantly longer duration to find hidden platform compared with WTsham mice (#p < 0.05, ##p < 0.01); §WT-TBI mice took a significantly longer duration to find hidden platform compared with PAD2-KOsham mice (§§p < 0.01, §§§p < 0.001).

Probe trials disclosed no differences between the experimental groups in swimming velocity or distance, nor time to reach platform’s prior location.

DISCUSSION

Previous studies implicate PAD-mediated protein citrullination in neurodegeneration and secondary brain injury after TBI, with most attention focused on the myeloid-enriched isoform PAD4. PAD4 catalyzes CitH3 and drives neutrophil extracellular trap formation.^{10,12,31–33} In contrast, PAD2, although also capable of triggering CitH3, is predominantly expressed in CNS cells and has been linked to neurodegeneration through both histone and nonhistone protein citrullination.^10–12 Our study shows that only PAD2, but not PAD4, was significantly upregulated 24 hours following TBI. Moreover, PAD2-KO and PAD2/4-DKO mice similarly showed approximately 40% smaller lesions than their WT and PAD4-KO counterparts post-TBI. This acute time point was selected based on prior work showing maximal group differences at 24 hours in clinically relevant TBI models, although other studies report larger lesions at later time points (e.g., 72 hours).³⁴ In addition, PAD2-KO mice also demonstrated improved longterm motor and sensory functions and visuospatial memory compared with WT control after TBI. These findings suggest that the neuroprotection observed with pan-PAD inhibitors may primarily result from blockade of PAD2 rather than PAD4. In addition, despite the canonical functions of PAD2 and PAD4 in triggering CitH3, CitH3 levels did not decrease, both in circulation and in the brain, in PAD2-KO–TBI. This suggests that PAD2’s role in TBI is arbitrated through mechanisms unrelated to CitH3-mediated extracellular trap formation.

PAD2 catalyzes the citrullination of various cerebral proteins under hypoxic conditions,³⁵ and abnormal accumulation of citrullinated proteins has been linked to Alzheimer’s disease and MS.³⁶ However, its specific role in TBI has not been established. Our study shows that PAD2 expression increases significantly following TBI, particularly in glial cells like astrocytes and microglia, mirroring the patterns seen in Alzheimer’s disease and MS.^36,37 Previous in vitro studies have demonstrated a significant upregulation of PAD2 in astrocytes subjected to mechanical stretching.³⁸ Given astrocytes’ key roles in debris clearance and CNS homeostasis, PAD2-mediated disruption of these functions may slow recovery and contribute to larger lesion size.³⁹

Although PAD2 could theoretically contribute to neurodegeneration via CitH3, we observed no differences in perilesional or plasma CitH3 levels between WT-TBI and PAD2-KO–TBI, suggesting that PAD2 is neither a major source of CitH3 nor the primary driver of its neuroprotective effects. This also raises the possibility that PAD2 acts through nonhistone citrullination in the context of TBI. Notably, perilesional CitH3 was higher in PAD2-KOmice post-TBI, yet lesion size was significantly larger in PAD4-KO group, indicating a lack of correlation between CitH3 levels and lesion size. While prior studies linked elevated CitH3 to worse neurological outcomes after TBI,^13,15 they used nonspecific PAD inhibitors targeting both PAD2 and PAD4, potentially lowering CitH3 and improving outcomes simultaneously. Consistent with this, we found that PAD2/4-DKO–TBI mice both reduced perilesional-plasma CitH3 and a smaller brain lesion size than WT-TBI. This suggests that pan-PAD inhibition may be responsible for both factors.¹³ However, the primary goal of this study was to understand the role of PAD isoforms in TBI rather than the impact of CitH3 on lesion size, and our CitH3 findings, contradictory to previous reports, warrant further investigations.

We also attempted to delineate the role of PAD2 in neuroinflammation via understanding its association with caspase-11, an important mediator of apoptosis and pyroptosis.⁴⁰ Our study showed a significant decrease in caspase-11 in PAD2-KO–TBI compared with WT-TBI and PAD4-KO–TBI. Previous studies have shown that caspase-11 plays a role in apoptosis following brain ischemia.^40,41 Another potential mechanism for PAD2-mediated brain injury could be pyroptosis-driven neuroinflammation. While pyroptosis and the role of caspase-11 have been studied in depth for sepsis and infection, their role in TBI warrants further investigation.⁴² A recent study by Yu et al.⁴³ on the downstream mechanisms of PAD2 in pneumonia-induced acute lung injury showed that PAD2 inhibition is associated with phenotypic change from M1 macrophages, which are pro-inflammatory, to M2 macrophages, which are anti-inflammatory. The potential mechanism suggested was PAD2 mediated for M1 polarization via citrullination and hence activation of NF-κB.⁴³ In the brain, NF-κB signaling system increases the acute posttraumatic mortality rate, worsens the neurological outcome, and promotes neuronal cell death by apoptosis.⁴⁴ Thus, an important next step in understanding the downstream mechanism of PAD2-mediated neurodegeneration is to understand its impact in NF-κB–mediated neuroinflammation.

Traumatic brain injury is associated with neurocognitive decline persisting up to 6 months, and no pharmacologic therapies currently exist to mitigate this.⁴⁵ In our study, PAD2-KO mice showed significantly improved spatial memory at 30 days postinjury, performing comparably with sham mice in the Morris water maze assay. This suggests that PAD2 deletion promotes substantial neurological recovery and accelerates return to baseline function. Strikingly, in the sham groups, PAD2-KO mice also performed similarly to WT, indicating no inherent motor or cognitive deficits from PAD2 loss and that observed benefits were injury specific. While prior work with the pan-PAD inhibitor Cl-amidine reported improved motor and sensory function, long-term neurocognition was not assessed.¹³ Our findings support further evaluation of PAD2-selective inhibitors for enhancing post-TBI neurocognitive recovery.

Although spatial memory improved significantly in PAD-KO group, cued and memory learning patterns were similar across all groups, possibly because visual memory had recovered by the time of testing. Indeed, visual and spatial memory recover along different trajectories, with visual working memory often rebounding faster after mild TBI.⁴⁶ As such, earlier testing of cued memory may better reveal PAD2’s impact. No differences were detected in the probe trial, consistent with previous TBI studies testing other interventions.⁴⁷ This lack of difference in the probe trials may reflect methodological factors such as variations in mouse positioning or data collection between probe and spatial trials.

Previous studies have shown that inhibition of PAD2 decreases mortality in a murine model of lethal hemorrhagic shock, suggesting a broader role for PAD2 in trauma.⁴⁸ Thus, a logical next step is to test selective PAD2 inhibitors in clinically relevant swine models of TBI, hemorrhagic shock, or polytrauma.^49,50

This study has certain limitations. First, mouse models may not fully represent the complexity of the disease in humans and lack the genetic diversity found in human populations. Consequently, results obtained from mouse models may not fully apply to humans. Second, the exact downstream mechanism of PAD2 in the CNS has not fully been elucidated. Future studies are warranted to establish the molecular basis of PAD2-mediated neurodegeneration, possibly with the involvement of nonhistone citrullination as reported in the field of cancer research.⁵¹ Third, because of dearth in publicly available data sets for TBI, for SN-RNA Seq, we used data sets from studies that incorporated a blast-model injury of TBI. Controlled cortical impact and blast injury in mice are nearly incomparable histologically and neurofunctionally. Moreover, while our bioinformatics-guided approach unveiled a potential key role of PAD2 dysregulation in glial cells, we cannot exclude the potential contributions from other cell types such as neutrophils. Previous research has demonstrated that neutrophils infiltrate the cortical brain tissue in the early stages (between 4 and 72 hours postinjury) and contribute to the lesion progression.⁵² Considering PAD4’s specific expression in myeloid cells, future studies at later time points are warranted. In addition, while the findings of PAD2-mediated neurodegeneration are novel, the fact that fetal knockout of PAD2 and PAD4 and its effect on neurodevelopment cannot be ruled out. Thus, our next project assessing the neurobehavioral outcomes after using specific PAD2 inhibitors such as AFM41a post-TBI can help us understand this phenomenon better.⁴⁹ Finally, we did not investigate the impact of PAD4-KO on long-term neurological recovery, and we acknowledge its potential role in the chronic sequelae of TBI. In the long-term neurological outcomes experiment, we decided to focus on PAD2-KO, as this group had showed the most promising results in the short-term study. However, it is possible that PAD4-KO animals, while not showing a reduction in brain lesion size, could have shown an improvement in long-term neurologic outcomes. Thus, future studies are warranted to understand the implication of PAD4 in post-TBI cognitive impairment.

CONCLUSION

In the current study, we identified PAD2 isoform to be involved in the development and progression of brain injury following CCI. While previous studies suggest a neuroprotective role of pan-PAD inhibition, our data suggest that inhibition of PAD2, instead of the widely studied PAD4, explains the observed benefits. These neuroprotective functions are likely through a mechanism different than its canonical functions independent of CitH3 and extracellular trap formation. These findings suggest that PAD2 is a promising therapeutic target in the setting of TBI and support additional mechanistic studies to elucidate its precise role.

Supplementary Material

Supplemental digital content is available for this article. Direct URL citations appear in the printed text, and links to the digital files are provided in the HTML text of this article on the journal’s Web site (www.jtrauma.com).