Endogenous GABA_A and GABA_B receptor-mediated electrical suppression is critical to neuronal anoxia tolerance

Abstract

Anoxic insults cause hyperexcitability and cell death in mammalian neurons. Conversely, in anoxia-tolerant turtle brain, spontaneous electrical activity is suppressed by anoxia (i.e., spike arrest; SA) and cell death does not occur. The mechanism(s) of SA is unknown but likely involves GABAergic synaptic transmission, because GABA concentration increases dramatically in anoxic turtle brain. We investigated this possibility in turtle cortical neurons exposed to anoxia and/or GABA_A/B receptor (GABAR) modulators. Anoxia increased endogenous slow phasic GABAergic activity, and both anoxia and GABA reversibly induced SA by increasing GABA_AR-mediated postsynaptic activity and Cl⁻ conductance, which eliminated the Cl⁻ driving force by depolarizing membrane potential (∼8 mV) to GABA receptor reversal potential (∼−81 mV), and dampened excitatory potentials via shunting inhibition. In addition, both anoxia and GABA decreased excitatory postsynaptic activity, likely via GABA_BR-mediated inhibition of presynaptic glutamate release. In combination, these mechanisms increased the stimulation required to elicit an action potential >20-fold, and excitatory activity decreased >70% despite membrane potential depolarization. In contrast, anoxic neurons cotreated with GABA_A+BR antagonists underwent seizure-like events, deleterious Ca²⁺ influx, and cell death, a phenotype consistent with excitotoxic cell death in anoxic mammalian brain. We conclude that increased endogenous GABA release during anoxia mediates SA by activating an inhibitory postsynaptic shunt and inhibiting presynaptic glutamate release. This represents a natural adaptive mechanism in which to explore strategies to protect mammalian brain from low-oxygen insults.

When deprived of oxygen, mammalian neurons are unable to produce sufficient ATP to meet cellular demands (1, 2). As a result, the Na⁺/K⁺ ATPase (Na⁺ pump) fails and neuronal membrane potential (V_m) becomes unsustainable and anoxic depolarization (AD) follows, causing electrical hyperexcitability, deleterious Ca²⁺ influx, and spreading depression in the penumbral region (2, 3). Numerous studies have focused on the role of glutamatergic N-methyl-d-aspartate receptors (NMDARs) in this mechanism, and although NMDAR blockade prevents glutamatergic excitotoxicity (4), it does not prevent AD-mediated injury or postinsult apoptotic cell death (5). Thus, it is not surprising that clinical interventions targeting glutamate receptors alone have been largely ineffective against anoxic or ischemic damage (6), and therefore examination of alternative mechanisms to limit excitability during such insults is necessary.

A potential therapeutic alternative to directly antagonizing excitatory pathways is to up-regulate inhibitory mechanisms such as those mediated by γ-aminobutyric acid (GABA), the primary inhibitory neurotransmitter in the mature mammalian CNS (7). GABAergic mechanisms are not strongly recruited in ischemic mammalian neurons. In fact, although [GABA] is elevated by ∼30% in ischemic murine brain, GABA_A receptor subunit mRNA expression is decreased by ∼85%, and GABA-evoked currents and [ATP] run down rapidly, suggesting endogenous GABAergic neuroprotection is transient and largely ineffective (8–10). Nonetheless, activating GABA receptors (GABARs) preinsult limits neuronal hyperexcitability in mammalian models of ischemic damage, and AD and cell death are not observed in the afflicted brain region (11, 12). Despite these promising results, GABA has received little attention as a clinical stroke intervention and the mode of GABAergic neuroprotection is poorly understood (13).

Unlike the modest increase observed in anoxic mammals, brain [GABA] is rapidly elevated in anoxia-tolerant species including freshwater turtles (Chrysemys picta bellii, Trachemys scripta elegans) and fish (Carassius carassius, Carassius auratus) (1, 14). These organisms achieve a large-scale reduction in cellular energy turnover and survive weeks to months of anoxic exposure without apparent detriment (1). The key to this tolerance may be an adaptive mechanism that combines depressed glutamatergic signaling with enhanced GABAergic signaling. Indeed, we have previously demonstrated that excitatory glutamatergic signaling (via NMDARs and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors; AMPARs) is down-regulated in these species during anoxia (15–17); however, the role of GABAergic signaling in anoxia tolerance has not been elucidated. In turtle brain, [GABA] is increased a remarkable 80-fold during anoxia and electrical activity is suppressed 75–95% (14, 18–20). Therefore, this organism provides a good model in which to examine the role of GABA in regulating electrical activity during anoxia. We hypothesized that increased GABAergic conductance to Cl⁻ mediates electrical suppression (i.e., spike arrest; SA) in the anoxic turtle cortex, and that in combination with glutamatergic channel arrest prevents hyperexcitability and provides neuroprotection against anoxic insults.

Results

Spike Arrest Is Comediated by GABA_A and GABA_B Receptors.

We examined SA in pyramidal neurons from intact cortical sheets because synaptic connectivity and endogenous neurotransmitter release are maintained in this preparation ex vivo (18, 20, 21). In normoxia, stepwise current injections elicited action potentials (APs) at a threshold (AP_th) of −49.1 ± 2.0 mV and spontaneous APs occurred with a frequency (AP_f) of 2.1 ± 0.3 Hz (n = 15), whereas during anoxia, AP_th depolarized to −31.1 ± 2.9 mV and AP_f decreased >70% (n = 19; Fig. 1 A, B, D, and E). These changes were reversed by reoxygenation and are consistent with previous reports of SA in anoxic turtle brain (15, 18, 20). GABA perfusion reversibly mimicked SA and depolarized AP_th in a dose-dependent fashion (n = 3–13 for each [GABA]; Fig. 1C), and at moderate (0.5 mM) or high (2.0 mM) [GABA], AP_th depolarized 20–35 mV and AP_f decreased 60–70% (n = 10–13 each; Fig. 1 A and D). Conversely, antagonism of GABA_ARs (with 25 μM gabazine; GZ) or GABA_BRs (with 5 μM CGP55845; CGP) each partially reduced the depolarization of AP_th and increased AP_f, whereas coantagonism of GABA_A+BRs abolished the anoxic depolarization of AP_th and increased AP_f 30- and 100-fold relative to normoxic and anoxic controls, respectively (n = 9–13 each; Fig. 1 A and D).

Fig. 1.

Spike arrest in the anoxic turtle cortex is GABA_A+BR-mediated. (A) Summary of neuronal AP threshold (AP_th) during a normoxic-to-anoxic transition with recovery (t = 10, 30, and 60 min). (B) Sample recordings of APs stimulated from electrically quiet neurons by stepwise current injection. (C) Dose–response relationship of [GABA] versus ΔAP_th. (D) Summary of AP_f expressed as fold change relative to pretreatment baseline. (E) Sample V_m recordings of APs from D. Treatments: GABA, 25 μM gabazine (GABA_AR antagonist), 5 μM CGP55845 (GABA_BR antagonist). Data are mean ± SEM from 9–19 separate experiments. Asterisks indicate significant difference from normoxic controls. Double daggers indicate significant difference from anoxic controls (P < 0.05).

Endogenous GABA Release Increases During Anoxia and “Resets” V_m to GABA Receptor Reversal Potential.

To directly assay anoxic changes in GABA release, we adapted techniques used elsewhere to examine GABAergic postsynaptic currents (PSCs) (9, 22), and used GABA_ARs to detect anoxic changes in extracellular [GABA] (n = 8; Fig. 2A) (for these experiments only, currents were isolated and enhanced as described in SI Materials and Methods and Fig. 2). Under these conditions, a significant tonic inhibitory current was not observed (normoxic I_tonic = 1.6 ± 3.9 pA vs. anoxic I_tonic = 3.1 ± 2.4 pA; n = 8; Fig. 2Aii); however, spontaneous GABAergic slow phasic PSCs were detectable during normoxia, and their amplitude doubled during anoxic perfusion (41.2 ± 1.6 to 82.1 ± 2.1 pA; n = 10; Fig. 2 Ai, B, and C), whereas PSC_f was unchanged (Fig. S1A). PSCs were abolished by GABA_AR antagonists (GZ or picrotoxin; PTX) and enhanced by GABA transporter antagonists (GAT; n = 4 each), confirming postsynaptic GABA_ARs activation by endogenous synaptic GABA release. To better understand this relationship, we measured native GABAergic postsynaptic potentials (PSPs) using the perforated-patch method to avoid perturbing intracellular [Cl⁻] and in the presence of 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and (2R)-amino-5-phosphonovaleric acid (APV) to isolate GABAergic events. Normoxic GABAergic PSP amplitude was 10.3 ± 1.8 mV, and events were depolarizing relative to V_m (n = 7; Fig. 2D and Fig. S1B). Similar to PSCs, GABAergic PSPs were abolished by GZ but not by CGP, indicating mediation by postsynaptic GABA_ARs (n = 10–13 each). PSPs occurred at a rate of 0.1 ± 0.01 Hz, and this frequency was unchanged during anoxia (0.09 ± 0.02 Hz); however, anoxic PSPs were markedly smaller in amplitude (−1.47 ± 0.7 pA), and their polarity was reversed such that PSPs generally became mildly hyperpolarizing relative to V_m, indicating that the relationship between V_m and the reversal potential of Cl⁻ (E_Cl) is reset during anoxia. Importantly, when we current-clamped neurons at −90 mV (near normoxic V_m) or −70 mV (depolarized relative to both normoxic and anoxic V_m), the resulting PSPs were not different between conditions, indicating that the mechanism underlying this change is an alteration of V_m homeostasis during anoxia.

Fig. 2.

Endogenous synaptic GABAergic activity is enhanced during anoxia and V_m resets to E_GABA. (A) (i) Raw spontaneous GABAergic current activity. (ii) Raw tonic GABAergic activity. To aide current detection, pipette [Cl⁻] was increased to 110 mM and neurons were clamped at −100 mV and cotreated with the NMDAR antagonist APV (25 μM) and the AMPAR antagonist CNQX (25 μM). (B) Summary of GABAergic PSC_amp. Neurons were examined per A. (C) Superimposed raw spontaneous GABAergic PSC recordings averaged from 30 events from three neurons per trace. (D) Sample recordings of GABAergic PSPs in normoxia and anoxia from neurons voltage-clamped at −70 mV or −90 mV (top and bottom traces, respectively) or free-running (no voltage-clamp; middle traces) in the presence of APV and CNQX. RMP, resting membrane potential [holding potential (HP)]. (E) Summary of paired V_m and E_GABA measured in the perforated-patch configuration. APs were prevented by the voltage-gated Na⁺ channel antagonist TTX (5 μM). (F) Summary of AP_rheo (I_inj to elicit an AP). Treatments were as in Fig. 1, with GABA transport inhibitors SKF 89976A hydrochloride (40 μM; GAT-1) and (S)-SNAP-5114 (20 μM; GAT-2 and -3). Data are mean ± SEM from 4–19 separate experiments. Asterisks indicate significant difference from normoxic controls. Double daggers indicate significant difference from anoxic controls (P < 0.05).

GABA_AR activation induces Cl⁻ flux in the direction of E_Cl (GABA_AR reversal potential; E_GABA) (7), which has not been previously measured in adult turtle brain. Therefore, to more closely examine the relationship between V_m and GABA_AR activity, we examined E_GABA. In normoxic whole-cell experiments, E_GABA was −80.9 ± 1.4 mV, and ∼8 mV depolarized relative to V_m in the same neurons (−88.8 ± 1.7 mV; n = 10; Fig. 2E). E_GABA was unchanged during anoxic perfusion (−80.8 ± 2.3 mV; n = 8; Fig. 2E and Fig. S1C); however, V_m depolarized to −80.2 ± 1.9 mV. Consistent with an underlying GABAergic mechanism, normoxic GABA perfusion depolarized V_m in a dose-dependent fashion (n = 3–13 each [GABA]; Fig. S1D), and 2 mM GABA depolarized V_m to E_GABA in normoxia and anoxia (−81.0 ± 2.1, n = 12, and −81.5 ± 1.2 mV, n = 7, respectively). In separate experiments, we examined the effect of normoxic GABAR modulation on V_m and E_GABA in the perforated-patch configuration. Consistent with our whole-cell results, E_GABA was depolarized relative to V_m (−85.6 ± 0.8 and −91.8 ± 0.7 mV, respectively; n = 7; Fig. S1E), whereas during GABA perfusion, V_m depolarized to −84.3 ± 1.3 mV (n = 10) in a GABA_AR-dependent fashion (n = 7). [Note: E_GABA could not be determined during GABA or GZ perfusion due to GABA_AR saturation or antagonism, respectively (Fig. 2E and Fig. S1 E and F).]

When E_GABA is depolarized relative to V_m, such as in the early stages of mammalian development and to a lesser degree in the anoxic turtle cortex, the Cl⁻ gradient is such that postsynaptic GABA_AR activation induces Cl⁻ efflux accompanied by reductions in cell volume (7). Therefore, to confirm the relationships between V_m and E_GABA observed in our electrophysiology experiments, we examined Cl⁻ movement and neuronal volume regulation using noninvasive assays [6-methoxy-N-ethylquinolinium iodide (MEQ) and calcein-AM fluorescence, respectively] (23, 24). Anoxia or GABA perfusion reversibly increased MEQ fluorescence 30–40% in a GABA_AR-dependent fashion, reflecting Cl⁻ efflux, because intracellular Cl⁻ quenches MEQ fluorescence (n = 5 each; Fig. S1 G and H). Similarly, calcein fluorescence increased 22.1 ± 4.0 and 9.0 ± 3.0% during anoxia or GABA perfusion, respectively, reflecting cell shrinkage, whereas GABA_A+BR antagonism reduced the anoxic increase by ∼80%, to 5.0 ± 7.7% (n = 4 each; Fig. S1 I–K). Therefore, unlike in most adult mammalian neurons, E_GABA is depolarized relative to normoxic V_m in adult turtle cortex. Furthermore, the GABA_AR-mediated increase in Cl⁻ conductance (G_Cl) is sufficiently large that E_Cl (E_GABA) becomes the primary determinant of V_m during anoxia, and V_m resets to E_GABA. Despite this initial mild depolarization of V_m, this mechanism strongly resisted further V_m depolarization beyond E_GABA because rheobase (AP_rheo; i.e., stimulation required to generate an AP) increased >15-fold during anoxia or GABA perfusion (n = 10–17 each; Fig. 2F).

GABA_A and GABA_BRs Mediate Distinct Post- and Presynaptic Mechanisms of SA.

Because anoxia or GABA perfusion induced Cl⁻ efflux and V_m depolarization, we hypothesized that GABAR activation increased neuronal conductance to Cl⁻. Not surprisingly, we found that anoxic perfusion reversibly increased neuronal whole-cell conductance (G_w) from 4.8 ± 0.3 to 6.7 ± 0.4 nS (n = 17 for each; Fig. 3 A and B) (see comment on previous G_w measurements in anoxic turtle neurons in SI Materials and Methods). This change was abolished by GZ but not by CGP (n = 7 each), whereas GABA perfusion increased G_w in a dose-dependent fashion (n = 3–13 for each [GABA]; Fig. 3C). Importantly, pretreatment with 0.5 mM GABA raised baseline G_w to 8.1 ± 0.6 nS (n = 9; Fig. 3A); however, following the transition to anoxia, G_w decreased to 6.6 ± 0.4 nS and this change was prevented by GZ (n = 9 each). This suggests the magnitude of the anoxic change in G_Cl was partially masked in our experiments, because G_w is a global cellular measurement that includes the conductance of all open channels and receptors. This is particularly relevant in this model because previous studies have demonstrated channel arrest of AMPARs (G_Na) and NMDARs (G_Na/Ca), decreased K⁺ leakage, and reduced K⁺ and Na⁺ channel and NMDAR density in anoxic turtle brain (1, 15), all of which would mask the GABA_AR-mediated G_Cl. To isolate ΔG_Cl from these confounding anoxic changes, we examined the effects of GABA on normoxic input resistance (R_in; the inverse of G_w) using the gramicidin perforated-patch technique. In support of our hypothesis that ΔG_Cl was partially obscured during anoxia, GABA reversibly decreased R_in ∼85%, from 314.5 ± 50.2 to 72.1 ± 13 MΩ (n = 10 and 7, respectively; Fig. S2 A and B), and this change was prevented by GZ but not by CGP (n = 8 and 6, respectively). Therefore, increased GABA release enhanced GABA_AR-mediated PSP activity and activated a large G_Cl, which was the primary determinant of G_w homeostasis during anoxia, causing V_m to depolarize to E_GABA.

Fig. 3.

GABA_AR activation induces a shunt-like increase in plasma membrane permeability to Cl⁻; GABA_BR activation decreases EPSP activity via a presynaptic mechanism. (A) Summary of G_w measured in the whole-cell configuration. (B) I–V relationships from A. (C) Dose–response relationship of [GABA] versus G_w. (D) Summary of EPSP_f expressed as fold change relative to pretreatment baseline. (E) Sample V_m recordings from neurons treated as indicated. Black bars represent duration of treatment. Treatments were as in Figs. 1 and 2. Data are mean ± SEM from 6–19 separate experiments. Asterisks indicate significant difference from normoxic controls. Double daggers indicate significant difference from anoxic controls (P < 0.05).

CGP perfusion increased action potential frequency approximately fivefold in both normoxia and anoxia (Fig. 1D) but, unlike GZ, had no effect on G_w (Fig. 3A). This suggests that GABA_BR-mediated inhibition occurs at the presynapse. Excitatory postsynaptic potentials (EPSPs) are mediated by presynaptic glutamate release, and unlike the dramatic increase characteristic of ischemic mammalian brain, [glutamate] decreases 47% in anoxic turtle brain due to a combination of sustained reuptake and reduced release, and glutamatergic EPSP_f and amplitude decrease (1, 2, 15). Glutamate release can be inhibited by agonism of presynaptic GABA_BRs, which activate K⁺ channels that hyperpolarize presynaptic cells and also inhibit Ca²⁺ channels that mediate vesicular glutamate release (25). Because [GABA] is elevated in anoxic turtle brain (14), such presynaptic mechanisms likely contribute to SA by decreasing postsynaptic glutamatergic drive. In support of this, we found that CGP enhanced AMPAergic EPSP_f in both normoxia and anoxia (n = 9–13 each; Fig. 3D). Furthermore, antagonizing NMDARs with APV prevented hyperexcitability in anoxic CGP-challenged neurons, and APs were prevented despite the CGP-mediated enhancement of EPSP activity during anoxia (n = 5; Fig. 3E).

GABAergic SA Is Critical to Anoxia Tolerance in Turtle Neurons.

To assess the importance of GABAergic mechanisms in anoxia tolerance, we measured hallmarks of excitotoxic cell death (ECD) characteristic of ischemic mammalian neurons, including hyperexcitability, AD, deleterious Ca²⁺ influx, and cell death (2, 3). Similar to previous examinations in turtle cortex, these parameters were unchanged in normoxia, whereas during anoxia, V_m depolarized to E_GABA (n = 12) and cytosolic calcium concentration ([Ca²⁺]_c) increased ∼30% (n = 5; Fig. 4 A and B and Fig. S3 A and B) (1, 17, 26). Anoxic changes were reversed by reperfusion, indicating turtle neurons were not significantly stressed. Similarly, propidium iodide (PI) uptake [assessed relative to maximum uptake in metabolically arrested neurons treated with iodoacetate (IAA) and sodium cyanide (NaCN); Fig. S3 D and E] was unchanged during 4 h of normoxia or anoxia (n = 3 each; Fig. 4 A and B and Fig. S3C).

Fig. 4.

Pharmacological inhibition of endogenous inhibitory GABAergic mechanisms during anoxia is excitotoxic to turtle neurons. Sample V_m recordings (upper trace in each grouping), fura-2 Ca²⁺ fluorescence recordings (lower trace), and paired raw PI fluorescence images before treatment onset (t = 0 h) and following 4 h of treatment from neurons treated with normoxia (A), anoxia (B), anoxia + GABA_A+BR antagonists (C), and anoxia + GABA_A+BR antagonists in the presence of TTX (D). Black bars represent duration of treatment. Treatments were as in Figs. 1 and 2.

In striking contrast, neurons cotreated with GZ and CGP underwent recurrent seizure-like events (SLEs) and terminal depolarization of V_m (n = 17), a fourfold increase in [Ca²⁺]_c during anoxia that became further elevated to >1 μM during reperfusion (n = 4), and an increase in PI uptake within 2 h of treatment onset that at 4 h was elevated four- and sevenfold above normoxic and anoxic controls (n = 3; Fig. 4C and Fig. S3 A–C). This suggests SA is critical to anoxia tolerance because blocking the inhibitory GABAergic activity sensitized V_m to depolarizing inputs, resulting in unsustainable electrical hyperexcitation and terminal depolarization of V_m. We hypothesized that SA is predominately mediated by a GABA_AR-dependent electrical shunt mechanism that is passively recruited to oppose spontaneous excitatory events. If correct, pharmacological inhibition of APs with voltage-gated Na⁺-channel antagonist tetrodotoxin (TTX) should limit electrical excitability in GABA_A+B R-antagonized neurons, eliminate the requirement of SA as critical for anoxia tolerance, and ameliorate cytotoxicity despite persistent GABA_AR blockade and enhanced presynaptic glutamate release. In support of this, TTX inhibited APs and prevented SLEs (n = 6), deleterious [Ca²⁺]_c accumulation (n = 4), and PI uptake by GABA_A+BR-antagonized anoxic neurons (n = 4; Fig. 4D and Fig. S3 A–C).

Discussion

Using a naturally anoxia-tolerant model, we show that GABA_A/BRs are critical to SA and neuronal survival against low-oxygen insult in an anoxia-tolerant species. Importantly, we also show through the use of GABA uptake blockers and GABA_A/BR antagonists that the anoxic response is sensitive to endogenous synaptic GABA release and activates distinct inhibitory mechanisms on pre- and postsynaptic cells. The primary mechanism of SA is an increase in slow phasic GABA_AR-mediated PSP activity, which opens a postsynaptic G_Cl that dominates G_w during anoxia. We propose that this conductance acts as an inhibitory electrical shunt that effectively clamps V_m at E_GABA during anoxia, opposing further V_m depolarization due to glutamatergic EPSPs. This is likely the primary mechanism underlying anoxic SA, because GABA_AR antagonism is significantly more excitatory than GABA_BR antagonism. Nonetheless, GABA_BR activation contributes to SA by decreasing postsynaptic EPSP_f during anoxia, presumably via inhibition of presynaptic glutamate release. Importantly, V_m depolarization, such as the shift to E_GABA we report here, is usually excitatory in neurons (7); however, the voltage separation between V_m and AP_th paradoxically increases during anoxia because AP_th depolarizes twofold further than V_m. Therefore, although anoxic neurons depolarize to E_GABA, V_m is hyperpolarized relative to AP_th, and the net effect of these changes is inhibitory. Indeed, despite V_m depolarization, rheobase increases ∼20-fold during anoxia and fewer APs are elicited by each supra-AP_th stimulus train.

E_GABA is primarily determined by the Cl⁻ gradient, which is in turn dependent upon the balance in expression and activity of Na⁺-K⁺-Cl⁻ (NKCC1) (Cl⁻ influx) and K⁺-Cl⁻ (KCC2) (Cl⁻ efflux) cotransporters (7). In developing mammalian brain, KCC2 expression is low relative to NKCC1, and GABA is the primary excitatory neurotransmitter. As a result, mammalian neonate brain is particularly prone to hyperexcitability and seizures, but can be protected by simultaneously inhibiting NKCC1 and antagonizing postsynaptic GABA_ARs to temporarily switch the Cl⁻ gradient and render GABA inhibitory (27). Conversely, in adult mammalian brain, KCC2 expression is high and E_GABA is generally hyperpolarized relative to V_m (7). Therefore, GABA perfusion opposes AP generation by hyperpolarizing V_m away from AP_th via GABA_AR-mediated Cl⁻ influx; however, this form of inhibition requires increased Na⁺-pump activity to maintain ion gradients and is therefore not an effective strategy (1).

In adult mammalian brain, endogenous synaptic GABA release in the first few minutes of ischemia slows the onset of AD; however, this limited inhibition is insufficient to preserve [ATP] and maintain ion pumping >2–3 min (9). As a result, ion gradients are lost and extracellular K⁺ concentration ([K⁺]_o) increases sharply, causing AD within ∼6 min of insult (2, 3, 9). Extracellular neurotransmitter levels also increase rapidly in the first few minutes of ischemia (2, 9, 22), and glutamatergic Na⁺ and Ca²⁺ influx quickly depolarizes V_m away from E_GABA, thereby enhancing AD and also the inward-driving force on Cl⁻. When extracellular GABA activates GABA_ARs, large-scale Cl⁻ and associated water influx occurs (9, 23). This abrupt osmotic shift causes ischemic cells to swell and, depending on the severity of AD, undergo necrotic membrane rupture (28). This problem is compounded by impaired cerebral blood flow (CBF), which limits the removal of deleterious ions and neurotransmitters, the extracellular accumulation of which are required for ischemia-related cell swelling (2, 29). In contrast, CBF is maintained during prolonged anoxia in adult turtle brain (1), and synaptic transmission persists in vivo and in vitro (18, 19, 30; present report). Furthermore, [K⁺]_o, and cellular [ATP] are maintained for at least 6 h in anoxic turtle brain (30), and cells do not exhibit a necrotic phenotype. Finally, whereas ischemic adult mammalian neurons swell, anoxic turtle neurons shrink because E_GABA is depolarized relative to normoxic V_m, and therefore anoxic activation of GABA_ARs causes Cl⁻ and water efflux and thus cell-volume reduction. To our knowledge, the inhibitory depolarized relationship between E_GABA and V_m we report in adult turtle neurons is a unique phenotype and may represent an evolutionary adaptation that allows this species to retain neuronal function while reducing energy expenditure during prolonged anoxia.

This unique phenotype of GABAergic inhibition contributes crucial energy savings by reducing the demand on ATPases and related cotransporters during anoxia. The bulk of these savings is attributable to the GABA-mediated decrease in AP firing (i.e., SA), which reduces activity-induced perturbations of ionic gradients and the requirement for compensatory Na⁺-pump activity (1). In addition, V_m resets to E_GABA during anoxia, and is depolarized relative to normoxia, which further reduces pump demand. Also, because GABAergic G_Cl dominates G_w during anoxia, the importance of G_K in determining V_m is reduced. As a result, K⁺-channel expression and G_K are decreased without detriment in anoxic turtle brain, and the pump activity required to maintain K⁺ gradients is inherently reduced (1, 30). Finally, glutamate reuptake is sustained during anoxia in turtle brain and is also primarily dependent on ATPase activity (31). GABA_BR-mediated inhibition of presynaptic glutamate release limits the workload on these pumps, further reducing ATP consumption. Indeed, turtle brain Na⁺-pump activity is known to decrease ∼35% during anoxia (1); therefore, when combined, these effects provide significant energy savings and suppress neuronal excitability sufficiently to maintain [ATP] and V_m despite a >90% reduction in ATP production during anoxia (1). This conclusion is supported by the observations that: (i) GABA_A+BR inhibition prevents anoxia-mediated SA and neurons exhibit hallmarks of ECD, indicating that in the absence of GABAergic inhibition, anaerobic ATP production is insufficient to sustain V_m homeostasis; and (ii) simultaneous inhibition of GABA_A+BRs and voltage-gated Na⁺ channels prevented the GABA_A+BR antagonist-mediated increase in excitatory activity and ECD during anoxia.

The lack of a measurable tonic inhibitory current in turtle cortical neurons is an important finding. Tonic currents are mediated by extrasynaptic GABA_ARs and set the excitability threshold for neurons in adult mammalian brain (32). Baseline tonic inhibition is typically >50 pA, and this activity increases greater than twofold postischemic insult in mammals and impairs functional recovery following stroke (9, 33). Conversely, anoxia-tolerant turtle cortical neurons likely do not express extrasynaptic GABA_ARs because they lack an observable tonic current. This is not uncommon; many cell types do not express these receptors (32). More importantly, because E_GABA is depolarizing relative to V_m in normoxic turtle brain, any tonic GABAergic activity would be excitatory, and therefore the absence of such a tone in this model is not surprising. Indeed, the lack of any tonic inhibition minimizes the cellular signal-to-noise ratio of synaptic GABAergic inputs, and would therefore acutely sensitize turtle neurons to slow phasic (i.e., synaptic GABA_AR-mediated) inhibition.

The phenotype of phasic GABAergic inhibition in this model is unique: Slow phasic GABAergic activity occurs during normoxia at a frequency that is ∼1/15th that of resting mammalian neurons (9). This activity doubles in amplitude but is unchanged in frequency throughout >30 min of anoxic exposure, whereas phasic frequency in ischemic mammalian neurons increases 20-fold (and 300-fold relative to anoxic turtle neurons) within seconds of insult onset, and GABA_ARs quickly become desensitized to extracellular GABA binding after ∼10 min due to NMDAR-mediated Ca²⁺ accumulation (9). In comparison, the phasic GABAergic response in anoxic turtle brain is sustained, likely because Ca²⁺-mediated GABA_AR desensitization is prevented in turtle by concurrent glutamatergic channel arrest and maintenance of [Ca²⁺]_c homeostasis during prolonged anoxia (15, 17), and because the lack of an underlying tonic baseline allows for considerably greater inhibition from a relatively modest increase in phasic activity.

It is important to note that although GABAergic pathways are pivotal to the anoxia tolerance of turtle neurons, several other mechanisms act in concert to provide this neuroprotection. Indeed, in turtle cortical neurons in which channel arrest of NMDARs is prevented, excitotoxicity is also observed (34), suggesting multiple mechanisms may provide critical contributions to anoxia tolerance in parallel. Nonetheless, these mechanisms combine to reduce whole-animal energy consumption >90% during anoxia, which matches the decrease in energy produced anaerobically relative to aerobic (normoxic) respiration (for a review, see ref. 1) and allows these animals to tolerate long-term anoxic insults. Many of these mechanisms cannot easily be adapted to mammals; however, cell-level mechanisms such as glutamatergic channel arrest and GABAergic up-regulation are more readily mimicked in mammalian brain and would provide significant energy savings during low-oxygen stresses. Indeed, in normoxic rat brain, AP generation and propagation and EPSPs account for ∼81% of neuronal energy expenditure (35).

In conclusion, we report that SA of turtle cortical neurons is key to prolonged anoxia tolerance and is mediated by unique inhibitory GABAergic mechanisms that combine increased GABA_A+BR activity with a mildly depolarized E_GABA to reduce AP firing and ATP demand. Although this phenotype appears to be unique to the adult turtle CNS, the elucidation of these endogenous inhibitory systems provides insight into effective neuroprotective strategies against low-oxygen insult that may be adapted to protect hypoxia-sensitive mammalian brain against stroke. Mimicking the combined GABAergic and glutamatergic phenotype of anoxic turtle brain by activating GABA_A+BRs while inhibiting KCC2 activity and AMPAR and NMDAR conductance may prove to be neuroprotective in mammalian stroke models, and this hypothesis deserves further evaluation. Finally, the turtle enters into a comatose state during prolonged anoxia and EEG activity is decreased to basal levels (19). Because general anesthetics are hypothesized to induce a comatose state through the inhibition of excitatory glutamatergic and enhancement of GABAergic neurotransmission (36), as we have shown in the anoxic turtle, it is tempting to speculate that this model represents a natural anesthetic mechanism.

Materials and Methods

This study was approved by the University of Toronto Animal Care Committee. Cortical slice dissection and anoxic whole-cell patch-clamp recording methods are described elsewhere (21, 26). Briefly, cortical sheets were dissected and placed in a bath with a flow-through perfusion system. Anoxia was achieved by gassing the perfusate with 95% N₂/5% CO₂. Perfusion lines were double-jacketed, the bath chamber was covered, and all air spaces were similarly gassed. Neurons were perfused with pharmacological modifiers and/or anoxia as specified in Results. Electrical activity was recorded for up to 2 h from pyramidal neurons. For perforated-patch experiments, electrodes were backfilled with a KCl solution containing 25 μg/mL gramicidin.

Methods for dye-loading and fluorescence experiments are described previously (17). Baseline fluorescence was recorded for 10–20 min and then the tissue was exposed to treatment for up to 80 min and then reperfused. Ca²⁺ changes were assessed using 5 μM fura-2. Cl⁻ changes were assessed using MEQ. Changes in MEQ fluorescence were calibrated using the Stern–Volmer equation as described elsewhere (23). Cell-volume changes were measured using calcein-AM (24). Cell death was assessed using a PI exclusion assay. PI fluorescence due to IAA plus NaCN treatment was assessed as maximal cell death, and each treatment group was normalized to this value.

Statistical Analysis.

Normalized data were analyzed using two-way ANOVA with a Student–Newman–Keuls post hoc test to compare within and against treatment and normoxic values. Significance was determined at P < 0.05.