Visual Outcomes in Experimental Rodent Models of Blast-Mediated Traumatic Brain Injury

Blast-mediated traumatic brain injuries (bTBI) cause long-lasting physical, cognitive, and psychological disorders, including persistent visual impairment. No known therapies are currently utilized in humans to lessen the lingering and often serious symptoms. With TBI mortality decreasing due to advancements in medical and protective technologies, there is growing interest in understanding the pathology of visual dysfunction after bTBI. However, this is complicated by numerous variables, e.g., injury location, severity, and head and body shielding. This review summarizes the visual outcomes observed by various, current experimental rodent models of bTBI, and identifies data showing that bTBI activates inflammatory and apoptotic signaling leading to visual dysfunction. Pharmacologic treatments blocking inflammation and cell death pathways reported to alleviate visual deficits in post-bTBI animal models are discussed. Notably, techniques for assessing bTBI outcomes across exposure paradigms differed widely, so we urge future studies to compare multiple models of blast injury, to allow data to be directly compared.


INTRODUCTION
Cases of traumatic brain injury (TBI) morbidity are increasing as people are more often surviving blast-mediated TBI (bTBI), an injury especially prevalent among military personnel. Over the past two decades, 417,503 U.S. service members sustained at least one TBI as active military (Defense and Veterans Brain Injury Center, 2020) with nearly 2/3 involving an explosive blast (McKee and Robinson, 2014). Unfortunately, bTBIs are typically classified as mild due to the lack of obvious acute macroscopic injury; consequently, affected service members often return to duty prematurely (McKee and Robinson, 2014;Bryden et al., 2019;Regasa et al., 2019). Many neuropathological processesmicrovascular injury, axonal injury, and neuroinflammationcan appear in the days to weeks after blast and have longterm effects on physical, cognitive, and emotional health (Hernandez et al., 2018).
Visual impairments are reported by some 75% of TBI patients, including blurry/double vision, difficulties reading, light sensitivity, and decreased peripheral vision (Armstrong, 2018;Frick and Singman, 2019). These visual impairments can arise due to optic neuropathy, axonal injury, and the loss of retinal ganglion cells (RGCs), which transmit visual stimuli to higher-level processing centers in the brain (Sen, 2017). In animal studies, decreased RGC survival and axonal integrity are strongly implicated with the activation of microglia and macrophages, with unregulated oxidative stress further contributing to RGC loss and optic nerve degeneration (Wang et al., 2013;Gupta et al., 2019).
Technological advancements in protective body armor and headgear have improved survival in combat, producing survivors with an increased number of co-morbidities. Polycarbonate eye protection does reduce the number of penetrating eye injuries, but does not prevent closed-globe damage to the eyes after a blast (Cockerham et al., 2011). Additionally, this protective gear in particular is not always worn, as dust and sweat can accumulate and reduce visibility, leaving the eye susceptible to injury (Cockerham et al., 2009). Furthermore, while many types of TBI produce visual impairments in humans, the variations in injury mechanics and pathophysiology necessitate studying blast-induced visual damage as its own entity. Experimental models of bTBI are critical in studying the mechanisms driving visual pathologies and can be utilized to identify and test novel therapeutic targets to prevent long-term visual dysfunction.
In humans, bTBI can be caused by a wide range of severities applied to multiple organ systems, eliciting various reparative responses from the body. Injury severity can depend on the subject's orientation to the blast wave, location and duration of impact, distance from the source, and protective equipment. It is further complicated by patient demographics such as sex, age, and coinciding co-morbidities (Cernak, 2017;Bryden et al., 2019). Mirroring the complexity of human injury, current murine experimental models vary in terms of injury location, level of protection from the blast wave, and the device and blast magnitude used to administer a bTBI. Many of these successfully model bTBI pathology, but lack of standardization makes direct comparison of data difficult. Here, we review the devices, exposure paradigms, and assessment criteria currently used in rodent models of bTBI-induced visual impairment.

Literature Search Process and Inclusion Criteria
To identify relevant literature, we used the default settings on PubMed Legacy edition, using three-part search terms: 1) the subject: mouse/rat/rodent; 2) the injury type: blast/TBI/traumatic brain injury/blast brain injury/brain injury; 3) the visual outcome: eye/vision. Different combinations yielded 28 search terms (e.g., "mouse traumatic brain injury vision"). The process was completed on June 9, 2020, and produced 1,152 results. A manual filtering process was used to exclude non-relevant results, ensuring that the included sources met three necessary inclusion criteria: 1) used a rodent TBI model, 2) evaluated visual outcomes associated with TBI, and 3) employed a blast-TBI model (Figure 1). 35 original literature sources were selected for the review, three from manual cross-referencing.

Blast Injury Devices
Devices used to administer bTBI fall into four classes (Figure 2): 51.43% (18) of studies used a modified paintball gun, 28.57% (10) used a compressed air shock tube, 17.14% (6) used an air pressure chamber, and 2.86% (1) used a TNT detonation model (injury parameters found in Table 2). Importantly, there is no commercially available standardized blast equipment; many groups have built or made adjustments to their own injuryinduction device, leading to slight variations even within each sub-type of bTBI. However, the majority of groups studying bTBI in this review generally accept that the blast wave created in their model should mimic the Friedlander waveform, modeling the primary blast exposure dynamics that victims are exposed to in the field (Friedlander, 1946;Kuriakose et al., 2016). The Friedlander waveform consists of the shock front, which is an immediate sharp rise in pressure, followed by the blast wind, which is an exponential decay in pressure (Cullis, 2001). All devices induced bTBI with a pressure wave between 3.916 and 80 psi, but target injury locations varied. Control mice underwent an identical process but were not exposed to a blast wave.
The air pressure chamber houses the subject lateral to the blast-wave origin and exposes the left cranium to a longduration (10-15 ms) blast wave between 3.92 and 20 psi (Figure 2A; Goldstein et al., 2012;Mohan et al., 2013). Anesthetized subjects are in a cushioned PVC pipe with their heads unrestrained to allow for full injury to the left side of the head while preventing injury to the right side. The shock tube is a long chamber that generates a short-duration (2-3 ms) blast overpressure wave, ranging from 9.14 to 80 psi, that travels the length of the tube ( Figure 2B; Panzer et al., 2012;Swietek et al., 2019). Impact was delivered to the right or left cranium, or left, right, or front side of the body in anesthetized subjects with various levels of protection.
Shock tube devices administer short (2 to 3 ms) blast overpressures (Panzer et al., 2012;Swietek et al., 2019) while air pressure chambers administer longer (10 to 15 ms) biphasic blast overpressures (Goldstein et al., 2012;Mohan et al., 2013). Whether from expanding gaseous chemicals in shock tubes or pressurized air in pressure chambers, both deliver an initial sharp rise in pressure. This difference in achieving air compression causes the peak in overpressure to be slightly delayed in the air pressure chamber compared to the shock tube. As the wave propagates away from its source, the pressure drops in both devices; in a shock tube this drop occurs exponentially, while in an air pressure chamber the drop is biphasic. The shock wave ends with subsonic particle velocities creating a vacuum with slightly negative pressure, until the wave dissipates (Panzer et al., 2012;Mohan et al., 2013;Swietek et al., 2019).
The modified paintball gun air-tank device emits a brief high pressure air blast that can be calibrated to a specific pressure by adjusting the output from a pressurized air tank, striking the target location on the mouse (Figure 2C; Hines-Beard et al., 2012). It administers bTBI with magnitudes between 15 and 70 psi, with blasts directed at the left eye, right eye, or left cranium. Anesthetized bTBI subjects were placed in a small chamber with a foam cushion to prevent secondary somatic injuries. One study from our literature search utilized 5 kg of 2,4,6-trinitrotoluene (TNT) to generate a blast wave and induce bTBI ( Figure 2D; Zou et al., 2013). Exposure conditions were manipulated by varying the distance from detonation. At two meters from the source, mice experienced a blast overpressure of 69.62 psi, whereas at three meters blast overpressure was 26.11 psi (blast high [BH] and blast low [BL], respectively). Anesthetized mice were positioned in metal cages facing the detonation site and secured to the cage loosely with Velcro to prevent movement.
Of note, mice were exposed to whole-body blast in only a handful of studies (bolded studies in Table 2), which can also serve as another variable within bTBI models. Damage to other organ systems can cause alterations in the body's response to injury and subsequently the systemic environment that the brain and eyes are exposed to after blood-brain-barrier and bloodretina-barrier damage due to blast.  (Yin et al., 2016) 20 (Dutca et al., 2014) 20 (Evans et al., 2020) 20 (Harper et al., 2019a) 20 (Mohan et al., 2013) 5 or 20 (Harper et al., 2019b)

Overall summary
Mouse models of bTBI include modified paintball guns, air shock tubes, air pressure chambers, or TNT explosives; while these devices vary, many aim to recreate the Friedlander waveform victims of blast injury experience.

Overall summary
Most pharmaceutical interventions aimed to reduce retinal inflammation or apoptosis after blast injury. Table 1)

Higher Visual Loci and Pathways
Blasted rodents demonstrated edematous forebrain enlargement (Guley et al., 2016), cerebral cortical neuronal loss (Petras et al., 1997;Guley et al., 2016;Yin et al., 2016), microglial activation (Reiner et al., 2014), and axonal transport defects (Bernardo-Colon et al., 2019). One study of repetitive 43.51 psi blasts to the right cranium reported significantly decreased glial fibrillary acidic protein (GFAP) protein expression in the prefrontal cortex after 30 days, while they found no difference in protein levels of ionized calcium binding adapter molecule 1 (IBA-1) or phosphorylated tau. This study found striatal neurotransmitter levels unchanged following repetitive injury, suggesting that neurons in this location were not damaged in their model (Mammadova et al., 2017). Likewise, another group reported no histological changes including tissue destruction or inflammatory cell accumulation 24 days after a single 39.02 psi blast to the left cranium (Evans et al., 2018). Of note, these groups did observe retinal changes described in other portions of this review.

Overall summary
The majority of structural changes after blast injury were seen in the eye, retina, and optic nerve, with RGC and optic nerve damage frequently reported.

Objective Functional Outcomes (Supplementary Table 2)
The functional integrity of photoreceptor and retinal bipolar cells is measured by electroretinogram (ERG), recorded as a- The and b-sine waves, respectively (Perlman, 1995). These metrics of retinal health and visual ability offer an in vivo diagnostic for both murine and human subjects, however, the results in these studies were inconsistent. When compared to baseline and/or sham controls, the a-wave and b-wave amplitudes in bTBI subjects either increased ( (Zhu et al., 2019). In terms of pharmacological interventions, raloxifene effectively restored ERG amplitudes (Honig et al., 2019), while galantamine partially prevented waveform reductions . While an ERG reflects the ability of the retina to respond to incoming light, a pattern ERG (PERG) is a functional readout of RGC signaling, providing information about visual transduction from the retina to the brain (Bach et al., 2013). Consistently decreased PERG amplitudes were seen in blast-injured rodents, for up to 16 weeks post-injury (Mohan et al., 2013;Dutca et al., 2014;Yin et al., 2016;Harper et al., 2019a,b;Evans et al., 2020). Interestingly, two studies reported temporary recovery in PERG amplitudes: one at 24 h post-injury (Mohan et al., 2013) and another at 4 weeks with a reoccurrence of impaired signaling again at 16 weeks (Dutca et al., 2014). The Wlds genotype (Yin et al., 2016), blast preconditioning (Harper et al., 2019b), and P7C3-S243 treatment (Dutca et al., 2014) preserved PERG amplitudes, hypothetically by promoting survival of the visual circuitry. The Wlds genotype protects against axonal degeneration and inflammatory proliferation at the site of injury (Yin et al., 2016), while blast preconditioning is thought to upregulate endogenous survival factors or downregulate harmful ones. In retinas of mice preconditioned with a small blast, RNA sequencing showed KMO was downregulated. Mice receiving daily oral treatments of Ro-61-8048, a KMO inhibitor, had improved PERG signaling (Harper et al., 2019b). Similarly, P7C3-S243 might preserve visual system integrity by activating metabolic cofactors (Dutca et al., 2014). The anti-inflammatory drug, anakinra, is also partially protective of impaired PERG signaling after blast injury via IL-1RI antagonism, preventing the propagation of inflammatory signaling through this pathway (Evans et al., 2020). An AD model developed worse PERG deficits after blast exposure, suggesting AD amplifies the pathologic retinal effects of bTBI (Harper et al., 2019a). Notably, all of the studies that measured PERG administered blast injury to the left cranium using an air pressure chamber with relatively low pressures (psi between 3.92 and 20).
Visually evoked potentials (VEP) via flash stimulation were also assessed. The VEP N1 amplitude, an early response to visual stimuli (Creel, 1995), was normalized by galantamine  and vitamin E treatment , suggesting these pharmacologic agents protect against VEP response deficits. One study examining RGC physiology reported temporary, spontaneous hyperactivity at 1 and 16 weeks after one 20 psi blast to the left cranium (Dutca et al., 2014). RGC hyperactivity is linked to photoreceptor dystrophic disorders and can significantly decrease the quality of vision (Barrett et al., 2015).

Overall summary
PERG was a functional outcome commonly investigated after bTBI, with many studies describing impaired PERG and RGC signaling due to blast injury. ERG changes were inconsistent when compared between studies.
Pupillary light constriction produced contrasting findings. One day post-20 psi blast to the left cranium, diminished pupillary constriction was seen that resolved after 10 months (Mohan et al., 2013). At 7 months post-injury, another group reported elevated pupillary constriction that was normalized by treatment with raloxifene (Honig et al., 2019). AD mice had an impaired pupillary light response, possibly due to amyloid deposits in the retina (Harper et al., 2019a).

Overall summary
Contrast sensitivity and visual acuity were frequently impaired after bTBI, while studies produced conflicting data on changes in pupillary light constriction.

Subjective and Behavioral General Outcomes (Supplementary Table 4)
Spatial-learning and memory was unaffected by bTBI, as assessed via Morris water maze, at 30 days (Mammadova et al., 2017) and Y maze at 3, 6, and 8 months (Allen et al., 2018). At 7 days postinjury, however, significant behavioral deficits were detectable in Barnes maze performance and could be rescued through a Wlds genotype, which prevents axonal degeneration and inflammatory infiltration to the injuries (Yin et al., 2016). That this genotype protects learning and memory implies potential for post-bTBI therapies that preserve axonal integrity and prevent inflammatory infiltration. Depressive behavior and contextual fear (at 6 to 8 weeks post-injury) were identified in blast rodents and were alleviated by SMM-189, which prevents blast-induced loss of Thy-1 fear-suppressing neurons (Reiner et al., 2014). Of note, blast-related vision loss can affect the assessment of cognitive function, as visual and spatial cues guide subjects throughout the task for many readouts.

Overall summary
Studies have found that blast causes deficits in a wide range of behavioral outcomes, but consistent trends and methodology have not been established.

Inflammatory Over-Activation Following bTBI
Post-bTBI inflammation was consistently detected in the included studies by identifying activated cellular inflammatory modulators or directly measuring inflammatory molecules (Supplementary Tables 1,5). In the inflamed retina, activated resident immune cells attract peripheral immune cells to the site of injury, amplifying the inflammatory response. Inflammatory cytokine levels rise and immunomodulatory cells are activated, triggering changes in cellular morphology or protein expression (Simon et al., 2017). Various post-bTBI-retina immunohistochemistry data detected inflammation-associated upregulation of IBA-1 Guley et al., 2016), GFAP (Zou et al., 2013;Choi et al., 2015;Allen et al., 2018;Honig et al., 2019), or both (Bricker-Anthony et al., 2014aBricker-Anthony and Rex, 2015;Mammadova et al., 2017;Jha et al., 2018;Guley et al., 2019;Evans et al., 2020), markers of microglial and macroglial activation, respectively, suggesting these inflammatory modulators are responding to stress and propagating inflammatory signals post-injury.
bTBI can activate resident immune cells to release proinflammatory cytokines, signaling for prolonged retinal inflammation and exacerbating visual damage (Li et al., 2015;Fehily and Fitzgerald, 2017). Normally, IBA-1 is expressed in quiescent microglia, but following a TBI, activated microglia proliferate, migrate to injured tissue, and exhibit morphological changes (Shapiro et al., 2009). Many studies detected upregulated IBA-1 in the bTBI retina, both as an acute and chronic indicator of ocular trauma and stress (Bricker-Anthony et al., 2014a,b;Bricker-Anthony and Rex, 2015;Guley et al., 2016Guley et al., , 2019Mammadova et al., 2017;Jha et al., 2018;Honig et al., 2019;Evans et al., 2020). The transition of microglia from the proinflammatory M1 state to the reparative M2 state is expressed as the M1/M2 ratio. A prolonged M1 state can damage retinal tissues due to its swelling injury response and downstream release of pro-inflammatory cytokines and free radicals (Loane and Kumar, 2016;Fehily and Fitzgerald, 2017). The M1/M2 ratio decreased after treatment with both raloxifene (Honig et al., 2019) and SMM-189 (Guley et al., 2019), suggesting microglia in blast retinas transitioned toward a reparative phenotype after pharmacologic treatment.
Several post-injury drug interventions suppressed activation of microglia and macroglia in the retina. EPO given at least 1 day after injury in DBA/2J mice decreased GFAP expression . Notably, Balb/c mice treated with rAAV EPO did not show a difference in retinal GFAP between sham and blast, suggesting the timing of EPO therapy is important. ASC-CCM (Jha et al., 2018), anakinra (Evans et al., 2020), and SMM-189 (Guley et al., 2019) downregulated GFAP and IBA-1 expression after blast.

Overall summary
Significant evidence in the literature suggests post-blast inflammation contributes to visual dysfunction after bTBI.
bTBI Upregulates Apoptotic, Necroptotic, and Pyroptotic Mediators (Supplementary Table 6) bTBI activates programmed cell death pathways, such as apoptosis, necroptosis, or pyroptosis. Caspase activation initiates apoptosis causing DNA cleavage and genome fragmentation (Reed, 2000). Modified DNA is packed into apoptotic bodies that await engulfment by phagocytes resulting in cellular death (Elmore, 2007). The terminal deoxynucleotidyl transferase (TdT) dUTP nick-end labeling (TUNEL) assay detects apoptotic DNA fragments. Caspase-3 functions as an effector caspase, killing the cell by cleaving specific intracellular targets (Pasinelli et al., 2000;Wang et al., 2014). The necroptotic pathway is characterized by plasma membrane rupture that renders the extracellular environment for nearby cells toxic with inflammatory cytokines (Dhuriya and Sharma, 2018;Chen et al., 2019). Necroptosis is associated with the family of receptor interacting protein kinases (RIPs); RIP1 and RIP3 are two critical signaling molecules and markers of necroptosis (Liu et al., 2019). Pyroptosis also features plasma membrane rupture and cytokine release affecting neighboring tissues, but caspase-1 initiates pyroptosis by cleaving the substrate gasdermin D, which creates pores and then ruptures the plasma membrane (Bergsbaken et al., 2009;Man et al., 2017).
Three studies investigated therapeutic reduction in apoptosis to increase retinal cell survival after blast injury. In the injured retina, Compound 49b stimulated β-adrenergic receptor activation and IGFBP-3 production, which in turn decreased the level of cleaved caspase-3 and decreased TUNEL labeling, reducing retinal apoptosis (Jiang et al., 2013). A ketogenic diet decreased inflammation and ROS levels, leading to a reduction in the level of cleaved retinal caspase-1 following blast . Finally, an acute increase in EPO in DBA/2J mice exacerbated retinal cell death after bTBI, possibly due to increased oxidative stress from amplified RBC formation and retinal iron levels. However, when analyzed 1 week after injury, this treatment was protective for cell death when compared to controls. EPO treatment with an rAAV with attenuated erythropoietic activity promoted retinal cell survival better if treatment was delayed .

Overall summary
Cell death pathways are frequently implicated in post-blast visual pathophysiology.

DISCUSSION
We reviewed the devices and exposure paradigms employed in bTBI research and found notable interstudy variations in techniques and assessment outcomes. The variability of bTBI experimental models' blast magnitude, location of injury, and device biomechanics makes comparing data on visual outcomes difficult; and the picture is further complicated by inconsistencies in outcome measures. Nevertheless, bTBI consistently resulted in increased inflammation, activation of resident inflammatory mediators, impaired PERG signaling, decreased visual acuity and contrast sensitivity, decreased RGC complex thickness, and ON degeneration. This suggests that these characteristics of visual dysfunction after bTBI are reproducible regardless of the technique employed.
Blast injury in humans, as in murine models, is an extremely heterogenous and multifactorial condition that can result in a wide range of consequences. Many of the outcomes measured in animal studies are not practically measured when assessing human injury, i.e., histology or measuring retinal inflammatory modulators at multiple time points. On the other hand, several parameters used to assess human ocular injury cannot be completely recapitulated in murine studies. For example, it is not feasible to measure specific reading issues or subtle changes in color vision in mice. However, the findings in this review of decreased visual acuity, impaired contrast sensitivity, and optic nerve dysfunction have been consistently seen in human blast injury (Cockerham et al., 2009;Scott, 2011;Saunders and Echt, 2012).

Pharmacologic benefits
FIGURE 3 | Summary of common molecular pathways, visual outcomes, and pharmacologic interventions following blast mediated traumatic brain injury (bTBI).
While retinal and optic nerve damage was frequently identified after blast injury, many groups did not find overt damage to brain tissue. This lack of consistent changes in the brain after blast could suggest that the retina and optic nerve are more sensitive indicators of mild injury in this model. The increased vulnerability of the eye and optic nerve to bTBI specifically is a unique aspect of this injury model, as other types of TBI can experience greater damage to brain tissue. Additionally, as impaired vision can be a confounding factor for cognitive testing, we recommend that future cognitive testing should be done in concert with tests assessing basic visual performance.
The data implicate inflammatory and apoptotic pathways as playing a causal role in long-term visual dysfunction after bTBI and several targeted pharmacological interventions show promise for manipulating those pathways. Generally, inflammatory blockade protected against deficits in contrast sensitivity, visual acuity, and RGC signaling. Additionally, antiinflammatory agents preserved RGC complex layer thickness and optic nerve integrity. Interestingly, these pharmacologic interventions targeted different portions of the inflammatory response and, in the studies that reported levels of individual inflammatory cytokines, varied in terms of the actual reduction of inflammatory molecules. This could suggest that multiple inflammatory pathways play a role after bTBI and that combination therapy using multiple agents would confer the most retinal protection after injury.
It is also clear that programmed cell death contributes to the retinal pathogenesis and subsequent visual disturbances following bTBI. Preventing cellular death is vital for vision preservation and would greatly improve outcomes post-injury. Together, these observations suggest that the overactivation of both inflammatory and apoptotic pathways contribute to visual dysfunction following blast injury (Figure 3). While the field still does not have a gold standard for a rodent blast model, making direct comparisons difficult at times, these common pathways could serve to bridge the gaps caused by variations in experimental techniques and outcome assessments.
Due to the lack of commercially available equipment for blast-induction, standardization of the equipment can be difficult and could contribute to issues with reproducibility across groups using the same type of injury device. However, studies should focus on recapitulating the Friedlander waveform, mimicking the primary blast experienced in the field, enhancing their direct translational potential, and improving cross-model standardization. We recognize the vital need for a study comparing the models directly across a range of blast magnitudes and injury locations to fully understand the commonalities and differences in visual outcomes following varied blast exposure. While this would be a massive undertaking for one group to conduct, the field would benefit greatly from the creation of a large-scale data repository. Individual labs could contribute data generated from their specific parameters, outcomes, and injury type, allowing for comparisons across, as well as within, blast devices, time points, and readouts. This would provide information concerning reproducibility in addition to identifying clear commonalities that could guide research in the search for an effective intervention. We urge that future studies focus on these pathways and their downstream targets to identify specific molecules that could mediate visual protection in patients suffering from bTBI. Experiments pinpointing the antiinflammatory mediators or survival factors that confer the greatest retinal protection would generate great strides toward translating these treatments to human use.

AUTHOR CONTRIBUTIONS
LE, NG, AR, and AB contributed to the conception and design of the study. NG conducted the literature search. LE, NG, and AR wrote sections of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.