A long-standing hypothesis predicts that pyramidal neurons of the cerebral cortex control the influx of sensory information at the level of primary sensory representations areas. Yet little is known about the cellular mechanisms governing selective attention to behaviorally relevant objects in space. Neurons in the superficial layers of the superior colliculus are notably involved in this process, and they are directly targeted by retinal and cortical afferents. To study long-term and short-term effects of the visual cortex (VC) on subcortical visual neurons we established an in vitro model of the developing cortico-tectal projection. To this end, cortical explants expressing Green Fluorescent Protein were allowed to form connections with non-labeled dissociated tectal neurons. The presence of VC explants led to an enhancement of tectal activity by 2 mechanisms. First, glutamatergic input was increased. Second, intrinsic GABAergic inhibition was suppressed. The latter effect was shown to be acute and mediated through postsynaptic metabotropic glutamate receptor activation, G-protein acitivity, and endocannabinoid receptor activation. The VC-induced disinhibition was readily reversed by application of an mGluR antagonist. However, high-frequency activation of the glutamatergic cortico-tectal input turned the labile disinhibition into a persistent suppression of inhibition.
During the so-called “critical period” of sensory development (Berardi et al. 2000; Hensch 2005; Majdan and Shatz 2006) maturation of the cortical efferent pathways may facilitate the experience-dependent rearrangement of neuronal connections in subcortical sensory areas, including the superior colliculus (SC). The axons from the visual cortex (VC) arrive rather late in the superficial layers of the SC (Thong and Dreher 1986) and meet there an already well-developed neuronal network formed by the axons of retinal ganglion cells, GABAergic interneurons and the collaterals of collicular projection neurons (for references see Leichnetz and Goldberg 1988; Grantyn et al. 2004; Mize and Salt 2004). Nonetheless, the resultant visual maps must eventually be in register (McIlwain 1975) for the SC to perform its well-established role in the initiation of attention shifts and orienting movements.
From the viewpoint of activity balance in the developing brain, addition of a strong excitatory input is likely to represent a major perturbation. Accordingly, one could expect a variety of adaptive responses of the preexisting neuronal network, ranging from a massive overall increase of the spontaneous spike discharge to homeostatic maintenance of the activity levels throughout development (Desai 2003; Turrigiano and Nelson 2004). Based on the temporal coincidence between the invasion of the SC by VC axons and changes in the properties of GABAergic synapses it has been suggested that cortico-tectal afferents may influence their target by regulating the strength of GABAergic inhibition (Aamodt et al. 2000; Henneberger, Jüttner, et al. 2005). Indeed, intact GABAergic inhibition has been shown to be required for the adjustment of multisensory activity patterns under condition of acute cross-modal activation (Skaliora et al. 2004) or persistent mismatches between visual and auditory input (Zheng and Knudsen 1999). However, direct evidence for a modification of GABAergic synaptic transmission by VC afferents is missing. In general, there is still little information available on the cellular and synaptic mechanisms underlying cortical efferent control of sensory activity in the SC, although this issue continues to attract attention (Hashemi-Nezhad et al. 2003; Bereshpolova et al. 2006).
One of the obstacles is that in slice preparations of the VC or SC cortico-tectal axons are necessarily damaged. We have therefore developed a method to overcome these limitations and to be able to directly evaluate the capacity of cerebral cortical input to modify the strength and regulatory range of GABAergic synaptic transmission in the SC. Axons from VC explants expressing Enhanced Green Fluorescent Protein (EGFP) were allowed to grow into a preformed network of dissociated neurons obtained from the murine tectum. As far as unitary GABAergic synaptic transmission is concerned, in vitro development of tectal neurons was shown to reproduce some major features of in vivo superficial SC development (Kirischuk et al. 2005). By comparing the effects of added VC with those of added SC or no added explants, this experimental model was expected to provide first insight into the mechanisms available for the execution of cortical efferent control, at least during the postnatal period. Specifically, we aimed at answering the following 3 questions: 1) How does the in-growth of VC axons affect the balance of excitation and inhibition in subcortical sensory neurons? 2) Do cortico-tectal afferents modulate tecto-tectal GABAergic inhibition and, if so, what is the site and duration of this modulation? 3) Which receptors/messengers underlie the possible transient or persistent actions of cortico-tectal afferents?
Material and Methods
Neuron Cultures and Explants
The cultures were prepared from mice. All animals were sacrificed according to the permit given by the Office for Health Protection and Technical Safety of the regional government of Berlin (Landesamt für Arbeitsschutz, Gesundheitsschutz und technische Sicherheit Berlin, T0121/03) and obeyed the rules laid down in the European Community Council Directive 86/609/EEC.
The SC was dissected on embryonic day (E) 17 from C57Bl/6J mouse fetuses, as described previously (Perouansky and Grantyn 1989; Henneberger, Kirischuk, et al. 2005). Cells were seeded at a density of 120 000 cells/cm2 on laminin-coated glass coverslips (diameter 15 mm) and cultured in 1 ml of minimal Eagle's medium (Gibco/Invitrogen, Karlsruhe, Germany) supplemented with 5% horse serum and 5% fetal calf serum (Gibco/Invitrogen), 30 mM glucose, 12 mM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid (HEPES), 21 mM NaHCO3, 2 mM glutamine, 2 mM pyruvic acid, and 37.5 μg/l insulin at 37 °C in 5% CO2/95% O2. Cytosine-β-D-arabinofuranoside (4 μM) was added to all cultures 7 days after plating to inhibit glial proliferation. If not stated otherwise, all chemicals were purchased from Sigma-Aldrich Chemie GmbH (Taufkirchen, Germany).
Explants of the VC or SC were prepared from E17 embryos expressing EGFP under control of the β-actin promoter. For that purpose, female C57/Bl6 and male heterozygous β-actin-EGFP (C57BL/6-Tg(ACTB-EGFP)1Osb/J, Jackson Laboratories, Bar Harbor, ME) were mated. The brains of fetuses were genotyped by using ultraviolet light. The validity of genotyping was verified by PCR using tail biopsies and primer sequences supplied by Jackson Laboratories. Three to 7 explants with diameters between 300 and 700 μm were placed 1 week after the start of collicular cultures. The distance between explants ranged from about 500 to 1000 μm. The day of collicular dissociation will be referred to as day in vitro (DIV) 1.
For the electrophysiological tests the cultures were transferred to the stage of an upright microscope (Axioskop 2 FS plus, Zeiss, Oberkochen, Germany) and maintained under continuous superfusion (flow rate 0.1 ml/min) with a standard salt solution (SSS) containing in (mM) 150 NaCl, 3 KCl, 20 HEPES, 30 glucose, 2 CaCl2, 1 MgCl2 (pH adjusted to 7.4 using NaOH). After removal from the incubator cultures were allowed to equilibrate for 30 min. All experiments were done at room temperature (20–22 °C). The cultures were used for no longer than 2 h.
Whole-cell currents were recorded from nonfluorescent collicular neurons at distances of up to 500 μm from an explant. To this end, patch electrodes were pulled from borosilicate glass capillary tubing (WPI, Sarasota, FL) using a P-87 puller from Sutter Instruments (Novato, CA). The pipette solution contained (mM): 45 K-gluconate, 100 KCl, 10 ethylene glycol-bis(β-aminoethyl ether)N,N,N′,N′-tetraacetic acid (EGTA), 5 NaCl, 20 HEPES, 1 CaCl2, 3 Mg-adenosine-5′ triphosphate, and 0.3 Na-guanosine-5′ triphosphate (pH adjusted to 7.2 using KOH). The calculated ECl− was −9 mV. If the Ca2+ buffer/Ca2+ concentration was varied an equimolar amount of K-gluconate was added or removed as necessary. Lidocaine N-ethyl bromide (QX314, 2 mM, RBI, Natick, MA) was included in the pipette solution to prevent action potential (AP) generation in the neuron under investigation except for recording of miniature postsynaptic currents (mPSCs). The pipette-to-bath direct current (DC) resistance ranged from 4 to 6 MΩ (tip diameter 1–2 μm). EGFP-negative neurons and single EGFP-positive/negative axons were selected for recording and stimulation using a 63× water immersion objective, phase contrast illumination, and epifluorescence at an excitation wavelength of 480 nm (Till Photonics, Munich, Germany). Synaptic currents were recorded in the whole-cell patch clamp configuration using the patch clamp amplifiers EPC-8 (HEKA Elektronik, Lambrecht, Germany) and Axopatch 200B (Axon Instruments Inc., Foster City, CA). The series resistance was monitored throughout the experiment and compensated (70–80%). Recordings were rejected if the initial RS was >25 MΩ. Neurons were held at −70 mV. Unitary evoked PSCs (ePSCs) were elicited by extracellular electrical stimulation of a solitary axon in the vicinity of the postsynaptic cell. The stimulation pipettes (tip diameter about 1 μm, pipette-to-bath DC resistance 5–10 MΩ) were filled with SSS and located at a distance of <1 μm from the axon. Isolated stimulation units were used to generate rectangular pulses of 0.5 ms duration. Stimulus intensities ranged from 0.5 to 1 μA. Synaptic responses were regarded as “unitary” PSCs if 1) latencies were short (<5 ms) and stable, 2) the shape of the postsynaptic response remained unchanged throughout the experiment, and 3) a more than 20% reduction of stimulus intensity resulted in a complete failure of the ePSCs. Stimulation configurations resulting in polysynaptic responses were discarded. A small (<1 μm) movement of the stimulating pipette away from the presumptive axon immediately resulted in a complete loss of the postsynaptic response.
Spontaneous spike discharge was recorded in the loose-patch cell-attached configuration at 0 mV. In this case pipettes were filled with extracellular solution. Signals were filtered at 3 or 5 kHz and sampled at a rate of 10 kHz using a 16-bit AD-converter (ITC-16, Instrutech, Port Washington, NY) and the software WinTida 4.11 (HEKA Elektronik, Lambrecht, Germany). PSCs were analyzed using WinTida and an in-house written software for semiautomatic detection of synaptic currents (C. Henneberger). The decay of ePSC, spontaneous PSC, and mPSC could well be fitted using a single exponential function yielding the decay time constant τ.
In a subset of experiments glutamatergic synaptic transmission was blocked by 6,7-dinitroquinoxaline-2,3(1H,4H)-dione (DNQX, 20 μM) and DL-2-amino-5-phosphonopentanoic acid (APV, 100 μM). The remaining PSCs were fully and reversibly blocked by the GABAAR antagonist gabazine (1 μM, Tocris) and therefore considered to be GABAAR-mediated IPSCs. The following substances were purchased from Tocris and used for pharmacological experiments (S)-3,5-dihydroxyphenylglycine (DHPG, 20 μM), (RS)-α-methyl-4-carboxyphenylglycine (MCPG, 200 μM), AM251 (0.5 μM), and WIN 55,212-2 mesylate (WIN, 1 μM).
Statistical analysis was performed using the software Origin 7.5 (OriginLab Corporation, Northampton, MA). Numerical data are reported as mean ± standard error of the mean, with n being the number of neurons studied. If not explicitly stated otherwise, analysis of variance followed by paired or unpaired Student's t-tests were used for statistical comparison. In the figures significance levels are indicated as *P < 0.05, **P < 0.01, ***P < 0.001, or n.s. for “not significant.” The number (n) of neurons or synaptic connections studied is given in brackets above the respective data points.
The Presence of VC Explants Increases the Strength of Glutamatergic Input to Collicular Neurons
Explants were either derived from the VC or the SC. One week after their placement on the preexisting collicular cultures, that is, by DIV 14, both types of explants exhibited substantial fiber outgrowth, as illustrated in Figure 1(A,B). The fibers were negative for the dendritic marker microtubule-associated protein 2 (supplementary Fig. 1), except for the immediate vicinity of explants. Neurons occasionally migrated out of the explants but due to their strong EGFP signal it was easy to distinguish between the fluorescent explant neurons and the nonfluorescent neurons of the preexisting SC culture. The activity of the latter was studied between DIV 14 and 21.
At the time of recording neuron differentiation and network activity were quite advanced, as indicated by the fact that virtually all tested neurons displayed spontaneous synaptic activity. The first question to be addressed was, to what extent does the spontaneous spike discharge depend on ionotropic transmitter receptors, and what is the role of GABA? It turned out that the spontaneous AP generation was 1) mainly driven by the activity of ionotropic glutamate receptors and 2) disinhibited by the block of GABAARs (Fig. 1C,D). Thus, a reduction in GABA release will increase the frequency of spontaneous spike discharge, as GABA acts as inhibitory neurotransmitter.
The incidence of spontaneous spike discharge was different in collicular neurons growing in the absence or presence of a VC explant. The respective values were 35 and 77%. Likewise, collicular neurons in VC coculture exhibited a significantly higher frequency of spontaneous spike discharge (Fig. 1E). To elucidate the basis for the increased activity level of SC neurons in VC coculture we first analyzed their spontaneous excitatory synaptic activity. AMPA receptor-(AMPAR-)-mediated spontaneous excitatory postsynaptic currents (sEPSCs) were distinguished from the bicuculline-sensitive GABAAR-mediated spontaneous inhibitory postsynaptic currents (sIPSCs) by their sensitivity to DNQX and their time course of decay (also see Fig. 2A,B). The average time constant of decay (τdecay) of sEPSCs was 2.91 ± 0.20 ms (n = 24), while τdecay of sIPSCs always exceeded 10 ms. It was found that the presence of VC explants resulted in a higher average amplitude and charge of sEPSC (Fig. 1F,G). The respective values were 26.5 ± 2.0 pA and 71.3 ± 6 fC in cultures with no explant, and 50.0 ± 7.4 pA and 113.1 ± 13.7 fC in the cortico-tectal cocultures (P < 0.01, both parameters). Interestingly, this upregulation of sEPSC amplitude or charge was specific for the VC explant because preparations carrying SC explants were similar to the controls without any explant (Fig. 1F,G; 28.7 ± 8.9 pA and 75.6 ± 7.5 fC, n.s.).
Next we analyzed mPSCs activity (Fig. 2A). As usual, AP-induced transmitter release was for this purpose blocked with tetrodotoxin (TTX, 1 μM). mPSCs were classed according to their decay kinetics (Fig. 2B, note logarithmic scale). It was found that the presence of VC explants neither affected the amplitude nor the frequency of mEPSCs (Fig. 2C,D). Together with the results of Figure 1(E,F) our data indicate that the differences in the spontaneous synaptic activity are likely to reflect differences in the level of AP-dependent glutamatergic synaptic activity.
GABAergic Synaptic Connections between SC Neurons are Suppressed Presynaptically
Once it became clear that the presence of a VC explant modifies the glutamatergic input to collicular neurons we asked whether this had any consequences for the GABAergic tecto-tectal input. The mIPSC amplitude distributions suggest that the presence of VC produced no changes on the postsynaptic side of GABAergic synapses (Fig. 2E). But there were marked differences in the mIPSC frequency (Fig. 2F). The average values of mIPSC frequency were 0.25 ± 0.5 Hz in cultures without explants and 0.08 ± 0.02 Hz in cultures with VC explants.
That GABAergic synaptic terminals are a target of modulation by cortical afferents is further supported by records of unitary eIPSCs (Fig. 3A). Nonfluorescent axons were stimulated in cultures without and with explants of VC or SC origin. Again, GABAergic “intrinsic” connections could be recognized 1) by the lack of EGFP-fluorescence and 2) by decay time constants >10 ms. The presence of VC explants resulted in a reduced amplitude of eIPSCs (Fig. 3B; with no explants: 173.3 ± 21.2 pA, with VC explants: 96.8 ± 14.9 pA, P < 0.01). The presence of SC explants failed to attenuate GABAergic eIPSCs (197.9 ± 47.1 pA, n.s. if tested versus control, P < 0.05 vs. VC explant).
A reduction of both mIPSC frequency and eIPSC amplitude could reflect 1) a reduced average probability of GABA release (prel) (Kirmse and Kirischuk 2006), with or without 2) a reduction in the readily releasable pool (Kirischuk et al. 2005), and/or 3) a decrease in the number n of synaptic terminals per connection (Singh et al. 2006). In GABAergic synapses between SC neurons a value of (prel) is typically associated with an elevated paired-pulse ratio (PPR) (Henneberger, Kirischuk, et al. 2005; Kirischuk et al. 2005). Indeed, if cultures were grown in the presence of a VC explant the PPR was significantly higher (Fig. 3C,D; with no explant: 0.39 ± 0.06; with VC explant: 0.84 ± 0.09; P < 0.001). Because the presence of SC explants failed to reduce the eIPSC amplitudes, we also expected that it had no effect on the PPR. This was the case (with SC explant: 0.46 ± 0.06; P = 0.39, comparison with no explant cultures; P < 0.01, comparison with VC explant culture). Thus, the proposal of explant-induced suppression of GABA release is supported by the reduction of mIPSC frequency, the decrease of eIPSC amplitudes, and the increase of PPR, and in all 3 cases the changes were due to the presence VC but not SC explants.
However, these findings do not completely rule out a contribution of reduced numbers of GABAergic synapses and/or release sites to the suppression of inhibition by VC explants. We addressed this by labeling GABAergic terminals with an antibody against the vesicular amino acid transporter (VIAAT) and calculating the coefficient of variation (CV) of eIPSCs (supplementary Fig. 2). The counts of VIAAT-positive synaptic terminals revealed no evidence for a decrease in the number of GABAergic terminals per postsynaptic cell (supplementary Fig. 2C) or the number of vesicles per terminal. Moreover, the intensity of the VIAAT signal of individual boutons was significantly stronger in preparations with VC (supplementary Fig. 2D). The CV of eIPSC amplitudes was not affected by the presence of VC explants (no explant: 0.35 ± 0.04; with VC explant: 0.34 ± 0.04; n.s.). As illustrated in the supplementary Figure 2(F), under condition of reduced (prel) an unchanged CV can only be obtained if the number of release sites/vesicles (n) increases (supplementary Fig. 2F). Thus, both the preliminary results of the counts and the theoretical relationship between CV, n, and (prel) hint at the possibility that the VC input may promote a compensatory increase in the overall GABAergic terminal/vesicle number rather than a decrease of n.
At this stage, it was still unclear whether the suppression of GABA release actually required the contact between cortical axons and collicular neurons. To address this question, we collected the medium conditioned by the presence of VC explants for 9 days and applied it to collicular cultures without explants. The paired pulse behavior of 10 tectal GABAergic connections was assessed on DIV16 and it was found to be not different from that of cultures in control medium (PPR: 0.33 ± 0.06, P = 0.55; amplitude: 152.2 ± 38.2, P = 0.62). We therefore conclude that a wide-range diffusible factor in the medium is, most likely, not responsible for the observed suppression of inhibition. The results are consistent with the hypothesis that cortical explants modulate GABAergic synaptic transmission via activity-dependent action of cortico-tectal connections.
The Depression of GABA Release Depends on VC-Induced Spontaneous Glutamatergic Synaptic Activity
If enhanced spontaneous glutamatergic synaptic activity was the cause of depressed GABA release, there should be a positive correlation between the average PPR of GABAergic eIPSCs and the average sEPSC charge provided that this charge can exert a generalized effect on all the GABAergic synaptic terminals of the shared postsynaptic neuron. The results of Figure 3(E) suggest that this might be true. The variability of the PPR in connections with the same postsynaptic cell was always smaller than the variability of the entire data set, and the range of PPR values per neuron did not increase with the average value of PPR. In any case, there was a clear positive correlation between the average PPR of GABAergic eIPSCs and the average sEPSC charge (Fig. 3F).
To further verify the hypothesis that tecto-tectal inhibition is modulated in dependence on the activity of the VC input it remains to be shown that in cultures with VC explants the release probability at GABAergic connections is indeed affected by alterations of spontaneous glutamatergic activity. Blocking AP generation in experiments with TTX reduced the average sEPSC charge by about 35% (Fig. 3G). Lowering the rate of spontaneous spike discharge in tectal neurons by recording in the presence of DNQX and APV (Fig. 1C,D) in cocultures with VC explants resulted in PPR levels characteristic of preparations without explants (Fig. 3H). It is therefore very likely that the larger AP-induced sEPSCs are required to initiate the suppression of GABA release in cortico-tectal cocultures.
Taken together, these findings can be interpreted as evidence for an altered probability of GABA release at tecto-tectal inhibitory synapses under the influence of spontaneously generated large glutamatergic EPSCs of cortical origin.
Postsynaptic Group I mGluRs and CB1Rs Link Glutamatergic and GABAergic Inputs of Collicular Neurons
Glutamatergic synaptic input and GABAergic release sites may be linked to each other through a variety of mechanisms. Among other possibilities, one might consider a signaling pathway via group I metabotropic glutamate receptors (group I mGluRs) (Morishita et al. 1998; Chevaleyre and Castillo 2003; Galante and Diana 2004). In cultured neurons from the neonatal rat SC group I mGluR expression has been demonstrated by PCR and immunocytochemistry (Meier et al. 2002). If group I mGluRs were indeed involved in the link between cortico-tectal glutamate release and tecto-tectal GABA release via the shared postsynaptic collicular neurons, activation of these receptors in control cultures without explant should mimic the effect of a VC explant.
Under the assumption that the measured PPR of GABAergic eIPSCs solely depends on the release probability, that is, if it is independent on the number of release sites and the quantal size, PPR will reflect changes in the release better than eIPSC amplitudes because PPR variability is not further increased by the substantial variability in the number of GABA release sites in this preparation (Henneberger, Kirischuk, et al. 2005). The following pharmacological analysis of the suppression of inhibition is therefore based on PPR data.
The specimen records of Figure 4(A) illustrate that in cultures without explant acute application of DHPG increased the PPR to the levels observed in cortico-tectal cocultures (control: 0.50 ± 0.13, n = 7; after 3 min DHPG: 0.85 ± 0.21; n = 7, P < 0.01). However, in contrast to observations with higher concentrations of DHPG applied to hippocampal slices (Edwards et al. 2006), the suppression of GABA release was not persistent, as the PPR returned to control levels within 3 min after washout start (Fig. 4B).
When postsynaptic G-protein function was suppressed by including S, guanosine-5′-O-2-(thiodiphosphate-βS) (GDP-βS) (2 mM) in the patch pipette, DHPG failed to modulate the PPR (Fig. 4C; control: 0.43 ± 0.06; DHPG + GDP-βS: 0.37 ± 0.08; n = 14, n.s.). This experiment is important because it shows that mGluR activation and GABA release are linked through the postsynaptic cell.
Because group I mGluRs have access to the phospholipase C/diacyl glycerol/inositol triphosphate signaling pathway we also considered a contribution of Ca2+ signaling in the postsynaptic neuron (for review see Schoepp et al. 1999). Although the standard intracellular solution contains 10 mM EGTA, it was not excluded that a fast Ca2+ signal escaped buffering. However, substituting 1,2-bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA) (20 mM) for EGTA did not prevent the DHPG-induced increase of the PPR (Fig. 4C, control 0.46 ± 0.14, DHPG + BAPTA: 0.86 ± 0.27, n = 5, P < 0.05), indicating that postsynaptic Ca2+ elevation was not required. It should be noted that the basal PPR values (in the absence of DHPG) were not affected either by using BAPTA as the intracellular Ca2+ buffer. In addition, lowering the intracellular EGTA concentrations to 0.1 mM and omitting Ca2+ did not affect the PPR values either (0.40 ± 0.10, n = 5, n.s.). We therefore conclude that postsynaptic Ca2+ signals neither affected the control values of PPR nor their changes under the influence of group I mGluR activation.
If group I mGluRs indeed initiated a signaling cascade responsible for the suppression of GABA release, addition of a group I mGluR antagonist should block the VC explant-induced suppression of inhibition. To verify this idea, we used MCPG. In general, MCPG is not an mGluR1/5-specific antagonist because it also blocks mGluR2 (Schoepp et al. 1999). However, collicular neurons were shown to lack mGluR2 (Meier et al. 2002). Therefore, for the sake of simplicity, we shall refer here to MCPG as a group I mGluR antagonist. In chronic experiments, exposure to VC explants was combined with the continuous presence of MCPG (200 μM). The presence of MCPG was also maintained throughout the recording session, and it was found that the PPR values went down to the level characteristic for preparations without explant (Fig. 4C, 0.42 ± 0.08, n = 13, P < 0.01, when compared with the untreated VC explant coculture). If MCPG was washed out from these chronically treated VC explant cocultures the PPR observed 5 min after washout start returned to high values, even overshooting the level observed in VC explant cocultures (Fig. 4C; in chronic MCPG: 0.42 ± 0.06; after washout: 1.59 ± 0.47; n = 6, P < 0.05).
Acute application of MCPG to VC cocultures reduced the PPR values to a level only slightly larger than those of cultures without explant (Fig. 4C; acute MCPG: 0.50 ± 0.08, n = 11, P = 0.25). The recovery from suppression of GABA release took no more than 3 min (not illustrated), which indicates that a continuous activity of group I mGluRs is necessary to maintain the VC-induced suppression. The overshoot seen after prolonged block of group I mCPGs suggests that the efficacy of the mechanisms underlying the suppression of inhibition is dynamically adjusted to the level of glutamatergic synaptic input.
An mGluR-dependent mechanism able to induce depression of GABA release is the release of endocannabinoids (eCBs) (Morishita et al. 1998; Varma et al. 2001; Chevaleyre and Castillo 2003; Galante and Diana 2004). Because CB1 receptors are present in the SC (Tsou, Brown, et al. 1998) and CB1, but no CB2 receptor messenger RNA (mRNA) was found in our collicular preparations without VC explants (supplementary Fig. 3), eCB release was regarded as a potential component of the link between the cortico-tectal glutamatergic and tecto-tectal GABAergic inputs. Indeed, pharmacological activation of CB1Rs by the unspecific CB agonist WIN proved to be very effective in reducing the frequency of sIPSCs in cultures without explant (Fig. 4D; control: 3.68 ± 2.17 Hz; WIN: 2.17 ± 2.04; n = 4; P < 0.05).
If the modulation of the PPR is both group I mGluR and CB1R dependent, the DHPG-induced increase of the PPR in cultures without explant should be prevented by the presence of the CB1R blocker AM251 (0.5 μM). Likewise, in cultures with VC explants, CB1 receptor block should prevent the increase in PPR. Both predictions were verified experimentally (Fig. 4E, no explant, DHPG: 0.85 ± 0.21; n = 7, DHPG + AM251: 0.25 ± 0.08, n = 5, P < 0.01; culture with VC explant, control: 0.84 ± 0.09; AM251: 0.47 ± 0.07, n = 7, P < 0.01).
The emerging hypothesis is that the spontaneous glutamatergic synaptic activity generated by cortico-tectal afferents depresses the GABA release in convergent tecto-tectal connections via mGluR signaling, eCB release, and presynaptic CB1R activation on the GABAergic synaptic terminals.
High-Frequency Activation of Single Glutamatergic Cortico-tectal Afferents Results in Persistent Suppression of GABA Release
The experiments with DHPG and acute mGluR block further support the idea that the depression of GABA release is the result of a continuous modulation rather than the result of structural reorganization of GABAergic synaptic connections. If so, and if the continuous modulation is submaximal, one should expect that acute high-frequency (HF) activation of the glutamatergic input will further enhance the depression at GABAergic terminals on the common postsynaptic neuron.
To examine this possibility, a glutamatergic cortico-tectal axon (EGFP positive, fast PSC) and a GABAergic tecto-tectal axon (EGFP negative, slow PSC) were stimulated as illustrated in Figure 5(A). Stimulation of cortico-tectal axons resulted in monosynaptic eEPSCs (Fig. 5B upper panel), polysynaptic eEPSCs, or polysynaptic mixed events in which the first event following the stimulation had fast decay kinetics. Only in one of more than 100 trials, cortico-tectal stimulation resulted in PSCs with a slow decay kinetic. This indicates that the vast majority of cortico-tectal connections were glutamatergic. Application of the AMPAR-blocker DNQX (20 μM) abolished all mono/polysynaptic responses, even if they were composed of both EPSCs and IPSCs suggesting that the stimulated glutamatergic cortico-tectal axons had divergent connections, which could result in polysynaptic effects. The percentage of responses with pure monosynaptic eEPSCs amounted to 30%, and only these cortico-tectal connections were used in the following experiments. A recording configuration suitable to place all 3 electrodes in one view field and resulting in stable monosynaptic eEPSCs and eIPSCs was found in about 1 of 10 SC cells tested.
The cortico-tectal eEPSCs were characterized by a short decay (τdecay: 3.0 ± 0.4 ms, n = 8) and a high interconnection variability, the average amplitudes ranging from 11.3 to 979.9 pA (on average 121.0 ± 57.0 pA, n = 17). After starting the PSC records in the whole-cell configuration, control paired stimulations were performed for 4 min. Then, the cortico-tectal axon was stimulated using an HF stimulation protocol (10 stimulus trains at 5 Hz, each train consisting of 4 stimuli at 100 Hz) while the investigated neuron was held in the current clamp mode. It should be noted, that in the presently used recording configuration this protocol did not, as a rule, induce a reproducible change in eEPSCs amplitudes (eEPSC amplitude ratio before/after: 0.99 ± 0.18, n = 8, one sample t-test P = 0.96).
To evaluate the possible adjustment of the tecto-tectal GABAergic synaptic transmission, we calculated the ratios of eIPSC amplitudes and PPRs at the end of the test (after HF stimulation) and the control period (before HF stimulation). They will be referred to as RAMP, RPPR, respectively. It was found that the amplitudes of eIPSCs of tecto-tectal GABAergic connections were significantly suppressed. Accordingly, the respective PPR increased (Fig. 5C; RAMP 0.63 ± 0.07, n = 11, P < 0.001; RPPR 3.07 ± 0.74, n = 9, P < 0.05).
To verify that this result was specific for the stimulation of cortico-tectal connections, the HF stimulation was also delivered to tecto-tectal GABAergic connections, and the eIPSC and PPRs were evaluated. In this case eIPSC amplitudes and PPR values were unchanged (Fig. 5D; RAMP 0.95 ± 0.03, n = 7, n.s.; RPPR 1.09 ± 0.23, n = 4, n.s.). It should be recalled that, due to the high intracellular Cl− concentration in the recording pipette, GABA would be depolarizing in the unclamped parts of the tested postsynaptic neuron, allowing for significant Ca2+ influx. Nonetheless, HF activation of GABAergic tectal afferents obviously failed to activate the mechanisms responsible for the depression of GABA release. We conclude that the suppression of GABA release was, again, a specific consequence of cortico-tectal HF activation.
We have already shown that the raised PPR observed in VC explant cultures required the activity of group I mGluRs and CB1Rs. If the acute suppression of single tecto-tectal eIPSCs and the increase of PPR upon stimulation of a unitary cortico-tectal glutamatergic connection were due to the same mechanism, the former should also be sensitive to group I mGluR and CB1R blockade. As shown in Figure 5(E), delivering a HF stimulus to a cortico-tectal input did not depress tectal GABAergic connections if the recording was performed in the presence of MCPG (Fig. 5E; RAMP 1.08 ± 0.26, n = 6, n.s.; RPPR 0.85 ± 0.20, n = 6, n.s.). On the contrary, in 2 out of 6 cases we even observed an increase of the eIPSCs immediately after HF stimulation. The RAMP for the first minute after HF stimulation ranged from 0.73 to 2.89, with an average of 1.39 ± 0.33 (n = 6). However, 10 min after HF stimulation eIPSCs amplitudes returned to the control levels (Fig. 5E, bar graph).
Finally, we tested whether the heterosynaptic depression of inhibition was mediated by CB1Rs and found that in the presence of AM251 HF stimulation of cortico-tectal axons failed to induce a change in eIPSC amplitudes and PPR values 10 min after the stimulation (Fig. 5E, bar graph, RPPR 1.20 ± 0.31, n = 8, n.s.).
These results indicate that the HF activation of a single cortico-tectal glutamatergic afferent can produce a persistent depression of tecto-tectal GABAergic synaptic transmission, associated with an increase in PPR. This heterosynaptic depression of inhibition is mediated by group I mGluRs as well as CB1 receptors.
The properties and actions of unitary connections of pyramidal tract neurons with subcortical sensory targets have not yet been subject to rigorous tests, mainly because in slice preparations these connections are not preserved (but see Tsvetkov et al. 2002). Yet efferent cortical control is very important, especially during the period for synapse refinement in the postnatal period. Previous developmental studies suggested that the onset of cortico-tectal signal transmission (in rat cortico-tectal connections are traced from postnatal days 4–5 on (Thong and Dreher 1986) may change the properties of synaptic inhibition in the superficial SC (Shi et al. 1997; Henneberger, Jüttner, et al. 2005) but direct evidence for an interaction between cortico-tectal glutamatergic and intrinsic GABAergic synapses was missing.
Our experiments showed that the presence of VC but not SC explants significantly enhanced the spontaneous glutamatergic synaptic activity in tectal neurons. However, this was not accompanied by a homeostatic upregulation of GABAergic inhibition. On the contrary, the presence of VC induced a continuous, but reversible suppression of GABA release, which could be mimicked by addition of the group I mGluR agonist DHPG. HF activation of unitary cortico-tectal connections resulted in a more pronounced and, in addition, persistent depression of GABAergic tecto-tectal transmission. The site of modulation was presynaptic, and both the nonstimulated VC tissue and the HF-stimulated single VC axons exerted their action via mGluRs and CB1Rs. We hypothesize that impulse activity in cortical axons activates eCB release in subcortical targets. Acting as retrograde messengers in tecto-tectal synapses eCBs induce local and graded disinhibition.
Cortico-tectal Synaptic Transmission In Vitro
Information about the properties of cortico-tectal synapses is scarce, in general. Recently it has become clear, that normal development of cortico-tectal connections relies on the activity of the layer V-specific transcription factors Otx1 (Weimann et al. 1999) and Fezl (Chen et al. 2005; Molyneaux et al. 2005). Earlier coculture experiments showed that, irrespective of the VC slice position in relation to the collicular target, cortico-tectal connections are preferentially made by layer V pyramidal neurons (Yamamoto et al. 1992; Novak and Bolz 1993; Cardoso de Oliveira and Hoffmann 1995). A pioneer study by Huettner and Baughman (1988) identified glutamate as the transmitter of retrogradely labeled layer V cortico-tectal pyramidal neurons. We can therefore assume that the EGFP-labeled axons in contact with unlabeled tectal neurons were layer V pyramidal neurons. Indeed, nearly all of them turned out to produce glutamatergic eEPSCs.
Activation of single cortico-tectal axons typically produced polysynaptic and, in part, mixed PSCs suggesting that each axon contacted more than one collicular neuron but each individual cortico-tectal connection was rather strong. Dividing the average maximal amplitude of cortico-tectal eEPSCs (239.4 ± 62.6 pA, n = 17) by the average amplitude of mEPSCs (24.1 ± 4.1 pA) rendered a maximal quantal content of 9.9. This means that each cortical axon can simultaneously release up to 10 vesicles, that is, it possessed at least 10 active zones. A wide-spread termination area of cortical axons and multiple innervation of individual target neurons has also been found in previous tracer studies (Rhoades et al. 1991) and in organotypical slice cocultures including the rodent SC (Cardoso de Oliveira and Hoffmann 1995).
As the cortico-tectal connection is not intact in slice preparations, the modulatory function of cortical afferents cannot properly be tested in a natural cellular environment, and the specific role of eCB signaling for the development of synaptic connections in the SC has remained unclear. Experiments on postnatal hippocampal slices emphasized the role of CB1Rs in the development of network activity (Bernard et al. 2005). The constituents of eCB signaling were already found in the rodent SC. Group I mGluRs are present in situ (Cirone et al. 2002) and in cultured tectal neurons (Meier et al. 2002). CB1R staining was demonstrated in situ (Tsou, Brown, et al. 1998), and CB1R mRNA was found in tectal cultures (supplementary Fig. 2). The hydrolysis of eCBs relies on fatty acid amide hydrolase (FAAH) and monoglyceride lipase (see Freund et al. 2003 for review). In the rodent SC, FAAH appears to be moderately expressed at the protein level (Tsou, Nogueron, et al. 1998).
The Cellular Mechanism of Cortical Control of GABAergic Inhibition in Tectal Neurons
The cellular basis of eCB actions in the brain is already well known (see Freund et al. 2003 for detailed review and Diana and Bregestovski 2005; Ohno-Shosaku et al. 2005; Edwards et al. 2006) for an update). It is generally assumed that the 2 predominant eCBs, anandamine and 2-arachidonoyl glycerol, are synthesized in the postsynaptic cell but act on the terminals of inhibitory or excitatory interneurons. In the CA1 region of the hippocampus and in other parts of the brain eCBs contribute to at least 3 basic forms of suppression of inhibition which differ with regard to induction and duration. (A) The depolarization-induced suppression of inhibition is reversible and requires postsynaptic Ca2+-elevation (Llano et al. 1991; Pitler and Alger 1992; Lenz and Alger 1999; Wilson and Nicoll 2001), in addition to CB1R activation (Kreitzer and Regehr 2001). (B) The suppression of inhibition mediated by the Gq/11-coupled mAChRs (Kim et al. 2002) and mGluRs (Maejima et al. 2001; Galante and Diana 2004) is also reversible but relatively insensitive to Ca2+. The ability of mAChR or mGluR agonists to induce suppression of inhibition is lost in PLCβ1-deficient mice (Hashimotodani et al. 2005). (C) A persistent Ca2+-insensitive form of depression has been termed I-LTD (Chevaleyre and Castillo 2003, 2004). It is initiated by mGluR, not mAChR activity and relies on the PLC-DAG lipase pathway but a cofactor might be required for the transition from the reversible to the persistent state of depression.
The suppression of synaptic GABA release presently observed in spontaneously active, nonstimulated cocultures, or in DHPG-treated tectal cultures without VC is similar to form (B) because it was readily reversed by acute application of an mGluR antagonist and, in case of DHPG-induced suppression, it was prevented by block of postsynaptic G-protein activity, but not BAPTA. The mGluR/CB1R-dependent suppression of inhibition after HF stimulation resembles I-LTD (C), as it persisted for at least 10 min after the end of stimulation. Considering that in this type of continuously perfused culture preparation, synaptically released glutamate (Taschenberger and Grantyn 1995), or exogenous glutamate agonists at mGluRs (present experiments) are rapidly washed-out, the mechanism responsible for the persistent character of suppression must reside downstream to the G-protein–coupled receptor.
Considering that tectal neurons challenged with an VC input had several days to compensate for increased glutamatergic input and decreased GABAergic inhibition in this spontaneously active preparation, it seems remarkable that there was obviously little change in the number of active postsynaptic receptors or the number of active zones. The former conclusion is based on the unchanged amplitudes of mPSCs, the latter on the similarity of changes in the eIPSC amplitudes and PPR. Thus, in contrast with the marked pre- or postsynaptic upregulation after continuous GABAAR block (for instance, Meier et al. 2003) or continuous treatment with the mGluR blocker MCPG (Fig. 4B), the basic characteristics of GABAergic synaptic connections remained the same. One has to conclude that the modulation exerted by the retrograde eCB release is mostly a short-living and local process. Thus, by graded disinhibition the activity level of the cortical targets is continuously adjusted, which represents an indirect, yet effective mechanism of gain control in a sensorimotor pathway.
Possible Function of Cortical Suppression of Tectal Inhibition
To fully understand the function(s) of the presently described mGluR/CB1R-dependent suppression of inhibition in the SC it will be necessary to further elucidate the role of GABAergic inhibition in the development of the visual system. It should also be recalled that, at least in rodents, the formation of the cortico-tectal pathway lags the formation of the retino-tectal pathway as well as the construction of an intrinsic GABAergic network. Tracer studies in rat revealed that the VC axons initially bypass their subcortical targets heading toward the spinal cord (O'Leary and Terashima 1988). Only after a considerable waiting period do collaterals grow out toward the SC from the main axon by “interstitial axon budding” (O'Leary et al. 1990). In the mouse SC, the temporal gap between the first electrophysiologically detectable GABAergic synapses (see Grantyn et al. 2004; Kirischuk et al. 2005) and the arrival of VC axon branches in the superficial layers (Inoue et al. 1992) could be 1 week. Based on results from anolphthalmic mice it has been argued that the refinement of cortico-tectal projections is aided by the preexisting retinal projection (Khachab and Bruce 1999). That the latter may also be modified by means of a local graded disinhibition of SC neurons receiving matched retino-tectal and cortico-tectal input is another new and attractive hypothesis.
To conclude, the presently described mGluR and eCB-dependent mechanism of graded efferent cortical control by means of target disinhibition appears to be well suited to validate a preexisting map of visual activity or to initiate its modification during postnatal development.
Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.
We thank Alexej Grantyn (Paris), Sergei Kirischuk (Berlin), and Dimitri Kullmann (London) for many helpful suggestions. We are particularly grateful to Bruce Walmsley who generously provided his lab facilities for the initial experiments of this study. The technical assistance of D. Betances, K. Rückwardt, and U. Neumann is highly appreciated. This work was supported by the German Research Council (Gr986/9-1 to RG), the European Regional Development Fund, and the Welcome Trust Foundation (grant to S.J.R.). Conflict of Interest: None declared.