Neural shutdown under stress: an evolutionary perspective on spreading depolarization

Abstract

Neural function depends on maintaining cellular membrane potentials as the basis for electrical signaling. Yet, in mammals and insects, neuronal and glial membrane potentials can reversibly depolarize to zero, shutting down neural function by the process of spreading depolarization (SD) that collapses the ion gradients across membranes. SD is not evident in all metazoan taxa with centralized nervous systems. We consider the occurrence and similarities of SD in different animals and suggest that it is an emergent property of nervous systems that have evolved to control complex behaviors requiring energetically expensive, rapid information processing in a tightly regulated extracellular environment. Whether SD is beneficial or not in mammals remains an open question. However, in insects, it is associated with the response to harsh environments and may provide an energetic advantage that improves the chances of survival. The remarkable similarity of SD in diverse taxa supports a model systems approach to understanding the mechanistic underpinning of human neuropathology associated with migraine, stroke, and traumatic brain injury.

INTRODUCTION

Under certain circumstances of metabolic stress, neural circuit operation in some animals shuts down via a process known as spreading depolarization (SD). Neural activity ceases due to synaptic failure and inactivation of action potentials; however, the key feature is the collapse of ion concentration gradients across neuronal and glial cell membranes, resulting in their near complete depolarization. The phenomenon has been studied in mammals for many years since its first description as a “spreading depression” of activity (Leao 1944), and it has been associated with the generation and exacerbation of neuronal damage arising from stroke and traumatic brain injury (Dreier and Reiffurth 2015; Shuttleworth et al. 2020). More recently, in insects, a similar ionic disturbance was first recognized in the context of motor pattern failure at high temperatures (Robertson 2004). It is particularly noteworthy that SD also occurs in insect nervous systems, where it has the identifying characteristics of mammalian SD (Rodgers et al. 2010; Spong et al. 2016a). This observation raises important questions about evolutionary origins, molecular mechanisms, and potential adaptive value that could help to set an agenda for future research. For example, much health-related research uses invertebrate models such as nematodes (Caenorhabditis elegans) and insects (Drosophila melanogaster). Drosophila is used as a model system for many human pathologies, including, but not limited to, cancer (Mirzoyan et al. 2019), neurodegeneration (Nainu et al. 2019), motor neuron disease (Walters et al. 2019), fragile X syndrome (Drozd et al. 2018), Parkinson’s disease (Aryal and Lee 2019), and Alzheimer’s disease (Tsuda and Lim 2018). Such comparative approaches rely on the fact that evolution of cellular mechanisms has been conservative and thus molecular genetic dissection in more tractable organisms can inform research that uses more clinically relevant organisms. If this is true for SD, then determining whether it arose in mammals and insects via conservative or convergent evolutionary processes is important. Regardless of its evolutionary origins, can neural circuit shutdown via SD ever be beneficial for an organism? The intent of this perspective is to consider the nature of SD in insects and mammals to examine their similarities and help to clarify the most important questions regarding mechanisms, origins, and consequences (damage or benefits).

PHENOMENOLOGY OF SPREADING DEPOLARIZATION

The triggers, electrophysiological properties, and nature of the spread of SD in insects and mammals appear similar (Rodgers et al. 2010; Spong et al. 2016a). SD is characterized by rapid changes in interstitial concentrations of ions and transmitters, generating a propagating wave of cellular depolarization associated with silencing of neural activity. It is commonly monitored electrophysiologically by recording extracellular potassium ion concentration ([K⁺]_o) in the interstitium and/or by recording a negative shift in the extracellular direct current (DC) potential. To date, these recordings from mammalian and insect preparations have been assumed to be similar. However, whereas the [K⁺]_o recordings are likely the same, a reconsideration of the different recording arrangements in insects and mammals suggests that the interstitial DC recordings are different representations of SD.

In mammalian experimental preparations and human patients (Dreier et al. 2019), the negative DC shift is thought to arise as a field potential resulting from the current sources and sinks associated with dynamic depolarization that vary according to the geometry of the neuronal array (Canals et al. 2005; Dreier et al. 2013; Makarova et al. 2007). The amplitude and shape of such a potential change depend on the location of the recording electrode in the array. However, in insects, the recordings are made by penetrating the sheath of the ganglion and reflect the DC potential developed across the blood-brain barrier (BBB). In insects, the sheath consists of an extracellular protective lamella on top of layers of perineurial and subperineurial glial cells, which are connected by gap junctions (enabling diffusion between glial cells) and septate junctions (limiting paracellular diffusion) forming the BBB (Limmer et al. 2014). This is similar to mammalian vascular endothelial and ventricular ependymal cells (Jarvis and Andrew 1988), and many BBB mechanisms are conserved between vertebrates and invertebrates (Hindle and Bainton 2014; Hindle et al. 2017). The glial layer forms the barrier to ion diffusion between the insect hemolymph and the interstitium of the nervous system (Schofield et al. 1984; Schofield and Treherne 1984; Treherne and Schofield 1979, 1981). The basolateral membrane of these cells faces the hemolymph, whereas the adglial membrane faces the nervous system and is expanded into folds to increase surface area and thus the capacity for ion transport (Fig. 1A). Modeling indicates that the transperineurial potential between the hemolymph and the neural interstitium (V_s; 20 mV; values are approximate) results from the difference between the basolateral potential (V_b; −70 mV; intracellular vs. hemolymph) and the adglial potential (V_a; −90 mV; intracellular vs. interstitium) (Schofield and Treherne 1984). With high [K⁺] in the hemolymph, V_b depolarizes toward 0 mV in a Nernstian fashion and V_s shifts to positive values (Fig. 1B), indicating a barrier to ion diffusion. Similarly, if interstitial [K⁺]_o increases in the nervous system, such as during SD, V_a will depolarize toward 0 mV and V_s will hyperpolarize toward −70 mV, generating a negative DC shift. Although field potentials synchronous with neural activity can be recorded in insect ganglia (e.g., Gwilliam and Burrows 1980), the depolarization of the adglial membrane of perineurial glial cells is without doubt the source of the large negative DC shift recorded during SD in insects. Whereas potentials do develop across the BBB in mammals [e.g., due to hypercapnia (Kang et al. 2013)], in brain slices the BBB is disrupted, so the negative DC shift cannot be generated across the BBB. It is important to note that although SD might present differently in insects and mammals due to the different recording arrangements, the mechanisms underlying the phenomenon itself (spreading collapse of ion gradients) could still be identical.

Fig. 1.

Origin of the negative direct current (DC) shift associated with spreading depolarization in insect ganglia. A: diagram of the perineurial layer of glial cells in the sheath of a cockroach ganglion illustrating microelectrodes recording the potential across the basolateral membrane of a perineurial cell (V_b) and the potential in the interstitial space (V_s), both relative to ground potential in the saline. E, electromotive force; R, resistance. B: effect of high K⁺ concentration on perineurial potentials. The potential across the adglial membrane (V_a) is obtained by subtraction (V_b − V_s). These recordings indicate the effectiveness and location of the insect blood-brain barrier at the perineurial glial layer. [Adapted with permission from Schofield and Treherne (1984)].

The near-complete depolarization of cortical neuron membrane potential (V_m) during SD is evident when an intracellular recording (Vi) is compared with an extracellular recording just outside the neuron (V_o) (Collewijn and Van Harreveld 1966; Czéh et al. 1993). In the gray matter of the mammalian higher brain, the negative shift in V_o is considered a field potential generated by local currents across neuronal membranes because it develops almost simultaneously with the depolarization of V_i (Fig. 2A) (Canals et al. 2005). The depolarization of insect neurons during treatments that evoke SD has been described as incomplete (10–25 mV from resting) (Armstrong et al. 2009; LeCorronc et al. 1997; Wu et al. 2001)]. However, these values were for V_i, relative to an indifferent electrode in the bathing solution at zero potential (Fig. 2Bi). When V_o is taken into consideration, insect neurons also depolarize nearly completely, although the changes in V_i and V_o are not necessarily simultaneous (Fig. 2Bii). For chilling-induced SD in locusts, the defining characteristics ([K⁺]_o surge and negative DC shift) occur after neurons in the ganglion have partly depolarized to silence (spike inactivation) (Robertson et al. 2017). The abrupt, near-complete depolarization of neuronal V_m is not reflected in the V_i recording because it is simultaneous with an abrupt depolarization of the adglial membrane of perineurial glial cells (resulting in the negative shift of V_o). V_i is recorded across neuronal membranes and perineurial glia, and simultaneous depolarization of V_m and V_a cancel each other out in a V_i recording. Nevertheless, some mammalian neurons also show two stages of depolarization, with the second often occurring after spike inactivation (Fig. 2C) (Czéh et al. 1993; see also Fig. 3A in Brisson et al. 2014). As noted above, the differences in the recordings could result from the same underlying mechanism (e.g., spreading failure of the Na⁺/K⁺-ATPase) presenting differently in insect and mammalian neuropil due to differences in tissue organization and electrode placement.

Fig. 2.

Spreading depolarization (SD) of mammalian and insect neurons. A: SD recorded at the apical dendrite of a pyramidal neuron in vitro. Simultaneous changes of intracellular voltage (V_i) and the extracellular field potential (V_o) result in neuronal membrane depolarization close to 0 mV throughout the negative direct current (DC) shift. [Adapted from Canals et al. (2005)]. Bi: chill-induced SD recorded from a flight motoneuron in a semi-intact preparation of the locust. The negative DC shift (V_o) occurs after the neuronal action potentials have been inactivated (V_i). Temp, temperature. [Adapted from Robertson et al. (2017)]. Bii: reanalysis of a recording from a different neuron illustrates that neuronal membrane potential (V_m) depolarizes close to 0 mV but that this is a consequence of the abrupt change in V_o. C: hypoxia-induced SD recorded from a pyramidal neuron in vitro. Note the similarity to Bii. [Adapted with permission from Czéh et al. (1993)].

Fig. 3.

Repetitive spreading depolarization (SD) in insects and mammals. A: recording of the transperineurial potential (V_s) of a cockroach abdominal ganglion during exposure to high-K⁺ bathing solution, which induces repeated negative direct current shifts. [Adapted with permission from Schofield (1990)]. B: repetitive extracellular K⁺ concentration ([K⁺]_o) surges recorded in the Drosophila brain in response to ouabain delivered by injections of 0.5 nL of 2 × 10⁻⁴ M ouabain every 5 min (arrows) until SD initiates. [Adapted from Spong et al. (2016b)]. C: repetitive [K⁺]_o surges recorded from the pyramidal layer of the brain of a rat in vivo during microdialysis of fluorocitrate (a gliotoxin). [Adapted from Largo et al. (1997)].

Mammalian SD is a phenomenon arising in gray matter and does not spread into white matter. In insects, “white” matter (axon tracts in the connectives between ganglia) is also clearly separated from “gray” matter (neuronal somata and neuropil within each ganglion), and SD ignited in one ganglion will similarly not spread through the connectives to an adjacent ganglion (Rodgers et al. 2010). The perineurial glia of both the connectives and the ganglia develop a transperineurial potential with transient or maintained high hemolymph [K⁺] (0.1 M), indicating the barrier to ion diffusion (Fig. 1B). However, it is an interesting observation that pulses of high [K⁺] delivered outside the sheath of ganglia (gray) can trigger a negative DC shift of interstitial potential (Schofield 1990) such as is characteristic of K⁺-induced SD. This is not true for connectives (white). Similarly, only in ganglia does maintained high [K⁺] in the bathing solution generate repetitive negative DC shifts (Schofield 1990) (Fig. 3A). This is strikingly reminiscent of the repetitive SD generated in insects by bath application of the Na⁺/K⁺-ATPase inhibitor ouabain (Rodgers et al. 2007, 2009; Spong et al. 2016b) (Fig. 3B) and in rat brain by treatment with fluorocitrate, which can be used as a gliotoxin (Largo et al. 1996, 1997) (Fig. 3C). The regional difference in the insect nerve cord is reflected in the ultrastructure of the associated perineurial glial cells and current flow across the sheath. At ganglia, where there is normally a net current flow from the nervous system into the hemolymph, the perineurial glial cells contain a very high density of mitochondria, whereas at connectives, where there is normally a net current flow from the hemolymph into the nervous system, mitochondria are much sparser and microtubular arrays are prominent (Smith and Shipley 1990). This extracellular current loop (out of ganglia and into connectives) is carried by K⁺ ions and is sensitive to metabolic inhibitors and ouabain, which increase K⁺ current out of the ganglion, indicating that the pumps are working against a maintained efflux of K⁺, presumably as a result of K⁺ flowing down its concentration gradient (Kocmarek and O’Donnell 2011). In addition, immunocytochemistry for the α-subunit of the Na⁺/K⁺-ATPase reveals a high density of pumps in the perineurial glia of locust ganglia (Hou et al. 2014).

In summary, SD in insects is associated with energetically demanding, perineurial glial cells of the BBB in the sheath surrounding the integrative centers of the nervous system. Moreover, activation of cellular signaling pathways indicating metabolic stress exacerbates repetitive SD induced by ouabain (Rodgers-Garlick et al. 2011), likely as a result of changes that could reduce the energetic cost of signaling (Money et al. 2014) but increase vulnerability to SD. The nature of the negative DC shift in the interstitium may be different in insects and mammals. However, these recordings are strongly influenced by the neural architecture, and the fundamental mechanism of SD appears to be similar in the different taxa.

MECHANISMS

Despite more than 70 years of research on mammalian SD, there is still no consensus on the mechanisms that trigger and propagate SD, although the unifying central feature is the inability of Na⁺/K⁺-ATPase pumps to maintain appropriate ion gradients as a result of energy limitation, pump impairment, or increased demand from greater ion fluxes (Andrew RD, Robertson RM, Revah O, Ullah G, Farkas E, Muller M, Ollen-Bittle N, Hartings JA, Shuttleworth CM, Brennan KC, Herreras O, Kirov SA, Ayata C, Dawson-Scully K, and Dreier JP, unpublished observations; Brisson et al. 2014; Major et al. 2017). This suggests that the regions of the nervous system most vulnerable to SD will be those heavily reliant on energetically demanding pump activity with high densities of mitochondria and pumps (such as the perineurial glial layer of ganglia in insects). To date, research focus has been on identifying one or more channels in neuronal membranes that could be triggered to mediate the large current that drives SD (as defined by V_m depolarization close to 0 mV, concurrent with a surge of [K⁺]_o and a negative DC shift of V_o). The fact that this is likely a nonspecific Na⁺/K⁺ channel (Czéh et al. 1993) has largely been ignored (Andrew RD, Robertson RM, Revah O, Ullah G, Farkas E, Muller M, Ollen-Bittle N, Hartings JA, Shuttleworth CM, Brennan KC, Herreras O, Kirov SA, Ayata C, Dawson-Scully K, and Dreier JP, unpublished observations).

One possible SD mechanism that could help to explain the lack of success in identifying an appropriate channel is that SD in response to metabolic stress is not just generated by pump overload but by actual conversion of the sodium pump into an open channel. It is now well established that palytoxin (PLTX), a potent poison concentrated by certain corals, exerts its lethal effect by converting the Na⁺/K⁺-ATPase pump into an open channel that nonspecifically conducts Na⁺ influx and K⁺ efflux (Artigas et al. 2000; Artigas and Gadsby 2004; Gadsby et al. 2009). This indicates definitively that a vulnerability of the pump exists. This toxin likely evolved in dinoflagellates to protect them, and subsequently the coral, from diverse predators by exploiting a conserved pump vulnerability. It is possible that insects have exploited this pump vulnerability such that metabolic stress evokes SD to temporarily shut down all or parts of the central nervous system (CNS) as part of a neuronal defense strategy. Whether mammals adopted a similar strategy to deal with traumatic brain injury is conjecture at this point.

The similarity of the SD trajectories of oxygen/glucose deprivation (OGD), ouabain, and PLTX in mammals suggests they may be generated by the same process and has prompted the intriguing suggestion that the elusive SD channel is in fact a converted pump (Brisson et al. 2014). In support of this idea, single-channel recording of cell-attached patches reveals opening of a novel channel in response to OGD that has similar properties to a channel opened by PLTX under the same conditions (Gagolewicz and Andrew 2017). If conversion of the pump to a channel is the SD mechanism, this could provide an evolutionary explanation for SD in diverse animals. The pump is a member of the family of P-type ATPases in the IIC subgroup (Horisberger 2004), which are evolutionarily ancient, being found in algae, fungi, and protists with likely a prokaryotic origin (Sáez et al. 2009). Mutations in PII ATPases can cause ion leakage associated with several human pathologies (Kaneko et al. 2014). The ancestral gene evolved to generate many ion-pumping ATPases, but only animals (metazoa) possess gene variants able to exchange Na⁺ for K⁺ (Okamura et al. 2003). Hence, there are good reasons for suspecting that a pump vulnerability could underlie SD and that evolutionarily conserved properties might explain the similarity between SD in mammals and insects. On the other hand, the phylogeny also shows that there has been considerable divergence from the ancestral pump protein subunits because of molecular adaptation and gene and genome duplication (e.g., Lohr et al. 2017). The biophysical effects of PLTX on the Na⁺/K⁺-ATPase have been investigated exclusively using vertebrate isoforms and, whereas PLTX has effects on invertebrate axons [cockroach and squid (Pichon 1982)], there is no information on PLTX and pumps constructed with invertebrate isoforms. In addition, insect pump isoforms have shown convergent and parallel evolution of molecular strategies to reduce sensitivity to plant secondary compounds, such as cardiac glycosides (e.g., ouabain) (Dobler et al. 2015; Taverner et al. 2019; Yang et al. 2019), suggesting that their vulnerabilities may vary. It remains to be determined whether PLTX will generate SD-like phenomena in insect ganglia.

The primary contenders to be mediators of SD propagation in the past have been K⁺ ions and glutamate because of their progressive release into the interstitium (Grafstein 1956; Van Harreveld 1959; Van Harreveld and Kooiman 1965). A critical examination questioning the role of K⁺ and glutamate as mediators of SD will be presented more fully elsewhere (Andrew RD, Robertson RM, Revah O, Ullah G, Farkas E, Muller M, Ollen-Bittle N, Hartings JA, Shuttleworth CM, Brennan KC, Herreras O, Kirov SA, Ayata C, Dawson-Scully K, and Dreier JP, unpublished observations). In insects, a plausible scenario is that a metabolic crisis or pump impairment triggers SD by preventing K⁺ clearance from the extracellular space. There is a continuous, energy-dependent flow of K⁺ out of ganglia. Presumably, this is via pumping K⁺ into glia and then spatial buffering to regions of lower [K⁺]_o (Spong and Robertson 2013). When pumps eventually stop working, due to energy starvation and/or crossing a [K⁺]_o threshold, there will be an immediate back-up of K⁺ in the interstitium. A primary role for the pump is in cell volume regulation (Kay and Blaustein 2019), and thus the effects will be amplified by inevitable cell swelling, which in a confined space with an inelastic ganglion sheath will tend to increase [K⁺]_o because the extracellular space is reduced, perhaps accounting for the positive feedback underlying this all-or-none event. This will depolarize adglial membrane of perineurial glia to 0 mV, generating the negative DC shift in the interstitium.

It is likely that glutamate is not the mediator of SD propagation in insects, but for different reasons than in mammals (Andrew RD, Robertson RM, Revah O, Ullah G, Farkas E, Muller M, Ollen-Bittle N, Hartings JA, Shuttleworth CM, Brennan KC, Herreras O, Kirov SA, Ayata C, Dawson-Scully K, and Dreier JP, unpublished observations). Glutamate has been established for many years as a neuromuscular excitatory transmitter in insects (Cull-Candy et al. 1981; Usherwood et al. 1968), but, in contrast with mammals (Reiner and Levitz 2018), the primary CNS excitatory neurotransmitter in insects is acetylcholine (Breer and Sattelle 1987). Neuroanatomical data localizing glutamatergic neurons immunocytochemically for glutamate (Bicker et al. 1988; Watson and Seymour-Laurent 1993) or the vesicular glutamate transporter (Mahr and Aberle 2006; Takamori et al. 2000) confirm a role for glutamate as a transmitter in the CNS but cannot distinguish whether it is excitatory or inhibitory. Extrajunctional receptors on insect muscle fibers respond to glutamate application with hyperpolarizing or depolarizing responses (Cull-Candy 1976), and similar hyperpolarizing (Dubas 1990) or depolarizing (Parker 1994; Sombati and Hoyle 1984) responses to glutamate are evident for neurons in the insect CNS. Genomic analysis has demonstrated that invertebrates possess chloride-specific pentameric ligand-gated ion channels responsive to glutamate that chordates lack. They are structurally related to vertebrate glycine and GABA receptors (Barbara et al. 2005; Dent 2006; Lynagh et al. 2015; Wolstenholme 2012). As a result, much glutamate-mediated CNS neurotransmission in insects is inhibitory (Liu and Wilson 2013; MacNamee et al. 2016), which is not the case for mammals.

Given the importance of glutamate N-methyl-d-aspartate (NMDA) receptors in the mammalian CNS, there is obvious interest in determining whether insect homologs have similar functions. Early experiments suggested that invertebrate preparations were relatively insensitive to NMDA agonists and antagonists (Glantz and Pfeiffer-Linn 1992). Since then we have found that the Drosophila genome encodes two NMDA receptor homologs (Chiang et al. 2002), which are involved in olfactory learning (Xia and Chiang 2009; Xia et al. 2005) and synaptic plasticity associated with sleep (Robinson et al. 2016). RNA interference of NMDA receptor expression disrupts memory formation in honeybees (Müssig et al. 2010) and long-term memory consolidation in Drosophila (Wu et al. 2007). In addition, NMDA antagonists can block locomotor pattern generation in Drosophila larvae (Cattaert and Birman 2001) and juvenile hormone synthesis in cockroach adults (Chiang et al. 2002) (but see Huang et al. 2015). Phylogenetic analysis has demonstrated a greatly expanded family of glutamate kainate receptors in insects compared with the smaller family in mammals (Ramos-Vicente et al. 2018). A heterologously expressed Drosophila kainate receptor (DKaiR1D) from this family has a novel pharmacological profile such that it is potently inhibited by NMDA and by the classic NMDAR antagonist 2(R)-amino-5-phosphonopentanoic acid (Li et al. 2016). Clearly, glutamate neurotransmission via ionotropic glutamate receptor homologs and paralogs has a role in CNS operation in insects, but the functions and molecular mechanisms are likely to be considerably different.

Hence, assuming that the similarity between insect and mammalian SD reflects a fundamental etiology, a focus on the potential role of a specific transmitter or receptor subtype in triggering SD may be short-sighted (except in the context of therapeutic drug discovery). To better understand SD, it may be more fruitful to consider what neural and CNS properties predispose to SD in insect and mammalian nervous tissue and how they differ from the properties of nervous systems that have well described CNS glutamate neurotransmission, such as in molluscs (Di Cosmo et al. 2006; Ha et al. 2006; Megalou et al. 2009), but for which there is no current evidence for the existence of a process resembling SD. Modeling approaches that incorporate generic properties of channels and pumps independently of species-specific molecular identities have much to offer (Ullah et al. 2015; Wei et al. 2014).

ORIGINS

Molecular signaling mechanisms have a long evolutionary history, and most neurochemicals predate the appearance of nervous systems and even the emergence of eukaryotes (Greer et al. 2017; Kristan 2016; Liebeskind et al. 2016; Plattner and Verkhratsky 2018; Varoqueaux and Fasshauer 2017). There is also much convergent and parallel molecular evolution of neural mechanisms to arrive at common solutions. This reflects the constraints imposed by the nature of the evolved function and the availability of substrates that can be appropriately evolved (Liebeskind et al. 2017; Lynagh et al. 2015; Nishikawa 2002). Thus it is indisputable that SD mechanisms in insects and mammals have common evolutionary origins. A common evolutionary origin could be relatively uninteresting if, for example, SD is a feature of all centralized nervous systems, i.e., with organization into the equivalents of gray and white matter. However, there is strong evidence that ouabain, metabolic inhibitors, and abiotic stressors, which ignite SD in insects and mammals, do not generate anything like SD in a variety of molluscan models (Alvarez-Leefmans et al. 1994; Nikolić et al. 2012; Russell and Brown 1972; Thomas 1974). This is also true for annelid nervous systems, such as in the leech (Frey et al. 1998; Schlue and Deitmer 1980). Recent deliberate attempts to trigger SD-like phenomena in the CNS of the pond snail, Lymnaea stagnalis, by using sodium azide or ouabain were unsuccessful (Wassef D, Magoski NS, and Robertson RM, unpublished observations). It is reasonable to conclude that insects and vertebrates generate SD but that slugs, snails, worms, and leeches do not. The major neural differences between insects and these other invertebrates, which make the insect CNS more like the vertebrate CNS, include the following:

1. The high speed and large capacity of information processing in insects due to the complexity and rapidity of behaviors such as visual control of flight, which requires neurons with high firing frequencies, short action potential durations, and a high demand for energy;

2. Increased neural complexity in insects in terms of numbers of neurons and compactness of neuropil with small interstitial volumes; and

3. Existence of an effective CNS/hemolymph barrier to ion flow in insects but not in annelids or most molluscs.

Of course, these properties are interconnected; the evolution of rapid, complex behaviors is enabled by neural complexity. Moreover, precise and repeatable neuronal information processing requires a stable ionic environment. It is likely not a coincidence that, broadly speaking, animals that exhibit SD also have highly regulated ion homeostasis in the nervous system, including an effective BBB, whereas animals in which there is unrestricted access of ions in the hemolymph to neuronal surfaces [e.g., gastropod molluscs (Sattelle and Lane 1972)] apparently do not exhibit SD. SD is a phenomenon associated with restricted extracellular spaces in the CNS: spaces that, when something compromises ion homeostasis (e.g., failure of pumps or of glial clearance), promote the rapid accumulation of K⁺ ions that have a limited opportunity to diffuse away in the constrained space. Comparative studies of invertebrates indicate that a BBB is absent in annelids and earlier evolved molluscs but well developed in insects and cephalopod molluscs (Abbott et al. 1986). In most vertebrates, the BBB is an endothelial barrier but the ancestral condition is glial (Bundgaard and Abbott 2008). The difference in BBB (glial in invertebrates; endothelial in all but the most ancient vertebrates) may account for some the differences in how SD presents in insects and mammals. These observations lead to the hypothesis that the BBB initially evolved due to the selective advantage of the improved neural integration in the CNS that is enabled by closely regulating the ionic environment of central synapses (gray matter) (Abbott 1992). By extrapolation, it is conceivable that SD is an emergent property arising because of anatomical features that create a confined space in combination with an evolutionarily conserved ion pumping vulnerability. A testable prediction is that the CNS of cephalopod molluscs, which have rich behavioral repertoires and tight BBBs, will exhibit SD when metabolically stressed.

ADAPTIVE VALUE

Regardless of how SD originated, the phenomenon may or may not have adaptive value (to increase lifetime fecundity). Survival of the individual is necessary to reproduce, but it does not thereby follow that fitness is improved.

Cost/benefit considerations mean that biological systems are vulnerable to extreme conditions of operation; it is wasteful for a system to maintain resources for rare events. Hence, system failure is predictable. Terminal SD occurs at brain death in humans (Carlson et al. 2018; Dreier et al. 2018), which is at least suggestive that rather than being a controlled trait, SD is just how a complex CNS fails and terminates. The features that define SD are the catastrophic onset (all-or-none ignition) and the possibility of recovery. In insects, the properties of SD induced by K⁺, ouabain, extremes of ambient temperature, and anoxia give it an appearance of being controlled and triggered, a mechanism that reversibly shuts down parts of the nervous system. The modeling suggests that the pattern of currents and potentials during SD may be a consequence of energy limitation in a system with a limited extracellular space that evolved to generate action potentials (Ullah et al. 2015; Wei et al. 2014). However, this does not necessarily mean that its ignition could not also be controlled.

What benefits could accrue from SD in response to metabolic stress? Possibilities include energy conservation or reallocation. It is self-evident that shutting down neural circuit operation by hyperpolarization or otherwise preventing action potential generation will conserve resources [e.g., via channel arrest and spike arrest (Robertson 2017)]. Much more energy will be conserved, or could be reallocated to different cellular and/or tissue operations, if membrane potentials are temporarily not maintained in neurons and glia (i.e., CNS arrest as a consequence of global neural depolarization enabling, or as a consequence of, shutdown of the pumps). This seems noncontroversial for anoxia or oxygen/glucose deprivation when oxidative phosphorylation is precluded and ATP supply to the pump is interrupted. With a brief SD event, the net effect could be increased energy consumption due to the energetic load required for repolarization, but as the time of shutdown increases, at least in insects, then the net effect will be reduced energy consumption. Energy partitioning may be equally important, i.e., allocating the energy available at any one time to processes for cell or tissue integrity that are more important to the animal than information processing (e.g., protection of more vital physiological processes or producing stress proteins for preconditioning).

Under harsh environmental conditions (e.g., extremes of temperature, anoxia), insects enter a reversible coma that is associated with SD in the CNS. It is clear that the SD-associated coma in insects reduces metabolic rate to a low level consistent with maintaining cellular energy balance (Campbell et al. 2018; Robertson et al. 2017), although this does not last indefinitely and eventually cellular damage accumulates. This suggests that insect SD is a secondary adaptation to prolong survival under harsh conditions. To determine that SD is an adaptive feature would require, at a minimum, evidence that SD (not merely quiescence) by itself improves fitness. One of the strongest arguments in insects is that larval Drosophila can behave for 20 to 30 min in anoxia but die after around an hour (i.e., coma and SD are suppressed to prolong larval behavior), whereas adult Drosophila enter a coma in less than 30 s and can survive as long as 8 h in anoxia (Callier et al. 2015). The coma, or hypometabolic paralysis, is generated by SD in the CNS (Armstrong et al. 2011) and is associated with adults maintaining low but non-zero ATP levels for longer and being able to tolerate extreme ionic variability (Campbell et al. 2018). A metabolomic investigation concluded that adult anoxia tolerance in Drosophila is not associated with substrate limitation but is assisted by the hypometabolism evident during the anoxic coma (Campbell et al. 2019). The extent of the contribution of SD, as opposed to mere quiescence, to hypometabolism has not been measured, but it certainly seems that the SD-associated coma is beneficial for adult survival. In locust preparations where SD is monitored during hyperthermia, neural circuit failure at a lower temperature (i.e., earlier SD) is associated with a faster recovery when temperature returns to normal (Rodgers et al. 2007). Moreover, chill coma temperatures of different Drosophila species are matched to their thermal environments (Andersen et al. 2018) suggesting an adaptive radiation of chill coma mechanisms and the underlying SD (Robertson et al. 2017).

Perhaps the strongest argument that SD may be beneficial in insects is that genetic and pharmacological manipulations known to promote circuit and synaptic thermotolerance (Dawson-Scully et al. 2007) and delay or reduce the severity of SD in locusts (Armstrong et al. 2009) and Drosophila (Spong et al. 2016b) also prolong behavior in anoxia, but at the cost of reduced survival (Dawson-Scully et al. 2010). In other words, prolonging circuit function under stress impairs subsequent survival. Experimental evolution of Drosophila over ~1,000 generations improved the resilience of adult flies and their ability to recover from N₂- or CO₂-induced comas (Xiao et al. 2019), indicating that there are traits associated with anoxic coma that are variable and heritable, requirements for a trait to be adaptive. An important research direction for insect SD induced by abiotic stressors is to tease out any specific contribution of SD to mitigating damage and/or improving survival; it is hard to separate SD from other consequences of exposure to extreme conditions or, indeed, to expose the insect to a sufficiently extreme stress without triggering SD. Nevertheless, with insects it may be possible to devise treatments that are stressful enough to be impairing, perhaps by increasing the duration of exposure, but that are around the threshold for generating SD, thus generating a data set of responses to abiotic stress with, and without, SD.

In mammals there is also differential tissue vulnerability to anoxia/OGD. Mammalian neocortex is more susceptible to OGD (SD) than brain stem (Andrew et al. 2017; Brisson et al. 2014). One possible explanation is that pump isoforms are differentially distributed in the CNS and some isoforms are more vulnerable to the conditions that promote SD. Perhaps vital functions of the brain stem are protected from SD by virtue of resistant pump isoforms. SD propagation from a focal site of contusion or ischemia could also represent a signaling mechanism informing neighboring tissue of nearby metabolic compromise. Therefore, SD duration and/or frequency might represent a gradient of stress. SD may also be beneficial in contused gray matter, where it promotes vasoconstriction. However, with the prevalence of cardiac arrest and occlusive stroke because of increased life span coupled with unhealthy lifestyles, SD in response to restricted blood flow in patients likely becomes deleterious rather than advantageous. Nevertheless, in a preclinical rodent model of global cerebral ischemia, earlier SD onset is predictive of improved neurological outcome (Wilson et al. 2019). There is also increasing willingness to consider that, under some circumstances, SD in mammalian brains could be beneficial by preconditioning tissue and/or limiting the expansion of damage (Shuttleworth et al. 2020).

CONCLUSIONS

The evolution of the Na⁺/K⁺- ATPase and the pump/leak mechanism for cell volume regulation freed animal cells from the strictures of a rigid cell wall (Kay and Blaustein 2019). This capacity for rapid changes in ion currents and membrane potential preadapted animals for the evolution of electrical signaling mechanisms. Selective pressure for complex behavior requiring the tight control of ion concentrations around neural integrating centers resulted in benefits in lineages that were able to physically constrain and chemically isolate the extracellular fluid bathing neurons using energetically expensive homeostatic mechanisms and effective BBBs. The combination of high neural performance and expensive ion homeostasis revealed a vulnerability in the Na⁺/K⁺-ATPase. It fails at the limits of performance imposed by extreme conditions (e.g., low oxygen, low ATP, low and high temperature), resulting in SD. At least in insects, increasing evidence suggests that this neural off switch has been co-opted to improve survival under stress, likely by enabling resource conservation or reallocation. SD can also be beneficial in mammalian nervous systems, and the strongest evidence is for preconditioning at the cellular level to limit damage from exposure to subsequent events.

It is reasonable to conclude that molecular components (pumps, channels, etc.) underlying SD are evolutionarily conserved in insects and mammals to the same extent as molecular components underlying action potential generation are conserved in all animals. However, given the absence of SD in species arising from the common ancestor of insects and mammals, it is likely that the tissue arrangements (restricted extracellular space, BBB, etc.) that give rise to SD represent evolutionarily convergent solutions promoting high neural performance. Thus, fundamentally, SD in insects and mammals is the same phenomenon generated by similar mechanisms. However, there are bound to be subtle differences resulting from, for example, molecular evolution of pump and channel isoforms during the 600 to 700 million years since their most recent common ancestor (Peterson et al. 2008). This does not mean that SD as a trait has arisen or been preserved by processes of natural selection. SD is a phenomenon that may only be an emergent property of nervous systems that have been selected for high performance and a tight BBB. Such a property can be beneficial or deleterious in different animals and/or under different circumstances without being adaptive in an evolutionary sense. From the perspective of a clinician, it is probably more relevant whether SD in insects and mammals has similar mechanistic underpinnings rather than the evolutionary origins reflecting conservation versus convergence. To demonstrate that natural selection shaped SD in insects or mammals would require at least a fine-grained phylogenetic analysis comparing the presence/absence and properties of SD across many taxa. In addition, it would be important to identify the genetic basis for SD and show that genetic variability confers differential fecundity in a population. Such time-consuming approaches are not necessary to understand the physiology of SD, including whether SD is beneficial to the health of an individual or not.

Given that there is still no consensus regarding the molecular mechanism driving the SD current or SD propagation in mammalian nervous systems (Andrew RD, Robertson RM, Revah O, Ullah G, Farkas E, Muller M, Ollen-Bittle N, Hartings JA, Shuttleworth CM, Brennan KC, Herreras O, Kirov SA, Ayata C, Dawson-Scully K, and Dreier JP, unpublished observations), research with insect model systems continues to be valuable, if only as a source of hypotheses for testing in preclinical studies. An important question that can profitably be investigated in a Drosophila model concerns the relative roles of neurons and glia in generating or suppressing SD. For example, in a Drosophila model of repetitive anoxia-induced SD, expression of HSP70 (a chaperone protein) in glial cells, but not neurons, delayed the permanent loss of CNS ion homeostasis (Armstrong et al. 2011). Similarly, tissue-specific genetic manipulation of the α-subunit of the metabolic sensor AMPK modulates anoxia-induced SD when targeted to glia but not when targeted to neurons (Evans et al. 2017). The array of rapid molecular genetic approaches to target different mechanisms in different subtypes of neurons and glia in Drosophila is impressive, but SD research has not yet sufficiently exploited this resource. The occurrence of SD in diverse animals ensures that we can choose the most convenient animal to study its different properties and mechanisms.

GRANTS

R. M. Robertson is supported by a Natural Sciences and Engineering Research Council (NSERC) of Canada Discovery Grant. K. D. Dawon-Scully is supported by National Institutes of Health Grant GM110651. R. D. Andrew is supported by an NSERC Canada Discovery Grant and an operating grant from the Heart and Stroke Foundation of Canada.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

R.M.R. prepared figures; R.M.R. drafted manuscript; R.M.R., K.D.D.-S., and R.D.A. edited and revised manuscript; R.M.R., K.D.D.-S., and R.D.A. approved final version of manuscript.