Astrocytes regulate developmental changes in the chloride ion gradient of embryonic rat ventral spinal cord neurons in culture

Abstract

1. Embryonic rat ventral spinal cord neurons were dissociated at day 15 and grown on: (i) poly-D-lysine (PDL); (ii) a confluent monolayer of type I astrocytes; or (iii) PDL in astrocyte-conditioned medium (ACM) to examine the influence of astroglia on the regulation of GABA_A receptor/Cl⁻ channel properties.

2. Potentiometric oxonol dye recordings of intact cells indicated that embryonic neurons were uniformly depolarized by muscimol. The depolarizing effects disappeared in cells dissociated during the early postnatal period and recovered in culture for 24 h. Similar recordings using the calcium-imaging dye fura-2 AM revealed that GABA or muscimol triggered a sustained rise in cytosolic Ca²⁺ (${\text{Ca}}_{\text{c}}^{\text{2+}}$) in embryonic neurons that was dependent on extracellular Ca²⁺, blocked by bicuculline and nifedipine and sensitive to changes in extracellular chloride. The incidence and amplitude of the Ca²⁺ response decreased with time in vitro and was accelerated in neurons cultured on astrocytes compared with those on PDL.

3. Perforated patch-clamp recordings revealed that GABA depolarized neurons in a Cl⁻-dependent and bicuculline-sensitive manner. Both the resting membrane potential and the GABA equilibrium potential became more hyperpolarized with time in vitro.

4. Astrocytes and ACM accelerated the transformation of GABAergic potential responses from depolarizing to hyperpolarizing. The change occurred over the first 4 days in co-culture or in ACM but took more than 2 weeks in neurons cultured on PDL alone.

5. The intrinsic, elementary properties of GABA_A receptor/Cl⁻ channels including open time and unitary conductance changed independently of the presence of astrocytes or ACM. Mean open time of the dominant kinetic component decreased and conductance increased with time in vitro.

6. In sum, astrocytes accelerate the developmental change in the Cl⁻ ion gradient extrinsic to GABA_A receptor/Cl⁻ channels, which is critical for triggering Ca²⁺ entry, without influencing parallel changes in the intrinsic properties of the channels.

Recent evidence indicates that amino acids like GABA and glutamate and their receptors are expressed throughout the embryonic mammalian central nervous system (CNS) where they are considered to play morphogenic roles in development (for review, see Lauder, 1993) in addition to their classic roles as fast-acting neurotransmitters. In this regard, the molecular components of a putative GABAergic signalling system, including transcripts encoding GABA-synthesizing enzymes and specific GABA_A receptor subunits have been detected in the embryonic rat spinal cord along with the corresponding gene products and GABA (Behar, Schaffner, Laing, Hudson, Komoly & Barker, 1993; Ma, Saunders, Somogyi, Poulter & Barker, 1993). The depolarizing effects of GABA and muscimol, an agonist at GABA_A receptor/Cl⁻ channels, on embryonic spinal cord neurons has been well established using potentiometric dye recordings (Walton, Schaffner & Barker, 1993) as well as perforated-patch (Wang, Reichling, Kyrozis & MacDermott, 1994) and on-cell patch techniques (Serafini, Valeyev, Barker & Poulter, 1995). The depolarizing effects of GABA and muscimol are blocked by bicuculline, which selectively antagonizes the activation of GABA_A receptor/Cl⁻ channels. Ca²⁺-indicator dye recordings of embryonic spinal cord cells dissociated from the dorsal horn reveal that both GABA and muscimol trigger bicuculline-sensitive, extracellular Ca²⁺-dependent elevations in cytosolic Ca²⁺ (${\text{Ca}}_{\text{c}}^{\text{2+}}$) (Reichling, Kyrozis, Wang & MacDermott, 1994; Wang et al. 1994). These are blocked by nifedipine, leading to the conclusion that the sustained elevation of ${\text{Ca}}_{\text{c}}^{\text{2+}}$ by GABA involves Cl⁻-dependent depolarization and activation of voltage-dependent Ca²⁺ channels. In vivo (Wu, Ziskind-Conhaim & Sweet, 1992) and in vitro (Reichling et al. 1994; Wang et al. 1994) the Cl⁻-dependent depolarization of spinal cord neurons and the elevation in their ${\text{Ca}}_{\text{c}}^{\text{2+}}$ levels gradually disappear as the Cl⁻ gradient across the membrane increases. The mechanisms involved in these developmental transformations in the polarity of GABA_A receptor-coupled, Cl⁻-dependent signals have yet to be elucidated.

Spinal astrocytes expressing glial fibrillary acid protein (GFAP) proliferate and differentiate during the embryonic and postnatal period (Yang, Lieska, Shao, Kriho & Pappas, 1993) when Cl⁻-dependent GABAergic signals switch polarity in vivo (Wu et al. 1992). Here we have used dissociated cell-culture techniques to study the effects of GFAP⁺ astrocytes on the differentiation of GABAergic signal polarity and associated channel properties expressed by embryonic ventral horn neurons maintained in vitro for several weeks. The results indicate that in vitro, astrocytes accelerate the time-dependent switch in the polarity of GABAergic signals independently of developmental changes in the elementary properties of GABA_A receptor/Cl⁻ channels and these astrocyte-induced effects can be reproduced using astrocyte-conditioned medium (ACM). Some of these results have been reported in abstract form (Li, Schaffner, Walton & Barker, 1995).

METHODS

Cell culture

Astrocyte monolayer

Rat pups (3-day-old) were rapidly decapitated. Cortices, free of hippocampi and striata and cleaned of meninges, were removed and placed in 10 ml of L-15 medium with 50 U ml⁻¹ gentamicin. The tissue was triturated through a 5 ml pipette followed by mechanical dissociation through a series of small bore needles (3 × 19 g, 3 × 22 g, 1 × 25 g) and filtered through 62 μm nylon mesh (Nitex, TETKO, Inc., Kansas City, MO, USA). The dissociated cells were centrifuged and resuspended in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10 % v/v fetal calf serum (FCS) and 50 U ml⁻¹ gentamicin and plated in 75 cm² flasks at the equivalent of two brains per flask (all tissue culture supplies from Gibco, Grand Island, NY, USA). The medium was changed after 72 h and twice each week thereafter. When a confluent monolayer was present (after ∼1 week) the flasks were tightly capped and placed overnight on a rotary shaker at 180 r.p.m. in an incubator at 37°C. The following morning the supernatant, containing microglia, loosely adherent O-2A progenitor cells and debris, was rapidly removed. Cultures were rinsed once with DMEM and re-fed with plating medium. To eliminate any surviving neurons and O-2A progenitor cells the flasks were then subjected to complement-mediated lysis, as described by Armstrong, Dorn, Kufta, Friedman & Dubois-Dalcq (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 h at 37°C. Cultures were rinsed twice in DMEM-1 % FCS and treated with rabbit complement diluted 1:8 in DMEM-1 % FCS for 1 h at 37°C. To reduce the amount of antibody and expose a greater surface area for antibody binding the cells could also be trypsinized off the flask and resuspended in a small (1-2 ml) volume of antibody solution since A2B5 is a trypsin-resistant surface antigen. After cytolysis, the cultures were rinsed twice in DMEM-1 % FCS and held in DMEM-5 % FCS. The cultures could be maintained in flasks for an indefinite period or trypsinized and transferred to 35 mm² dishes precoated with 5 μg ml⁻¹ poly-D-lysine (PDL; 53 K, Sigma, St Louis, MO, USA). When cells reached confluency in the dishes they were exposed to 10 μM cytosine arabinoside for 2 days. If flasks or dishes were to be maintained for several weeks the medium was changed to DMEM-2.5 % FCS or Minimum Essential Medium (MEM)-5 % horse serum. Cultures prepared in this way contained ≤ 95 % type-1 astrocytes as determined by GFAP, S100β, and A2B5 immunocytochemistry and morphological examination. Greater than 95 % of the cells on the monolayer were GFAP/S100β positive and A2B5 negative (data not shown).

Astrocyte-conditioned medium

ACM was prepared by the addition of MEM-5 % horse serum to flasks or dishes of ‘purified’ astrocytes for 24 h; 5 % FCS was added to ACM for incubation with newly dissociated neurons but was not added to ACM intended for neuronal cultures on or after day 4.

Neurons

Pregnant Sprague-Dawley rats were narcotized by CO₂ inhalation followed by cervical dislocation. E15 embryos were removed from the uteri by Caesarean section, rapidly decapitated and placed in phosphate-buffered saline at room temperature (20-22°C). Ventral spinal cords were dissociated and placed into a solution of Earle's Balanced Salt Solution (EBSS) containing 20 U ml⁻¹ papain (Worthington Biochemical, Freehold, NJ, USA), 0.01 % w/v DNase (Boehringer Mannheim, Indianapolis, IN, USA), 0.5 mM EGTA, and 1 mM L-cysteine for 45 min at 37°C (Huettner & Baughman, 1986) to initiate dissociation of the cells. After trituration the cells were spun at 300 g for 5 min, and resuspended in EBSS with 1 mg ml⁻¹ bovine serum albumin (BSA; Sigma) and 1 mg ml⁻¹ ovomucoid trypsin inhibitor (Sigma). The cell suspension was layered over 5 ml of EBSS with 10 mg ml⁻¹ each of BSA and ovomucoid and centrifuged at 80 g for 7 min. The neurons were plated in 35 mm diameter plastic dishes (NUNC, Thomas Sci., Swedesboro, NJ, USA) that had been previously coated with high molecular weight PDL or on a confluent layer of astrocytes. Plating medium consisted of modified Eagle's medium (MEM) with 3.7 g l⁻¹ sodium bicarbonate, 6 g l⁻¹ glucose (Gibco), 5 % fetal calf serum and 5 % v/v horse serum. In some experiments medium conditioned by astrocytes for 24 h was used as the culture medium. Cultures were kept at 36°C in a CO₂ incubator. The medium was changed twice weekly and the cells were maintained in MEM and 5 % horse serum.

Cultures used for digital videomicroscopy were in 35 mm diameter glass-bottom microwell dishes (MatTek Corp., Ashland, MA, USA).

Microelectrode recording

Cultures were removed from the incubator, medium was removed and replaced with recording saline. Recordings were made on the stage of an inverted phase microscope (Nikon). Patch-clamp recordings were carried out at room temperature either in the whole-cell configuration or with the perforated-patch technique using gramicidin. Ionic currents and voltages were measured with a patch-clamp amplifier (Axopatch 200A), band-pass filtered at 0.1 Hz to 2 kHz and monitored on a storage oscilloscope and displayed on a pen recorder. The electrical signals were recorded at 9 kHz on a four-channel VCR PCM system (Instrutech VR-100B, Great Neck, NY, USA). For fluctuation analysis of GABA-activated current responses, the currents were low-pass filtered at 0.4-250 Hz, digitized (12 bits, National Instrument LAB-PC card) at 2 kHz, stored on a 386 PC microcomputer and analysed using SPAN (software courtesy of J. Dempster, Strathclyde University, Glasgow, UK).

The patch pipettes for whole-cell recordings were filled with (mM): 145 CsCl, 5 Hepes, 2 Mg-ATP, 5 BAPTA, 5 sodium phosphocreatine, pH 7.2, and the osmolarity was adjusted to 310 mosmol l⁻¹ with sucrose. The K₂SO₄ pipette solution contained (mM): 10 Hepes, 70 K₂SO₄, 10 KCl, 5 EGTA, 0.5 CaCl₂, pH 7.2, with osmolarity at 310 mosmol l⁻¹. Electrode resistance measured in the recording saline was between 4 and 6 MΩ. Series resistance was ∼15 MΩ and was compensated for by 80-90 %. The standard external solution contained (mM): 145 NaCl, 10 Hepes, 10 D-glucose, 5.4 KCl, 1.8 CaCl₂, 0.8 MgCl₂, titrated to pH 7.4 with NaOH, and osmolarity adjusted to 330 mosmol l⁻¹ with sucrose. The low Cl⁻ solution was made by equimolar replacement of NaCl with sodium gluconate.

For perforated-patch recordings, the patch pipette solution contained (mM): 140 potassium gluconate, 10 KCl, 10 Hepes, titrated to pH 7.2 with NaOH, and osmolarity adjusted to 320 mosmol l⁻¹ with sucrose. Gramicidin (Sigma) was initially dissolved in dimethyl sulphoxide (DMSO, 50 mg ml⁻¹). The pipettes were first filled with gramicidin-free pipette solution by brief immersion of the tip. The remainder of the pipette was then backfilled with the solution containing gramicidin diluted to a final concentration of 100 μg ml⁻¹. Series resistance was ≤ 20 MΩ and was compensated for by 80-90 %.

Dishes were superperfused continuously with recording medium at a rate of ∼1.2 ml min⁻¹. Drugs were prepared and stored at -20°C in 10 mM aliquots and were applied from a puffer pipette.

To estimate intracellular Cl⁻ concentration ([Cl⁻]_i) a version of the Nernst equation was used:

where 155.6 is the concentration of extracellular Cl⁻ and E_GABA is the reversal potential for GABA-induced responses.

Digital video microscopy

Oxonol dye potentiometry

Details are described in Walton et al. (1993). Briefly, a Nikon Diaphot microscope was equipped with a xenon source and epi-illumination with a rhodamine filter was set appropriate for the oxonol dye DiBaC₄(5) (Molecular Probes, Eugene, OR, USA). Dye (50 nM) was added to all superfusion media which consisted of a physiological saline with and without various test substances. Two glass inserts were placed inside the culture dish to decrease the volume to approximately 0.2 ml thus ensuring a rapid change in solutions. An image intensifier coupled to a CCD video camera was attached to the video output port of the microscope and fluorescent images were captured using a frame-grabber at the rate of one per minute. The anionic oxonol dyes diffuse across the membrane and achieve a Nernstian equilibrium (Apell & Bersch, 1987). Depolarization is recorded as an increase in fluorescence. Since fluorescence is directly related to the intracellular dye concentration and altered via diffusion, the fluorescence changes are quantitatively continuous rather than exhibiting a threshold. However, due to intrinsic electronic noise, the analysis procedure accepted only changes of at least 0.1 log units as an unequivocal change in membrane potential. Manipulation of extracellular [K⁺] indicated that a fluorescence change in the order of 0.1 log units corresponded to a potential difference of ∼8 mV, evidence that the technique is sensitive to discrete changes in membrane potential (Walton et al. 1993).

Calcium imaging

Neurons were loaded with the calcium indicator dye fura-2 by exposure to 4 μM fura-2 AM (Molecular Probes) in standard bath solution for 30 min and then washed and maintained for 45 min for ester hydrolysis at 37°C. As with oxonol recordings, rapid solution changes were obtained by decreasing the dish volume to 0.2 ml with glass inserts. Digital video imaging fluorescence microscopy was used for measuring the fluorescence of a chosen field of cells (× 40, using a Nikon Diaphot inverted microscope), and images using excitation wavelengths of 340 and 380 nm were captured and stored. The ratio of fluorescence at the two exciting wavelengths was calculated for each pixel within a cell boundary. Calibration of the ratio in terms of Ca²⁺ was carried out by adding ionomycin to the dish and recording images in calcium-free medium (no Ca²⁺ plus 5 mM EGTA) and in normal medium (1.8 mM Ca²⁺). The calcium concentration was derived from:

where K_D is the fura-Ca²⁺ binding constant (∼220 nM), R is a ratio of fluorescence at two wavelengths and R_min and R_max are values of R in calcium-free and normal medium, respectively. F₀/F₀ is the ratio of fluorescence at 380 nm in calcium-free medium versus 1.8 mM Ca²⁺.

Statistical tests

A two-tail t test was used to assess statistical significance in Figs 5 and 7. Differences were considered significant if P < 0.05 and have been indicated by *, while P < 0.01 are denoted by ** in these figures. The data points reflect means ± standard error of the mean. For electrophysiological studies n represents the number of neurons tested; for optical recordings n is the number of fields tested.

Figure 5. E_GABA hyperpolarization in long-term culture occurs more rapidly in the presence of astrocytes or astrocyte-conditioned medium (ACM).

A, the resting membrane potentials (V_m) recorded with perforated-patch techniques become more negative over time in culture independently of the presence of astrocytes. All the shift in E_GABA occurs during the first week. B, E_GABA hyperpolarizes during long-term culture and the rate and extent is greater on astrocytes or in ACM relative to that recorded in neurons on PDL. C, the differences between the absolute values of E_GABA and V_m are plotted, with positive values representing the depolarizing responses and the negative values representing hyperpolarizing responses. D, intracellular chloride ([Cl⁻]_i) derived from measurements of E_GABA, changes over time in culture more rapidly in neurons on astrocytes and in neurons exposed to ACM compared with changes occurring on PDL alone. The change in [Cl⁻]_i occurs over the first week with differences becoming statistically significant at 2 days. The data points represent the means ± s.e.m. (n = 5-7 cells). *P significant at the 0.05 level, **P significant at the 0.01 level. Significant differences were present between cells on PDL versus astrocytes and ACM in B, C and D.

Figure 7. Unitary properties of GABA-activated Cl⁻ channels change in culture independently of astrocytes.

Unitary properties of channels were estimated at different times in neurons cultured under three different experimental conditions: on PDL; on astrocytes; or in astrocyte-conditioned medium (ACM). Each data point reflects the mean ± s.e.m. of 5-7 cells. Biophysical properties were very similar in all three conditions. A, the estimated mean open time of the long-lasting component (τ_long) shortens from ≈75 to ≈50 ms in 7 days, then remains stable. B, the contribution of power in this component to the fluctuating signal remains relatively constant at ≈ 80 % in all three conditions. C, the unitary conductance increases from ≈16 to ≈21 pS in 7 days, then remains stable. D, the estimated mean open time of the short-lasting component (τ_short) remains constant at 3-4 ms throughout the culture period.

A Fisher exact probability test using 2 × 2 contingency tables was used to determine significance between responders and non-responders on the two surfaces in Fig. 3A;* denotes P < 0.05.

Figure 3. GABA-evoked [Ca²⁺]_c responses disappear more rapidly in neurons co-cultured on astrocytes.

A, the percentage of neurons responding to 10 μM GABA with an increase in [Ca²⁺]_c of at least 15 nM is plotted as a function of days in culture. The number of cells responding decreases with time in culture on both PDL (▪) and astrocytes () but astrocytes accelerate this change. The numbers above each column are the number of responding cells/number of cells tested. * denotes P < 0.05. B, the mean amplitude of the ${\text{Ca}}_{\text{c}}^{\text{2+}}$ response evoked by 10 μM GABA decreases with time in culture on both surfaces, but the decrease is magnified in the presence of astrocytes. *P significant at the 0.05 level, **P significant at the 0.01 level.

Animal procedures

All animals procedures were done in accordance with Public Health Service Policy on Humane Care and Use of Laboratory Animals and the Guide for the Care and Use of Laboratory Animals in the USA.

RESULTS

Muscimol-induced depolarization disappears in vivo during the early postnatal period

We used oxonol dye potentiometry to record from intact ventral spinal cord cells dissociated at late embryonic and early postnatal days to assess the time course over which the GABA_A agonist muscimol is able to depolarize neurons. We found that about 80 % of embryonic ventral horn cells dissociated and recorded within 2-3 h of plating on PDL (79 ± 2 %; n = 3 fields/129 cells) or after 24 h in culture (75 ± 6 %; n = 6 fields/201 cells) depolarized to 2 μM muscimol (Fig. 1A). All the responding cells exhibited processes and were TuJ1 or tetanus-toxin positive (data not shown), identifying them as neurons (Koulakoff, Bizzini & Berwald-Netter, 1983; Moody, Quigg & Frankfurter, 1989). At this age and subsequent ages it is unlikely that non-responders are dead or traumatized. These states should result in a very high level of baseline oxonol fluorescence due to leaky plasma membranes and this was not found to be the case. A similar percentage of neurons dissociated at birth or at different postnatal days depolarized to muscimol when the cells were recorded within 2-3 h of plating. However, progressively fewer neurons depolarized after recovery in culture for ∼24 h (Fig. 1A). Thus, by postnatal day 14 few neurons recovered in culture for 24 h depolarized to muscimol (4 ± 3 %; n = 7 fields/98 cells). This was not due to selective loss of muscimol-responsive neurons. When the same neuron was recorded after several hours in culture and again after 24 h, the disappearance of a depolarizing response was demonstrated in the same cell. In contrast, the percentage of neurons that depolarized to 50 μM kainate remained high at all times whether they were derived from the embryonic or postnatal period or whether they were recorded several hours or 24 h after dissociation (Fig. 1A). A similar percentage of embryonic neurons depolarizing at ∼2 and ∼24 h in response to kainate strongly suggests that the depolarizing effects are not due to the trauma of cell dissociation, but reflect the physiological state of the cells that exists in vivo. These potentiometric results recorded in vitro on ventral spinal cord neurons indicate that in vivo the depolarizing effects of muscimol and, by inference GABA, disappear during the early postnatal period. The results with acutely recorded postnatal neurons also imply that the trauma of cell dissociation leads to Cl⁻ loaded cells whose intracellular Cl⁻ decreases dramatically during ∼24 h recovery in vitro.

Figure 1. GABA_A agonists elicit depolarization and an increase in intracellular Ca²⁺.

A, depolarizing responses to muscimol, examined with the potentiometric dye oxonol, disappear in postnatal dissociates recovered for 1 day in vitro (1 Div). Eighty percent of the embryonic cells depolarize to muscimol after 2-3 h in vitro and after a 1 day recovery in vitro. Progressively fewer postnatal cells depolarize to muscimol after 24 h with only rare cells from postnatal day (PN) 14 responding. The percentage of neurons that depolarize to kainate remains high at all times independent of age at dissociation or time after dissociation. Each data point represents the mean ± s.e.m. for 3-7 fields (11-54 cells per field). B-D, GABA induces Ca²⁺-dependent increases in [Ca²⁺]_c in ventral spinal cord neurons dissociated at embryonic day (E) 15 and recovered overnight in culture. B, 10 μM GABA induces a sustained increase in ${\text{Ca}}_{\text{c}}^{\text{2+}}$ that is completely and reversibly blocked by co-application of 100 μM bicuculline. C, the ${\text{Ca}}_{\text{c}}^{\text{2+}}$ increase is completely eliminated in Ca²⁺-free saline and recovers within 10 min. Note that Ca²⁺-free saline also lowers the steady-state ${\text{Ca}}_{\text{c}}^{\text{2+}}$ level in a reversible manner. D, 5 μM nifedipine blocks the ${\text{Ca}}_{\text{c}}^{\text{2+}}$ increase induced by 10 μM GABA in a reversible manner. The left-hand panels illustrate representative recordings and the right-hand panels are plots of mean values and their standard errors (s.e.m.) from multiple recordings (n = 31-36 cells).

GABA triggers a sustained rise in ${\text{Ca}}_{\text{c}}^{\text{2+}}$ in embryonic ventral spinal cord neurons

We recorded the effects of GABA on ventral spinal cord neurons dissociated at E15 and recovered for ∼24 h in culture on PDL. GABA triggered a rise in ${\text{Ca}}_{\text{c}}^{\text{2+}}$ from about 50 nM to about 200 nM (Fig. 1B-D) in the same percentage of neurons (80 %) as depolarized to muscimol. The ${\text{Ca}}_{\text{c}}^{\text{2+}}$ response to GABA was blocked almost completely and in an entirely reversible manner by co-application of bicuculline, which by itself had no effect (Fig. 1B). The steady-state level of ${\text{Ca}}_{\text{c}}^{\text{2+}}$ was noticeably decreased in ${\text{Ca}}_{\text{c}}^{\text{2+}}$-free medium while the response to GABA was completely eliminated in a reversible manner (Fig. 1C). The ${\text{Ca}}_{\text{c}}^{\text{2+}}$ rise to GABA was also markedly, but not completely, blocked by nifedipine in a reversible manner (Fig. 1D). Virtually identical effects were recorded regarding bicuculline and nifedipine sensitivity and ${\text{Ca}}_{\text{c}}^{\text{2+}}$ dependency when muscimol was used instead of GABA (data not shown). Co-application of 5 μM ω-conotoxin GVIA, 3 μM ω-conotoxin MVIIC or 100 nM tetrodotoxin with muscimol did not affect the ${\text{Ca}}_{\text{c}}^{\text{2+}}$ response (n = 4; data not shown). These results implicate bicuculline-sensitive GABA_A receptor/Cl⁻ channels, which, when activated, depolarize neurons and trigger extracellular Ca²⁺ entry via nifedipine-sensitive, voltage-dependent Ca²⁺ channels.

${\text{Ca}}_{\text{c}}^{\text{2+}}$response is sensitive to changes in extracellular chloride

Since our results implicated Cl⁻-dependent depolarization in the GABA- and muscimol-evoked Ca²⁺ responses, we changed extracellular Cl⁻ (${\text{Cl}}_{\text{o}}^{\text{−}}$) and recorded ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses. Virtually all neurons that exhibited ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses of ∼100 nM to muscimol in normal saline responded to muscimol co-incidentally delivered with low ${\text{Cl}}_{\text{o}}^{\text{−}}$ saline with about a 50 % increase in the amplitude (∼150 nM) of the sustained signal (Fig. 2A,B and D). When neurons were allowed to equilibrate in low ${\text{Cl}}_{\text{o}}^{\text{−}}$ saline and then exposed to muscimol delivered in normal saline, most neurons (∼75 %) no longer responded (Fig. 2A and B (▪)). Those that did respond manifested markedly reduced responses (Fig. 2B ()). Neurons were also recorded that did not respond to muscimol in normal saline but upon switching to low ${\text{Cl}}_{\text{o}}^{\text{−}}$ did express quite detectable ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses indistinguishable from neuronal responses in normal saline (Fig. 2C). These results are consistent with the idea that the observed Ca²⁺ responses are triggered via voltage-dependent Ca²⁺ channels activated by the depolarization of the cell in response to opening of GABA_A receptor/Cl⁻ channels.

Figure 2. Muscimol-induced changes in [Ca²⁺]_c are sensitive to extracellular Cl⁻.

Embryonic neurons were plated on poly-D-lysine (PDL), then recorded for their ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses to 10 μM muscimol under control conditions and in salines with altered [Cl⁻]. Representative recordings are shown on the left and plots summarizing all the results from multiple recordings are depicted on the right (means ± s.e.m.). A and B, neurons that exhibited increases in ${\text{Ca}}_{\text{c}}^{\text{2+}}$ of ≈100 nM in response to muscimol in normal saline (150 mM) responded to muscimol coincidentally delivered in low ${\text{Cl}}_{\text{o}}^{\text{−}}$ saline (50 mM) with ≈50 % increase in the amplitude of the sustained signal (≈150 nM). Neurons equilibrated in low ${\text{Cl}}_{\text{o}}^{\text{−}}$ saline, and then exposed to muscimol delivered in normal saline, no longer responded (A or ▪in B) or exhibited responses of markedly reduced amplitude (, B). C, neurons were also recorded that did not respond to muscimol in normal saline but did so when exposed to muscimol delivered coincidentally in low ${\text{Cl}}_{\text{o}}^{\text{−}}$ saline. These responses were indistinguishable from those typically recorded in normal saline. D, all the cells that responded to muscimol delivered in standard saline with an increase in ${\text{Ca}}_{\text{c}}^{\text{2+}}$ were defined as the starting population (100 %). The vast majority of these cells responded similarly when the saline was changed to one with low Cl⁻. However, when neurons were allowed to equilibrate in low ${\text{Cl}}_{\text{o}}^{\text{−}}$ saline a much smaller percentage responded with an increase in ${\text{Ca}}_{\text{c}}^{\text{2+}}$ when muscimol was subsequently delivered in normal saline. In A and C, smaller ticks are drawn at 5 s intervals, taller ticks at 1 min intervals.

Astrocytes accelerate time-dependent change in GABA-induced ${\text{Ca}}_{\text{c}}^{\text{2+}}$responses occurring in vitro

We recorded ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses to GABA in E15 neurons cultured on a commonly used surface (PDL) in a standard medium and compared them to responses of neurons plated on confluent cortical astrocytes in the same medium. After one day in culture, similar percentages of neurons (∼80 %) differentiating on PDL or on astrocytes exhibited ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses of identical amplitude (Fig. 3A). At 4 days, both the percentage of cells exhibiting GABA-induced ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses and their mean amplitude had decreased considerably more in neurons differentiating on astrocytes compared with neurons growing on PDL. At 1 week, more than 60 % of neurons cultured on PDL still exhibited ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses to GABA. In those neurons that responded the mean elevation of ${\text{Ca}}_{\text{c}}^{\text{2+}}$ was greater than 90 nM. Less than 20 % of neurons on astrocytes expressed ${\text{Ca}}_{\text{c}}^{\text{2+}}$ signals and in those neurons the mean ${\text{Ca}}_{\text{c}}^{\text{2+}}$ response was less than 60 nM (Fig. 3A and B). By 2 weeks in culture, neurons on astrocytes did not respond to GABA with an elevation in ${\text{Ca}}_{\text{c}}^{\text{2+}}$ while one-third of those grown on PDL still manifested ${\text{Ca}}_{\text{c}}^{\text{2+}}$ signals with a mean about 60 nM. Thus, relative to cultivation of ventral spinal neurons on PDL, co-culture of neurons on astrocytes accelerated the developmental disappearance in the cellular distribution of GABA-induced ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses.

Astrocytes enhance developmental changes in E_GABA that involve diffusible factors

We used gramicidin-perforated patch recordings to record the resting membrane potentials and equilibrium potential for GABA-induced conductance and polarization (E_GABA). GABA depolarized neurons cultured for 1 day on either PDL or astrocytes in a bicuculline-sensitive manner (n = 7) (Fig. 4A). Bicuculline by itself did not alter resting membrane potential. Depolarizing responses evoked under these experimental conditions usually did not provoke regenerative action potential activity (Fig. 4A) except in several neurons cultured on PDL for 1-2 weeks (data not shown). We recorded brief pulses to GABA under voltage clamp in normal saline and low ${\text{Cl}}_{\text{o}}^{\text{−}}$-containing saline. In the cell illustrated (Fig. 4B and C), GABA-induced current reversed polarity at approximately -38 mV in normal saline and -2 mV in low ${\text{Cl}}_{\text{o}}^{\text{−}}$, consistent with a Nernstian shift in E_GABA. Similar results were obtained in six out of six cells tested.

Figure 4. GABA depolarizes cells in gramicidin-perforated patch recordings via activation of GABA_A receptor/Cl⁻ channels.

A, a depolarizing response to GABA was recorded under current clamp (I = 0) and could be reversibly blocked by 100 μM bicuculline. Note that neither depolarizing response triggers an action potential. B, the reversal in polarity of superimposed current responses activated by 10 μM GABA (bar at bottom) occurs at different potentials in the two salines. C, current-voltage curves constructed from the data shown in B(same cell) reveal the Cl⁻dependency of the reversal potential in current polarity.

We recorded steady-state membrane potential and GABA-induced polarization of neurons cultured on PDL or on astrocytes using gramicidin-perforated patch-recording techniques. Resting membrane potential recorded in this manner progressively increased from an initial level of about -40 mV within hours of plating to about -55 mV at 1 week in vitro in all neurons studied (Fig. 5A; PDL, -40 ± 0.6 to -56 ± 1.0 mV; astrocytes, -42 ± 1.6 to -55 ± 1.6 mV). E_GABA was about -30 mV at 1 day in culture in all neurons recorded (Fig. 5B). Astrocytes accelerated the time course of changes in E_GABA, which hyperpolarized to -60 ± 2.5 mV by 1 week in culture, eventually reaching -64 ± 1.0 mV by 3 weeks. In contrast, E_GABA recorded in neurons on PDL, was between -48 ± 0.8 mV by 1 week, eventually reaching -53 ± 1.8 mV by 3 weeks (Fig. 5B). Medium conditioned by astrocytes (ACM) completely mimicked the initial rate of change in E_GABA recorded in co-cultures of neurons with astrocytes (Fig. 5B). However, the effectiveness of ACM in promoting hyperpolarization became limiting such that E_GABA levelled off at -58 ± 1.4 mV, a value intermediate between those recorded in the two other conditions. When the data were plotted relative to the resting potential, both astrocytes and ACM stimulated changes in E_GABA that, by 4 days in vitro, led to no potential response or hyperpolarization relative to the resting potential. Hyperpolarizing responses took 3 weeks to manifest in neurons on PDL alone (Fig. 5C). Intracellular chloride, [Cl⁻]_i (estimated from E_GABA), was calculated to change from ∼45-50 mM at the time of plating to ∼12-15 mM after 4-7 days in culture in neurons on astrocytes or in ACM. A similar decrease in [Cl⁻]_i was evident in neurons on PDL but took 3 weeks to drop to ∼20 mM (Fig. 5D).

GABA_A receptor/Cl⁻ channel properties change in neurons independently of astrocytes

We used the whole-cell patch-clamp recording technique with Cl⁻-filled pipettes to set E_Cl near or at 0 mV and clamped cells at -80 mV to optimize 10 μM GABA-induced Cl⁻-dependent current signals. Macroscopic GABA-induced currents were always superimposed with microscopic fluctuations characteristic of underlying Cl⁻ channel activity (Fig. 6A and B). We used fluctuation analysis methods to estimate the elementary properties of the Cl⁻ channels activated by GABA in the whole cell (Neher & Stevens, 1977). At the recording bandwidth used under our experimental conditions we consistently calculated two components in every spectrum of current fluctuations in all neurons (Figs 6 and 7). We computed the relative areas of the two components and found that the long-lasting component accounted for about 80-85 % of the spectrum in all neurons at all time points (Fig. 7B). At day 1 for example, PDL = 82 ± 3.5 %, astrocytes = 79 ± 4.3 %, ACM = 79 ± 1.7 %; and by day 21, PDL = 77 ± 3.0 %, astrocytes = 81 ± 2.1 %, ACM = 80 ± 2.0 %; n = 6-8 cells. Since the estimated mean channel open time, or burst duration, of the greater component (τ_long) was considerably longer than the shorter component at all times in culture, more than 95 % of the pharmacologically induced current reflects Cl⁻ channel activity with these kinetics. τ_long decreased progressively in all neurons from ∼75 ms in acutely recorded cells to ∼50 ms in cells cultured for 3 weeks (Fig. 7A). In contrast, a short-lasting component averaging 3-4 ms was recorded in all neurons at all times (Fig. 7D). The estimated elementary conductance progressively increased from ∼16 to ∼22 pS over the first week in culture in all neurons recorded (Fig. 7C). These results demonstrate that unitary GABA_A receptor/Cl⁻ channel properties inferred from analysis of fluctuations in GABA-evoked Cl⁻ current recorded from whole cells change in a concerted manner primarily during the first week in vitro and that these changes occur independently of astrocytes.

Figure 6. The long-lasting component of GABA-activated Cl⁻ channel activity shortens over time in culture.

Cells were recorded within hours of plating (0 Div) or after 7 days in culture (7 Div). At both times GABA macroscopic currents superimposed with microscopic fluctuations. Representative traces reveal the presence of low frequency fluctuations at 0 Div (A), which are less apparent at 7 Div (B). Spectral analysis of the fluctuations show that both spectra can be adequately explained with two components indicated by the corner frequencies, f_c1 and f_c2. The mean open times (τ) of the underlying channel activity estimated from τ = (2πf_c)⁻¹ shorten from ≈70 ms to ≈50 ms for the long-lasting openings, but remain at about 3 ms for the short-lasting component.

DISCUSSION

The present study shows that: (1) pharmacological activation of GABA_A receptor/Cl⁻ channels depolarizes cultured ventral spinal cord neurons and triggers sustained elevation of ${\text{Ca}}_{\text{c}}^{\text{2+}}$; (2) both responses disappear in long-term culture, with GABA_A receptor/Cl⁻ channel activation eventually hyperpolarizing cells relative to the resting potential; (3) astrocytes in co-culture and medium conditioned by them for 24 h accelerate these changes relative to the rate recorded in neurons cultured on PDL; and (4) the unitary properties of GABA_A receptor/Cl⁻ channels change independently of astrocytes, as channels shorten in open time and increase in conductance. Thus, in vitro, astrocytes regulate changes in the Cl⁻ gradient extrinsic to GABA_A receptor/Cl⁻ channels without modulating their intrinsic properties, which change in parallel.

GABA depolarizes embryonic neurons

Depolarization via GABA_A receptor/Cl⁻ channels has been recorded in cells dissociated from the cervical region of the embryonic rat spinal cord using dye potentiometry (Walton et al. 1993). Depolarizing responses have also been recorded electrically using perforated-patch techniques in the majority of dorsal horn neurons plated at E15-16 and studied during the first week in culture (Wang et al. 1994). Embryonic, rat lumbar motoneurons in slices of spinal cord prepared from E16 and then recorded at PN2 with high resistance microelectrodes also depolarized in response to GABA (Wu et al. 1992). In the latter preparation, depolarizing synaptic potentials mediated by GABA can be evoked in motoneurons by electrical stimulation of primary afferents in dorsal roots. These depolarizing GABAergic potentials did not trigger action potentials. This may be due to the peak of the depolarizing potential lying subthreshold to action potential activation as well as to the associated conductance increase to Cl⁻ ions. Depolarizations evoked pharmacologically in perforated-patch recordings of cultured dorsal horn neurons (Wang et al. 1994) or ventral cervical cord neurons (this study) only rarely triggered action potentials. However, in on-cell patch-clamp recordings of spinal neurons, GABA's activation of GABA_A receptor/Cl⁻ channels enclosed in the patch triggered action potentials and promoted their occurrence in the remainder of the membrane (Serafini et al. 1995). These excitatory effects of single-channel openings may result from charge transfer throughout the intact membrane that effectively depolarizes the cell outside the patch. Thus, differences in experimental conditions and/or electrical recording configurations may account for the variability in recording action potentials.

The Cl⁻-dependent depolarization of embryonic and early postnatal spinal cord neurons, and presumably embryonic neurons from brain as well, is due to high intracellular [Cl⁻]_i and disappears with age as the Cl⁻ gradient across the membrane increases. In spinal motoneurons GABA generates membrane depolarization until at least 1-2 days after birth but it has not been determined when the response becomes hyperpolarizing (Wu et al. 1992). CA3 and CA1 neurons of the hippocampus show the transition from Cl⁻-dependent depolarizations to hyperpolarizations at the beginning of the second postnatal week (Ben-Ari, Cherubini, Corradetti & Gaiarsa, 1989). In the present study the number of ventral spinal cord neurons exhibiting depolarizations to muscimol (after a 24 h recovery) dropped from ∼80 % at E17 to ∼40 % at PN0/PN7 and to 4 % at PN14. Neurons dissociated at E15 and grown on astrocytes or PDL exhibit the transition to GABA-induced hyperpolarizations after 4 days in vitro or 3 weeks in vitro, respectively. These results indicate that the transition occurring in vitro follows roughly the same time course as has been seen for spinal and hippocampal neurons in vivo.

Cl⁻-dependent depolarization activates Ca²⁺ entry

Previous studies on cultured embryonic dorsal horn neurons have revealed that GABA evokes a sustained elevation in ${\text{Ca}}_{\text{c}}^{\text{2+}}$, which has been attributed to voltage-dependent Ca²⁺ channels activated by depolarization (Reichling et al. 1994). In this study, we have found similar effects of GABA_A receptor/Cl⁻ channel activation on ${\text{Ca}}_{\text{c}}^{\text{2+}}$ levels in embryonic ventral cervical spinal cord neurons. The sensitivity of the ${\text{Ca}}_{\text{c}}^{\text{2+}}$ response to nifedipine in the present study implicates voltage-dependent L-type Ca²⁺ channels as the primary pathway for Ca²⁺ entry. Thus, GABA induces ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses indirectly via its Cl⁻-dependent depolarizing actions. Quite similar conclusions regarding L-type Ca²⁺ channel involvement in GABA-evoked ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses have been reported in cultured embryonic hypothalamic neurons (Obrietan & van den Pol, 1995). Perforated-patch recordings of hypothalamic neurons cultured for several days show that in about half of the cells GABA evoked depolarizing responses that triggered one or several action potentials. Bicuculline blocked many of these action potentials occurring spontaneously together with most, but not all of the depolarizing synaptic activity (Chen, Trombley & van den Pol, 1996). The coupling of bicuculline-sensitive, GABA-mediated depolarization and ${\text{Ca}}_{\text{c}}^{\text{2+}}$ elevation has also been reported in recordings of cultured rat hippocampal neurons (Segal, 1993), pyramidal neurons in early postnatal hippocampal slice preparations (Leinekugel, Tseeb, Ben-Ari & Bregestovski, 1995) and in tissue print preparations of embryonic neocortex (Owens, Boyce, Davis & Kriegstein, 1996). In the hippocampal slice preparation, the coupling between GABA_A receptor/Cl⁻ conductance, depolarization and activation of voltage-dependent Ca²⁺ channels disappears during the postnatal period coincident with a progressive shift in E_Cl to more hyperpolarized potentials.

Astrocyte effects

Changes in the functional effects of GABA_A receptor activation recorded in slice preparations during the postnatal period parallel the developmental appearance of astrocytes in the hippocampus (Nixdorf-Bergweiler, Albrecht & Heinemann, 1994). In the developing rat spinal cord, GFAP⁺ elements appear at E16 and their numbers increase during the late embryonic and early postnatal period (Yang et al. 1993). In the present study, accelerated changes in the polarity of the GABA response and the progressive loss of a concomitant elevation in ${\text{Ca}}_{\text{c}}^{\text{2+}}$ occur in ventral spinal cord neurons dissociated at E15 and cultured over a 3 week period in the presence of astrocytes. This is the same time period during which these neurons would be subject to the influence(s) of proliferating astrocytes in vivo. Hence, in vivo astrocytes could play a role in regulating these developmental changes. There have been numerous studies on the profound effects of astrocytes on neuronal proliferation, morphology, electrophysiological properties and homeostasis (Silver, Lorenz, Wahlsten & Coughline, 1982; Sivron, Eitan, Schreyer & Schwartz, 1993; Travis, 1994; Wu & Barish, 1994; Barish, 1995; Sontheimer, 1995; Tsacopoulos & Magistretti, 1996).

The effects of astrocytes on GABA-induced, chloride-dependent potential in vitro were mimicked by exposure to medium conditioned for 24 h by astrocytes. Thus, confluent astrocytes derived from postnatal cortex secrete substances independently of the presence of embryonic neurons that serve to signal developmental changes in E_Cl and, in turn, ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses. There are several possible targets for astrocyte factor regulation that could influence levels of intracellular Cl⁻. These include a Cl⁻/HCO₃⁻ exchanger driven by the intracellular production of CO₂/HCO₃⁻ (Kobayashi, Morgans, Casey & Kopito, 1994), a Cl⁻/cation transporter driven by the transmembrane gradient for Na⁺ (Misgeld, Deisz, Dodt & Lux, 1986; Rohrbough & Spitzer, 1996), or an ATP-dependent Cl⁻ pump (Inagaki, Hara & Inoue, 1992). Interestingly, brain astrocytes and their conditioned media have recently been shown to stimulate Na⁺–K⁺-Cl⁻ cotransporter activity in brain endothelial cells (Sun, Lytle & O'Donnell, 1997). In the latter study the cytokine, IL-6 appeared to be the stimulatory factor. Further work should help to identify which transporter, if any, is a target for astrocyte regulation in developing spinal cord neurons.

Developmental changes in GABA_A receptor/Cl⁻ channel properties

We found that the unitary properties of GABA_A receptor/Cl⁻ channels inferred from fluctuation analyses of Cl⁻ current responses changed in parallel with the hyperpolarization of E_Cl and loss of GABA-induced ${\text{Ca}}_{\text{c}}^{\text{2+}}$ responses. However, these changes occurred with the same time course whether or not neurons were cultured with astrocytes. Thus, in vitro, astrocytes do not regulate the intrinsic biophysical properties of GABA_A receptor/Cl⁻ channels. The developmental changes in Cl⁻ channel properties could involve a switch in the subunit composition of the putative pentameric GABA_A receptors. In this regard, we have carried out a preliminary study of GABA_A receptor subunit transcript expression using in situ hybridization applied to cells dissociated at E17 and P12 (data not shown). We found cells expressing transcripts for α2, α3, α4, α5, β1, β2, β3, γ1 and γ2 in E17 dissociates, and in P12 dissociates, transcripts for α2, α3, α4, β2, β3 and γ2, consistent with a previous study in vivo (Ma et al. 1993), were found.

It is significant that GABA_A receptor subtypes have been shown to change during development (Poulter, Barker, O'Carroll, Lolait & Mahan, 1992; Fritschy, Paysan, Enna & Mohler, 1994; Chang, Luntz-Leybman, Evans, Rotter & Frostholm, 1995; Ma et al. 1993) and different combinations of subunits are associated with different physiological properties (Mathews et al. 1994; Saxena & Macdonald, 1994). Hence, the shortening in the open time derived from the major component in the spectral analyses of GABA-induced fluctuations and the coincident increase in estimated unitary conductance occurring over time in vitro may reflect changes in subunit composition. This remains to be determined.

Conclusion

In summary, astrocytes facilitate developmental changes in the Cl⁻ gradient of spinal neurons extrinsic to GABA_A receptor/Cl⁻ channels that profoundly influence the functional effects of GABA_A receptor/Cl⁻ channel activation. These astrocyte effects are mediated by diffusible substances and occur independently of parallel changes in the intrinsic properties of the receptor.