Corresponding author S. Picaud: Laboratoire de Physiopathologie Rétinienne, Médicale A, BP426, 1 Place de l'Hôpital, 67091 Strasbourg cedex, France. Email: firstname.lastname@example.org
1Edge contrast enhancement is an integrated visual function based on the complex centre-surround organization of the cone photoreceptor light response. While centre responses result from direct light activation, surround responses are thought to result from lateral inhibition mediated by horizontal cells. This feedback signal has been attributed to GABAA receptors which have been found in lower vertebrate cones.
2In order to study the GABA response of adult mammalian photoreceptors, we designed a culture system consisting of isolated photoreceptors seeded on a layer of retinal glial cells. Mature rods and cones required the presence of Müller glial cells to survive and develop neurites; they degenerated in the absence of glial cells.
3Cone photoreceptors generated large GABA responses whereas rod photoreceptors did not respond to GABA applications.
4Cone GABA responses consisted of two distinct components, one suppressed by the GABAA receptor blockers bicuculline and SR95531, and the second by the GABAC receptor antagonists TPMPA and imidazole-4-acetic acid (I4AA). Pentobarbital greatly increased the GABAA receptor component whereas it did not affect, or even reduced, the GABAC receptor component. During long GABA applications, GABAA receptor currents desensitized by 78 %, contrasting with the sustained GABAC response.
5Expression of GABAC receptors in cone photoreceptors was confirmed by anti-ρ-subunit immunolabelling of porcine retinal sections.
6These results indicate that both GABAA and GABAC receptors may participate in the feedback synapse from horizontal cells to cone photoreceptors in the mammalian retina.
GABA is the major inhibitory neurotransmitter in the brain as well as in the vertebrate retina. Its actions on bicuculline-sensitive ionotropic GABAA receptors and metabotropic GABAB receptors have been well documented. GABAC receptors, a third receptor type less common in the CNS (Johnston, 1996), have recently drawn much attention to the retina because the GABAC receptor ρ-subunits were cloned from a retinal library (Cutting et al. 1991), and were expressed in oocytes from the corresponding cDNA (Shimada et al. 1992) or after injecting bovine retinal mRNA (Polenzani et al. 1991). GABAC receptors were subsequently recorded from retinal neurons: bipolar cells (rat: Feigenspan et al. 1993; salamander: Lukasiewicz et al. 1994; fish: Matthews et al. 1994), horizontal cells (fish: Qian & Dowling, 1993) and ganglion cells (salamander: Zhang & Slaughter, 1995). Although these receptors gate Cl− channels similarly to GABAA receptors, they have distinct pharmacologies, GABA sensitivities and modulations (Bormann & Feigenspan, 1995; Johnston, 1996). Their pharmacological profile includes a lack of bicuculline sensitivity and their ρ-subunits exhibit only 30–38 % homology with those of GABAA receptors (Cutting et al. 1991).
In mammals, GABAC receptor-gated currents were not recorded in horizontal cells (Blanco et al. 1996). Their presence in rat bipolar cells was first detected in cultured neonatal cells (Feigenspan et al. 1993), and confirmed with freshly dissociated adult cells (Feigenspan & Bormann, 1994) and retinal slices (Euler et al. 1996). In situ hybridization on rat retinal sections and isolated cells confirmed the expression of ρ1- and ρ2-subunits in bipolar cells (Enz et al. 1995). Furthermore, immunostaining for ρ-subunits was mostly located in the subdivision of the outer and inner plexiform layers containing bipolar cell processes (Enz et al. 1996; Koulen et al. 1997).
No information is yet available on the GABA sensitivity of mammalian photoreceptors despite the major role described for GABA in the feedback signal from horizontal cell to cone photoreceptors in lower vertebrates (for review see Piccolino, 1995). This mechanism, which generates the surround component of the cone complex antagonistic centre- surround receptive field, is thought to be mediated by GABAA receptors found at cone terminals in turtle (Kaneko & Tachibana, 1986). This hypothesis is only supported in the mammalian retina by GABAA receptor immunolabelling, irregularly found in cone pedicles (Hughes et al. 1989, 1991).
Here we provide physiological and histological evidence for the expression of GABAC receptors in adult mammalian cone photoreceptors. Physiological measurements were obtained from a new photoreceptor-glia co-culture system that we specifically developed to record photoreceptors.
Pig eyes were obtained from the local slaughterhouse immediately after death. After sterilization of the whole eye in 70 % alcohol, the cornea with lens and vitreous attached was removed. Retinae separated from the posterior eye cup were placed in Dulbecco's modified Eagle's medium-Hams F12 (DMEM/F12, Gibco, Life Technologies, Paisley, UK).
In order to prepare retinal glial feeder layers, retinal cells were dissociated as described previously (Gaudin et al. 1996). Müller cells were isolated from the cell suspension following the procedure described by Guidry (1996). The suspension was laid on a 10 ml continuous density gradient composed of 0–50 % Percoll in normal saline for centrifugation at 500 g for 5 min. The band located at the middle of the tube contained the purified Müller cells, which were seeded at 0.4 × 105 cells cm−2 in DMEM plus 10 % fetal calf serum onto tissue culture plates containing coverslips pre-coated with polylysine and laminin. Glia were allowed to grow to near confluence prior to the addition of photoreceptors.
Photoreceptors were isolated as described previously (Dreyfus et al. 1996) by horizontal sectioning of the retina with a vibratome series 1000 (Technical Product International, Saint Louis, MO, USA). A square piece of central retina (∼ 1 cm2) was isolated and flattened scleral surface down on a 20 % gelatin block using 4 % gelatin warmed to 40°C in CO2-independent DMEM. When the razor blade just touched the vitreal surface of the retinal tissue, it was lowered by 150 μm in order to remove the inner half of the retina (preliminary microscopic studies enabled the determination of the cut depths required to remove cleanly the inner nuclear and ganglion cell layers). The remaining photoreceptor layer was then retrieved with excess attached gelatin which was eliminated by warming the tissue at 37°C. The vibratome setting for blade advance was adjusted to level 2 (level 10, Vmax= 1.25 mm s−1) and the oscillation frequency set to maximum. The isolated photoreceptor layer was then treated following the dissociation treatment described above but with a shorter (10 min) papain incubation and without mechanical trituration. Following this procedure, more than 90 % of the cells were viable as assessed by Trypan Blue exclusion. Cells were then seeded into two wells of the 6-well tissue culture plates containing Müller cell cultures. Culture medium was renewed weekly and such cultures could be maintained for up to 2 weeks in vitro. Recordings were performed between 2 and 10 days in culture.
Recording pipettes were pulled from thin-walled borosilicate glass (TW150F, World Precision Instruments, Sarasota, FL, USA) using a Brown and Flaming-type puller (P-87, Sutter Instruments, Novato, CA, USA). Pipettes had a measured resistance in Ringer solution of 5–10 MΩ and a series resistance during recordings estimated at 14.7 ± 0.7 MΩ (mean ±s.e.m., n= 10). An RK400 patch-clamp amplifier (Biologic, Grenoble, France) was used to voltage clamp the recorded cells. Data were filtered at 3 kHz and digitized at 2 kHz using a data acquisition Labmaster board (Scientific Solutions, Solon, OH, USA) mounted in an IBM-compatible personal computer. Experimental data were acquired and analysed using the Patchit and Tack software packages, respectively (Grant & Werblin, 1994). Data are expressed as means ±s.e.m., of n recordings.
Solutions and drug application
The standard recording solution contained (mm): 135 NaCl, 5 KCl, 1 CaCl2, 1 MgCl2, 10 glucose and 5 Hepes, pH titrated with NaOH to 7.75. GABA was applied to the cells by puff applications (50 ms, 20 p.s.i.; Picospritzer II, General Valve Corporation, Fairfield, NJ, USA) every minute or 30 s, while other drugs were delivered by a general perfusion system driven by gravity (∼ 2 ml min−1) in the recording chamber that contained a volume of less than 1 ml. Drugs were continuously applied until GABA responses stabilized indicating that equilibrium conditions had been reached.
The pipette solution used in most whole-cell recordings contained (mm): 140 KCl, 1 MgCl2, 0.5 EGTA, 5 ATP (disodium salt) and 4 Hepes, adjusted to pH 7.4 with KOH. In the solution with a calculated Cl− equilibrium potential (ECl) of −30 mV, 96 mm potassium gluconate was substituted for 96 mm KCl.
The general procedure for double immunolabelling was performed as described previously (Gaudin et al. 1996). The antibodies used in the present study were anti-arrestin polyclonal antibody (generous gift of Dr I. Gery, NIH, Bethesda, MD, USA), specific for rod and cone photoreceptors (Wacker et al. 1977), and anti-rhodopsin monoclonal antibody (rho4D2) specific for rods (Hicks & Molday, 1986).
Immunostaining with the anti-ρ-subunit antibody (generous gift of Dr R. Enz & Professor H. Wässle, MPI, Frankfurt/M, Germany; Enz et al. 1996) was performed on thin frozen sections (< 10 μm) following 3 min fixation with 2 % paraformaldehyde in phosphate-buffered saline at room temperature. Sections were incubated with the primary polyclonal antibody for 24 h at 4°C and the secondary anti-rabbit antibody, coupled to Oregon Green 514 (Molecular Probes), for 5 h at 4°C.
Chemicals and proteins
Unless otherwise stated in the text, all chemicals, enzymes and secondary antibodies were obtained from Sigma, except for 2-(carboxypropyl)-3-amino-6,4-methoxyphenyl)pyradazynium bromide (SR95531; Research Biochemicals Inc.) and (1,2,5,6-tetrahydropyridine-4-yl)methylphosphinic acid (TPMPA; Tocris, Bristol, UK).
Photoreceptor cell culture
Figure 1 illustrates the aspect of photoreceptors cultured either on a layer of Müller glial cells (PR + Glia) or directly on laminin-polylysine-coated coverslips (PR). On the glial sheet, pig photoreceptors were present as small clusters of birefringent cells, exhibiting a neuronal morphology with short neurites (Fig. 1A–C). Rods could be distinguished from cones by their immunoreactivity, rods being stained by both rho4D2 and anti-arrestin antibodies, whereas cones were uniquely labelled by anti-arrestin antibody. Many cones were thus observed in the cultures (arrow in Fig. 1A–D), these cells usually being larger in size than rods. When seeded directly onto laminin-polylysine-coated coverslips, photoreceptors became irregular in shape and never grew processes (Fig. 1E–G). Their rapid degeneration was indicated by the absence of DAPI nuclear staining in many cell bodies (arrow in Fig. 1E–H). The selectivity of rho4D2 and anti-arrestin antibodies for photoreceptors in the porcine retina was assessed on retinal sections (Fig. 1I–L). These observations indicate that glial cells can promote the survival in culture of fully differentiated adult photoreceptors.
GABA-elicited Cl− current in cones but not rods
Whole-cell recordings from such cultured photoreceptors showed two types of response to voltage steps that were similar to responses obtained from freshly dissociated rods and cones (data not shown). When GABA was puff applied on photoreceptors voltage clamped at −70 mV, no currents or a current < 5 pA was observed in rods (data not shown) whereas it elicited a very large current in cones (Fig. 2). To determine whether the GABA-elicited current in cones was carried by a Cl− channel as in turtle (Kaneko & Tachibana, 1986), its reversal potential was measured in the presence of different Cl− concentrations in the recording pipette (Fig. 2). GABA puffs were delivered while cells were voltage clamped at different potentials ranging from −70 to +30 mV. The reversal potential for GABA currents shifted from −6.5 ± 0.7 mV (n= 4) when the Cl− concentrations were almost symmetrical in the pipette to −33.1 ± 3.3 mV (n= 3) when the calculated equilibrium potential for Cl− was brought to −30 mV by substituting gluconate ions for Cl− in the recording pipette. This result was consistent with GABA gating Cl− channels in cones.
GABAA receptor pharmacology
The pharmacology of the GABA-gated Cl− currents was analysed with specific blockers of GABAA receptors (Fig. 3). Picrotoxin suppressed the GABA-gated currents by 92 ± 3 % (n= 6), whereas bicuculline (100 μM) only decreased the response by 55 ± 5 % in the same cells (Fig. 3A and B). The reversal potential of the bicuculline-resistant current also approximately followed the equilibrium potential for Cl− (ECl= 0 mV, Vrev= 1.8 ± 1.0 mV, n= 4; ECl=−30 mV, Vrev=−22.5 ± 3.1 mV, n= 3). SR95531 (20 μM), another specific blocker of GABAA receptors (Lukasiewicz et al. 1994; Matthews et al. 1994), similarly suppressed a fraction of the GABA response (Fig. 4A). It decreased the amplitude of the GABA response by 70.1 ± 4.1 % (n= 5), while bicuculline suppressed the response by 76.7 ± 2.3 % in the same cells. Of all the cells recorded so far, only a few were found to generate GABA responses fully sensitive to either bicuculline or SR95531. These results suggested that the GABA response is not solely mediated by GABAA receptors.
GABAC receptor pharmacology
To assess the origin of the bicuculline- and SR95531-resistant current, we applied TPMPA (50 μM), a selective antagonist of GABAC receptors (Ragozzino et al. 1996), to the cells. TPMPA decreased the GABA-elicited response by 32.9 ± 0.1 % while SR95531 (100 μM) suppressed it by 84.7 ± 3.1 % in the same cells (n= 5; Fig. 4A). TPMPA and SR95531 had additive effects in suppressing the GABA response (99.4 ± 0.2 %, n= 5). When applied separately, even high concentrations of SR95531 (1 mm) or TPMPA (100 μM) could not completely suppress the GABA-elicited response. Similarly, imidazole-4-acetic acid (I4AA, 200 μM), another selective GABAC receptor antagonist (Kusama et al. 1993), decreased the GABA-elicited current by 44.4 ± 8.5 % (Fig. 4B), while bicuculline suppressed the response by 79.9 ± 1.2 % in the same cells (n= 5). Bicuculline and I4AA had additive effects in blocking the GABA-elicited current, totally suppressing the response (94.2 ± 2.7 %, n= 3). In Fig. 4B, it is shown that the sum of the bicuculline- and I4AA-resistant currents nearly matched the control GABA response. A similar result was obtained (Fig. 4A) when comparing control GABA responses with SR95531- and TPMPA-resistant currents. These results suggested that the GABA response was composed of a GABAA and a GABAC receptor-elicited current.
Pentobarbital modulations of GABAA and GABAC receptors
Since pentobarbital has been shown to increase selectively GABAA-elicited currents in contrast to GABAC-gated currents, we used this selective sensitivity to verify whether GABAA and GABAC receptors were expressed in cones. Pentobarbital (100 μM) was found to increase the GABA response in amplitude (by 129.2 ± 4.5 %, n= 8) and duration. A similar increase of the GABA-elicited current (136.6 ± 6.1 %, n= 4) was also observed during bath application of the GABAC receptor antagonist TPMPA (50 μM) (Fig. 5A). In contrast, when GABAA receptors were blocked with SR95531 (100 μM), pentobarbital never increased the GABA response; the SR95531-resistant current was instead slightly decreased in some cells (82.0 ± 5.4 %, n= 6; Fig. 5B). The absence of upregulation by pentobarbital of the SR95531-resistant current is consistent with the notion that this current is generated by GABAC receptors.
Differential GABA receptor desensitization
Since neurons in the outer retina do not fire action potentials but communicate at graded potential synapses, one may wonder whether GABAA receptors, which desensitize during prolonged GABA application, could contribute significantly in situ to tonic GABA inhibition of cones. To address this question, long GABA applications (8 s) were delivered to the cells. GABA responses showed a transient component followed by a sustained component (Fig. 6). When the GABAA receptor-elicited current was isolated with TPMPA, the response exhibited a large decrease in amplitude (78.7 ± 1.4 %, n= 4). In contrast, the GABAC receptor-elicited current was either strictly constant or exhibited a slight and slow decrease in amplitude over time (17.2 ± 6.8 %, n= 5; Fig. 6). The GABAA receptor desensitization also explained the faster decay observed during short puff applications in the presence of GABAC receptor antagonists (I4AA and TPMPA, τ= 604 ± 30 ms, n= 7) than that recorded in the presence of GABAA receptor blockers (bicuculline and SR95531, τ= 1786 ± 242 ms, n= 7; Fig. 4A, inset). Despite this desensitization, GABAA receptors contributed 43.7 ± 1.7 % (n= 3) of GABA responses even after an 8 s application.
Expression of GABAC receptors in situ
To determine whether cones express GABAC receptors in vivo, porcine retinal sections were stained with an antibody directed against the GABAC receptor ρ-subunit (Enz et al. 1996). Immunopositive cell bodies were observed at the scleral face of the outer nuclear layer (arrow in Fig. 7), where cones are located, and in the inner nuclear layer. No staining was observed in rods or in photoreceptor outer segments. When the primary antibody was omitted, no immunopositive cells were found. This specific immunolabelling is consistent with in situ expression of GABAC receptors in cone photoreceptors.
Adult mammalian cone photoreceptors were shown here to generate large GABA-elicited currents mediated by GABAA and GABAC receptors. In contrast, rods had no GABA-elicited responses. These results support the notion that mammalian cones but not rods receive a GABA-mediated feedback from horizontal cells. They also challenge the idea that GABAC receptors are only expressed in bipolar cells within the mammalian retina (Enz et al. 1996). The expression of GABAC receptors in mammalian cones (this study), fish horizontal cells (Qian & Dowling, 1993), bipolar cells from various species (rat: Feigenspan et al. 1993; salamander: Lukasiewicz et al. 1994; fish: Matthews et al. 1994), and salamander ganglion cells (Zhang & Slaughter, 1995) is surprising in contrast to the sparse localization of GABAC receptors in the CNS. This distribution in the retina is suggestive of the suitability of GABAC receptor features to the constraints in processing tonic visual information.
Glial dependence for adult photoreceptor survival
Glial cells are not required for the differentiation of embryonic chick precursor cells into photoreceptors and their subsequent survival in vitro for several days (Adler et al. 1984). We similarly found that purified immature rat photoreceptors (postnatal day 5–15) can survive alone for several days in vitro and develop neurites (V. Fontaine & D. Hicks, unpublished observations). However, retinal Müller cells have been shown to provide a preferred substrate for in vitro neurite extension by both neonatal and adult mammalian rods (Kljavin & Reh, 1991; Hicks et al. 1994). Our results indicate further that Müller cells are required for the survival of adult photoreceptors. Further studies are needed to evaluate whether this survival induction relies upon glial surface adhesion molecules or release of diffusible factors.
GABA receptor pharmacology
The additive effects of GABAA and GABAC receptor antagonists on mammalian cones suggested that their GABA response was generated by both receptor types. However, these antagonist effects were not purely additive, since the combination of their individual effects was greater than 100 % (TPMPA, 32.9 %; SR95531, 84.7 %). This distortion is also visible when adding the TPMPA- and SR95531-resistant recordings (a+b, Fig. 4A and B) and comparing the resulting trace with the initial control GABA response; the sum is slightly smaller than the initial response. The amplitude decrease in GABA responses that is seen over long recording periods (Fig. 3B) may contribute partly to this difference. However, it is more likely that this effect is due to the incomplete selectivity of the GABA receptor antagonists. GABAC receptor antagonists also act as weak antagonists at GABAA receptors and might therefore reduce GABAA receptor currents. For instance, TPMPA has an apparent antagonist dissociation constant (Kd) for GABAA receptors of 320 μM, in contrast to 2.3 μm for GABAC receptors (Ragozzino et al. 1996). Despite this, the use of TPMPA (50 μM) and SR95531 (100 μM) seemed appropriate to isolate GABAA and GABAC receptor currents, even if their respective amplitudes were reduced. This conclusion was supported by the consistency between observed properties of the GABAA and GABAC receptor components and known features of these receptors. The GABAC receptor component was not affected or was slightly reduced by pentobarbital whereas the GABAA receptor component was greatly increased by pentobarbital. The GABAA receptor component was desensitizing whereas the GABAC receptor current remained sustained during long puff applications. This pharmacological isolation of the two receptor currents should prove very useful for further characterization of GABAC receptors in mammalian cones and for demonstrating their physiological expression in situ.
Evidence for GABAA and GABAC receptors in mammalian cones
Kaneko & Tachibana (1986) showed that GABA acts at axon terminals of turtle cones at GABAA receptors. In this study, we recorded both GABAA and GABAC receptor-gated currents in adult porcine cones. This result is consistent with the GABAA receptor immunolabelling found in cat cone pedicles (Hughes et al. 1991) although such immunolabelling was not observed in monkey (Hughes et al. 1989). This apparent discrepancy in immunolabelling might be due to interspecies differences in the proportion of GABAA/GABAC receptors in mammalian cone terminals.
In the immunocytochemical localization of GABAC receptor ρ-subunits, Enz et al. (1996) observed a punctate labelling in the outer plexiform layer of all species investigated (rat, rabbit, cat and macaque monkey). At high power magnification, the labelling seemed to be aggregated on dendritic tips of rod bipolar cells in rats, whereas it coincided with the flat tops of putative cone bipolar cells in monkey. This outer plexiform layer staining was attributed to bipolar cells, since GABAC receptor-gated current had only been reported in these mammalian retinal cells (Feigenspan et al. 1993; Feigenspan & Bormann, 1994). In view of our physiological and immunocytochemical results, the punctate staining in the outer plexiform layer could be attributed to both cone terminals and bipolar cell dendrites.
Physiological role of GABA receptors in cones
In the mammalian retina, horizontal cells may release GABA as in lower vertebrates, since they can be stained by both anti-GABA antibodies and antibodies directed against glutamic acid decarboxylase (GAD), the GABA-synthesizing enzyme (Nishimura et al. 1985). The presence of GABA receptors in cultured adult porcine cones suggests that mammalian cones are sensitive to this horizontal cell GABA release. Similarly to bipolar cell terminals (Lukasiewicz & Werblin, 1994; Matthews et al. 1994), GABAA and GABAC receptor activation may reduce depolarization-induced Ca2+ influx in cone pedicles, thereby reducing their glutamatergic transmission to postsynaptic cells. This GABA feedback signal from horizontal cells is usually considered to generate the cone surround response in lower vertebrates (see, for review, Piccolino, 1995). Rods, which do not bear a centre- surround light response similar to that of cones, appear to express neither GABAC nor GABAA receptors. The incomplete block of the surround response by bicuculline in salamander cones (Wu, 1991) is consistent with the notion proposed by Piccolino (1995) that GABAC receptors may also participate in the production of this feedback signal in lower vertebrates. Our data do not exclude the possibility, however, that GABA receptors function instead as a switch from nocturnal to diurnal cone vision as proposed by Verweij et al. (1996), who reported that cone surround response is produced by a GABA-independent modulation of Ca2+ currents in goldfish retina. The absence of desensitization makes GABAC receptors particularly well adapted to process this tonic visual information.
This work was supported by INSERM, AFRP/retina-France, ADRET-Alsace, DRET/DGA, Fédération des Aveugles de France, Fondation de l'Avenir, Fondation de France, IRIS Servier, and MGEN/INSERM. B. P. received a fellowship from AFRP/retina France.