key: cord-0314483-snkpbgsp authors: Calabrese, Barbara; Jones, Steven; Yamaguchi-Shiraishi, Yoko; Lingelbach, Michael; Manor, Uri; Svitkina, Tatyana M; Higgs, Henry N; Shih, Andy Y; Halpain, Shelley title: INF2-mediated actin filament reorganization confers intrinsic resilience to neuronal ischemic injury date: 2021-10-01 journal: bioRxiv DOI: 10.1101/2021.10.01.462790 sha: d95c2aa4a0e438c6228a8e45b37555c550f8244e doc_id: 314483 cord_uid: snkpbgsp During early stages of ischemic brain injury, glutamate receptor hyperactivation mediates neuronal death via osmotic cell swelling. Here we show that ischemia and excess NMDA receptor activation – conditions that trigger neuronal swelling -- cause actin filaments to undergo a rapid and extensive reorganization within the somatodendritic compartment. Normally, F-actin is concentrated within dendritic spines, with relatively little F-actin in the dendrite shaft. However, beginning <5 min after incubation of neurons with NMDA, F-actin depolymerizes within dendritic spines and polymerizes into long, stable filament bundles within the dendrite shaft and soma. A similar “actinification” of the somatodendritic compartment occurs after oxygen/glucose deprivation in vitro, and in mouse brain after photothrombotic stroke in vivo. Following transient, sub-lethal NMDA exposure these actin changes spontaneously reverse within 1-2 hours. A combination of Na+, Cl-, water, and Ca2+ entry are all necessary, but not individually sufficient, for induction of actinification. Spine F-actin depolymerization is also required. Actinification is driven by activation of the F-actin polymerization factor inverted formin-2 (INF2). Silencing of INF2 renders neurons more vulnerable to NMDA-induced membrane leakage and cell death, and formin inhibition markedly increases ischemic infarct severity in vivo. These results show that ischemia-induced actin filament reorganization within the dendritic compartment is an intrinsic pro-survival response that protects neurons from death induced by swelling. Ischemic stroke results from occlusion of cerebral blood vessels, causing brain tissue injury due to loss of oxygen and glucose supply 1 . It is a leading cause of death and chronic disability, and has enormous public health implications, especially recently in the context of Covid-19, which significantly increases a severely ill patient's risk of stroke [2] [3] [4] [5] . Beyond anticoagulant therapies, there are relatively few emergency treatment options for stroke. This highlights the need for a deeper understanding of the biological cascades ensuing from an ischemic event, including cellular pro-survival mechanisms that could minimize the ultimate brain tissue damage. Although apoptosis accounts for much of the delayed neuronal death that occurs over days and weeks following a stroke, most of the neuronal death that occurs in the early hours after a stroke is due to a pathological swelling of neurons that leads to disruption of plasma membrane integrity [6] [7] [8] [9] . Neuronal swelling, which is also called cytotoxic edema, is triggered when catastrophic ATP depletion perturbs ionic balance, leading to a massive influx of ions through multiple entry routes, with cation entry through NMDA receptors and chloride entry through the SLC26A11 ion exchanger playing major roles 10 . Neuronal depolarization spreads in waves from the site of initial ischemia via local release of glutamate, thereby exacerbating and extending the initial damage 11 . NMDA receptor hyperactivation plays an especially critical role, as administration of NMDA receptor antagonists before or even just after the onset of ischemia results in significantly reduced infarct volume in experimental models [11] [12] [13] . Neurons, like most other cells, respond to cellular stress and injury, including ischemic injury, by mounting a series of pro-survival responses, with changes in organelles, reduction in protein synthesis, and activation of pro-survival genes 14 . In contrast, the pro-survival functions of the cytoskeleton are less well characterized. Here we describe an extensive reorganization of neuronal filamentous actin (F-actin) that occurs in response to stroke, oxygen and glucose deprivation, or NMDA receptor hyperactivation. F-actin is rapidly depolymerized within dendritic spines [14] [15] [16] . Simultaneously, we show here, it polymerizes extensively within the soma and dendrites. This result was surprising given the usual ATP dependence of actin polymerization and the reduced ATP availability during ischemia. We find that this F-actin response is pro-survival and selectively triggered by conditions that elicit neuronal cytotoxic edema. The F-actin build-up in the soma and dendrites results in long, slowly turning over actin filaments that persist while the stress is present. However, the F-actin reorganization spontaneously reverses if the stress is transient. We demonstrate that activation of inverted formin-2 (INF2) is a key mediator of this neuronal pro-survival response. To investigate actin filament responses to stroke-like conditions we exposed cultured rat brain hippocampal neurons to oxygen and glucose deprivation (OGD). OGD is a well-established in vitro model for investigating neuronal cellular and subcellular responses to hypoxia and ischemia 17 . OGD induced a dramatic reorganization of F-actin within the soma and dendrites of neurons. Over 2-6 hours an increasing fraction of neurons showed a substantial loss of phalloidin staining for F-actin within dendritic spines, where F-actin is normally concentrated, and an aberrant accumulation of F-actin within the somatodendritic compartment (Fig 1A-C) . Filaments accumulated throughout the interior of the soma and proximal dendrites, and to various extents within more distal dendrites. Because in most spiny neurons the soma and shaft of the dendrite is fairly devoid of F-actin relative to dendritic spines, we call this novel phenomenon the "actinification" of the neuron. OGD-induced somatodendritic actinification was completely blocked in the presence of the NMDA receptor antagonist (2R)-amino-5-phosphonovaleric acid (APV; Fig 1C) . Interestingly, oxygen-deprivation alone was insufficient to induce actinification within the same time frame (Fig 1C) , perhaps because neurons sustain partial energy production via glycolysis for extended periods 18 . Together, these results suggest that actinification is induced by a catastrophic loss of ATP, leading to excess glutamate release and NMDA receptor hyperactivation. We investigated whether ischemia in vivo also induces neuronal actinification using either wildtype mice or transgenic Thy1 promoter-driven YFP expressing mice, the latter of which allowed us to readily identify dendritic arbor morphology in layer 2/3 and 5/6 cortical pyramidal neurons. Photothrombotic occlusion in single penetrating arterioles within the somatosensory cortex of mouse or rat brain induces small infarcts with well-defined borders 12, 19 . Such strokes are greatly attenuated by application of NMDA antagonists before or immediately following blood vessel occlusion 12 . Within 2-6 hours the infarct region induced by unilateral single vessel photothrombosis was identifiable using a variety of markers, including fluorescent hypoxyprobe to detect severely hypoxic tissue, anti-IgG to detect vascular leakage, and reduced immunostaining for the neuron-specific proteins MAP2 and NeuN 20, 21 (Fig 1D; Suppl Fig S1) . We noted that, consistent with a previous report 21 , the strongly reduced immunoreactivities for NeuN at this early time point reflected a loss of antibody staining, i.e., epitope loss, rather than a reduction in overall cell number, since DAPI staining for nuclear DNA and cytoplasmic YFP labeling were retained in neurons lacking NeuN (Suppl Fig S2) . Actinification was detected within infarcts induced in either wildtype or Thy1-YFP mice. Phalloidin staining revealed that in control brain tissue, as in control cultured neurons, the majority of F-actin was concentrated in dendritic spines (Fig 1E, Suppl Fig S3) . By 4-6 hours after arteriole occlusion, the dendrites of GFP-expressing layer 5/6 pyramidal neurons showed a substantial loss or shrinkage of dendritic spines and the appearance of dystrophic dendrites, (Suppl Fig S4) . Low-magnification imaging of the infarct region indicated there was a timedependent decrease in overall F-actin concentration, consistent with a loss of dendritic spine Factin (Suppl Fig S5) . However, high magnification imaging of individual layer 5/6 YFP-positive neurons showed aberrant accumulations of F-actin bundles (Fig 1E) . Quantitative analysis revealed that within the ischemic core of the infarct (defined by the region of NeuN loss) there was a significant increase in the number of actinified neurons compared to control neurons in either sham treated mouse brains, or in the stroke condition within contralateral cortex, or within ipsilateral temporal lobe cortex distant from the infarct region. Within the core of the infarct, nearly 90% of the YFP-labeled neurons showed actinification. (Fig 1F) . Dendritic spine shrinkage and spine F-actin loss have been described previously as an early response to stroke or to NMDA receptor hyperactivation, 15, 16, 22, 23 and mechanisms contributing to spine F-actin loss following strong NMDA receptor activation have been investigated previously 16, [24] [25] [26] . However, the significant accumulation of F-actin in the somatodendritic compartment was unexpected, given that F-actin polymerization typically involves ATP consumption, and would logically seem to be disfavored in conditions of hypoxic cellular stress, when energy supplies become severely limited. We therefore focused on understanding the mechanism and relevance of somatodendritic actinification. Since an NMDA receptor antagonist completely prevented somatodendritic actinification following OGD, we asked whether actinification could be triggered via direct activation of NMDA receptors. Incubation of cultured hippocampal neurons with 50 µM NMDA induced a timedependent increase in neuronal actinification, accompanied by extensive loss of dendritic spine F-actin and shrinkage of spines, similar to that seen following OGD in vitro and stroke in vivo ( Fig 2A) . However, the time course for the actinification response was greatly accelerated. Based on a time series collected using fixed cultures, half-maximal actinification (the time at which 50% of all neurons were actinified) occurred after 5-10 minutes (average t1/2 = 7 min) in the sustained presence of NMDA, reaching nearly 100% by 60 minutes (Fig 2B) . We carried out time-lapse imaging of individual hippocampal pyramidal neurons using either Lifeact-mRFP, mApple-F-tractin, or the small molecule dye far-red silicon-rhodamine actin. The time course of live neurons undergoing actinification was consistent with the time course determined using fixed populations of neurons. Dendrite actinification occurred in a time frame that closely overlapped with the observed decrease in dendritic spine F-actin (Fig. 2C, D) . As observed in fixed samples, live imaging showed that neurons underwent actinification after variable delays ranging from 3-20 minutes post-NMDA. However, once detectably initiated, actinification proceeded rapidly and with a similar time course. The rapid time course of actinification initiated by NMDA suggests that the aberrant accumulation of F-actin filaments is likely due to an enzymatically-driven process, rather than the slower aggregation type process that governs build-up of pathological aggregates seen in neurodegenerative diseases such as Alzheimer's. To investigate the characteristics of actinified dendrites at the ultrastructural level, we carried out platinum replica electron microscopy (PREM) on control cultures and compared them to cultures incubated with 50 µM NMDA for 5 or 30 min ( Fig 2E) . PREM images from control neurons revealed numerous clusters of short, moderately branched actin filaments localized within dendritic spine-like structures along the dendrite shaft, consistent with previous studies 27 . Within the dendrite shaft of control neurons, actin filaments were relatively sparse. Long filaments running parallel to the main dendrite axis were rarely observed, and when detected they usually exhibited occasional branches. In contrast, and in agreement with our light microscopy observations, PREM images of dendrite shafts from NMDAtreated cultures revealed numerous instances of long, mostly unbranched filaments, which often formed irregular bundles. In parallel, there was a substantial decrease in the frequency of spinelike protrusions containing branched actin filaments that were common in control neurons. These observations suggest that dendrite actinification is driven by a process that disassembles branched actin filaments in spines and polymerizes F-actin into unbranched, rather than branched, filaments in dendrite shafts. Once formed, actin filaments in the actinified neuronal somata and dendrites appeared to be highly stable. We tested whether incubation with latrunculin A would accelerate the removal of the actinified F-actin in the dendrite shaft. Actin filaments within dendritic spines turnover with a half-time of < 1 minute 28, 29 . We reasoned that if a similarly high turnover rate of F-actin characterizes the actinified dendritic compartment, we should observe that application of the Gactin sequestering compound Latrunculin A -applied after actinification has occurred --would greatly speed the net disassembly of the filaments. However, our observations did not support this hypothesis. Incubation with 2 µM LatA for up to 2 hours after actinification induction caused no detectable decrease in the percentages of actinified neurons, nor loss of F-actin staining intensity within individual neurons in the soma and dendrites (Fig 3 A-C) . Note that we first confirmed that LatA would prevent actinification, as expected, when applied before actinification was induced by NMDA (Fig. 3A) , consistent with actinification being an actin monomerdependent polymerization process. The resistance of filaments to latrunculin added postactinification suggests that the newly polymerized F-actin induced by NMDA is extremely stable in the continued presence of NMDA. Despite the apparently non-dynamic nature of the F-actin induced by NMDA, the actinification filaments were able to spontaneously disassemble upon removal of NMDA. When cultures were incubated with NMDA for 5 min, followed by addition of MK-801 to prevent further NMDA receptor activation, the fraction of neurons in fixed cultures showing actinification dropped from about 60% to less than 20% within 1 hour (Fig 3D) . By 3 hours only 8% of neurons remained actinified, and by 12 hours only 4% of neurons remained actinified. Over this same time frame there was no significant decrease in the number of live neurons, although there was a trend toward decreased neuronal viability by 24 hours (Fig. 3D ). On average, the time for half-maximal recovery after a 5 min incubation with NMDA was approximately 30 min. Time lapse imaging of individual neurons similarly documented that actinification was reversible, and that F-actin within the somatodendritic compartment spontaneously returned to a controllike distribution following arrest of the ongoing NMDA stress. Fig 3E shows a neuron expressing mApple-F-tractin, illustrating that the NMDA-induced accumulation of F-actin began to detectably clear from the dendrite shaft within 17 minutes of the addition of NMDA antagonists. Simultaneously, F-actin re-emerged in dendritic spines, thus returning to a qualitatively normal distribution within 1 hour (Fig 3E) . These observations indicate that actinification persists only while the NMDA receptor hyperactivation persists, and that endogenous mechanisms support the spontaneous depolymerization of F-actin in the dendrite shaft/soma and the repolymerization of F-actin in dendritic spines. Most of the early neuronal cell death induced during ischemia occurs via pathological cell swelling (also called cytotoxic edema) 30, 31 . We observed that somatodendritic actinification occurred in parallel with swelling of the cell body and adjacent dendrite (Fig 4A) . We therefore asked whether actinification was provoked by the same conditions that cause osmotic cell swelling, namely influx of both sodium and chloride, followed by water. First, we observed that replacement of extracellular sodium chloride by N-methyl-D-glucamine chloride completely prevented NMDAinduced actinification (Fig 4D) , consistent with the hypothesis that sodium flux across the plasma membrane is critical for actinification. Recent studies demonstrated that voltage-gated chloride influx through the solute carrier family 26 member 11 (SLC26A11) ion exchanger drives neuronal Clentry during glutamate-induced neuronal cell swelling 10 . We found that actinification exhibited a similar pharmacological profile for chloride influx inhibitors as that reported for cytotoxic edema in brain tissue 10 . DIDS and GlyH-101, which inhibit Clentry via SLC26A11 10 , significantly inhibited actinification, but neither bumetanide, which blocks the brain NKCC1 cation-chloride cotransporter, nor 5-nitro-2-(3-phenylpropylamino) benzoic acid (NPPB), which blocks volume regulated Clchannels and Ca 2+ -activated Clchannels, were effective in preventing actinification ( Fig 4B) . This indicates that actinification might require chloride entry via the same routes as reported for neuronal cell swelling. We next tested whether water entry was necessary for induction of actinification. Incubation of cultures with the non-cell permeable sugar mannitol, to reduce the osmotic driving force for water entry, indicated that swelling was indeed contributing to NMDA-induced actinification ( Fig 4C) . NMDA receptors are highly permeable to calcium ions, and increased intracellular calcium is a trigger for many biochemical cascades elicited by NMDA receptor activation. Removal of extracellular calcium completely prevented neuronal actinification ( Fig 4D) . However, depletion of intracellular stores using thapsigargin (1 µM and 20 µM) had no effect in preventing actinification (data not shown). These data indicate that calcium influx across the plasma membrane is required to trigger actinification. However, calcium flux alone was insufficient, because although 10 min incubation with the calcium ionophore ionomycin induced substantial actinification, even in the presence of NMDA receptor antagonists, this effect was completely blocked when extracellular sodium was removed ( Fig 4E) . Collectively, our results indicate that sodium, chloride, and calcium influx are all required to induce actinification, and that water entry and consequent neuronal swelling is also a critical factor. We next turned to identifying the key actin mechanisms involved in catalyzing neuronal actinification. The polymerization of most actin filaments in cells are initiated via two distinct mechanisms. The Arp2/3 complex nucleates and elongates daughter filaments from the side of an existing actin filament, thereby forming branched F-actin networks, as seen in lamellipodia and dendritic spines 32-34 . Conversely, formin-driven F-actin polymerization induces formation of unbranched actin filaments, as seen in filopodia and other structures where straight filaments predominate 32, 35, 36 . We applied small molecule inhibitors of formin-mediated and Arp2/3mediated actin polymerization, respectively. Only the formin inhibitor prevented NMDA-induced actinification ( Fig 5A) . This result is consistent with the ultrastructural observation that NMDA induced long, unbranched actin filaments (Fig 2D) . Formins constitute a large superfamily of molecules, with fifteen mammalian formin genes identified to date 37 . Our attention was drawn in particular to inverted formin 2 (INF2), a member of the diaphanous subclass of formins. We found that immunoreactivity for INF2 is present throughout the somatodendritic domain of cultured hippocampal neurons, and distributes in a punctate fashion within both proximal and distal dendrites ( Fig 5B) . No such staining was observed when endogenous INF2 was depleted from individual neurons via RNA interference (Suppl Fig. S6 ). We observed a gradient of INF2 immunoreactivity within the dendritic arbor, with highest intensity in the soma and proximal dendrites, which diminished gradually toward the more distal dendrites (Fig. 5B ). This distribution gradient in control neurons resembles that of actinification itself. Interestingly, we observed some variability in the relative concentrations of INF2 immunoreactivity across neurons within the same culture. We treated cultures for 5 min with NMDA (a time when typically 35-60% the neurons have become actinified) and quantified the level of immunoreactive staining for INF2 in the soma of actinified neurons versus non-actinified neurons ( Fig 5C) . We observed a significantly higher degree of INF2 staining in the neurons that were the "early responders" -i.e., those that underwent actinification at this early time point. This bias in endogenous INF2 level implies that the concentration of INF2 may at least partly determine the strength or timing of the actinification response in individual neurons. To test this hypothesis directly, we transfected neurons with a constitutively active form of INF2 and observed that neuronal soma and dendrites showed actinification even in the absence of NMDA exposure (Fig. 5D ). More importantly, neurons transfected to ectopically express a wildtype form of INF2 showed no increase in actinification in the absence of NMDA, but showed a dramatically enhanced actinification response when stimulated with NMDA for 5 min, with nearly all the neurons becoming actinified within this short time frame (Fig 5D) . Conversely, silencing of endogenous INF2 expression using an shRNA completely blocked the actinification of the neuron, an effect that was rescued in the presence of an shRNA-resistant form of wildtype INF2 (Fig 5E) . Both CAAX and non-CAAX variants of INF2-wt 38 were tested, and as we observed no statistical difference in the degree of rescue between them we combined the data into a single group. Together, these observations implicate INF2 as a critical driver of NMDAinduced actinification. They also indicate that INF2 activity normally remains low under control conditions regardless of INF2 concentration, and must be induced by a stimulus, since overexpression of wildtype INF2 had no detectable effect on actinification under resting conditions. Recent studies showed that INF2 is maintained in an inactive conformation in cells 39 and is then activated by stimuli that raise intracellular calcium 40-42 . One mechanism for this activation might be an increase in G-actin levels, since INF2 is known to be stimulated by elevated actin monomer concentrations 39 . We therefore hypothesized that the G-actin released during NMDA-induced depolymerization of spine F-actin might activate INF2. In support of this hypothesis, adding jasplakinolide during the 5 min NMDA incubation completely prevented both spine F-actin disassembly and dendritic actinification (Fig 5F) . INF2 has also been proposed to be negatively regulated by its binding to a complex consisting of cyclase-associated protein (CAP) and acetylated-G-actin. This complex is disrupted when G-actin becomes deacetylated, thereby allowing activation of INF2 43, 44 . We observed that incubation of cultured neurons with the acetyltransferase inhibitor C646 strongly induced actinification, an effect that was not blocked by NMDA antagonists, consistent with a role for acetylation in regulating actinification ( Fig 5G) . Notably, the C646-induced actinification was almost completely blocked by preincubation with the formin inhibitor SMIFH2 (Fig. 5G ), similar to actinification induced by NMDA (Fig 5A) , suggesting they are mediated by the same INF2-dependent mechanism. We therefore hypothesized that INF2 activity might become induced in response to NMDA via the activation of cytosolic deacetylases (which are typically called histone deacetylases, or HDACs, even though it is now known that they can deacetylate numerous cytosolic substrates). As shown in Table 1 , pre-incubation with several inhibitors of Class 1 or Class 2 HDACs modestly reduced NMDA-induced actinification, but none were robustly effective, including the HDAC6 inhibitor tubastatin, which blocked INF2 activity in non-neuronal cells 44 . Moreover, combinations of multiple inhibitors qualitatively showed little or no additive effect toward inhibiting actinification. Therefore, while our data suggest that de-acetylation is a factor in regulating neuronal actinification, the precise mechanisms that lead to INF2 activation in response to NMDA-induced cellular edema might not require deacetylase activity. Previous studies have reported that excess neuronal glutamate or hypoxic stress induce the accumulation of aberrant F-actin containing structures called cofilin-actin rods, which are defined by the concentrated presence of the actin severing protein ADF/cofilin along bundles of F-actin 45 . Such actin rods are not detectable by phalloidin staining, since the prominent binding of cofilin prevents the binding of phalloidin in such filaments 46 . Four lines of evidence indicate that the actinification of the dendrite compartment is a different process than the formation of cofilinactin rods. First, unlike cofilin-actin rods, the actin filaments we observe are clearly labeled by phalloidin. Second, these filaments do not contain high concentrations of cofilin immunoreactivity (Suppl Fig S7A) . Third, the spatiotemporal dynamics of actinification and the emergence of cofilin-actin rods differ qualitatively. Cofilin-actin rods reportedly appear in dendrites only after 30 min or more of continuous exposure to glutamate or NMDA 45 . Although we can detect cofilin-actin rods after a lengthy exposure of our cultures to glutamate (Suppl. Fig. S7B1 and B2), typically they appear mainly in the distal dendrites and along axons surrounding the dendritic arbor. In contrast, actinification occurs within less than 5 minutes and is preferentially detected in the cell body and the proximal dendrites. Finally, even cofilin rods induced without NMDA by ectopic expression of wild type cofilin together with one of its activating phosphatases, chronophin do not promote neuronal actinification (Suppl Fig S7C) . Together, these observations strongly argue that somatodendritic actinification is a novel process that is distinct from cofilin-actin rod formation. We next investigated the functional impact of INF2-dependent actinification. Because the distribution of actin filaments spontaneously returned to control values after cessation of the stressful stimulus, we reasoned that actinification might be a pro-survival response. To test this hypothesis, we incubated cultures with NMDA for 1 or 4 hours and compared neuronal survival using the VivaFix cell viability assay in neurons that were either transfected with shRNA against INF2 or with an empty vector ( Fig 6B) . For both conditions, neurons were co-transfected with eGFP to identify transfected cells, and we quantified the percentage of transfected neurons that took up the VivaFix dye as an indicator of cell death ( Fig 6A) . We observed that after either a 1 hour or 4 hours incubation with NMDA the prior silencing of INF2 approximately doubled the fraction of non-viable neurons. By 24 hours after incubation with NMDA for 1-4 hours, the vast majority of neurons had died. However, a very small number of live neurons were still reliably detected (i.e., they excluded the VivaFix dye), even after this strong excitotoxic stimulus. Remarkably, 100% of these late-surviving neurons displayed actinification (Fig 6C-D) . Conversely, none of the neurons that were identified as dead showed actinification (Fig 6C-D) , and, indeed, phalloidin staining was depleted in dead neurons. We next asked whether INF2-driven actinification plays a role in neuronal survival following ischemic stroke. Focal stroke was induced using single vessel photothrombosis in mouse cortex. Brains were fixed 4 hours after induction of vessel occlusion, and Fluorojade C was used as a marker of early cell death in post-fixed histological sections to evaluate the effect of inhibiting formin activation using the compound SMIFH2. SMIFH2 or vehicle were applied 4 hours prior to stroke induction using an agar-saturated plug gently placed over the thinned skull. We found that pre-treatment with the formin inhibitor induced a near doubling of cell death after stroke, compared to that observed with vehicle treatment ( Quantification of the ischemic volume indicated that formin inhibition induced a non-significant trend toward increased infarct volume at 4 hours post-occlusion (Fig 7C and D) . Moreover, we observed a substantial increase in apparent damage to cortical tissue within the core of the infarct, with variably sized cavities appearing after strokes induced in the presence of SMIFH2. Such cavitation was never observed in non-infarcted brain regions nor in vehicle treated brains either with or without stroke. Despite these indicators of enhanced tissue damage following stroke, SMIFH2 did not induce a significant increase in the leakage of immunoglobulin from compromised blood-brain barrier following vessel occlusion (Fig 7C and D ). In addition, direct measurement of unoccluded arteriole diameters in the vicinity of the occluded vessel showed no differences between vehicle or SMIFH2 treatment (Fig 7E) , indicating that the drug did not alter local blood flow dynamics prior to stroke induction. These data indicate that INF2 normally helps protect ischemic brain tissue during the acute phase of infarct development. Cellular pro-survival responses engage multiple subcellular events to enable cells to endure periods of transient stress 47 . Excitable cells like neurons and cardiomyocytes are especially vulnerable to osmotic stress because perturbed flux through various ion channels can lead to osmotic imbalance. Swelling in these cells and tissues endangers the survival of the organism. Due to the critical role of ATP-dependent ion pumps in osmoregulation, catastrophic reduction in cellular ATP throws such homeostatic mechanisms into disarray. Here we show that the neuronal actin cytoskeleton undergoes a rapid and dramatic reorganization in response to ischemia or excitotoxic levels of glutamate --conditions that trigger cytotoxic edema. This neuronal response is reversible and pro-survival. Our data suggest a model in which a strong influx of sodium and chloride, and subsequent water entry (i.e., the key drivers of cell swelling), along with an influx of calcium ion, are necessary and sufficient to induce a fundamental reorganization of the actin cytoskeleton within the somatodendritic compartment of neurons. This convergence of ion influx leads to the activation of the diaphanous family formin INF2. We postulate that INF2 is required either to nucleate new filaments or to elongate existing short filaments within the soma and dendrite. The identification of a formin-based mechanism for actinification is consistent with several lines of evidence, including the long, unbranched filaments observed using both light and electron microscopy, the complete inhibition of actinification by the formin inhibitor SMIFH2, and the complete prevention of actinification by genetic silencing of INF2, which was rescued by ectopic expression of INF2. Interestingly, although we could observe actinification in live cells using multiple fluorescent reporters of actin filaments (Lifeact-mRFP, mApple-Ftractin, and the small molecule probe SiRactin), we were unsuccessful in observing actinification using GFP-actin. This also is consistent with a formin based mechanism, since various reports indicate that large moiety-terminal tags on G-actin interfere with formin-dependent F-actin assembly 48-52 . Importantly, several lines of evidence establish that the actinification phenomenon is distinct from the formation of cofilin-actin rods, which can also form in response to excess glutamate 45 . First, the spatial and temporal features of these events are different, with actinification occurring rapidly within 5 minutes and predominantly in the soma and proximal dendrites, while cofilinactin rods reportedly appear after at least one hour, preferentially in distal dendrites 53 , an observation that we confirmed in our own culture system. Secondly, we found that cofilin immunoreactivity did not colocalize with the actinified filaments as would be expected for cofilinactin rods. Finally, multiple means of preventing or inducing cofilin activation neither prevented or favored actinification. The accumulation of F-actin within the dendrites also did not resemble typical actin stress fibers, since it was not prevented by blebbistatin or by the Rho-kinase inhibitor fasudil (data not shown). We therefore conclude that the actin filaments that accumulate within the somatodendritic compartment following osmotic stress in neurons are distinct, and characterized here for the first time. The precise mechanisms by which swelling and calcium entry converge to activate INF2 require further investigation. Higgs and colleagues have shown that, in non-neuronal cells, an inactive conformation of INF2 is maintained via binding of a complex of lysine acetylated G-actin bound to cyclase-associated protein (CAP) to the INF2 DID and DAD domains, respectively, and that INF2 becomes activated when HDAC6 activity deacetylates G-actin 44 . In our studies of primary neurons, however, none of the various deacetylase inhibitors we tested, including the HDAC6 inhibitor tubastatin, robustly inhibited NMDA-induced actinification. Nevertheless, some mechanism involving acetylation/deacetylation activity does seem to influence actinification driven by INF2, since the compound C646, which broadly inhibits acetyltransferases, by itself caused actinification, even in the presence of NMDA antagonists. This effect of C646 was blocked in the presence of the formin inhibitor SMIFH2, indicating it may act via INF2, similar to NMDA. Taken together, we cannot rule out that NMDA induces INF2 activity via a pathway involving deacetylation, but our results suggest that NMDA possibly activates INF2 through alternative mechanisms. Studies have demonstrated in vitro and in cells that elevated actin monomer concentration can compete with INF2 autoinhibition in addition to its role as nucleation substrate 39, 54 . We postulate that the abrupt rise in G-actin driven by NMDA-induced depolymerization of spine actin generates a burst of soluble monomers that may facilitate the activation of INF2. Indeed, prevention of F-actin disassembly by jasplakinolide completely blocked NMDA-induced actinification ( Fig 5F) . The filaments formed during actinification appear to exhibit an unusually slow turnover. The failure of GFP-tagged forms of exogenous actin to participate in actinification precluded our ability to directly quantify actin turnover rates following NMDA. However, when latrunculin A was applied to sequester G-actin after actinification had occurred, we observed no enhanced clearance of actinified filaments for up to 2 hours, suggesting that there is very little turnover of these filaments in the continued presence of NMDA. We therefore conclude that the half-time for filament turnover probably exceeds ~1 hour, meaning they are highly stable. Despite this remarkable apparent stability, the filaments are able to spontaneously disassemble upon cessation of NMDA receptor activation. Although detailed characterization is required, we estimate that the half-time for filament disassembly (when NMDA antagonists are applied after 5 min of NMDA) is on the order of 15-45 minutes. Interestingly, F-actin also reassembled in dendritic spines during the same time frame for F-actin disassembly in the dendrite shaft, a phenomeon that also deserves further investigation. Initially, we had assumed that the massive reorganization of F-actin represents an early step in excitotoxic neuronal cell death. Indeed, previously we reported that NMDA-induced F-actin reorganization was attenuated by lithium 14 , which has been implicated in neuroprotection 55 . However, as discussed below, we determined that actinification is pro-survival. Given that actin filament assembly is typically an energy-consuming process, it is curious that neurons would engage in large-scale F-actin polymerization during a time of cell stress, especially during hypoxia when ATP is in short supply. The ATP-dependence of actinification remains to be determined; however, polymerization of ADP-actin into filaments has been described 56, 57 . The reversible nature of somatodendritic actinification suggested that it might be beneficial to the cell. Subsequent experiments convincingly demonstrated the pro-survival nature of this pathway (Fig 6B) . Prolonged incubation with NMDA induced a portion of cells to die within 1-4 hours, probably via swelling and necrosis rather than apoptosis, due to its rapidity. We chose lengthy exposures to NMDA for these experiments in order to maximize cell death induction, since most neurons incubated with NMDA for shorter periods do not succumb to stress immediately, but rather over 24-48 hours 58 . Note that even after 1 hour of continuous exposure to NMDA only ~25% of control, empty vector-transfected neurons were dead at this time point, and only ~35% were dead after 4 hours of continuous exposure to NMDA (Fig 6B) . Silencing of INF2 significantly enhanced this rapid neuronal death. Moreover, pharmacological inhibition of formin activity in vivo confirmed that blocking this pathway significantly worsened ischemic infarct severity. Given that at early times (i.e., hours) following a stroke most of the neuronal cell death that generates the infarct is mediated by pathological cell swelling 31, 59 , we postulate that INF2-mediated actinification attenuates the effects of cell swelling and reduces cell death in the early stages after a stroke. Consistent with this hypothesis, we observed the appearance of small cavities within the core of the infarct in our experimental model of stroke, but only when formin activity was inhibited ( Fig 7) . Cavitation within infarct zones has been described previously as cavities which appear many days or weeks after a stroke, evolving from a cystic, fluid filled core 60, 61 . The early appearance of tissue cavities was therefore unexpected, and this observation is worthy of follow up investigation. One possibility consistent with our model is that blockage of formin activity renders neurons so susceptible to edema that they undergo cytolysis, leading to rapid tissue damage and cavitation in the acute ischemic core. The pro-survival function of INF2-mediated actinification might be selective for neuronal edema or related types of osmotic stress. We observed that actinification required a convergence of calcium entry together with an ionic imbalance and water entry. This may imply that actinification selectively protects neurons from cytotoxic edema, but not other stressful conditions. It will be of interest to determine whether actinification is a relevant pro-survival mechanism in widely-occurring injury conditions involving glutamate receptor overload, including traumatic brain injury and seizures in addition to hypoxia/ischemia. Previous studies have also implicated INF2 in mechanosensitive responses of cells in culture, but the in vivo relevance of these responses has not been explored. A study in XTC cells demonstrated mechanosensitive activation and processive F-actin polymerization by diaphenous formins, including INF2 62 . A study in NIH 3T3 cells showed that INF2 mediated the formation of a perinuclear actin rim in response to mechanical stress or calcium ionophore, but did not determine the function of this actin rim structure 41 . Another study similarly reported that INF2 mediates a transient response to cell damage or strong calcium entry, with actin polymerization occurring along the ER, simultaneous with actin depolymerization at the cell periphery, a response the authors termed "calcium-mediated actin reset" 42 . Because the temporal dynamics, sodium and chloride dependence, and other key features of the actin polymerization events described in these prior studies appear to differ from the neuronal actinification we show here, further investigations are needed to examine the relationship among these various cytoskeletal responses. Nevertheless, it is worth postulating that INF2-driven actin polymerization may function in a general pro-survival capacity to protect many types of cells from a variety of mechanical stressors, including swelling during ischemic, hypoxic, or osmotic episodes. The authors declare no competing interests. Detailed methods are provided in the online version of this paper and include the following: Rat hippocampal neurons were isolated according to Calabrese and Halpain 1 The animal's head was affixed to a stable imaging apparatus under the two-photon microscope. The We generated acute, skull-removed cranial windows. Under 4% isoflurane anesthesia, we first injected (2) a study showing that ~10% of similarly-sized molecules at the meninges enters the brain parenchyma through transcranial diffusion 5 . Photothrombosis was initiated 30 min after starting the SMIFH2/vehicle loaded agarose application to allow diffusion through the cortex. Animals were perfused with PFA 4h post-stroke to harvest the brain. Three week old dissociated hippocampal cultures were incubated with either vehicle (H2O) or 50 µM NMDA added directly to their conditioned medium at 37°C for 5 min or other times, as indicated. Neurons were kept in the culturing incubator prior to fixation. This approach was chosen over replacing conditioned media with fresh media containing NMDA due to the established toxicity of this manipulation 6 To induce oxygen and glucose deprivation 2-deoxyglucose was added to the cultures just before placing them into a hypoxic chamber XVIVO system (Biospherix, Parish, NY) at 37°C containing 1%O2. Lipofectamine 2000 was used to transfect 3 weeks old dissociated hippocampal cultures, using various plasmids listed in the above key resources table. Fresh neurobasal media was used to prepare the mixture of Lipofectamine and cDNA. 1 µl of Lipofectamine and 1.5 µg of cDNA per 50µl of NBM were mixed, incubated for 30 min and then added dropwise to 500 µl NBM in which neurons had been growing for 3 weeks. Washing was not required and neurons showed no overt signs of toxicity. Cultures were fixed with 3.7% formaldehyde in phosphate-buffered saline (PBS) plus 120 mM sucrose for For immunohistochemical analysis, animals were perfused transcardially with 4% paraformaldehyde in Sample preparation for platinum replica electron microscopy was performed as described previously 8, 9 . In brief, detergent-extracted samples were sequentially fixed with 2% glutaraldehyde in 0. Unless otherwise indicated, all images shown in this article represent maximum projection images derived from a z-stack. To acquire images of fixed dissociated cultures or brain sections we used an Olympus IX-70 microscope µm for 60X images of dissociated cultures and brain sections. A single plane of focus was used to acquire low magnification images of brain sections using a 1.25X objective. A Zeiss LSM 880 Rear Port Laser Scanning Confocal with Airyscan FAST Microscope with a 63X/1.4 oil objective was used to acquire super-resolution images of actinified neurons (Fig 2A) . Live images were acquired every 30 s before and after NMDA-induced actinification or every 15 minutes when recovery from actinification was monitored with image acquisition times of 0.01-0.2 s using a Nikon Ti-E microscope with perfect focus system (Nikon) and an iXon X3 DU897 EM-CCD camera (Andor VivaFix reagent was added directly to the conditioned media and cultured neurons for 20 min at 37°C, modifying manufacturer's protocol (Biorad). Samples were rinsed once with conditioned media from sister cultures before live imaging or fixing the cells, and co-staining with phalloidin or other markers where indicated. Free-floating brain sections were first incubated with the anti NeuN antibody for 24 h. Then they were mounted onto charged slides, which were dried for 30 min at 50°C, then rinsed for 5 min in distilled water, incubated in 0.06% potassium permanganate solution for 5 min and rinsed again in water. Slides were then transferred for 10 min to a 0.0001% solution of Fluoro-Jade C dissolved in 0.1% acetic acid. DAPI was added at this step. The slides were then rinsed through three changes of distilled water for 1 min per change. Excess water was drained, and slides were then air dried on a slide warmer at 50°C for at least 5 min. The air-dried slides were cleared in xylene for at least 1 min and then coverslipped with DPX non-fluorescent mounting media. In all experiments digital images were acquired using identical parameters and settings (e.g., laser excitation power, acquisition time, time-lapse interval, exposure time, etc.) across experimental conditions. All images displayed in this paper use identical image display settings whenever experimental groups are compared to one another. Sample sizes are provided in all figure legends. For analyses we used Fiji, the open source image software 10 . Adobe Systems Inc. software Photoshop was used for image display, and for analysis of fixed cultures using the binary scoring system described below. To quantify the degree of neuronal F-actin reorganization under various conditions, we devised a convenient assay based on the percentage of neurons that exhibited altered F-actin distribution. We used a binary scoring system that identified individual neurons as being either "actinified" or "nonactinified," as visualized in fixed cultures using fluorescent phalloidin as a label for F-actin. "Actinified" neurons were defined as those that showed robust accumulation of filamentous-appearing phalloidin staining within the soma and proximal dendrites, with little or no phalloidin accumulation in a spine-like punctate pattern; "non-actinified" neurons showed robust accumulation of phalloidin staining in dendritic spine protrusions, and only faint, non-filamentous phalloidin staining in the soma and dendrites. This binary mode of designation was valid because, although different individual neurons became actinified at various times following the onset of a stimulus (e.g., NMDA), once actinification was initiated it proceeded rapidly, and the filaments that accumulated in actinified neurons were stable for long periods. This subjective, binary approach to quantification was applied in an unbiased manner -observers assigned to collect and quantify the images were blind to the experimental manipulation via randomized encoding of the samples. Sample identification was not unveiled until after all data for a given experiment were collected and analyzed. In Fig 3B and C actinification was quantified within the proximal region of the dendrite outlined by Fiji. Briefly, the proximal region of the shaft was cropped after subtracting background. The cropped region of the shaft was then thresholded so that it would be uniformly highlighted. The thresholded area was then binarized and any feature outside the shaft was removed using the erase tool. The plugin Macros Macros was used to outline the binarized thresholded area to generate a ROI (region of interest), which could be saved and added to the original cropped image using ROI Manager, as shown in Fig 3C. This allowed Fiji to measure intensity only within the generated ROI. In Fig 4A, To quantify the effect of the formin inhibitor SMIFH2 versus vehicle on formation of cavitites within the infarct, we calculated the ratio of the area of the cavities (regions lacking neuropil) to the area of ischemic damage, defined as the region with robustly depleted NeuN immunostaining (the "core" of the infarct, as described above). We also quantified the area of vascular leakage as the region staining for mouse immunoglobulin G. The area of tissue damage was measured in adjacent tissue sections and the total infarct volume, Vt, was calculated by Vt = (A 1 + A2 + ...+ An) h, where An was the area of damage in the nth slice, and h was the distance between adjacent sections. All results reported here were observed reproducibly in at least two to three independent culture preparations; similarly, stroke experiments in vivo were repeated across multiple days using mice from multiple litters. Prior to quantitative analysis, sample identity was either encoded for blinding of the experimental group prior to analysis, or image acquisition and analysis were conducted by different people to avoid observer bias. Statistical significance was set at the 95% confidence level (two tailed) and calculated using Prism (Graphpad Software). Values are presented as the mean ± S.E.M. To assess whether the data were normally distributed we used the D-Agostino-Pearson test to determine deviation by skewness or kurtosis. When normality was not met an appropriate non-parametric test was stroke-induced mice, within the corresponding contralateral cortex; stroke-induced mice within temporal cortex ipsilateral to the infarct region; stroke-induced mice, within the ischemic core of the infarct; stroke-induced mice within the penumbral region of the infarct (as defined in Materials and Methods). Data are represented as mean ± SEM; n=3 animals each, sham vs. stroke; *** p<0.001;****p<0.0001, one-way ANOVA, Tukey's multiple comparison post-hoc test. Data are mean ± SEM from two independent culture preparations, with 4 coverslips in total per group; ****p<0.0001; one-way ANOVA, followed by Tukey's multiple comparison post-hoc test. (E) Calcium-induced actinification requires extracellular sodium, but is not inhibited by NMDA receptor antagonists. Intracellular calcium was elevated using the calcium ionophore ionomycin (5 µM, 30 min) in the absence (-) or presence (+) of either extracellular sodium depletion or a cocktail of the NMDA antagonists APV and MK-801. Data are mean ± SEM from two independent culture preparations, with 4 coverslips in total per group; ** p<0.01; one-way ANOVA, followed by Tukey's multiple comparison post-hoc test. Inhibition of NMDA-induced actinification by the specific broad-spectrum formin inhibitor SMIFH2, but not the specific Arp2/3 inhibitor CK666. Data are mean ± SEM from two independent culture preparations, with 4 coverslips in total per group; ****p<0.0001; one-way ANOVA, followed by Tukey's multiple comparison post-hoc test. (B) Representative images illustrating the distribution of endogenous INF2 (green) detected using immunostaining. Cultures were co-stained for NeuN (blue) and MAP2 (red), to identify neurons and their dendritic arbor, respectively. Arrows indicate dendritic regions that lie distal (yellow) and proximal (blue) to the cell soma, and are shown in higher magnification in the lower panels. Note the proximal-to-distal gradient of INF2 immunostaining intensity, with a higher overall signal within the soma and proximal dendrite. those that did take up the VivaFix dye were never observed to be actinified, whether fixed and stained immediately after the indicated NMDA incubation, or fixed after an additional 24 hours when the 1 or 4 hour NMDA incubation was followed by a potential "recovery period" imposed by the addition of a cocktail of APV and MK-801 to prevent ongoing excitotoxic stress. Data are mean ± SEM from two independent culture preparations; **** p<0.0001; two-way ANOVA, followed by Tukey's multiple comparison post-hoc test. ; mice receiving the same drug delivery but no infarct were used as controls for the drug delivery manipulations. Brain section were blinded to treatment group, and stained to evaluate infarct volume (based on NeuN signal loss), the fraction of the ischemic volume subsumed by cavitations (i.e., holes in the brain parenchema), and vascular leakage (based on immunoreactivity for mouse IgG). Data are mean ± SEM. *p<0.05; one-way ANOVA, followed by Dunnett's multiple comparisons posthoc test vs. veh+stroke); number of brains: veh./no stroke = 3, SMIFH2/no stroke = 3, veh + stroke = 7, SMIFH2 + stroke = 8. (E) Left, Representative in vivo two-photon imaging of cortical pial vessels from vehicle-treated (upper) and SMIFH2-treated (lower) mice. Red lines placed across the width of a pial arteriole denote the position of diameter measurements that were averaged to provide a single data point. Scale bars, 50 µm. Right, Scatter plot of pial arteriole diameters from vehicle and SMIFH2-treated groups, including 93 arterioles over 6 vehicletreated mice, and 90 arterioles over 7 SMIFH2-treated mice; mean and SEM are indicated adjacent to each scatter plot; no significant differences were detected between groups, statistical analysis performed by Wilcoxon test. Supplemental Figure S1 . Selected histological markers of ischemic brain injury. Representative images of a, individual section of a mouse brain subject to single vessel photothrombotic stroke, immunostained with some of the indicated markers used in this study to characterize ischemic changes: anti-NeuN (specifically identifies neuronal cell bodies); anti-MAP2 (specifically identifies neuronal dendrites); hypoxyprobe (which uses an antibody to detect the presence of pimonidazole, which accumulates in tissues where pO2 < 10mmHG); and mouse anti-IgG (to assess the extent of hypoxia-induced vascular leakage and/or blood-brain-barrier disruption). Scale bar, 700 µm. Supplemental Figure S2 . Reduction of NeuN immunofluorescence in the ischemic region does not reflect post-stroke neuronal loss. (A) Representative Thy1-GFP mouse brain section stained for NeuN and DAPI, 6h post-stroke, displaying a clear loss of NeuN signal in the ischemic core (indicated by the blue arrow), but relatively little reduction in DAPI staining. Scale bar, 1000 µm. (B) Selected regions of cortical layer 5/6 of somatosensory cortex (from the region indicated by the blue arrow in A) comparing at high magnification brain tissue ipsilateral vs. contralateral to the stroke. In the control, contralateral region (top row) DAPI and NeuN signals are strong and present together tin all neuronal cell bodies. Arrows indicate YFP-positive neurons. In contrast, within the ischemic core (middle row), NeuN staining is lost from individual neurons within the ischemic core, while DAPI staining persists. Thy1-driven soluble GFP is detectable in neurons both within and outside the ischemic core. Dashed yellow line in the bottom row indicates the boundary where NeuN staining is lost in the ischemic core, relative to staining in the surrounding presumptive penumbra, which can be seen at this magnification as a depletion of NeuN immunoreactivity in individual neuronal cell bodies within the core, but not the penumbra. Scale bar, 30 µm Supplemental Figure S3 . Enrichment of F-actin within dendritic spines in cortex of control mouse brain. A Thy1-YFP mouse brain was processed and stained for F-actin as described, using Alexa Fluor 647-phalloidin. Here we focus on a single YFP-positive pyramidal neuron and surrounding neuropil from Layer 5/6 in somatosensory cortex, imaged by confocal microscopy in coronal brain sections using a 60x (N.A. 1.42) oil immersion objective. Images in the left column show phalloidin staining in reverse grayscale; images in the center column show the YFP fluorescence in reverse grayscale; the right column shows a colorized merged image of the two. The top row of images show a single plane from the z-stack of collected images. Note that dense, punctate phalloidin staining is present throughout the neuropil, but the neuronal cell body and proximal apical dendrite are relatively devoid of phalloidin staining. The middle row of images shows the same x-y field of view but as a maximum projection image from the z-stack. Note the marked increase in the number and density of phalloidin puncta, and the fact that many such puncta now overlay portions of the soma and apical dendrite. We interpret this pattern to indicate that phalloidin-positive puncta densely surround the dendrite in three dimensions, but that the interior of the somatodendritic compartment itself is relatively non-enriched for F-actin under control conditions. (We elsewhere show that stroke substantially alters this pattern, as Factin accumulates within the somatodendrite interior; see Figure 1 ). The bottom row shows the boxed region at higher magnification, illustrating our consistent observation that all dendritic spines colocalize with phalloidin-positive puncta; cyan arrows point to specific examples. Given the high density of dendritic spines within mammalian cortical neuropil, we conclude that the majority of phalloidin puncta in the size range of ~0.2-1 µm in diameter that we observe are likely to correspond to dendritic spines, where F-actin is enriched. Scale bar, 4 µm upper & middle rows; 2 µm, bottom row. Supplemental Figure S4 . Dendritic spines decrease in size and number after stroke. Selected dendritic regions proximal to the cell bodies of Thy1-YFP-positive layer 2/3 pyramidal neurons from sham-operated vs. ischemic mouse brain. YFP fluorescence is displayed in reverse grayscale. Scale bar, 5 µm. Supplemental Figure S5 . Overall F-actin signal within the ischemic core decreases over time following stroke. Representative tissue sections from sham or ischemic Thy1-YFP brains fixed at 2h and 6h after single vessel photothrombotic stroke and co-stained for NeuN and F-actin. Cyan arrows indicate the ischemic region. Note that a dramatic decrease in NeuN immunoreactivity is already observed by 2h after stroke, accompanied by a very modest decrease in phalloidin staining. By 6h post-stroke we observe that, as expected, the average infarct volume (as measured using NeuN or other markers) increases, and it is accompanied by a consistent decrease in overall phalloidin staining within the infarct zone. Nevertheless, within the infarct zone (core and penumbra) we observe numerous neuronal cell bodies with accumulations of somatodendritic F-actin that are suggestive of actinification (see Fig 1) . Scale bar, 700 µm. Supplemental Figure S7 . Somatodendritic actinification is a novel process distinct from cofilinactin rod formation. (A) Left, representative black and white images of control and NMDAtreated neurons co-stained for F-actin, and cofilin. Endogenous cofilin shows the expected punctate distribution in control neurons (cyan arrows), and little detectable immunoreactivity in NMDA-treated actinified neurons (blue arrows), while remaining detectable and punctate in a neighboring astrocytes (orange arrows). Scale bar, 20 µm. Right, selected dendritic regions from a control neuron and an actinified NMDA-treated neuron. Scale bar, 4 µm. (B) Cofilin rod formation does not occur in parallel with actinification. Cofilin rods were not detected after incubation for 5 min with NMDA, and only rarely after 30 min of NMDA, even in the actinified neurons; in contrast, we reliably detect cofilin rods after incubation with 200 µM glutamate for 1 hr or longer. Data are mean ± SEM from three independent culture preparations; **** p< 0.0001; two-way ANOVA, followed by Sidak's multiple comparisons post-hoc test, to selectively compare neurons with cofilin rods and actinified neurons within each experimental group. (C) Two representative images demonstrating the lack of colocalization between the actinified somatodendritic regions (endogenous actin stain detected using an anti-actin antibody (magenta); endogenous cofilin rods, detected using an anti-cofilin antibody (green). Scale bar, 10 µm (left image); 18 µm (right image). (D) Ectopic co-expression of cofilin and chronophin, one of its Ser-3 phosphatases, induces cofilin rods without inducing actinification. Representative neuron transfected with cofilin-GFP (Cof) and chronophin-HA (CIN) displaying the distribution of cofilin rods in distal regions of thin secondary dendrites (arrows). Scale bar, 24 µm. Representative cortical regions from the hemipshere contralateral to the infarct (upper half), and along the border (yellow line) between the infarct core and the penumbra (lower half), in the absence (left) or presence (right) of formin inhibitor SMIFH2 delivered by passive diffusion from a saturated agar plug placed over the thinned skull. Each color-combined image is juxtaposed to the correspondent individual grayscale channels (which are rotated 90 deg relative to the merged image). Nuclei are labeled with DAPI (in blue), NeuN (in red), and the cell death marker FluoroJade-C (FJ-C; green). Scale bar, 60 µm. Cerebral tissue oxygenation and oxidative brain injury during ischemia and reperfusion Neurobiology of COVID-19 Effects of COVID-19 on the Nervous System COVID-19: A Global Threat to the Nervous System The emerging association between COVID-19 and acute stroke Pathophysiology and therapy of experimental stroke Pathogenic mechanisms following ischemic stroke The science of stroke: mechanisms in search of treatments Molecular Communication of a Dying Neuron in Stroke The cellular mechanisms of neuronal swelling underlying cytotoxic edema Spreading depolarization is not an epiphenomenon but the principal mechanism of the cytotoxic edema in various gray matter structures of the brain during stroke The smallest stroke: occlusion of one penetrating vessel leads to infarction and a cognitive deficit Targeting NMDA receptors in stroke: new hope in neuroprotection Lithium prevents aberrant NMDA-induced F-actin reorganization in neurons Rapid morphologic plasticity of peri-infarct dendritic spines after focal ischemic stroke Regulation of F-actin stability in dendritic spines by glutamate receptors and calcineurin Modeling Ischemic Stroke In Vitro: Status Quo and Future Perspectives Glucose, glycolysis, and neurodegenerative diseases Microvascular basis for growth of small infarcts following occlusion of single penetrating arterioles in mouse cortex Acute focal ischemia-induced alterations in MAP2 immunostaining: description of temporal changes and utilization as a marker for volumetric assessment of acute brain injury Essential role for the PKC target MARCKS in maintaining dendritic spine morphology Imaging neuronal subsets in transgenic mice expressing multiple spectral variants of GFP INF2 is an endoplasmic reticulumassociated formin protein Two-photon microscopy as a tool to study blood flow and neurovascular coupling in the rodent brain Transcranial amelioration of inflammation and cell death after brain injury Excitotoxicity triggered by Neurobasal culture medium A genetically encoded reporter for real-time imaging of cofilin-actin rods in living neurons Electron microscopic analysis of the leading edge in migrating cells Imaging cytoskeleton components by electron microscopy Fiji: an open-source platform for biological-image analysis We are grateful to the laboratory of Mana Parast (UC San Diego) for use of their hypoxic chamber, and Christian A. Olsen for the generous gift of specific SRT5 inhibitors. We thank Vaidehi Gupta for generating dissociated neuronal cultures and other support, and Agnieszka Brzozowska-Prechtl for brain sectioning. We thank David Kleinfeld (UC San Diego) and Halpain lab members for helpful discussions. We greatly appreciate the following peerless undergraduate lab volunteers for conducting blinded data acquisition and analysis and general lab support: Kyra Rashid, Jeremy Aung, Liam Huber, Shyam Patel, Youjia Guo, Molly Thapar, Huanqiu Zhang. We thank the UCSD Nikon Imaging Center for access to the NIS Elements analysis software. This research was supported by grants from the U.S. National Institutes of Cultured hippocampal neurons were incubated for 1h with HDAC inhibitors (HDI) before incubating the cells for 5 min with 50µM NMDA. Concentrations below and above the ones listed were also tested to reach either no or toxic effect. n.d.q. = no inhibition detected in qualitative analyses; n = number of independently repeated experiments.