Spinal dorsal horn astrocytes release GABA in response to synaptic activation

Abstract

Key points

• GABA is an essential molecule for sensory information processing. It is usually assumed to be released by neurons.

• Here we show that in the dorsal horn of the spinal cord, astrocytes respond to glutamate by releasing GABA.

• Our findings suggest a novel role for astrocytes in somatosensory information processing.

Abstract

Astrocytes participate in neuronal signalling by releasing gliotransmitters in response to neurotransmitters. We investigated if astrocytes from the dorsal horn of the spinal cord of adult red‐eared turtles (Trachemys scripta elegans) release GABA in response to glutamatergic receptor activation. For this, we developed a GABA sensor consisting of HEK cells expressing GABA_A receptors. By positioning the sensor recorded in the whole‐cell patch‐clamp configuration within the dorsal horn of a spinal cord slice, we could detect GABA in the extracellular space. Puff application of glutamate induced GABA release events with time courses that exceeded the duration of inhibitory postsynaptic currents by one order of magnitude. Because the events were neither affected by extracellular addition of nickel, cadmium and tetrodotoxin nor by removal of Ca²⁺, we concluded that they originated from non‐neuronal cells. Immunohistochemical staining allowed the detection of GABA in a fraction of dorsal horn astrocytes. The selective stimulation of A∂ and C fibres in a dorsal root filament induced a Ca²⁺ increase in astrocytes loaded with Oregon Green BAPTA. Finally, chelating Ca²⁺ in a single astrocyte was sufficient to prevent the GABA release evoked by glutamate. Our results indicate that glutamate triggers the release of GABA from dorsal horn astrocytes with a time course compatible with the integration of sensory inputs.

Key points

• GABA is an essential molecule for sensory information processing. It is usually assumed to be released by neurons.

• Here we show that in the dorsal horn of the spinal cord, astrocytes respond to glutamate by releasing GABA.

• Our findings suggest a novel role for astrocytes in somatosensory information processing.

Introduction

Astrocytes are present throughout the CNS. In addition to potassium and neurotransmitter clearance, they actively contribute to synaptic integration. Indeed, astrocytes respond to neurotransmitters such as glutamate or GABA with an increase in intracellular Ca²⁺ concentration (Porter & McCarthy, 1996; Fellin & Carmignoto, 2004) that triggers the release of gliotransmitters acting both pre‐ and postsynaptically (Araque et al. 1999, 2014). This interaction, which led to the concept of the tripartite synapse (Araque et al. 1999, 2014), has been reported in several regions of the CNS, including the hippocampus (Perea & Araque, 2007; Henneberger et al. 2010), cerebellum (Brockhaus & Deitmer, 2002), neocortex (Halassa et al. 2009) and spinal cord (Carlsen & Perrier, 2014). It is now widely accepted that astrocytes release glutamate, ATP, d‐serine or S100β (Zhang et al. 2003; Perea & Araque, 2007; Henneberger et al. 2010; Araque et al. 2014; Morquette et al. 2015). A few studies also suggest that GABA is released from astrocytes (Lee et al. 2010, 2011; Le Meur et al. 2012; Jo et al. 2014; Woo et al. 2018). This idea is supported by the fact that astrocytes store GABA, as shown by immunohistochemistry (Blomqvist & Broman, 1988; Bull & Blomqvist, 1991; Gonzalo‐Ruiz et al. 1993). In addition, the expression of glutamic acid decarboxylase (GAD) suggests that they can synthetize GABA (Martinez‐Rodriguez et al. 1993). Moreover, astrocytes express GAT‐1 and GAT‐3 isoforms of GABA transporters (Durkin et al. 1995; Minelli et al. 1995; De Biasi et al. 1998), which take up GABA released from inhibitory synapses. Under physiological conditions, the reversal potential of GAT is so close to the resting membrane potentials that marginal changes of intracellular GABA concentration or membrane potential are sufficient to reverse the direction of GABA transport (Richerson & Wu, 2003). In line with this result, it was demonstrated that GAT‐1 from Bergmann glial cells artificially loaded with a high concentration of GABA produced a tonic efflux of the amino acid (Barakat & Bordey, 2002). A tonic release of GABA from Bergmann glia was also ascribed to the volume‐ and Ca²⁺‐activated anion channel Bestrophin 1 (Lee et al. 2010). In the olfactory bulb studied under physiological conditions, the mechanical stimulation of astrocytes induced a phasic release of GABA, which by activating GABA_A receptors from different neurons promoted their synchrony (Kozlov et al. 2006). Together these studies suggest that at least some astrocytes can release GABA (Kozlov et al. 2006).

In the dorsal horn of the spinal cord, primary afferent fibres are under a tonic inhibitory control mediated by GABA_A receptors (Sivilotti & Woolf, 1994; Takazawa & MacDermott, 2010). Interestingly, dorsal horn astrocytes respond to nociceptive input by coordinated calcium excitation (Sekiguchi et al. 2016). Studies have also shown that chronic pain is correlated with the appearance of reactive astrocytes (Zhuang et al. 2006; Hansen & Malcangio, 2013) and that reactive astrocytes release more gliotransmitters (Agulhon et al. 2012). These observations prompted us to assess whether astrocytes from the dorsal horn of the adult spinal cord release GABA. By combining patch‐clamp recordings, Ca²⁺ imaging and GABA detection with a custom‐made sensor, we show that dorsal horn astrocytes respond to glutamate by releasing GABA in the extracellular space.

Methods

Ethical approval

The surgical procedures complied with Danish legislation and were approved by the controlling body under The Ministry of Justice.

Slice preparation

Adult turtles (Trachemys scripta elegans) were obtained from Nasco, Fort Atkinson, WI, USA. Animals had free access to water and were fed once a day. Turtles were anaesthetized by intravenous injection of propofol (0.3 ml/100 g; Propolipid Frenesius Kabi, Sweden) and killed by decapitation. The spinal cord was removed after intracardiac perfusion of a high Mg²⁺ solution (in mm: 120 NaCl, 5 KCl, 15 NaHCO₃, 20 glucose, 20 MgCl₂, 3 CaCl₂). All experiments were performed at room temperature (20–22°C). The cervical enlargement (C1–C8) and lumbar enlargement (D8–S2) and slices (300–1500 μm thick) were cut with a vibratome (MicroM slicer HM 650 V equipped with cooling unit CU65 set at 2°C). For some experiments, a dorsal root filament was left in continuity with the slice and mounted on a suction electrode. The preparation was continuously perfused with Ringer's solution (in mm): 120 NaCl, 5 KCl, 15 NaHCO₃, 20 glucose, 2 MgCl₂, 3 CaCl₂, saturated with 98% O₂ and 2% CO₂ to obtain a pH of 7.6.

Calcium imaging

Sulforhodamine 101 (SR101, Sigma‐Aldrich, St Louis, MO, USA) was dissolved in dimethyl sulfoxide plus Pluronic F‐127 (BASF Global, Ludwigshafen, Germany) and diluted in Ringer's solution to a concentration of 7 μm. The membrane‐permeant Ca²⁺ indicator Oregon Green BAPTA‐1/AM (Invitrogen, Carlsbad, CA, USA) was dissolved in dimethyl sulfoxide with Pluronic F‐127 (BASF Global) and diluted in Ringer's solution at a concentration of 0.8 mm. Both solutions were puff applied in the dorsal horn of spinal cord slices. We verified the specificity of SR101 for astrocytes by immunohistochemistry (see below). The slices had an intact dorsal root filament and were preincubated in TTX (50 nm). After a resting time > 1h, fluorescent cells were observed under a 20× (1.0 numerical aperture) water‐immersion objective of an SP5 multiphoton/confocal laser scanning microscope (Leica, Wetzlar, Germany) equipped with a Mai Tai HP Ti:Sapphire laser (Millennia Pro, Spectra Physics, Sweden). The excitation light was set to 800 nm. The emitted light was filtered to retain red and green signals using a TRITC/FITC filter. Frames were acquired at sample rates of 11.3–11.8 Hz. Each measurement was repeated three times at 5 min intervals.

Image analysis

The responses of cells loaded with SR101+ and Oregon Green BAPTA to dorsal root filament stimulation were quantified with the public domain ImageJ software (National Institutes of Health). Regions of interest (ROIs) were defined offline. The average amplitude of each ROI was calculated for each frame and exported in Origin 8.6 (Origin Lab corporation, Northampton, MA, USA) for further analysis. The values obtained from three consecutive measurements were averaged. Potential bleaching was corrected by subtracting the linear regression of the signal. The relative change of fluorescence signal ΔF/F was calculated as (F _(t) − F _rest)/F _rest, where F _rest is the mean value of the signal calculated before any stimulation. A response was considered significant when it exceeded twice the standard deviation of the signal at rest. Signals were smoothed with a percentage filter (5 points).

Patch clamp recordings

Visually‐guided patch clamp recordings were performed with a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA, USA) in current and voltage clamp modes. Cells were visually identified in the slices by means of a BX51WI microscope (Olympus, Tokyo, Japan) equipped with an oblique illumination condenser or with a differential interference contrast system. The pipette solution (in mm: 122 potassium gluconate, 2.5 MgCl₂, 5.6 magnesium gluconate, 5 K‐Hepes, 5 H‐HEPES, 5 Na₂ATP, 1 EGTA, 2.5 biocytin, KOH to adjust the pH to 7.4) contained the fluorescent dye Alexa 488 (250 μm, Sigma‐Aldrich) to stain the recorded cells and visualize their processes. Calcium‐clamp pipette solution contained (in mm): 40 potassium gluconate, 30 K₄‐BAPTA, 50 sucrose, 2.5 MgCl₂, 5.6 magnesium gluconate, 5 K‐Hepes, 5 H‐Hepes, 5 Na₂ATP, 1 EGTA, 2.5 biocytin, 0.25 Alexa 568 hydrazide sodium salt and the necessary amount of KOH to adjust the pH to 7.4. Osmolarity was measured to 275 mm. Free [Ca²⁺] was calculated to 0.0017 nm using WEBMAX chelator (Stanford.edu). Electrodes had a resistance ranging from 4 to 8 MΩ. Recordings were sampled at 10–20 kHz. Data were sampled with a 16‐bit analog‐to‐digital converter (DIGIDATA 1440, MDS) and displayed by means of Clampex 10.2 software (MDS).

Focal application of drugs

Electrodes made from borosilicate capillaries (G150F‐3; Warner Instruments, Hamden, CT, USA) (tip diameter < 1.5 μm) were filled either with GABA (1 mm in Ringer's solution) or glutamate (1 mm in Ringer's solution). Drugs were pressure‐ejected at 14–34 Pa by a homemade time‐controlled pressure device.

GABA sniffer

Human Embryonic Kidney cells (HEK) 293 constitutively expressing GABA_A receptors (gift from NeuroSearch A/S, Ballerup, Denmark) were grown in T75 flasks (VWR, Herlev, Denmark) in full Dulbecco's modified Eagle medium (DMEM) (DMEM supplemented with 100 U/ml penicillin, 100 mg/ml streptomycin and 10% fetal calf serum; Sigma‐Aldrich, Copenhagen, Denmark) at 37°C, in a humidified atmosphere with 5% CO₂. Confluent cells were rinsed with phosphate buffered saline (in‐house, University of Copenhagen), trypsinized (1%, Sigma‐Aldrich) and plated on 9 mm coverslips (VWR). Cells were allowed to settle on coverslips for a minimum of 30 min before being transferred to a recording chamber, where they were positioned next to a spinal cord slice. HEK cells were recorded in whole‐cell or outside‐out configurations in the voltage‐clamp mode. The recording pipette was mounted on a three‐axis motorized micromanipulator (Luigs and Neumann, Ratingen, Germany). Their membrane potential was held at 0 mV (the driving force for chloride was around 50 mV). The sensitivity of HEK cells to GABA was systematically tested by puffing GABA (1 mm). Responsive cells (95% of the total) were moved to the dorsal horn of the slice.

Online identification of astrocytes

Astrocytes were identified on the basis of their morphology and membrane properties. The cells had a diameter below 10 μm. Their processes visualized after dialysis of Alexa 488 were restricted to a spherical volume. They had an average membrane potential of –76.1 ± 4.8 mV (after correction of the liquid junction potential), and input resistance of 138 ± 94.2 MΩ (median ± SD; n = 14). Astrocytes were voltage clamped from –90 to +50 mV with incremental steps of 10 mV. Current voltage relationships were either linear (n = 11/14) or displayed a minor inward rectification (n = 3/14).

Immunohistochemistry

Animals were anesthetized (50 mg/kg pentobarbital; i.p.) and fixed by intracardiac perfusion with 4% paraformaldehyde and 0.5% glutaraldehyde in 0.1 m phosphate buffer (PB, pH 7.4). Tissues were sectioned with a vibrating microtome (60–80 μm, Leica VT1000S) and placed in PB with 0.5% bovine serum albumin for 30 min and then incubated with the primary antibodies diluted in PB with 0.3% Triton X‐100 (Sigma‐Aldrich). The slices were then incubated with anti‐GABA (rabbit polyclonal, 1:100; # 20094, Immunostar, Hudson, WI, USA) or anti‐S100 (β subunit) (mouse monoclonal, 1:500, #S2532, Sigma‐Aldrich or S‐100 polyclonal antibody produced in rabbit, Sigma‐Aldrich; 1:500). After washing in PB, tissues were incubated in secondary antibodies conjugated with Alexa 488, 633 or 405 (Invitrogen). Slices were then mounted in anti‐fade medium (Vectashield, Vector Laboratories, Burlingame, CA, USA). Control experiments were performed by omitting or replacing primary or secondary antibodies with pre‐immune normal serum. The material was visualized by confocal microscopy (FV 300; Olympus or LSM 700; Zeiss, Oberkochen, Germany) and images were acquired with Fluoview 5 (Olympus) or ZEN pro 2012 (Zeiss). Orthogonal projections were used to test co‐localization. Fluorescence intensity profiles were obtained with ImageJ (NIH).

For co‐staining with S100β and SR101, slices were bulk‐loaded with SR101 (5 μm) dissolved in pluronic F‐127 acid (0.25%) and DMSO (5%). The tissue was then snap frozen and incubated with the primary anti‐S100β antibody for 24 h and then with a secondary antibody for 30 min. Slices were then mounted and imaged immediately.

In situ hybridization

Coronal cryostat sections (12 μm) of the spinal cord from three different animals were mounted on Superfrost Plus slides. Sections were hybridized as previously described (Rath et al. 2013) with a DNA oligo probe specific for GAT3 (5′‐CAAGTAGAATTTTATTCCTTCAGAAGCCCCTGGTAA‐3′) corresponding to position m.1144–1109 on SLC6A11 (XM_005279910.2). After hybridization and washing, the sections were covered with a photographic emulsion (Kodak, Rochester, NY, USA) and exposed for 6 weeks. The emulsion was developed in amidol and the sections were counterstained in cresyl violet.

Drugs

Glutamate, GABA, CdCl₂, NiCl₂ and sulforhodamine101 were from Sigma‐Aldrich (St. Louis, MO, USA). MNI‐caged l‐glutamate, K₄‐BAPTA, 6‐cyano‐7‐nitroquinoxaline‐2,3‐dione (CNQX), dl‐2‐amino‐5‐phosphonopentanoic acid (AP5), SR 95531 hydrobromide (gabazine) and picrotoxin were from Tocris Bioscience (Bristol, UK). TTX was from Alomone Labs (Jerusalem, Israel).

Statistical analysis

Data were analysed by means of Origin 8.6 (Origin Lab, Northampton, MA, USA) and Matlab (MathWorks, Natick, MA, USA). The normality of distribution of each sample was tested with a one‐sample Kolmogorov–Smirnov test. Non‐parametric tests were used when a normal distribution could not be approximated. Data are represented as mean ± SEM or median with 10, 25, 75 and 90 quartiles in case of a non‐parametric distribution. Statistical significance was assessed by Student's t test, non‐parametric Wilcoxon signed‐rank test, Wilcoxon rank‐sum test or Mood's median test (^* P<0.05, ^** P<0.01, ^*** P<0.001).

Results

We investigated whether mature dorsal horn astrocytes release GABA in a slice preparation from the spinal cord of adult turtles (Trachemys scripta elegans). We chose this model for its exceptional resistance to anoxia (Buck et al. 2012). In contrast to mammals, it allows reliable experiments performed in the adult spinal cord. This unique resistance to hypoxia allows the use of large in vitro preparations of fully mature nervous tissue comprising one or more spinal cord segment with dorsal roots attached (Hounsgaard et al. 1988; Russo et al. 2000; Perrier & Delgado‐Lezama, 2005; Cotel et al. 2013). To monitor the release of GABA, we designed a sensor consisting of HEK293 cells expressing α1, β2 and ɣ2 subunits of the GABA_A receptor.

HEK cells recorded in voltage‐clamp mode transduced extracellular GABA into an electrical current (Christensen et al. 2014). When held at 0 mV, the GABA sniffer responded to brief GABA puff (1 mm) by strong transient outward currents that were blocked by gabazine (10 μm) and picrotoxin (50 μm) (Fig. 1 A; n = 3). By contrast, glutamate did not induce any response (Fig. 1 B; n > 20). The sniffer remained very stable over time as repeated puffs of GABA produced similar responses for more than 4 h (Fig. 1 C; n = 3). By varying the concentration of GABA in the puff pipette, we found that the responses increased monotonically with GABA and saturated at concentrations above 1 mm (Christensen et al. 2014). These results demonstrate that our sensor can be used to detect GABA transients. Next, we tested if the sniffer was sensitive enough to detect GABA released from astrocytes in the spinal cord. After positioning the sniffer on the surface of the dorsal horn of a slice preparation, we moved it carefully down to a depth of 10–50 μm. To prevent the release of transmitters from neurons, we blocked all voltage‐gated Ca²⁺ channels with non‐selective blockers (CdCl₂, 0.5–1 mm; NiCl₂, 0.1–0.5 mm). We also prevented the influx of Ca²⁺ via NMDA receptors by blocking them with AP5 (50 μm). Under these conditions, a glutamate puff (0.5–4 s) evoked strong outward currents in the sniffer (Fig. 2 A, B; n = 7 slices). Neither addition of the Na⁺ channel blocker TTX (500 nm; tested in 5 out of the 8 cells), nor removal of the extracellular Ca²⁺ (tested in 3 out of the 7 cells) prevented the response (Fig. 2 A). The amplitude and frequency of events were not affected by the compounds (Fig. 2 D: frequency of events per trial mean compounds: 0.86 ± 1.56, mean control: 0.55 ± 0.74, P = 0.170, unpaired Student's t test; amplitude compounds: 248 ± 392 pA, amplitude control 180 ± 200.0 pA SEM, P = 0.396, unpaired t test). The high variability of the events under our experimental conditions may reflect a low probability of GABA release. These observations suggest that glutamate evoked the release of GABA, either from neurons through an unknown Ca²⁺‐independent mechanism or from non‐spiking cells such as glial cells. Interestingly, the glutamate‐induced events decayed very slowly compared to spontaneously occurring IPSCs (Fig. 2 C, D; n = 6 slices). On average, the glutamate evoked responses decayed with a time constant of 260 ± 307 ms (Fig. 2 D), which is one order of magnitude slower than the average for IPSCs (17 ± 9.8 ms; P = 8.96 × 10⁻¹¹, unpaired Student's t test). This result strengthens the idea that GABA was not released by a vesicular synaptic mechanism. Altogether, our data suggest that glutamate evoked the release of GABA from non‐spiking cells.

Figure 1. Detection of ambient GABA with a sniffer.

A, whole‐cell patch‐clamp recording of a HEK cell expressing α1, β2 and γ2 subunits of the GABA_A receptor. Recording in voltage‐clamp mode with a holding potential of 0 mV. A 1 s puff of GABA (1 mm) evoked a strong outward current that was abolished in the presence of gabazine (10 μm) and picrotoxin (50 μm). B, left: response induced by a 1 s puff of GABA (1 mm). Right: a 1 s puff of glutamate (1 mm) did not evoke any response. C, amplitude of the current peaks evoked by repetitive GABA puffs (1/min) as a function of time. The response remained stable over 4 h. Addition of gabazine (10 μm) and picrotoxin (50 μm) abolished the response. [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 2. Glutamate evoked GABA release in the dorsal horn of the spinal cord.

A, GABA sniffer positioned on the dorsal horn of a spinal cord slice preparation. Glutamate pipette (1 mm) positioned near the sniffer. Synaptic transmission blocked by CdCl₂ (0.5–1 mm), NiCl₂ (0.1–0.5 mm); NMDA receptors blocked by AP5 (50 μm). In 3/8 cells (right column) Na⁺ channels were blocked by TTX (500 nm) and after removal of Ca²⁺ from the extracellular medium. Glutamate puffs (0.5–4 s) evoked outward currents in the GABA sniffer. B, in another experiment, the glutamate evoked events were blocked by gabazine (10 μM). C, spontaneously occurring IPSCs in a dorsal horn interneuron recorded in voltage‐clamp mode (V _h = 0 mV). D, left: superimposed glutamate‐evoked events recorded in the sniffer with the compounds used in A or in normal Ringer's solution. Right: examples of IPSCs recorded in C. The boxplots indicate the median, first and fourth quartiles with 10–90% percentile whiskers while circles indicate the means. Number of events per trial mean compounds: 0.86 ± 1.56, mean control: 0.55 ± 0.74, P = 0.170, unpaired Student's t test; amplitude compounds: 248.2 ± 392 pA, amplitude control: 180 ± 200.0 pA (no significant difference; P = 0.396, unpaired Student's t test). IPSC amplitude: 55.4 ± 54.1 pA (significantly smaller than events recorded with compounds: P = 9.7 × 10⁻⁶, unpaired t test). Rise time (from 20 to 80% of peak): mean rise time compounds: 72.2 ± 162.2 ms, rise time control: 91.5 ± 152.3 ms (no significant difference: P = 0.726, unpaired Student's t test). Rise time IPSCs: 6.8± 11.6 ms (significantly shorter than events measured with compounds: P = 1.19⋅10⁻⁵, unpaired Student's t test). Event decay time constant: mean tau compound: 260 ± 307 ms, mean tau control: 447.1 ± 480.0 ms (no difference: P = 0.073, unpaired Student's t test). Mean IPSC tau: 17 ± 9.8 ms (significantly shorter than tau for events with compounds: P = 8.9 × 10⁻¹¹, unpaired Student's t test). [Color figure can be viewed at http://wileyonlinelibrary.com]

To determine if GABA was released from astrocytes, we performed three different sets of experiments. First, we verified that astrocytes contain GABA by means of immunohistochemistry. We combined the labelling of GABA and astrocytes with the astrocyte‐specific marker S100β. We found an intense staining for GABA in cells that were negative for S100β (asterisks in Fig. 3 A), presumably corresponding to inhibitory interneurons. In addition, a fraction of the S100β+ cells (33 out of 164, i.e. 20%) was positive for GABA (Fig. 3 A). These results indicate that a subpopulation of dorsal horn astrocytes contain GABA. We also verified that, as in the mammalian brain, GAT‐3 was specific for glial cells in the turtle spinal cord by performing in situ hybridization with a specific DNA oligo probe. A hybridization signal present in both the dorsal and the ventral horns was localized above glial cells, whereas the signal intensity of neurons, as identified in Nissl stainings, was comparable to background levels (Fig. 3 B).

Figure 3. Dorsal horn astrocytes store GABA.

A, optical section of a cervical spinal cord section immunostained to detect S100β (green), GABA (red) and DAPI (blue). From left to right: merged image; immunostaining against GABA; immunostaining against S100β. The arrow indicates a GABA‐positive astrocytic process. ^* indicate GABAergic interneurons. Orthogonal planes showing co‐localization of S100β and GABA. Stack of 23 sequentially scanned slices in 0.5 μm steps. Scale bar: 5 μm. B, radiochemical in situ hybridization for detection of Gat3 (SLC6A11) transcripts in the turtle spinal cord. A hybridization signal was detectable in glial cells in both the ventral horn (left image) and the dorsal horn (right image). Arrows indicate glial cells and arrowheads indicate neurons. Sections were cut from the lumbar part of the spinal cord. Scale bar: 20 μm. [Color figure can be viewed at http://wileyonlinelibrary.com]

In a second series of experiments performed in a slice preparation, we tested whether astrocytes recorded in voltage‐clamp mode responded to glutamate (Fig. 4 A–C). In agreement with others (Ziak et al. 1998; Hefferan et al. 2007), we found that a puff of glutamate induced an inward current (n = 13) that was reduced by the AMPA receptor blocker CNQX (20 μm) (Fig. 4 B, C; significant decrease for 8 of 13 cells that responded to glutamate (paired t test; P = 0.0126). CNQX also inhibited the responses recorded in TTX (10 μm) (Fig. 4 C). Next, we tested if synaptically released glutamate was sufficient to activate dorsal horn astrocytes. We labelled astrocytes by loading slices with the astrocyte‐specific dye SR101 (Fig. 4 D) (Nimmerjahn et al. 2004). We verified the specificity of the labelling by performing immunohistochemical staining against the astrocytic protein S100β. We found that all S100β‐positive cells were positive for SR101 (see Methods) and that 35/43 (i.e. 81%) of SR101‐positive cells were labelled by antibodies against S100β (Fig. 4 D). We monitored intracellular Ca²⁺ in astrocytes in response to the stimulation of nociceptive afferents in thick slices with a dorsal root attached (Fig. 4 E). Unlike central neurons and thick peripheral myelinated axons, nociceptive afferents express sodium channels resistant to low concentrations of TTX (Yoshida et al. 1978; Yoshimura & Jessell, 1990; Ogata & Tatebayashi, 1993). In the turtle nervous system, TTX (50–100 nm) inhibits action potentials in all central and peripheral axons except in C and Aδ afferents (Russo & Hounsgaard, 1996; Russo et al. 2000). This property allowed us to selectively stimulate a fraction of primary afferent fibres while blocking the firing of all central neurons. To monitor the activity of astrocytes, we bulk loaded cells with the Ca²⁺ indicator Oregon Green BAPTA‐AM (Fig. 4 F) and observed the variations in fluorescence intensity in cells positive for both dyes by means of two‐photon microscopy. In the presence of TTX (50–100 nm), a single electric shock applied to the dorsal root filament (Fig. 4 E) evoked a significant Ca²⁺ increase in dorsal horn astrocytes (Fig. 4 G; n = 7 cells from 4 slices). The Ca²⁺ response was reduced by addition of CNQX (20 μm) (Fig. 4 H; median control: 2.85 AU; median after block: 0.94 AU; P = 0.015; Wilcoxon signed‐rank test; n = 6). These findings suggest that glutamate released by non‐myelinated fibres induces a response in dorsal horn astrocytes by activating AMPA receptors.

Figure 4. Spinal astrocytes respond to glutamate and dorsal root stimulation.

A, glutamate puff pipette positioned near an astrocyte recorded in whole cell configuration. B, a puff of glutamate evoked a transient inward current that was blocked by CNQX. Inset: current response to increasing voltage steps. C, summary of all observations. Blue dots: response evoked in TTX (10 μm). CNQX significantly decreased the response evoked in six astrocytes. D, labelling of 3 mm thick slices with SR101 and immunohistochemical staining with an antibody directed against the astrocytic protein S100β. Upper panels, cells positive for both S100β and SR101. Middle panels, high magnification showing a cell positive for both S100β and SR101. The images correspond to the projection of a Z‐stack. The orthogonal planes (below and right) passed through the centre of each image. Lower panels, control picture taken from a slice cut ∼2 mm below the area where SR101 was applied. No soma‐like structures were found positive for SR101. Stars indicate S100β‐positive somas. The red colour corresponds to auto‐fluorescence. In total, 35/43 (81%) of S100β‐positive cells were also positive for SR101. E, spinal cord slice preparation positioned under the objective of a multi‐photon microscope. Dorsal root filament (*) stimulated by means of a suction electrode. Preparation incubated in TTX (50 nm). F, two astrocytes were co‐labelled with SR‐101 and the Ca²⁺ indicator Oregon‐Green BAPTA‐AM. G, a single shock (^*) applied to the dorsal root filament induced an increase in Ca²⁺ concentration (sample images in pseudo‐colour and associated ΔF/F plots) (average of three consecutive traces). H, summary observations (seven cells recorded in TTX). Boxes and whiskers indicate median with 10, 25, 75 and 90 quartiles; ^* P < 0.05, Wilcoxon signed‐rank test. [Color figure can be viewed at http://wileyonlinelibrary.com]

Although the mechanism by which astrocytes release gliotransmitters remains of debate, it is well accepted that it occurs via a Ca²⁺‐dependent mechanism (Araque et al. 2014). To determine if the activation of astrocytes by glutamate evoked the release of GABA, we performed a third set of experiments. We induced the release of GABA evoked by puff application of glutamate as in Fig. 2 (Fig. 5 A) and verified that repeated stimulation produced responses that did not attenuate with time (Fig. 5 B, D). To test whether GABA was released by astrocytes, we buffered Ca²⁺ in a single astrocyte located near the glutamate pipette (Fig. 5 C). For this, we used a patch pipette containing an intracellular solution enriched with the Ca²⁺ buffer BAPTA (30 mm) (Fig. 5 A). Establishing a gigaseal on the astrocyte did not affect the release of GABA (Fig. 5 B–E; n = 6; P = 1.00 Wilcoxon signed‐rank test). However, a few minutes after switching from cell‐attached to whole‐cell configuration, the release of GABA was strongly decreased (Fig. 5 B, D–F; n = 6; P = 0.031; Wilcoxon signed‐rank test). This final result convincingly demonstrates that upon glutamate receptor activation, dorsal horn astrocytes release GABA via a Ca²⁺‐dependent mechanism.

Figure 5. Glutamate induces the release of GABA from dorsal horn astrocytes.

A, GABA sniffer positioned in the dorsal horn of a slice near a glutamate puff pipette. Patch pipette with a high concentration of BAPTA positioned near an astrocyte. B, glutamate puff evoked an outward current in the sniffer (black traces). Establishing a gigaseal on the astrocyte did not affect the response (grey traces). A few minutes after recording the astrocyte in whole‐cell configuration, the response of the sniffer was strongly reduced (blue traces). Lower traces: average of the first seven responses before patching the astrocyte (black) and of the seven last responses after patching the astrocyte (blue). C, left: bright field micrograph of cells. Right: epifluorescence image showing the astrocyte after breaking in with the pipette containing the fluorophore Alexa 658. D, amplitudes of glutamate‐induced currents during the experiment. E, summary of all observations. Chelating Ca²⁺ in astrocytes significantly decreased the release of GABA evoked by glutamate (error bars indicate median ± SD). Beige line: linear regression of the points corresponding to trials before pipette positioning the BAPTA electrode vs. after establishing a gigaseal on the astrocyte. Blue line: regression of the dots after recording in whole‐cell configuration. F, number of events per time bin (3 min) before and after establishing the whole‐cell configuration on the astrocyte (time 0). [Color figure can be viewed at http://wileyonlinelibrary.com]

Discussion

Our study demonstrates that astrocytes from the dorsal horn of the spinal cord of adult turtles can release GABA upon activation by glutamate. Three arguments favour this interpretation. Firstly, after blocking synaptic transmission and in the absence of extracellular Ca²⁺, glutamate evoked the release of GABA. Secondly, glutamate released from central terminals of Aδ and C fibres evoked a Ca²⁺ increase in astrocytes. Lastly, and most importantly, the release of GABA evoked by glutamate was prevented by chelation of Ca²⁺ in a single astrocyte.

The glutamate‐induced responses were partially inhibited by CNQX (Fig. 4), suggesting the involvement of AMPA receptors. In support of this, studies have reported the presence of AMPA receptors in the membrane of astrocytes (Kettenmann et al. 1984; Muller et al. 1992; Hoft et al. 2014). Interestingly, some AMPA receptors, such as those lacking the GluR2 subunit, are permeable for Ca²⁺. Such receptors are expressed in Bergmann glial cells from the cerebellum (Muller et al. 1992), in astrocytes of the olfactory bulb (Droste et al. 2017) and in the grey matter of spinal cord (Brand‐Schieber & Werner, 2003). AMPA receptors could therefore be partly responsible for the Ca²⁺ increase in astrocytes induced by glutamate. The fraction of the response that was not blocked by CNQX could be mediated by other receptors such as NMDA or metabotropic glutamate receptors, which are also present in astrocytes (Bezzi et al. 1998; Palygin et al. 2010).

Glutamate could also bind to receptors in cell types other than astrocytes. Indeed, glutamate receptors including AMPA and metabotropic subtypes are present not only in neurons but also in microglial cells (Biber et al. 1999; Noda et al. 2000). In addition, in the hippocampus, the selective activation of microglia triggers the release of ATP that activates astrocytes, which in turn release glutamate thereby modulating synaptic transmission (Pascual et al. 2012). However, the delay reported between microglial activation and modulation of synaptic transmission was within the range of minutes. By contrast, we found that synaptic release of glutamate triggered an increase in Ca²⁺ in astrocytes and that glutamate elicited GABA release within a fraction of a second (Figs 2, 4 and 5). It is therefore unlikely that microglial cells are involved in the glutamate‐induced release of GABA, even though we cannot rule out this possibility. We excluded the possibility of neuronal contribution because the release of GABA persisted after blocking action potentials with a high concentration of TTX and after complete block of synaptic transmission by Ni²⁺ and Cd²⁺ and removal of extracellular Ca²⁺.

Even though the release of gliotransmitters from astrocytes is well established, the mechanisms involved remain of debate (Araque et al. 2014). Here, we did not investigate the mechanism used by astrocytes to release GABA. However, we can speculate on three possibilities. It has been shown that astrocytes play an active role in nociception via a SNARE‐mediated release of gliotransmitters (Foley et al. 2011). GABA could be released by a reverse uptake mechanism, as demonstrated in other systems (Wu et al. 2007). GABA release could occur via a non‐selective ion channel such as the bestrophin‐1 receptor (Lee et al. 2010). The exact mechanisms will need to be addressed in further studies.

In agreement with a recent study (Sekiguchi et al. 2016), we found that dorsal horn astrocytes responded to Aδ and C afferent stimulation, raising the possibility that GABA released from astrocytes may play a role in the processing of nociceptive information. The duration of GABA release evoked by brief puffs of glutamate was in the same range (from hundreds of milliseconds to seconds) as the slow outward currents induced by the release of GABA from astrocytes in the olfactory bulb (Kozlov et al. 2006). This time span, which exceeds the duration of GABA‐induced IPSP by one order of magnitude, is similar to the duration of primary afferent depolarizations (PADs) underlying presynaptic inhibition (Eccles et al. 1961). Three arguments suggest that GABA released from astrocytes could contribute to the generation of PADs. First, the PAD of primary afferents including non‐myelinated fibres is mediated by the activation of GABA_A receptors (Eccles et al. 1961; Zimmermann, 1968; Janig & Zimmermann, 1971; Calvillo, 1978). Second, in the turtle spinal cord, a slow component of the PAD is generated by a non‐spiking microcircuit (Russo et al. 2000). Third, the time course of PADs is in the same range as the duration of GABA release induced by brief puffs of glutamate (Figs 2 D and 5). Within this context, inflammatory insults leading to chronic pain activate microglial cells, which release specific mediators leading to the appearance of reactive astrocytes (Agulhon et al. 2012). Interestingly, reactive astrocytes have been reported to release more gliotransmitters (Agulhon et al. 2012) raising the possibility that the release of GABA by spinal astrocytes described in this study may be potentiated in some forms of pain.

Additional information

Competing interests

The authors declare no competing financial interests.

Author contributions

RKC, RDL, RER and JPF designed the work. RKC, RDL, RER, BLL, ELA, MFR, GF, NS, ML. AVP, EMC and JPF acquired, analysed or interpreted of data for the work. RKC, RDL, RER and JPF drafted the work or revised it critically for important intellectual content. All authors approved the final version of the manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

The project was funded by the Owensenske Foundation, Simon Fougner Hartmanns Family Foundation, Agnes and Poul Friis Foundation, Carlsberg Foundation, and the International Association for the Study of Pain.

Biography

Rasmus Kordt Christensen received his PhD in neuroscience in the laboratory of Associate Professor Jean‐François Perrier in Copenhagen, working towards understanding of fundamental mechanisms underlying peripheral and spinal cord processing. He then studied auditory processing at the cortical level in the group of Tania Rinaldi Barkat at the University of Basel, Switzerland, identifying neuronal activity features relevant for behaviourally derived percepts. At present he is investigating hippocampal function in the group of Julija Krupic at the University of Cambridge, UK. His long‐term goal is to reach a circuit‐based understanding of sensory percepts.