Mechanistic insight into the initiation of spreading depolarizations: Is it really all about potassium?

Abstract

Cortical tissue, when subjected to certain electrical, mechanical, thermal or chemical disturbances, can undergo a widescale disruption of neuronal ionic electrochemical gradients, termed spreading depolarization (SD). SDs propagate slowly across the brain (∼2–9 mm min−1), coinciding with the depression of spontaneous and evoked electrophysiological activity; hence, the near synonymous term spreading depression (Somjen, 2004). It was this suppression of activity, discerned in rabbit models of experimental epilepsy, that first alerted Aristides Leão to the phenomenon in (Leao, 1944). Since then, various technological developments have provided new ways to identify and characterize SDs. Intracellular recordings show the near complete breakdown in ionic concentration gradients during SD, causing a sustained shift in neuronal intracellular potential to near 0 mV (Somjen, 2004). The disturbance propagates outward from the point of stress in a regenerative, all-or-nothing wave, independent of the stimulus. The neuronal depolarization is accompanied by glial depolarization as extracellular potassium, [K+]o, increases and, importantly, normally reverses after a few seconds to minutes. Transient alterations in blood flow and metabolic rate, cellular swelling, and the release of most neurotransmitters and neuromodulators within the depolarized tissue occur concomitantly. The cellular swelling causes subtle changes in reflectance and light scattering, allowing SDs to be easily visualized using intrinsic optical imaging (Somjen, 2004). Clinically, the recent introduction of wide-band recording facilities is helping record SDs in humans, whereas previously, most EEG machines used high-pass filtering to prevent drift in the recordings from overloading the amplifier, excluding SD events identified from DC shifts and low frequency components of the local field potential. The recognition that SDs are a prominent feature in many human pathologies has greatly increased interest in understanding these events. SDs have now been implicated in multiple different clinical neurological conditions, including migraine with aura, traumatic brain injury, intracranial and subarachnoid haemorrhage, ischaemic stroke, seizure and sudden unexpected death in epilepsy (SUDEP). Identifying the distinct molecular and electrophysiological mechanisms underlying SD initiation is essential to understanding their role in neuronal health and disease and, more importantly, what therapeutic approaches may be taken to reduce brain injury resulting from these various neuropathologies. In this review, we discuss how our understanding of SD initiation has been shaped by the choice of experimental models, and how new models of SD induction invite us to reexamine the underlying mechanism. During an SD, K+ is extruded from the cells, increasing [K+]o significantly, sometimes reaching 30–60 mm. It was realized quite early that direct application of extracellular K+, with a threshold concentration of ∼12–15 mm, is a reliable way to trigger SDs, making this a popular experimental method. The fact that increased [K+]o both triggers and is a consequence of SDs provides one potential explanation as to why SDs propagate as regenerative waves (Andrew et al., 2022). During SDs, Na+, Ca2+ and Cl− are also redistributed into neurons from the extracellular space: [Na+]o decreases from ∼150 mm to 50–70 mm and [Ca2+]o decreases from 1.2 mm to 0.1–0.2 mm. The neuronal depolarization at the SD wavefront is accompanied by transient alkalinization followed by sustained acidification that presumably reflects the increased metabolic demands of recovery. These ionic shifts occur in tandem with various cell morphology changes, including cell swelling, dendritic beading and an apparent transient loss of dendritic spines alongside significant shrinkage of extracellular space, which collectively suggest that water is moving into neurons. The precise causal link between the osmotic changes and ionic redistribution eludes us still, yet, we argue, is pivotal for understanding SDs. It is noteworthy that changes in light scattering appear to precede the increase in [K+]o in SDs that follow a hypoxic insult, suggestive of swelling prior to SD induction (Somjen, 2004). SDs may also be induced by sustained depolarization of neurons; for example, through synaptic drive or by pan-neuronal optogenetic stimulation (Masvidal-Codina et al., 2021). The most important case, in that it happens naturally and commonly, is when SDs follow seizures; we will return to the relationship between seizures and SDs later in this review. Significant glutamate release is evident in SD both in vivo and in vitro and is suggested to contribute to the positive feedback cycle involving increased [K+]o. Consistent with this idea, antagonists of NMDA receptors (but not antagonists of AMPA or kainate receptors) decrease the rate of SD onset, occasionally inhibiting SDs entirely (Charles & Brennan, 2009); SDs, however, are still readily induced in a Ca2+-free solution, which blocks conventional synaptic glutamate release (Somjen, 2004). One final pertinent observation is that holding neurons in a persistently depolarized state using voltage clamp does not appear to trigger the cellular responses seen during SDs. In summary, although depolarization is a key feature of SDs, it does not appear to be the means of inducing them. The involvement of inhibitory synaptic drive in SDs is less clear cut. GABAA receptors normally facilitate chloride influx, which limits the SD propagation rate (Aiba & Shuttleworth, 2014), but can also contribute to dendritic beading (Steffensen et al., 2015). Although removal of Cl− can reduce dendritic beading, it does not prevent cell swelling because other anionic compounds, such as bicarbonate, probably accompany Na+ influx when Cl− is absent (Somjen, 2004). Various pharmacological agents that interfere with chloride levels, including furosemide, which inhibits both the NKCC1 and KCC2 cotransporters, or the voltage-dependent anion channel blockers, DIDS or DNDS, had little impact on the magnitude of the depolarizing shift and onset time for SD (Somjen, 2004; Steffensen et al., 2015). Adding to this substantial body of literature focusing upon increased [K+]o or depolarization, we recently discovered a different way of inducing SDs that prompts us to rethink the basic model of how SDs occur. This paradigm-shifting observation was a serendipitous finding in a study conceived initially to determine the pattern of ionic redistribution during activation of the optogenetic chloride pump halorhodopsin, which was expressed exclusively in the excitatory neurons (Parrish et al., 2023b). Using a combination of patch clamp recording, extracellular field potential and [K+]o measures, we showed that the primary, light-driven movement of Cl− into neurons causes a secondary K+ inward movement through the open leak channels (which are predominantly K+ channels) as hyperpolarization alters the potassium electrochemical gradient (note that hyperpolarization affects the electrochemical gradients for all ions, but K+ shifts dominate because the main conductance at rest is gK). Consequently, during the actual period of illumination, [K+]o decreased from a baseline of 3.5 mm to ∼2.5 mm. After the end of illumination, there was a large rebound increase in [K+]o, and, not surprisingly, SDs often ensued at this point. The real surprise, however, was that SDs also arose, on occasions, during the period of illumination when halorhodopsin was active, most neurons were hyperpolarized and [K+]o was lower than baseline: this combination of conditions is seemingly the very opposite of conventional experimental SD induction protocols. Another optogenetic pump, archaerhodopsin, hyperpolarizes cells by pumping protons out rather than by pumping Cl− ions in as halorhodopsin does. We found that archaerhodopsin activation also caused [K+]o to decrease during illumination and to increase after illumination ended, although to a much lesser extent than halorhodopsin as Cl− clearance is coupled to K+ extrusion. Archaerhodopsin, however, did not elicit SDs in any of our recordings (Parrish et al., 2023a). This novel optogenetic SD induction method, by showing that SDs can be triggered in low [K+]o just as well as by increased [K+]o, invites us to consider how the various induction models differ and what they share in common. The halorhodopsin model directs our attention away from depolarization and high [K+]o as the causative mechanisms of SDs. Instead, we propose a different explanation of SD initiation, prompted by the realization that the various triggers of SDs all involve osmotic stress. Both halorhodopsin activation and KCl application lead to a large inward flux of both Cl− and K+ ions, as previously described. The intra-to-extracellular distributions of these two ions are tightly coupled by the cation-chloride cotransporter KCC2 expressed strongly in most adult neurons. Although close to equilibrium under normal physiological conditions, the K+ electrochemical gradient is fractionally larger than the oppositely oriented Cl− gradient. Consequently, KCC2 normally supports outward movement of KCl, but, when KCl is added to the extracellular media, the point of equilibrium shifts, KCC2 reverses direction and KCl enters the cell. Experimental demonstrations of KCC2 reversing its direction have used extracellular application of rubidium or thallium ions (Zhang et al., 2010), both of which can substitute for K+ ions. Thus, both models are characterized by ionic loading of neurons. Intriguingly, shifting osmotic balance is a pattern replicated in other models of SD induction (Fig. 1). The absence of aquaporins in neurons (Andrew et al., 2007), in contrast to glia which do express aquaporins, has been interpreted to mean that neurons are impervious to osmotic challenges. In fact, this arrangement simply means that any osmotic challenge is absorbed primarily by the glial population; phospholipid membranes are not waterproof (Pohl et al., 1997). Ergo, even without aquaporins, neurons are not exempt from osmotic stress and may still swell or shrink under osmotic pressure (Glykys et al., 2019; Murphy et al., 2017). Tellingly, cultured neurons in isolation readily swell when subjected to hypoosmotic medium (Inoue et al., 2005). We suggest that, in the intact network, glia serve as a first line of defence against osmotic stress, buffering excess water, consistent with their high expression levels of aquaporins. The halorhodopsin model of SD induction circumvents this protective glial function by directly loading ions into neurons. Overloading neurons with a different ion, Na+, may be the mechanism by which veratridine, a drug that prevents Na+ channel inactivation, triggers SDs in brain slices. Interestingly, tetrodotoxin, which blocks voltage-gated Na+ channels, restricted the propagation of SDs, although it did not stop them from occurring (Suryavanshi et al., 2022). More physiologically, sustained neuronal activity is associated with a net inward movement of ions, which, during seizures, is accompanied by water movement leading to measurable increase in the total intracellular volume and shrinkage of the extracellular space (Glykys et al., 2019; Somjen, 2004). As previously mentioned, SDs often occur as seizures terminate, possibly playing a causal role (Norby et al., 2025). Finally, perhaps the most important natural SD triggers involve reduced activity of the Na+/K+ ATPase pump arising from hypoxia or ischaemia, mimicked in oxygen-glucose deprivation models or by blocking the pump using ouabain. The Na+/K+ ATPase pump removes 3 Na+ ions at the same time as taking just 2 K+ ions in, thus constituting a net outward movement of osmotic particles. Without this action, there is a net build up of intracellular ions, causing cells to swell (note that the Gibbs–Donnan ‘equilibrium’ is not a stable steady-state). Simply put, metabolic stress generates an environment conducive to SD, and all these manipulations can cause neurons to swell, indicative of changes in osmotic balance leading to water entry (Fig. 1). Having recognized that increased intracellular osmotic tension appears common to many of these SD triggers, how does this help our understanding? The input resistance of neurons plummets during SDs, indicating that the phenomenon involves opening high conductance ion channels, although there remains uncertainty about the exact nature of the pores. In non-neuronal cells, swelling leads to the opening of volume-regulated anion channels (VRACs, also termed volume-sensitive organic anion channels, VSOACs), very large channels allowing Cl−, bicarbonate and small organic anions to move freely. Their role in SDs is under-investigated; one study showed that the VRAC antagonist NPPB inhibits the release of glutamate and delays SD onset (Basarsky et al., 1999), whereas another study reported that SD ampliltude was unaffected by 300 µm DIDS, which should block VRACs (Steffensen et al., 2015). The picture remains incomplete because VRACs are not reported to conduct K+ and presumably would also facilitate further Cl− entry, as indicated by measurements of extracellular [Cl−] showing a decrease during SDs (Somjen, 2004). So what is the mechanism by which the osmotic crisis is dissipated? A major contribution is K+ leaving neurons during neuronal depolarization. It is helpful to consider how many different K+ channels may facilitate this: leak channels, voltage-dependent K+ channels (e.g. IA channels), Ca2+-gated K+ channels and TRPM7 channels, which can be activated by membrane stretch and have been implicated in osmotic responses in epithelial cells (Numata et al., 2007), as well as in various neurodegenerative disorders. The other osmotic counterbalance is probably the exodus of small organic osmolytes, including amino acids and sugars (Okada, 2024). Pannexins, channel pores closely related to VRACs, are opened by reactive oxygen species production (Weilinger et al., 2023) and, notably, have been implicated in post-seizure and post-hypoxic decrease in pyramidal cell membrane resistivity (Thompson et al., 2006). On the other hand, pharmacological agents that block pannexins, including carbenoxolone, lanthanum and mefloquine, did not block anoxia-induced depolarizations (Madry et al., 2010), which were instead attributed to glutamate release locally. Another candidate pore may be Piezo1 channels. When any of these channels open, small neuroactive organic molecules exit the cell through these pores, potentially evoking paracrine effects that could either act in a negative feedback manner to stop the event, or reinforce changes leading to an acceleration of the event. Along these lines, Andrew et al. (2022) postulated an SD activator compound, possibly acting on the K+/Na+ ATPase pump at the depolarization wavefront. All these effects would move osmotically active particles following the induction of SDs, ultimately aiding in rebalancing the osmotic gradient. Characterizing these potential paracrine effects will be key to understanding SDs and may lead to therapeutic avenues for controlling SDs at the same time as bolstering any protective benefits. The origin of this opinion piece was our discovery that SDs can be triggered even when extracellular [K+] is low (Parrish et al., 2023b), forcing us to consider other explanations of SD induction. That is not to dismiss a role for increased [K+]o in SD propagation; the fact that this can be both a cause and a consequence of SD defines it as a positive feedback mechanism, which is a key element of regenerative events. An argument against this model of propagation, however, is that electrode measurements show only small increases in [K+]o ahead of the SD, with the prominent increase in [K+]o occurring only after other indicators of the local SD (field potential recordings and evidence of cell swelling). On the other hand, this is exactly what would be expected if the K+ released by cells already incorporated into the SD is buffered, initially, by uptake in neighbouring territories as K+ enters cells through the reversed action of KCC2 transporters. The very sharp increase in extracellular K+ would only occur later once the osmotic crisis triggers the full cellular response, opening VRACs and other channels, and causing K+ to leave neurons by the mechanisms already described. Interestingly, increased [K+]o is not the only positive feedback mechanism apparent; both ATP and glutamate are released through VRAC, from both neurons and astrocytes (Chen et al., 2024), and both molecules open VRAC even under isotonic conditions (Okada, 2024). Glutamate release may also occur by reversing the Na+-dependent transporter, which ordinarily takes up glutamate from the extracellular space (Rossi et al., 2000). The relative importance of these different sources of glutamate has been debated, in part because we lack clean pharmacological blockers of VRAC. Notably, though, knockdown of LRRC8A in astrocytes reduced glutamate release induced by a hypotonic challenge by ∼80% (Hyzinski-García et al., 2014). The consequences of SDs are not trivial. SDs are accompanied by substantial ionic redistribution incurring high energetic costs for correction and the risks of increased [Ca2+]i. At the network level, SDs have been suggested to underlie both migraines and SUDEP. On the other hand, we propose, in this opinion piece, that the cellular mechanism that underlies SDs provides critical protection against cell swelling, and that the risk of cell swelling may supersede other dangers. Water intake associated with dendritic beading is frequently terminal for neurons (Takeuchi et al., 2005), although not inevitably so; interestingly, dendritic beading that occurs in K+-induced SDs is transient and neurons recover fully (Steffensen et al., 2015). Many cell classes do tolerate intermittent or even cell death because they can but this is not the for most neurons. if cells do not swelling will neuronal function simply by the of the the space is to the of the and swelling would neurons more increasing and potentially positive with the risk of We that neurons are of osmotic and, hence, to prevent swelling are the in against these Consistent with this idea, glia are the protection against osmotic in the by their water through aquaporins. An secondary of the neuronal to osmotic stress is that it may network activity, both by further neuronal swelling directly and by other cells into depolarization both effects may help to (Norby et al., 2025). of are also in of osmotically induced swelling because the Na+/K+ ATPase ordinarily a critical role in an SD during an ischaemic may further swelling at the of the neuronal however, may be if the in a In summary, we suggest that the of SD are as protection at the cellular the however, is that extruded K+ or other key neuronal may other neurons into a regenerative regenerative events may to that not be (e.g. brain that or and, the cellular consequences (e.g. ionic high release of ATP and other neuroactive when there were prior risks to brain the will be by the relative volume of the extracellular space, which the impact of extracellular ions or neurotransmitters released from a cell and how well the glial network may the transient Interestingly, the and have low extracellular volume that ions concentration and do more than in other brain These are the most conducive to It has also been that high SD propagation and that SD is by with (Somjen, 2004). This new of SD initiation in to osmotic stress was out of the to two experimental models of SD KCl application and halorhodopsin We further that this provides a model for a of natural and experimental triggers of SDs. remains a major in the because the exact nature of the neuronal membrane pore is A pharmacological can be used to the of this pore because blocking the pore would the cellular leading instead to cell of this common in SD induction osmotic stress, will perhaps the of SD initiation and point to new avenues for therapeutic The is not for the or of any by the than should be to the for the All have the final of the and to be for all of the All as for and all for are This was by the and the

Topics & Concepts