Thirty years later: Revisiting the glial failure hypothesis as the homeostatic organizing principle of ischemic injury Authors: Oscar Herreras, Santiago Canals Published: 2026 ABSTRACT Thirty years after the glial failure hypothesis was proposed, ischemic neuronal death still lacks a unifying mechanism. Here we revisit the hypothesis in light of evidence positioning astrocytes as central regulators of metabolic, ionic, and vascular homeostasis, unifying mechanisms often considered independently in ischemic injury. Early astrocytic responses, including adenosine-mediated synaptic suppression, may transiently protect tissue by reducing energetic demand. Progressive glial dysfunction, however, drives spontaneous spreading depolarizations and impairs their termination, creating a bifurcation toward immediate or delayed neuronal terminal depolarization depending on energetic capacity. Thus, glial failure emerges as the upstream disturbance linking metabolic compromise to neuronal injury. Introduction Thirty years after the formulation of the glial failure hypothesis, 1 the mechanisms governing ischemic neuronal death remain incompletely understood. Multiple mechanistic frameworks have been developed over this period, yet a unifying explanation of tissue failure is still lacking. 2 Astrocytes have often been viewed as parallel victims of ischemia rather than primary determinants of neuronal fate, 3 with their relative resistance attributed to metabolic resilience. Accordingly, neuronal vulnerability has largely been interpreted within a neuron-centric perspective of cell-autonomous molecular and metabolic determinants.4,5 When originally proposed, the glial failure hypothesis challenged this view by suggesting that progressive astrocytic metabolic dysfunction precedes and drives neuronal injury. 1 Since then, astrocytes have emerged as central regulators of ionic, metabolic, synaptic, and vascular homeostasis, integrating neuronal and vascular function into a dynamic homeostatic network.6–10 This evolution provides a compelling rationale to revisit whether failure of this integrated astrocytic system represents the common upstream event from which multiple downstream mechanisms arise. Although termed the “glial failure hypothesis,” its mechanistic focus is primarily on astrocytes, whose homeostatic functions provide the principal substrate for the proposed transition from metabolic stress to neuronal injury. Spreading depolarizations (SDs) have also undergone a major conceptual shift.11,12 Once regarded largely as epiphenomena of metabolic failure, they are now recognized as major determinants of lesion progression in ischemic stroke and other acute brain injuries.13–16 Thus, while their spontaneous emergence remains poorly understood, the factors governing their transition from reversible electrophysiological events to neuron-lethal depolarizations are increasingly linked to the duration of the depolarized state and failure of recovery mechanisms. In the glial dysfunction hypothesis, SDs are not merely downstream epiphenomena of metabolic collapse. Rather, they are emergent manifestations of a failing glial–neuronal metabolic unit and represent a critical bifurcation point in tissue fate. When astrocyte-dependent homeostatic mechanisms remain competent, SD can be terminated and ionic gradients restored. When glial metabolic support fails, however, SD becomes prolonged or recurrent, increasing the energetic burden of recovery and pushing the tissue toward terminal depolarization. More broadly, glial metabolic failure may constitute a convergent upstream state that biases whether excitotoxicity, ionic imbalance, energy failure, oxidative stress, inflammation, and downstream cell-death pathways remain transient stress responses or evolve into self-amplifying processes of irreversible ischemic injury. In this view, neuronal fate is ultimately determined at the level of astrocyte-dependent homeostatic control. The original formulation of the glial failure hypothesis The hypothesis did not emerge in isolation. During the 1980s and early 1990s, studies in cultured astrocytes showed that single perturbations relevant to ischemia could trigger others, including acidosis, glutamate accumulation, ionic imbalance, production of free radicals, and cell swelling, revealing a tightly coupled system of interdependent variables.17,18 While these findings suggested glia at the crossroads of the ischemic cascade, they did not support a central causal role for glia in vivo. Yet, they provided a mechanistic rationale for experimentally probing whether selective glial impairment could reproduce key features of ischemic dysfunction in intact tissue, including the conditions under which SDs arise. Evidence from in vivo models Building on these observations, preferential impairment of glial oxidative metabolism while preserving neuronal metabolism in situ revealed a highly stereotyped sequence of events.1,19 Excitatory synaptic transmission failed early, whereas neuronal membrane excitability remained preserved for several hours despite severe glial depolarization, swelling, and metabolic impairment. During this early phase, extracellular amino-acid dynamics indicated disruption of the glutamate–glutamine cycle and a shift toward glial anaerobic metabolism, accompanied by significant extracellular acidification. The onset of recurrent SDs marked a clear transition to a lethal phase. Each event was associated with a stepwise loss of population electrical viability, accompanied by increases in extracellular potassium and GABA, and progressive acidification, consistent with massive release of intracellular contents following loss of membrane integrity. Histological analysis revealed a striking dissociation between neuronal and glial fate: neurons exhibited widespread irreversible injury, with shrunken and hyperchromatic morphology consistent with terminal depolarization and collapse of ionic and metabolic homeostasis. In contrast, astrocytes and pericytes remained structurally preserved despite marked swelling, consistent with profound functional impairment in the absence of overt cell death. Subsequent studies in more controlled slice preparations reproduced the core sequence of early synaptic failure and glial swelling preceding SD generation.20,21 In slices that did not develop SD, neuronal electrical viability persisted for several hours and declined gradually, indicating that glial metabolic impairment alone ultimately leads to neuronal death, albeit through a slower process that permits the progressive engagement of additional injury mechanisms. In contrast, when SDs occurred, one or two events were sufficient to abolish virtually all neuronal electrical activity, dramatically shortening the time to irreversible failure. Importantly, these studies further showed that the earliest synaptic failure was not simply a passive exhaustion of transmitter stores, as initially proposed, but an active adenosine A1-receptor-mediated suppression of synaptic transmission triggered by glial metabolic stress. 21 Blocking A1 receptors restored synaptic transmission, but at a cost: SDs appeared earlier, occurred more frequently, and produced a more severe loss of neuronal electrical viability. These findings place astrocytes at the center of an adaptive trade-off during energy imbalance: by releasing adenosine, they suppress synaptic activity, reduce energetic demand, and regulate blood flow to match with metabolic activity, thereby preserving ionic homeostasis and delaying lethal depolarization.22–24 If this protective mechanism is removed or overwhelmed, persistent synaptic signaling accelerates the transition from reversible dysfunction to SD-associated neuronal death. These observations demonstrate that progressive glial metabolic impairment alone is sufficient to generate spontaneous SDs in the absence of any additional triggering stimulus. Both the original experiments and subsequent independent studies increasingly identified astrocytes as major determinants of SD susceptibility and outcome, and have progressively shifted the view of astrocytes from passive participants to active determinants of both SD dynamics and its pathological consequences.25–28 A single-neuron perspective of the glial-mediated ischemic sequence Slice preparations enabled continuous intracellular recordings during selective glial metabolic impairment, providing a cellular-scale view of the transition from preserved neuronal viability to terminal depolarization and death (Figure 1). Single-neuron perspective of the glia-mediated ischemic sequence. Continuous electrophysiological recording from a CA1 pyramidal neuron during glial poisoning in a hippocampal slice. Neuronal viability was operationally assessed by the amplitude of the aPS. Upper and lower continuous traces show simultaneous extracellular and intracellular DC recordings, respectively. Insets show evoked responses (population: upper row and unitary: middle row) and I–V response series (lower row). Minutes after application of the glial toxin fluoroacetate to the bath (triangle #1), orthodromic synaptic transmission was rapidly and completely abolished (triangle #2), coinciding with neuronal hyperpolarization. Antidromic unitary and population spike responses (a1), and I–V responses were preserved, indicating maintained electrical viability. After a spontaneous SD swept across the recorded neuron site (triangle #3), the neuron fully recovered its membrane potential (Vi), as well as the antidromic and I–V responses, indicating preserved viability, although recovery of the population was only partial (a2: asterisk). Minutes later, Vi depolarized, initially slowly and then more rapidly. For a minutes-long period, the neuron maintained electrical viability while the aPS progressively declined (a3) until complete blockade, finally the entire population becoming electrically silent (a4) during the rapid phase of Vi depolarization. Injection of negative current (−0.3 nA) failed to preserve the neuron’s electrical viability. The extracellular slow negative potential shift was not followed by the terminal DC depolarization shift, which is absent in regions where all neurons have already lost electrogenesis. 17 Parts of this experiment were previously published in Canals et al. The state transitions during metabolic glial poisoning shown at the bottom illustrate the bifurcation point determined by the occurrence of an SD. aPS: antidromic population spike. The recording captures a neuron that remains electrically viable despite progressive glial metabolic dysfunction. Excitatory synaptic transmission is rapidly blocked, whereas membrane excitability is preserved, as evidenced by antidromic spikes. A first SD-wave invades the tissue and is followed by a fast recovery of the recorded neuron, while the antidromic population spike reveals a ~50% reduction in neuronal population viability. Only later does the recorded neuron undergo a minutes-long irreversible depolarization that culminates in terminal membrane failure, coinciding with an accelerated loss of viability of the neuronal fraction that survived the SD. This delayed collapse suggests that although SD does not immediately eliminate all neuronal viability, it aggravated destabilization of the remaining metabolic support system. What makes this recording particularly compelling is that it provides a rare real-time observation of the functional transition from neuronal viability to irreversible electrical failure. Unlike histological endpoints, which identify cellular damage retrospectively, the loss of membrane potential defines the moment at which neuronal viability is irreversibly lost. This single-cell perspective condenses the sequence of events originally inferred from population electrophysiology, extracellular ion dynamics, amino-acid measurements, and histological analyses. Why revisit the glial failure hypothesis now? The original glial failure hypothesis was proposed at a time when astrocytes were viewed largely as supportive cells with limited influence on neuronal survival. Over the past three decades, however, this perspective has changed profoundly. Astrocytes are now recognized as active regulators of extracellular ion homeostasis, glutamate recycling, metabolic substrate allocation, neuromodulation, and neurovascular coupling, integrating neuronal and vascular function into a dynamic homeostatic network.6,7,9,10,29 Rather than independent homeostatic processes, these functions are increasingly understood as coordinated components of tissue resilience during metabolic stress. More recently, astrocytes have also been shown to sense changes in oxygen and glucose availability and to initiate rapid metabolic, vascular, and neuromodulatory responses to energetic stress.30–33 This emerging role as metabolic sentinels provides a framework for the early adaptive responses observed during selective glial metabolic impairment. During metabolic stress, astrocytes release adenosine that rapidly suppress excitatory synaptic transmission through A1 receptor activation while preserving neuronal membrane excitability.21,34 Preventing this adaptive response restores synaptic activity but accelerates SD onset, increases SD frequency, and worsens neuronal electrical failure. Thus, astrocytic adenosine signaling illustrates an essential principle of the glial failure hypothesis: before neuronal function collapses, astrocytes actively reduce network activity to preserve energy and maintain ionic homeostasis. The transition to irreversible injury therefore reflects failure of an initially protective homeostatic program rather than the immediate consequence of neuronal metabolic exhaustion. The critical issue is therefore not whether astrocytes are protective or deleterious, but when their protective responses remain sufficient and when their failure becomes a driver of tissue collapse. A second major development concerns spreading depolarizations (SDs). Initially regarded largely as transient electrophysiological phenomena, SDs are now recognized as major determinants of lesion progression in ischemic stroke and other acute brain injuries.14,35 In this view, SD is not an independent pathogenic process but the emergent electrophysiological expression of failure within the glial–neuronal homeostatic unit. A bifurcation model: Metabolic failure and SD lethality SD represents the bifurcation point in this model. In healthy tissue, SD is usually reversible because neurons and astrocytes can restore ionic gradients and metabolic balance after the depolarizing wave has passed.1,13,16 Yet, although initially considered innocuous, it was later shown to cause a profound, hours-lasting disruption of dendritic electrogenesis, suggesting persistent structural and/or functional alterations of the dendritic compartment. 36 In metabolically compromised tissue, however, SD occurs spontaneously and becomes prolonged or recurrent.1,19 As the SD-associated depolarization persists, the ATP cost of repolarization rises, mitochondrial stress accumulates, and the probability of recovery declines. If astrocytes can no longer sustain K+ clearance, substrate allocation, and acid–base regulation, SD ceases to be a transient electrophysiological event and becomes the immediate cause of irreversible injury. This framework addresses two unresolved questions. First, SDs may not require a specific triggering molecule in compromised tissue, but may emerge as an intrinsic consequence of progressive astrocyte-dependent homeostatic failure. 1 This possibility is consistent with observations that spontaneous SDs can develop following blood brain barrier opening during systemic hyperkalemia without detectable increases in interstitial K+. 37 Second, SDs become neuron-lethal because the same glial failure that promotes their initiation also impairs their termination. Selective glial dysfunction prolongs SD duration several-fold, 19 and prolonged SDs are commonly observed in the ischemic penumbra and in conditions associated with glial dysfunction.38,39 This view also reframes the ischemic core and penumbra model. At the time the glial failure hypothesis was formulated, the distinction between ischemic core and penumbra was rooted in electrophysiological observations, with critical perfusion thresholds derived from the occurrence of terminal versus nonterminal spreading depolarizations and the loss of evoked and spontaneous electrical activity.40,41 Yet the functional state of astrocytes within these regions remained poorly defined. The penumbra was generally viewed as a region threatened by pathological processes originating in the ischemic core, including SDs and other diffusible injurious factors. Rather than representing sharply demarcated regions, however, the ischemic core and penumbra may be better understood as points along a continuum of metabolic and electrophysiological dysfunction, with sustained depolarization and continuous energy expenditure predominating at the core, while progressively less severe depolarization and intermittent SDs characterize the transition toward the penumbra. 42 In this dynamic continuum, recurrent SDs may progressively extend and exacerbate ischemia, depolarization, and energy consumption, in part through SD-induced hypoperfusion (“spreading ischemia”), rather than simply marking the boundary between regions of distinct vulnerability. In this view, SDs would not initiate tissue failure de novo, but rather accelerate the collapse of an already unstable metabolic condition. More broadly, the present framework proposes that reduced perfusion may itself be sufficient, from the onset of ischemia, to compromise astrocyte-dependent homeostatic control while initially preserving neuronal electrical viability. This interpretation is consistent with studies showing that hypoxia or oxygen–glucose deprivation profoundly impair astrocyte-dependent homeostatic mechanisms and increase susceptibility to prolonged SD.43,44 There is also a protective side to this bifurcation. Under energy limitation, astrocytic adenosine signaling suppresses synaptic activity, reducing energetic demand and delaying entry into SD. This protective response may occur across the ischemic continuum, including in the developing ischemic core, where adenosine-mediated hyperpolarization may initially limit neuronal energy expenditure but is ultimately overwhelmed by sustained depolarization and severe energy failure. 45 In this way, the tissue sacrifices synaptic communication to preserve the ATP required for ionic homeostasis and neuronal survival. This interpretation is consistent with profound synaptic depression in electrically viable penumbral neurons, 46 as well as with evidence that selective glial metabolic stress triggers an adenosine-dependent suppression of synaptic transmission. 21 Viewed in this light, early synaptic silencing is not merely a secondary consequence of tissue failure, but may represent an intrinsic astrocytic energy-conservation response to local metabolic stress. As metabolic stress intensifies, however, this protective mechanism may become insufficient to prevent the transition from reversible SD to sustained depolarization and irreversible injury, shifting the tissue toward the lethal branch of the bifurcation (Figure 1). Recovery from SD ultimately depends on ATP-driven restoration of ionic gradients. Under glial dysfunction, prolonged depolarization may exceed the mitochondrial capacity of neurons to sustain ATP production. Consistent with this, SD has been associated with mitochondrial fragmentation, raising the possibility that impaired mitochondrial bioenergetics progressively limits recovery, manifested as persistent loss of dendritic electrogenesis in intact tissue and impaired restoration of SD-induced ionic gradients under penumbra-like conditions.47,48 What would falsify the glial failure hypothesis today? The glial failure hypothesis generates clear, experimentally testable predictions and would be challenged if neuronal death occurred despite preserved astrocytic metabolic and homeostatic function or if ischemic lesion expansion proved independent of glial metabolic integrity. Likewise, the hypothesis would be weakened if the initiation, propagation, or recovery of SDs were found to be insensitive to astrocytic energetic status, or if neurons consistently recovered from SD despite sustained energetic limitation and without astrocyte-dependent metabolic support. Direct experimental testing of these predictions will be essential to establish whether glial failure represents the upstream event that determines neuronal vulnerability or a secondary response to tissue injury. Glial failure as a common upstream event Within this framework, ischemic injury reflects a single upstream disturbance in astrocyte-dependent homeostatic control. Table 1 summarizes traditional and revised interpretations of glial involvement across the major pathways leading to neuronal death. Although therapies targeting downstream mechanisms have often shown robust neuroprotection in experimental models, none has translated into effective clinical therapy for ischemic stroke. Beyond methodological limitations, this failure may reflect a deeper conceptual issue: therapeutic strategies have largely targeted consequences rather than the upstream failure of astrocyte-dependent homeostatic control. Thirty years after its formulation, the glial failure hypothesis therefore offers not only a reinterpretation of ischemic pathophysiology but also a conceptual basis for therapeutic discovery. Rather than viewing excitotoxicity, oxidative stress, inflammation, or SDs as independent therapeutic targets, these processes may instead represent downstream consequences of a common homeostatic failure. Preserving or restoring astrocytic homeostatic competence before irreversible network collapse may therefore provide a more effective direction for future neuroprotective strategies. Table 1.Summary of traditional and revised interpretations of glial involvement across the major pathways leading to neuronal death.Observed phenomenonTraditional interpretationGlial failure interpretationFunctional consequence for network viabilityExcitotoxicityExcess glutamate release and receptor overactivationFailure of astrocytic glutamate clearance and metabolic recyclingProgressive destabilization of synaptic homeostasisIonic imbalance (↑ K+o)Neuronal depolarization and channel dysfunctionFailure of astrocytic K+ buffering and spatial potassium redistributionLoss of membrane potential stability and excitability controlMetabolic acidosisAnaerobic glycolysis and lactate accumulationCollapse of astrocyte-supported metabolic coupling and acid–base regulationImpaired ionic gradients and enzymatic dysfunctionSpreading depolarizationsPathological mass depolarization eventsFailure of astrocytic termination of network depolarization statesTransition from reversible activity to sustained energetic crisisNeuroinflammationMicroglial activation and immune responseSecondary response to glial network dysfunction and tissue distress signalsAmplification of secondary injury cascadesSynaptic failureNeurotransmitter exhaustionAdenosine A1-receptor-mediated suppression of transmission triggered by glial metabolic stressEnergy preservation delays SD initiation and prolongs cellular viabilityBlood–brain barrier breakdownEndothelial dysfunction and inflammatory injuryFailure of multicellular neurovascular unit homeostasisLoss of compartmentalization and edema progression Summary of traditional and revised interpretations of glial involvement across the major pathways leading to neuronal death.