Anti-homeostatic synaptic plasticity of glycine receptor function after chronic strychnine in developing cultured mouse spinal neurons

Authors


Address correspondence and reprint requests to Luis G. Aguayo, Department of Physiology, University of Concepción, PO Box 160-C, Concepción, Chile.
E-mail: laguayo@udec.cl

Abstract

In this study, we describe a novel form of anti-homeostatic plasticity produced after culturing spinal neurons with strychnine, but not bicuculline or 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX). Strychnine caused a large increase in network excitability, detected as spontaneous synaptic currents and calcium transients. The calcium transients were associated with action potential firing and activation of γ-aminobutyric acid (GABAA) and alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors as they were blocked by tetrodotoxin (TTX), bicuculline, and CNQX. After chronic blockade of glycine receptors (GlyRs), the frequency of synaptic transmission showed a significant enhancement demonstrating the phenomenon of anti-homeostatic plasticity. Spontaneous inhibitory glycinergic currents in treated cells showed a fourfold increase in frequency (from 0.55 to 2.4 Hz) and a 184% increase in average peak amplitude compared with control. Furthermore, the augmentation in excitability accelerated the decay time constant of miniature inhibitory post-synaptic currents. Strychnine caused an increase in GlyR current density, without changes in the apparent affinity. These findings support the idea of a post-synaptic action that partly explains the increase in synaptic transmission. This phenomenon of synaptic plasticity was blocked by TTX, an antibody against brain-derived neurotrophic factor (BDNF) and K252a suggesting the involvement of the neuronal activity-dependent BDNF-TrkB signaling pathway. These results show that the properties of GlyRs are regulated by the degree of neuronal activity in the developing network.

Abbreviations used
BDNF

brain-derived neurotrophic factor

CNQX

6-cyano-7-nitroquinoxaline-2,3-dione

GABAA

γ-aminobutyric acid

GlyRs

glycine receptors

KS

Kolmogorov–Smirnov

mIPSCs

miniature inhibitory post-synaptic currents

PCR

polymerase chain reaction

ROI

regions of interest

sIPSCs

spontaneous inhibitory glycinergic currents

TTX

tetrodotoxin

The phenomenon of homeostatic synaptic plasticity in hippocampal and cortical neurons leads to synaptic stabilization in the network after chronic changes on neuronal activity (Turrigiano and Nelson 2004). The protocols to increase or decrease excitatory neurotransmission involve the use of blockers such as tetrodotoxin (TTX) and 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) or bicuculline, respectively. On the contrary, it is largely unknown whether chronic blockade of a neurotransmitter receptor has an impact on its own function after blockade, or whether all neurotransmissions in a particular network have the same strength to induce this modulation. Although the phenomenon has been described in higher brain regions, where glutamatergic transmission predominates (Watt et al. 2000; Turrigiano and Nelson 2004), it is unknown whether a similar response can be induced in spinal neurons, which primarily display glycinergic neurotransmission (Wu et al. 1992; Nishimaru et al. 1996; Tapia et al. 2001).

Local Ca2+ entry mediated by extrasynaptic glycine receptor (GlyR) activation was shown to regulate post-synaptic assembly and synaptogenesis of glycinergic synapses (Kirsch and Betz 1998). Further studies indicated that GlyR activation was necessary for the maintenance of membrane GlyRs (Levi et al. 1998, 1999; Rasmussen et al. 2002). Additionally, modification of afferent inputs or network activity was shown to alter glycinergic neurotransmission in developing auditory circuits and spinal cord networks (Kotak and Sanes 1996; Chub and O’Donovan 1998; Leao et al. 2004), but detailed characterization of these changes is lacking. Thus, it was postulated that neuronal afferent activity, as well as GlyR activation, was important for GlyR stabilization. Despite the reported changes of receptor clustering in the post-synaptic membrane, no studies on the function of glycinergic transmission after chronic blockade of GlyRs are available. In addition, as previous studies were initiated at a time prior to the appearance of synaptic GlyRs [5 days in vitro (DIV)], they implicate the participation of extrasynaptic more than synaptic receptors on the maintenance of these membrane receptors (Dumoulin et al. 2000; Meier et al. 2000; Tapia et al. 2001). In addition, it is also unknown whether other neurotransmissions, namely GABAergic and glutamatergic, which are present at this stage of neuronal development, have enough strength to induce plastic changes on glycinergic spinal networks. Therefore, in the present study, we examined the effect of chronic blockade of GlyRs with strychnine. Strychnine, but not bicuculline or CNQX, induced a phenomenon that we named anti-homeostatic plasticity because it caused a marked network hyperexcitability, together with an enhancement in glycinergic transmission. Finally, we tested the hypothesis that this was mediated by brain-derived neurotrophic factor (BDNF), shown to play a central role in activity-dependent modification of inhibitory synapses (Seil 2003; Elmariah et al. 2004; Turrigiano and Nelson 2004).

Methods

Neuronal cultures

The animals were treated and handled according to NIH and institutional guidelines. Timed (13–14 days) pregnant mice (C57BL/J6) were terminally anaesthetized with ether and euthanized by cervical dislocation. Embryonic spinal neurons (350 000 cells/mL) were plated onto 33-mm culture plates containing glass coverslips coated with poly-l-lysine (mol. wt. >350 kDa, Sigma Chemical Co., St Louis, MO, USA). The neuronal feeding medium consisted of 90% minimal essential medium (MEM, BRL Technologies, Rockville, MD, USA), 5% heat-inactivated horse serum, 5% fetal bovine serum and a mixture of nutrient supplements (Tapia and Aguayo 1998). The culture medium was changed every 3 days. Grown under this condition, spinal neurons display a significant level of glycinergic transmission at 5 DIV (Tapia et al. 2001; van Zundert et al. 2002). In addition, with the use of antibodies such as calbindin, calretenin and Islet-1, the neurons in these cultures were identified mainly as glycinergic interneurons (van Zundert et al. 2002; Gonzalez-Forero and Alvarez 2005). Neurons were treated with 1 μmol/L strychnine for 48 h (5–7 DIV), alone or in the presence of a polyclonal antibody against BDNF with the capacity to neutralize BDNF biological activity (1 : 100; Chemicon International, AB1779). After this period, the cells were extensively washed with control solution for 30 min to remove strychnine and prepared for electrophysiological recordings, calcium imaging or immunocytochemistry.

Recordings and data analysis

The culture medium in the dish was changed to an external solution containing (in mmol/L): 150 NaCl, 5.4 KCl, 2.0 CaCl2, 1.0 MgCl2, 10 HEPES (pH 7.4, 330 mOsm), and 10 glucose. The electrode internal solution contained (in mmol/L): 120 CsCl, 4.0 MgCl2, 10 BAPTA, 10 HEPES and 2 ATP-Na2 (pH 7.35, 310 mOsm). Whole-cell patch-clamp recordings were obtained using an Axopatch-1D (more recently a 200B) amplifier (Molecular Devices, Sunnyvale, CA, USA) in voltage clamp mode at a holding potential of -60 mV (22–24°C). Electrodes were pulled from borosilicate capillary glass (WPI, Sarasota, FL, USA) in two stages on a vertical puller (Sutter Instruments, Novato, CA, USA). The resistance of the fire-polished patch pipettes was below 4 MΩ when filled with the internal solution. The series resistance (<10 MΩ) was continuously monitored and compensated to >90%. The current signal was filtered at 2 kHz and stored for off-line analysis using a PC interfaced with a Digidata 1200 acquisition board (Molecular Devices). Glycine-evoked currents were obtained by local application of glycine (1–200 μmol/L) using a series of glass flow pipettes fed by gravity. The data are presented normalized to cell capacitance. Analysis of cell capacitance was performed with ClampFit software. Spontaneous post-synaptic currents were recorded in 2 min segments. Glycinergic neurotransmission was pharmacologically isolated in the presence of CNQX (2 μmol/L) and bicuculline (2 μmol/L). TTX (100 nmol/L) was added when miniature inhibitory post-synaptic currents (mIPSCs) were recorded. To enhance the resolution of mIPSC kinetics, these recordings were performed at 18°C in parallel control and treated cultures. The data shown in Figs 1c and 3d were obtained with application of the antagonist to the bath. In all other cases, the receptor antagonists were applied locally to the neuron under study using a small glass fiber (200 μm in diameter). The data were analyzed with MiniAnalysis 5.0 Program (Synaptosoft, Inc., Leonia, NY, USA) to obtain mean averages of peak amplitude, frequency, rise time (10–90%) and decay time constant (between 90–10% of decay). Amplitude histograms were plotted with a bin width of 2–5 pA. For automated analysis, events with peak amplitudes larger than 1.5 times the background noise (8–10 pA) were selected. Otherwise, synaptic events having a monotonic rising phase (>10 pA) were manually selected. From the resulting data, cumulative or frequency histograms were generated. Recordings from gramicidin perforated cells were performed with a similar protocol to that previously described (Tapia and Aguayo 1998). Briefly, gramicidin (Sigma) was dissolved in methanol and used at 100 μg/mL. Recording of input resistance was performed using current steps of varying amplitude (±40 pA, 10 ms) and measuring the resulting lineal change in membrane potential.

Figure 1.

 Effects of receptor blockade on spinal neuron excitability. (a) Voltage traces obtained with gramicidin-perforated recordings with a KCl-filled electrode. Each trace represents the overall level of synaptic potentials and spikes mediated by glycine, GABAA, and AMPA receptors. The control traces were obtained at resting potentials (≈-50 mV) and show that most of the activity was depolarizing. Application of 1 μmol/L strychnine to the bath induced a large increase in the frequency of synaptic potentials and spikes, without changes on resting potential. Bicuculline (2 μmol/L) did not increase neuronal excitability. The application of CNQX (2 μmol/L) reduced the frequency of overall post-synaptic potentials and spikes (p < 0.05, paired test). (b–e) All-point voltage amplitude distributions obtained before and in the presence of bath applied strychnine and CNQX in these group of neurons (n = 3). The solid line is a fit to a single or double Gaussians to the data points.

Figure 3.

 Enhanced spontaneous calcium transients during actual GlyR blockade. The panels illustrate the spontaneous change in fluorescence intensity associated with the calcium indicator fluo-3 in neuronal ensembles. The upper traces represent intensities obtained in the absence (a) and continuous presence (b) of 1 μmol/L strychnine in 7 DIV neurons. Each trace shows the intensity recorded in a single neuron. Compared with the control neurons (n = 4), neurons in the presence of strychnine (n = 4) for 40 h displayed a large increase in the number of calcium transients. The lower panels show that the calcium spikes were completely blocked by acute application of 1 μmol/L TTX (c) or by a combination of bicuculline and CNQX applied to the bath (d). The acute effect of bicuculline and CNQX was reversed after 60 s of wash with normal solution. The bars (e, f) represent the percent of neurons displaying spontaneous calcium transients after culturing the neurons with chronic strychnine or bicuculline (n = number of independent fields analyzed). At least 80 neurons per group were recorded under the indicated conditions (* denotes p < 0.05, Student’s t test).

Analysis of intracellular calcium transients with fluo-3

Neurons were loaded with fluo-3 AM (1 μmol/L in pluronic acid/DMSO, Molecular Probes, Eugene, OR, USA) for 30 min at 36.5°C. The neurons were then washed twice with external solution and incubated for 30 min at 36.5°C. The cells were mounted in a perfusion chamber that was placed on the stage of an inverted fluorescent microscope (Eclipse TE, Nikon, Inst. Inc., Melville, NY, USA) equipped with a xenon lamp and a 40× objective (22–24°C). The cells were briefly (200 ms) illuminated using a computer-controlled Lambda 10-2 filter wheel (Sutter Instruments, Noveto, CA, USA). Regions of interest were simultaneously selected on neuronal somata containing fluo-3 fluorescence (excitation 480 nm, emission 510 nm) in a field having usually more than 10 cells. Images were collected at 2–5 s intervals during a continuous 5-min period. The imaging was carried out with a 12-bit cooled SensiCam camera (PCO, Kelheim, Denmark). The calcium transients, defined by their TTX sensitivity (Gu et al. 1994), were acquired and analysed off line with Axon Instruments Workbench 2.1 software.

FM1-43 loading and de-staining

The styryl dye FM1-43 [N-(3-triethylammoniumpropyl)-4-(4-(dibutylamino)styryl)pyridinium dibromide, Molecular Probes] was dissolved in external solution (2 mmol/L) and the vesicles were labeled with 15 μmol/L during a high-K+ depolarization for 5 min. Similarly, depolarization-evoked de-staining was induced by bath perfusion with elevated K+ (30 mmol/L with equiosmolar replacement of Na+). The depolarization with K+ can induce either dye loading or unloading because the incorporation of FM to the vesicles depends on the previous state of release. FM1-43 fluorescence was acquired using an epifluorescence microscope Nikon (Eclipse TE, Nikon Inst., Inc.) equipped with a 100× objective (Neofluor, oil immersion, NA 1.0 (Zeiss, Jena, Germany). FM1-43 data were collected with appropriate filters (excitation 540 nm, emission 620 nm). Fluorescence intensity was measured using a 2 × 2 binning with a CCD camera (SensiCam; PCO). Images were digitized and processed with Imaging Axon Workbench 2.2 (Axon instruments, Forest City, CA, USA).

Immunocytochemistry and detection of synaptic vesicles

Not all the pre-synaptic markers are associated to recycling vesicles and this is apparent because they will not co-localize with pre-synaptic terminals loaded with AM1-43, a fixable form of FM1-43 (15 μmol/L, Biotium Inc, Hayward, CA, USA). For this, the dye was immediately washed in dye-free solution with nominal Ca2+ to minimize spontaneous dye loss, fixed for 30 min with 4% paraformaldehyde and permeabilized with 0.1% triton X-100 in PBS. Following this, the neurons were stained with monoclonal synapsin I antibody (1:100, Santa Cruz Biotechnology, Santa Cruz, CA, USA), incubated overnight at 22–24°C, followed by incubation with anti-goat secondary antibody conjugated with FITC (1:500, Jackson ImmunoResearch Laboratories, PA, USA). Non-specific immunoreactivity was blocked with 10% horse serum for 1 h at 22–24°C. The coverslips were mounted in fluorescent mounting medium (DAKO, CA, USA). Optical sections of 0.5 μm for two separate channel fluorescence (AM1-43 and FITC) were acquired using a C1 Nikon confocal microscope (60×, water immersion, NA 1.4; 488 and 543 nm, respectively). After acquisition, images were processed with ImageJ (NIH, Maryland, USA).

Quantitative real-time polymerase chain reaction analysis

cDNA was synthesized as previously described (van Zundert et al. 2004). Real-time PCR was carried out using 2 μL of cDNA per sample in a final volume of 25 μL with 20 mmol/L Tris-HCl (pH 8.4), 3 mmol/L MgCl2, 50 mmol/L KCl, 200 μmol/L of each dNTP, 1X SYBR Green buffer and 0.25 U of platinum Taq DNA polymerase. Standard curves, which convert the measured threshold cycle number into the starting quantity of DNA, were constructed using serial tenfold dilutions of plasmids for the α1 and α2 GlyR subunits (GlyRα1 and GlyRα2) and β-actin (pBactin-231). PCRs were performed in an iCycler iQ real–time PCR system (BioRad Labs, Hercules, CA, USA) in 96 well format. All measurements were performed in triplicate and three different experiments were considered per condition. The relative levels of GlyR subunits were calculated as ratios obtained between starting quantities of DNA versus those obtained for β-actin. To avoid saturation errors, α1 samples were run at a low cycle number and in the presence of excess reagents in comparison to cDNA. The primers used, sequence and size have been previously described (van Zundert et al. 2004).

Data analysis

The data from control and strychnine-treated sister cultures are presented as mean ± standard error of the mean (SEM). Pair-wise comparisons were performed using Student’s t-test and anova. Cumulative probability distributions were compared with Kolmogorov–Smirnov (KS) tests. Each set of data presented (experimental vs. control) was performed in sister cultures to reduce variability. Some comparisons were performed before and during applications of antagonists to reduce the statistical variability (Fig. 1). The experiments were performed in triplicate. Ten randomly selected optical fields were used for analysis of neuronal survival under the different conditions. The number of neurons was expressed in relation to the surface area used for this analysis (105 μm2). The symbols represent *p < 0.05, **p < 0.01 and ***p < 0.005.

Results

Effects of acute strychnine on the activity of spinal neuronal cultures

In the first series of experiments, we wanted to examine the impact of acute receptor blockade on excitability. The recordings in Fig. 1, obtained with gramicidin-perforated patches, show that 5 DIV spinal neurons under control conditions displayed depolarizing synaptic responses, which were able to elicit overshooting spikes. These potentials are associated to synaptic currents induced by activation of glycine, GABAergic, and AMPAergic receptors (Tapia et al. 2001). In addition, application of glycine (10–30 μmol/L) induced increases in intracellular calcium (not shown) indicating its depolarizing nature. Acute strychnine applied to the recording chamber evoked an increase in synaptic activity that resulted in increased neuronal firing (Fig. 1a, upper traces, b and c; p < 0.05). This increased firing is likely associated to a reduction in glycine-mediated membrane shunting because the input resistance of the neurons increased from 382 ± 36 in control conditions to 430 ± 51 MΩ after the application of strychnine (n = 5). The effect of strychnine on excitability was not reproduced by bath application of bicuculline (Fig. 1a, middle traces). On the contrary, CNQX reduced the overall neuronal excitability (Fig. 1a, lower traces, d and e; p < 0.05). The data in Fig. 1 also show that the resting membrane potential was not altered during the application of the antagonists.

It was recently reported that blockade of GABAA receptors with bicuculline produced a long-lasting increase in hippocampal neuron excitability reflected as periodic and synchronous activity discharge patterns (Arnold et al. 2005). Therefore, after culturing spinal neurons for 48 h with strychnine (1 μmol/L), we re-examined the synaptic activity (Fig. 2). Whole-cell voltage clamp recordings of spontaneous synaptic currents were performed in normal external solution (without strychnine) to monitor the overall level of spontaneous synaptic activity in these cultures (Fig. 2, upper traces). The average frequency of synaptic currents in control neurons was 0.9 ± 0.5 Hz (n = 5) and increased by 234% ± 32% in neurons cultured in the presence of strychnine (Fig. 2b, n = 5; p < 0.05). Similarly, the amplitude of these currents increased from 39 ± 3 to 68 ± 16 pA (n = 5). Similar chronic treatment of the neurons with bicuculline did not change synaptic excitability (Fig. 2b). Chronic CNQX (2 μmol/L), on the contrary, reduced the level of synaptic transmission (Fig. 2b, p < 0.05). The lack of GABAA effects on receptor blockade on overall network activity is consistent with its small effect in acute experiments (Fig. 1). Culturing the neurons with TTX (1 μmol/L) to block evoked synaptic transmission did not have any effect (92% ± 13%, n = 5; Fig. 2b) on the overall synaptic transmission, contrary to cortical neurons (Turrigiano et al. 1998). All together, these data show the absence of homeostatic plasticity in spinal neurons.

Figure 2.

 Chronic blockade of glycine, but not GABAA and AMPA receptors, produced a large enhancement on overall synaptic transmission. (a) The current traces show total neurotransmissions mediated by glycine, GABAA, and AMPA receptors. The responses were recorded in control and after culturing the neurons for 48 h with the antagonists. (b) The graph illustrates the effect of receptor and neuronal activity blockade on the frequency of overall synaptic transmission. Each bar is the mean ± SEM obtained from a least five neurons (* denotes p < 0.05). The line represents the control level of synaptic transmission.

As it was surprising to find an increased level of synaptic activity after chronic strychnine (Fig. 2), we decided to examine the behavior of these neurons during the actual strychnine treatment (i.e. in presence of strychnine). For this, calcium imaging was used to simultaneously examine the behavior of ensembles of neurons in the presence of the GlyR antagonist. This type of analysis permits examination of neuronal assemblies and allows for better inferences about the behavior of neuronal networks and the impact of chronically blocking GlyRs in vitro (Tapia et al. 2001). Neurons were loaded with fluo-3 and the fluorimetric activity of neurons cultured in the presence of strychnine (1 μmol/L, 40 h) was measured. Control neurons showed calcium transients that were not highly synchronized (Fig. 3a). Spontaneous calcium transients in neurons still in the presence of strychnine, on the contrary, showed an organized rhythmic pattern (Fig. 3b). Application of TTX abolished calcium transients indicating that they were mediated by action potential firing (Fig. 3c). Subsequent blockade of these calcium transients with acute bicuculline and CNQX applied to the bath showed that these spontaneous events depended on the remaining synaptic transmission mediated by the other two receptors expressed in the neurons (Tapia et al. 2001), GABAA and AMPA receptors (Fig. 3d). Finally, the number of neurons displaying spontaneous calcium transients increased significantly after blocking glycine, but not GABAA receptors (Figs 3e and f).

Neuronal activity enhanced glycinergic synaptic activity

The previous results show that chronic blockade of neuronal glycine receptors with strychnine results in remarkable increases in overall synaptic activity. Using this model, we decided to examine the glycinergic synapses. We measured glycinergic currents in neurons treated for 48 h with strychnine (1 μmol/L) and then extensively washed for 30 min with external solution before starting recordings. Both the frequency and amplitude of glycinergic spontaneous inhibitory glycinergic currents (sIPSCs) were increased after chronic strychnine (Fig. 4a). In the presence of bicuculline and CNQX, the frequency of pharmacologically isolated glycinergic sIPSCs was 0.55 ± 0.25 Hz and their average peak amplitude was 37 ± 3.7 pA in control neurons (n = 8). In treated neurons, the frequency of glycinergic events increased to 2.4 ± 0.42 Hz (n = 5; p < 0.001) and their amplitude to 68 ± 7.6 pA (n = 5; p < 0.05). Similarly, the cumulative probability curves for inter-event interval and amplitude of glycinergic events showed significant (p < 0.001) shifts toward higher frequencies and amplitudes (Figs 4b and c). Additional experiments were performed treating 1 DIV neurons with either strychnine alone or in combination with CNQX and bicuculline. Neurons treated with strychnine for 7 days showed a small increase in frequency (0.59 ± 0.15 Hz; p > 0.05) accompanied by a slight increase in current amplitude (44% ± 19% pA; p > 0.05). This lack of increase in current amplitude is in agreement with previous studies that show some degree of receptor internalization with strychnine in 1 DIV neurons (Kirsch and Betz 1998). Additionally, we performed experiments where all the neurotransmission was blocked by a cocktail containing strychnine, CNQX and bicuculline for 7 days. This treatment did not affect either the frequency (0.77 ± 0.32 Hz) or the amplitude (48 ± 6 pA) of glycinergic transmission.

Figure 4.

 Chronic blockade of GlyRs induced an increase in glycinergic synaptic activities. (a) Spontaneous glycinergic currents recorded in the presence of bicuculline and CNQX after culturing the neurons with strychnine. The third trace shows that these synaptic currents were reversibly blocked by strychnine indicating a glycinergic nature. The right record was obtained after 2 min of wash. (b) The cumulative distributions of inter-event interval curves were obtained from control (open squares, n = 5) and treated (closed squares, n = 6) neurons and show that strychnine increased the frequency of synaptic events [Kolmogorov–Smirnov (KS) test, p < 0.001]. (c) Plot of cumulative peak current amplitude showing that strychnine shifted the curve toward larger peak amplitudes (KS test, p < 0.001).

A number of mechanisms including increased number of synapses or active zones, increased probability of release at single sites or enhanced post-synaptic actions of released neurotransmitter could explain the observed changes in glycinergic sIPSC frequency and amplitude (Lim et al. 1999; van Zundert et al. 2004). To obtain further information about possible mechanisms that contribute to the potentiation of glycinergic synapses after strychnine, we recorded spontaneous glycinergic miniature IPSC (mIPSCs) isolated with bicuculline and CNQX in the presence of acute TTX (1 μmol/L) to block action potentials. mIPSCs are generally thought to be the post-synaptic response to the release of a single vesicle (quanta) of neurotransmitter. Assuming that the packing of glycine in pre-synaptic vesicles is relatively uniform (but see Gomeza et al. 2003), increases in the peak amplitude of mIPSCs are usually interpreted as an increase in post-synaptic sensitivity, while increases in frequency are usually associated with increases in release probability at individual sites, increases in the number of release sites or both. The average frequency and peak amplitude of glycinergic mIPSCs significantly increased after strychnine treatment (Figs 5a and b). mIPSCs frequency increased from 0.15 ± 0.05 Hz in control to 0.5 ± 0.1 after strychnine (n = 6; p < 0.001). Average mIPSC peak amplitude was 13 ± 0.9 pA in control (n = 6) and 26 ± 4.4 pA (n = 6; p < 0.05) in treated neurons. In addition, we found that the time course of mIPSCs was accelerated after chronic strychnine. For example, the decay time constant of glycinergic events were 42 ± 4 and 18 ± 1.4 ms in control and treated neurons (Fig. 5c), respectively (n = 6; p < 0.001). The acceleration in the decay of mIPCS is consistent with the idea that the receptor switched from α2 to α1 subunits containing GlyRs (Takahashi et al. 1992). In agreement, analysis with quantitative real-time PCR showed that the content of α2 subunit mRNA was reduced to 7% ± 4% of control after strychnine treatment (Fig. 5e). On the contrary, the content of mRNA for the α1 subunit was not significantly altered in three independent experiments (95% ± 25% of control).

Figure 5.

 The properties of glycinergic mIPSCs were modified by chronic blockade with strychnine. (a, b) The average values for event frequency (a) and peak amplitude (b) of mIPSCs (isolated with application of bicuculline and CNQX, in the presence of TTX) were significantly increased in the treated neurons (n = 6). (c) The graph shows the values for the mIPSC decay time constant (single exponential adjusted to points between 90% and 10%) obtained in control (n = 6) and after strychnine (n = 6). (d) Shows normalized amplitude (maximal 1.0) of mIPSCs recorded in whole-cell configuration obtained from control and strychnine-treated neurons (16–18°C). Each bar is the mean ± SEM from 6 neurons. (e) The graph shows the effect of strychnine on expression levels of α1 and α2 mRNA measured with quantitative PCR using SYBR-Green I (see ‘Methods’). Data were obtained from three independent experiments.

We then decided to study the effect of strychnine in pre-synaptic function and morphology using dynamic and static indicators. The colocalization of AM1-43 and synapsin-I indicated the presence of active and inactive releasing sites (Fig. 6a). Therefore, to avoid interferences from non-specific dye staining (not colocalized with synapsin-I), we used repetitive loading and de-staining of FM1-43 and only the spots that went through the cycle were considered for further analysis. An enhanced rate and magnitude of the FM1-43 de-staining was found in treated neurons (Fig. 6b). Furthermore, we found a significant increase in the intensity of FM1-43 fluorescence in treated neurons when compared with the control condition (Figs 6c and d). These changes are in agreement with the increased frequency of synaptic events found in these neurons (Fig. 5a).

Figure 6.

 Chronic strychnine increased K+ evoked release in spinal cord neurons. (a) The confocal images show synapsin I immunoreactivity and its co-localized AM1-43 fluorescence label in control and treated neurons (bar is 10 μm). Only the co-localized (yellow) markers, higher in treated neurons, should be associated to active vesicle recycling. (b) The FM1-43 fluorescence de-staining evoked by high-K+ was significantly higher in the treated neurons (p < 0.05; n = 150 puncta from three experiments). (c) The histograms show FM1-43 fluorescence values for labeled terminals in control and treated neurons (p < 0.05, n = 74 puncta). (d) Shows a higher magnification of dashed boxes in A for control (A1) and treated (A2) neurons. Scale bar is 2 μm. (e) The scheme is the protocol for FM1-43 loading.

The effect of strychnine was dependent on BDNF-trkB signaling pathway

It is well known that an increase in electrical activity within a neuronal network can enhance the release of BDNF (Balkowiec et al. 2000; Schinder and Poo 2000), which exerts a powerful influence on the development of inhibitory synapses (Vicario-Abejon et al. 1998; Baldelli et al. 2002). Until now, however, much of the information obtained has been related to GABAergic synapses and little is known about BDNF effects on glycinergic neurotransmission. The effects of BDNF on GABAergic synapses can be effectively blocked by antibodies against this neurotrophin (Marty et al. 2000; Seil and Drake-Baumann 2000). To test the idea that strychnine effects on GlyRs are via BDNF, we used an antibody known to block the biological activity of BDNF (Fig. 7a). Culturing spinal neurons with anti-BDNF (1:100 dilution) blocked the increase in event frequency (Fig. 7b) and current amplitude (Fig. 7c) of glycinergic sIPSCs produced by chronic strychnine treatment. The blocking effect of the antibody was not related to differences in survival because the number of neurons/105 μm2 was similar in the different conditions; 28 ± 2.3 in control, 29 ± 3.5 in strychnine and 28 ± 1.7 in the presence of the antibody. In addition, culturing the neurons with Ab-BDNF alone did not affect the amplitude of glycinergic currents (37 ± 3 vs. 44 ± 12 pA, n = 6). Supporting the involvement of TrkB receptor activation, the effect of strychnine on the enhancement of glycinergic transmission was blocked by K252a (200 nmol/L), a tyrosine kinase inhibitor (Knusel and Hefti 1992). The data in Fig. 8 show that culturing the neurons with this kinase antagonist blocked the increase in frequency produced by strychnine. Similar to the data with the antibody, application of K252a alone produced no changes on the amplitude of glycinergic transmission (22 ± 0.8 pA for control vs. 21 ± 3 pA for K252, n = 4). In addition, chronic TTX (1 μmol/L) was also able to block the increase in frequency of glycinergic transmission with strychnine. On the contrary, in agreement with the idea that BDNF can enhance glycinergic transmission, culturing the neurons with this neurotrophin (10 ng/mL) produced a positive effect on glycinergic transmission (Fig. 8).

Figure 7.

 The effect of strychnine on spontaneous glycinergic currents was blocked by an antibody against BDNF. (a) Spontaneous glycinergic currents obtained in control neurons and those neurons treated with strychnine alone or with anti-BDNF. (b, c) Graphs illustrate the effect of anti-BDNF on the cumulative probability for inter-event interval and peak amplitude of glycinergic spontaneous currents in control (open squares, n = 4), strychnine (closed squares, n = 6) and strychnine plus anti-BDNF (open triangles, n = 4; Kolmogorov–Smirnov test, p < 0.001). The effect of chronic strychnine was blocked by the antibody treatment.

Figure 8.

 The effect of strychnine on synaptic plasticity was dependent on neuronal activity and BDNF-trkB pathway. The effect of strychnine was inhibited by co-culturing the neurons with TTX (1 μmol/L, 48 h) or K252a (200 nmol/L). The effect of strychnine was mimicked by culturing the neurons with BDNF alone (10 ng/mL, 5 days). Each data point was obtained from at least four neurons. ** denotes p < 0.05.

The increase in the amplitude of spontaneous synaptic glycinergic events could be related to an increase in the number of post-synaptic receptors. In line with this possibility, the direct application of extracellular glycine indicated a 2.6-fold enhancement of the whole-cell current density after strychnine treatment (Fig. 9). Although some small changes in the current amplitude were found, strychnine did not affect the affinity of the receptor to glycine (p > 0.05; Fig. 9b). Thus, the changes in sIPSCs and mIPSC amplitude appear to correlate with an increase in the total number of glycine receptors in the membrane, leading perhaps to a higher number of clustered receptors at post-synaptic (synaptic and extrasynaptic) sites. In line with the data for synaptic currents, the increase in evoked post-synaptic glycinergic current density after chronic strychnine was blocked in the presence of anti-BDNF (Fig. 9c).

Figure 9.

 The density of the post-synaptic glycine-evoked current increased with strychnine. (a) Whole-cell current traces illustrate the increase in membrane current amplitude obtained in the presence of increasing concentrations of glycine (10–200 μmol/L). (b) The graph shows the relationship between current amplitude and concentration of glycine for control and treated neurons. The EC50 was 39 ± 1.0 µmol/L in control and 35 ± 1.0 µmol/L in strychnine-treated neurons (p > 0.05). (c) The graph shows that the effect of strychnine on current density was blocked by Ab-BDNF. Each data point was obtained from at least six neurons. ** denotes p < 0.01 by Student’s t test.

Taken together, these data strongly suggest that BDNF is a likely mediator of some of the activity-dependent changes that produce an increase in glycinergic neurotransmission after early chronic blockade with strychnine.

Discussion

Response of cultured neurons to synaptic blockade

Cortical and hippocampal neurons in culture form abundant excitatory, AMPA-mediated networks that contribute to spontaneous synaptic activity and display homeostatic plasticity (Turrigiano et al. 1998; Watt et al. 2000; Turrigiano and Nelson 2004). Classical experiments to demonstrate homeostatic plasticity in cortical and hippocampal neurons showed a marked increase in network excitability after chronic inhibition of excitatory glutamatergic synapses with CNQX. On the contrary, cortical networks exposed to bicuculline, a GABAA receptor blocker, responded with an acute increase in activity that returned to control levels after 2 days. Overall, these studies show that the homeostatic phenomenon adjusts the cellular and synaptic properties such as strength and synaptic formation of central networks to compensate for changes in synaptic drive.

The present results obtained in spinal neurons showed a very different pattern of responses after chronic receptor blockade. Unlike hippocampal neurons, blockade of glycine receptors in spinal neurons induced a marked and stable increase in synaptic transmission. In addition, long-term blockade of GABAA or AMPA receptors was unable to induce the expected homeostatic responses (Fig. 2). Although our results are not in agreement with the phenomenon of homeostatic plasticity, they are in line with the long lasting increase in excitability recently observed after the inhibition of GABAA receptors by application of bicuculline to hippocampal neurons (Arnold et al. 2005), indicating that homeostatic plasticity induction is more complex than previously reported (Turrigiano et al. 1998; Turrigiano and Nelson 2004).

Effects of strychnine on synaptic plasticity

This study showed that strychnine produced a strong enhancement on overall neurotransmission in spinal neurons (Fig. 2). Thus, the data disclose the presence of a novel form of synaptic plasticity in a central synapse, a phenomenon that we nominated anti-homeostatic synaptic plasticity because it tended to stabilize the network in a more excitable state (Fig. 2). In addition, after the removal of strychnine, we found a large increase in the levels of glycinergic synaptic transmission. We believe that the different effects of blocking glycinergic or GABAA/AMPA receptors can be explained at the neuronal network level. For example, it is known that glycinergic transmission predominates over GABAA and AMPA neurotransmissions in immature neurons (Wu et al. 1992; Nishimaru et al. 1996; Tapia et al. 2001). This predominance explains the large impact on network excitability that follows the acute blockade of glycinergic, but not GABAergic transmission. The present data also show an important, not previously described, characteristic of immature GlyRs. Thus, although GlyR activation caused membrane depolarization and increased intracellular calcium (Tapia and Aguayo 1998; not shown), the increased Cl- conductance appeared to shunt excitability because acute strychnine increased excitability (Fig. 1).

Maturation of post-synaptic GlyRs in spinal neurons

Previous studies indicate that GlyRs are recruited to the synapse to establish functional synaptic activity (Bechade et al. 1996; Colin et al. 1996; Dumoulin et al. 2000; Meier et al. 2000). Thus, clustering and perhaps maturation of functional GlyRs might be modulated by neuronal activity, but this has not been physiologically examined. In structural studies, GlyRs appeared sequestered inside the cytoplasm after chronically culturing spinal neurons (1 DIV) with strychnine (Kirsch and Betz 1998; Levi et al. 1998). Additional analysis showed that GlyR accumulation was related to a change in the insertion/degradation equilibrium of the receptor (Rasmussen et al. 2002). Additionally, post-synaptic gephyrin/GlyR clusters were not destabilized by treatment with strychnine.

Our results indicate that activation of GlyRs is not necessary for its accumulation in the membrane as under the present experimental conditions it should be fully antagonized with strychnine (EC50 = 20 nmol/L). Rather, inhibition of synaptic GlyRs was able to enhance the expression and maturation of these inhibitory receptors at a stage when GlyRs and GABAARs are both depolarizing (Reichling et al. 1994; Tapia and Aguayo 1998; Kulik et al. 2000). It is most likely that the discrepancy with the early morphological studies is due to differences in the experimental protocols used. For example, the previous treatment with strychnine started at 1 DIV and the neurons were cultured for 8 days (Kirsch and Betz 1998; Levi et al. 1998). At this stage of development, the inhibitory effect of strychnine on GlyRs was likely related to actions on extracellular receptors, because the glycinergic neurotransmission is only established at 5 DIV (Tapia et al. 2001). In our hands, treatment of 1 DIV spinal neurons with strychnine was unable to produce any change on synaptic transmission suggesting that the time of stimulation is critical for the up regulation of glycinergic transmission. Additionally, another study (Rasmussen et al. 2002) used a higher concentration of strychnine to block GlyRs and this opened the possibility of inducing non-specific effects on other membrane receptors (Jonas et al. 1998).

Influence of neuronal activity on the function of GlyRs

How are GlyRs induced to mature? After blocking GlyRs, spinal neurons responded with a large increase in the level of neurotransmission detected by electrophysiological and cell imaging methodologies. It was significant to find that TTX-dependent calcium transients were enhanced in the presence of strychnine (Fig. 3). This finding supports the idea that spinal neurons are in a state of enhanced excitability while in the presence of the alkaloid. Analyses of glycinergic mIPSCs suggest increases in both pre-synaptic (higher frequency of events) and post-synaptic (higher current amplitude) mechanisms. Therefore, an increase in glycine release after strychnine was most likely due to increased release probabilities at individual sites. This was directly demonstrated by increases in evoked FM1-43 de-staining. On the contrary, the increase in current amplitude to exogenous glycine suggests an augmentation in post-synaptic receptors. The effect of strychnine was not a simple scaling up of the GlyR-mediated current initiated in the post-synaptic membrane, but it also included a significant acceleration in the decay phase of the mIPSC. Similar effects of neuronal activity on the time course of mIPSCs were reported in auditory glycinergic synapses (Leao et al. 2004). These changes are likely related to a significant reduction in α2 subunits, similar to that previously described in other maturing neurons (Malosio et al. 1991; Singer et al. 1998; Ali et al. 2000). The lack of changes in the level of α1 mRNA after strychnine suggests that the enhancement in glycinergic transmission is related to post-translational modifications, such as receptor stabilization in the cell membrane.

Mechanisms associated to the blockade of GlyRs in developing neurons

It is well known that the synthesis and release of BDNF depends on the level of synaptic activity and intracellular calcium (Zafra et al. 1992; Marty et al. 1996). Therefore, the effect of strychnine on the function and morphology of post-synaptic GlyRs was likely associated with an enhancement of BDNF release produced by the increase in neuronal activity. The possibility that GlyRs were modulated by activity-dependent increases in BDNF was confirmed by blocking the effect of strychnine with TTX and an antibody against BDNF. Additionally, the effect of strychnine was blocked by a tyrosine kinase inhibitor and mimicked by exogenous BDNF. These findings are in line with previous reports showing that BDNF was implicated in the development of GABAA transmission (Mizuno et al. 1994; Vicario-Abejon et al. 1998; Bolton et al. 2000; Marty et al. 2000; Yamada et al. 2002; Elmariah et al. 2004; Jovanovic et al. 2004). Furthermore, overexpression of BDNF resulted in an enhancement of GABAergic inhibition in visual cortex (Huang et al. 1999) and cerebellar slices (Bao et al. 1999). Similarly, neuronal activity and BDNF regulated the density of GABAA inhibitory synapses in developing hippocampal interneurons (Marty et al. 2000).

In conclusion, the present results indicate that blockade of GlyRs with strychnine during a highly dynamic period of initial synaptogenesis induced an increase in glycinergic transmission, via pre- and post-synaptic mechanisms, mediated by BDNF and activation of TrkB receptors.

Acknowledgements

The authors would like to thank Laurie J. Aguayo for her technical assistance. Research support was provided by Fondecyt 2000135, Fondecyt 1980106, 1020475, GIA-DIUC and NIAAA AA15150 (LGA), Fondecyt doctoral grants (MAC and JCT).

Ancillary