Spontaneous Synaptic Activity in an Organotypic Culture of the Mouse Retina

Abstract

purpose. Many strains of mutant mice die at birth, when the retina is still very immature. The retinas of such mice can be studied in organotypic cultures. After a preceding anatomic study of the synaptic development, the electrical activity of the synaptic circuits within such cultures was studied in wild-type and gephyrin-deficient mice.

methods. Organotypic cultures of newborn mouse retinas were grown for 14 days in vitro. Spontaneous postsynaptic currents (sPSCs) of amacrine cells were measured by using the whole-cell configuration of the patch-clamp technique. GABAergic and glycinergic currents that were isolated with specific antagonists, and retinas from wild-type (geph ^+/+) and gephyrin-deficient (geph ^−/−) mice were compared.

results. Rapidly decaying sPSCs that were blocked by kynurenic acid were mediated by ionotropic glutamate receptors, whereas sPSCs with significantly higher peak amplitudes and slow-decay kinetics were identified as spontaneous inhibitory postsynaptic currents (sIPSCs) mediated by γ-aminobutyric acid type A receptors (GABA_ARs) and glycine receptors (GlyRs). In gephyrin-deficient (geph ^−/−) cultures, we found no sIPSCs mediated by GlyRs. sIPSCs mediated by GABA_ARs expressed in amacrine cells of geph ^−/− retinas decayed significantly faster than GABAergic sIPSCs recorded in amacrine cells of geph ^+/+ retinas.

conclusions. The different decay kinetics of GABA_ARs expressed in amacrine cells of geph ^+/+ and of geph ^−/− retinas suggests that these cells express at least two types of GABA_AR subtypes. In amacrine cells of geph ^−/− mice, a specific GABA_AR subtype that may contain the α2 subunit, is impaired by the absence of gephyrin, whereas other GABA_ARs appear to function normally.

The mouse retina is becoming an important object for study of mammalian retinal organization, because of the ever-growing availability of genetically manipulated mice. However, many mutants die at birth (postnatal day [P]0), at a time in development when the retina is still very immature. ^{1 2} Organotypic tissue cultures can rescue the retinas of such mutants and offer the possibility to study the maturation of the mouse retina and impairments by the genetic manipulations in vitro. Both in retinal slice cultures ^{3 4} and organotypic retinal cultures, the formation of nuclear and synaptic layers have been observed. ^{4 5 6 7} Immunocytochemical markers have also demonstrated the differentiation of the major retinal cell types in such cultures; however, the maturation of some cells appears to require the addition of neurotrophic factors. ⁸  

The postnatal development of the retinal circuitry is critically dependent on neuronal activity. ⁹ Initially in development, even in the absence of functional photoreceptors, waves of spontaneous electrical activity spread between retinal neurons. ^{10 11} In addition, ganglion cells and amacrine cells undergo synchronized oscillations in the intracellular Ca²⁺ concentration, which spread in a wavelike fashion tangentially across the retina. ¹² The segregation of the ganglion cell dendrites in the inner plexiform layer (IPL) into an OFF- and ON-sublamina depends on the electrical activity. ¹³ Recordings from ganglion cells have shown that there is correlated, spontaneous firing among ON-ganglion cells and among OFF-ganglion cells. ¹⁴ During the first postnatal month, the correlated spontaneous bursting activity between amacrine cells and ganglion cells is modulated by γ-aminobutyric acid (GABA). ¹⁵ The expression of GABA_A receptors (GABA_ARs) precedes the formation of functional synapses and then changes along with cellular differentiation of the mammalian retina. ¹⁶ Immunocytochemical studies suggested the presence of a functional GABAergic network in the inner plexiform layer by birth, coinciding with the appearance of correlated bursting activity in the inner retina. ¹⁵ Thus, the establishment of functional GABAergic synapses appears to be crucial for development of orderly connections in the mammalian retina. 

Despite the wealth of anatomic evidence for the apparently normal maturation of the retina in organ cultures, hardly any recordings of the electrical activity within such cultures are available. Previous studies were focused on transmitter gated currents of the neurons in the cultures and only briefly reported synaptic currents. ³ In the present study we concentrated on the spontaneous synaptic currents in these cultures, which represent ongoing, intrinsic synaptic activity. We applied the patch-clamp technique to record spontaneous inhibitory postsynaptic currents (sIPSCs) mediated by GABA_ARs and glycine receptors (GlyRs). We concentrated on the inhibitory synaptic events, because we have studied those recently also in acutely isolated slices of the adult mouse retina. ¹⁷ We thus can directly compare sIPSCs recorded in cultures with those recorded in situ. An additional reason for the concentration on sIPSCs is the availability of organotypic cultures of retinas of gephyrin-deficient mice. It has recently been shown that gephyrin (geph) is a clustering molecule for both glycine and GABA receptors in postsynaptic densities of inhibitory synapses. ^{1 18} We prepared organotypic cultures from wild-type (geph ^+/+ ), heterozygous (geph ^+/− ), and homozygous (geph ^−/− ) gephyrin-deficient mice. By applying immunocytochemistry, we observed a dramatic impairment of the clustering of GABA_ARs and GlyRs. ⁵ In the present study we investigated whether this impairment results in functional changes of the synaptic transmission in the retina. 

Material and Methods

Animals

All experiments were performed according to the guideline for the welfare of experimental animals issued by the Federal Government of Germany (Tierschutzgesetz) and fulfill the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Two heterozygous geph ^+/− mice were mated and timed-pregnant mice at 20 days of gestation were deeply anesthetized with halothane and killed by cervical translocation to obtain the embryos. Genotyping was performed by polymerase chain reaction. Altogether, litters of five dams were used. 

Organotypic Culture

Preparation of the organotypic cultures adhered to protocols previously described. ^{3 5 19} Briefly, retinal explants were mounted photoreceptor-side-down onto sterile nitrocellulose filters. Filters with explants were attached to coverslips, inserted into screw-top culture tubes, and incubated in a roller incubator at 37°C without CO₂. Explants were harvested after 15 to 30 days in culture. 

Immunocytochemistry

Explants on the filters were fixed in 4% (wt/vol) paraformaldehyde (PFA) for 3 to 30 minutes, depending on the antibody used. The tissue was then cryoprotected and sectioned vertically at 12 μm. Immunocytochemical labeling was performed with the indirect fluorescence method. ²⁰ For light microscopy analysis, the sections were photographed with a photomicroscope (Axiophot; Zeiss, Jena, Germany), using the appropriate fluorescence filters. The confocal micrographs (Figs. 1B 1C) were taken with a laser scanning microscope (LSM Pascal; Zeiss). For GABA_AR immunocytochemistry (Fig. 1E) a polyclonal antibody ²¹ against the γ2 subunit (raised in guinea pigs, diluted 1:3000) was used (kind gift of Hanns Möhler, Institute of Pharmacology, Zürich, Switzerland). For GAD65 immunocytochemistry (Fig. 1B) , a mouse anti-GAD65 antibody was used (Roche Diagnostics, Mannheim, Germany; diluted 1:100). For the detection of bassoon (Fig. 1C) a rabbit anti bassoon antibody ²² was used (diluted 1:8000; kindly provided by Eckhard Gundelfinger, Leibniz Institute for Neurobiology, Magdeburg, Germany). 

Electrophysiological Recordings

Vertical slices (thickness ∼200–300 μm) of organotypic cultures were cut with a scalpel or a microslicer (DSK, Tokyo, Japan), stored at room temperature in extracellular saline and bubbled with 95%O₂ and 5% CO₂. Electrophysiological experiments were started 15 minutes after preparation of the slices. Neurons were visualized by differential interference contrast microscopy using a 40× water immersion objective and a digital camera (PCO Computer Optics GmbH, Kelheim, Germany) mounted on an upright microscope (Zeiss, Oberkochen, Germany). Patch pipettes were pulled from borosilicate glass tubing (2.0 mm outer diameter, 0.5 mm wall thickness; Hilgenberg, Malsfeld, Germany). When filled with internal solution, they had a resistance of 6 to 8 MΩ. Cells were approached under visual control by maintaining a moderate positive pressure in the patch pipette. Membrane currents were recorded in the whole-cell configuration of the patch-clamp technique using an amplifier (EPC-7; List, Darmstadt, Germany). Patch pipette capacitance and cell capacitance were canceled and series resistance was compensated by approximately 80% using the internal compensation circuits of the amplifier. Recordings were obtained at a holding potential (V_h) of −60 mV. A recording was discarded, if the holding current exceeded ±25 pA. Currents were sampled at 10 kHz and filtered at 3 kHz before digitization by using the internal low-pass filter of the amplifier. Data were stored on-line (PClamp software; Axon Instruments, Foster City, CA.). All recordings were made at room temperature (20–to 24°C). Recordings were taken from visually identified amacrine cells in the inner nuclear layer (INL) and cells located in the ganglion cell layer of the retina. All cells were filled with Lucifer yellow during recording to confirm their identity, based on their cell body position in the INL and on morphologic criteria such as the dendritic ramification pattern (Fig. 2A) . In addition, electrical membrane properties were checked (Fig. 2B) . Cells of the ganglion cell layer were either ganglion cells surviving the transsection of the optic nerve, or displaced amacrine cells. They were not further distinguished here, but most of the cells located in the ganglion cell layer exhibited small voltage-dependent Na⁺ currents. Bipolar cells were also present in the inner INL, but they were easily distinguished by their characteristic morphology. 

Solutions and Chemicals

Slices were continuously superfused with a physiological extracellular saline that contained (in mM): NaCl 125, KCl 2.5, NaHCO₃ 26, NaH₂PO₄ 1.25, glucose 25, CaCl₂ 2, and MgCl₂ 1. To maintain the extracellular pH at 7.4 the saline was bubbled with 95% O₂ and 5% CO₂. The patch pipette solution contained (in mM): KCl 140, CaCl₂ 1, MgCl₂ 1, HEPES 10, EGTA 11, with pH adjusted to 7.2 with KOH. Stock solutions of strychnine (1 mM in distilled water) and bicuculline (10 mM in distilled water) were prepared shortly before the experiments and added to the extracellular saline in defined concentrations. To block spontaneous excitatory postsynaptic currents (sEPSCs) mediated by ionotropic glutamate receptors, 1 mM kynurenic acid was added to the extracellular saline in some experiments. 

Analysis of Postsynaptic Currents and Statistics

Spontaneous postsynaptic currents (sPSCs) were detected (threshold, 8–10 pA) by computer (MiniAnalysis software; Synaptosoft, Leonia, NJ). Peak amplitudes, rise times and decay time constants were estimated for single sPSCs and further analyzed on a personal computer (MiniAnalysis; Synaptosoft; and Origin; MicroCal, Northampton, MA). Mean amplitudes and frequencies of sPSCs were computed from all sIPSCs and all sEPSCs observed in an amacrine cell. Only events that did not show any signs of multiple peaks (i.e., contamination of rise or decay phases by subsequent events) were selected for analysis of the kinetics and for exponential fitting. The rise times of the sPSCs were determined by calculating the time in which the current increased from 10% to 90% of the peak amplitude denoted as T10/90. Decay kinetics of single sIPSCs were determined by least-square fits of the decay phase after the peak current using a monoexponential and a biexponential decay function. The number of exponentials necessary for a good fit of the data were determined by visual inspection. To compare events best fitted with a different number of exponentials, we calculated the amplitude weighted time constant, τ_w = (A ₁τ₁ + A ₂τ₂)/(A ₁ + A ₂), where A ₁ and τ₁ are the amplitude and the time constant of the fast component, and A ₂ and τ₂ are the amplitude and the time constant of the slow component of the biexponential fit, respectively. To calculate the decay kinetics of sEPSCs, at least 50 sEPSCs were randomly selected, averaged, and fitted using a biexponential decay function. Unless specified differently, data were denoted as statistically significant, if P < 0.05, using the two-tailed Student’s t-test. The correlation between two samples (i.e., between τ_w and T10/90), was determined by linear regression analysis. 

Results

Development of Neurons and Synapses in Organotypic Cultures of the Mouse Retina

The mouse retina is still very immature at birth (P0). Only ganglion cells appear to be differentiated, but the nuclear and plexiform layers are just segregating from the neuroblast layer, and synaptogenesis has not yet started. ⁵ However, after 15 days in vitro (15 DIV) nuclear and synaptic layers had differentiated. Figure 1A shows a vertical cryostat section through an organotypic culture (15 DIV), and, with Nomarski (differential interference contrast) optics, the retinal layers become apparent. They appeared to be normal. However, close inspection showed that only rudimentary photoreceptor inner and outer segments were above the outer limiting membrane (OLM) in Figure 1A (arrows). The inner plexiform layer (IPL) is sometimes split (arrowheads in Fig. 1A ), and the cells in the ganglion cell layer are relatively small and of uniform size. Because the optic nerve is cut when the retina is taken into culture, most ganglion cells appear to undergo retrograde degeneration and most cells surviving in the ganglion cell layer are likely displaced amacrine cells. ²³ However, bipolar and amacrine cell development is seemingly normal. ^{3 5} Because we recorded sIPSCs preferentially we also studied the inhibitory synaptic circuits of these organotypic cultures. The confocal micrographs of Figures 1B and 1C show the IPL of a section (15 DIV) that was double labeled for the GABA-synthesizing enzyme glutamic acid decarboxylase (GAD65, Fig. 1B ) and for the synapse-associated protein bassoon (Fig. 1C) . GAD65 is a marker of most GABAergic amacrine cells, ²⁴ and their processes formed a dense plexus in the IPL (Fig. 1B) . The bodies of amacrine cells in the INL and of putative displaced amacrine cells in the ganglion cell layer were also GAD65 immunoreactive. Bassoon was localized to the presynaptic terminals of conventional, inhibitory synapses of the rodent IPL. ²⁵ The dense punctate labeling of the IPL in Figure 1C suggests that there are numerous such synapses in the organotypic cultures. Both Figures 1B and 1C indicate that the presynaptic sites of inhibitory synapses in the IPL of the organotypic cultures were differentiated by 15 DIV. In a preceding study we immunolabeled such cultures for GABA and GlyRs. ⁵ In cultures of wild-type mice we found apparently normal development and synaptic clustering of GABA_ARs and GlyRs. However, in cultures of geph ^−/− mouse retinas, we observed a complete absence of GlyR synaptic clusters and also a substantial impairment of GABA_AR clusters. Figures 1D and 1E show a section through the IPL of a wild-type mouse organotypic culture (15 DIV) that was labeled for the GABA_AR subunit γ2 (GABA_Aγ2). The Nomarski micrograph in Figure 1D shows the IPL with amacrine cells in the INL and putative displaced amacrine cells in the ganglion cell layer. GABA_Aγ2 immunoreactivity has a punctate appearance throughout the IPL, and we have shown previously by electron microscopy that these puncta represent a clustering of the receptors in postsynaptic densities. ⁵ Taken together, these anatomic results suggest that numerous inhibitory synapses are present in the IPL of these organotypic cultures. 

Analysis of Postsynaptic Currents in Organotypic Cultures of the Mouse Retina

Next we investigated whether the synaptic structures that develop in this organotypic culture system are functional. Therefore, we used the whole-cell configuration of the patch-clamp technique to record sPSCs from amacrine cells and cells located in the ganglion cell layer of organotypic cultures obtained from wild-type geph ^+/+, geph ^+/−, and geph ^−/− mice. Altogether, the litters of five dams were used. From these, 86 explant cultures were prepared (22 from geph ^+/+, 34 from geph ^+/− and 30 from geph ^−/− mice). Each explant culture was cut into three to five slices, and approximately 300 slices were investigated. Successful recordings were collected from 132 slices. This is also the number of successfully recorded cells, because only one such recording was performed in a given slice. All cells were filled with Lucifer yellow, and 52 cells identified by the position of their cell bodies and their arborization pattern as amacrine cells or as cells found in the ganglion cell layer were analyzed in the present study (Fig. 2) . In wild-type cultures 20 cells were analyzed (12 amacrine cells; 8 cells found in the ganglion cell layer), in geph ^+/−cultures 14 cells (6/8), and in geph ^−/− cultures 18 cells (11/7). All 29 amacrine cells (Fig. 2A) showed a typical set of outwardly rectifying K⁺ currents (Fig. 2B) as previously described. ^{26 27} All 23 cells found in the ganglion cell layer (Fig. 2C) also showed such outwardly rectifying K⁺ currents. However, 18 of these cells also exhibited fast activating, spontaneously inactivating inward currents (Fig. 2D) that could be blocked by tetrodotoxin in eight of eight cells tested (not shown), indicating that they are most likely voltage-dependent Na⁺-currents. Such currents can be recorded in the in vivo retina from both displaced amacrine cells and ganglion cells. ²⁸ We did not further characterize these cells and refer to them in the remainder of the article as cells located in the ganglion cell layer. However, as mentioned earlier, most of the cells surviving the transsection of the optic nerve when the retinal cells were obtained for culture appeared to be displaced amacrine cells. 

Excitatory Postsynaptic Currents

In cultures of wild-type retinas, all 12 amacrine cells and all 8 cells located in the ganglion cell layer showed spontaneous synaptic activity occurring at a mean frequency of 1.1 ± 0.92 Hz (20 cells). A typical recording of sPSCs from a cell found in the ganglion cell layer is shown in Figure 3A , and the sPSCs of this cell are analyzed in detail in Figures 3B 3C 3D . Two types of sPSCs can be distinguished. One group (asterisks in Fig. 3A ) was characterized by small amplitudes and fast-decay kinetics. Another group was significantly larger and had slow-decay kinetics (Fig. 3A , arrows). We applied selective antagonists to separate spontaneous excitatory postsynaptic currents (sEPSCs) from sIPSCs. The fast sPSCs persisted in the presence of 0.5 to 1 μM strychnine and 10 μM bicuculline, but were completely absent in the presence of 1 mM kynurenic acid (data not shown), indicating that they were sEPSCs mediated by ionotropic glutamate receptors. The slow-decaying sPSCs were blocked by bicuculline or strychnine or a combination of both, indicating that they were sIPSCs mediated by GABA_ARs and GlyRs (described in detail later). Figure 3B shows a comparison of the amplitude distribution of sEPSCs and sIPSCs recorded from this cell. sEPSC amplitudes showed a narrow distribution ranging from 5 to 30 pA, with a mean peak amplitude of 15.8 ± 6.0 pA (151 events). The amplitude distribution of sIPSCs was strongly skewed with a mean of 20.2 ± 16.3 pA (±SD; 737 events). To compare the decay kinetics of sEPSCs and sIPSCs, selected single events were superimposed and averaged (Fig. 3C ; note the different scaling of the ordinate). The decay time constants were determined by fitting these events individually with a biexponential function. The distributions of the corresponding τ_ws and the cumulative fraction plots (Fig. 3D) indicate that sIPSCs and sEPSCs also differed significantly in their decay kinetics. Whereas sEPSCs decayed with a mean τ_w of 1.4 ± 0.6 ms (±SD; 60 events), sIPSCs decayed significantly slower, with a mean τ_w of 26.9 ± 18.2 ms (531 events). Similar sEPSC decay kinetics were observed in all neurons that showed both, sEPSCs and sIPSCs (data not shown). Thus, sEPSCs and sIPSCs, could be reliably distinguished by their decay kinetics, with those having a τ_w of 4 ms or more classified as sIPSCs. It was not necessary to superfuse the cultures with kynurenic acid to block sEPSCs. Such application of kynurenic acid blocks the excitatory synaptic drive of amacrine cells and causes a substantial reduction of the frequency of sIPSCs as an unwanted side effect. ¹⁷  

GABAergic sIPSCs in Organotypic Cultures of Retinas of geph^+/+ Mice

To characterize the sIPSCs in amacrine cells and cells located in the ganglion cell layer we used the selective antagonists bicuculline and strychnine to separate sIPSCs mediated by GABA_ARs and GlyRs, respectively. Figure 4A shows the amplitude–time plot of a typical recording from an amacrine cell. Each dot represents the occurrence and peak amplitude of a single sIPSC. The original recording is shown in Figure 4B . In the absence of the antagonists (control) sIPSCs occurred at a frequency of 0.2 Hz with a mean amplitude of 19.8 ± 12.7 pA (368 events). Bicuculline (10 μM) blocked the sIPSCs completely, but strychnine (0.5 and 1 μM) did not induce an inhibition of the sIPSCs, indicating that they were mediated by GABA_ARs (Figs. 4A 4B) . In contrast, in the presence of strychnine, the frequency of sIPSCs significantly increased to 1.0 Hz (P < 0.05), most likely because of a disinhibition of the surrounding network (Fig. 4A) . The amplitude distributions of sIPSCs recorded in the absence and presence of strychnine were skewed to lower values (Fig. 4C) . However, the amplitude distribution of sIPSCs during strychnine application was noticeably different from that of the control sIPSCs (Figs. 4C 4D) , and the mean peak amplitude during application of strychnine increased to 27.4 ± 14.1 pA (2360 events; P < 0.01). To find out whether the sIPSCs in the presence of strychnine represent a different subpopulation, sIPSCs that did not show any signs of multiple peaks were selected and individually fitted with two exponentials. The τ_ws were calculated (see Methods section), resulting in a mean τ_w of 9.8 ± 7.6 ms (217 events) for control sIPSCs and 7.9 ± 6.6 ms (1557 events) for those in the presence of strychnine. Superimposed averaged sIPSC traces from both groups are displayed in Figure 4E , and the cumulative fraction plot is shown in Figure 4F . There was a tendency for an increase in amplitude and a slight decrease in τ_w when sIPSCs were compared between control records and during application of strychnine. 

We analyzed the GABAergic sIPSCs of the 12 amacrine cells and of the 8 cells located in the ganglion cell layer in recordings from organotypic cultures of geph ^+/+ mice. The mean peak amplitude recorded in amacrine cells was 30.4 ± 17.8 pA (12 cells) and the mean τ_w was 28.0 ± 4.9 ms (12 cells). In cells of the ganglion cell layer we found a mean peak amplitude of 26.0 ± 13.9 pA (8 cells) and a mean τ_w of 24.0 ± 7.6 ms (8 cells). Neither the mean peak amplitudes nor the mean τ_ws of GABAergic sIPSCs recorded in amacrine cells and cells located in the ganglion cell layer were significantly different from each other (P > 0.05). 

Glycinergic sIPSCs in Organotypic Cultures

Synaptic sIPSCs sensitive to strychnine (i.e., mediated by GlyRs) were found in only 4 of 29 cells. Two cells were found in the ganglion cell layer of geph ^+/+ cultures and two further were recorded in the amacrine cell layer of geph ^+/− cultures. The sIPSCs sensitive to strychnine showed a mean peak amplitude of 47.1 ± 24.9 pA (4 cells) and a mean τ_w of 18.5 ± 4.9 ms (4 cells). Two cells expressed both GABAergic and glycinergic sIPSCs simultaneously and one of them, an amacrine cell is analyzed in detail in Figure 5 . The amplitude–time plot is shown in Figure 5A . Before the drug application, sIPSCs occurred at a frequency of 0.9 Hz with a mean amplitude of 74.2 ± 41.9 pA (592 events). Strychnine (0.5 μM) strongly reduced the mean amplitude, but also induced an increase in sIPSC frequency to 3.8 Hz. which was probably due to disinhibition of the local network. By the application of 10 μM bicuculline in addition to 0.5 mM strychnine the remaining sIPSCs nearly completely disappeared, indicating that they were mediated by GABA_ARs. The withdrawal of strychnine evoked the reappearance of a subpopulation of sIPSCs. These sIPSCs occurred at a frequency of 0.24 Hz and persisted in the presence of bicuculline, indicating they were mediated by GlyRs (Fig. 5A) . Thus, this amacrine cell expressed two types of sIPSCs, one mediated by GABA_ARs and another by GlyRs. Indeed, we found two types of sIPSCs that differed in their decay kinetics (Fig. 5B) . In the absence of the antagonists, fast- and slow-decaying sIPSCs were found, the first persisted in the presence of strychnine, whereas the latter remained in the presence of bicuculline, suggesting that the fast-decaying sIPSCs were mediated by GABA_ARs and the slow-decaying by GlyRs. For a detailed analysis, several sIPSCs traces were selected from each group, superimposed, and averaged (Fig. 5C) . Single sIPSCs were fitted with two exponentials, and the τ_ws were calculated (see the Methods section), resulting in a mean τ_w for GABAergic sIPSCs of 5.6 ± 3.9 ms (1019 events) and of 10.1 ± 10.8 ms (223 events) for glycinergic sIPSCs, which was significantly different (P < 0.01). The superimposed and normalized averaged traces and the cumulative fraction plot of the τ_ws (Fig. 5D) indicate that GABA_ARs expressed in this cell decayed significantly faster than GlyRs. In addition, we found that the GABAergic sIPSCs were characterized by a significantly larger amplitude (GABAergic sIPSCs: 24.6 ± 14.7 pA, 1019 events; glycinergic sIPSCs: 17.1 ± 10.4 pA, 223 events; P < 0.01). The amplitude distribution histograms and corresponding cumulative fraction plots are displayed in Figure 5E . Because the differences in the decay kinetics between GABAergic and glycinergic sIPSCs could have arisen from different locations of the two types of synapses in a different electrotonic distance from the recording electrode, we plotted the τ_ws of both groups of sIPSCs against the corresponding increase times (T10/90s; see the Methods section; Figs. 5F 5G ). We found no correlation between τ_ws and T10/90s, indicating that the differences in the decay kinetics of GABA_ARs and GlyRs is not due to dendritic filtering. The second cell that showed glycinergic and GABAergic sIPSCs simultaneously was found in the ganglion cell layer. In this cell glycinergic sIPSCs were characterized by a mean peak amplitude of 41.6 ± 20.4 pA (545 events) and a mean τ_w of 25.4 ± 17.5 ms (545 events) and they occurred at a frequency of 1.3 Hz. GABAergic sIPSCs showed a mean peak amplitude of 17.6 ± 9.6 pA (2880 events) and a mean τ_w of 22.7 ± 15.7 ms (2880 events) and they occurred at a frequency of 1.7 Hz. 

sIPSCs in Organotypic Cultures of Gephyrin-Deficient Mice

Gephyrin is essential for the clustering of GlyRs at postsynaptic densities. ²⁹ Recently, evidence was found that gephyrin is also essential for the clustering of many GABA_ARs. ^{5 18 20 30 31} We therefore studied sIPSCs in cultures from geph ^+/− and geph ^−/− mice. We observed ongoing, sIPSC activity in cultures from both geph ^+/− and geph ^−/− mice, indicating that gephyrin-deficient mice still express functional inhibitory synapses. The mean frequency of these sIPSCs was 1.0 ± 1.5 Hz (14 cells) in cultures from geph ^+/− mice and 0.6 ± 0.8 Hz (18 cells) in cultures from geph ^−/− mice. They were not significantly different from the mean sIPSC frequency observed in wild-type cultures (P > 0.05). However, a more detailed analysis revealed that sIPSCs mediated by GlyRs were absent from the 11 amacrine cells and from the 7 cells located in the ganglion cell layer in cultures of geph ^−/− mice. This is in line with the anatomic findings of a total absence of synaptic GlyR clusters in geph ^−/− mice ⁵ and supports the conclusion that gephyrin is essential for the development of functional glycinergic synapses in the retina. 

Bicuculline-sensitive, strychnine-resistant (i.e., GABAergic) sIPSCs were found in both geph ^+/− and geph ^−/− cultures. Typical recordings of such sIPSCs obtained from three amacrine cells of similar developmental stages (22–29 DIV) of the three genotypes are compared in Figure 6 . Whereas the peak amplitudes recorded in amacrine cells of geph ^+/+ and geph ^+/− cultures were similar, those recorded in geph ^−/− cultures were smaller (Fig. 6A) . The amplitude distributions were strongly skewed to lower values in all three genotypes (Fig. 6B) . However, although sIPSCs recorded from geph ^+/+ and geph ^+/− cultures showed a broad distribution of peak amplitudes, those of amacrine cells recorded from geph ^−/− cultures were all smaller than 40 pA (Figs. 6B 6C) . The mean peak amplitudes of sIPSCs recorded from all amacrine cells of geph ^+/− cultures were 37.5 ± 12.0 pA (6 cells), and those from all amacrine cells of geph ^−/− cultures were 19.5 ± 6.1 pA (11 cells). The mean τ_ws of sIPSCs recorded from all amacrine cells of geph ^+/− cultures were 25.8 ± 6.7 ms (6 cells) and those from all amacrine cells of geph ^−/− cultures were 23.4 ± 5.7 ms (11 cells). The results from geph ^+/+ cultures were listed earlier in the article. 

The statistical analysis of these data revealed three interesting findings. First, sIPSCs from amacrine cells of geph ^+/+ and geph ^+/− retinas were not significantly different in peak amplitude and τ_w (P > 0.05). This suggests that in geph ^+/− mice the absence of one gephyrin gene may be compensated for by the other allele. Second, the mean peak amplitude of sIPSCs of amacrine cells of geph ^−/− cultures was significantly smaller than those of geph ^+/+ and geph ^+/− (pooled data; 18 cells; P < 0.01), suggesting a diminished release of GABA and/or a loss of clustering of GABA_ARs expressed at some synapses of geph ^−/− mice. ^{5 31} Third, the mean τ_w determined from sIPSCs recorded from amacrine cells in geph ^−/− cultures was significantly smaller than the mean τ_w measured from geph ^+/+ cultures (P < 0.05). The different decay kinetics of GABA_AR-mediated sIPSCs of the different genotypes may have been caused by dendritic filtering. However, we found no significant correlation between the τ_ws or the peak amplitudes and the corresponding increase times (T10/90s). Sample diagrams plotting the relationship between τ_ws and T10/90s peak amplitudes and T10/90s are shown in Figure 7 . Thus, the different GABA_AR decay kinetics observed in the amacrine cells of geph ^−/− retinas may reflect an impaired clustering of a slower decaying GABA_AR subtype. 

We also analyzed the properties of sIPSCs recorded in cells found in the ganglion cell layer of gephyrin-deficient mice. In geph ^+/− cultures mean peak amplitudes and mean τ_ws were 25.9 ± 10.3 pA and 27.5 ± 7.0 ms (eight cells), respectively, and in geph ^−/− cultures 21.7 ± 7.7 pA and 26.3 ± 6.3 ms (seven cells), respectively. Peak amplitudes and τ_ws of cells found in the ganglion cell layer were not significantly different between the three genotypes. 

Effect of TTX

We recorded sIPSCs in the absence and presence of tetrodotoxin (TTX) to determine the contribution of GABA and glycine release triggered by action potentials (data not shown). In the presence of TTX (0.5 μM) the sIPSC mean frequency changed from 1.0 ± 1.0 Hz (control, without TTX) to 0.5 ± 0.6 Hz (eight cells; paired t-test, not significant, P > 0.05; pooled data from three geph ^+/+, two geph ^+/−, and three geph ^−/− cells). The mean peak amplitude of sIPSCs decreased significantly from 41.0 ± 26.5 pA (control) to 22.4 ± 8.6 pA (eight cells; paired t-test, P < 0.05; pooled data from three geph ^+/+, two geph ^+/−, and three geph ^−/− cells) in the presence of TTX, suggesting that a part of the GABA and glycine release is triggered by action potentials (data not shown). 

Discussion

Synaptic Development in Organotypic Cultures of the Mouse Retina

Only a few synapses are found in the retina of newborn rodents, and most synaptic development occurs during the first two postnatal weeks. ³² Several mutants with defects in synapse-associated proteins die at birth because of respiratory and/or feeding problems and the synaptic organization of the retina of such mutants cannot be studied. In a preceding study, we established an organotypic culture of the mouse retina to rescue gephyrin-deficient retinas and to study their phenotype. ⁵ We observed in wild-type and gephyrin-deficient cultures a relatively normal development of the most major retinal cell types and of their synaptic connections. The focus of that study was on glycinergic and GABAergic synapses. Using immunocytochemical staining, we observed synaptic clustering of GABA_A and GlyR subunits, suggesting a normal development of inhibitory synapses. In the present study, we have added further evidence for synaptic integrity in these cultures. We have shown that GABAergic amacrine cells formed a dense plexus of processes within the IPL (Fig. 1B) . The presynaptic cytomatrix protein, bassoon, was found to be clustered at numerous synapses (Fig. 1C) . Together with previous findings ^{4 5 6 7} these results suggest that structurally intact synapses formed in the organotypic retinal cultures. In the current study, we showed that these synapses were functional. 

Electrophysiological Characterization of Synaptic Activity in Organotypic Cultures of the Mouse Retina

Both amacrine cells and cells located in the ganglion cell layer of organotypic cultures of the mouse retina revealed vigorous spontaneous synaptic activity. Comparable to our recordings from amacrine cells of acutely isolated retinal slices of adult mice ¹⁷ we found two classes of postsynaptic currents: fast-decaying, kynurenate-sensitive sEPSCs that were mediated by ionotropic glutamate receptors and slow-decaying sIPSCs. These sIPSCs were blocked either by bicuculline or strychnine or a combination of both, indicating that they were mediated by GABA_ARs and GlyRs. The obvious differences in the decay kinetics (Fig. 3) provided an easy method to discriminate sEPSCs from sIPSCs by visual inspection. On this basis, we were able to perform most experiments in the absence of kynurenate, which would have inhibited much of the spontaneous synaptic activity by blocking excitation mediated by ionotropic glutamate receptors. The mean τ_w of the sEPSCs that we recorded in organotypic cultures was similar to that recorded in amacrine cells of the adult mouse retina, ¹⁷ in rat retinal ganglion cells, ^{33 34 35} and in rat hippocampal and cortical pyramid cells. ^{36 37} The sEPSC decay kinetics resembled that of recombinant α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs), ^{38 39} but was significantly faster than that of N-methyl-d-aspartate receptors (NMDARs). ⁴⁰ Thus, we conclude that the sEPSCs we observed in organotypic cultures were mediated by AMPARs. We may have missed NMDAR-mediated sEPSCs that were observed previously in AII amacrine cells, ⁴¹ because they may have been inhibited in the presence of Mg²⁺ at the V_h of −60 mV which we applied in our study. 

GABAergic sIPSCs

GABAergic sIPSCs were observed in cells from all three genotypes. First of all, we found that the mean amplitude of GABAergic sIPSCs of amacrine cells (30.4 pA) and cells located in the ganglion cell layer (26.0 pA) from organotypic cultures of geph ^+/+ mice was smaller than that found in amacrine cells of adult mice retinas (42.1 pA) ¹⁷ or ganglion cells of the adult rat retina (42 pA). ³⁴ If we assume that a single sIPSC results from the activation of a single synapse, this decrease in amplitude could reflect a reduction in the number of postsynaptic receptors and/or in the number of vesicles simultaneously released from cells in organotypic cultures. There was no significant difference between the mean peak amplitude and the mean τ_ws of GABA_ARs-mediated sIPSCs in amacrine cells of geph ^+/+ and geph ^+/− retinas, suggesting that geph ^+/− mice may express GABA_AR subtypes similar to those of the amacrine cells of geph ^+/+ retinas (i.e., they possess a mechanism to compensate for the partial gephyrin deficit). The mean τ_w of GABA_ARs we observed in amacrine cells of geph ^+/+ organotypic cultures was in the same range as that of adult mouse amacrine cells from acutely isolated slices, where the simultaneous expression of two GABA_AR subtypes, fast-decaying and slow-decaying, was shown. ¹⁷ Thus, amacrine cells of geph ^+/+ organotypic cultures may express a mixture of GABA_AR subtypes similar to that of amacrine cells of the adult mouse retina. The mean decay time constant (τ_w = 21.5 ms) of GABA_ARs expressed in amacrine cells of geph ^−/− retinas was comparable to the fast-decaying GABA_AR subtype of adult amacrine cells (20.3 ms). ¹⁷ Therefore, the absence of slow-decaying GABAergic sIPSCs and the decrease of the mean sIPSC peak amplitude in amacrine cells of geph ^−/− retinas may reflect that in these cells the clustering of a slow-decaying GABA_AR subtype is impaired. Because we did not find a correlation between amplitudes and decay or increase times, the different decay kinetics of GABAergic sIPSCs in the different genotypes probably reflect the intrinsic properties of distinct GABA_AR subtypes rather than dendritic filtering. 

Subunit Composition of the Synaptic GABA_A Receptors of Amacrine Cells

Our experiments also provide some insights into the possible subunit composition of GABA_AR subtypes expressed in amacrine cells. From experiments with recombinant receptors, it is known that currents mediated by GABA_A receptor subtypes that contain the α1 subunit decay significantly faster than those mediated by subtypes containing the α2 or α3 subunits. ^{42 43} The τ_ws of synaptic currents measured in amacrine cells of geph ^−/− mice are close to those of recombinant α1β1γ2 receptors (τ_w = 20 ms; calculated from data of Ref. ⁴³ ); however, they are significantly faster than those of α2β1γ2 receptors (τ_w = 198.7 ms; calculated from data of Ref. ⁴³ ). In our preceding studies we have shown for geph ^−/− hippocampal and retinal cultures that postsynaptic clusters containing the GABA_ARα2 subunit are most severely reduced. ^{5 30} Thus, we suggest that amacrine cells in cultures of geph ^+/+ mice express at least two types of GABA_A receptor subtype, a slowly-decaying one that may contain the α2 subunit and a fast-decaying one that possibly contains the α1 subunit. In amacrine cells of geph ^−/− mice, synapses containing the α2 subunit appear to be impaired by the absence of gephyrin, whereas synapses containing the α1 subunit appear to function normally. 

Glycinergic sIPSCs

Although we found sIPSCs mediated by GABA_ARs in all cells investigated, only four cells showed glycinergic sIPSCs. The mean peak amplitude of glycinergic sIPSCs that we observed in organotypic cultures (47 pA) was larger than that found in amacrine cells of the adult mouse (28 pA) ¹⁷ and ganglion cells of the adult rat retina (36 pA). ³⁴ Glycinergic sIPSCs recorded in organotypic cultures were faster than those observed in amacrine cells of adult mice. ¹⁷ However, the low number of cells with glycinergic sIPSCs prohibited a more detailed statistical analysis. 

We found a discrepancy between the apparently low number of cells that showed glycinergic IPSCs and the disinhibitory effect of strychnine (Fig. 5) . There are several possible explanations to reconcile the finding. First, it is possible that internally released glycine acts on extrasynaptic receptors. Such extrasynaptic receptors form the pool, from which GlyRs are aggregated at synapses. ⁴⁴ Second, it is also possible that there is a basal, nonvesicular release of glycine that would not show up in sIPSCs. ⁴⁵ Moreover, glycinergic amacrine cells have smaller dendritic fields than GABAergic amacrine cells, ^{46 47 48} suggesting that there is a lower number of glycinergic synapses. Indeed, there are four times more GABAergic than glycinergic synapses in the plexiform layers of the mammalian retina, as has been shown with antibodies specific for GABA_AR and GlyR subunits. ⁵ As expected from the previously reported absence of synaptic GlyR clusters in geph ^−/− retinas, ⁵ we found no glycinergic sIPSCs in organotypic cultures from geph ^−/− mice. Thus, the expression of gephyrin appears to be mandatory for the clustering of GlyRs at functional synapses in amacrine cells and ganglion cells of the mouse retina. 

Our results show that the organotypic culture of the retina provides a powerful model to study synaptic development and function in vitro, because spontaneous activity, known as a prerequisite for neuronal development, is comparable to that observed in the acutely isolated tissue. In particular, this is of importance if transgenic mice that carry severe or lethal mutations are to be investigated. 

 Supported by a fellowship from the Mexican National Council of Science and Technology (JP-L) and German Research Foundation Grants Schwerpunkt 1026 BA 1311/6-2, (KHB) and Sonderforschungsbereich 269/B4 (HW).

 Submitted for publication July 11, 2002; accepted October 8, 2002.

 Disclosure: J. Pérez-León, None; M.J. Frech, None; J.E. Schröder, None; F. Fischer, None; M. Kneussel, None; H. Wässle, None; K.H. Backus, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

(A) Unstained vertical cryostat section obtained with Nomarski optics through an organotypic culture of a mouse retina that was grown for 15 DIV. Arrows: rudimentary inner and/or outer segments of the photoreceptors; arrowheads: split in the IPL. (B, C) Vertical cryostat section through the IPL of an organotypic culture of a mouse retina (15 DIV) double immunostained for GAD65 (B) and for bassoon (C). (B, arrows) Cell bodies of amacrine and displaced amacrine cells. Bassoon immunoreactivity (C) was punctate throughout the IPL; the puncta represent the clustering of bassoon in presynaptic terminals. (D, E) Vertical cryostat section through the IPL of an organotypic culture of a mouse retina (15 DIV) that was immunostained for the γ2 subunit of the GABA_A receptor. (D) Nomarski micrograph showing the IPL, innermost INL, and GCL. (E) Fluorescence micrograph showing numerous γ2-immunoreactive puncta, which represent a clustering of the receptors in postsynaptic densities. OLM, outer limiting membrane; ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL; ganglion cell layer. Scale bars, 20 μm.

Figure 2.

 

Morphologic identification and electrophysiological characterization of an amacrine cell and of a cell found in the ganglion cell layer of organotypic cultures of the mouse retina. (A) Amacrine cell filled with Lucifer yellow. Fluorescence microscopy shows the cell’s dendritic arbor in the proximal IPL. (B) Voltage-activated whole-cell currents from the cell shown in (A). Depolarizing voltage steps evoked mainly noninactivating potassium currents; V_h= −60 mV, step increment 10 mV from −100 to 50 mV, duration 150 ms. (C) Morphology of a cell in the ganglion cell layer. (D) The current-response pattern of these cells was characterized by fast-activating and -inactivating inward currents and outwardly directed noninactivating K⁺ currents. V_h = −60 mV, step increment 10 mV from −100 to 50 mV, duration 150 ms.

Figure 3.

 

Spontaneous synaptic activity in a geph ^+/+ culture. Whole-cell recordings from a cell in the ganglion cell layer of a 32-DIV organotypic culture. (A) Two types of spontaneous synaptic currents (sPSCs) were recorded: fast-decaying sEPSCs, (asterisks) which were blocked by kynurenic acid (not shown) and slow-decaying sIPSCs (arrows). Inset: Trace shows an enlargement of a sEPSC as indicated. (B) Amplitude histograms of sIPSCs and sEPSCs. Excitatory events were restricted to amplitudes smaller then 30 pA. Inset: corresponding cumulative fraction plot. Mean ± SD: sEPSCs 15.8 ± 6.0 pA; 151 events; sIPSCs 20.2 ± 16.3 pA; 737 events. (C) Superimposed sEPSCs and sIPSCs, bottom: averaged traces. (D) Distribution of the τ_ws of sEPSCs and sIPSCs. Note the restricted distribution of sEPSCs to very fast values (sEPSCs: τ_w = 1.4 ± 0.6 ms; 60 events; sIPSCs: τ_w = 26.9 ± 18.2 ms; 531 events). Inset: corresponding cumulative fraction plot.

Figure 4.

 

Characteristics of sIPSCs mediated by GABA_ARs in an amacrine cell of a 9-DIV organotypic culture. (A) Amplitude–time plot of sIPSCs recorded in the whole-cell configuration. Each dot represents the peak amplitude of a single sIPSC. All events were blocked by the application of bicuculline (10 μM) to the bath solution. The application of strychnine (0.5 and 1 μM, as indicated by the bar) did not block sIPSCs, but evoked an increase in the sIPSC frequency. (B) Current traces showing sIPSCs in the control, in the presence of bicuculline, after washout of bicuculline, and in the presence of strychnine. (C) Amplitude histograms of sIPSCs in the control and in the presence of strychnine. (D) Corresponding cumulative fraction plots of the sIPSCs in (C), indicating that strychnine evoked an increase in the peak amplitudes. (E) Averaged and normalized traces of sIPSCs recorded in the absence (control) and presence of strychnine were superimposed to demonstrate their similar kinetics. (F) Cumulative fraction plot of τ_ws of sIPSCs in control and strychnine-containing solutions. The similar decay kinetics and the sensitivity to bicuculline, but not to strychnine, indicated that all sIPSCs were mediated by GABA_ARs.

Figure 5.

 

sIPSCs mediated by GABA_ARs and GlyRs in an amacrine cell from an organotypic culture (28 DIV). (A) Amplitude–time plot of sIPSCs recorded in the whole-cell mode. Strychnine (Stry; 0.5 μM) and bicuculline (Bic; 10 μM) were applied as indicated. Only the simultaneous application of both drugs produced a complete suppression of the sIPSCs. (B) Current traces in the absence and presence of bicuculline and/or strychnine show two types of sIPSCs that differed in their decay kinetics (a and b). Fast-decaying sIPSCs (a) persisted in the presence of strychnine, indicating that they were mediated by GABA_ARs. Slow-decaying sIPSCs (b) remained in the presence of bicuculline, indicating that they were mediated by GlyRs. The coapplication of strychnine and bicuculline suppressed both types of sIPSCs. (C) Events recorded in the presence of strychnine (a) and in the presence of bicuculline (b) were averaged and superimposed, showing their different decay kinetics (mean values ± SD: a, GABA_ARs: τ_w = 11.9 ± 10.9 ms, 823 events; b, GlyRs: 14.9 ± 12.2 ms, 149 events; P < 0.01). (D) Corresponding cumulative fraction plot of the τ_ws of GABA_ARs (a) and GlyRs (b). (E) Amplitude distribution and the corresponding cumulative fraction plot show that GABAergic sIPSCs found in this amacrine cell were characterized by larger peak amplitudes (mean ± SD: a, GABA_ARs: 24.6 ± 14.7 pA, 1019 events; b, GlyRs: 17.1 ± 10.4 pA, 223 events; P < 0.01). (F, G) T10/90s of GABAergic (F) and glycinergic (G) sIPSCs were plotted against the corresponding τ_ws. The absence of correlation (GABAergic sIPSCs: r = 0.027, n = 180, P = 0.496; glycinergic sIPSCs: r = −0.045, n = 100, P = 0.663) indicates that the different decay kinetics is not due to dendritic filtering and reflects the functional difference of these receptors.

Figure 6.

 

Comparison of GABAergic sIPSCs recorded from amacrine cells of organotypic cultures of geph ^+/+, geph ^+/−, and geph ^−/− mice embryos. (A) Continuous whole-cell recordings from three amacrine cells of comparable age (22–29 DIV) of the different genotypes. All these sIPSCs were inhibited by 10 μM bicuculline (not shown), indicating that they were mediated by GABA_ARs. (B) Peak amplitude distributions of GABAergic sIPSCs of the cells shown in (A). (C) Corresponding cumulative fraction plots. The amplitude distributions of geph ^+/+ and geph ^+/− cells were similar but differed significantly from that of geph ^−/− cells (mean ± SD: geph ^+/+: 35.5 ± 24.2 pA, 982 events; geph ^+/−: 32.2 ± 19.4 pA, 122 events; P = 0.153; geph ^−/−: 14.3 ± 5.35 pA, 639 events; P < 0.01 for geph ^+/+ versus geph ^−/− and geph ^+/− versus geph ^−/− results). (D) To demonstrate the different decay kinetics of sIPSCs of the different genotypes single events from the cells shown in (A) were averaged, normalized, and superimposed. (E) Cumulative fraction plot of the τ_ws from the recordings in (A).

Figure 7.

 

The different properties of GABAergic sIPSCs from the different genotypes were not due to dendritic filtering. (A) Plot of the T10/90s against the corresponding τ_ws of GABAergic sIPSCs recorded in amacrine cells geph ^+/+ (left), geph ^+/− (middle), and geph ^−/− (right) organotypic cultures (same cells as displayed in Fig. 6 ). Lines: data fitted by linear regression. No correlation was found between these parameters (geph ^+/+: r = 0.0499, 245 events, P = 0.437; geph ^+/−: r = 0.0069, 122 events, P = 0.939; geph ^−/−: r = −0.016, 115 events, P = 0.883). (B) Plot of the T10/90s against the corresponding peak amplitudes of GABAergic sIPSCs from cells shown in (A): geph ^+/+ (left), geph ^+/− (middle), and geph ^−/− (right). Lines: data fitted by linear regression (geph ^+/+: r = 0.084, 245 events, P = 0.173; geph ^+/−: r = −0.106, 122 events, P = 247; geph ^−/−: r = 0.025, 115 events, P = 0.791).

The authors thank Felicitas Boij and Dagmar Magalai for excellent technical assistance and Angelica Robles for continuous encouragement.