Acute neuroinflammation induces AIS structural plasticity in a NOX2-dependent manner

Background Chronic microglia-mediated inflammation and oxidative stress are well-characterized underlying factors in neurodegenerative disease, whereby reactive inflammatory microglia enhance ROS production and impact neuronal integrity. Recently, it has been shown that during chronic inflammation, neuronal integrity is compromised through targeted disruption of the axon initial segment (AIS), the axonal domain critical for action potential initiation. AIS disruption was associated with contact by reactive inflammatory microglia which wrap around the AIS, increasing association with disease progression. While it is clear that chronic microglial inflammation and enhanced ROS production impact neuronal integrity, little is known about how acute microglial inflammation influences AIS stability. Here, we demonstrate that acute neuroinflammation induces AIS structural plasticity in a ROS-mediated and calpain-dependent manner. Methods C57BL/6J and NOX2−/− mice were given a single injection of lipopolysaccharide (LPS; 5 mg/kg) or vehicle (0.9% saline, 10 mL/kg) and analyzed at 6 h–2 weeks post-injection. Anti-inflammatory Didox (250 mg/kg) or vehicle (0.9% saline, 10 mL/kg) was administered beginning 24 h post-LPS injection and continued for 5 days; animals were analyzed 1 week post-injection. Microglial inflammation was assessed using immunohistochemistry (IHC) and RT-qPCR, and AIS integrity was quantitatively analyzed using ankyrinG immunolabeling. Data were statistically compared by one-way or two-way ANOVA where mean differences were significant as assessed using Tukey’s post hoc analysis. Results LPS-induced neuroinflammation, characterized by enhanced microglial inflammation and increased expression of ROS-producing enzymes, altered AIS protein clustering. Importantly, inflammation-induced AIS changes were reversed following resolution of microglial inflammation. Modulation of the inflammatory response using anti-inflammatory Didox, even after significant AIS disruption occurred, increased the rate of AIS recovery. qPCR and IHC analysis revealed that expression of microglial NOX2, a ROS-producing enzyme, was significantly increased correlating with AIS disruption. Furthermore, ablation of NOX2 prevented inflammation-induced AIS plasticity, suggesting that ROS drive AIS structural plasticity. Conclusions In the presence of acute microglial inflammation, the AIS undergoes an adaptive change that is capable of spontaneous recovery. Moreover, recovery can be therapeutically accelerated. Together, these findings underscore the dynamic capabilities of this domain in the presence of a pathological insult and provide evidence that the AIS is a viable therapeutic target.


Background
The axon initial segment (AIS) is a highly specialized axonal domain responsible for action potential initiation and modulation [1]. The AIS is characterized by a unique assembly of cytoskeletal and scaffold proteins [2] and densely packed voltage-gated ion channels, which are recruited to and clustered at the AIS via the scaffolding protein ankyrinG (ankG) [3]. ankG is considered the "master organizer" of the AIS and is essential for AIS function [4,5]. Accumulating evidence suggests that the AIS is a dynamic domain capable of structural plasticity, undergoing changes in length [6], location [2,7,8], and ion channel clustering [9,10] in response to neuronal pathology and altered activity.
AIS plasticity is characterized by the relocation of cytoskeletal-associated proteins such as ankG, βIV spectrin, neurofascin, and voltage-gated sodium (NaV) channels [2,7,[11][12][13]. Although plasticity can be triggered by both pathologic and non-pathologic stimuli, the mechanisms and cell types that drive plasticity remain largely unknown. Schafer et al. [14] were the first to implicate the calcium-dependent protease calpain as a mediator of AIS plasticity with recent studies confirming these findings [10,15]. Consistent with calpain activation, Evans et al. [8] reported that AIS plasticity is triggered by calcium channel activation with downstream activation of calcineurin. Recently, it has also been shown that microglia may influence neuronal activity through specific association with the AIS [16]. Microglia-AIS contact was found to occur early in development and persist throughout adulthood in the uninjured brain [16] as well as during chronic inflammation present in an animal model of multiple sclerosis known as experimental autoimmune encephalomyelitis (EAE) [17], suggesting an important interaction that may influence AIS integrity.
Microglia, the resident immune cells of the central nervous system (CNS), are dynamic cells that survey, respond, and shape neuronal networks through neuronal contact and synaptic pruning [18][19][20][21]. Microglia are critical for maintaining tissue homeostasis in the CNS, rapidly activating and eliminating pathogens and cellular debris in response to infection or insult [22][23][24]. Upon activation, microglia display an enhanced pro-inflammatory response and a dampened resolving phenotype [25][26][27]. This is typified by increased expression of inflammatory mediators such as tumor necrosis factor alpha (Tnf-α), cyclooxygenase-2 (COX-2), and NADPH oxidase 2 (NOX2), elevated production of reactive oxygen species (ROS), and reduced expression of resolving factors such as transforming growth factor beta (TGF-β), mannose receptor, C type 1 (Mrc1), and resistin-like beta (Fizz-1) [28][29][30]. Though reactive microglia play an important role in pathogen clearance and CNS homeostasis, amplified ROS production or aberrant activation of the inflammatory phenotype has been implicated in a number of neuronal pathologies [31][32][33][34][35] where AIS disruption is observed [14,16,17,36,37]. Previous studies from our lab demonstrated that chronic neuroinflammation in EAE resulted in changes in AIS length and protein clustering and this disruption corresponded with increased microglial reactivity and production of pro-inflammatory factors [17]. Furthermore, AIS disruption corresponded with increased contact between reactive microglia and the AIS, suggesting that in a chronic inflammatory environment, proinflammatory microglia may drive AIS disruption [17].
The microglial inflammatory response is amplified by the enzyme NOX2, which is responsible for the microglial respiratory burst and extracellular production of ROS [38]. NOX2 activity has been implicated in the chronic activation of microglia and its deleterious effects both through the production of extracellular ROS and through amplification of the pro-inflammatory response [39][40][41]. Inhibition of NOX2 reduced microglial ROS production and reduced microglia-mediated neurotoxicity [40,42,43]. Here, we investigate the role of microglial inflammation and the ROS-producing enzyme NOX2 on AIS integrity. Using a lipopolysaccharide (LPS)-induced model of neuroinflammation, we demonstrate that in the presence of acute microglial inflammation, AIS ankG clustering is disrupted and upon resolution of inflammation, AIS changes are reversed. Furthermore, ablation of NOX2 preserved AIS integrity. These data underscore the dynamic capabilities of the AIS in the presence of a pathological insult.

Animals
Six-to eight-week-old C57BL/6J mice and NOX2-deficient (B6.129S-Cybb tm1Din /J, NOX2 −/− ) mouse breeding pairs were purchased from Jackson Laboratories (Bar Harbor, ME). All animals were maintained in the AAALACaccredited McGuire Veterans Affairs Medical Center (VAMC) vivarium with access to food and drink ad libitum. NOX2 −/− mice have a targeted mutation of the 91-kD subunit of the oxidase cytochrome b and lack phagocyte superoxide production [44]. NOX2 −/− mice are maintained on a C57BL/6J background; therefore, age-matched C57BL/6J mice (NOX2 +/+ ) were used as controls. All procedures were conducted in accordance with the National Institutes of Health guidelines for the care and use of laboratory animals and were approved by the McGuire VAMC Institutional Animal Care and Use Committee.

Imaging and analysis
Imaging was performed on a Zeiss LSM 710 confocal laser scanning microscope (Carl Zeiss Microscopy, LLC, Thornwood, NY) housed in the VCU Department of Anatomy and Neurobiology Microscopy Facility. For AIS number analysis, images were collected as previously described [17]. Briefly, confocal z-stacks spanning an optical thickness of 25 μm, using a pinhole of 1 Airy disc unit and Nyquist sampling (optical slice thickness, 0.48 μm), were collected from neocortical layer V for each of six sections (spanning 1.1 mm anterior to the bregma to 2.5 mm posterior to the bregma) per mouse resulting in 12 images per animal (n = 4-6 animals per treatment group). Images were then processed and analyzed using FIJI (NIH ImageJ software). Settings were optimized by comparing manual AIS tracings (previously described by [17]) and FIJI automated counts; no significant difference was found between methods (data not shown). Once established, settings remained constant throughout analysis. Thresholds of maximum intensity projections of ankG labeling were automatically set using the Otsu threshold method [54], and AISs were quantified using the "Analyze Particles" plugin (FIJI) (size 0-infinity μm 2 ; circularity 0-0.5; objects touching edges excluded). ankG-positive structures measuring <10 μm were excluded from analysis consistent with previous studies [17,36,37].
For analysis of microglial NOX2 immunoreactivity, confocal z-stacks spanning an optical thickness of 25 μm were collected from neocortical layer V for each of six sections (spanning 1.1 mm anterior to the bregma to 2.5 mm posterior to the bregma) per mouse (n = 3 animals per treatment group). Images were blinded, and NOX2 immunoreactivity in Iba-1 + cells was quantified using Volocity™ 3D Image Analysis Software version 6.3 allowing 3D confirmation of double immunolabeling in each Iba-1 + cell. The total number of microglia and the number of NOX2 + microglia were counted manually for each double-immunolabeled z-stack. Data are presented as the percent of NOX2 + microglia (Iba-1 + ) per field of view.
For neuronal nuclei analysis (NeuN labeling), four confocal images per mouse were collected using a ×20 objective with a numerical aperture of 1.4 and a pinhole of 1 Airy disc unit. Images were processed and analyzed using FIJI (n = 3 mice per treatment group). Settings were optimized by comparing manual NeuN counts (previously described [17]) and FIJI automated counts; no significant difference was found between methods (data not shown). Thresholds of maximum intensity projections of NeuN labeling were automatically set using the Otsu threshold method [55], and neuronal nuclei were quantified using the "Analyze Particles" plugin (size 10-150 μm 2 ; circularity 0-1; objects touching edges excluded). No differences in NeuN + cell counts were detected among any treatment groups (NOX2 +/+ Saline, NOX2 +/+ LPS-injected, NOX2 −/− Saline, or NOX2 −/− LPS-injected; Table 1).
Cortical volume analysis was performed using the Cavalieri principle as previously described (modified [55,56]). Briefly, unbiased stereology was performed using every 15th section from the total sections spanning the cortical region 1.1 mm anterior to the bregma to 2.5 mm posterior to the bregma and analyzed to estimate cortical volume. Each reference space was outlined with a ×2 objective and analyzed using a point-grid analysis, sampling 100% of the regions of interest. Samples were counted in a blind manner and volumes calculated using an Olympus BX51 microscope (Center Valley, PA) and newCAST software (Visiopharm, Hoersholm, Denmark) (n = 3-4 mice per treatment group). No differences in cortical volumes were detected among any treatment groups (NOX2 +/+ Saline, NOX2 +/+ LPS-injected, NOX2 −/− Saline, or NOX2 −/− LPS-injected; Table 1).

Microglia isolation
Adult cortical microglia were isolated using MACS magnetic bead separation (Miltenyi Biotec, San Diego, CA) as described previously [17,45]. Briefly, salinetreated and LPS-treated mice were deeply anesthetized and transcardially perfused with 50 mL ice-cold PBS. After removal of the meninges, the cerebral cortices of two mice were harvested and pooled per sample (2 mice = 1 n) and suspended in Hank's balanced salt solution (HBSS) without CaCl 2 and MgCl 2 (Corning, Corning, NY). A single-cell suspension was prepared using the Miltenyi Neural Tissue Dissociation Kit according to the manufacturer's instructions. The cells were depleted of myelin by suspension in 3 mL of 30% isotonic Percoll™ (GE Healthcare Life Sciences, Pittsburgh, PA) followed by a 10-min centrifugation at 700 x g at 4°C. The cell pellet was washed in 5 mL HBSS without CaCl 2 and MgCL 2 , and isolation of microglia was performed with magnetic CD11b microbeads (Miltenyi) and MACS magnetic separator (Miltenyi) according to the manufacturer's instructions.

RNA isolation and RT-qPCR analysis
Total RNA was extracted from isolated CD11b + cells or whole cortical tissue using a Qiagen RNeasy mini kit (Qiagen, Germantown, MD) and treated with Ambion DNase I (Invitrogen Life Technologies, Grand Island, NY) (n = 3 samples per treatment group). RNA concentrations were determined using a NanoPhotometer (Implen, Los Angeles, CA), and purity was assessed by the ratio of absorbance at 260 and 280 nm (OD 260/280 > 1.8). Oligo-dT-primed complementary DNAs (cDNAs) were synthesized from 0.25 μg of RNA for each sample using the iScript Reverse Transcription Supermix (Bio-Rad) according to the manufacturer's guidelines. RT-qPCR reactions with at least two technical replicates per sample were performed on a CFX96 real-time PCR detection system (Bio-Rad) using 1 μL of cDNA, SsoFast Evagreen Supermix (Bio-Rad), and forward and reverse primers (500 nM). Cycling parameters were 1 cycle of 95°C (5 min), 40 cycles of 95°C (5 s), and 56°C (5 s) followed by a melt curve measurement consisting of 5-s 0.5°C incremental increases from 65 to 95°C. Relative changes in gene expression were calculated by the 2 −ΔΔCt method [57] using cyclophilin A and phosphoglycerate kinase 1 (PGK1) as an endogenous reference gene. Gene-specific primers were designed and checked for specificity using National Center for Biotechnology Information/Primer-BLAST (basic local alignment search tool [58]) ( Table 2). Primers were generated by Integrated DNA Technologies (San Diego, CA).

Calpain activity assay
To quantify the levels of calpain activity and to determine the effect of Calpeptin on inhibition of calpain activity, vehicle-, LPS + vehicle-, or LPS + Calpeptin-treated mice were deeply anesthetized and transcardially perfused with 50 mL ice-cold 0.9% saline at 3 days post-LPS injection. Cerebral cortices (10 mg) were harvested and immediately homogenized in ice-cold extraction buffer (Calpain activity kit). Samples were centrifuged for 5 min at 4°C at 15,000 x g to remove insoluble material. Calpain activity was quantified using a fluorometric calpain activity assay kit (ab65308, Abcam, Cambridge, MA) according to the manufacturer's protocol. All samples were analyzed in triplicate, and calpain activity was measured using a Tecan M1000 PRO microplate reader (Männedorf, Switzerland). Changes in calpain activity were normalized to saline control levels and expressed as relative fluorescent units (RFU).

Statistical analysis
All graphing and statistical analyses were performed using GraphPad Prism version 6.03 (GraphPad Software, San Diego, CA). Data were analyzed by a one-way or two-way analysis of variance and, where mean differences were significant, assessed using Tukey's honest significance difference post hoc analysis. Treatment groups were presented as percent of saline control (% Control ± SEM), and p < 0.05 was considered statistically significant.

LPS-induced inflammation alters AIS protein clustering
Studies from our lab [17] and others [14,16,59] have shown that AIS protein clustering is disrupted during chronic inflammation and disease. To determine if acute neuroinflammation alters AIS integrity, we used a LPSinduced neuroinflammatory model and assessed AIS ankG protein clustering (Fig. 1). AISs were immunolabeled for ankG in saline-and LPS-treated mice at 6 h, 24 h, 3 days, and 1 week post-injection. Disruption of ankG labeling was first observed 24 h post-LPS injection (71.6% ± 3.7, p < 0.01) compared to saline controls (100% ± 5.8) (Fig. 1c, f ). The number of AISs detected in LPS-injected mice remained significantly decreased at both 3 days and 1 week post-injection (72.7% ± 2.5, p < 0.01 and 65.3% ± 5.1, p < 0.0001, respectively) compared to saline controls (Fig. 1d-f ). AIS disruption as indicated by a loss of ankG immunolabeling was confirmed by immunolabeling for NaV1.6 (data not shown). To determine if AIS disruption was a consequence of neuronal loss or changes in cortical volume, we quantified NeuN immunolabeling and cortical volume and found no difference among saline or LPS groups (Table 1). These data suggest that acute neuroinflammation caused a significant disruption in AIS protein clustering, but altered ankG detection was not associated with neuronal loss.

AIS disruption is reversible
The AIS is the site of action potential initiation and thus is critical for neuronal function [60]. Studies have shown that the AIS can undergo structural plasticity in development and in response to pathological insults to sustain proper signaling within neuronal networks [2,6,61,62]. To determine in vivo if inflammation-induced AIS disruptions are reversible, we assessed ankG clustering of AISs in saline-and LPS-treated mice 2 weeks post-injection (Fig. 1). ankG immunolabeling revealed that the number of AISs in LPS-treated mice 2 weeks post-injection returned to baseline and was not significantly reduced compared to saline controls (91.3% ± 2.8) (Fig. 1f, g). Furthermore, AISs at 2 weeks post-LPS injection were significantly increased compared to LPS 1 week treated mice (mean difference 25.9% ± 5.7, p < 0.01) ( Fig. 1e-g). Thus, LPS-induced disruption of AIS ankG clustering is reversible.

AIS integrity coincides with microglial inflammatory response
Previous studies demonstrated that chronic neuroinflammation in EAE resulted in disruption of the AIS, and this disruption coincided with microglial reactivity and increased microglial-AIS contact [17]. Therefore, to better understand how microglial inflammation contributes to AIS disruption in LPS-induced neuroinflammation, we examined microglial reactivity and gene expression of inflammatory mediators in saline-and LPS-treated mice. AIS disruption was first observed 24 h post-LPS injection and remained significantly disrupted until recovery 2 weeks post-injection. Therefore, we assessed microglial reactivity 6 h, 24 h, 3 days, 1 week, and 2 weeks post-injection. Iba-1 immunolabeling revealed that at 6 h post-LPS injection, prior to AIS changes, microglia display a reactive phenotype which is maintained 1 week post-LPS injection (Fig. 2b-e). By 2 weeks post-LPS injection (Fig. 2f ), microglia morphology returned to a surveying phenotype similar to that of saline-injected mice (Fig. 2a). qPCR analysis of isolated microglia from LPS-and saline-injected mice revealed that gene expression of inflammatory mediators Tnf-α, COX-2, and NOX2 was significantly upregulated 6 h post-LPS injection (p < 0.05, Fig. 2g-i). Messenger RNA (mRNA) expression of NOX2, the enzyme responsible for extracellular release of ROS and amplification of microglial pro-inflammatory response [38], was significantly increased at 6 h (p < 0.01) and remained elevated 3 days post-LPS injection (p < 0.05), returning to control levels prior to AIS recovery (Fig. 2i). Furthermore, gene expression of resolving factors Mrc1, TGF-β, and Fizz-1 was significantly decreased by 24 h post-LPS injection and returned to control levels by 1 week post-injection, coincident with AIS disruption and recovery, respectively (p < 0.05, Fig. 2j-l). Thus, microglial inflammation preceded disruption of the AIS, while AIS recovery followed the resolution of microglial inflammation.

Treatment with anti-inflammatory Didox reverses AIS disruption
In LPS-induced inflammation, microglia rapidly respond, displaying dramatic morphological alterations and significantly increasing expression of pro-inflammatory genes while significantly downregulating expression of proresolution factors (Fig. 2). mRNA expression of microglial inflammatory markers remained elevated and was not resolved until 1 week post-LPS injection. However, AIS clustering remained disrupted until 2 weeks post-LPS injection (Fig. 1). To determine if AIS recovery could be accelerated by a therapeutic approach, we treated LPS-injected mice with the anti-inflammatory and free radical scavenger Didox [47,[63][64][65]. Didox administration was initiated 24 h post-LPS injection (Fig. 3e), when microglia were reactive and AISs were significantly disrupted. Following treatment of saline-and LPS-injected mice, AISs were immunolabeled for ankG at 24 h and 1 week post-injection. At 24 h post-injection, ankG clustering in LPS-treated mice (71.6% ± 3.7, p < 0.01, Fig. 3b, f ) was significantly disrupted compared to saline controls. The number of AISs in LPS-injected mice remained significantly decreased 1 week post-injection (65.3% ± 5.1, p < 0.001, Fig. 3c, f ) compared to saline controls. However, in LPS + Didox mice, the number of AISs was significantly higher compared to that in LPS 1 week untreated mice (96.3% ± 2.8, p < 0.01) and was not significantly different from saline controls (Fig. 3d, f ). Thus, treatment with Didox reversed AIS disruption and increased the rate of AIS recovery. Maintenance of this pro-inflammatory phenotype is supported by the persistence of significantly elevated NOX2 expression which does not return to baseline levels until 1 week post-injection. Similarly, markers associated with resolution of inflammation are reduced following LPS injection, and expression of these markers does not return to control levels until 1 week post-injection. mRNA expression was evaluated by quantitative RT-PCR; values were normalized using the 2 −ΔΔCT method and were reported as mean expression ± SEM. An asterisk indicates significant difference (p < 0.05) from saline, and a ¥ indicates a difference between time points post-LPS injection

Treatment with Didox alters microglial NOX2
Didox is a ribonucleotide reductase inhibitor which modulates the inflammatory response through inhibition of NFκβ activation, reduction in ROS-producing enzymes, and reduction in oxidative injury [47,49,66]. To determine the effect of the treatment with the anti-inflammatory and free radical scavenger Didox on microglial inflammatory response and AIS integrity following LPS treatment, we analyzed microglial NOX2, which is dependent on, and induced by, NF-κβ [67,68]. NOX2 is a ROS-producing enzyme primarily expressed by microglia and has been implicated as the primary producer of extracellular ROS and oxidative stress in the CNS [39,40,69,70]. Quantitation of NOX2 immunolabeling (Fig. 4A-D) in cortical microglia from saline-and LPS-treated mice with or without Didox treatment revealed that NOX2 immunoreactivity was significantly enhanced in microglia 24 h post-LPS injection (Fig. 4B, E) and remained significantly elevated 1 week post-LPS injection (Fig. 4C, E) compared to saline controls (p < 0.0001, Fig. 4A, E). However, NOX2 immunoreactivity was decreased in LPS-injected mice treated with the anti-inflammatory and free radical scavenger Didox (p < 0.0001, Fig. 4D, E). Thus, Didox treatment significantly decreased microglial NOX2 back to saline control levels, and this decrease corresponded with the reversal of AIS disruption.

Ablation of NOX2-derived ROS production prevents AIS disruption
Although Didox is a known free radical scavenger and our data demonstrate a reduction in NOX2 expression resulting from Didox treatment, it is possible that Didox targets other inflammatory factors and that the observed AIS recovery was coincidental with inhibition of NOX2 expression. Therefore, to more specifically investigate the role that NOX2 plays in AIS disruption, we exploited NOX2 −/− mice. We injected NOX2 +/+ and NOX2 −/− mice with saline or LPS and assessed AIS integrity 24 h and 1 week post-injection. ankG clustering in NOX2 +/+ LPSinjected mice was significantly disrupted at 24 h (Fig. 5b) and 1 week (Fig. 5c) post-injection resulting in an~30% loss of AISs, compared to saline controls (p < 0.01, Fig. 5g). Strikingly, ankG clustering in LPS-injected NOX2 −/− mice was not significantly different at either 24 h or 1 week post-injection compared to saline-injected NOX2 −/− mice (Fig. 5e, f). However, ankG clustering was significantly higher at both 24 h and 1 week compared to NOX2 +/+ LPS-injected mice (p < 0.01, Fig. 5g). Thus, ablation of NOX2-derived ROS prevented inflammation-induced AIS disruption.

Inhibition of calpain prevents AIS disruption
To further elucidate the mechanism of inflammationinduced AIS disruption, we investigated the calciumactivated protease calpain. Calpain activity has been implicated in AIS structural changes and the targeted proteolysis of AIS proteins [9,10,14]. To determine if calpain activity is involved in inflammation-induced loss of AIS ankG clustering, we treated LPS-injected mice with the calpain inhibitor Calpeptin. Calpeptin administration was initiated 30 min prior to LPS injection and continued once daily for 2 days (Fig. 6d). Following treatment, AISs were immunolabeled for ankG at 3 days post-LPS injection. ankG clustering in LPS-treated mice (72.7% ± 2.5, p < 0.01, Fig. 6b, e) was significantly disrupted compared to saline controls (100% ± 5.8, Fig. 6a, e). However, in LPS + Calpeptin-treated mice, the number of AISs was significantly higher compared to that in LPS 3 days vehicle-treated mice (90.2% ± 2.3 and 72.8% ± 2.5, respectively, p < 0.05, Fig. 6c, e) and was not significantly different from saline controls. Inhibition of calpain activity by Calpeptin was determined by a fluorometric calpain activity assay on mouse cortical homogenates 3 days post-LPS injection (Fig. 6f ). Calpain activity was significantly increased in LPS 3 days mice (132.9% ± 3.0, p < 0.01, Fig. 6f ) compared to saline controls (100% ± 5.3), and treatment with Calpeptin significantly reduced calpain activity in cortical homogenates (84.1% ± 3.1, p < 0.001, Fig. 6e). Thus, inhibition of calpain activity prevents inflammation-induced disruption of AIS ankG clustering in cortical neurons.

Discussion
In this study, we demonstrate that LPS-induced neuroinflammation disrupts protein clustering at the AIS concomitant with the microglial inflammatory response resulting in an~30% loss of AIS detection. Importantly, we found that inflammation-induced AIS disruptions were reversed following resolution of microglial inflammation and changes in AIS ankG clustering are NOX2mediated and dependent on calpain activity. Thus, in the presence of acute microglial inflammation, the AIS undergoes an adaptive change that is capable of The AIS has the capacity to adapt and recover The AIS is targeted for disruption in injury and disease emphasizing its need for homeostatic adaptations. Indeed, many studies [16,17,37,[71][72][73] have shown that the AIS is plastic, undergoing change in response to various stimuli. However, few studies have demonstrated that these changes are reversible. Alterations in AIS length [12] and location [2] caused by changes in neural activity were reversible in vitro; however, loss of AIS protein clustering due to ischemic insults in vitro were not, even in the absence of cell death [14]. A previous study examining AIS integrity after stroke observed axonal sprouting resulting in an increase in small, immature AISs demonstrating reparative potential of this domain [59]. Furthermore, our lab previously reported that shortening of AIS length is reversible following treatment with the anti-inflammatory Didox [17].
Here, we provide evidence that loss of AIS protein clustering is spontaneously reversible, independent of axonal sprouting. Moreover, we show that by modulating the neuroinflammatory response using therapeutic intervention, the rate of AIS recovery can be increased, even after significant AIS disruption has occurred. These data suggest that while insults at the AIS, such as ischemia [14], can cause irreversible damage, the AIS has the capacity to adapt and recover after insult. The mechanism by which this occurs or what the extent of injury is after which the AIS cannot recover remains to be determined.

Microglial phenotype influences AIS integrity
Although AIS plasticity can be triggered by both pathological and non-pathological stimuli, the events that drive plasticity remain largely unknown. Recently, Baalman et al. [16] established a relationship between microglia and the AIS, revealing that microglia contact AISs early in development and throughout adulthood in the uninjured brain, suggesting an important interaction that may influence neuronal excitability. In a model of chronic neuroinflammation, reactive microglia increased contact with AISs, and this contact both preceded AIS disruption and increased with disease progression, suggesting that in a chronic inflammatory environment, increased microglial contact may drive AIS disruptions [17]. Consistent with previous findings, we found that reactive microglia contact the AISs during LPS-induced neuroinflammation. However, contrary to findings in the chronic inflammatory model, the amount of contact made by microglia did not increase throughout the course of inflammation and did not correlate with AIS disruption (data not shown). The microglial inflammatory profile, however, did correspond with AIS disruption and recovery. Furthermore, modulation of the inflammatory profile using anti-inflammatory treatment increased the rate of AIS recovery. Though our findings suggest that changes in microglial inflammation correspond with AIS alterations, it is possible that these changes do not directly influence AIS integrity. However, the direct association of microglia with the AIS suggests this axonal domain may be particularly vulnerable to changes in microglial reactivity. Thus, our findings suggest that AIS integrity may be influenced by microglial phenotype, with a pro-inflammatory phenotype driving AIS disruption while a resolving phenotype hastens repair. Consistent with this premise, Klapal et al. [74] showed that incubation of hippocampal cultures with activated microglia or the pro-inflammatory cytokine Tnf-α increased neuronal excitability. In contrast, incubation with the pro-resolution factor TGF-β decreased Na + current density to control levels. Together, these findings suggest that neuroactive factors released by microglia augment neuronal excitability, which drives AIS structural changes [2,6,12]. Here, we demonstrate that AIS structure is altered following significant increases in microglial expression of Tnf-α. Furthermore, this AIS pathology is reversed after expression of microglial TGF-β is enhanced. Thus, our findings are consistent with microglial neuroactive factors driving changes in neuronal activity and AIS structural plasticity.

NOX2-mediated ROS, calpain, and AIS changes
During insult, pro-inflammatory microglia increase expression of inflammatory mediators and ROS-producing enzymes [30,40,45,75]. ROS are highly reactive and diffuse signaling molecules that regulate cell functions through redox modification of target proteins. ROS can result in further production of reactive species [76,77] and elevated calcium levels [78], which have been implicated in AIS disruption. In this study, we show that changes in microglial expression of ROS-producing enzymes correspond with AIS disruption and recovery, suggesting a role for microglial ROS in inflammationdriven AIS disruption. Consistent with this premise, ablation of NOX2 prevented AIS disruption. Though NOX2 is primarily expressed by microglia, NOX2 is also present in cortical neurons, where it plays a role in ROS regulation and calcium dynamics [79]. Therefore, NOX2 ablation may preserve AIS integrity through both the prevention of microglial ROS release and neuronal NOX2 ROS production, both of which may converge on pathways resulting in AIS changes.
Reactive species such as hydrogen peroxide and nitric oxide influence calcium-permeable channels including L-type Ca( 2+ ) [78,80,81] and TRPM channels [82,83]. Upon activation, intracellular calcium concentrations rise, resulting in the subsequent activation of calcium-regulated proteins such as calpain [9,10,14], CamKII [8], and calcineurin [8,12], which have been implicated in AIS disruption. Consistent with previous studies [9,10,14], our data implicate the calciumdependent protease calpain as a mediator of AIS structural changes. In this study, we demonstrate that acute neuroinflammation increases calpain activity consistent with disruption of AIS ankG clustering and inhibition of calpain prevents inflammation-induced disruptions. Together, our data suggest that NOX2derived ROS and calpain activity are drivers of AIS structural plasticity during acute neuroinflammation.

Conclusions
In conclusion, we demonstrate that in the presence of acute neuroinflammation, protein clustering at the AIS is altered. Importantly, our data demonstrate that this AIS disruption is reversible and that the AIS has the capacity to adapt and spontaneously recover. Furthermore, we reveal that inflammation-driven plasticity at the AIS is mediated by NOX2 and calpain activity.