Brain derived neurotrophic factor (BDNF) promotes the formation, maturation and stabilization of inhibitory synapses in the central nervous system. In addition, BDNF has been suggested to regulate the critical period for ocular dominance plasticity in the visual system. Here we further evaluated the role of BDNF in the visual cortex by studying the GABAergic synaptic transmission under conditions of chronically reduced levels of BDNF. Whole-cell patch-clamp recordings were performed from pyramidal neurons located in layers II/III of visual cortical slices in heterozygous BDNF knockout mice (BDNF (+/−)) and their wild-type littermates at the age of 21–25 days. The BDNF (+/−) mice showed a decreased frequency and amplitude of miniature inhibitory postsynaptic currents (mIPSCs) as well as a reduced amplitude and prolonged decay time constant of evoked IPSCs. Further analyses indicated an impaired presynaptic GABAergic function in BDNF (+/−) mice, as shown by the decreased release probability, steady-state release and synchronous release of GABA. However, the number of functional release sites remained unchanged. In line with these observations, an impaired glutamate-driven GABA release was observed in BDNF (+/−) mice. Furthermore, the overall balance in the strength of cortical excitation to inhibition shifted towards a decreased inhibition. Finally, the reversal potential for chloride-mediated evoked IPSCs was not affected. These findings suggested that chronically reduced levels of BDNF strongly impair the GABAergic inhibitory function in visual cortex by altering postsynaptic properties and by reducing presynaptic GABA release as well as the overall strength of inhibition onto pyramidal neurons within the cortical network. These impairments of inhibitory function are compatible with a rather immature status of the GABAergic system in BDNF (+/−) mice, which supports the hypothesis that the level of expression for BDNF critically affects maturation and function of the GABAergic inhibition.
The neurotrophin ‘brain derived neurotrophic factor’ (BDNF) has been intensively studied in the brain over the last decade. Along with its permissive role in survival and differentiation, BDNF is suggested to modulate the efficacy of basal synaptic transmission and synaptic plasticity at central excitatory synapses (for review see: Poo, 2001; Lu, 2004). Acute or chronic application of the neurotrophin led to enhanced glutamatergic neurotransmission as shown by experiments in vivo (Messaoudi et al. 1998) as well as in hippocampal and cortical cultures (Lessmann et al. 1994; Levine et al. 1995; Li et al. 1998) and in acute slices in vitro (Kang & Schuman, 1995; Carmignoto et al. 1997). In accordance, an impaired glutamatergic neurotransmission could be observed under conditions of long-term reduced levels of BDNF by use of BDNF knockout (KO) mice models in hippocampus (Korte et al. 1995; Pozzo-Miller et al. 1999), in cerebellum (Carter et al. 2002) and in the neocortex (Bartoletti et al. 2002; Abidin et al. 2006). The mechanisms by which BDNF enhances the glutamatergic neurotransmission included an enhanced phosphorylation of synaptic vesicle-associated proteins (Jovanovic et al. 2000), causing enhanced transmitter loading, vesicle mobilization or secretion at excitatory synapses.
For the inhibitory system it is suggested that acutely applied BDNF reduces the efficacy of the inhibitory transmission, as shown by impaired miniature- and evoked- inhibitory postsynaptic currents (mIPSCs and eIPSCs, respectively) in hippocampus slices (Tanaka et al. 1997; Frerking et al. 1998) and in cell cultures of the cerebellum (Cheng & Yeh, 2003) and hippocampus (Brunig et al. 2001). In contrast, a chronic surplus application of BDNF rather enhanced the efficacy of the inhibitory system as shown by an increased presynaptic release probability for GABA in hippocampal cultures (Baldelli et al. 2005) and an increased asynchronous GABA release in cell cultures of the superior colliculus (Henneberger et al. 2005). It also led to increased staining for glutamic acid decarboxylase 65 (GAD65) in cell cultures of the hippocampus (Ohba et al. 2005) and superior colliculus (Henneberger et al. 2005). Furthermore, chronic application of BDNF was shown to modulate the composition of presynaptic Ca2+ channels in hippocampal cultures, thereby increasing the release probability for GABA (Baldelli et al. 2005). On the structural level, the chronic levels of surplus BDNF led to an elevated size of GABAergic cell somata (Yamada et al. 2002) and to an increase in the number of axonal branches and in the total length of the axons of GABAergic neurons (Vicario-Abejon et al. 1998). In agreement, Rutherford et al. (1997) observed a reduced number of GABAergic neurons after pharmacological blockade of the tyrosine kinase (Trk) receptors in cell cultures of the visual cortex. In these experiments K252A blocked all BDNF–receptor interactions, thereby preventing all modulatory effects of BDNF on the GABAergic neurons.
For the visual system, BDNF has been suggested to modulate functional plasticity via regulation of GABAergic innervation and inhibition (Huang et al. 1999). For example, transgenic mice with an accelerated postnatal rise of BDNF revealed a precocious development of cortical GABAergic inhibition in the visual cortex (Huang et al. 1999; Gianfranceschi et al. 2003). In addition, these transgenic animals showed an earlier termination of the critical period for ocular dominance plasticity (Huang et al. 1999). If it is true that BDNF controls the inhibitory neurotransmission in the visual cortex, then one could expect a reduced strength of the GABAergic system under conditions of chronically reduced endogenous levels of BDNF. Since this has not been tested yet, the present study used a heterozygous BDNF KO mouse model (Korte et al. 1995), which expresses 40–50% reduced levels of BDNF in the visual cortex (Abidin et al. 2006). The model has been successfully used to describe changes in glutamatergic synaptic transmission in the hippocampus (Patterson et al. 1996; Pozzo-Miller et al. 1999), the cerebellum (Carter et al. 2002), and the visual cortex (Abidin et al. 2006; Bartoletti et al. 2002). This heterozygous KO mice model was used in the present study to evaluate the role of the neurotrophin on GABAergic inhibition in the visual cortex. In summary, the visual cortex of the BDNF (+/−) mice showed a reduced frequency and amplitude of miniature inhibitory postsynaptic currents (mIPSCs). Together with the prolonged decay time constants of evoked inhibitory postsynaptic currents (IPSCs), the reduced presynaptic release probability for GABA, the diminished glutamate driven GABA release, and the reduced overall efficiency of GABAergic inhibition, the present study provides evidence that the neurotrophic factor BDNF plays a critical role for the GABAergic function in the visual cortex.
The transgenic mouse model of the present study was originally established by Korte et al. (1995). In heterozygous knockout mice one allele of the BDNF coding region is replaced by a neomycine-resistance gene (BDNF (+/−)), which results in fully viable and fertile animals. Wild-type littermates served as controls. The presence of the transgene was verified in each experiment by polymerase chain reaction (PCR) from tail tissue. PCR primers were used to recognize BDNF (5′-ACC ATA AGG ACG CGG ACT TGT AC-3′) and neomycine (5′GAT TCG CAG CGC ATC GCC TT-3′), while 5′-GAA GTG TCT ATC CTT ATG AAT CGC-3′ was used as a reverse primer. Treatment of the animals was in accordance with the guidelines of the local animal research committee and with the laws of the European Union (EU).
Slice preparation and electrophysiology
A detailed description of the slice preparation and electrophysiological recordings has been described elsewhere (Abidin et al. 2006). In brief, wild-type (n= 29) and BDNF (+/−) mice (n= 26) at the age of 21–25 days were deeply anaesthetized with ether and decapitated. Coronal slices were prepared from the visual cortex (thickness: 350 μm) by use of a vibratome (Leica, VT 1000 S, Germany). The tissue was incubated in artificial cerebrospinal fluid (ACSF) containing (mm): 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 25 d-glucose, 2 CaCl2 and 1.5 MgCl2, bubbled with 95% O2 and 5% CO2 to pH 7.4. The tissue recovered at room temperature for 1.5 h before individual slices were transferred to a submerged recording chamber superfused with oxygenated ACSF at 30 ± 1°C. The recording chamber was mounted on the stage of an upright microscope (Olympus BX50-WI, Olympus, Japan) equipped with 2.5× and 40× water-immersion-type objectives. Whole-cell patch-clamp recordings were performed from layers II/III pyramidal neurons under visual guidance using DIC optics. Patch pipettes (4–6 MΩ) were pulled from borosilicate glass capillaries (GB 150F-8P, Science Products, Germany). All recorded neurons in layers II/III responded to depolarizing current injection with regular, frequency-adapting spikes (not shown), which is characteristic for cortical pyramidal neurons (McCormick et al. 1985; Connors & Gutnick, 1990). A recent study from our lab has shown similar spiking properties of action potentials in layer II/III neurons of the visual cortex in BDNF (+/−) mice compared to wild-type controls (Abidin et al. 2006). In addition, some recorded cells were intracellulary labelled with biocytin, and the immunohistochemical analysis of these fixated slices revealed a pyramidal shape of the biocytin labelled neurons (Abidin et al. 2006). Voltage-clamp recordings were not corrected for liquid junction potentials and series resistance. Data were digitized at 5 kHz and filtered at 3 kHz using a Digidata-1200 system with pCLAMP9 software (Molecular Devices, Sunnyvale, CA, USA). Only cells with high seal resistance of > 1 GΩ and series resistance of < 25 MΩ were included in the analysis. The series- and input-resistance was controlled before and after each recording, and cells were discarded if one or both parameters changed more than 20%. Miniature inhibitory postsynaptic currents (mIPSCs) were pharmacologically isolated by bath application of 20 μm DNQX (Tocris, Biotrend, Cologne, Germany) and 0.5 μm tetrodotoxin (TTX, ICS, Munich, Germany). The intracellular solution contained (mm): 120 CsCl, 2 MgCl2, 2 CaCl2, 2 Mg-ATP, 0.3 Na-GTP, 10 Hepes, 10 EGTA and 5 QX-314. The cells were voltage clamped to −80 mV and mIPSCs were recorded for at least 5 min.
Evoked GABAA receptor-mediated inhibitory postsynaptic currents (IPSCs) were pharmacologically isolated by bath application of the AMPA receptor antagonist DNQX (20 μm) and the NMDA receptor blocker d-AP5 (50 μm) (Tocris). Signals were evoked by placing a glass electrode (resistance 1 MΩ) in cortical layer IV to stimulate afferent fibres projecting onto pyramidal neurons in layers II/III, which were voltage clamped to −70 mV. The stimulus duration was constantly set to 50 μs, while the stimulus intensity was adjusted to yield a maximum IPSC amplitude. The rise time of IPSCs was calculated between 10% and 90% of the peak amplitude onset, and the decay time constant was obtained by fitting the decaying current to the following monoexponential function:
An average of six to eight paired IPSCs with an inter-stimulus interval of 50 ms was used to analyse the paired-pulse ratio (PPR), which was calculated from the amplitude ratio of the second IPSC to the first one. The failure/success rate of GABAergic synapses was evalulated by minimal synaptic stimulation through a glass electrode located in layer IV to evoke quantal IPSCs (meIPSCs) in neurons located in layers II/III. The position of the stimulating electrode was carefully adjusted to fulfil three criteria: (1) meIPSC signals should be evoked all or none; (2) the responses were similar in amplitude compared to mIPSCs recorded from the same cells (see Fig. 1), and (3) within each recording the latencies of meIPSCs were always within 5–10 ms following the stimulation artifact.
Further analysis of postsynaptic signals was based on the assumption that IPSCs can be approximated by a binomial model of synaptic transmission originally described by Katz (1969). This model suggests that (1) there is a constant number of release sites (Nsyn) that liberates vesicles with an average probability of (Pr), (2) a single vesicle produces an invariant (quantal) IPSC (q), (3) all release sites are independent, and (4) each release site liberates either a single vesicle or nothing in response to an action potential. It should be noted that ‘Nsyn’ reflects the number of synaptic release sites activated by the stimulus. It is different from the number of sites made by a single interneuron onto a pyramidal cell. In this binomial model the mean eIPSC amplitude is defined by the following equation:
Two approaches were applied to estimate the quantal size (q). First, miniature IPSCs (mIPSCs) were recorded in the presence of tetrodotoxin (TTX; 0.5 nm). Second, spontaneous IPSCs (sIPSCs) were recorded following trains of pulses (40 stimuli at 50 Hz) in the absence of TTX in the bathing solution (see Fig. 3). Since these sIPSCs contained events larger than the presumed quantal size, this would skew the distribution of sIPSCs. Thus, the quantal size (q) was estimated by calculating the median of sIPSCs rather than the arithmetic mean (Kirischuk et al. 2005). This method allowed us to estimate q from sIPSCs that were recorded from the same cells treated with the high frequency synaptic stimulation (HFS) protocol. This resulted in a better estimation of the quantal amplitude for each recorded neuron.
It is difficult to directly measure the number of release sites. Instead, the size of the readily releasable pool (RRP) can be used as its approximate (see Schneggenburger et al. 1999; Lu & Trussell, 2000). To estimate the RRP, high-frequency stimulation (50 Hz, 40 pulses, HFS) was used followed by the acquisition (duration: 4–5 min) of spontaneous IPSCs (sIPSCs) within the same recording. Stimulus intensities were adjusted to 1.2 times of the level for maximum response amplitudes. Repetitive stimulation led to a decrease in the eIPSC amplitudes (see Fig. 3). Assuming, that the eIPSC depression is primarily caused by a transient decrease in the number of readily releasable quanta, it is possible to estimate the RRP size on the basis of the cumulative eIPSC amplitude plot (Schneggenburger et al. 1999; Kirischuk & Grantyn, 2003). In accordance with this, we plotted the cumulative eIPSC amplitudes versus the stimulus number. After 20 pulses, the cumulative eIPSCs reached a steady state, as indicated by the linear slope dependence of the cumulative eIPSC amplitude on the pulse number (see Fig. 3D). Assuming that (1) the number of release sites remains constant throughout the experiment and (2) the linear component reflects vesicle recycling, the cumulative IPSC amplitude in the absence of pool replenishment can be estimated by back-extrapolation to the start of the train (Baldelli et al. 2005). Therefore, the release probability (Pr) could be estimated according to eqn (2) by dividing the measured amplitude of the first response of the train (first eIPSC) to the cumulative IPSC amplitude in the absence of pool replenishment (RRP):
After obtaining Pr, q and the eIPSC amplitudes, the number of release sites ‘Nsyn’ can be assessed by eqn (2). We measured the amplitudes of stimulus-locked eIPSCs to rule out a possible contribution of spontaneous IPSCs originating from non-stimulated synaptic contacts.
High frequency stimulation was also used to calculate the steady-state release of GABAergic synapses. The eIPSC amplitudes were normalized to the RRP (quantal size and the number of release sites), and signal responses were calculated for stimulation pulses 20–40 to finally calculate the ‘steady-state’ release of eIPSC. Thus, we estimated the total GABA release (synchronous plus asynchronous) by calculating the total charge transfer (Q) from eIPSCs recorded under steady-state conditions during the last 20 stimuli of the HFS (Fig. 4A). In this calculation ‘Q’ is defined as:
where qc represented the quantal charge (calculated from the sIPSCs recorded after HFS stimulation), Nsyn is the number of release sites, F represents the mean frequency of vesicle liberation at a single release site, and ISI is the inter-stimulus interval.
To compare the relative contribution of the excitatory input onto inhibitory neurons to the overall inhibitory strength in the cortical network, we analysed the glutamate-driven GABA release onto layer II/III neurons. Spontaneous and evoked IPSCs were recorded at 0 mV before and after application of glutamate antagonists (20 μm DNQX and 50 μm d-AP5). Here the intracellular solution contained (mm): 130 caesium gluconate, 8 KCl, 2 MgCl, 1 CaCl, 10 EGTA, 10 Hepes, 5 QX-314, 3 Na-ATP and 0.2 GTP. Relative changes in the frequency of sIPSCs and in the amplitude of eIPSCs were calculated compared to signals acquired in the absence of DNQX and d-AP5. The monophasic shape and the stimulus-to-onset latency of eIPSCs (14.1 ± 1 ms, n= 14) indicated signal responses from monosynaptic connections.
The ratio of the maximum inhibitory to maximum excitatory postsynaptic currents (maxIPSC/maxEPSC) was determined by recordings of eIPSCs and eEPSCs in layers II/III with an identical intracellular solution as used in the previous experiment. The electrical stimulation in layer IV was gradually increased (40, 80, 120, 160 and 200 μA) to reach maximum signal amplitudes at 200 μA (Morales et al. 2002). While eIPSCs were recorded at a holding potential of 0 mV, the eEPSCs were acquired in the same recording at a holding potential of −60 mV.
Whole-cell gramicidine-perforated patch-clamp recordings were performed with an intracellular solution containing (mm): 130 potassium gluconate, 4 MgCl, 1 CaCl, 0.5 EGTA, 20 Hepes, 5 QX-314, 3 Na-ATP and 0.2 GTP (pH = 7.3). Gramicidine (Sigma, Germany), 25 μg ml−1 dissolved in dimethylsulphoxide, was added to the intracellular solution, since it forms pores permeable to monovalent cations and small uncharged molecules but not to Cl−, thereby permitting reliable recordings of GABAergic currents (Kyrozis & Reichling, 1995; Owens et al. 1996). The perforated patch-clamp conditions were controlled by the Na+-channel blocker lidocaine N-ethyl bromide (QX-314, 5 mm) diluted in the intracellular solution. Since QX-314 is pore impermeable, an occurrence of Na+ channel-mediated action potentials confirmed the perforated membrane conditions at the recorded cell. The synaptic reversal potential of IPSCs was determined by varying the Vh of the postsynaptic cell in 10 mV increments from −100 to −40 mV and measuring the resulting IPSC amplitude. A best-fit line for the current–voltage (I–V) relationship was calculated using a linear regression, and the interpolated intercept of this line with the abscissa was taken as the reversal potential.
Parametric Student's t test was performed for statistical evaluation of the data. Results are presented as mean ±s.e.m., and differences were considered significant, if P < 0.05.
First, the basal GABAergic synaptic neurotransmission was studied in BDNF heterozygous KO mice by the recording of mIPSCs in pyramidal cells located in layers II/III. Application of picrotoxin (PTX) abolished the signals indicating that they were dominated by activation of GABAA receptors (Fig. 1A). The analysis of the cumulative probability of inter-event intervals indicated a right-ward shift in the curve for the BDNF (+/−) animals. In accordance, calculation of the frequency of mIPSCs showed a significant (P < 0.0001) reduction in slices from BDNF (+/−) mice (2.7 ± 0.1 Hz, n= 8) compared to wild-type controls (4.8 ± 0.2 Hz, n= 17) (Fig. 1Ba and b). Also, the mean amplitude of mIPSCs was significantly lower in heterozygous KO mice (BDNF (+/−): 27.2 ± 4.8 pA, n= 8; wild-type: 39.1 ± 2.8 pA, n= 17; P= 0.0074) (Fig. 1Ca and b). These results suggest an impairment of the basal GABAergic neurotransmission in the heterozygous BDNF (+/−) mice. In addition, we have analysed the kinetics of eIPSCs. The rise time of these signals was not different (P= 0.513), but we observed significantly prolonged decay time constants of eIPSCs in slices of BDNF KO mice (BDNF (+/−): 13.5 ± 0.9 ms, n= 9; wild-type: 10.4 ± 0.6 ms, n= 17; P= 0.006; Fig. Da and b, E and F). In accordance, the recorded mIPSCs also revealed an increased decay time constant in the BDNF (+/−) group (11.8 ± 0.7 ms, n= 9; wild-type: 9.8 ± 0.3 ms, n= 17; P= 0.003). These prolonged decay time constants could link to changes of the postsynaptic GABAergic function in BDNF (+/−) mice. However, it does not account for the reduced frequency of mIPSCs in the BDNF (+/−) mouse (see Fig. 1).
To disclose the possible presynaptic changes in GABAergic function of the heterozygous KO mice, we analysed the paired-pulse relation (PPR) of eIPSCs during repetitive synaptic stimulation (Fig. 2A). The PPR, defined as the ratio of the amplitude of the second eIPSC to the first one, was significantly (P= 0.013) larger in transgenic animals (BDNF (+/−): 0.86 ± 0.1, n= 13; wild-type: 0.66 ± 0.03, n= 7; (Fig. 2B). This suggested, in agreement with the reduced frequency of mIPSCs (see Fig. 1), an altered presynaptic GABA release in the BDNF (+/−) mice. Since the PPR can also be meditated by postsynaptic mechanisms, we additionally recorded eIPSCs under conditions of minimal synaptic stimulation (meIPSCs) (Fig. 2C). The success rate of these meIPSCs was significantly (P= 0.0011) reduced in BDNF (+/−) mice (0.365 ± 0.031, n= 16) compared to wild-type (0.503 ± 0.02, n= 26) animals. This result strongly supports a reduced presynaptic GABA release in BDNF (+/−) mice. In accordance with mIPSCs, the amplitudes of the meIPSCs were significantly (P= 0.0376) reduced in recordings of BDNF (+/−) mice (14.3 ± 1.7 pA) compared to wild-type littermates (20.2 ± 1.8 pA).
In general, the GABA release is influenced by the quantal size (q), the readily releasable pool (RRP), number of release sites (Nsyn), the steady-state release and/or the asynchronous GABA release. We tried to separate these parameters by use of the binominal model of synaptic plasticity (Katz, 1969) through the analysis of recordings of spontaneous IPSCs (sIPSCs) right after high frequency synaptic stimulation (HFS) within the same experiment (see Fig. 4A and Methods). During HFS already the first eIPSC showed a smaller signal amplitude in slices from BDNF (+/−) mice (247.5 ± 36.3 pA, n= 11) compared to controls (531.9 ± 72.8 pA, n= 12, P= 0.00272) (Fig. 3A and B). This might be caused by a reduced mean quantal size at GABAergic synapses in the KO animals. For verification we estimated the quantal size by calculating the median of the spontaneous IPSCs (sIPSCs), which were recorded immediately after the HFS within the same recording (Kirischuk et al. 2005). The analysis revealed, as expected, a significantly reduced quantal amplitude obtained by analysis of the median of sIPSCs in the BDNF (+/−) mice (BDNF: 16.1 ± 1.3 pA, n= 11; wild-type: 21.2 ± 1.6 pA, n= 12, P= 0.0266; Fig. 3C), which could explain the reduced amplitude of the first eIPSC during HFS. From the same experiment we calculated the readily releasable pool (RRP) by back-extrapolating the linear phase of the cumulative amplitude plot of eIPSCs to the ordinate (Fig. 3D) (Schneggenburger et al. 1999; Kirischuk & Grantyn, 2003). When the first amplitude of the eIPSCs was divided by the RRP, it disclosed the release probability (Pr, see Methods). The Pr was significantly reduced in transgenic mice (BDNF (+/−): 0.20 ± 0.01, n= 11; wild-type: 0.31 ± 0.02, n= 12; P= 0.000514; Fig. 3E). We then normalized the RRP to the estimated quantal amplitudes to disclose the number of release sites (Nsyn). As shown in Fig. 3F, this parameter remained unchanged between the two animal groups (BDNF (+/−): 85.6 ± 14.3, n= 11; wt: 91.4 ± 13.4, n= 12; P= 0.769).
The HFS induced an initial depression of the eIPSCs in both groups followed by steady-state-like eIPSC amplitudes after about 20 stimuli. Since these steady-state eIPSCs provide information on the level of GABAergic vesicle recycling, we normalized the signals by dividing the absolute eIPSC amplitude to the parameter ‘q’ and to the number of release sites (Nsyn) derived from the same cell (Fig. 4A). This allowed us to define the level of vesicle recycling while taking into account a reduced level of q and the possible variability of RRP in the BDNF heterozygous heterozygous KO mice (see Methods). As shown in Fig. 4B, the resulting normalized steady-state eIPSCs of the last 20 stimuli were significantly larger in wild-type cells (0.09 ± 0.01, n= 12) compared to (P= 0.0018) data from the BDNF (+/−) mice (0.05 ± 0.01, n= 11) indicating a reduced recycling of vesicles in the transgenic animals.
The impaired expression of BDNF might have changed the presynaptic Ca2+ concentration, thereby affecting both the stimulus-locked (synchronous) as well as the asynchronous GABA release. Thus, we estimated the total GABA release (synchronous plus asynchronous) by calculation of the total charge transfer (Q) from eIPSCs recorded under steady-state conditions during the last 20 stimuli of HFS (see Methods). As a result, cells from wild-type mice reached a total charge transfer of 75.7 ± 7.6 pA * s that was significantly (P < 0.0001) higher compared to the BDNF (+/−) mice (24.7 ± 3.7 pA * s; Fig. 4C). Also the frequency of the steady-state GABA release was significantly reduced (P= 0.001) in the transgenic animals (BDNF (+/−): 15.1 ± 1.7 Hz, wild-type: 7.3 ± 1.2 Hz; Fig. 4D. In summary, both the reduced steady-state vesicle release and the impaired frequency of steady-state release suggest that the GABAergic synapses in BDNF (+/−) visual cortex are characterized by a reduced recycling of vesicles.
The strength of GABAergic inhibition in visual cortex is also affected by the glutamate-driven GABA release onto the pyramidal neurons of layers II/III. This was tested by estimating the relative proportion of sIPSCs which are driven by glutamate to the total number of sIPSCs within the same neuron. As expected, bath application of the glutamate receptor antagonists DNQX and d-AP5 decreased the frequency of sIPSCs in both groups. However, the frequency of sIPSCs was reduced to a lesser extent in neurons from BDNF (+/−) mice (8.0 ± 4.6%, n= 7) compared to wild-type tissue (21.7 ± 4.1%, n= 7; P= 0.0222, Fig. 5A and B). In addition, the peak amplitudes of eIPSCs measured in the presence of DNQX and d-AP5 were also reduced more prominently in slices from wild-type animals (wild-type: 27.1 ± 3.5%, n= 7; BDNF (+/−): 11.7 ± 3.9%, n= 7; P= 0.0106; Fig. 5C and D). These results indicate that a chronic reduction in the expression of BDNF not only directly affects inhibitory function by impairing the presynaptic release at GABAergic synapses, but it also reduces indirectly the strength of the GABAergic system by reducing glutamatergic input to the GABAergic neurons.
Since we have recently shown an impaired glutamatergic transmission at the same cortical neurons by use of the BDNF (+/−) model (Abidin et al. 2006), we wanted to evaluate the implications of a reduced excitation and inhibition for the function of the visual cortical network. Therefore, we tested the strength of the excitatory and inhibitory drive, as well as the maximum strength of excitatory and inhibitory synaptic inputs onto the pyramidal neurons in layers II/III. As shown in Fig. 6A, the gradual increase in stimulation intensity of ascending fibres located in layer IV led to eEPSCs and eIPSCs, depending on the holding membrane potential, with a rising signal amplitude. While the amplitudes of maximum evoked glutamatergic EPSCs showed no differences between the two groups (wild-type: 1152.8 ± 101.9 pA, n= 6; BDNF (+/−); 1150.9 ± 79.4 pA, n= 8), the eIPSC signals revealed a 36% smaller amplitude under maximum stimulation conditions in the BDNF (+/−) mice group (wild-type: 1244.4 ± 100.6 pA, n= 6; BDNF (+/−): 792.26 ± 37.4 pA, n= 8). The slow component of the excitatory currents was mediated by NMDA receptors, since it is abolished by application of d-AP5 (50 μm) (not shown). This obvious imbalance between the inhibitory and excitatory synaptic inputs onto layer II/III cells in the BDNF (+/−) mice was quantified by plotting the ratio of the maximum amplitude of eIPSCs versus the maximum amplitude of eEPSC. This ratio was significantly (P= 0.0073) reduced in cells from BDNF (+/−) mice (0.69 ± 0.1 versus wild-type: 1.08 ± 0.1, Fig. 6B). The reduction in the amplitude of IPSCs under maximum stimulation conditions might be compensated by a prolonged decay time constant (see Fig. 1). Therefore, we calculated the total mediated charges during maximum stimulation conditions. However, the ratio of these inhibitory to excitatory charges was still significantly (P= 0.0306) smaller in the BDNF (+/−) group (1.18 ± 0.08) compared to wild-type mice (1.55 ± 0.14, Fig. 6C).
This imbalance might be compensated through possible postsynaptic mechanisms, for example through an altered intracellular Cl− concentration in the recorded layer II/III neurons. It is known that BDNF can change the strength of inhibition by regulating the postsynaptic Cl− concentration (Wardle & Poo, 2003; Rivera et al. 2004). This could counteract the relatively imbalanced and excitatory dominated synaptic inputs. Since changes in postsynaptic [Cl−]i would shift the reversal potential of GABAA-mediated currents (EIPSC), we calculated EIPSCs in both experimental groups by the recording of synaptically evoked GABAA-mediated currents (eIPSCs) at different membrane holding potentials in the whole-cell perforated patch-clamp mode (Fig. 7A and B). The resulting current–voltage relationship of GABAA receptor-mediated chloride currents was similar in both experimental groups, and consequently no changes were observed in the EIPSC (wt: −67.3 ± 2.2 mV, n= 17; BDNF (+/−): −68.8 ± 5.5 mV, n= 8; P= 0.7702) (Fig. 7C). In accordance with the initial experiments shown in Fig. 1F the decay time constants of eIPSCs under perforated patch-clamp conditions were significantly (P= 0.0283) longer in BDNF(+/−) neurons (16.1 ± 0.7 ms, n= 8) as compared to wild-type controls (11.6 ± 0.7 ms, n= 17; Fig. 7D).
While there is undisputed evidence that reduced levels of BDNF impair the efficacy and plasticity of excitatory synaptic transmission in the visual cortex (Bartoletti et al. 2002; Abidin et al. 2006), so far no data are available regarding the effects of long-term reduced levels of BDNF on the development and maturation of GABAergic inhibition in this brain area. Therefore, the present study investigated the effects of chronically reduced levels of BDNF on GABAergic synapses in the visual cortex of heterozygous BDNF knockout mice. This model has been used successfully to describe changes in glutamatergic synaptic transmission in hippocampus (Pozzo-Miller et al. 1999), cerebellum (Carter et al. 2002) and visual cortex (Bartoletti et al. 2002; Abidin et al. 2006). As we have recently shown, the level of BDNF protein in these mice is reduced by about 45% in all neocortical areas of 21-day-old heterozygous BDNF (+/−) mice.
The function of GABAergic synapses was initially tested through the analysis of mIPSCs. The frequency of these signals is known to gradually increase during postnatal cortical development (Dunning et al. 1999). Therefore, we verified that the frequency of mIPSCs was unchanged in neurons from wild-type mice within the selected time window of 21–25 days (data not shown). Compared to control data, the neurons from BDNF (+/−) mice showed a decreased frequency and amplitude of mIPSCs, which indicated a reduced basal inhibitory synaptic neurotransmission. At first, this result seemed to be contradictory to data from others, who showed a reduced frequency of mIPSCs following acute application of BDNF (Tanaka et al. 1997). However, BDNF-mediated effects critically depend on the protocol of BDNF elevation/reduction that is employed: most of the experiments that dealt with acute applications of BDNF resulted in a reduced frequency and size of mIPSCs (Tanaka et al. 1997; Frerking et al. 1998), whereas chronic application of surplus BDNF increased the frequency of mIPSCs and the size of spontaneous and evoked IPSCs (Vicario-Abejon et al. 1998; Bolton et al. 2000; Baldelli et al. 2002, 2005). Thus, in this context, our results are in line with others who studied the effects of chronically altered levels of BDNF. The impaired mIPSC signals might be explained by either a reduced anatomical arborization of dendritic and axonal processes and/or by a reduced presynaptic release probability. The latter interpretation is supported by our results of an increased paired-pulse ratio (PPR) and enhanced failure rate of meIPSCs observed in the BDNF (+/−) visual cortex.
The smaller amplitudes of mIPSCs and eIPSCs in BDNF (+/−) mice were caused by a reduced quantal size that is assembled by (1) the transmitter content of individual vesicles and (2) the total number of postsynaptic GABAA receptors. (1) The chronically reduced expression of BDNF could change the transmitter content of individual vesicles, since it has been shown that the neurotrophin affected presynaptic expression of glutamic acid decarboxylase (GAD), a rate-limiting enzyme in GABA synthesis, as well as the GABA uptake at the GABAergic terminals (Huang et al. 1999; Bolton et al. 2000; Rico et al. 2002; Yamada et al. 2002). Hence, a decreased GABA content in the vesicles would in turn reduce the quantal size. (2) Other studies showed a BDNF-mediated change in the number of functional cell surface GABAA receptors (Yamada et al. 2002; Mizoguchi et al. 2003). Therefore, both parameters could have influenced the quantal size at GABAergic synapses in the present KO animals. The data of the present study do not disclose any information as to if the reduced quantal size in the BDNF (+/−) animals are caused by pre- and/or postsynaptic mechanisms. However, the reduced release properties of GABAergic vesicles is one important factor for the overall impaired efficiency of inhibition in the visual cortex of BDNF (+/−) mice.
Both evoked and miniature IPSCs disclosed alterations in the functional properties of GABA receptors as shown by the prolonged decay time constants. This could originate from changes in the composition of GABAA receptor subunits. Increasing expression of the α1 and α6 subunits of the GABAA receptor is associated with faster kinetics of spontaneous and miniature IPSCs (Farrant et al. 1999; Vicini et al. 2001). During postnatal development the expression of the α1 and α6 subunits increases, while the α2 and α3 subunits are down regulated (Laurie et al. 1992; Poulter et al. 1992). As a result, faster decay time constants of IPSCs were observed with increasing age of the animals (Tia et al. 1996). In this regard, the prolonged decay time constants observed in the present study are in agreement with a delayed postnatal development in the GABAergic system of the BDNF (+/−) mouse. In addition, the decay time constant of IPSCs is influenced by GABA transporters. Activity of these transporters regulates the presence of GABA in the synaptic cleft, thereby limiting the duration of the neurotransmitter action. This was shown in recordings of knockout mice lacking the GAT1 GABA transporter, in which GABAA receptor-mediated sIPSCs revealed prolonged decay time constants (Chiu et al. 2005) and since it is known that tyrosine kinase regulates the activity of these transporters (Law et al. 2000), an impaired activity of GAT1 transporters might account, at least in part, for the prolonged decay time constants of IPSCs in our BDNF (+/−) mice. In regard to the overall function of GABAergic synapses it should be noted that the prolonged decay time constants would increase the total charge of GABAergic currents, thereby acting as a compensatory mechanism for the overall reduced GABAergic efficacy due to the reduced quantal size and peak amplitude of eIPSCs.
The characteristics of presynaptic GABA release were further examined by analysis of the readily releasable pool (RRP). The calculation of RRP at GABAergic synapses using a similar methodical approach has been employed successfully in hippocampus (Baldelli et al. 2002, 2005), superior colliculus (Kirischuk & Grantyn, 2003; Kirischuk et al. 2005), cerebellum (Saitow et al. 2005), and also in the cerebral cortex (Kirmse & Kirischuk, 2006). The calculation of presynaptic release probability (Pr) through analysis of high frequency stimulation is based on the assumption that a binomial model can approximate IPSCs. For inhibitory synapses, the model predicts that each release site liberates either a single vesicle or nothing in response to an action potential (see Methods). Although GABAergic synapses might release more than one GABA-containing vesicle, as shown in cerebellum (Auger & Marty, 1997) and hippocampus (Biro et al. 2006), this multivesicular release is rather asynchronous and not exactly coinciding. In contrast, we always analysed the stimulus locked, synchronous GABA release from HFS-mediated eIPSCs. Furthermore, our independent experiment on the minimal stimulation of GABAergic synapses further supports our result of a reduced presynaptic release probability for GABA in the BDNF (+/−) mice. Therefore, the HFS experiment provides evidence that a chronic reduction of BDNF levels leads to a marked decrease in the amplitude of evoked IPSCs, and a decrease in the GABA release probability. These results are compatible with two other studies, in which exogenously added, elevated BDNF induced an increase in the release probability of inhibitory synapses at cultured hippocampal neurons (Baldelli et al. 2002, 2005). The authors suggested that the release probability was altered through changes in the activity of presynaptic N- and P/Q type Ca2+ channels. In accordance with this, chronic application of BDNF led to an increase in the release probability of GABA through regulation of presynaptic glutamic acid decarboxylase (GAD) in cultured neurons of the superior colliculus (Henneberger et al. 2005) and in hippocampus (Ohba et al. 2005).
Our electrophysiological data argue against significant changes in the number of GABAergic synaptic contacts onto layer II/III pyramidal neurons in the visual cortex of the BDNF (+/−) mice. First, the number of release sites (Nsyn) remained unchanged. One study showed an enhanced arborization of GABAergic neurons after BDNF application that gave rise to more synaptic contacts (Vicario-Abejon et al. 1998). However, compared to our study the authors used a different methodical approach by use of cell cultures and a chronic application of the neurotrophin. In this regard, their results might be due to different experimental conditions, especially to non-comparable BDNF concentrations. In contrast, Kohara et al. (2007) reported that pyramidal neurons, which are unable to synthesize BDNF at all, received less inhibitory contacts in organotypic cultures of the visual cortex. They suggested that BDNF, as released from postsynaptic target neurons, promotes the formation or proliferation of GABAergic synapses through its local actions in layers II/III of visual cortex. This contradictionary result might be explained by the different experimental models. We have recently shown that heterozygous BDNF (+/−) mice, as used in the present study, still synthesize ∼50% of the BDNF protein compared to wild-type (Abidin et al. 2006) and here we observed no changes in the number of inhibitory release sites. This suggests that a ∼50% reduced level of BDNF expression might still be enough to sustain a similar amount of GABAergic contact as in wild-type animals. In this regard, there might be a critical threshold of BDNF expression, where an expression level below this threshold could affect the number of GABAergic terminals. Yet, our result is in good agreement with anatomical data derived from the same animal model. Genoud et al. did not find any alterations in the density of synapses in the somatosensory cortex of BDNF (+/−) mice (Genoud et al. 2004). According to these data it seemed unlikely that an altered number of GABAergic synaptic contacts accounted for the impaired GABAergic function in the BDNF (+/−) mice.
In general, cells recorded from heterozygous KO mice showed a reduced amplitude of eIPSCs. This might be due to differentially regulated subtypes of interneurons. In the visual cortex, distinct types of interneurons showed different release properties (Hefft & Jonas, 2005). Since BDNF is known to selectively increase the number (Alcantara et al. 1997; Huang et al. 1999) and arborization (Berghuis et al. 2006) of parvalbumin-expressing interneurons, we cannot exclude that the reduced expression of BDNF in our mice model affected the composition of specific subtypes of GABAergic interneurons. In particular, the degree of modification by BDNF could vary between different types of interneurons. Since the specific types of interneurons in the cortex are generated at different points during brain development (Cavanagh & Parnavelas, 1988, 1989), they get unequally exposed to the neurotrophin. However, our electrophysiological result of an unaltered number of release sites (Nsyn) should be carefully interpreted in regard to the possible morphological modifications in our animal model. One can expect changes in the number of release sites upon reduced levels of BDNF; however, we found that Nsyn remained unchanged in the BDNF (+/−) mice. It must be noted that Nsyn reflects the number of functional release sites, and we cannot exclude that the number of immature or silent synapses is different in the BDNF (+/−) mice. In addition, Nsyn was rather indirectly calculated from a model of binomial release through electrophysiological recordings.
It should be noted that the inhibitory signals recorded in neurons from cortical layers II/III are originating from possibly different populations of interneurons. The eIPSCs were exclusively generated by electrical stimulation of afferent fibres located in layer IV, while mIPSC signals originated from all (vertical and horizontal) inhibitory inputs that project onto neurons in layers II/III. As a consequence, eIPSCs and mIPSCs were generated in part by possibly different types of interneurons, which could differ with respect to their quantal IPSC amplitude, release probability and number of contacts. However, the differences might not be as significant, because the strong majority of inhibitory inputs received by neurons in layers II/III is mediated via projections from layer IV (Morales et al. 2002)
The GABAergic inhibition in the visual cortex is also mediated by the strength of the excitatory drive onto the inhibitory neurons. Although we did not record directly the pyramidal neuron-driven excitatory input (E) onto interneurons (I) (E→I), our indirect methodical approach is sufficient to calculate the strength of glutamatergic synaptic inputs onto the GABAergic neurons. The reduced excitatory drive onto inhibitory neurons observed in the heterozygous KO mice is in good agreement with data on chronically applied BDNF in cortical cell cultures, which led to an enhanced excitatory synaptic transmission onto GABAergic neurons (Nagano et al. 2003). Our results imply that the visual cortical interneurons in BDNF (+/−) mice receive a weaker excitatory input and this can explain, at least in part, the observed reduction of GABAergic neurotransmitter release. In addition, we have recently shown that the presynaptic excitatory glutamate release was also impaired in BDNF (+/−) mice (Abidin et al. 2006). Hence, the overall cortical network activity in BDNF (+/−) mice might be described by a reduced presynaptic GABA release, which is simply counteracted by the reduced efficacy of the glutamatergic system. We directly tested this hypothesis, and the direct comparison of the strength of excitation versus inhibition by use of maximum synaptic stimulation conditions showed a relatively stronger impairment of the GABAergic inhibition in the visual cortical network of the heterozygous KO mice (see Fig. 6). It should be noted that the reduced strength of inhibition was exclusively measured through eIPSC amplitudes while ignoring the prolonged decay time constants of eIPSCs in the KO animals (see Fig. 1F). These prolonged decay time constants might compensate the reduced signal amplitudes during short bursts of transmission by increasing the total charge transfer, thereby mediating inhibition to levels as high as in wild-type animals, but with a different temporal distribution. Indeed, the reduction in charge transfer was not as severe as for the eIPSC amplitudes (Fig. 6C), but the analysis still disclosed a reduced strength of inhibition in terms of charge transfer.
The results of the present study are in line with the hypothesis of a delayed functional GABAergic maturation in the visual cortex due to reduced expression of BDNF in heterozygous KO mice. Since activity from chloride (Cl−) transporters and the resulting postsynaptic [Cl−]i plays a critical role during development of inhibitory synapses (Owens et al. 1996; Rivera et al. 1999; Ehrlich et al. 1999; Ganguly et al. 2001), one might speculate that neurons in the premature visual cortex of BDNF (+/−) mice may contain a relatively high [Cl−]i, as typically observed in animals older than postnatal day 12 (P12). This would result in more depolarized reversal potentials for IPSCs (EIPSCs) (Thompson & Gähwiler, 1989; Kaila, 1994). However, we did not observe any differences in the EIPSC values between the two groups. This is not surprising, since (1) one other study also failed to show any effects of BDNF on postsynaptic [Cl−]i in excitatory neurons of hippocampal cultures (Wardle & Poo, 2003) and (2) the expression of the neurotrophin is reduced to about 45% in the present BDNF (+/−) mice at the age of 21–25 days (Abidin et al. 2006). Possibly depolarized EIPSCs might be visible much earlier in development only, e.g. around P12 or earlier (Rivera et al. 2005).
In summary, our findings strongly suggest that chronically reduced levels of BDNF impair the GABAergic inhibitory function in the visual cortex, preferentially by reducing the presynaptic GABA release as well as the overall strength of inhibition onto pyramidal neurons within the cortical network. Since growing evidence suggests that the maturation of synaptic inhibition controls the timing of the critical period for visual cortical plasticity (Kirkwood & Bear, 1994; Hensch et al. 1998; Hensch, 2005), the present BDNF (+/−) mice could represent a model of delayed cortical postnatal development. Although Bartoletti et al. (2002) reported an unchanged critical period for visual cortical plasticity in a different BDNF (+/−) mouse model (Bartoletti et al. 2002), others have shown that overexpression of BDNF caused precocious development of visual acuity and earlier termination of the critical period (Huang et al. 1999). In this regard BDNF promotes the maturation of the GABAergic inhibition, thereby controlling the timing for critical period plasticity in the visual cortex during early postnatal life.
The authors thank Petra Kuesener for excellent technical assistance. This work was supported by the Deutsche Forschungsgemeinschaft (SFB 509, C4) and The International Graduate School of Neurosciences (IGSN) Bochum.