Tonic release of glutamate by a DIDS-sensitive mechanism in rat hippocampal slices

Abstract

Tonic release of glutamate into the extracellular space of the hippocampus and striatum is non-vesicular, and has been attributed largely to a cystine–glutamate exchanger which is blockable by the glutamate analogue (S)-4-carboxyphenylglycine (CPG). Tonic glutamate release may be functionally important: modulation of this release in the striatum has been suggested to underlie relapse in the use of cocaine. We monitored tonic glutamate release in area CA1 of hippocampal slices by measuring the glutamate receptor-mediated current evoked in pyramidal cells on block of Na⁺-dependent glutamate uptake with dl-threo-β-benzyloxyaspartate (TBOA). Superfused cystine increased tonic glutamate release, and this increase was blocked by CPG, but CPG did not affect tonic glutamate release in the absence of superfused cystine. Tonic glutamate release was not affected by blocking gap junctional hemichannels with 18α-glycyrrhetinic acid, blocking ATP receptors with pyridoxal-phosphate-6-azophenyl-2′,4′-disulphonic acid (PPADS), blocking Ca²⁺-dependent exocytosis from neurones with Cd²⁺ or bafilomycin, blocking Ca²⁺-dependent release from glia with indomethacin, or blocking anion channels with 5-nitro-2-(3-phenylpropyl amino) benzoic acid (NPPB) or tamoxifen. However tonic glutamate release was reduced by 4,4′-diisothiocyanostilbene-2,2′-disulphonic acid (DIDS), and was potentiated by inhibiting astrocytic conversion of glutamate to glutamine with methionine sulfoximine. These data suggest that although cystine–glutamate exchange is present in the hippocampus it does not generate significant tonic release of glutamate when the extracellular [cystine] is at a physiological level, and that tonic glutamate release is at least partly from astrocytes and is mediated by a DIDS-sensitive mechanism. Theoretical calculations suggest that a significant fraction of tonic glutamate release in hippocampal slices could occur via diffusion of glutamate across lipid membranes.

In the brain, a low extracellular concentration of neurotransmitter is normally maintained between times of exocytotic transmission, in order to prevent continual activation or desensitization of receptors and, for the excitatory transmitter glutamate, to avoid excitotoxicity (reviewed by Cavelier et al. 2005). However, tonic activation of receptors by the ambient transmitter concentration has been reported for high affinity GABA_A receptors in the cerebellum and hippocampus, where it decreases excitability and alters information flow through the tissue (Kaneda et al. 1995; Tia et al. 1996; Brickley et al. 1996; Wall & Usowicz, 1997; Stell & Mody, 2002; Hamann et al. 2002; Rossi et al. 2003; Semyanov et al. 2003), and for NMDA and metabotropic glutamate receptors in hippocampus, where it increases excitability and modulates presynaptic release of transmitter (Sah et al. 1989; Forsythe & Clements, 1990; McBain et al. 1994; Dalby & Mody, 2003).

What generates this tonic activation of receptors? In the absence of glutamate release, the ionic stoichiometry of glutamate transporters provides sufficient accumulative power to lower the extracellular glutamate concentration, [glu]_o, to ∼ 2 nm (Zerangue & Kavanaugh, 1996; Levy et al. 1998), but microdialysis experiments in vivo report a [glu]_o of ∼ 2 μm (Benveniste et al. 1984; Hagberg et al. 1985; Phillis et al. 1994; Wahl et al. 1994) suggesting a constant release of glutamate. Surprisingly, this tonic glutamate release does not reflect either action potential evoked or spontaneous exocytosis of glutamate since, in cultured hippocampal slices, it is not blocked by TTX nor by blocking exocytosis with botulinum A or tetanus neurotoxin (Jabaudon et al. 1999).

Baker et al. (2002) have suggested, from microdialysis experiments, that most (60%) of the tonic glutamate release in the striatum is generated by cystine–glutamate exchange (in which glutamate is released in exchange for cystine taken up to make glutathione: Bannai, 1984; Cho & Bannai, 1990; Warr et al. 1999). Further, Baker et al. (2003) proposed that a reduction of cystine–glutamate exchange after withdrawal from cocaine use causes relapse into drug-seeking behaviour, highlighting the possible functional significance of this tonic release. However, cystine–glutamate exchange requires a high [cystine]_o (its EC₅₀ for cystine is ∼ 100 μm: Warr et al. 1999), while in the extracellular space of the striatum and in the CSF [cystine]_o is only around 0.13 μm (Perry et al. 1975; Baker et al. 2003) and will activate cystine–glutamate exchange to less than 0.2% of its maximum rate. Apart from cystine–glutamate exchange, there are several other unusual modes of transmitter release that are also candidates for mediating the tonic release of glutamate. In astrocytes, a rise of [Ca²⁺]_i can release glutamate by a prostaglandin-dependent mechanism (Bezzi et al. 1998), and three distinct ion channels have been shown to release glutamate: swelling-activated anion channels (Rutledge et al. 1998), gap junctional hemichannels (Contreras et al. 2002; Ye et al. 2003), and P2X₇ receptors gated by ATP (Sperlagh et al. 2002; Wang et al. 2002; Duan et al. 2003). In this paper we have re-investigated the possible contribution of cystine–glutamate exchange to setting the ambient [glu]_o, and tested the role of these novel astrocyte release mechanisms.

Methods

Preparation of brain slices

Coronal brain slices (225 μm) containing the hippocampus were obtained from Sprague-Dawley rats (11–13 days old), killed by cervical dislocation in accordance with the UK Animals (Scientific Procedures) Act 1986. The slices were made in an oxygenated (95% O₂–5% CO₂) solution at 4°C containing (mm): 126 NaCl, 2.5 KCl, 2 MgCl₂, 26 NaHCO₃, 1 NaH₂PO₄, 2.5 CaCl₂, 10 glucose and 1 sodium kynurenate (to block glutamate receptors), pH 7.4; they were stored for 1 h at 34°C before they were used for the experiments.

Electrophysiology

CA1 pyramidal cells were whole-cell voltage-clamped with a pipette solution containing (mm): 135 CsCl, 4 NaCl, 0.5 CaCl₂,10 Hepes, 5 Na₂EGTA, 2 MgATP and 0.5 Na₂GTP, 10 QX-314 (to suppress voltage-gated sodium currents), pH set to 7.2 with CsOH. Electrode junction potentials (−3 mV) were corrected for. Electrodes were pulled from thin-walled borosilicate tube and had a resistance in external solution of 2–4 MΩ. Cells were considered acceptable if the holding current at −33 mV was < 400 pA, with an access resistance < 10 MΩ. External solution, as for brain slicing but without sodium kynurenate, was superfused at 25°C, or in some experiments at 35°C as stated in the text. Excitatory postsynaptic currents were evoked by placing a concentric stimulating electrode at a standard location among the Schaffer collaterals and stimulating at 100 V.

Drugs and chemicals

Tetrodotoxin (TTX, 0.1–1 μm, see text), picrotoxin (300 μm), strychnine (5 μm) and glycine (100 μm) were normally present in the bathing fluid to block action potentials, to block GABA_A and glycine receptors, and to fully activate the glycine-binding site on NMDA receptors, but TTX was omitted when studying synaptic currents. Glutamate (Glu), dl-threo-β-benzyloxyaspartate (TBOA), cadmium (CdCl₂), indomethacin, d-2-amino-5-phosphonopentanoic acid (AP5), pyridoxal-phosphate-6-azophenyl-2′,4′-disulphonic acid (PPADS), 4,4′-diisothiocyanostilbene-2,2′-disulphonic acid (DIDS), 5-nitro-2-(3-phenylpropyl-amino) benzoic acid (NPPB), 18α-glycyrrhetinic acid, cystine, (S)-4-carboxy-phenylglycine (CPG), tamoxifen and l-methionine sulfoximine (MSO) were applied as indicated. TBOA, NPPB, CPG and bafilomycin were purchased from Tocris (Bristol, UK); picrotoxin, glutamate, DIDS, PPADS, CdCl₂, 18α-glycyrrhetinic acid, cystine, tamoxifen, veratridine and MSO were from Sigma; TTX was from Alomone labs. In order to maintain a constant extracellular chloride concentration in experiments involving propionate, we used a solution which was obtained from the normal solution by substituting 20 mm sodium methanesulphonate for 20 mm NaCl. Sodium propionate was then substituted for an equivalent amount of sodium methanesulphonate. To pretreat slices with bafilomycin to deplete glutamate from synaptic vesicles, slices were incubated in a solution (without sodium kynurenate) to which 4 μm bafilomycin was added for over 2.5 h, and for 5 min at the start of this period 10 μm veratridine was also present to depolarize neurones and increase vesicle turnover. Control slices were incubated for the same period in solution containing sodium kynurenate (and were also exposed to veratridine for 5 min).

Analysis

Data are presented as means ± s.e.m. Glutamate release was assessed as the current produced by activation of glutamate receptors which occurred when glutamate uptake was blocked with TBOA. TBOA-evoked currents in the presence of a putative glutamate release-blocking drug were compared with the average of the values of the current in control solution before applying and after washout of the drug. Levels of significance were assessed using Student's two-tailed t test. Differences in means were considered significant when the P-value was less than 0.05.

When putative blockers of glutamate release mechanisms were applied, we carried out control experiments to check that they did not affect the glutamate receptors being used to sense the glutamate concentration, by assessing the effect of the drug on the response to exogenous glutamate. Only if a drug blocked the response to TBOA more than it blocked the response to exogenous glutamate did we conclude that glutamate release had been reduced.

We considered the possibility that exogenously applied glutamate might act on cells in the slice to release further glutamate. This does not invalidate our approach of comparing the effect of a putative release-blocking drug on the responses to TBOA and to exogenous glutamate, as is shown in the following analysis, because any secondary glutamate release evoked by exogenous glutamate will also be produced by the glutamate that accumulates (as a result of tonic release) when transporters are blocked with TBOA. To see this, suppose that a glutamate concentration ([glu]) produced in the extracellular space either exogenously ([glu]_applied) or by inhibiting uptake with TBOA ([glu]_TBOA), produces extra glutamate release through two mechanisms, one mechanism which is blocked by the drug being investigated and one mechanism which is not affected by the drug. If, in response to the [glu]_applied, the [glu] in the extracellular space is amplified by a factor F_b via the mechanism which is blocked (b) by the drug, and amplified by a factor F_nb via the mechanism which is not blocked (nb) by the drug, then in the absence of the drug the total glutamate concentration produced by applied glutamate is:

and in the presence of the drug (when F_b = 0):

Thus, the ratio of the glutamate concentrations produced in the presence and absence of drug is:

(1)

When TBOA is applied the same equations apply except that [glu]_applied is replaced by a glutamate concentration proportional to the glutamate release rate. We express the tonic glutamate release as:

where R_b is the rate of release by the mechanism that is blocked by the drug under consideration, and R_nb is the rate of release by the non-blocked release mechanism. These release mechanisms will generate a glutamate concentration that we assume to be proportional to the release rate (because diffusion and transporters operate with a linear dependence on [glu] at the low concentrations being considered), which will then be amplified as described above for exogenous glutamate application. Thus, in the absence of the blocking drug:

where k is a constant, and in the presence of the blocking drug (when R_b and F_b are zero):

Thus, the ratio of the glutamate concentrations produced in the presence and absence of drug is:

(2)

Comparing eqns (1) and (2), we see that the blocking drug reduces the extracellular glutamate concentration produced by TBOA application more than it reduces the [glu] produced by exogenous glutamate application (and so will reduce the glutamate receptor generated response more as well). The extra factor by which the TBOA response is multiplied is:

(3)

i.e. the decrease (relative to the decrease of the response to exogenous glutamate) reflects only the fraction of the glutamate release which is suppressed (and is not affected by the parameters F_b and F_nb describing secondary release of glutamate). Thus, if a drug blocks glutamate release, then it will reduce the response to TBOA by a larger factor than it reduces the response to externally applied glutamate. Consequently, we can determine whether a drug blocks glutamate release, even if there is secondary glutamate release, by comparing the effect of the drug on the response to TBOA with the effect on the response to applied glutamate.

Transient currents caused putatively by transient glutamate release from astrocytes (Angulo et al. 2004; Fellin et al. 2004) were defined as inward currents with an amplitude > 20 pA, a rise time > 10 ms and a decay time > 100 ms, and their rate of occurrence in 1-min periods in control solution or around the peak of the response to 1.5 min TBOA exposure was quantified. TBOA did not significantly increase the rate of these events (0.33 ± 0.13 events min⁻¹ in 8 cells in TBOA versus 0.22 ± 0.05 events min⁻¹ in 10 control cells, P = 0.4).

Results

Blocking Na⁺-dependent glutamate transporters activates an NMDA receptor current in pyramidal cells

We monitored tonic glutamate release into the extracellular space of hippocampal slices by applying a non-transported blocker of Na⁺-dependent glutamate transporters, dl-TBOA (Shimamoto et al. 1998, 2000),and detecting the resulting rise of extracellular glutamate concentration, [glu]_o, from the membrane current it produced in whole-cell clamped CA1 pyramidal cells (Jabaudon et al. 1999). All experiments were done with action potentials blocked by TTX, GABA_Aand glycine receptors blocked with picrotoxin and strychnine, and the glycine site of NMDA receptors activated with glycine. Cells were clamped to −33 mV to remove Mg²⁺ block of NMDA receptor channels and thus maximize the current generated in response to a rise of [glu]_o. dl-TBOA blocks the hippocampal glutamate transporters GLAST/EAAT1, GLT-1/EAAT2 and EAAC1/EAAT3 with K_i values of 42, 5.7 and 8.0 μm, respectively (Shimamoto et al. 1998, 2000), so the concentration of TBOA used here (200 μm) should reduce uptake of a low concentration of extracellular glutamate by these transporters by 83%, 97% and 96%, respectively.

Figure 1A shows that TBOA produced a slowly developing inward current, which reached 84.4 ± 3.6 pA after 1.5 min in 100 cells. The magnitude of the current was not significantly reduced by increasing the concentration of TTX present from 0.1 to 1 μm TTX (75.9 ± 5.3 pA in 34 cells and 88.8 ± 4.6 pA in 66 cells, respectively, P = 0.07), ruling out incompletely blocked action potentials in the lower TTX concentration as a source of tonic glutamate. The TBOA-evoked current was almost completely absent (blocked by 98.4 ± 1.6% in 6 cells, P = 1.3 × 10⁻⁶) when TBOA was applied in the presence of the NMDA receptor blocker d-AP5 (50 μm), and recovered when the d-AP5 was washed out (Fig. 1B), implying that the current is generated by tonic release of glutamate activating NMDA receptors (Jabaudon et al. 1999).

Figure 1. Blocking Na⁺-dependent glutamate transporters reveals tonic glutamate release that evokes an NMDA receptor-mediated current.

A, blocking glutamate uptake with TBOA evokes an inward current at −33 mV that is blocked by d-AP5, in an area CA1 pyramidal cell. B, mean current amplitude data for experiments as in A, in 6 cells, normalized to initial amplitude before d-AP5. C, NMDA (200 μm) evokes a desensitizing inward current at −33 mV. D, the effect of the voltage-gated calcium channel blocker Cd²⁺ on the response of pyramidal cells to glutamate (10 μm, chosen to generate a current similar in amplitude to that evoked by TBOA). E, effect of Cd²⁺ on the response to TBOA. F, mean amplitude of currents evoked by glutamate (4 cells) and TBOA (4 cells) in Cd²⁺ (normalized to values without Cd²⁺ present). G, effect of bafilomycin pretreatment (see Methods) on the amplitude of the EPSC evoked at −33 mV by stimulating the Schaffer collaterals (at 100 V). H, effect of bafilomycin pretreatment on the response of pyramidal cells to 10 μm glutamate. I, effect of bafilomycin pretreatment on the response to 200 μm TBOA.

d-AP5 itself has been reported to block a large inward current in hippocampal pyramidal cells (Sah et al. 1989: ∼ 200 pA at −35 mV in 400 μm thick slices (at 30°C, age not specified); < 60 pA in 100 μm slices), which was attributed to activation of NMDA receptors by the ambient glutamate level present. However, we found (in 225 μm slices, at −33 mV) that d-AP5 only blocked an inward current of 5.4 ± 1.4 pA in seven cells at 25°C and a current of 20.8 ± 4.0 pA in six slices at 35°C (data not shown).

Applying a saturating concentration of NMDA (200 μm; Fig. 1C) activated a peak current of 3.6 ± 0.4 nA in six cells (at 25°C), which decreased to a slowly declining plateau current of 2.7 ± 1.2 nA after 1.5 min when receptor desensitization had occurred (as will presumably also occur in response to the ambient [glu]_o and when raising [glu]_o with TBOA). Thus, the AP5-suppressible current (5.4 pA) produced by the baseline [glu]_o is ∼ 0.2% of the current produced (after 1.5 min) by a saturating NMDA concentration (2.7 nA), or 0.13–0.17% of the current that would be produced by a saturating glutamate concentration, since glutamate produces a maximum current that is 1.2- to 1.5-fold larger than that produced by NMDA (via NR1/NR2A or NR1/NR2B receptors: Priestley et al. 1995). Similarly, the rise of [glu]_o produced by TBOA activates NMDA receptors to produce approximately 3.3% (84.1 + 5.4 pA = 89.5 pA) of the maximum NMDA-evoked current (after 1.5 min), or about 2.2–2.8% of the current that would be produced by a saturating glutamate concentration. Using the dose–response curve for NMDA receptors, I = I_max[glu]ⁿ/{[glu]ⁿ+ EC₅₀ⁿ}, measured for cultured mouse hippocampal neurones by Patneau & Mayer (1990), who found a Hill coefficient, n, of 1.5 and an EC₅₀ of 2.3 μm, these fractional activations translate into an estimated baseline extracellular glutamate concentration of 27–33 nm, and a concentration after 1.5 min TBOA of 0.18–0.22 μm. These experiments were at 25°C. Carrying out similar experiments at 35°C (at which glutamate release, but also glutamate uptake, may be faster), and assuming for simplicity no change in the EC₅₀ for glutamate activating NMDA receptors, we estimated a baseline extracellular glutamate concentration of 77–89 nm, and a concentration after 1.5 min TBOA of 0.60–0.71 μm. Thus, increasing the temperature increases glutamate release somewhat more than it raises the rate of glutamate uptake, and (in the absence of TBOA) a new equilibrium is reached at a higher [glu]_o.

Exocytosis does not contribute to tonic glutamate release

The experiment of Fig. 1A and B was done with action potentials blocked, but spontaneous Ca²⁺-dependent exocytosis might still release glutamate, so we tested the effect of blocking voltage-gated Ca²⁺ channels with 200 μm Cd²⁺, with the aim of confirming the conclusion of Jabaudon et al. (1999), for cultured slices, that tonic glutamate release is not via Ca²⁺-dependent exocytosis. This concentration of Cd²⁺ abolishes somatic Ca²⁺ currents in cultured pyramidal cells (Jabaudon et al. 1999) and reduces transmitter release at the CA3–CA1 synapse (assessed from the magnitude of the field EPSP) by about 70% (Wu & Saggau, 1994).

Cd²⁺ is known to block the NMDA receptors we are using to sense the released glutamate (Riveros & Orrego, 1986; Mayer et al. 1989), so we first tested the effect of Cd²⁺ on the response to superfusion of a glutamate concentration (10 μm) chosen to generate a current similar to that evoked by TBOA. Cd²⁺ reduced the response to glutamate (Fig. 1D), on average by 47 ± 7% (Fig. 1F, P = 0.005, n = 4). Cd²⁺ also reduced the current evoked by TBOA (Fig. 1E) but by approximately the same fraction as it blocked the NMDA sensor current (Fig. 1F, 40 ± 5%, P = 0.008). There was no significant difference between the inhibition of the glutamate-and of the TBOA-evoked currents (P = 0.46). Thus, the effect of Cd²⁺ on the TBOA-evoked current is entirely attributable to its inhibition of the NMDA receptors used to sense glutamate release, and we conclude that Cd²⁺ had no effect on the tonic release of glutamate. Subsequent experiments use this format for displaying the results, i.e. showing the effect of a drug first on the glutamate sensitivity of the NMDA receptor sensors, and then the effect on tonic release detected as a TBOA-evoked current mediated by the NMDA receptors. In order to conclude that a drug blocks tonic glutamate release we would need to demonstrate a significantly greater reduction of the TBOA-evoked current than of the glutamate-evoked current (see Methods).

To rule out a contribution of Ca²⁺-independent exocytosis to tonic glutamate release, we used bafilomycin pretreatment to block the vesicular H⁺-ATPase: this depletes vesicles of transmitter and thus prevents spontaneous and evoked exocytotic transmitter release (Zhou et al. 2000; Rossi et al. 2003; Allen et al. 2004; Allen & Attwell, 2004). To check that bafilomycin pretreatment was indeed depleting glutamate from vesicles, we evoked EPSCs in pyramidal cells by stimulating the Schaffer collateral input from CA3. Bafilomycin pretreatment reduced the amplitude of the EPSC evoked by a 100 V stimulus by 91% (Fig. 1G). In contrast bafilomycin had no effect on the cells' response to glutamate (Fig. 1H) or to TBOA (Fig. 1I). We conclude that exocytosis does not contribute significantly to tonic glutamate release onto hippocampal pyramidal cells.

Jabaudon et al. (1999) also found that, in cultured slices, inhibition of exocytosis with botulinum A or tetanus neurotoxin (which abolished spontaneous miniature EPSCs) did not reduce tonic glutamate release, again ruling out the possibility that tonic release is by Ca²⁺-independent spontaneous vesicular release.

Ca²⁺-evoked glial glutamate release does not contribute significantly to tonic activation of glutamate receptors

Transient release of glutamate from glia, probably by exocytosis, can activate NMDA receptors in hippocampal pyramidal cells (Fellin et al. 2004; Angulo et al. 2004). These transient release events were reported to occur at a very low rate (0.16–0.82 min⁻¹ at room temperature to 34°C) and to have an amplitude of ∼ 100 pA and a decay time of ∼ 600 ms. We also observe these rare transient events, at a mean frequency of 0.22 ± 0.05 min⁻¹ in 10 slices at 25°C (mean amplitude 61.7 ± 1.1 pA, mean rise time 169 ± 42 ms, mean decay time 891 ± 380 ms), but they contributed only a small fraction to the time averaged tonic activation of NMDA receptors (3 ± 1% in 10 cells).

Cystine–glutamate exchange does not generate the tonic glutamate release

To test the possible contribution of cystine–glutamate exchange to tonic glutamate release, we first activated the exchange by applying 300 μm exogenous cystine. The EC₅₀ for activation of the exchange by external cystine is ∼ 100 μm (Wyatt et al. 1996; Warr et al. 1999) so the concentration used is sufficient to activate the exchange to about 75% of its maximum level. As reported previously for cerebellar Purkinje cells and neocortical pyramidal cells, cystine sometimes evoked a small inward current itself (arrow in middle panel of Fig. 2A), reflecting a rise of glutamate concentration activating glutamate receptors (Warr et al. 1999). Cystine did not significantly alter the response to 10 μm glutamate (reduced by 6 ± 4% in 6 cells, P = 0.22, Fig. 2D), but increased the response to TBOA (Fig. 2A) by a factor of 2.26 ± 0.49 in 11 cells (P = 0.025, Fig. 2A and D). Thus, cystine–glutamate exchange is present in the hippocampus, and can contribute significantly to glutamate release when the extracellular cystine concentration is high enough.

Figure 2. Cystine–glutamate exchange is present in hippocampal slices, but does not mediate tonic glutamate release.

A, cystine potentiates the TBOA-evoked current (without affecting the response to glutamate: see D), implying that it increases glutamate release. Arrow shows a small inward current produced by cystine. B, the cystine–glutamate exchange blocker (S)-4-CPG abolished the potentiation produced by cystine, consistent with cystine releasing glutamate via cystine–glutamate exchange. C, in the absence of cystine, CPG has no effect on the TBOA-evoked current, implying cystine–glutamate exchange does not normally release a significant amount of glutamate. D, mean effects on the glutamate- and TBOA-evoked currents of cystine (6 cells for glutamate, 11 cells for TBOA), cystine in the presence of CPG (6 cells for TBOA; glutamate not studied), and CPG (5 cells each for glutamate and TBOA). Data are normalized to the value with no drug (for cystine alone or CPG alone), or to the value in CPG (for application of cystine in CPG).

To determine the contribution of cystine–glutamate exchange to glutamate release in the absence of superfused cystine (in which situation the [cystine]_o is presumably closer to its in vivo value (measured in the striatum) of 0.16 μm: Baker et al. 2003), we employed the non-transported cystine–glutamate exchange blocker CPG ((S)-4-carboxyphenylglycine, 50 μm: Ye et al. 1999; Patel et al. 2004). The K_i of CPG is 5 μm (Patel et al. 2004),so for a low external cystine and glutamate concentration 50 μm CPG should inhibit the exchange by 91%. First we verified that CPG did block cystine–glutamate exchange, by repeating the experiment of Fig. 2A in the presence of CPG. CPG abolished the increase of tonic glutamate release evoked by exogenous cystine (Fig. 2B and D; in the presence of CPG, cystine increased the TBOA-evoked current by 0.8 ± 4.5%, P = 0.76, in 6 cells). Then we tested the effect of CPG on the TBOA-evoked current in the absence of exogenous cystine. CPG had no effect on the NMDA receptors used to sense the rise of [glu]_o evoked by TBOA (the response to 10 μm glutamate was increased insignificantly by 6.9 ± 3.4% in 5 cells, P = 0.1, Fig. 2D). CPG also had no significant effect on the TBOA-evoked current (increased by 3.8 ± 13.3%, P = 0.77, Fig. 2C and D). Thus, cystine–glutamate exchange does not contribute significantly to tonic glutamate release in the absence of an artificially raised extracellular cystine concentration.

Effect of blocking glutamine synthetase on tonic glutamate release

The astrocyte-specific enzyme (Norenberg & Martinez-Hernandez, 1979) glutamine synthetase maintains a low glutamate concentration in astrocytes by converting glutamate to glutamine (Ottersen et al. 1992). Jabaudon et al. (1999) found that blocking this enzyme, and presumably raising astrocyte [glu]_i, increased tonic glutamate release in cultured slices, suggesting that at least some tonic glutamate release is from astrocytes. Since astrocytes can change their properties in culture (Thio et al. 1993; Schultz-Suchting & Wolburg, 1994; Kimelberg et al. 1997; Zelenaia et al. 2000), we wanted to check that this result holds in acute slices, before investigating the possible release mechanism of glutamate from astrocytes.

We tested the effect of applying the glutamine synthetase blocker l-methionine sulfoximine (MSO, 5 mm) for 1 min on the response to glutamate and TBOA. MSO had no effect on the response to glutamate (Fig. 3A and C, increased by 10.1 ± 7.7%, P = 0.3 in 3 cells), but produced a small inward current itself in the absence of TBOA (12.8 ± 3.0 pA in 5 cells), as expected if a rise of [glu]_i in astrocytes increased glutamate release, and increased the response to TBOA by 53.5 ± 5.6% (P = 0.003, n = 4, Fig. 3B and C). The increase in the response to TBOA was significantly larger than that for the response to glutamate (P = 0.01). Thus, as in culture, methionine sulfoximine increases the tonic release of glutamate.

Figure 3. The glutamine synthetase inhibitor methionine sulfoximine (MSO) increases tonic glutamate release.

A, MSO has no effect on the NMDA receptors used to sense glutamate. B, MSO potentiates tonic glutamate release, assessed as a TBOA-evoked current. C, mean data for experiments shown in A and B, normalized to the current value in the absence of MSO (3 cells for glutamate, 4 cells for TBOA).

We were surprised that even a minute's application of MSO was sufficient to potentiate glutamate release, since this agent is often applied for hours. The rapid effect of exposure to MSO suggests that it can enter the cells and inhibit glutamine synthetase, and allow [glu]_i to rise, relatively rapidly. A similar rapid regulation of transmitter release by modulation of intracellular transmitter concentration has been demonstrated for GABAergic synaptic transmission: blocking uptake of precursor glutamate decreases intracellular [GABA] and reduces postsynaptic IPSCs within minutes (Mathews & Diamond, 2003).

Tonic indomethacin-sensitive glutamate release from astrocytes is not significant

A possible contribution to tonic glutamate release of prostaglandin- and Ca²⁺-dependent glutamate release from astrocytes was assessed by blocking this release mechanism with 100 μm indomethacin, which inhibits glutamate release by > 80% (Bezzi et al. (1998) showed that 1 μm indomethacin blocks 80% of the release). Indomethacin was applied at least 2 min before glutamate or TBOA (10–100 μm indomethacin produces 50–66% of its prostaglandin-blocking effect in 2 min (Ko et al. 2002; Yerxa et al. 2002)). Indomethacin did not significantly alter the response to glutamate (increased by 8.8 ± 8.2% in 6 cells, P = 0.35) and did not reduce the response to TBOA (increased by 38 ± 13% in 6 cells, P = 0.03, Fig. 4A), and the effects on the response to glutamate and TBOA were not significantly different (P = 0.094), implying that the prostaglandin-dependent mechanism does not contribute significantly to tonic glutamate release.

Figure 4. Tonic glutamate release is not via a prostaglandin-dependent mechanism, swelling-activated channels, gap junctional hemi-channels or P2X₇ receptor channels.

Effects on the currents evoked by glutamate or TBOA of A, indomethacin (6 cells for glutamate, 6 cells TBOA), B, NPPB (5 cells glutamate, 4 cells TBOA), C, tamoxifen (3 cells glutamate, 3 cells TBOA), D, 18α-glycyrrhetinic acid (18α-GA, 7 cells glutamate, 4 cells TBOA), and E, PPADS (5 cells glutamate, 5 cells TBOA). Data are normalized to currents in the absence of the drugs.

Tonic glutamate release via astrocyte swelling-activated anion channels, gap junctional hemichannels and P2X₇ receptors is not significant

Glutamate release from astrocytes can occur via swelling activated anion channels that are blocked by NPPB (5-nitro-2-(3-phenylpropylamino)benzoic acid) and tamoxifen. These drugs were applied at least 2 min before glutamate or TBOA. NPPB (100 μm) blocks the channels mediating swelling-evoked release by 90% within 1 min (Rutledge et al. 1998; Crepel et al. 1998) and tamoxifen (15 μm) blocks these channels by 80% within 1.5 min: (Rutledge et al. 1998; Ransom et al. 2001). Neither NPPB (100 μm, 4 cells) nor tamoxifen (pooled data from 15 μm on 2 cells and 30 μm on 1 cell) had a significant effect (< 4% change, P > 0.35) on the pyramidal cell response to either glutamate or TBOA (Fig. 4B and C).

To test for tonic glutamate release via gap junctional hemichannels, which are opened by low [Ca²⁺]_o and metabolic inhibition but could have a non-zero open probability under normal conditions (Contreras et al. 2002; Ye et al. 2003), we applied 18α-glycyrrhetinic acid (25 μm, applied at least 2 min before glutamate or TBOA was applied), which blocks hemichannel permeability by > 80% (Ye et al. 2003), and acts in ∼ 10 s (Takeda et al. 2005). This did not significantly alter the pyramidal cell response to glutamate (decreased by 8.3 ± 6.0% in 7 cells, P = 0.22) and increased (rather than decreased) the response to TBOA (increased by 22.2 ± 4.0% in 4 cells, P = 0.01), implying that there is no significant contribution to tonic glutamate release from these channels (Fig. 4D).

ATP-gated P2X₇ receptors can mediate transmitter release, either via transmitter diffusing through the large pore that these receptors form or by activation of a transporter mechanism (Sperlagh et al. 2002; Wang et al. 2002; Duan et al. 2003). In the presence of 30 μm PPADS (pyridoxal phosphate-6-azophenyl-2′, 4′-disulphonic acid, applied 2 min before glutamate or TBOA), which reduces P2X₇ receptor activation by more than 75–86% in 1 min (Duan et al. 2003; Sun et al. 1999), the response to glutamate was reduced by 27.9 ± 8.6% (P = 0.047, 5 cells), implying a small inhibition of NMDA receptors (Fig. 4E), but there was no significant change in the response to TBOA (reduced by 10.2 ± 7.8% in 5 cells, P = 0.19), and there was no significant difference when comparing the effects of PPADS on the glutamate and TBOA evoked currents (P = 0.16), implying that PPADS did not inhibit tonic glutamate release.

Tonic glutamate release is reduced by DIDS

4,4′-Diisothiocyanatostilbene-2,2′-disulphonic acid (DIDS) has been found to block numerous anion transporters and channels via which glutamate might exit the cell. Applying 1 mm DIDS evoked an outward current in pyramidal cells (69.1 ± 9.5 pA in 9 cells at −33 mV), and increased the response to glutamate by 32.7 ± 10.9% (P = 0.04) in 5 cells (Fig. 5A and C). Tauskela et al. (2003) previously suggested that 1 mm DIDS decreased NMDA receptor-mediated currents; however, they applied DIDS only in the presence of NMDA and the apparent decrease of the NMDA-evoked current that they reported was apparently an artefactual result caused by the outward DIDS-evoked current shown in Fig. 5A.

Figure 5. Effects of DIDS and propionate on glutamate sensing and tonic glutamate release.

A, DIDS evokes an outward current and increases the current evoked by glutamate. B, DIDS decreases the TBOA-evoked current. C, mean data for experiments shown in A and B, normalized to current values in the absence of DIDS (5 cells for glutamate, 6 cells for TBOA). D, propionate (which, like DIDS, is expected to evoke an intracellular acidification) also evokes an outward current and increases the glutamate-evoked current. E, unlike DIDS, propionate slightly increases the TBOA-evoked current. F, mean data from experiments shown in D and E (5 cells each for glutamate and TBOA).

DIDS inhibits pH-regulating transporters and makes the intracellular pH go acid (Schwiening & Boron, 1994; Baxter & Church, 1996; Bonnet et al. 2000). Sodium propionate (20 mm), which enters across the cell membrane in neutral form and then releases a proton intracellularly, is expected to mimic this acidification (Rorig et al. 1996; Lee et al. 1996). We found that sodium propionate also induced an outward current (37.7 ± 2.7 pA in 7 cells at −33 mV) and increased the response to glutamate by 13.4 ± 3.6% in five cells (P = 0.016; Fig. 5D and F), suggesting that the outward current and the potentiation of the glutamate-evoked NMDA response seen in DIDS are produced by intracellular acidification.

Propionate also increased the response to TBOA (by 28.5 ± 4.7% in 5 cells, P = 0.0017, Fig. 5E and F). By contrast, DIDS decreased the response to TBOA by 47.2 ± 5.3% in six cells (P = 0.0002, Fig. 5B and C), which was statistically significantly different from the DIDS-evoked change in the glutamate response (P = 0.00067), implying that DIDS decreases the tonic release of glutamate. The fact that DIDS decreases the response to TBOA, while propionate increases it, suggests that the suppressive action of DIDS on tonic glutamate release (unlike its actions on the baseline current and the response to glutamate, which mimic those of propionate) is not produced by an intracellular acidification.

DIDS has been reported to inhibit a glutamate transporter (probably GLAST) in Bergmann glial cells (Ruiz & Ortega, 1995). We considered the possibility that the inhibition of the response to TBOA in Fig. 5B and C was a result of DIDS having already inhibited glutamate transporters present in the slice, so that superimposed TBOA produced less uptake inhibition. If this were the case we would expect that application of DIDS would generate a rise of [glu]_o that would partially activate NMDA receptors. This cannot be assessed from the change of current evoked by DIDS, because that change clearly also involves an outward (probably acidification-evoked) component. We therefore applied 50 μmd-AP5 to determine whether there was more tonic activation of NMDA receptors present in DIDS. The current change measured (2.7 ± 1.2 pA in 6 cells) was not larger (P = 0.18) than that seen in absence of DIDS (5.4 ± 1.4 pA in 7 cells, see above), showing that inhibition of glutamate transporters by DIDS cannot explain the results in Fig. 5B and C.

Discussion

The value of the baseline extracellular glutamate concentration, [glu]_o, in the CNS could have a major effect on neuronal excitability and synaptic transmission, but what sets this value is poorly understood. The data in this paper from acutely isolated hippocampal slices extend earlier results on tonic non-synaptic glutamate release in cultured slices (Jabaudon et al. 1999) which showed using tetanus toxin that this release is not via exocytosis, a result we have confirmed in acute slices using bafilomycin to deplete vesicles of glutamate. We have demonstrated that this release is not via cystine–glutamate exchange, contrary to current ideas based on microdialysis work in striatum (Baker et al. 2002), and that it is not via four other glutamate release mechanisms previously characterized in astrocytes, but that it is inhibited by DIDS. Below, we assess the implications of these results. Throughout this discussion it should be remembered that, although we discuss release of glutamate, it is possible that the release of other agents acting on NMDA receptors, such as aspartate, might also contribute to the currents recorded.

Cystine–glutamate exchange does not significantly contribute to the ambient [glu]_o

In our experiments the cystine–glutamate exchange blocker CPG had no effect on tonic glutamate release unless an unphysiologically high concentration of external cystine was superfused. This is consistent with the low level of cystine present in the extracellular space of the brain. Baker et al. (2003) measured an extracellular [cystine] value of ∼ 0.16 μm, which is much less than the EC₅₀ for external cystine activating cystine–glutamate exchange (∼ 100 μm, Wyatt et al. 1996; Warr et al. 1999) and will activate the exchanger to less than 0.2% of its maximum rate.

There are three possible explanations for the difference between our results and those of Baker et al. (2002). First, in our experiments on slices the diffusional access to the bulk solution (and lack of a cystine/cysteine supply from the blood) might reduce the cystine concentration in the extracellular spaces of the preparation to a value even smaller than the 0.16 μm that Baker et al. (2003) measured in vivo, leading to cystine–glutamate exchange being deactivated. If so, then the cystine–glutamate exchange could be important in vivo (despite the very low [cystine]_o), as Baker et al. (2002, 2003) suggest, but not in slice experiments. Alternatively, in the experiments of Baker et al. (2002) the microdialysis cannula might have caused leakage of cystine into the extracellular space from the blood where the cystine concentration is 80 μm (Battistin et al. 1971), artefactually raising the local [cystine]_o (Westergren et al. 1995). In this case, slice experiments might better assess the true role of cystine–glutamate exchange, which could be negligible as a source of glutamate in vivo. Finally, Baker et al. (2002) relied on CPG being a specific blocker of cystine–glutamate exchange, but this glutamate analogue is also an antagonist at mGluR1 and mGluR5 metabotropic glutamate receptors and an agonist at mGluR2 (Hayashi et al. 1994; Kingston et al. 1995) and is expected to reduce neuronal excitability and glutamate release by both of these mechanisms (Lingenhohl et al. 1993; Ugolini & Bordi, 1995; Moroni et al. 1998). Baker et al. (2002) checked that an action of CPG at group I mGluRs was not how CPG inhibited cystine uptake in their experiments, but they did not check that CPG still reduced tonic glutamate release when mGluRs were blocked. Thus, CPG may have reduced [glu]_o in their experiments by acting on mGluRs and reducing spontaneous exocytotic glutamate release rather than by blocking cystine–glutamate exchange.

What mechanism mediates tonic glutamate release other than cystine–glutamate exchange?

We found, like Jabaudon et al. (1999), that the glutamine synthetase inhibitor methionine sulfoximine (MSO) increased tonic glutamate release. MSO is expected to raise the glutamate concentration in astrocytes, so the increase in tonic release could be explained if at least some of the release occurs from astrocytes and a raised astrocyte [glutamate] potentiates this release. However, our experiments ruled out a contribution to tonic glutamate release from four previously characterized glutamate release mechanisms present in astrocytes: prostaglandin- and Ca²⁺-dependent release, swelling-activated anion channels, gap junctional hemichannels and P2X₇ receptors. The only agent we found to significantly reduce tonic glutamate release was DIDS, which blocks many transporters and channels that allow the movement of negatively charged molecules like glutamate.

In the Appendix, we consider the possibility that some of the tonic glutamate release results from passive diffusion of glutamate across cell membranes. Using the measured diffusion coefficients for charged and neutral amino acids crossing lipid bilayers, we estimate the baseline [glu]_o that could be produced by diffusive glutamate release when Na⁺-dependent uptake is functioning normally ([glu]_o∼ 31 nm) or when it is inhibited with TBOA ([glu]_o∼ 0.5 μm). Although these estimates depend on parameters for diffusion and uptake which are only approximately known, they are quite similar in magnitude to the values we estimated experimentally (30 nm and 0.2 μm, respectively, at 25°C, see Results). Thus, it appears that diffusive glutamate efflux may contribute significantly to determining the baseline [glu]_o in hippocampal slices. Potentiation of the TBOA-evoked current by MSO (Fig. 3) could then just reflect more diffusive efflux when the intracellular glutamate concentration is raised in astrocytes, but it is unclear how DIDS might modulate diffusive glutamate efflux (Chakrabarti & Deamer, 1992 reported that pH changes had little effect on the diffusion of amino acids across membranes), suggesting a release mechanism in addition to passive diffusion across the lipid bilayer.

Functional significance of tonic glutamate release

The micromolar levels of ambient glutamate found in microdialysis experiments should activate NMDA and mGluR receptors, and induce desensitization of NMDA, AMPA and kainate receptors, implying that alteration of the ambient [glu]_o may have profound consequences for the information processing carried out by neurones. Indeed, in cultured hippocampal neurones, Forsythe & Clements (1990) found that less than 1 μm glutamate depressed the EPSC amplitude by 40%, and Zorumski et al. (1996) found similar effects of low micromolar levels of glutamate on both the AMPA and the NMDA components of the EPSC. Sah et al. (1989) and Dalby & Mody (2003) reported that NMDA receptors in cells in hippocampal slices could be tonically activated by the ambient glutamate level, generating an inward current which increases the excitability of the neurones. However, the size of the tonically activated NMDA receptor (d-AP5 suppressible) current found in the present study was much smaller than that found in pyramidal cells by Sah et al. (1989), suggesting a resting [glu]_o of around 30–80 nm at 25–35°C (see Results) rather than the 2 μm found in microdialysis experiments. The reason for the difference in d-AP5 suppressible current is unclear, but it could reflect the sharp electrode recordings made by Sah et al. (1989) being from cells very deep in the slice, possibly at a location with compromised oxygen supply, while those we studied with patch-clamp techniques are from more superficial cells. Whether our resting [glu]_o is artefactually lowered in the slice preparation, or the microdialysis value is artefactually raised by damage caused by the microdialysis procedure (see Westergren et al. 1995), remains an important question to be answered.

The small size of the d-AP5-sensitive current we record without TBOA present (∼ 5 pA at −33 mV at 25°C), compared to that seen when uptake is blocked by TBOA (84 pA), highlights the fact that the functional significance of the tonic release of glutamate will depend critically on the prevailing level of Na⁺-dependent glutamate uptake.

Appendix

Estimate of the contribution of diffusion across lipid membranes to the ambient [glu]_o

We considered the possibility that some of the tonic glutamate efflux detected when TBOA is applied results from passive diffusion of glutamate across cell membranes. Charged and neutral amino acids cross lipid bilayers with a permeability coefficient of P = 5–20 × 10⁻¹⁴ m s⁻¹ (depending on the saturation of the membrane lipid molecules: Chakrabarti & Deamer, 1992). The membrane area per unit volume of hippocampus is 14 μm²/μm³, so in 1 litre there is A= 1.4 × 10⁴ m² membrane area (Lehre & Danbolt, 1998). Thus, for a mean intracellular glutamate concentration of [glu]_i = 3 mm (Attwell et al. 1993), the diffusive glutamate flux into the extracellular space will be R_diffusion =PA[glu]_i = 2.1−8.4 nmol l⁻¹ s ⁻¹. The [glu]_o this would produce can be estimated by noting that, if this is the only glutamate release mechanism, the release rate must equal the uptake rate by Na⁺-dependent transporters. Equating an R_diffusion of (say) 5 nmol l⁻¹ s⁻¹ to an uptake rate, described by Michaelis-Menten kinetics as U_max[glu]_o/([glu]_o+K_m), where the maximum uptake rate, U_max, in brain slices is ∼ 0.2 μmol ml⁻¹ min⁻¹ or 3.33 μmol l⁻¹ s⁻¹ and K_m is ∼ 20 μm (Hertz, 1979; Schousboe, 1981), we predict a baseline [glu]_o of 31 nm, which is similar to the 30 nm we estimate from the size of the AP5-suppressible current. When TBOA is added, the [glu]_o rise will depend critically on the magnitude of glutamate uptake remaining unblocked in TBOA. Assuming that 21% of hippocampal uptake is due to GLAST and 79% to GLT-1 (Lehre & Danbolt, 1998), i.e. ignoring the contribution of EAAC1 (Haugeto et al. 1996), the expected 83% and 97% inhibition of GLAST and GLT-1 in the presence of 200 μm TBOA (see Results) will lead to only 6% of glutamate uptake remaining functional in TBOA. Equating an R_diffusion of 5 nmol l⁻¹ s⁻¹ to an uptake rate of 0.06.U_max[glu]_o/([glu]_o+K_m) predicts an equilibrium value for [glu]_o in TBOA of 0.5 μm. This is slightly higher than the 0.2 μm[glu]_o we estimate to be produced when [glu]_o is still rising after 1.5 min of TBOA exposure (see Results). Aspartate could be released by the same mechanism, and would contribute to the activation of NMDA receptors detected.