Embryonic rat hippocampal neurons were cultured on poly-d-lysine (PDL) or a monolayer of postnatal cortical astrocytes to reveal putative changes in neuronal physiology that involve astrocyte-derived signals during the first 4 d of culture. GABA-induced Cl− current (IGABA) was quantified using outside-out and whole-cell patch-clamp recordings beginning at 30 min, when cells had become adherent. The amplitude and density (current normalized to membrane capacitance) of IGABA in neurons grown on astrocytes became statistically greater than that recorded in neurons grown on PDL after 2 hr in culture (HIC). Although the current density remained unchanged in neurons on astrocytes, that in neurons on PDL decreased and became statistically lower beginning after 2 HIC. The differences in amplitude and density of IGABA in the two groups of neurons were maintained during the 4 d experiment. The upregulation effect of astrocytes on neuronal IGABA required intimate contact between the neuronal cell body and underlying astrocytes. Suppression of spontaneous Cac 2+ elevations in astrocytes by bis(2-aminophenoxy)ethane-N,N,N′,N′-tetra-acetic acid that was loaded intracellularly decreased their modulatory effects on IGABA. IGABA in all cells was blocked completely by bicuculline and exhibited virtually identical affinity constants, Hill coefficients, and potentiation by diazepam in the two groups. Outside-out patch recordings revealed identical unitary properties of IGABA in the two groups. More channels per unit of membrane area could explain the astrocyte enhancement of IGABA. The results reveal that cortical astrocytes potentiate IGABA in hippocampal neurons in a contact-dependent manner via a mechanism involving astrocyte Cac 2+elevation.
Astrocytes can directly influence neuronal electrical activity by regulating extracellular ions (mainly K+) and neurotransmitter concentrations (such as GABA and glutamate) in the extracellular compartment that bathes these cells (Schon and Kelly, 1974; Lieberman et al., 1989; Hertz, 1990). Astrocytes also have significant effects on the neuronal expression and distribution of ion channels and neurotransmitter receptors (for review, see Barish, 1995). For example, experiments by Bostock et al. (1981), Shrager (1988), Ritchie et al. (1990), Joe and Angelides (1992), and Waxman and Ritchie (1993) suggest that astrocytes in close contact with axonal membranes might influence the distribution of Na+ and K+ channels in the latter. Wu and Barish (1994) found that direct contact with astrocytes induces the appearance of transient A-type K+currents and depresses that of sustained D-type K+ current in embryonic mouse hippocampal neurons when the two cells are grown in co-culture. These effects of astrocytes on voltage-dependent K+ currents seem to involve a surface- or extracellular matrix-associated mechanism, rather than a free diffusion of soluble factors in the extracellular space (Wu and Barish, 1994). Intimate contact between axon and vital physiologically intact astrocytes and not simply the plasma membrane of the astrocyte are required for these effects (Joe and Angelides, 1992; Wu and Barish, 1994). Glial regulation of transient K+ current density in chick lumbar sympathetic ganglion neurons also occurs via direct contact, but the glial membrane surface remaining in nonvital cells suffices (Raucher and Dryer, 1994). In rat sympathetic neurons, astrocyte regulation of K+ currents can be mimicked by astrocyte-conditioned culture medium or by ciliary neurotrophic factor implicating a soluble substance or substances (McFarlane and Cooper, 1993). The exact mechanisms and factors involved in astrocyte regulation of neuronal excitability, the intercellular communication, if any, between neurons and astrocytes, and the role such intercellular signaling plays in the codifferentiation of the two cell types remain to be elucidated. These initial results suggest heterogenous forms of communication and signal transduction.
In the present study, we have compared GABA-induced Cl− currents (IGABA) in embryonic rat hippocampal neurons cultured on poly-d-lysine (PDL) and on confluent monolayers of postnatal rat cortical astrocytes. We found that relative to PDL cultures, cortical astrocytes upregulate, or maintain, the neuronal expression of GABAAreceptor/Cl− channels. This modulatory effect requires intimate contact between neuronal cell bodies and astrocytes, indicating the involvement of astrocyte membrane- or extracellular matrix-associated factors and/or ultra-short-range soluble factors. Suppression of spontaneous intracellular Ca2+ fluctuations in astrocytes by bis(2-aminophenoxy)ethane-N,N,N′,N′-tetra-acetic acid (BAPTA) loaded intracellularly significantly reduced these effects on IGABA, suggesting that they involve elevation in cytosolic Ca2+(Cac 2+).
Parts of this paper have been published previously (Liu et al., 1995).
MATERIALS AND METHODS
Preparation of astrocyte monolayers. Cortices, free of hippocampi and striata, were removed from 3-d-old rat pups, cleaned of meninges, and placed in 10 ml L-15 medium with 50 U/ml gentamicin. The tissue was triturated through a 5 ml pipette, dissociated mechanically through a series of small gauge (G) needles (3 × 19–20 G, 3 × 22–23 G, 1 × 25 G), and passed through 62 μm Nitex. Cells were centrifuged at low speed, resuspended in plating medium consisting of DMEM supplemented with 10% fetal calf serum (FCS) and 50 U/ml gentamicin, and plated at the equivalent of two brains per flask in 75 cm2 flasks that had been pretreated with 5 μg/ml PDL (30–70 K; Sigma, St. Louis, MO). Medium was changed completely after 72 hr and twice weekly thereafter. When a confluent monolayer was present (after ∼1 week), the flasks were capped tightly and placed overnight on a rotary shaker at 180 rpm at 37°C. The supernatant, containing microglia, loosely adherent O2A progenitor cells, and debris, was removed rapidly and completely after ∼12 hr. Cultures were rinsed once with L-15 or DMEM and then refed with plating medium. To select further for astrocytes in the culture by removing any surviving neurons and O2A progenitor cells, the flasks were subjected to complement-mediated lysis of cells expressing A2B5 surface antigen, as described by Armstrong et al. (1992). Briefly, the cultures were incubated with a 1:50 dilution of A2B5 ascites in DMEM with 1% FCS or full-strength A2B5 culture supernatant for 1 hr at 37°C. Cultures were then rinsed twice in DMEM-1% FCS and treated for 1 hr at 37°C with rabbit complement diluted 1:8 in DMEM-1% FCS. To reduce the amount of antibody necessary and expose a greater surface area for antibody binding, the cells were trypsinized off the flask and resuspended in a small (1–2 ml) volume. A2B5 is a trypsin-resistant surface antigen. After cytolysis, cultures were rinsed twice in DMEM-1% FCS and refed with plating medium. The cultures were then trypsinized, and the cells were transferred into 35 mm culture dishes precoated with 5 μg/ml PDL (30–70 K, Sigma). Astrocytes for calcium imaging were plated in 35 mm culture dishes with glass coverslip bottoms (MatTek Corporation, Ashland, MA). When cells reached confluence they were exposed to 10 μm cytosine arabinoside for 2 d and then maintained in minimal essential medium (MEM) with 5% horse serum before being used. Cultures prepared in this way contained ≥95% type-1 astrocytes as determined by glial fibrillary acidic protein (GFAP) (positive) and A2B5 (negative) immunocytochemistry and morphological examination.
Preparation and culture of hippocampal neurons. Hippocampal neurons were dissociated from embryonic day 18 rat embryos by papain digestion, according to the method of Huettner and Baughman (1986). Briefly, the hippocampal tissues were cut carefully into small pieces, transferred into 5 ml Earle’s balanced salt solution containing 20 U/ml papain, 0.01% DNase (both from Boehringer Mannheim, Indianapolis, IN), 0.5 mm EDTA, and 1 mm l-cysteine, and kept in an incubator for 35–40 min at 37°C. Single neurons, obtained by triturating the tissue with a Pasteur pipette, were washed twice with culture medium containing 90% MEM (Gibco, Grand Island, NY), 5% FCS, and 5% horse serum (Biofluid, Rockville, MD). The neurons were then plated at a density of ∼5 × 105 cells/dish in 35 mm plastic culture dishes coated with high molecular weight PDL (Sigma) and with or without a monolayer of astrocytes. In some experiments, astrocytes were exposed to culture medium containing 10 μmBAPTA-acetoxymethyl ester compound (BAPTA-AM) for 30 min at 37°C and then washed twice before neurons were plated. Astrocytes were loaded with BAPTA to suppress Cac 2+elevations (Furuya et al., 1994; Gu and Spitzer, 1995). The cultures were kept at 37°C in a water-saturated atmosphere containing 10% CO2. Cells were studied initially at 30 min in culture, the minimum time when they adhered firmly enough to be useful for electrophysiological study. Neurons with relatively large cell bodies bearing visible processes were selected for study. Neurons cultured on PDL were recorded only if they were not in visible contact with astrocytes, which after a variable delay spread progressively and limited useful study of astrocyte-dependent changes in neurophysiology to the first 4 d.
Current recording and analysis. Before recording, dishes were removed from the incubator, and the culture medium was replaced with either a solution containing (in mm) 142 NaCl, 8.1 CsCl, 1 CaCl2, 6 MgCl2, 10 Glucose, 10 HEPES-CsOH, pH 7.3, and 310 mOsm for single-channel recordings in outside-out patch-clamp mode, or Tyrode’s solution containing (in mm) 140 NaCl, 5.4 KCl, 1.8 CaCl2, 0.8 MgCl2, 10 glucose, 10 HEPES-NaOH, pH 7.4, and 310 mOsm for whole-cell recordings. Standard patch-clamp recordings (Hamill et al., 1981) were made with pipettes pulled in three stages from 1.5 mm outer diameter glass capillary tubes (WPI, Sarasota, FL) with a computer-controlled pipette puller (BB-CH-PC, Mecanex SA). These pipettes had a resistance of 3–5 MW when filled with internal solution. Pipettes for single-channel recordings were fire-polished, coated with Sylgard-184, and filled with a solution containing (in mm) 153 CsCl, 1 MgCl2, 5 EGTA, 10 HEPES-CsOH, pH 7.3, and 290 mOsm. For whole-cell recordings, the pipettes were used without being coated or fire-polished and were filled with a solution composed of (in mm) 145 CsCl, 2 MgCl2, 0.1 CaCl2, 1.1 EGTA, 5 HEPES, 5 ATP, 5 phosphocreatine, pH 7.2, and 290 mOsm. Both whole-cell and single-channel currents were recorded using an L/M EPC-7 patch-clamp amplifier (Medical Systems, Greenvale, NY) at different gains: 5 mV/pA for whole-cell and 200 mV/pA for single-channel recordings. Series resistance was compensated for >70% in whole-cell recordings. Current signals were filtered at 10 KHz and stored on videocassettes via a videocassette recorder (VCR) and a VR-100 digital recorder (Instrutech) for later off-line analysis. Whole-cell currents also were recorded simultaneously on a pen recorder (Gould, Glen Burnie, MD). Total membrane capacitance was determined by integrating capacity transients evoked by 10 mV, 15 msec hyperpolarizing pulses (from a holding potential of −80 mV), which were recorded immediately after entering the whole-cell configuration. Input resistance was determined from the steady-state current recorded after the capacitive transients had settled. Single-channel currents were played back from the VCR system as analog signals, filtered at 2 KHz, digitized with a Labmaster-TL-1 DMA interface (Axon Instruments, Burlingame, CA), and then sampled (10 KHz) and analyzed with Pclamp V.6.02 program (Axon Instruments) on a 486 personal computer. Openings and closings of the channels were detected by applying a 50% threshold criterion (Colquhoun and Sigworth, 1983). Openings and closings were considered valid only if their durations were greater than twice the rise time of the system. GABA-activated Cl− channels exhibit multiple current levels (Bormann et al., 1987; Smith et al., 1989). In the present study, only the dominant openings to the 30 pS conductance level were analyzed. Records that contained infrequent double openings (<5% of open events) were used for kinetic analysis. All recordings were carried out at room temperature (22–25°C) on a Nikon inverted microscope. Recorded cells and cell-free membrane patches were superfused continuously with a perfusion system composed of a locally made perfusion controller and miniature electric solenoid valves (The Lee Company, Essex, CT) that allows fast switching (<200 msec complete solution exchange time) among different solutions. Nine inputs converge into a common channel positioned 100–350 μm away from the recorded cell. The perfusion rate (∼0.3–0.5 ml/min) was controlled by the air pressure applied to the solution reservoir.
Calcium imaging. Astrocytes were loaded with the calcium indicator dye Fura-2 by exposure to 4 μm Fura-2 AM (Molecular Probes, Eugene, OR) in standard bath solution (in mm: 145 NaCl, 10 HEPES, 5.4 KCl, 1.8 CaCl2, 0.8 MgCl2, titrated to pH 7.4 with NaOH, and osmolarity adjusted to 330 mOsm with sucrose) for 30 min at 37°C, washed, and then maintained at 37°C for 45 min for ester hydrolysis. Fluorescence microscopy at room temperature (22–24°C) was used for measuring the fluorescence of a chosen field of cells (at 40×, using a Zeiss inverted microscope) with an Attofluor RatioVision digital imaging system (Atto Instruments, Bethesda, MD, and Carl Zeiss, Thornwood, NY). Images using excitation wavelengths of 340 and 380 nm were captured and stored every 2 sec. The ratio of fluorescence at the two exciting wavelengths was calculated for each pixel within a cell boundary to index free intracellular Ca2+ levels.
Statistical tests. Two-tailed t tests for paired and unpaired data were used to assess significance. Differences were considered significant if p < 0.05 (indicated by *) orp < 0.01 (indicated by **).
Astrocytes modulate progressive changes in steady-state membrane properties of hippocampal neurons
Embryonic rat hippocampal neurons plated on both PDL and confluent cortical astrocytes survived and differentiated into visible networks during the experimental period (Fig. 1). Some neurons retained remnants of processes after papain digestion as shown in cultures of 2 hr (Fig. 1, left). These process-bearing neurons were selected for electrophysiological study. During the 4 d period in culture, noticeable differences in the apparent diameters of cell bodies or the extent and numbers of associated processes were not detected using light microscopy. Hence, neurons survived as well on PDL as on astrocytes. After 2–3 d, stellate and epithelioid GFAP+ astrocytes began to appear in the culture of cells growing on PDL (not shown), because the plated population was heterogenous. This limited our investigation of the neuronal properties altered by confluent cortical astrocytes with those persisting in the neuronal population grown on PDL in the presence of progenitor cells and immature, nonconfluent hippocampal astrocytes.
Resting membrane properties recorded at −80 mV in the whole-cell mode including membrane capacitance (C m), input resistance (R in), and specific conductance were quantified in all neurons tested for GABA-evoked current responses to investigate possible astrocyte-derived contributions to steady-state membrane properties. C m andR in values were 8.5 ± 0.6 pF and 4.2 ± 0.6 gW (n = 7), respectively, after 0.5 HIC for neurons grown on PDL, and 8.4 ± 1.7 pF and 4.9 ± 0.8 gW (n = 7) in neurons cultured on astrocytes (Fig. 2 A,B). There were no significant differences in theC m and R invalues between the two groups of neurons (p > 0.05).
After 1 d in culture (DIC), C m increased and R in decreased significantly compared with 2 HIC in neurons grown on both astrocytes (p < 0.01) and PDL (p < 0.01). Furthermore, differences between neurons grown on PDL and on astrocytes inC m and R invalues recorded at 1 DIC were statistically significant (p < 0.01 for both parameters).C m continued to increase in both groups of neurons, but values recorded in neurons on astrocytes remained significantly higher during the 4 d study (Fig. 2 A). The differences between the two groups of neurons inR in remained significant at 2 DIC but became insignificant thereafter (Fig. 2 B). Specific steady-state membrane conductances were both ∼30 pS/pF during the initial 0.5–2 HIC. At 1 DIC, specific membrane conductance had increased in neurons on astrocytes to ∼50 pS/pF, which was significantly different from corresponding values recorded in neurons on PDL that had increased modestly (Fig. 2 C). Thus, the decrease in R in recorded at 1–2 DIC, which was significantly different between the two experimental groups, was independent of progressive increases in C m. After 2 DIC, the difference in specific conductance, like that inR in, became statistically insignificant.
Astrocytes increase the amplitude and density of GABA-activated Cl− current
GABA activated IGABA responses in all tested neurons cultured from 0.5 hr to 4 d on either PDL (n = 167) or astrocytes (n = 182) and recorded in Cl−-loaded cells at a holding potential of −80 mV. Peak current amplitudes evoked by brief applications of GABA in neurons cultured for 0.5 hr on astrocytes (1331.7 ± 287.3 pA;n = 7) were already ∼20% greater than those cultured on PDL (1088.6 ± 119.6; n = 7), but this difference was not statistically significant (p > 0.05; Fig.3 A). After just 2 HIC, however, IGABA had increased to an average of 1736.8 ± 153.2 pA in neurons on astrocytes (n = 20; p > 0.05), whereas IGABA recorded in neurons on PDL had decreased to 839.9 ± 87.5 pA (n = 27; p > 0.05). This difference between mean values of IGABA was highly significant (p < 0.01). During the following 2 d (1 and 2 DIC), there were few or modest changes in the average amplitudes of IGABA recorded in neurons on PDL, and they were not different significantly from IGABA recorded initially (p > 0.05 compared with that at 0.5 HIC). In marked contrast, IGABA recorded in neurons on astrocytes became significantly greater at 1 and 2 DIC when compared with that at 0.5 HIC: 2219.6 ± 95.5 pA (n = 48) and 3177.5 ± 322.1 pA (n = 19), respectively (p < 0.01 compared with that at 0.5 HIC). At 1 and 2 DIC, the differences in average amplitudes of IGABA recorded in the two experimental conditions were highly significant (p < 0.01 for both days). This difference was maintained throughout the whole period of observation (Fig. 3 A).
To account for differences in current amplitude that could be correlated with growth in membrane surface area, specific current densities (IGABA divided by whole-cell membrane capacitance as an index of membrane surface area) were calculated and compared (Fig. 3 B). The average current density recorded at 0.5 HIC in neurons on astrocytes was ∼20% greater (157.8 ± 11.7 pA/pF; n = 7) than that in neurons on PDL (131.1 ± 16.2 pA/pF; n = 7); however, this difference was not statistically significant (p > 0.05). The average current density did not change significantly during the 4 d period in neurons cultured on astrocytes, because bothC m and IGABAincreased in parallel. The current density decreased in neurons grown on PDL during the first 2 hr, continued to decrease during the first 2 d, and then increased slowly thereafter, yet it never reached the level recorded at 0.5 HIC. The rapid and lasting decrease in the density of IGABA in those neurons led to statistically significant differences between the two groups beginning at 2 HIC. This was sustained for the experimental period.
Astrocyte modulation involves direct contact at the cell body level
In some dishes with a layer of confluent astrocytes, there were regions devoid of astrocyte covering (Fig.4 A). We took advantage of this to compare IGABA in neurons whose cell bodies were growing on or off astrocytes in the same dishes to test whether the astrocyte effects required direct contact (Fig. 4 B1,B2). In many cells, the “off-astrocyte” neurons contacted astrocytes through processes. In nine neurons whose cell bodies were growing on astrocytes for 2 d, IGABA averaged 2150.9 ± 255.7 pA, significantly greater than that recorded in neurons in the same dishes whose cell bodies were growing off astrocytes (992.1 ± 164.1 pA;n = 8; p < 0.05; Fig. 4 B,C). Furthermore, IGABA in off-astrocytes neurons was not significantly different from that in neurons cultured on PDL (p > 0.05). The density of IGABA in these off-astrocyte neurons was also significantly less than that in on-astrocyte neurons (92.4 ± 15.8 pA/pF and 157.6 ± 19.1 pA/pF, respectively; p < 0.05) (Fig. 4 D). The results demonstrate that direct contact of the neuronal cell body with astrocytes is a prerequisite for the modulatory effects of the latter on IGABA.
BAPTA-loaded astrocytes do not modulate IGABA
Spontaneous changes in Cac 2+ levels in cultured astrocytes have been reported in various studies (Cornell-Bell and Finkbeiner, 1991; Fatatis and Russell, 1992), and BAPTA-AM loaded intracellularly has been used to suppress spontaneous and evoked changes in Cac 2+ elevations in astrocytes and in many other cell types (Martin et al., 1992; Ballerini et al., 1993; Koyama et al., 1993; Furuya et al., 1994; Gu and Spitzer, 1995). To test whether astrocytes have spontaneous Ca2+ changes in our culture conditions and whether the modulatory effects of astrocytes on IGABA involve elevations in astrocyte Cac 2+, we compared astrocytes unloaded and preloaded with BAPTA-AM in terms of intracellular Ca2+ signals and their effects on neuronal IGABA. In 25 unloaded astrocytes, 15 exhibited spontaneous intracellular Ca2+ elevations during a 600 sec recording period (Fig. 5, left). When loaded with BAPTA-AM and examined for Ca2+signals immediately thereafter, no astrocyte showed spontaneous Cac 2+ elevations (data not shown). Furthermore, only 4 of 28 astrocytes that were exposed to BAPTA-AM for 30 min and then washed and cultured for 20 hr showed spontaneous intracellular Ca2+ elevations (Fig.5, right). Hence, there were clear and persistent differences between the two groups of astrocytes in terms of spontaneous Cac 2+ signals. The frequency of spontaneous Cac 2+signals was the same or higher in unloaded astrocytes (ranging from one to seven fluctuations) than in BAPTA-AM-treated astrocytes (one elevation in each of four astrocytes). IGABA and its density in neurons grown on BAPTA-treated astrocytes for 20 hr (1000.5 ± 185.8 pA and 79.2 ± 7.2 pA/pF; n = 11) were both significantly less than corresponding values recorded in neurons on untreated astrocytes (1909.6 ± 84.6 pA and 118.5 ± 7.9 pA/pF;n = 14; p < 0.01) (Fig.6 A,C). These results suggest that spontaneous Cac 2+ spikes and/or waves are important for the direct contact observed effects of astrocytes on neuronal GABAAreceptor/Cl− channels. Neurons grown on BAPTA-treated astrocytes also had lower C mon average (13.6 ± 1.4 pF) compared with that recorded in neurons grown on unloaded astrocytes (17.0 ± 0.8 pF; p < 0.05) (Fig. 6 B). No significant differences were detected between either R in or specific steady-state conductance measured in the two groups. The effects of BAPTA-AM are unlikely to be attributable to generation of toxic metabolites, because treatment with calcein-AM that does not respond to Ca2+ did not produce similar effects.
Pharmacological properties of IGABA are identical in neurons cultured on PDL and astrocytes
IGABA was highly sensitive to bicuculline (Fig. 7 A) and picrotoxin (data not shown), antagonists at GABAAreceptor/Cl− channels, in both groups of neurons. Ten micromolar bicuculline blocked IGABAby 95.3% (from 1.36 ± 0.26 nA to 0.06 ± 0.01 nA; n = 6;p < 0.01) in neurons on PDL and by 96.4% (from 2.75 ± 0.37 to 0.10 ± 0.03 nA; n = 7; p < 0.01) in neurons on astrocytes. Fifty micromolar bicuculline blocked GABA-induced current responses recorded in neurons on both PDL (n = 6) and astrocytes (n = 7) in a virtually complete manner. The results indicate that all of the IGABA involves bicuculline-sensitive currents characteristic of GABAA-type receptors.
Current responses could be detected consistently with brief pulses of 1–3 μm GABA in neurons grown on either PDL (n = 7) or astrocytes (n = 6) for 2 d. IGABA increased in a sigmoidal fashion with GABA concentration. When experimental values in the two groups were normalized to the maximum response, the two curves virtually superimposed and were well fitted with the continuous theoretical curves calculated from the following equation: where I is the amplitude of IGABA, I max is the maximum current, [GABA] is the concentration of GABA,K d is the dissociation constant of GABA with its receptors, and n is the Hill coefficient (Fig.7 B). The Hill coefficients andK ds were virtually identical for neurons grown on PDL (1.65 and 9.9 μm) and on astrocytes (1.56 and 9.1 μm). Thus, at the macroscopic level, IGABA recorded in the two sets of neurons exhibited indistinguishable [GABA]-dependent properties.
IGABA evoked in neurons cultured on PDL and astrocytes were all assumed to be Cl−-dependent because they reversed polarity at ∼0 mV, the equilibrium potential for Cl− under these recording conditions (Fig.7 C). The slope conductance at the peak of the response (normalized according to whole-cell C m) was significantly greater in neurons grown on astrocytes (1964.5 ± 202.6 pS/pF; n = 5) than in neurons on PDL (1249.4 ± 74.5 pS/pF;n = 3; p < 0.05).
We studied some of the pharmacological properties of the GABAA receptor/Cl−channels expressed at 2 HIC and 24 HIC, using a clinically relevant drug that affects GABAAreceptor/Cl− channels (diazepam) and a naturally occurring divalent cation (Zn2+). Diazepam increased IGABA by 158.8 ± 28.8% after 2 HIC (n = 8; p < 0.01) and by 140.9 ± 20.5% after 24 HIC (n = 7; p < 0.01) in neurons on PDL, and IGABA was increased by 109.9 ± 7.3% after 2 HIC (n = 5; p < 0.01) and 181.9 ± 61.4% after 24 HIC (n = 5; p < 0.01) in neurons on astrocytes. The differences in the enhancing effects of diazepam in neurons cultured on PDL and astrocytes were not significant for both 2 HIC and 24 HIC data sets (p > 0.05 for both groups).
After 2 HIC, 100 μm Zn2+blocked 25.4 ± 3.2% of the GABA-evoked current in 13 neurons on PDL (p < 0.01) and 23.6 ± 3.2% in 12 neurons on astrocytes (p < 0.01). The differences in the blocking effects of Zn2+ between the two groups of neurons was not significant (p > 0.05). After 24 HIC, Zn2+ blocked 15.9 ± 2.2% of IGABA in neurons on PDL (n = 11;p < 0.01) and 36.5 ± 2.9% in neurons on astrocytes (n = 12; p < 0.01). The difference was significant (p < 0.01) between the two groups.
IGABA decay involves both redistribution of Cl− and conductance decay
The current response to GABA recorded at negative holding potentials decreased with time (Fig. 8 A) during continuous applications. IGABA recorded in neurons grown on astrocytes decayed faster and to a greater extent than that recorded in neurons grown on PDL (Fig. 8, inset). In 15 neurons cultured on astrocytes, IGABA decayed by 87.3 ± 1.1% with a half-decay time (T ½) of 3.6 ± 0.2 sec, whereas that recorded in 11 neurons grown on PDL decayed by 78.4 ± 2.1% (p < 0.01) with aT ½ of 7.9 ± 1.5 sec (p < 0.01) during 1 min GABA applications.
The current decay could be attributable to GABAAreceptor desensitization at the level of the receptor/channel complex (decrease in the activation of receptor-coupled conductance) and/or by Cl− ion redistribution across the membrane (indicated by a shift in the reversal potential of the current response) (Akaike et al., 1987; Frosch et al., 1992). To test the contributions of conductance decay and Cl− ion redistribution to the decay of IGABA, fast ramp voltage-commands (1 S, 100 mV/S) were applied every ∼10 sec during the 1 min applications of GABA to generate data (Fig. 8 B,insets) on reversal potential (Fig. 8 B,C) and maximum slope conductance (g Cl) (Fig.8 D). g Cl was measured over the positive potential range when Cl− ions moved into the cell from an infinite reservoir. The reversal potentials of the peak currents recorded during the first second were ∼0 mV in neurons grown on both PDL (−0.1 ± 1.5 mV; n = 6) and astrocytes (−0.6 ± 0.5 mV; n = 5; p > 0.05), whereas those measured ∼10 sec later had shifted consistently to negative potentials in both groups. The mean values in both groups were fitted adequately with bi-exponential functions with time constants of 8.3 S and 111.3 S in neurons on astrocytes and 3.0 S and 498.1 S in neurons on PDL. The maximal extent in the shift was significantly greater in neurons grown on astrocytes than in neurons on PDL (Fig.8 B,C). In six neurons grown on astrocytes, the reversal potential revealed by the ramp command applied just before the end of GABA application averaged −24.6 ± 3.5 mV, whereas that in five neurons grown on PDL was −10.9 ± 1.2 mV (p < 0.01). Furthermore, the cumulative shift in reversal potential closely correlated with the peak current amplitude (p < 0.01). g Cl in both groups decayed in a bi-exponential manner during the prolonged GABA application (Fig.8 D). The mean time constants were 5.1 S and 84.1 S in neurons on astrocytes and 2.1 S and 69.6 S in neurons on PDL. The extent of decrease in g Cl was virtually identical in the two groups: 67.1% in neurons on astrocyte (from 60.5 ± 7.8 nS at the peak current to 19.9 ± 2.5 nS just before the end of the GABA application) and 67.3% in neurons on PDL (from 40.1 ± 2.8 nS to 13.1 ± 1.9 nS) (p > 0.05). The results indicate that the faster and more complete decay of IGABA in neurons grown on astrocytes reflects a greater degree of rapid Cl− ion redistribution rather than a faster rate of desensitization.
Unitary properties of GABA-activated Cl− channels do not differ
The enhancement of GABA-activated Cl−current by astrocytes may involve detectable changes in the biophysical properties of GABAAreceptor/Cl− channels. To investigate these possibilities further, Cl− channels activated by GABA were recorded in outside-out patches obtained from neurons cultured for 2 d on PDL or astrocytes. Application of 3 μm GABA evoked transitions in microscopic currents, which occurred as discrete single openings or interrupted bursts (Fig. 9 A) and were highly sensitive to bicuculline (not shown). Only the main conductance (∼30 pS) was studied further. The reversal potentials of single-channel currents were ∼0 mV, the equilibrium potential for Cl−under these recording conditions (data not shown). In four patches in neurons on PDL, the mean value of open channel conductance was estimated to be 29.4 ± 0.9 pS, which was not significantly different from that derived from three patches in neurons on astrocytes (29.9 ± 1.8 pS; p > 0.05). These values are similar to those reported previously for GABA-activated channel openings recorded from cultured embryonic hippocampal, spinal cord, and cortical cells (Ozawa and Yuzaki, 1984; Bormann et al., 1987; Smith et al., 1989; Orser et al., 1994).
The open-time histograms were fitted adequately with bi-exponential functions having time constants of 0.58 ± 0.08 and 5.37 ± 0.62 msec in neurons on PDL (n = 4) and 0.64 ± 0.13 and 6.04 ± 0.10 msec in neurons on astrocytes (n = 3; p > 0.05 for both time constants) (Fig. 9 B). The mean open times were 3.05 ± 0.41 msec in neurons on PDL and 3.35 ± 0.46 msec in neurons on astrocytes (p > 0.05). These values are similar to mean open-time durations reported previously for GABA-activated Cl− channels (Mienville and Vicini, 1989; Ma et al., 1994; Orser et al., 1994). The closed-time histograms were fitted adequately with the sum of three exponential functions (Fig.9 C). The closed-time constants were 1.71 ± 0.11, 8.83 ± 1.55, and 48.63 ± 10.34 msec, and the mean closed time was 24.9 ± 8.9 msec in neurons on PDL, which were not significantly different from those in neurons on astrocytes (2.42 ± 0.58, 9.15 ± 2.9, and 58.46 ± 8.53 msec in time constants and 23.6 ± 13.7 msec in mean closed time;p > 0.05). Furthermore, the open probabilities were also not significantly different between neurons on PDL (0.08 ± 0.02) and astrocytes (0.12 ± 0.04; p > 0.05).
Differences in IGABA in neurons cultured on astrocytes and PDL occur before changes in cell capacitance. The results show that embryonic rat hippocampal neurons cultured for hours/days on a monolayer of confluent cortical astrocytes express IGABA responses that are consistently greater in amplitude during brief exposure and faster in decay during sustained application than those recorded in neurons cultured in the same medium under identical conditions on PDL. A statistically significant difference in IGABA was recorded as early as 2 HIC, before differences in C mbecame significant. Direct contact of neuronal cell bodies with astrocytes was essential for a difference in IGABA because “off-astrocyte” neurons recorded in the same fields exhibited significantly smaller IGABA, even though they visibly contacted astrocytes via their processes (Fig. 4). IGABAnormalized to C m, which may be an index of the specific density of GABAAreceptor/Cl− channels, and remained relatively stable (∼150 pA/pF) in neurons cultured on astrocytes, but decreased significantly during the first 2 d in culture in neurons on PDL and never recovered to its original value recorded at 0.5 HIC (Fig. 3). The delayed increase in GABA-activated Cl− current density in PDL dishes coincided with the gradual appearance of GFAP+ astrocytes (not shown).
Differences in IGABA cannot be accounted for by obvious differences in pharmacological and biophysical properties
Structure–function studies of GABAAreceptor/Cl− channels composed of different subunit proteins expressed in recombinant systems have revealed variabilities in biophysical and pharmacological properties associated with different subunits (Burt and Kamatchi, 1991; Burt, 1994). For example, γ2-subunits influence the enhancement by benzodiazepines (Puia et al., 1991) and depression by Zn2+ (Smart et al., 1991); single-channel conductance and kinetics differ with different subunit combinations (Verdoorn et al., 1990; Angelotti and Macdonald, 1993). Although definitive experiments involving in situ hybridization and immunocytochemical detection of different subunits have not been performed in the present set of experiments, there were few differences in the pharmacological and biophysical properties of IGABA recorded in neurons on PDL and astrocytes. They were enhanced by diazepam and blocked by bicuculline. Dose–response curves in the two groups virtually superimposed when normalized to the maximum response (Fig. 7 B), indicating similar if not identical affinity constants and Hill coefficients. The single-channel conductance and kinetics were also similar if not identical in membrane patches obtained from cell bodies (Figs. 9). IGABA in neurons on astrocytes was significantly more Zn2+-sensitive at 1 DIC. Thus, the differences in properties recorded in the two groups of neurons were subtle and modest at best, and at present do not explain the approximately twofold difference in IGABA. Because the elementary properties are similar, if not identical, we infer that their subunit composition is similar. It is possible that more GABAA receptor/Cl−channels with identical properties are present at the neuronal cell surface in neurons in contact with astrocytes.
Cl− ion redistribution rather than desensitization explains the difference in IGABA decay
An important phenomenon, which seems to be intrinsic to functional combinations of GABAA receptor subunits, is marked decay in response to prolonged application of GABA that recovers over a few minutes (Numann and Wong, 1984; Thalmann and Hershkowitz, 1985; Akaike et al., 1987; Oh and Dichter, 1992; Celentano and Wong, 1994). In most cases, this decay is the mixed result of desensitization of GABAA receptor and a decrease in the transmembrane Cl− ion gradient, because of Cl− redistribution (Huguenard and Alger, 1986;Akaike et al., 1987; Frosch et al., 1992). It is not yet clear how ECl could change so dramatically during a prolonged exposure to GABA when Cl− ions are present in a virtually infinite supply in both the patch pipette and extracellular solutions. The dramatic, hyperpolarizing shift of ECl at negative holding potentials (as in our case) could be attributable to local depletion of intracellular Cl− and/or local accumulation of extracellular Cl− in the immediate vicinity of the orifices of GABAA receptor/channels. Perhaps unstirred layers adjacent to the intracellular and extracellular openings of the channel transiently form discrete compartments with EClthen being determined by the Cl− concentrations in these compartments. Cl− in these undisturbed compartments may have a limited exchange rate with extracellular or cytosolic and, by inference, pipette Cl− so that during prolonged activation of GABAAreceptor/channels, Cl− concentrations change locally, resulting in the observed shift in ECl.
The molecular mechanisms of desensitization of GABAA receptors also are not entirely clear, but seem to be influenced by GABAA receptor phosphorylation state (Moss et al., 1992) and subunit composition (Verdoorn et al., 1990; Moss et al., 1992). The greater hyperpolarizing shift in the reversal potential of IGABA in neurons on astrocytes (Fig. 8 B,C), coupled with the fact that GABA-activated maximal slope conductance decays to the same extent and in a parallel manner in both groups (Fig. 8 D), indicates that the faster, more complete degree of IGABAdecay in neurons on astrocytes is the result of faster and more complete Cl− redistribution. The statistically significant correlation between the extent of IGABA decay and the initial peak amplitude is consistent with the hypothesis that GABA rapidly moves the intracellular Cl− ions out of the cell, thus depleting Cl− ions in the immediate vicinity of GABAA receptor/channels and shifting the reversal potential of IGABA in a hyperpolarizing direction.
Astrocyte modulation of IGABA occurs in parallel with increases in Cm
It is well known that astrocytes support neuronal survival and neurite extension (Lindsay, 1979; Noble et al., 1984; Fallon, 1985;Manthorpe et al., 1986; Alliot et al., 1988; Le Roux and Reh, 1994). After 1 DIC, neurons on astrocytes exhibited measurably moreC m than those on PDL, although in both groups of neurons C m increased significantly when compared with initial values (Fig.2 A). One explanation for larger-amplitude IGABA recorded in neurons cultured on astrocytes would be that functional GABAA receptor/channels are inserted as new membrane differentiates, so that neurons with more plasma membrane and greater C m will exhibit greater IGABA. In fact, the specific density of GABA-activated Cl− current never changed significantly in neurons grown on astrocytes (Fig. 3 B). In neurons on PDL, IGABA remained unchanged during the first 2 d in culture, whereas whole-cell capacitance increased by >60%, so that density of the current decreased (Figs.2 A, 3). Despite morphological differentiation (Fig.1) and an increase in plasma membrane in neurons cultured on PDL, the rate of insertion of new receptors at the cell surface is less, compared with the rate in those on astrocytes. Alternatively, without the support of astrocytes, GABAA receptors are degraded faster so that although new receptors are inserted at a constant rate, their density actually decreases. Independent assessment of receptor turnover rate will help to reveal how this is related to the observed phenomenology.
Mechanisms of the astroglial modulation of neuronal IGABA
Astrocytes synthesize neurotrophic factors, including some extracellular matrix macromolecules and soluble substances that are released into the extracellular space (Lindsay, 1979; Banker, 1980;Hatten and Mason, 1986; Manthorpe et al., 1986; Pixley et al., 1987;Alliot et al., 1988; Sanes, 1989; Le Roux and Reh, 1994). These neurotrophic factors may affect neuronal survival, migration, differentiation, neurite extension, and, as recently reviewed by Barish (1995), expression, distribution, and function of ion channels. No factor or factors that generally influence ion channels have been defined clearly yet. For example, the induction of A-type transient potassium current and depression of D-type potassium current in mouse hippocampal pyramidal neurons differentiating in culture required active synthesis of a factor or factors transmitted by direct contact or short-range diffusion (Wu and Barish, 1994), whereas diffusible factors were believed to influence the potassium currents in rat sympathetic ganglion neurons (McFarlane and Cooper, 1993). Similarly, no specific factors have been identified yet for the contact-mediated astrocyte modulation of GABA-activated Cl−current in embryonic rat hippocampal neurons. It is most likely that either factors associated with cell surface or extracellular matrix or those that diffuse only a very limited distance are involved. Furthermore, the effect of astrocytes may not be specific to GABAA receptors but rather may be part of a more general change in the physiology of cultured hippocampal cells involving other (such as glutamate) receptors. Cultured astrocytes show spontaneous Cac 2+ elevations (Cornell-Bell and Finkbeiner, 1991; Fatatis and Russell, 1992). The fact that suppression of spontaneous Cac 2+ elevation in astrocytes by loading BAPTA-AM intracellularly significantly reduced their modulatory effects on IGABA (Fig. 6) suggests that either active Cac 2+-dependent secretion of regulatory factors from astrocytes or their elaboration on the cell surface is involved in the modulation of IGABA.
We thank the Instrumentation and Computer Section, National Institute for Neurological Disorders and Stroke, National Institutes of Health, for fabricating the nine-channel perfusion controller complete with solenoid valves. We thank Atto Instruments, Bethesda, Maryland, and Carl Zeiss, Inc., Thornwood, New York for letting us use their Attofluor RatioVision digital imaging system.
Correspondence should be addressed to Qi-Ying Liu, Laboratory of Neurophysiology, NINDS, National Institutes of Health, Building 36/Room 2C02, 9000 Rockville Pike, Bethesda, MD 20892.