Brain slices are widely used to investigate basic processes of brain function. Although being a reduced preparation (i.e., there is no blood flow, oxygen levels are non-physiological, most in vivo metabolites are not present in the artificial cerebrospinal fluid), brain slices provide easier access to cellular phenomena than in vivo models. Many results obtained in vitro (and reproduced by different laboratories) have been verified in vivo, giving ground to the general thought that in vitro results can be generalized to the intact organism. However, although adequate in many cases, this approach may lead to misinterpretation in many others. The concept of the excitatory action of GABA at early postnatal stages of development provides a particular example of correct observations performed in vitro which may not apply to the in vivo situation.

GABA, the main inhibitory neurotransmitter in vertebrates, activates GABA_A receptors (GABA_AR) resulting in opening of anion-selective channels and transmembrane fluxes of chloride (Cl) and bicarbonate. Normally, the direction of Cl current determines the hyperpolarizing or depolarizing effect of GABA_AR activation on the membrane. If the reversal potential for Cl (E_Cl) is above (below) the resting membrane potential, Cl leaves (enters) the cell. An outward (inward) flux of negative charges depolarizes (hyperpolarizes) the membrane.

It is important to clarify here the difference between depolarizing and excitatory actions of GABA since there is a widespread misunderstanding of these notions. The concentration of intracellular Cl⁻ measured in different cell types varies from 3 to 60 mM and in mammalian neurons in vitro it is generally low (<10 mM, see Khirug et al., 2008; Bregestovski et al., 2009). As a result, the reversal potential of GABAergic currents, E_GABA, is close to the resting membrane potential and activation of GABA_AR causes hyperpolarization or weak depolarization. Meanwhile, GABA_AR channel opening decreases the input membrane resistance inducing “shunting inhibition” (see Andersen et al., 1980; Staley and Mody, 1992; Tang et al., 2011; Wright et al., 2011) that lowers the neuron’s firing probability. Therefore, a weakly depolarizing GABA may exert an inhibitory effect. In contrast, the “excitatory” GABA action means that GABA_AR activation induces a depolarization large enough to generate action potentials.

The inhibitory/hyperpolarizing effects of GABA have been extensively verified in juvenile and adult animals in vivo. At earlier stages of development, the picture appears to be different. In vitro experiments have shown an excitatory action of GABA at early stages of development in kittens (Schwartzkroin and Altschuler, 1977), rabbits (Mueller et al., 1983), and rats (Dunwiddie, 1981; Harris and Teyler, 1983; Mueller et al., 1984; Ben-Ari et al., 1989) in a large number of subsequent studies (for review, Ben-Ari et al., 2007). Experiments performed in rodent brain slices indicated that the switch from the excitatory to inhibitory action of GABA takes place during the second postnatal week (P12–P13; Ben-Ari et al., 2007). The mechanism of this switch was explained as the increased age-dependent expression of KCC2 chloride exporter which takes over the leading role in Cl homeostasis from NKCC1 chloride importer (Blaesse et al., 2009). A hypothesis on the leading role of excitatory GABA in development was proposed by Ben-Ari and co-authors who claimed it as a universal rule: “In all developing animal species and brain structures investigated, neurons have a higher intracellular chloride concentration at an early stage leading to an efflux of chloride and excitatory actions of GABA in immature neurons” (Ben-Ari et al., 2007). These in vitro findings obtained in brain slices or cell cultures were frequently taken for granted. However, several lines of evidence challenge the extrapolation of these conclusions to the intact brain.

First, the early study performed in vivo, using intracellular recordings of hippocampal neurons in young kittens, suggested that inhibition is a predominant form of synaptic activity at early postnatal ages (Purpura et al., 1968). However, a high concentration of KCl was used in the pipette solution, which can alter ionic homeostasis.

Second, in vivo recordings using GABA_AR antagonists contradict the in vitro observations. A study based on the analysis of more than 200 rat pups at the age of P3–P5 demonstrated that the injection of bicuculline triggered seizures in these pups (Baram and Snead, 1990). Another in vivo study reported that cerebellar Purkinje cells inhibit each other as early as at P5 and that bicuculline abolishes their interaction and increases their spontaneous firing activity (Bernard and Axelrad, 1993). Also, several more recent in vivo studies using specific agonists or antagonists of GABA_ARs clearly demonstrated the inhibitory action of GABA during the first postnatal week (Minlebaev et al., 2006, 2011; Isaev et al., 2007). For instance, Minlebaev et al. (2006) wrote that in P3–P5 rats: “Blockade of GABAA receptors by gabazine significantly increased spontaneous cortical activity by almost doubling the occurrence of spontaneous spindle-bursts…” However, these results were not mentioned in the subsequent review by the same main authors (Ben-Ari et al., 2007), who instead claimed that GABA “…excites immature neurons and generates primitive oscillations.” It is difficult to state that GABA exerts an excitatory action when GABA_AR blockade leads to an increased activity in vivo.

Third, observations on the “intact hippocampus” preparation (in toto) where cellular integrity and connectivity are maintained, also suggest the inhibitory action of GABA. Using recordings from the CA1 area in isolated hippocampus, Wong et al. (2005) showed that synaptically released GABA causes inhibition. Moreover, in contrast to observations made in brain slices (Figure 1A; Ben-Ari et al., 1989), application of bicuculline resulted in epileptiform discharges (Figure 1C; Wong et al., 2005). Interestingly, similar effects were observed by Ben-Ari’s group in the very first study on the intact immature hippocampus (Figure 1B; Khalilov et al., 1997), but they were not discussed in their later publications. Recent experiments using the same preparation from P5–P7 mice confirmed these observations (Dzhala et al., 2010, 2012). Isoguvacine, a selective agonist of GABA_ARs, transiently reduced spontaneous neuronal activity. Thus, the net effect of GABA_AR activation in the intact hippocampal network is inhibitory.

Together, these results strongly suggest that GABA is inhibitory in the immature intact brain. On the other hand, the excitatory action of GABA has been observed in a number of studies on brain slices (Ben-Ari et al., 2007). What mechanisms may underlie this apparent discrepancy?

Using brain slices implies that brain tissue will be cut, i.e., that cell processes (dendrites, axons etc.) will be severed, generating a model of traumatic brain injury. According to early histological observation in slices, there is a 50- to 100-μm deep zone of severely disrupted tissue (Garthewaite et al., 1979; Bak et al., 1980; Frotscher et al., 1981). As a consequence of mechanical injury, microglial cells in slices are rapidly activated and become highly mobile (Petersen and Dailey, 2004). This may trigger a cascade of detrimental processes due to the release of a number of neurotoxic substances including cytokines, chemokines, nitric oxide, and superoxide free radicals that generate reactive oxygen species and reactive nitrogen species (Loan and Byrnes, 2010).

While in more recent studies microtomes/vibratomes are used for slices preparation, still the regions close to the surface (30–80 μm deep) contain a large amount of damaged cells (Dzhala et al., 2012). Since most electrophysiological and imaging studies of cell body layers (like hippocampal pyramidal cells) are performed in this region, the results may be biased by the inclusion of these injured cells, thus reflecting pathological rather than physiological processes. Indeed, slicing through brain tissue invariably leads to pathological reorganizations (Hoffman et al., 1994; McKinney et al., 1997).

As mentioned above, the net action of GABA_AR activation depends upon E_Cl. Hence, the depolarizing action of GABA in slices may result from intracellular Cl accumulation in traumatized neurons located close to the surface. Indeed, after neuronal trauma, GABA, both synaptically released and exogenously applied, induced depolarization of neurons, and increased intracellular Ca²⁺ (van den Pol et al., 1996). Using gramicidin perforated-patch recordings, Nabekura et al. (2002), demonstrated that E_GABA was more depolarized in axotomized than in intact neurons of the vagus dorsal motor nucleus. The authors concluded that: “axotomy led to …elevation of intracellular Cl, and an excitatory response to GABA. A switch of GABA action from inhibitory to excitatory might be a mechanism contributing to excitotoxicity in injured neurons” (Nabekura et al., 2002). Direct non-invasive measurements of intracellular Cl concentration in Clomeleon-expressing mice (Dzhala et al., 2010, 2012) clearly demonstrated that axotomized and dendrotomized cells proximal to the slice surface have a much higher intracellular Cl concentration than in deeper situated and less injured cells (Figure 2A). In contrast, Cl levels were much lower in the intact hippocampus preparation (Figure 2A), in which a direct activation of GABA_AR decreased neuronal firing – an observation consistent with an inhibitory/shunting action of GABA (Dzhala et al., 2012).

Finally, it is important to note that the intracellular Cl concentration may be cell type-dependent (Rohrbough and Spitzer, 1996; Sauer et al., 2012) and location-dependent in a given cell (Duebel et al., 2006). An uneven distribution of Cl ions has been described in hippocampal neurons using electrophysiological recordings (Szabadics et al., 2006; Khirug et al., 2008) and non-invasive monitoring of intracellular Cl (Waseem et al., 2010). Future studies on GABA action in the immature brain should take these factors into account.

Thus, the slicing procedure is clearly associated with damaged cells, which accumulate chloride. Slice quality critically depends upon the slicing procedure and equipment. Recent studies described conditions for better preparation (with microtomes/vibratomes) and preservation of acute slice preparations (Schurr et al., 1989; Hájos and Mody, 2009; Hájos et al., 2009; Maier et al., 2009; Ivanov and Zilberter, 2011). Still, even state-of-the-art procedures do not prevent damage inherent to slicing. For instance, using a vibratome, Taylor et al. (1999) wrote: “Light microscopy of slices fixed immediately after Vibroslice preparation indicated significant swelling of pyramidal neurons, i.e., cell bodies, mitochondria, dendrites, and nuclei were enlarged and hydropic.” While experimentators try to achieve recovery as much as possible after slicing (Taylor et al., 1999; Bischofberger et al., 2006), even after 1.5 h incubation in artificial cerebrospinal fluid (ACSF; typical experimental procedure for recovery of slice integrity) neurons and glial cells are still functionally and energetically defective. This point is supported by the observations of Dzhala et al. (2012) who demonstrated Cl accumulation in slice surface-proximal neurons (Figures 2A,B).

Abnormalities induced by tissue trauma in brain slices are exacerbated by several additional factors. The lack of blood flow in slices dramatically changes the way energy substrates and oxygen are delivered to cells. Energy substrates and O₂ are instead supplied exogenously by artificial extracellular solution (ACSF), which must diffuse passively from the surface. In the intact brain, blood vessels, astrocytes, and neurons form a complex system supporting and adjusting brain metabolism (Pellerin, 2010; Turner and Adamson, 2011; Zilberter and Bregestovski, 2012) while in brain slices metabolism depends entirely on the experimental conditions. Although experimentalists are trying to create conditions maximally close to the in vivo environment, they are obviously far from ideal. Support normally provided by blood is not entirely compensated by perfusion of ACSF. Glucose-based composition of ACSF was empirically adjusted more than 60 years ago for relatively long-lasting preservation of neuronal function in brain slices and is, obviously, not physiological (Hájos and Mody, 2009; Zilberter et al., 2010). Slices exposed to ACSF exhibit severe abnormalities in energy metabolism. For instance, the rate of glycolysis is reduced by more than 50% in brain slices (Rolleston and Newsholme, 1967; Benjamin and Verjee, 1980) as compared to the in vivo estimates (Ghajar et al., 1982). In addition, the total adenine nucleotide pool is decreased by 30–50% in slices as compared to that observed in vivo (Whittingham et al., 1984) and this effect become less important with increasing of slice thickness (Zur Nedden et al., 2011). Remarkably, the slicing procedure causes a decrease to about 50% of the total content of ATP, creatine, and adenylate, as well as a strong change in intracellular pH from about 6.6–7.2 (Whittingham et al., 1984). Such a deficit in the cell energy supply may directly affect GABAergic action.

To test this hypothesis, Zilberter and collaborators analyzed whether improving energy supply to neurons with glucose oxidative energy substrates (OES) can modulate the response to GABA. In neocortical and hippocampal slices from neonatal (P3–P8) rats and mice, supplementing ACSF with β-hydroxybutyrate, lactate, or pyruvate significantly hyperpolarized E_GABA, switching the GABA action from excitatory to inhibitory (Holmgren et al., 2010). Moreover, OES inhibited giant depolarizing potentials (GDPs; Holmgren et al., 2010; Mukhtarov et al., 2011), a spontaneous network activity pattern characteristic for neonatal hippocampal slices (Ben-Ari et al., 2007). The beneficial effect of OES on energy metabolism status in neurons was confirmed by direct simultaneous measurements of oxygen consumption and NADH fluorescence during neuronal activity (Ivanov and Zilberter, 2011; Ivanov et al., 2011). For instance, in the presence of glucose, lactate was effectively utilized as an energy substrate (Ivanov et al., 2011), causing an augmentation of oxidative metabolism (Figure 1D). Moreover, in the absence of glucose, lactate was fully capable of maintaining synaptic function (Schurr et al., 1988; Ivanov et al., 2011). These observations demonstrate that neuronal function can definitely be improved in both neonatal (Ivanov et al., 2011) and adult (Ivanov and Zilberter, 2011) slices by supplementing glucose with OES. Glucose alone, even at strongly hyperglycemic concentrations as in standard ACSF (10 versus 1–2 mM in the brain extracellular fluid (Abi-Saab et al., 2002; Zilberter et al., 2010) cannot fully cover energy demands during neuronal activation.

These studies have ignited a controversy (Kirmse et al., 2010; Ruusuvuori et al., 2010; Tyzio et al., 2011). However, although Tyzio and co-authors failed to reproduce the effects of b-hydroxybutyrate on E_GABA, they did reproduce the E_GABA-hyperpolarizing effect of 5 mM pyruvate. Kirmse et al. (2010) did not find any effect of β-hydroxybutyrate or pyruvate on GABA-induced Ca²⁺ fluorescent transients; but measurements for control and BHB-treated cells were performed on different slices with a slow ACSF perfusion rate leading to improper oxygenation (see Ivanov et al., 2011; Ivanov and Zilberter, 2011). Ruusuvuori et al. observed the inhibitory effect of lactate on GDP generation but suggested that this effect is induced by intracellular acidification. Indeed, OES caused pH_i changes of less than −0.05 pH units (Ivanov et al., 2011; Mukhtarov et al., 2011). However, the 0.25–0.35 reduction in pH_i obtained by substituting bicarbonate-containing solution with HEPES-based HCO₃-free solution did not eliminate GDPs (Mukhtarov et al., 2011). Therefore, a significant contribution of pH_i to the effects of OES on GDPs is unlikely (Ivanov et al., 2011; Mukhtarov et al., 2011). Certainly, the controversy needs to be resolved by independent groups. But the results clearly demonstrate that metabolic processes are central to the reorganization of cell function after making brains slices.

Altogether, these observations demonstrate that the slicing procedure injures cells and disrupts brain metabolism, leading to intracellular Cl accumulation in neurons and rendering GABA strongly depolarizing or even excitatory as has been reported during the first postnatal week in rodents.

This, however, does not rule out the possibility that GABA may be depolarizing, in particular at very early stages of development. For example, treatment of mice with bumetanide during the period of embryonic cortical developmental results in disruption of excitatory synapse formation (Wang and Kriegstein, 2011). As bumetanide antagonizes the Na⁺–K⁺–2Cl⁻ cotransporter (NKCC1), which accumulates intracellular Cl, these observations suggest that Cl in embryonic neurons is elevated and plays an important signaling role in developmental processes.

Oscillations/correlated neuronal discharges are a hallmark of network activity at any stage of development (Buzsáki, 1986, 2002; Spitzer, 1994; Chrobak and Buzsáki, 1998; Leinekugel et al., 2002; Khazipov et al., 2004; Adelsberger et al., 2005; Sipilä et al., 2006). At early stages of development, this synchronized activity may be important for brain maturation, regulating multiple processes including neuronal migration (Komuro and Rakic, 1998) and directing neuronal differentiation (Gu and Spitzer, 1997; Spitzer et al., 2000), dendritic growth and patterning (Katz and Shatz, 1996; Wong and Ghosh, 2002), activation of transmitter receptors (Liao et al., 2001), and the pattern of specific connections (Penn et al., 1998). The most prominent synchronized activity, early network oscillations (ENOs) associated with changes in neuronal intracellular Ca²⁺ concentration, were observed in small groups of neurons and in large populations in vitro (Garaschuk et al., 1998, 2000; Corlew et al., 2004) and in vivo (Adelsberger et al., 2005). Spindle-bursts were described in the neonatal rat neocortex in vivo (Khazipov et al., 2004). Thus, waves of spontaneous electrical activity propagating across many regions of the brain are a hallmark of developing networks, and actively contribute to cortical development and plasticity (Katz and Shatz, 1996; Mizuno et al., 2007). Distinct mechanisms underlie generation of synchronized events, including synaptic interaction, gap junction communication, the presence of pacemaker-like neurons as well as activation of metabotropic glutamate and ACh receptors (Kandler and Katz, 1998; Flint et al., 1999; Blankenship and Feller, 2010).

However, the reports that GABA is depolarizing/excitatory in slices from the immature brain led to a very popular theory, which inspired many researches in the neurodevelopment field and provided a conceptual framework to explain early network activities recorded in vivo. Excitatory GABA (i.e., its ability to drive the membrane potential to firing threshold) would be essential for developing networks. In vitro experiments revealed the occurrence of spontaneous network events involving large populations of neurons. This phenomenon was first described by Harris and Teyler (1983) who called it “spontaneous unison firing.” It was also observed by Mueller et al. (1984), who wrote: “Immature neurons often demonstrated spontaneous depolarizations of up to 30 mV amplitude and 30 to 60 sec duration.” Several years later, Ben-Ari et al. (1989) also described this phenomenon in immature brain slices, which they named GDPs. GDPs were infrequent or absent after P12. It was proposed that depolarizing GABA plays a key role in the generation of GDPs and that this spontaneous activity results from the synergistic excitatory activities mediated by GABA_A and glutamate N-methyl-d-aspartate (NMDA) receptors (Ben-Ari et al., 1997). Since GDPs were not observed after postnatal days 10–11, at the time close to the “excitation/inhibition switch,” it was postulated that GDPs represent a primitive activity pattern of the developing brain and that it is “largely based on excitatory GABA” (Ben-Ari et al., 2007).

These observations led to the broadly accepted idea that the excitatory action of GABA underlies neuronal maturation of immature neuronal networks. According to this concept, the elevated Cl concentration and, consequently, the excitatory action of GABA, represent necessary steps in the development of the nervous system. This viewpoint is epitomized in the recent review of van Welie et al. (2011), who wrote: “Depolarizing GABA is required for normal brain development, as it contributes to the morphological maturation of neurons (Cancedda et al., 2007) and neuronal circuits (Ben-Ari, 2001; Akerman and Cline, 2006). Depolarizing GABA can drive juvenile neurons to fire action potentials (Ben-Ari, 2002) and conversely, neuronal activity can regulate E_GABA, by either specific patterns of synaptic activation (Woodin et al., 2003; Balena and Woodin, 2008), or alterations in postsynaptic activity levels via changes in intracellular Ca²+ (Fiumelli et al., 2005).”

This statement relies on the axiom that the nature of GDPs observed in brain slices correlates with network activities recorded in vivo in developing networks. While the general patterns of this activity may be similar in vitro and in vivo, the underlying mechanisms may be different. The presence and character of oscillatory activity in brain slices highly depend upon energy support, oxygenation, and perfusion rate (Hájos and Mody, 2009; Hájos et al., 2009; Holmgren et al., 2010; Mukhtarov et al., 2011). For instance, sharp wave (SPW) oscillations are a hallmark of hippocampal activity in developing and adult hippocampus in vivo (Leinekugel et al., 2002). SPWs are usually not observed or very infrequent in slices when using slow perfusion rates of ACSF (1.6–2.4 ml/min; Hájos et al., 2009; Maier et al., 2009). However, SPWs appear (or become more frequent) at high speed of perfusion (Figure 2C), suggesting that a proper delivery of oxygen to the whole slice is critical for the genesis of SPWs in vitro (Hájos et al., 2009). The importance of oxygen delivery at elevated flow rates was further demonstrated by Ivanov et al. (2011). A decrease from 15 to 3.25 ml/min in the perfusion rate resulted in strong decrease of oxygen and a two-fold reduction of the local field potential amplitude in brain slices from P6 mice (Figures 2D,E).

Particularly convincing arguments were obtained in a recent study demonstrating that while GDPs can be recorded both in slices and the intact hippocampus during the first postnatal week, the mechanism of their genesis is different (Dzhala et al., 2012). Isoguvacine application dramatically increased GDP frequency in brain slices (in keeping with the excitatory action of GABA); whilst in the intact hippocampus isoguvacine completely abolished GDPs (in keeping with the inhibitory action of GABA).

Since the slicing procedure also lesions superficial neurons that leads to Cl accumulation in mature networks (Dzhala et al., 2012), one would expect GDPs to occur in adult slices. However, the study by Dzhala et al. (2012) shows that, whilst superficial neurons remain connected to the network in immature slices, they are functionally disconnected in mature slices. Hence, superficial cells with high internal Cl do not contribute much to network activity in mature slice.

Together, these observations strongly suggest that ENOs do not rely upon excitatory GABA. Hence, the mechanistic insights regarding GDP genesis/propagation/function gained from slice studies should be re-evaluated. As underlined in the recent review: “Usage of brain slice preparations has significantly contributed to a deeper understanding of neuronal functions at the cellular and network level in the recent decades. However, given factors such as absence of blood circulation, longer diffusion distances, steep interstitial pO2 gradients, and composition of the recording solution have to be kept in mind when interpreting data from slice preparations” (Kann, 2011).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.