Dysfunctional ER-mitochondrion Crosstalk Leads to Subsequent ER Stress-induced Neuronal Apoptosis and Neurological Decits in an Experimental Rodent Model of Severe Head Injury

Introduction: Previous studies have shown proper interorganelle communication between the endoplasmic reticulum (ER) and mitochondria (ER-mitochondrion crosstalk) is crucial for cellular homeostasis. Mitochondria-associated membranes (MAMs) provide an excellent platform and play an essential role in different signaling pathways to maintain cellular viability. However, the time course and potential pathological effects of this ER-mitochondrion physical and functional tethering in traumatic brain injury (TBI) remain unknown. We tested the hypothesis that dysfunctional ER-mitochondrion crosstalk at the acute phase of injury results in ER stress-induced neuronal apoptosis and neurological decits in a mouse model of TBI. Methods: Male C57BL/6 mice were subjected to severe TBI (sTBI) using a controlled cortical impact (CCI) device. Transmission electron microscopy, western blot, immunouorescence staining, terminal deoxynucleotidyl transferase (TdT)-mediated dUTP nick end labeling (TUNEL), and MitoSOX assays were used to analyze posttraumatic ER-mitochondrion physical contact, Ca 2+ transfer from the ER to mitochondria, mitochondrial reactive oxygen species (ROS) production, the unfolded protein response (UPR), the neuroinammatory response and ER stress-mediated apoptosis in the perilesional cortex. The brain water content (BWC) and Evans blue dye extravasation were used to assess the blood-brain barrier (BBB) integrity, and a modied neurological severity score (mNSS) to evaluate neurological function in mice. Results: We showed sTBI led to signicant MAM reorganization in the mouse cerebral cortex during the rst 24 hr after injury. Enhanced ER-mitochondrion crosstalk peaked at 6 hr post injury and was signicantly correlated with increased ER-mitochondrion Ca 2+ transfer, mitochondrial ROS overproduction, elevated ER stress and UPR levels, and augmented levels of proinammatory cytokines. In vivo experimental downregulation of PACS2, a protein essential for ER-mitochondrion tethering, restored mitochondrial Ca 2+ homeostasis and alleviated mitochondrial oxidative stress by downregulating the IP 3 R 1 -GRP75-VDAC1 axis, inhibited ER stress and suppressed the inammatory response through the PERK/eIF2α/ATF4/CHOP signaling pathway, blocked Caspase 12-dependent ER stress-mediated apoptosis, and reduced BBB permeability, resulting in signicantly improved neurological function in mice subjected to sTBI. Conclusions: These results indicate that dysfunctional ER-mitochondrion crosstalk at the acute stage of injury might be primarily involved in the neuronal apoptosis and neurological decits following sTBI, and specic modulation of ER-mitochondrion crosstalk might be a novel promising therapeutic strategy for sTBI. ER stress and UPR signaling as well as with an augmented neuroinammatory response. (3) This strengthened ER-mitochondrion crosstalk at the acute phase of injury was followed by BBB leakage and Caspase 12-dependent ER stress-mediated neuronal apoptosis. (4) More importantly, diminishing the early ER-mitochondrion connection led to reduced Ca 2+ accumulation, oxidative stress and ROS production in mitochondria and to alleviated ER stress, UPR signaling and neuroinammation, in turn leading to restoration of the impaired BBB permeability, Caspase 12-dependent ER stress-induced neuronal apoptosis, and neurological function. These results indicate that dysfunction in acute ER-mitochondrion crosstalk might be primarily involved in the subsequent neuronal apoptosis and neurological decits following sTBI, and specic modulation of acute ER-mitochondrion crosstalk might be a novel promising therapeutic strategy for patients with TBI.


Introduction
Traumatic brain injury (TBI) remains a leading cause of death and disability among children and young adults in China, with an estimated annual cost of $20 billion for medical expenses. Despite extensive basic and clinical research on TBI during the last 150 years, promising preclinical data have not translated into successful clinical trials, and no effective pharmacological interventions are available to date [1]. TBI mainly consists of primary mechanical insult and multifactorial secondary injury that evolves over hours to days and ultimately leads to neuronal cell death [2]. This secondary injury cascade provides a window of opportunity for therapeutic intervention [3].
Although the molecular mechanism that underlies secondary brain injury is not completely de ned, dysfunctional cellular interorganelle communication has emerged as a critical event in alterations following primary insult [4][5][6]. The endoplasmic reticulum (ER) and mitochondria play essential physiological roles in cells through the control of multiple signal transduction pathways [7]. The ER physically and biochemically interacts with mitochondria via mitochondria-associated ER membranes (MAMs), which serve as an excellent scaffold for crosstalk and allow rapid exchange of biological molecules to maintain cellular viability [8]. MAMs are conserved dynamic structures found across eukaryotic phyla and are characterized by a unique lipid pro le and the expression of a speci c group of proteins involved in Ca 2+ signaling, phospholipid biosynthesis, protein folding, membrane tethering, and stress signals transferring [9][10][11]. Accordingly, MAMs are crucial for not only the e cient transfer of Ca 2+ from the ER to mitochondria, proper mitochondrial bioenergetics, mitochondrial dynamics, lipid synthesis, and autophagosome assembly but also ER stress, an accumulation of unfolded proteins in the ER lumen, initiating a defensive the unfolded protein response (UPR) to safeguard cellular survival [12]. MAMs were recently shown to not only have numerous physiological and metabolic functions but also to contribute substantially to a variety of pathological conditions, including obesity, insulin resistance, aging, tumorigenesis, axon regeneration, Alzheimer's disease, Parkinson syndrome and other neurodegenerative disorders [13][14][15][16][17]. Although the pathological impacts of ER and mitochondrial dysfunction have primarily been observed and investigated independently in TBI [18][19][20], the time course and potential pathological effects of ER-mitochondrion physical and functional crosstalk on severe TBI (sTBI) remain poorly understood.
Endoplasmic reticulum (ER) is the largest cellular organelle, where all secreted and membrane proteins are synthesized and properly folded. The accumulation of unfolded and misfolded proteins causes ER stress and induces the unfolded protein response (UPR) to maintain cellular homeostasis [21]. ER stress activates the UPR signaling network and eventually promotes cell death in a wide range of pathologies including cerebral ischemia, traumatic brain injury, Alzheimer's disease, multiple sclerosis, and amyotrophic lateral sclerosis [22].
Here, we report that sTBI drives an abnormal acute increase in MAM formation in the mouse cerebral cortex, with the most enhanced ER-mitochondrion crosstalk in neurons occurring at 6 hr post injury as an early event in the course of secondary brain injury development. Such increased ER-mitochondrion tethering correlates with increased Ca 2+ ux from the ER to mitochondria, increased mitochondrial reactive oxygen species (ROS) production, elevated ER stress levels, increased unfolded protein response (UPR) signaling, and an augmented neuroin ammatory response. Suppression of phosphofurin acidic cluster sorting protein 2 (PACS2), a distinct protein critical for ER-mitochondrion associations, improves neuronal homeostasis and reduces subsequent ER stress-mediated apoptosis and blood-brain barrier (BBB) leakage, resulting in improved neurological function in mice subjected to sTBI. These results indicate that dysfunctional ER-mitochondrion crosstalk at the acute stage of injury might be primarily involved in the subsequent neuronal apoptosis and neurological de cits following TBI, and speci c modulation of ER-mitochondrion crosstalk might be a novel promising therapeutic strategy for sTBI.

Materials And Methods
Animal care and experiments were conducted following ethical approvals provided by the Small Animal Protection Board of Tianjin Medical University.

Animals
Adult male C57BL/6 mice weighing 22 -25 g (8 -10 weeks old) at the time of surgery were purchased from the Experimental Animal Laboratories of the Academy of Military Medical Sciences (Beijing, China).
All mice were housed individually in a temperature-(20 ± 2 °C) and humidity-controlled (55 ± 5%) vivarium and maintained on a standard 12-hr light/dark cycle (7:00 a.m. to 7:00 p.m) with access to food and water ad libitum. All efforts were made to minimize the number of mice used and their suffering. In all experiments, the data were obtained by investigators blinded to the experimental design.

Experimental design
In the present study, the following separate experiments were conducted: In Experiment 1, we investigated the dynamic changes in neuronal ER-mitochondrion physical contacts after TBI. Thirty-six mice were randomly assigned to the following six groups: sham, TBI 1 hr, TBI 3 hr, TBI 6 hr, TBI 12 hr, and TBI 24 hr. We used transmission electron microscopy (TEM) to examine the time pro le of ER-mitochondrion physical tethering, as described below.
In Experiment 4, we examined cleaved Caspase 12-dependent ER stress-mediated apoptosis after TBI.
In Experiment 6, we investigated the potential effects of PACS2 knockdown on BBB functions, ER stressmediated apoptosis, and neurological outcomes. One hundred twenty mice were randomly allocated into the following four groups: sham, TBI, TBI + PACS2 siRNA, and TBI + NC siRNA. Neurological scores were used to evaluate the neurological functions of the mice at preinjury and at postinjury days 1, 3, 5, 7, and 14. At 72 hr post injury, extravasation of Evans blue (EB) and the brain water content (BWC) were measured in all groups. cleaved Caspase 12, cleaved Caspase 3, cleaved PARP1, CHOP, Bcl-2, BAX, and Cytc expression levels were also determined by western blot. The terminal deoxynucleotidyl transferase (TdT)-mediated dUTP nick end labeling (TUNEL) assay and immuno uorescence staining were performed in each group at 72 hr after injury.
All mice in the sham group were subjected to the same surgical procedure but were not treated with the controlled cortical impact (CCI) device.
Severe traumatic brain injury model CCI-induced brain injury is an extensively characterized and broadly used preclinical model of head injury [21,23,24]. In brief, a digital electromagnetic CCI device (eCCI Model 6.3; Custom Design, Richmond, VA, USA) was used to establish the sTBI model in male C57BL/6 mice. The mice were allowed to adapt to their new environment for 1 week before surgery and then anesthetized with 10% chloral hydrate. The depth of anesthesia was assessed by monitoring the pedal withdrawal re ex and respiratory rate. Then, the mice were placed in a stereotaxic apparatus, and the surgical site was clipped and cleaned with Nolvasan scrubs. A 4.0-mm hole was drilled into the right parietal bone to expose the dura. The CCI device subsequently impacted the skull at a depth of 2.5 mm and a velocity of 5 m/s over a period of 200 ms. The incision was closed immediately following injury, and the mice were then placed in heated cages to allow recovery from anesthesia at room temperature (RT). Fig. 1 shows the hematoxylin and eosin (H&E) staining of the cerebral cortex at 24 hr after CCI, which con rmed the severe injury of the mice used in this study. ER-mitochondrion imaging and quanti cation were performed according to a previous study [25]. Brie y, 60 images of mice in each experimental group (6 mice from each group) were obtained at ×6,800 magni cation, and ImageJ (National Institutes of Health, Bethesda, Maryland, USA) was used to analyze ER-mitochondrion contacts and mitochondrial morphology. We delineated the mitochondria and ER membranes using the free-hand tool. Two independent investigators blinded to the experimental design calculated the ratio of ER adjacent to mitochondria to mitochondrial perimeter, and the total number and area of mitochondria.

Immuno uorescence and image analysis
At designated time points, mice were sacri ced with an overdose of 10% chloral hydrate and then immediately perfused through the heart with PBS followed by 4% paraformaldehyde. The brain was rapidly dissected and embedded in OCT medium (Sakura, Oakland, CA, USA). Coronal sections of 8-μm thickness were cut on a cryostat at -20°C and imprinted on poly-L-lysine-coated slides. The sections were stained for NeuN (neuronal marker), glial brillary acidic protein (GFAP, astrocyte marker), and PACS2 and Mfn2 (markers of ER-mitochondrion contacts).
In brief, the sections were xed with 2% paraformaldehyde lysine periodate (PLP), rinsed three times with PBS (pH 7.4), and blocked with 1% normal donkey serum in PBS containing 0.1% Triton X-100 PBST at RT for 1 hr. The sections were then incubated with a rabbit anti-PACS-2 antibody diluted at 1:1000 (Abcam, Cambridge, MA, USA), a rabbit anti-Mfn2 antibody diluted at 1:1000 (Abcam), a mouse anti-NeuN antibody diluted 1:100 (Cell Signaling Technology, Danvers, MA, USA), and a mouse anti-GFAP antibody diluted at 1:1000 (Abcam, Cambridge, MA, USA) in PBST containing 1% normal donkey serum at 4°C overnight followed by extensive washing with PBS. Finally, the sections were incubated with Alexa Fluor-conjugated anti-mouse or anti-rabbit IgG (1:1000, Invitrogen, Grand Island, NY, USA) for 3 hr at RT. The nuclei were counterstained with Hoechst for 5 min.
Images of each section were captured using a uorescence microscope (Olympus IX81, Tokyo, Japan), and the data were analyzed from 15 randomly selected microscopic elds ( ve elds per section x three sections per mouse) with ImageJ (National Institutes of Health, Bethesda, Maryland, USA).

Western blotting
Mice were sacri ced by transcardiac perfusion with cold PBS to eliminate the proteins expressed by blood cells at designated time points. Their brain tissues were homogenized in ice-cold RIPA buffer (Beyotime) containing phenylmethylsulfonyl uoride (PMSF, 1 mM nal) for 30 min and then centrifuged for 10 min (12,000 rpm, 4℃). After centrifugation, the supernatants were collected and boiled with 4x sample buffer at 95℃ for 10 min. The total protein content was determined by the BCA protein assay kit (Thermo). Proteins (8 μg per lane) and prestained molecular weight markers (Thermo) were separated by SDS/PAGE and transferred to PVDF membranes (Roche, Canada), which were then blocked with 5% nonfat dry milk in Tris-buffered saline (TBS) for 2 hr at RT. After blocking, the blots were incubated overnight at 4℃ with primary antibodies (Table 1), rinsed with TBS, incubated with the appropriate HRPconjugated secondary IgG for 1 hr at RT and then developed with the ECL system (Millipore, Billerica, MA, USA). Protein expression was quanti ed by ImageJ (National Institutes of Health, Bethesda, Maryland, USA) according to the mean pixel density of each protein band, and β-actin was employed as a loading control.

Mitochondrial reactive oxygen species content
MitoSOX-based assays were used to detect mitochondrial ROS production according to the mice 48 hr prior to TBI. In terms of the i.c.v injection [30], a 1-mm cranial burr hole was drilled into the skull, and a 30-gauge needle on a Hamilton syringe was implanted into the lateral ventricle using the following stereotactic coordinates: 1.5 mm posterior to bregma, 1.0 mm right lateral to the midline, 2 mm in depth, with an injection speed of 1μL/min (total volume=5μL). The transfection rate is approximately 85 ± 5%.
Modi ed neurological severity scores The modi ed neurological severity score (mNSS) was used to evaluate neurological function, as described previously [31]. Neurological assessments were performed at baseline before the injury and at post injury days 1, 3, 5, 7, and 14 using the mNSS. The assessments included motor, sensory, re ex, and balance tests. These scores were used 1) to ensure the relative uniformity in injury severity and 2) to compare neurological impairments among mice receiving different treatments. The tests were performed by two independent observers who were blinded to the experimental conditions and treatments.
Brain water content Hemispheric cerebral edema was determined by measuring the BWC, as previously described [31]. Mice were sacri ced with an overdose of chloral hydrate at 72 hr post injury. Their brains were promptly removed, and the hemispheres were immediately weighed (wet weight) and then placed in an incubator at 100°C for 24 hr. The samples were weighed again to determine the dry weight, and the BWC was calculated as follows: (wet weight -dry weight)/wet weight x 100%.

Evans blue dye extravasation
The BBB permeability of the cerebral hemispheres was assessed by measuring the extravasation of EB dye 72 hr after TBI. EB dye injected intravenously binds instantaneously to albumin and other plasma proteins and serves as a marker for plasma exudation. In brief, EB (2% in PBS, Sigma) was injected slowly through the jugular vein (4 ml/kg) and allowed to circulate for 1.5 hr. Then, mice were sacri ced and transcardially perfused with PBS followed by 0.9% saline. The hemispheres were removed, frozen in -55°C isopentane and freeze-dried. The freeze-dried specimens were homogenized in formamide (1:20) and incubated at 60°C overnight, and the homogenates were then centrifuged at 14000 rpm for 30 min to collect the supernatant. The EB content in the supernatant was determined spectrophotometrically at OD 620 nm (Thermo Scienti c). The tissue EB concentration was quanti ed using a standard linear curve and expressed as micrograms per gram of brain tissue.
Terminal deoxynucleotidyl transferase dUTP nick end labeling assay The TUNEL assay is a well-de ned method for detecting apoptotic DNA fragmentation in a cell [32]. We used the In Situ Cell Death Detection Kit, POD (Roche, Germany) to detect apoptosis in the perilesional cortex of the mouse brain at 72 hr after TBI. According to the manufacturer's instructions, the sections were xed in acetone for 8 min at 4°C. After being rinsed with PBS, the brain sections were treated with 3% BSA for 30 min at 37°C and then incubated with the TUNEL reaction mixture in the dark for 90 min at 37°C. The nuclei were counterstained with Hoechst for 5 min. Images of each section were captured using a uorescence microscope (Olympus IX81, Tokyo, Japan), and the data were analyzed from 15 randomly selected microscopic elds ( ve elds per section x three sections per mouse) with ImageJ (National Institutes of Health, Bethesda, Maryland, USA).

Data analysis
The data are presented as the mean ± standard deviation (SD) and were analyzed using Prism 8.3.1 (GraphPad Software, San Diego, CA). Parameters were compared by ANOVA followed by Tukey's multiple comparisons test. Pearson's correlation coe cients (r) were calculated to assess the strength of relationships. P values < 0.05 were considered statistically signi cant.

Results
Dynamic changes in neuronal ER-mitochondrion contacts in the mouse cerebral cortex after TBI The ER and mitochondria interact to form dynamic contact sites, which are responsible for the integration of several cellular functions, including Ca 2+ homeostasis, oxidative stress, ER stress, neuroin ammation, and survival [15,33]. For this reason, we speculated that ER-mitochondrion contacts induce characteristic changes in cerebral cortex neurons to orient different signaling pathways for effective crosstalk and cellular viability after sTBI.
To investigate ER-mitochondrion physical interactions and mitochondrial morphology, we rst used TEM to examine neurons in the mouse perilesional cortex at preinjury and at 1, 3, 6, 12, and 24 hr post injury. The neuronal cell bodies were distinguished from glial cell bodies using previously described morphologic criteria [20,34]. We observed that prior to injury, mitochondria exhibited a moderately dense matrix and a normal cristae architecture (Fig. 2a, g, m). Signi cant structural changes and a reduced number of neuronal mitochondria were observed immediately after injury, as expected ( Fig. 2b-f, n-r). Some mitochondria exhibited disorganized morphologies with low to high extents of swelling and an electron-lucent matrix. At 6 hr post injury, neuronal mitochondria showed severe swelling and inner membrane-associated dense granular inclusions, which is a distinguishing mitochondrial feature of calcium overload and irreversible injury, along with a markedly high degree of apposition to the ER (Fig. 2d, j, p). Detailed quantitative analysis of cerebral sections from each time point demonstrated that the proportion of the ER in close contact with mitochondria to the mitochondrial perimeter began to increase at 1 hr, peaked at 6 hr, and returned to slightly above baseline at 24 hr (Fig. 2B). The neuronal mitochondrial number was signi cantly reduced after injury (Fig. 2C), but the mitochondrial area was signi cantly elevated at 6 hr ( Fig. 2D), indicating mitochondrial morphological changes and swelling. Together, these results suggest a high amount of neuronal MAM formation at the acute phase of sTBI than in the phase prior to injury. This ER-mitochondrion interface might recruit and orient various signaling proteins needed to regulate productive crosstalk and cellular homeostasis.
Altered expression of MAM-resident proteins and mitochondrial oxidative stress in the mouse cerebral cortex after TBI Given that more ER-mitochondrion physical contacts were observed at the acute phase of TBI than in the phase prior to injury and that the MAM subdomain is enriched for speci c proteins, such as the ERmitochondrion tether proteins PACS2 and Mfn2, we next examined the expression of these proteins in the mouse perilesional cortex before injury and at 1 3, 6, 12, and 24 hr after injury. The protein levels of PACS2 and Mfn2 increased rapidly after injury, peaking at 6 hr, declining gradually, and returning to the baseline level by 24 hr (Fig. 3B, C). These ndings were further substantiated by double immuno uorescence labeling of PACS2 with NeuN (a neuronal marker) and GFAP (an astrocyte marker) in the ipsilateral cerebral cortex, and the ER was more frequently attached to the neuronal mitochondria at 6 hr following TBI than at the other time points (Fig. 3A).
It is well known that within the ER-mitochondrion tethering complex, IP 3 R 1 located on the ER membrane is physically and functionally linked to the VDAC located on the mitochondrial outer membrane via chaperone GRP75. Sigma-1R is a Ca 2+ -sensitive and ligand-operated receptor chaperone that regulates ER-mitochondrion interorganelle Ca 2+ signaling. Because any disruption of mitochondrial Ca 2+ homeostasis can contribute to increased mitochondrial ROS production and oxidative stress, we explored the possible functional consequence of the increased ER-mitochondrion connectivity in sTBI. The levels of IP 3 R 1 , GRP75, VDAC1, and Sigma-1R were markedly elevated and the highest amounts of mitochondrial ROS production were observed at 6 hr after injury, when MAM formation peaked simultaneously ( Fig. 4A-C). Altogether, these results not only further support our morphological and microscopic observations of a signi cant number of ER-mitochondrion contacts in the acute stage of sTBI but also suggest that numerous anchoring proteins and functional components involved in Ca 2+ transport are expressed at these sites, and the crosstalk between ER and mitochondria is associated with mitochondrial oxidative stress.
Alteration of ER stress, UPR signaling and neuroin ammatory responses in the mouse cerebral cortex after TBI MAMs are crucial for not only the e cient transfer of Ca 2+ from the ER to mitochondria, proper mitochondrial bioenergetics, mitochondrial dynamics, lipid synthesis, and autophagosome assembly but also ER stress, the UPR, and neuroin ammatory responses [12,[35][36][37]. MAMs have been considered a critical hub to transfer stress signals from the ER to mitochondria, most notably under the condition of loss of ER proteostasis, by engaging the UPR[38]. Given that RNA-dependent PERK, a key ER stress sensor, is located in the ER membrane within the ER-mitochondrion tethering complex [39,40], we assumed that altered crosstalk between neuronal ER and mitochondria affects ER stress, the PERK branch of the UPR and neuroin ammation. We analyzed the protein expression pro les of ER stressrelated molecules and proin ammatory cytokines in the pericontusional cortex at different time points up to 24 hr after TBI. We showed that the expression of the 78 kDa glucose-regulated protein (GRP78), also referred to as BiP, a classical marker of ER stress, was signi cantly elevated at 3 hr and peaked at 6 hr following TBI (Fig. 5D, P < 0.05), while the levels of p-PERK/PERK, p-eIF2α/eIF2α, and ATF4 exhibited comparable dynamic changes with peak expression at 6 hr ( Fig. 5A-C, P < 0.05). The production of the procytokines cleaved IL-1β and TNFα also peaked at 6 hr ( Fig. 5E-F, P < 0.05). Furthermore, Pearson's correlation analysis revealed that PACS2 expression was signi cantly positively correlated with the levels of Mfn2, IP 3 R 1 , VDAC1, Sigma-1R, GRP75, ROS, GRP78, p-PERK/PERK, p-eIF2α/eIF2α, ATF4, cleaved IL-1β, and TNFα within 24 hr of injury (Fig. 6, P < 0.05).
These data, together with the altered ER-mitochondrion connection, suggest that the enhanced ER and mitochondria coupling in the acute stage of TBI is associated with increases in the Ca 2+ ux from the ER to mitochondria, ROS production, ER stress and UPR signaling as well as with an augmented neuroin ammatory response.
Caspase 12-dependent ER stress-mediated apoptosis following increased ER-mitochondrion coupling in the mouse cerebral cortex ER stress-induced apoptosis involves triggering the PERK-eIF2α-ATF4 branch of the UPR, upregulating CHOP expression, and activating Caspase 12, a regulator speci c to ER stress-induced apoptosis [21,41,42]. To explore the possible pathological consequence of increased ER-mitochondrion crosstalk on subsequent ER stress-mediated apoptosis, we assessed the protein expression of cleaved Caspase-12, cleaved Caspase-3, cleaved PARP1, CHOP, Bcl-2, Bax and Cytc at different time points from 6 to 72 hr post injury. As expected, the levels of cleaved Caspase-12, cleaved Caspase-3, cleaved PARP1, CHOP, and Cytc began to rise immediately after the enhancement of ER-mitochondrion crosstalk at 6 hr, reached signi cance at 12 hr, and peaked at 72 hr after injury, when the Bcl-2/Bax ratio decreased signi cantly to its lowest level (Fig. 7). Taken together, our data demonstrated that enhanced ER-mitochondrion communication preceded ER stress-induced apoptosis, suggesting that ER-mitochondrion crosstalk may play an essential role in the secondary injury cascade after TBI.
Silencing Pacs2 alleviated ER-mitochondrion Ca 2+ transfer and mitochondrial oxidative stress, reduced ER stress and UPR activation, and suppressed neuroin ammatory responses If upregulated MAM functioning in neurons is related to ER-mitochondrion Ca 2+ transfer and mitochondrial oxidative stress, ER stress and UPR activation, and neuroin ammatory responses in the acute stage of sTBI, then decreased contacts between the two organelles should have the opposite effects. For this reason, to explore the mechanistic relationship among ER-mitochondrion crosstalk, cellular homeostasis and neuroin ammation, we suppressed the ER-mitochondrion physical connection by silencing cerebral PACS2 expression. PACS2 siRNA treatment signi cantly reduced the expression levels of ER-mitochondrion tethers PACS2 and Mfn2 by approximately 50% at 6 hr post injury (Fig. 8A-B).
Silencing Pacs2 resulted in a modest but signi cant reduction in IP 3 R 1 , GRP75, VDAC1, Sigma-1R expression along with a signi cant reduction in ROS production at 6 hr after injury (Fig. 8C-E). In addition, the expression levels of GRP78, p-PERK/PERK, p-eIF2α/eIF2α, ATF4, cleaved IL-1β, and TNFα, which were signi cantly elevated in TBI mice at 6 hr post injury, were also signi cantly reduced in mice treated with PACS2 siRNA (Fig. 8F-K). These results demonstrate that reducing ER-mitochondrion crosstalk at the acute stage of TBI abolishes ER-mitochondrion Ca 2+ transfer and mitochondrial oxidative stress, ER stress and UPR activation, and neuroin ammatory responses.
Knocking down Pacs2 reduced TBI-associated BBB permeability, reversed ER stress-mediated apoptosis, and improved neurological function In an attempt to further mechanistically link ER-mitochondrion coupling to secondary brain injury and neurological de cits after TBI, we attempted to mitigate BBB leakage, alleviate ER stress-mediated apoptosis, and improve neurological function by diminishing ER-mitochondrion crosstalk. Knocking down PACS2 led to signi cant reductions in EB extravasation and the BWC at 72 hr post injury compared to the injury control group (Fig. 9A-C). The neurological scores of mice preconditioned with PACS2 siRNA were signi cantly lower than those of injury control mice during a 14-day follow-up period (Fig. 9C). PACS2 siRNA successfully reduced the protein expression of cleaved Caspase 12, CHOP and Cytc and increased the Bcl-2/Bax ratio (Fig. 9G-J). More importantly, these phenomena were accompanied by a signi cant reduction in the percentage of TUNEL-positive neuronal cells in the perilesional cortex at postinjury day 3 compared to the injury control ( Fig. 9E-F). These results demonstrated that reducing ER-mitochondrion crosstalk at the acute stage of TBI improved cerebrovascular function, increased cellular viability, and ameliorated neurological de cits. In total, this loss-of-function approach supports the concept that targeting the structural components of MAMs may be an effective strategy to improve neurological outcomes in the context of sTBI.

Discussion
In the present study, we investigated changes in acute ER-mitochondrion crosstalk in the cerebral cortex and mechanistically linked such changes to subsequent ER stress-mediated neuronal apoptosis and neurological de cits after sTBI. Indeed, we demonstrated the following: (1) sTBI drove an abnormal acute increase in MAM formation in the cerebral cortex, with the most enhanced ER-mitochondrion tethering occurring in neurons at 6 hr post injury as an early event in the pathological course of secondary brain injury development.
(2) Such increased ER-mitochondrion coupling was signi cantly positively correlated with increases in the elevated Ca 2+ ux from the ER to mitochondria, mitochondrial oxidative stress and ROS production, ER stress and UPR signaling as well as with an augmented neuroin ammatory response.
(3) This strengthened ER-mitochondrion crosstalk at the acute phase of injury was followed by BBB leakage and Caspase 12-dependent ER stress-mediated neuronal apoptosis. (4) More importantly, diminishing the early ER-mitochondrion connection led to reduced Ca 2+ accumulation, oxidative stress and ROS production in mitochondria and to alleviated ER stress, UPR signaling and neuroin ammation, in turn leading to restoration of the impaired BBB permeability, Caspase 12-dependent ER stress-induced neuronal apoptosis, and neurological function. These results indicate that dysfunction in acute ERmitochondrion crosstalk might be primarily involved in the subsequent neuronal apoptosis and neurological de cits following sTBI, and speci c modulation of acute ER-mitochondrion crosstalk might be a novel promising therapeutic strategy for patients with TBI.
Increasing evidence suggests that the dysfunctional ER-mitochondrion crosstalk occurs in a variety of neurological diseases. Mutations in presenilins upregulate MAM function and increase ER-mitochondrion communication in patients with both the familial and sporadic forms of Alzheimer's disease (AD), indicating that AD is fundamentally a disorder of ER-mitochondrion communication [37,43]. Furthermore, dysfunctional MAM signaling impairs neuronal calcium homeostasis, mitochondrial dynamics, ER function, and autophagy, eventually leading to axonal degeneration in amyotrophic lateral sclerosis (ALS) and hereditary motor and sensory neuropathy (HMSN) [7]. The loss and impairment of Sigma-1R, a MAM protein, leads to axonal and motor neuron degeneration by affecting calcium homeostasis, ER stress, mitochondrial dynamics and transport [44], and mutation of Mfn2 alters the interplay between ER and mitochondria, contributing to the development of Charcot-Marie-Tooth type 2A (CMT2A), a dominant axonal form of peripheral neuropathy [45]. Research and clinical interest have been increasingly focused on understanding the critical role of ER-mitochondrion crosstalk in the pathological course of sTBI [46][47][48]. To the best of our knowledge, no data on the spectrum of ER-mitochondrion interactions in TBI are available to date. Because MAM dysfunction might be the common denominator underlying disease development, we propose that increased MAM activity and ER-mitochondrion communication lie at the heart of TBI pathogenesis. In support of this view, we note that both the physical connection and the functional crosstalk between these two organelles increase rapidly after injury, peaking at 6 hr post injury.
The early onset of enhanced MAM functioning and ER-mitochondrion communication are the initial events in the pathological course of sTBI, leading to disruption of mitochondrial homeostasis and subsequent neuronal apoptosis and neurological de cits.
Mitochondria are in close proximity to the ER and form an elaborate platform to ensure precise modulation of pathophysiological Ca 2+ signal transfer from the ER to mitochondria [7,49]. It has been shown that ER-resident IP 3 Rs physically associate with the cytosolic fractions of the mitochondrial chaperone GRP75 and the VDAC1 of the outer mitochondrial membrane (OMM) to form a Ca 2+ -transfer complex. This multiprotein structure is crucial for effective IP 3 -dependent ER-mitochondrion Ca 2+ coupling [50]. Three IP3R isoforms were shown to be explicitly engaged in Ca 2+ shuttling from the ER to mitochondria, and IP 3 R 1 was predominantly enriched at MAMs and more e cient at sustaining Ca 2+ delivery to mitochondria from store-operated Ca 2+ entry (SOCE), a central mechanism in cellular calcium signaling and in maintaining cellular calcium balance [51,52]. Sigma-1R is a chaperone protein residing at MAMs, where it interacts with several partners, such as IP 3 Rs, regulating the Ca 2+ exchange between the ER and mitochondria. Consistent with this, we showed that acute MAM formation at 6 hr post injury led to increased activity of the IP 3 R 1 -GRP75-VDAC1 complex and the Sigma-1R chaperone as well as to excessive mitochondrial ROS production in the cerebral cortex. Inhibition of acute MAM formation restored mitochondrial calcium homeostasis and oxidative stress.
It has been recognized that ROS overproduction causes oxidative damage to essential cellular components, including neurons, astrocytes, and vascular elements, and is the primary cause of secondary brain injury after TBI [19,53]. Mitochondria not only initiate the generation of such ROS but also enhance cellular and mitochondrial dysfunction by becoming targets of their damaging products through a vicious cycle of oxidative toxicity [54]. Excessive ROS is believed to be a major stimulus that triggers ROS-dependent ER stress, which occurs when the capacity of the ER to fold proteins becomes saturated under oxidative stress [11,[55][56][57]. ER stress activates a signaling network called the UPR to alleviate this stress and restore ER homeostasis, promoting cell survival and adaptation [58]. Here, we indeed found that TBI induced ROS production followed by UPR signaling. However, suppression of acute MAM formation alleviated ROS-dependent ER stress and reduced BBB leakage, which is also supported by our previous study on subarachnoid hemorrhage (SAH) that found that ER stress inhibitors attenuated acute brain injury by rescuing cerebrovascular dysfunction [41]. Recent reports note that inhibiting ER stress restores the dysfunction of retinal endothelial cells by decreasing NO production and downregulating the expression of ICAM-1, NOS, NF-κB, and VEGF [59] and improves endotheliumdependent vasorelaxation through restoration of endothelial ER calcium homeostasis and Sirt1 activation [60] [61]. Together, these results suggest that acute MAM formation is the primary pathological cascade element of secondary brain injury following TBI.
The UPR is mediated through three ER-transmembrane effector proteins, PERK, inositol-requiring enzyme 1 (IRE1), and ATF4. PERK is uniquely enriched at MAMs [42] and has been considered not only a central regulator of ER stress but also a major transcription factor pathway to convey apoptosis after ROS-based ER stress [39]. Under ER stress, PERK dissociates from GRP78 and undergoes oligomerization and autophosphorylation, which leads to phosphorylation of the eukaryotic initiation factor eIF2α. Together with our previous study on SAH [41], this observation validates a causal role of Caspase 12dependent ER stress in the induction of neuronal apoptosis following TBI.

Conclusions
In conclusion, dysfunctional ER-mitochondrion crosstalk at the acute stage of injury might be primarily involved in the neuronal apoptosis and neurological de cits following sTBI, and speci c modulation of ER-mitochondrion crosstalk might be a novel promising therapeutic strategy for patients with TBI.

Figure 9
Experimental knockdown of MAM formation improved cerebrovascular function, increased cellular viability, and ameliorated neurological de cits. (A) Representative photographs of the brains from the