Stellate and pyramidal neurons in goldfish telencephalon respond differently to anoxia and GABA receptor inhibition

During periods of oxygen deprivation, ATP synthesis in the anoxia-intolerant brain (mammals) halts and compromises the ability of the Na⁺/K⁺-ATPase pump to maintain neuronal ion gradients, leading to membrane depolarization (Hansen, 1985). The release of excitatory neurotransmitters (glutamate, aspartate) follows, leading to activation of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and N-methyl-d-aspartate (NMDA) receptors (Sattler and Tymianski, 2000). Subsequent calcium (Ca²⁺) influx through NMDA receptors causes further depolarization of membrane potential (V_m), and activates intracellular Ca²⁺-dependent signaling cascades, which leads to neuronal hyper-excitability and, eventually, excitotoxic cell death (Sattler and Tymianski, 2000). As anoxia eliminates oxidative phosphorylation, the glycolytic pathway becomes the major source of ATP produced by the brain, generating less than 1/10th of the ATP produced under normoxic conditions (Hochachka, 1986). The ATP shortfall is particularly problematic for neurons, as action potential propagation, consequent activation of output synapses and maintenance of neuronal ion gradients are metabolically expensive events, consuming approximately 60% of the total brain ATP turnover (Erecińska and Silver, 1994). Anoxia-tolerant organisms, such as the western painted turtle (Chrysemys picta), avoid anoxia-mediated cellular injury by not only reducing ATP synthesis by over 90% (Bickler and Buck, 2007) but also reducing ATP demand; indeed, in the case of turtle cortical pyramidal neurons, an anoxia-mediated 60–70% reduction in action potential frequency (AP_f) has been reported (Pamenter et al., 2011).

There are three hypotheses that encompass the events that lead to a large decrease in metabolic rate such as that exhibited by anoxic-tolerant species; these are the metabolic arrest (MA), channel arrest (CA) and spike arrest (SA) hypotheses (Hochachka, 1986; Lutz, 1992; Buck and Hochachka, 1993). The MA hypothesis generally describes events leading to a reduction in ATP supply, while the CA hypothesis is specific to a reduction in ion channel conductance leading to a reduction in ATP demand. The SA hypothesis is specific to neurons or excitable tissues and entails a reduction in AP_f that leads to a reduction in ATP demand and therefore a reduction in ATP supply (Hochachka et al., 1996). Neurons function within an integrated network; therefore, mechanisms regulating metabolic rate are also likely to occur at the network level. Specifically, with the onset of anoxia, some neurons may be activated in order to inhibit other neurons. We found evidence of this in anoxia-tolerant turtle cortical neurons, where glutamatergic pyramidal neurons undergo spike arrest as a result of an increase in inhibitory γ-aminobutyric acid type A (GABA_A) receptor currents (Pamenter et al., 2011). However, we have not recorded directly from the likely source of the increased GABA levels – stellate interneurons.

Stellate neurons are smaller than pyramidal neurons and are more difficult to patch clamp in the turtle cortex (Connors and Kriegstein, 1986); therefore, our present goal was to develop a brain slice model in which stellate neurons could be reliably patched to obtain recordings during a normoxic to anoxic transition. As the reversal potential (E_GABA) for the GABA_A receptor is depolarized relative to V_m in the turtle pyramidal neurons, we also set out to determine E_GABA in both goldfish stellate and pyramidal neurons. We hypothesized that, with the onset of anoxia, stellate neuron V_m will depolarize towards E_GABA, and AP_f will increase. Furthermore, we hypothesized that pyramidal neuron V_m will depolarize towards E_GABA, and whole-cell conductance will increase but AP_f will decrease.

This study was approved by the University of Toronto Animal Care Committee and conforms to the relevant guidelines issued by the Canadian Council of Animal Care regarding the care and use of experimental animals. Common goldfish, Carassius auratus (Linnaeus 1758), weighing 50–100 g were purchased from a commercial supplier (AQUAlity Tropical Fish Wholesale, Mississauga, ON, Canada) and held in flowing dechlorinated City of Toronto tap water at 18°C at the University of Toronto.

Whole goldfish brain was removed from the cranium following decapitation and the entire brain underwent Golgi staining. This technique reveals fine structures of neurons, glia and their processes, allowing a better understanding of cellular morphology along with their orientation and location in cortical slices (Melendez-Ferro et al., 2009). Tissue was fixed in 10% formalin for 2 days followed by fixation in 3% K₂Cr₂O₇ for 1 day and 2% AgNO₃ for another 2 days (Olude et al., 2015). The telencephalon was separated from the rest of the brain and sliced into 100 μm-thick axial slices using a Leica VT1200S vibratome (Leica, Richmond Hill, ON, Canada). Slices were collected using a soft brush, placed on a microscope slide, cover-slipped and stored at 4°C.

Telencephalon dissection, tissue preparation and whole-cell patch-clamp recording methods are described elsewhere (Wilkie et al., 2008). In short, the whole brain of the goldfish was removed from the cranium after decapitation and was placed in oxygenated ice-cold 4°C artificial cerebrospinal fluid (aCSF) containing (in mmol l⁻¹): 20 NaHCO₃, 118 NaCl, 1.2 MgCl₂.6H₂O, 1.2 KH₂PO₄, 1.9 KCl, 10 Hepes-Na, 10 d-glucose and 2.4 CaCl₂ (pH 7.6, adjusted with HCl); osmolarity 285–290 mOsm. Both lobes of telencephalon were then dissected from the rest of the brain while in the chilled aCSF. Each telencephalon lobe was glued to a sectioning plate on a vibratome slicing chamber using cyanoacrylate glue (Krazy Glue) and was filled with 4°C aCSF. Isolated goldfish telencephalon was cut into 300 µm slices using a VT1200S vibratome. Slices were lifted out of the chamber and were stored in vials of oxygenated aCSF for no longer than 48 h. Slices remained physiologically viable up to 48 h in the preparation (Wilkie et al., 2008). During each experiment, individual telencephalon slices were perfused with either oxygenated (99% O₂, 1% CO₂) or anoxic (99% N₂, 1% CO₂) aCSF (composition described above), and whole-cell recordings were performed using the voltage-clamp method with 9–11 MΩ borosilicate glass pipette electrodes (Harvard Apparatus Ltd, Holliston, MA, USA) containing the following (in mmol l⁻¹): 120 potassium gluconate, 10 KCl, 15 sucrose, 2 Na₂ATP, 0.44 CaCl₂, 1 EGTA and 10 Hepes-Na (pH 7.6 adjusted with KOH); osmolarity 285–290 mOsm. The electrode was connected to a CV-4 headstage (gain: 1/100 U, Axon Instruments, Sunnyvale, CA, USA) and an Axopatch 1D patch-clamp amplifier (Axon Instruments). A whole-cell patch configuration was established by voltage clamping the recording potential to −60 mV and applying negative pressure after a 1–20 GΩ seal was established. All seals were obtained using the blind-patch technique (Blanton et al., 1989). Whole-cell capacitance was compensated, typical whole-cell access resistance was 30–40 MΩ and series resistance was not greater than 50 MΩ. Access resistance was monitored periodically and data were discarded if it changed by 10–15 MΩ. Electrophysiological recordings were performed on pyramidal neurons and stellate neurons in goldfish telencephalon. The cerebral cortex of fish, amphibians and reptiles is characterized by three layers, in which the pyramidal neurons are excitatory glutamatergic cells and the stellate neurons are primarily GABAergic and inhibitory, providing feed-forward and feedback inhibition to pyramidal cells (Shepherd, 2011; Larkum et al., 2008). Electrophysiologically, the two cell types were differentiated in goldfish telencephalon by their responses to somatic current injections which produce trains of action potentials. If accommodation occurred quickly, the cell was distinguished as a pyramidal neuron and if no accommodation was observed, the cell was differentiated as a stellate neuron. Further, neural morphology was determined by filling the microelectrode solution with 5% Lucifer Yellow in 2 mol l⁻¹ LiCl. Lucifer Yellow was ejected by steady hyperpolarizing current for 5 min after completion of the experiment (Connors and Kriegsten, 1986; Shin and Buck, 2003). Spontaneous electrical activity was recorded for approximately 1 h (5–10 min normoxia, 30 min anoxia, 20 min recovery) from pyramidal and stellate neurons. Action potential threshold (AP_th) was determined by injecting current in a ramp manner in 1 nA steps until an AP was elicited. Whole-cell conductance (G_w) was measured by clamping neurons at voltage steps from −90 to −30 mV in 10 mV increments, lasting 200 ms each. Current amplitudes were measured between 135 and 145 ms and a slope conductance according to the current–voltage (I/V) relationship was determined (Pamenter et al., 2011).

We utilized the gramicidin perforated-patch technique to determine E_GABA without perturbing neural Cl⁻ gradients. Micropipette tips were front-filled with electrode recording solution by applying suction to the back of the pipette for 30 s, then back-filled with a solution of 13 µg ml⁻¹ gramicidin in DMSO dissolved in electrode solution. G_w recordings and I/V relationships were analyzed by taking the linear regression of each line before and during 2 mmol l⁻¹ GABA application. E_GABA was estimated by measuring the voltage at which the I/V relationships before and during GABA application intersected (Watanabe et al., 2009). Using the Nernst equation and the known extracellular [Cl⁻], we estimated the intracellular [Cl⁻] from E_GABA. E_GABA was calculated by deducting conductance values pre- and post-GABA application, and plotting the results on a I/V curve.

Goldfish telencephalon slices were perfused with pharmacological modifiers as specified in the Results through a fast-step perfusion system (VC-6 perfusion valve controller and SF-77B fast-step perfusion system; Warner Instruments, Hamden, CT, USA). Possible involvement of the GABAergic system in the anoxic response was investigated by application of GABA_A and GABA_B receptor antagonists, 25 μmol l⁻¹ SR95531 gabazine (GZ), 5 μmol l⁻¹ CGP-55848 (CGP) and 100 μmol l⁻¹ bicuculline (BIC) during anoxia. Changes in V_m and AP_f post-drug application were compared. AMPA and NMDA receptors were blocked using 25 μmol l⁻¹ of 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and 25 μmol l⁻¹ (2R)-amino-5-phosphonovaleric acid (APV), respectively. Voltage-gated sodium channels were blocked with 1 μmol l⁻¹ tetrodotoxin (TTX). E_GABA was obtained by perfusing cortical slices with 2 mmol l⁻¹ GABA. All chemicals were obtained from Sigma-Aldrich Canada Ltd (Oakville, ON, Canada) except CGP and TTX, which were obtained from Tocris Cookson (Bristol, UK).

Electrophysiological data were analyzed using SigmaPlot (version 11.0; Systat Software, Inc., San Jose, CA, USA). All data were tested for normality and equal variance and an analysis of variance (repeated measure ANOVA) was used for between-group comparisons where appropriate. Post hoc analyses consisted of Holm–Sidak or Tukey's test. Data are expressed as means±s.e.m.; n values represent a recording from a single cell from a single telencephalic slice of a single goldfish. P<0.05 was considered statistically significant.

Golgi staining and fluorescence imaging revealed that goldfish telencephalon neurons were located in two distinct regions. The first neural cell population was located at the lateral regions and cells were spread into distinct layers (Fig. 1A). For simplicity, we divided the cells into three types. Type one cells had a triangular shaped soma with dendritic fields extending away from the cell body in both apical and basal directions (Fig. 1B,D). Their dendritic extensions were large with medial projections towards the center of the slice. With application of a suprathreshold current pulse, these neurons showed rapid adaptation and fired distinct action potentials followed by hyperpolarization (Fig. 1F). Cells with these characteristics were identified in painted turtle cortex as pyramidal neurons (Connors and Kriegsten, 1986). Type two cells had a small (∼25 µm) and circular soma with bi-lateral projections extending away from the cell body (Fig. 1C). It appeared that these cells were more localized and had little branching on their dendritic tree (Fig. 1Eii). Type three cells had a more oval-shaped cell body, which was larger than type two cells (∼50 µm) (Fig. 1Ei). Similar to pyramidal neurons, they sent their projections medially and had large dendritic fields. Physiologically, both type two and type three cells showed little or no spike frequency accommodation in response to suprathreshold current. Action potentials were generated continuously and had equal amplitude (Fig. 1G). This physiological morphology has been associated with stellate neurons (Connors and Kriegstein, 1986).

The second population was part of the area dorsalis telencephali pars centralis located at the center of the slice and was isolated from the lateral region (Peter and Gill, 1975). The area dorsalis telencephali pars centralis contained both pyramidal and type three stellate neurons; however, no type two stellate neurons were observed. Perhaps because of their small size, type two stellate neurons were encountered only twice. Hence, data collected from these cells were excluded from the data set. Both pyramidal and stellate cells had large dendritic trees and were highly interconnected. In this experiment, the term ‘stellate neurons’ refers to the type three stellate cells.

Pyramidal neurons are the primary excitatory neurons in the turtle cortex and are the major consumers of ATP (Howarth et al., 2010). A reduction in excitability of these neurons is crucial to tolerance of anoxia and ischemia, and energy conservation (Pamenter et al., 2011; Hogg et al., 2015). Inhibition is not independent from the rest of the cortex and if GABAergic stimulation is responsible for suppression of neuronal excitability, the inhibitory stellate cells should show greater excitation with anoxia. Under passive recording conditions (I=0), it was observed that pyramidal neurons were mostly quiescent and V_m was maintained at −72.18±2.3 mV during normoxia. V_m depolarized to −57.7±1.6 mV after 30 min of anoxia (P=0.01, n=14) and repolarized to −64.22±2.6 mV after 20 min of normoxic recovery (P=0.08, n=14). Despite V_m depolarization, these cells did not generate an action potential (Fig. 2A,E,F). However, a small population of pyramidal neurons were more depolarized and active during normoxia (V_m=−60.5±1.6 mV, AP_f=0.51±0.03 Hz). In these cells, anoxic treatment caused a depolarization of V_m to −53±1.08 mV (P=0.04, n=4) and AP_f decreased to 0.1±0.06 Hz (P<0.001, n=4). With normoxic recovery, V_m hyperpolarized to −55±2.6 mV; however, this was not significantly different from anoxic V_m; AP_f increased to 0.24±0.1 Hz (P=0.003 from anoxia to recovery, n=4) (Fig. 2D,G,H). Similar to findings in pyramidal neurons, V_m depolarized from −64.54±1.05 mV to −46.86±1.2 mV and repolarized to −53±2.65 mV in stellate neurons during a normoxic to anoxic transition and recovery (P=0.02 for normoxia to anoxia and P=0.03 for anoxia to recovery, n=11). Unlike pyramidal neurons, stellate neurons generated action potentials with anoxia and AP_f increased from 0.03±0.009 Hz to 2.04±0.12 Hz (P<0.001, n=11). AP_f was reduced to 0.13±0.03 Hz with normoxic recovery (P<0.001, n=11) (Fig. 2B,E,F). Application of AMPA and NMDA receptor antagonists (25 μmol l⁻¹ CNQX and 25 μmol l⁻¹ APV, respectively) did not suppress action potentials; however, 1 μmol l⁻¹ TTX abolished them (data not shown). Furthermore, both pyramidal and stellate neurons showed an ∼60% increase in G_w with anoxia. These changes were from 1.75±0.34 nS to 2.83±0.17 nS in pyramidal neurons (P=0.04, n=7) and 1.27±0.15 nS to 2.01±0.16 nS in stellate cells (P=0.07, n=8) (Fig. 2G).

Activation of GABA_A receptors leads to a Cl⁻ flux in the direction of the reversal potential (E_Cl­) (Blaesse et al., 2009). As GABA_A receptors are predominantly permeable to Cl⁻, E_GABA and E_Cl should be similar (Errington et al., 2014). We measured E_GABA in goldfish telencephalic neurons to better understand the relationship between V_m and E_GABA. In order not to perturb neuronal Cl⁻ gradients, we used the gramicidin perforated-patch technique to determine E_GABA. Addition of 2 mmol l⁻¹ GABA during normoxia increased V_m in telencephalic neurons by 20±2.6 mV (n=8 for pyramidal neurons and n=8 for stellate neurons, P<0.001), which returned to baseline approximately 10 min post-GABA application (Fig. 3A). In pyramidal neurons, stepwise current injection generated action potentials in the absence of GABA but not in the presence of GABA (Fig. 3B). Similarly, stellate neurons showed suppressed activity with GABA application; however, with increased current injection, these neurons were capable of generating action potentials (Fig. 3C). Furthermore, G_w was recorded before and after GABA application and a I/V relationship was constructed. GABA increased G_w and the intersection between the two I/V curves was taken as E_GABA (Fig. 3D). E_GABA was more depolarized in stellate neurons than in pyramidal cells: −42.1±2.3 and −53.80±2.63 mV, respectively. By obtaining the E_GABA values and utilizing the Nernst equation, we were able to calculate intracellular [Cl⁻], which was ∼15.1 and ∼24 mmol l⁻¹ in pyramidal and stellate neurons, respectively. As demonstrated previously, G_w increased with anoxia from 3.96±0.95 nS to 6.84±0.61 nS (P<0.001, n=5). Addition of GABA appeared to increase G_w during normoxia; however, this increase was not significant (G_w=5.49±0.86 nS, P=0.3, n=5). With anoxia, G_w increased to 8.49±0.2 nS, which was significantly different from anoxia without GABA (P=0.02, n=5) and from normoxia with GABA (P<0.001, n=5) (Fig. 3E).

In order to investigate the possible role of GABA in the defense against anoxia-mediated damage in goldfish telencephalic neurons, we blocked GABA_A and GABA_B receptors under anoxic conditions. During anoxia, pyramidal neuron V_m hyperpolarized with application of the GABA_A receptor inhibitor GZ (25 μmol l⁻¹) from −56.54±1.7 mV to −64.77±1.9 mV and depolarized back to −55.15±1.36 mV with normoxic recovery (P<0.001, n=6) (Fig. 4A,D). In anoxic stellate neurons, GZ also hyperpolarized V_m, from −46.8±2.1 mV to −58.54±2.1 mV (P<0.001, n=5) (Fig. 4B,D) and AP_f decreased from 2.04±0.12 Hz to 0.155±0.01 Hz (P<0.001, n=5) (Fig. 4E). These changes were reversed with the termination of GZ application. The GABA_A receptor inhibitor BIC also hyperpolarized V_m to −68.82±1.7 mV and −56.8±1.97 mV in pyramidal and stellate neurons, respectively (n=3, P<0.001; data not shown). No significant difference was observed between BIC and GZ in reversal of the anoxic response. Even though the GABA_B receptor antagonist CGP alone did not change V_m or AP_f in either cell type (V_m=−53.97±2.24 mV for pyramidal and V_m=−47.74±2.83 mV, n=4), co-antagonism of GABA_A and GABA_B receptors with GZ plus CGP significantly influenced neuronal excitability (Fig. 4E). GZ+CGP initially hyperpolarized V_m similar to what was observed with GZ perfusion. However, shortly after, V_m depolarized significantly (V_m=−36.58±1.62, P<0.001, n=4, and −34.25±5.9 mV, P<0.001, n=5 for pyramidal and stellate neurons, respectively) and generated sudden, seizure-like action potentials (Fig. 4C,D). Furthermore, AP_f increased in pyramidal neurons from quiescent to 2.23±0.09 Hz (n=4, P<0.001); the increase in AP_f was not significantly different in stellate neurons (AP_f=2.13±0.02 Hz, n=5) (Fig. 4E). Even though AP_f drastically increased with GZ+CGP application, in some neurons (n=5) V_m hyperpolarized to the pre-anoxia level, neurons became quiescent and stepwise current injection failed to generate an action potential (data not shown). The rest (n=4), however, did not recover after GZ+CGP application and continued to generate seizure-like action potentials until loss of V_m, which is associated with neuronal death (Fig. 4C).

GABA_A/B receptor antagonism significantly affected G_w in both pyramidal and stellate neurons. Anoxia increased G_w from 3.96±0.95 nS to 6.84±0.62 nS (P<0.001). Blocking the GABA_A receptor with GZ had no significant effect on normoxic G_w (2.83±0.33 nS, P=0.071) but it did significantly block the anoxia-mediated increase (3.87±0.39 nS, P<0.001). Normoxic application of CGP alone did not have an effect on G_w (4.1±0.56 nS, P=0.81) and application of CGP alone during the anoxic period did not suppress the anoxia-mediated increase (5.43±0.59 nS, P=0.27). Co-application of GZ+CGP did not affect normoxic G_w (3.48±0.18 nS, P=0.43) but did significantly block the anoxia-mediated increase in G_w (3.7±0.26 nS, P<0.001). This suggests that the increase in conductance during anoxia is dependent on the activation of GABA_A receptors and that GABA is likely responsible for anoxic depolarization (Fig. 4F). G_w measurements were conducted on anoxic and normoxic controls and following treatment with GZ, CGP and GZ+CGP, each with 5 replicates. As the change in G_w was similar for the two neuronal types, the pharmacological responses were combined. Data are presented as the combined responses of pyramidal and stellate neurons.

In this study we investigated the role of GABAergic neurotransmission in the telencephalon of the naturally anoxia-tolerant goldfish, C. auratus. We show that anoxia depolarizes the V_m of both stellate and pyramidal neurons towards the GABA_A receptor reversal potential (E_GABA), which is reversed by application of a GABA_A receptor blocker. Interestingly, depolarization by increasing GABA_A receptor currents decreases pyramidal cell AP_f while it increases the AP_f of stellate neurons. Furthermore, we show GABA_B receptors are also important in anoxic survival as inhibition of both GABA_A and GABA_B receptors induce seizure-like activities in telencephalic neurons and apparent cell death. GABA is the primary inhibitory neurotransmitter in the mature mammalian central nervous system and an increase in the extracellular [GABA] seems to be a common response to anoxia in anoxia-tolerant organisms (Krnjevic, 1997; Nilsson and Lutz, 2004; Bickler and Buck, 2007). In the anoxia-tolerant turtle brain (Chrysemys picta bellii, Trachemys scripta elegans), [GABA] increases by 80-fold, suppressing whole-brain electrical activity by 75–95% (Lutz and Nilsson, 1997). This is strongly mediated by both GABA_A and GABA_B receptors, which are critical to anoxic survival of turtle cortical neurons (Pamenter et al., 2011). While turtles become quiescent and enter a stage of minimal activity, goldfish and crucian carp remain slightly active in anoxic waters (Bickler and Buck, 2007). Compared with levels in turtle brain, the anoxia-mediated increase in extracellular [GABA] in crucian carp brain is quite modest and doubles after 5 h of anoxia (Hylland and Nilsson, 1999). Furthermore, with anoxia, GABA removal from the extracellular space is significantly reduced because the expression of GABA transporter proteins decreases by 80%, increasing the effective extracellular [GABA] further (Ellefsen et al., 2009). We observed a significant increase in neuronal V_m with anoxia. This depolarization led to action potential propagation in stellate neurons while pyramidal neurons remained silent. An increase in AP frequency of stellate neurons seems counter to energy conservation within a low oxygen environment; however, inhibitory neurons consume significantly less energy than excitatory neurons and there are about 10-fold fewer (Howarth et al., 2010). In a study by Hylland and Nilsson (1999), running a high [K⁺] Ringer solution through a microdialysis probe to depolarize the surrounding tissue caused the extracellular [GABA] to increase 14 times, while extracellular [glutamate] doubled. This suggests that the potential for GABA release in crucian carp telencephalon is significantly greater than that for glutamate, resulting in greater GABA release in response to equal amounts of stimulation. If we consider the similarities between the telencephalon of teleost fish and mouse hippocampus, the number of excitatory synapses is approximately 18-fold greater than the number of inhibitory ones (Megías et al., 2001; Muller, 2012). Therefore, as a defense mechanism in goldfish telencephalic neurons, a small population of neurons with high release potential and lower energy requirements are activated to suppress the larger population of excitatory neurons and conserve energy.

In our experiment, blocking NMDA and AMPA receptors failed to suppress action potential firing in stellate neurons while TTX blocked action potentials completely. Therefore, we hypothesized that pathways other than excitatory glutamatergic signaling, such as GABAergic pathways, are responsible for V_m depolarization during anoxia. Indeed, blocking GABA_A receptors with GZ reversed anoxic depolarization of V_m and prevented the anoxia-mediated increase in G_w. Under physiological conditions, GABA_A receptors are highly permeable to both Cl⁻ and HCO₃⁻; however, the channels are considerably more permeable to Cl⁻ than to HCO₃⁻ (Cl⁻/HCO₃⁻ permeability: 5/1) and, thus, the majority of charge transfer following channel activation is Cl⁻ mediated (Errington et al., 2014). Therefore, we conclude that the anoxic increase in G_w is largely due to the movement of anions.

Measurements of E_GABA revealed that V_m shifts towards a depolarizing E_GABA with anoxia and induces action potentials in stellate neurons but not in pyramidal cells. This is consistent with a previous report showing exogenous GABA can depolarize goldfish telencephalic gonadotropin-releasing hormone (GnRH) neurons (Nakane and Oka, 2010). By changing V_m to E_GABA in pyramidal neurons, the potential difference between the intracellular and extracellular environments is decreased, which significantly reduces energy in the form of ATP required to maintain ionic gradients (Edwards et al., 1989). Although GABA is depolarizing, it remains inhibitory because it activates an inhibitory shunt that maintains V_m below the threshold potential (Pamenter et al., 2011). E_GABA is primarily determined by the Cl⁻ gradient where K⁺/Cl⁻-cotransporter 2 (KCC2) and Na⁺/K⁺/Cl⁻-cotransporter 1 (NKCC1) set the reversal potential and driving force of anion currents mediated by GABA_A receptors (Blaesse et al., 2009). For GABA to have a depolarizing effect, the intracellular [Cl⁻] must be relatively high (20–25 mmol l⁻¹) so that an efflux of Cl⁻ occurs via GABA_A receptors (Ben-Ari, 2002). We found that intracellular [Cl⁻] is relatively high in goldfish telencephalic neurons and approximately 1.5 times greater in stellate neurons than in pyramidal neurons. As a result, activation of GABA_A receptors leads to a greater efflux of Cl⁻ and further depolarization in stellate neurons compared with pyramidal neurons. Cl⁻ flow occurs through both synaptic (phasic) and extrasynaptic (tonic) GABA receptors (Blaesse et al., 2009). Localization of GABA_A receptors has not been investigated in goldfish neurons; however, crucian carp neurons express the gamma subunit, which is necessary for phasic currents, and delta subunits, which in mammalian brain are extrasynaptic (Mody, 2001; Ellefsen et al., 2009). By applying GZ, we likely inhibited the activity of phasic currents as GZ has a greater affinity for synaptic GABA_A receptors (Mody, 2001). Application of BIC hyperpolarized V_m of both pyramidal and stellate neurons and reduced AP_f in stellate neurons similar to GZ. However, BIC blocks both phasic and tonic GABA_A receptors (Ueno et al., 1997), suggesting that the anoxic response is primarily mediated through the activity of the synaptic receptors. Anoxia-tolerant turtle brain differs from goldfish brain as anoxia induces large phasic GABA currents and a 50% increase in tonic current (Hogg et al., 2015). Large phasic GABAergic currents were only observed in a single goldfish pyramidal neuron recordings (1/47 attempts). Perhaps formation of such currents requires an intact circuitry or is localized to a specific cortical region. As our electrophysiological recordings were from brain slices, the circuitry might have been disrupted (Dukoff et al., 2014; Hogg et al., 2015). Furthermore, all the recordings utilized the blind-patch technique, which prevents clear localization of the areas populated with neurons that induce large GABA-mediated currents (Blanton et al., 1989). As a result of such limitations and the inability to frequently record GABA-induced currents, it is not possible to state to what degree tonic GABA_A receptors are involved in the anoxic response in the goldfish brain.

Blocking GABA_B receptors alone did not affect V_m or AP_f and changes in G_w were not as strong as those during GABA_A receptor inhibition. These receptors are slow acting (100–500 ms) G-protein coupled inwardly rectifying potassium channels (GIRKs) and mediate prolonged neural inhibition by increasing K⁺ efflux via K⁺ channels post-synaptically and inhibiting voltage-gated Ca²⁺ channels pre-synaptically (Gassmann and Bettler, 2012). GABA_B receptor activity is strongly mediated by intracellular messenger mechanisms (Kuramoto et al., 2007). With anoxia or a shortage of ATP, certain intracellular signaling proteins, such as AMP-dependent protein kinase (AMPK), are activated due to an increase in [AMP] (Carling, 2004). AMP has a critical role in maintaining cellular and systemic energy balance and promoting glycolysis (Stensløkken et al., 2008). Through phosphorylation, AMPK strongly enhances GABA_B receptor activity, allowing a greater K⁺ efflux via GIRKs (Kuramoto et al., 2007). Anoxic exposure significantly increases AMPK phosphorylation in crucian carp brain (Stensløkken et al., 2008).

In this study, we were able to show that inhibition of both GABA_A and GABA_B receptors leads to sudden, uncontrollable electrical activity and a paroxysmal depolarizing shift. The paroxysmal depolarizing shift is a manifestation of neural epilepsy that is in contrast to endogenous burst firing as it is network driven (Johnston and Brown, 1984). In some neurons, V_m did not return to baseline and seizure-like activities persisted. In others, V_m hyperpolarized to a pre-anoxic stage after hyperactivity and cells became quiescent. This does not necessarily imply that these neurons recovered from seizure as this can occur as a result of a depolarization block in which seizure can be maintained even in the absence of continuous neuronal firing (Bikson et al., 2003). While co-antagonism of GABA_A and GABA_B receptors induced paroxysmal depolarizing shift, blocking either GABA_A or GABA_B receptors alone did not cause catastrophic depolarization. GABA_A and GABA_B receptors have a complementary role in reducing neural excitation (Mann et al., 2009). GZ significantly reduced conductance and reversed the anoxic depolarization; however, GABA_B receptors might prevent the release of excitatory neurotransmitters and excitotoxic cell death even when GABA_A receptors are inactive. In the anoxic turtle brain, GABA_B receptors predominantly act pre-synaptically and decrease excitatory post-synaptic potentials (EPSPs), especially AMPAergic EPSPs (Pamenter et al., 2011). Furthermore, we show that inhibiting GABA_B receptors alone does not significantly change the anoxic G_w while blocking GABA_A receptors does. This suggests that the major change in conductance brought on by anoxia must occur through GABA_A receptors and could explain why application of CGP alone causes no catastrophic depolarization. Activity of GABA_B receptors may be crucial to neural survival and limiting excitotoxicity, especially if GABA_A receptors are inactive, perhaps by decreasing EPSPs and reducing neurotransmitter release (Kuramoto et al., 2007; Pamenter et al., 2011; Tao et al., 2013).

In conclusion, we report that in anoxic goldfish telencephalic stellate neurons, AP_f increases while pyramidal neuron AP_f or the potential to generate action potentials is decreased. In both neuronal types, V_m depolarizes towards the E_GABA but because E_GABA is more depolarizing in stellate neurons, it is responsible for the increased activity. Inhibiting both GABA_A and GABA_B receptors induces terminal depolarization of goldfish neurons, highlighting the importance of GABA in anoxia tolerance.

The authors would like to thank Dr Kaori Takehara-Nishiuchi and Dr Martin Wojtowicz for their valuable insights and critiques during every stage of this study.