GABA blockade unmasks an OFF response in ON direction selective ganglion cells in the mammalian retina

Abstract

One unique subtype of retinal ganglion cell is the direction selective (DS) cell, which responds vigorously to stimulus movement in a preferred direction, but weakly to movement in the opposite or null direction. Here we show that the application of the GABA receptor blocker picrotoxin unmasks a robust excitatory OFF response in ON DS ganglion cells. Similar to the characteristic ON response of ON DS cells, the masked OFF response is also direction selective, but its preferred direction is opposite to that of the ON component. Given that the OFF response is unmasked with picrotoxin, its direction selectivity cannot be generated by a GABAergic mechanism. Alternatively, we find that the direction selectivity of the OFF response is blocked by cholinergic drugs, suggesting that acetylcholine release from presynaptic starburst amacrine cells is crucial for its generation. Finally, we find that the OFF response is abolished by application of a gap junction blocker, suggesting that it arises from electrical synapses between ON DS and polyaxonal amacrine cells. Our results suggest a novel role for gap junctions in mixing excitatory ON and OFF signals at the ganglion cell level. We propose that OFF inputs to ON DS cells are normally masked by a GABAergic inhibition, but are unmasked under certain stimulus conditions to mediate optokinetic signals in the brain.

Over a dozen subtypes of retinal ganglion cell deliver visual signals to higher brain centres (Wässle & Boycott, 1991; Rockhill et al. 2002; Coombs et al. 2006). Each subtype receives a mixture of spatially and temporally distinct synaptic inputs and thereby carries different neural representations of the visual world (Barlow et al. 1964; Caldwell & Daw, 1978; Roska & Werblin, 2001; Roska et al. 2006). These include the fundamental segregation of the signalling of luminance increment and decrement into the parallel ON and OFF pathways (Famiglietti & Kolb, 1976; Bloomfield & Miller, 1986).

One unique subtype of ganglion cell is the direction selective (DS) cell, which responds vigorously to stimulus movement in a preferred direction, but weakly to movement in the opposite or null direction (Barlow et al. 1964; Oyster, 1968). While the ON-OFF DS cells project to the lateral geniculate nucleus, the ON subtype of DS cell targets the accessory optic system (AOS) where their directional signals underlie the optokinetic response (OKR), a compensatory eye movement that stabilizes an image on the retina during slow head rotation (Oyster et al. 1980; Simpson, 1984).

The mechanism by which DS ganglion cells can derive the direction of stimulus motion has been a classic problem of neural computation over the last 40 years. Numerous studies have implicated both GABAergic and cholinergic inputs from presynaptic starburst amacrine cells as critical to the generation of direction selective responses (reviewed by Taylor & Vaney, 2003; Demb, 2007). It has been posited that an asymmetric release of GABA and possibly acetylcholine from starburst cell dendrites gives rise to a null inhibition or preferred excitation critical to direction selective ganglion cell responses (Barlow & Levick, 1965; Wyatt & Daw, 1975; Fried et al. 2002; Borg-Graham & Grzywacz, 1991; Grzywacz et al. 1997, 1998).

Consistent with an inhibitory mechanism, application of the GABA receptor blocker picrotoxin results in the loss of the direction selective responses of DS cells (Caldwell et al. 1978; Kittila & Massey, 1995; Ackert et al. 2006). Interestingly, Ariel & Daw (1982) reported that application of picrotoxin also revealed an OFF response in ON DS ganglion cells, although they attributed this to technical artifact. Recently, however, a late, secondary response to a moving slit of light has been reported for ON DS cells in the mouse retina (Sun et al. 2006). Further, GABAergic blockade has been shown to unmask an OFF response in other subtypes of ON ganglion cells in the rabbit retina (Roska & Werblin, 2001).

We therefore reexamined the effects of GABA blockade on the responses of ON DS cells. Here, we report that application of picrotoxin regularly unmasks a robust centre-mediated OFF response in ON DS ganglion cells. Interestingly, we find that the OFF response is direction selective, but, surprisingly, its preferred direction is opposite to that of the ON response of the same cell. Moreover, we find that the OFF response is abolished by application of a gap junction blocker, suggesting that it arises from electrical synapses between ON DS and neighbouring polyaxonal amacrine cells. These results suggest a novel role for gap junctions as conduits for mixing of excitatory ON and OFF signals at the ganglion cell level. Further, they reveal an unexpected complexity in the circuitry subserving direction selectivity in the retina.

Methods

Flattened retina–sclera preparation

The animal preparation used in this study has been described previously (Hu et al. 2000). Briefly, 36 adult New Zealand White rabbits (Oryctolagus cuniculus) were anaesthetized with an intraperitoneal injection of 40% ethyl carbamate (2.0 g (kg body weight)⁻¹) and a local subcutaneous injection of 2% lidocaine hydrochloride (1 ml) to the eyelids and surrounding tissue. The eye was then removed under dim red illumination and hemisected ∼1 mm posterior to the ora serrata. The vitreous humour was removed with ophthalmic sponges, and the resultant retina eyecup was flattened by making radial cuts at the periphery in a Maltese-cross configuration. The retina eyecup was then placed in a superfusion chamber, which was mounted on the stage of an upright light microscope (BX50 WI; Olympus, Center Valley, PA, USA) within a light-tight Faraday cage. The eyecup was superfused at a rate of 25 ml min⁻¹ with a mammalian Ringer solution (in mm): NaCl (120.0), KCl (5.0), NaHCO₃ (25.0), Na₂HPO₄ (0.8), NaH₂PO₄ (0.1), glucose (10.0), ascorbate (0.01), MgSO₄ (1.0) and CaCl₂ (2.0). The superfusate was kept at a constant temperature of 34°C, with oxygenation and pH 7.4 maintained by bubbling with a gaseous mixture of 95% O₂–5% CO₂. After enucleations, animals were killed with an intracardial injection of ethyl carbamate (15 ml of 40% solution). All animal procedures were in compliance with the NIH Guide for the Care and Use of Laboratory Animals and approved by the Institutional Animal Care and Use Committee at NYU School of Medicine.

To visualize cells for recording, five drops of 0.1% Azure B were applied to the retina to label cells in the ganglion cell layer. A 780 nm cut-off filter allowed transmission of infrared (IR) light from below the stage and then up through a condenser and the glass coverslip mounted in the superfusion chamber. An IR sensitive CCD camera (VE-1000; Dage-MTI, Michigan City, IN, USA) captured the retinal image that was displayed on a video monitor outside the Faraday cage. This protocol allowed retinas to remain in a dark-adapted state during viewing.

Electrophysiology

Single or dual extracellular recordings were obtained from neurons using carbon fibre microelectrodes (World Precision Instruments, Sarasota, FL, USA) attached to an isolated AC differential amplifier (DAM80i, World Precision Instruments). It should be noted that the carbon fibre electrodes effectively recorded activity from cells only when placed directly atop their somata. Thus, there was no ambiguity in identification of the cell being recorded.

For pharmacological experiments, drugs (Sigma, St Louis, MO, USA) were applied to the retina by switching from the control Ringer solution to one containing 25–100 μm picrotoxin, 25 μm 18β-glycyrrhetinic acid, 5 μm neostigmine, or 100 μm hexamethonium. All recording data were digitized online with an analog-to-digital board (Digidata 1200; Axon Instruments, Sunnyvale, CA, USA) and stored for off-line analyses.

Following extracellular recordings, cells were penetrated with intracellular glass microelectrodes under visual guidance for labelling with Neurobiotin. The microelectrodes were filled at their tips with 4%N-(2-amino-ethyl)-biotinamide hydrochloride (Neurobiotin, Vector Laboratories; Burlingame, CA, USA) in 0.1 m Tris buffer, pH 7.4, and then back-filled with 4 m KCl. Final DC resistances of these electrodes ranged from 350 to 450 MΩ. Neurobiotin was injected into the cell with a combination of sinusoidal (3 Hz; 0.4 nA; peak-to-peak) and direct current (0.4 nA) applied simultaneously. This method allowed for passage of tracer through the microelectrode without polarization. All injections were made in and around the visual streak region corresponding to 1.0–3.2 mm from the optic disk.

Light stimulation

Light intensity of visual stimuli was measured using a portable photometer/radiometer (Ealing Electro-Optics; Hollistan, MA, USA). Intensities are expressed as the time-averaged rate of photoisomerizations per rod (Rh* rod⁻¹ s⁻¹). For these calculations it was necessary to determine the effective collecting area (ECA) of a rod at the peak absorption wavelength. Rod outer segments in the rabbit retina are typically 11 μm long and 1 μm in diameter (Prince & McConnell, 1964). Taking a specific axial density of rhodopsin of 0.015 μm⁻¹ and a quantum efficiency of photoisomerization of 0.67 (Liebman & Entine, 1964; Dartnell, 1968), the ECA can be calculated by the following equation:

Thus, 5.8 photons μm⁻² s⁻¹ was approximately equivalent to 1 Rh* rod⁻¹ s⁻¹. The ECA was reduced by 0.8% and 1.0% to account for the lower absorption of the 468 nm and 480 nm test stimuli from the peak of 502 nm based on the Dartnell nomogram for the rod spectral sensitivity in the rabbit (Dodt & Elenius, 1960; Nuboer, 1971; DeMonasterio, 1978).

Light stimulus intensities were kept within the scotopic range and retinas were maintained in a dark-adapted state. For initial surveys of cells’ light-evoked responses, a green light emitting diode (λ_max= 468 nm) focused onto the retinal surface provided full-field illumination in the low scotopic intensity range (4.7 Rh* rod⁻¹ s⁻¹). To test direction selectivity, a 50 μm-wide/275 μm-long rectangular slit of light was projected onto the retina using a 100 W halogen lamp in a dual-channel Newtonian light bench with a narrowband chromatic filter (λ_max= 480 nm). The intensities of these stimuli were also kept within the scotopic range (2.5–10.0 Rh* rod⁻¹ s⁻¹). In initial experiments, the slit was swept back and forth across the retinal surface at four orientations, 0, 45, 90, or 135 deg, where 90 deg corresponded to the orientation of the visual streak. This corresponded to eight directions of movement: 0, 45, 90, 135, 180, 225, 270 and 315 deg. The slit speed ranged from approximately 90 to 180 μm s⁻¹. Since we found that the direction selectivity of ON DS cells were predominantly in the dorsal (0 deg), ventral (180 deg), or nasal (90 or 270 deg dependent on right or left eye) directions, subsequent experiments used slits moving only along the two cardinal axes (4 directions) (see Fig. 4).

Figure 4. The OFF response of ON DS ganglion cells is direction selective.

A, polar plot summarizing the directions for which the ON response of 22 ON DS cells were selective under control conditions and the OFF response unmasked by PTX. The stimulus was a 50 μm wide/275 μm long rectangular slit of light moved in 8 directions. Cells could be divided into three groups in which the ON response was selective roughly to stimulus movement in the dorsal, ventral, or nasal directions. Arrows indicate average summed vector. Grey arrows indicate that summed vectors of directional preference for individual cells, whereas the black arrows indicate the average summed vector for each cell population. Clearly, the OFF response showed a directional preference opposite to that of the ON response for the 3 ON DS cell subtypes. B, responses of a ON DS cell to a 50 μm wide slit of light swept across the retina in 4 directions: dorsal (D), ventral (V), nasal (N), and temporal (T) under control conditions. The cell was direction selective for temporal-to-nasal (leftward) movement along the horizontal axis. After application of PTX, both ON and OFF responses were visible. However, while the direction selectivity of the ON response was abolished by PTX (50 μm), the OFF response showed direction selectivity, but to movement in the nasal-to-temporal (rightward) direction. Application of PTX (50 μm) and l-AP4 (50 μm) abolished the ON response and isolated the OFF response. Slit speed was approximately 180 μm s⁻¹.

Concentric spots of green light (λ_max= 480 nm) were delivered by the second channel of the light bench at the same range of intensities as for the slit stimuli. The spots ranged from 75 to 1750 μm in diameter. For area summation measures, the smallest spot of light was centred over the somata of a recorded cell and larger diameter spots were then presented in sequence while maintaining the same intensity for all stimuli.

Analysis of spike activity

Spikes were sorted and time-stamped off-line using commercially available software (Off-Line Sorter; Plexon, Dallas, TX, USA). A direction selectivity ratio was computed for ON DS ganglion cells by averaging the spikes in a 1 s bin for 20 sweeps in one direction and then in the opposite direction. The ratio was then taken as the quotient of the two averages. It should be noted that since ON DS ganglion cells showed little spontaneous activity, light-evoked activity was easily differentiated from background firing.

Histology and immunocytochemistry

After electrophysiological experiments, retinas were fixed in a cold (4°C) solution of 4% paraformaldehyde–0.1% gluteraldehyde in 0.1 m phosphate buffer, pH 7.3, for 12 min. Retinas were then detached, trimmed, fixed onto a gelatinized glass coverslip, and left in fixative overnight at 4°C. They were washed for 4–5 h in 10 mm sodium phosphate buffered saline (PBS) (9% saline, pH 7.6) and then reacted with the Elite ABC kit (Vector Laboratories) and 1% Triton X-100 (Sigma) in PBS overnight at 4°C. Retinas were processed for peroxidase histochemistry using 3,3′-diaminobenzidine (DAB) with cobalt intensification, dehydrated, cleared, and flat-mounted in Permount (Fisher Scientific, Pittsburgh, PA, USA). Images of labelled neurons were captured by a cooled CCD camera (Spot 2, Diagnostic Instruments) followed by software manipulation of brightness and contrast (Adobe Photoshop).

Alternatively, Neurobiotin injections in some retinas were visualized using a Cy3-conjugated streptavidin reagent (Sigma). After labelling with streptavidin-Cy3, retinas were washed in PBS for 1 h. Subsequently, retinas were incubated in a primary antibody solution of goat anti-choline acetyltransferase (ChAT) antibody (Chemicon; Temecula, CA, USA) at a concentration of 1: 100, and mouse anti-CtBP2 (Ribeye; BD Biosciences; San Jose, CA, USA) at a concentration of 1: 1 000 for 72 h at 4°C. Retinas were then washed for 2 h in PBS, before incubation in a secondary antibody solution of donkey anti-goat Cy2 (1: 500) and donkey anti-mouse Cy5 (1: 500) overnight at 4°C. Secondary antibodies were purchased from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Retinas were subsequently washed in PBS for 3 h and mounted in Vectashield mounting medium (Vector Laboratories). Retinas were imaged using a Zeiss 510 Meta confocal microscope (Zeiss; Thornwood, NY, USA).

Results

Visual targeting of ON DS ganglion cells

To target ganglion cells for recordings and injections, we utilized a flattened retina–sclera preparation in which neurons in the ganglion cell layer (GCL) can be visualized under IR illumination (Hu et al. 2000). Cells were recorded in and around the visual streak region. As we reported recently (Ackert et al. 2006), ON DS ganglion cells could be visualized as a regular mosaic of relatively large somata, approximately 15–20 μm in diameter, which had a crescent-shaped appearance following Azure B labelling. ON DS cells were labelled with Neurobiotin following electrophysiological recording. Recorded ON DS cells labelled in the present experiments displayed a soma/dendritic morphology similar to that described previously for the AOS-projecting ganglion cells in the central rabbit retina (Buhl & Peichl, 1986; Pu & Amthor, 1990; Ackert et al, 2006) (Fig. 1A). Consistent with their ON physiology, the dendrites of ON DS ganglion cells monostratified within sublamina-b of the IPL (Fig. 1B). As we have reported previously (Ackert et al. 2006), the ON DS ganglion cells in the rabbit retina were tracer coupled to a dense array of small, round cell bodies, < 10 μm in diameter that lay in the inner nuclear layer (INL). These amacrine cells displayed two distinct systems of processes characteristic of polyaxonal cells (Dacey, 1989; Famiglietti, 1992a,b,c; Völgyi et al. 2001). These included a dendritic field of highly branched and relatively thick processes that terminated proximally (Fig. 1A). The other system was formed by an extensive arbour of multiple thin axons that showed little or no branching). A total of 62 ON DS ganglion cells were recorded and labelled in this study.

Figure 1. Morphology and tracer-coupling pattern of ON DS ganglion cells.

A, drawing of a Neurobiotin-labelled ON DS ganglion cell (black) and the somata of tracer-coupled amacrine cells (red). The morphology of one coupled amacrine cell is drawn in detail (blue). Long, straight axon-like processes can be seen extending beyond the field of view (AC axons). In addition, the amacrine cells maintain curvy dendritic processes that end proximally. Scale bar, 25 μm. These morphological features indicate that the amacrine cells are polyaxonal cells with dendritic and axonal systems. B, confocal vertical section showing the level of dendritic stratification of ON DS cell dendrites. Immunolabelling for ChAT (green) shows the two bands of starburst amacrine cell processes defining sublaminae-a and -b of the IPL. The dendrites of the ON DS cell monostratify along the ChAT band in sublamina-b.

Picrotoxin unmasks an OFF response in ON DS cells

A GABAergic inhibition induced by null stimulation is thought to underlie the direction selectivity of DS cells (Caldwell et al. 1978; Kittila & Massey, 1995). Previous studies have shown that application of the non-selective GABA receptor blocker picrotoxin (PTX) increased the light-evoked and spontaneous activity of ON DS cells and abolished their direction selectivity to moving light stimuli (Caldwell et al. 1978; Kittila & Massey, 1995; Ackert et al. 2006). However, we found that, at concentrations of 50–100 μm, PTX had another clear effect on the light-evoked responses of ON DS cells. Application of PTX unmasked a robust OFF response at the termination of a steady spot of light for 58 ON DS ganglion cells tested in this study (Fig. 2A and B). The average latency of the unmasked OFF response (73 ± 8 ms) from light offset was almost identical to that of the ON response (72 ± 12 ms) from light onset for these cells.

Figure 2. Application of PTX unmasks an OFF response in ON DS ganglion cells.

A, typical response of an ON DS cell to a focal light stimulus (275 μm diameter) under control superfusate conditions. Trace below record shows onset and offset of the light stimulus. B, application of PTX (50 μm) reveals a robust OFF response at stimulus offset. C, application of PTX (50 μm) and l-AP4 (50 μm) blocks the ON response, but not the OFF response, indicating that the latter is generated by the OFF retinal pathway.

Overall, we found that 25 μm PTX was insufficient to unmask the OFF response in ON DS cells (n= 22). However, 50 μm PTX unmasked the OFF response in 83% of cells (n= 50), whereas 100 μm unmasked the OFF response in nearly all cells (98%, n= 18). At a concentration of 50 μm, picrotoxin has been shown to activate glycine receptors in retinal neurons (Wang & Slaughter, 2005; Li & Slaughter, 2007), suggesting that at least part of our results may have reflected blockade of glycinergic inhibition. However, we found that application of strychnine in concentrations up to 5 μm never unmasked an OFF response in ON DS ganglion cells (data not shown).

To determine whether the unmasked OFF response was generated by the OFF pathways, we applied the mGluR6 agonist l-AP4, to block the ON pathway (Slaughter & Miller, 1981), together with PTX. Whereas l-AP4 (50 μm) eliminated the ON response of ON DS cells, the unmasked OFF response remained (Fig. 2C). This result indicates that the OFF response is indeed generated by the OFF pathway and that it is not a surround-mediated response of the ON pathway.

We also varied the duration of the light stimulus to determine whether the OFF response was truly evoked by the termination of the stimulus and did not occur at a fixed latency from its onset (n= 5). As illustrated in Fig. 3A, the OFF response unmasked by PTX occurred at stimulus offset, irrespective of its duration. The average latencies for the ON and OFF responses of this cell were 67 ± 7 ms and 70 ± 9 ms measured from the onset and offset of the four stimuli, respectively.

Figure 3. Latencies and area summations of the ON and OFF response of ON DS ganglion cells.

A, responses of an ON DS cell after application of PTX to unmask the OFF response. Onset and offset of the full-field light stimulus is shown below each record. Descending down the figure, the response of the cell is shown to stimuli with increasing duration of 1.5, 4.5, 7.0 and 11.0 s. The OFF response always occurs at light offset, irrespective of the stimulus duration, indicating that its latency is not linked to the time of stimulus onset. This result is consistent with idea that the unmasked OFF response is generated by the OFF retinal pathway. B, area summation profiles of the ON and OFF responses to concentric spots of light of increasing diameters (75, 150, 275, 425, 750, 1250 and 1750 μm), but constant luminance (2.5 Rh* rod⁻¹ s⁻¹). Each point is the average of 5 cells; bars indicate standard errors of the mean.

Finally, we computed the area summation profiles of the ON and OFF responses by evoking them with concentric spots of light with changing diameter, but constant intensity (n= 5) (Fig. 3B). Both the ON and OFF responses showed the greatest amplitude to a 275 μm-diameter spot of light. Larger diameter spots produced an attenuated response, presumably due to activation of the antagonistic surround. The similar profiles generated by the ON and OFF responses support our conclusion that they both reflect centre receptive field mechanisms.

The unmasked OFF response is direction selective

To further evaluate the properties of the OFF response unmasked by PTX, we presented a moving light stimulus classically used to evoke the direction selective response of ON DS ganglion cells. In initial experiments, we used a 50 μm wide/275 μm long rectangular slit of light moved across its minor axis along four axes (8 directions) (see Methods). We found that the ON DS cells (n= 22) could be divided into three groups based on their selectivity for movement in the general dorsal, ventral and nasal directions (Fig. 4A), consistent with those reported by Oyster (1968) showing selectivity principally along these same directions. Further, we found that the average summed vector of direction selectivity for the ON response for each group was tilted from the cardinal axes, again consistent with the findings of Oyster (1968). Since the selectivity of all ON DS cells fell generally along three principal directions, in most subsequent experiments we limited movement of the slit to the two cardinal axes (dorsal–ventral and temporal–nasal) corresponding to four directions of movement, which was adequate to characterize their direction selectivity. Figure 4B shows the typical response of an ON DS cells to a rectangular slit of light moved slowly across the retinal surface. Under control conditions, the ON response of the cell was selective for stimulus movement in the temporal to nasal direction, whereas movement along the dorsal–ventral axis was non-selective. For this cell, then, the preferred direction of movement that produced a robust response was temporal to nasal, while the null direction was the opposite nasal to temporal. The selectivity ratio (see Methods) computed for this cell was 5.3 ± 0.9 (for 20 sweeps). Application of PTX (50 μm) had two clear effects on the response of the DS cell. First, as expected, the direction selectivity of the ON response was abolished as there was now equal activity for temporal-to-nasal and nasal-to-temporal movement of the rectangular slit of light; selectivity ratio of 1.0 ± 0.1 (for 20 sweeps). Second, an OFF response was unmasked, corresponding to the trailing edge of the moving slit of light passing across the ON DS cell receptive field. Moreover, the OFF response was direction selective, but its preferred direction was opposite to that of the ON response seen under control conditions (selectivity ratio of 5.7 ± 0.4 for 20 sweeps). Application of l-AP4 blocked the ON response and thus revealed the direction selectivity of the isolated OFF response more clearly.

We moved the slit along four axes (8 directions) to accurately measure the axis of the direction selectivity of the OFF response (n= 22). As shown in the polar plot in Fig. 4A, the average summed vector of the OFF responses of ON DS cells showed selectivity slightly tilted from the dorsal, ventral, or temporal directions. Moreover, the vector of the direction selectivity of the OFF response was clearly opposite to that of the ON response of ON DS cells. This was true for individual cells as well as the average vector calculations across the entire population.

A selectivity ratio was computed for the ON and OFF responses of 22 ON DS ganglion cells. Under control conditions, spike activity in response to stimulus movement in the preferred direction was 6.4 ± 0.9 times greater than that in response to movement in the null direction (Fig. 5A). As expected, movement along the non-preferred, orthogonal axis was 1.3 ± 0.1 under control conditions. Application of PTX clearly abolished the direction selectivity of the ON responses, resulting in a selectivity ratio of 1.1 ± 0.1 for stimulus movement along the preferred/null axis, similar to that found for the non-preferred orthogonal axis (Fig. 5B). In contrast, the OFF responses unmasked by PTX showed a selectivity ratio of 5.5 ± 0.7, comparable to that of the ON responses under control conditions (Fig. 5B). Movement along the non-preferred orthogonal axis under control conditions was 1.2 ± 0.2 for the OFF response (Fig. 5B). Thus, while the direction selectivity of the ON and OFF responses of individual cells was always opposite, they showed a similar robust difference in the spike count evoked by stimulus movement in their respective preferred and null directions.

Figure 5. Direction selectivity ratios of the ON and OFF responses of ON DS ganglion cells.

A, bar graph showing the average direction selectivity ratio of ON DS cells under control conditions. These cells show a clear direction selectivity along the preferred/null axis for stimulus movement, but no selectivity along the orthogonal axis. Error bars indicate standard errors of the mean. B, left, bar graph showing the average selectivity ratio for the ON response of the same cohort of cells as in A during PTX application. PTX clearly abolishes the direction selectivity of the ON response. B, right, bar graph showing the average selectivity ratio for the OFF response of the same cohort of ON DS cells unmasked by PTX. The selectivity ratio of the OFF response is comparable to that shown by the ON response for stimulus movement along the preferred/null axis. As for the ON response, the OFF response shows no significant direction selectivity along the orthogonal axis.

A role for cholinergic circuitry in OFF response direction selectivity

As mentioned, a GABAergic inhibition, likely to be from presynaptic starburst amacrine cells, is thought to provide the null inhibition essential for the direction selective responses of DS ganglion cells (Caldwell et al. 1978; Kittila & Massey, 1985). Clearly, this scenario cannot hold for the OFF responses, as they were unmasked by the blockade of GABAergic inhibition with PTX. Starburst cells also release acetylcholine and it has been proposed that asymmetric nicotinic input to DS cells also plays a role in generating the direction selectivity of DS cell responses (Vaney, 1990; Borg-Graham & Grzywacz, 1991; Grzywacz et al. 1997, 1998). We therefore examined the effects of cholinergic drugs on the direction selective properties of the OFF responses unmasked by PTX.

In control experiments, we found that application of the nicotinic acetylcholine receptor blocker hexamethonium (Hex; 50 μm) reduced spiking, but had little effect on the selectivity ratio of the ON response of DS cells (5.7 ± 0.8 in control vs. 5.9 ± 1.0 in HEX, n= 8). In contrast, application of Hex abolished the direction selectivity of the unmasked OFF responses as indicated by a reduction of the selectivity ratio to about 1.0 (5.7 ± 2.1 under PTX vs. 0.9 ± 0.1 under PTX and Hex, n= 8) (Fig. 6A). We found that application of the cholinesterase inhibitor neostigmine (Neo; 5 μm) produced a slight increase in light-evoked spike activity, but had no significant effect on the selectivity ratio of the ON responses of DS ganglion cells (6.3 ± 0.6 under control conditions vs. 6.1 ± 0.6 under Neo, n= 9). However, at this concentration, Neo reduced the selectivity ratio of the OFF responses to nearly 1.0, indicating an abolition of its direction selectivity (6.6 ± 0.8 under PTX vs. 0.9 ± 0.2 under PTX and Neo, n= 9) (Fig. 6B). Taken together, these data suggest that cholinergic circuitry plays a crucial role in maintaining the direction selective properties of the OFF responses in ON DS ganglion cells.

Figure 6. Effect of cholinergic drugs on the direction selectivity of the OFF response of ON DS ganglion cells.

A, ON response of an ON DS cell to a 50 μm wide slit moving along the horizontal axis. The cell prefers movement in the temporal to nasal direction. Application of PTX (50 μm) and l-AP4 (50 μm) unmasks and isolates the OFF response in the ON DS cell. The OFF response shows direction selectivity to the moving slit of light, but its preferred direction is nasal to temporal, opposite of that of the ON response. The addition of hexamethonium (Hex) abolishes the direction selectivity of the OFF response, resulting in similar responses to stimulus movement in either direction along the horizontal axis. Slit speed was approximately 80 μm s⁻¹. B, ON response of another ON DS cell to a 50 μm wide slit moving along the vertical axis. The cell prefers movement in the upward direction. Application of PTX (50 μm) and l-AP4 (50 μm) unmasks and isolates the OFF response in the cell. The OFF response shows direction selectivity to the moving slit of light, but its preferred direction is downward, opposite of that of the ON response. The addition of neostigmine (Neo) abolishes the direction selectivity of the OFF response, resulting in similar responses to stimulus movement in either direction along the horizontal axis. Slit speed was approximately 100 μm s⁻¹.

Possible role of gap junctions in delivering OFF signals to ON DS ganglion cells

As shown in Fig. 1C, the dendrites of ON DS ganglion cells are restricted to sublamina-b of the IPL and are thus positioned to receive excitatory input from ON bipolar cells, but not from OFF bipolar cells whose axons terminate exclusively in sublamina-a (Nelson et al. 1978). This morphology raises an important question: how do ON DS cells receive excitatory signals from the OFF pathway? It has been reported that, on occasion, some dendrites of ON DS cells can extend into sublamina-a (Buhl & Peichl, 1986; He & Masland, 1998). We therefore carefully examined the dendritic structure of five ON DS cells that showed an OFF response unmasked by PTX. We found that all dendrites of these cells were restricted to sublamina-b and never crossed into the distal sublamina-a of the IPL. Figure 1B illustrates the restricted dendritic arborization of one of these ON DS cells.

Since direct chemically meditated synaptic inputs from OFF bipolar cell axons in sublamina-a can apparently be ruled out, an alternative explanation is that the gap junctions between ON DS cells and polyaxonal amacrine cells (Fig. 1) may form the conduit for the OFF signals. To test this hypothesis, we used 18β-glycyrrhetinic acid (18β-GA; 25 μm) to disrupt the gap junctions (Davidson & Baumgarten, 1988). In initial experiments, we assessed effects of 18β-GA on the coupling of ON DS cells by fixing the retina at different time points after application of the drug. We found that it took approximately 20 min of exposure of 18β-GA to completely uncouple ON DS ganglions from neighbouring polyaxonal amacrine cells (Fig. 7A). In control experiments, we found that application of 18β-GA for up to 30 min had no significant effect on the ON response of ON DS cells to stationary spots of light (n= 10) (Fig. 7B). In contrast, 18β-GA completely abolished the OFF response unmasked by PTX and this blockade was reversed following washout of the gap junction blocker (n= 12) (Fig. 7B). In subsequent experiments, we tested the effects of 18β-GA on the OFF response to moving light stimuli (Fig. 8). Blockade of the gap junctions reversibly abolished the OFF response to a moving slit of light in both the preferred and null directions (n= 8). Consistent with the timing of the uncoupling action of 18β-GA on ON DS cell gap junctions, we found that it took approximately 20 min of exposure for the drug to completely block the OFF response.

Figure 7. 18β-Glycyrrhetinic acid eliminates the OFF response in ON DS ganglion cells to a stationary light stimulus.

A, photomicrograph of an ON DS cell labelled with Neurobiotin after 20 min incubation in 18β-glycyrrhetinic acid (18β-GA; 25 μm). Tracer coupling to neighbouring amacrine cells is completely eliminated. Scale bar, 100 μm. B, response of an ON DS cell to full-field illumination under control conditions. Subsequent application of the gap junction blocker has no effect on the ON response of the ON DS cell. Light trace below shows onset and offset of the light stimulus. C, response of an ON DS cell to full-field illumination under control superfusate conditions. Subsequent application of PTX (50 μm) and l-AP4 (50 μm) unmasks an OFF response and eliminates the ON response. Application of 18β-GA (25 μm) completely abolishes the OFF response. The effect of 18β-GA is reversed upon washout. The onset and offset of the full-field light stimulus are shown by the light trace at bottom of the panel.

Figure 8. 18β-Glycyrrhetinic acid eliminates the OFF response in ON DS ganglion cells to a moving light stimulus.

A, ON response of an ON DS cell to a 50 μm wide slit moving along the vertical axis at approximately 180 μm s⁻¹. The cell prefers movement in the downward direction. B, application of PTX (50 μm) and l-AP4 (50 μm) unmasks and isolates the OFF response in the ON DS cell. The OFF response shows direction selectivity to the moving slit of light, but its preferred direction is upward. C, application of 18β-GA completely eliminates the OFF response to the moving slit of light. D, the direction selective OFF response is completely recovered after washout of 18β-GA.

Polyaxonal amacrine cells coupled to DS cells form a substrate for possible delivery of OFF signals

Our pharmacological results indicate that gap junctions are critical to the delivery of excitatory OFF signals to ON DS ganglion cells. Gap junctions exist at all levels of the retina and are blocked during application of 18β-GA, but it is most likely that the OFF signals are received directly via electrical synapses made with neighbouring polyaxonal amacrine cells. If so, these polyaxonal cells must send processes to sublamina-a of the IPL so that they can receive excitatory OFF signals, which are then passed on to the DS cells via the gap junctions in sublamina-b.

To test this idea, we examined the dendritic/axonal morphology of the polyaxonal cells that were tracer coupled to Neurobiotin-injected DS ganglion cells. Figure 9 shows a series of confocal images in which polyaxonal cells were labelled with Neurobiotin that crossed the gap junctions with injected ON DS cells. In addition, cholinergic starburst amacrines cells were labelled with an antibody against ChAT (Fig. 9A) and the ribbon synapses in bipolar cell axon terminals were also labelled using an antibody against the ribbon protein Ribeye (Fig. 9B and C). The somata of polyaxonal amacrine cells lay in the proximal region of the INL and emitted both axonal and dendritic processes that entered the IPL (Fig. 9). The amacrine cell dendritic processes were found to costratify with many synaptic ribbons found in the axon terminals of OFF bipolar cells in sublamina-a (Fig. 9B and C). In addition, the processes also were found to cofasciculate with the dendrites of OFF starburst amacrine cells in sublamina-a (Fig. 9D and E). The processes then descended to sublamina-b, where they cofasciculated with the dendrites of the ON starburst amacrine cells and formed gap junctions with the dendrites of ON DS ganglion cells (Fig. 9F). These results thus suggest that a morphological substrate exists for polyaxonal amacrine cells to receive excitatory OFF signals from OFF bipolar cells (glutamatergic) and OFF starburst cells (cholinergic) and to transfer them to ON DS ganglion cells via gap junctions made in sublamina-b of the IPL.

Figure 9. Amacrine cells coupled to ON DS ganglion cells cofasciculate with OFF bipolar and starburst amacrine cell dendrites in sublamina-a of the IPL.

A, a 1 μm-thick confocal section showing somata of polyaxonal amacrine cells (red) in the proximal INL that are coupled to a Neurobiotin-injected ON DS ganglion cell. The somata of starburst-a amacrine cells are also visible (green) following immunolabelling for ChAT. Scale bar, 20 μm. B, dendrites of polyaxonal amacrine cells cofasciculate with the synaptic ribbons of OFF bipolar cell axon terminals immunolabelled for Ribeye in sublamina-a of the IPL. Scale bar, 20 μm. C, higher magnification of the micrograph in B. Scale bar, 10 μm. D, dendrites of polyaxonal amacrine cells (red) cofasciculate with the dendrites of starburst-a amacrine cells in sublamina-a of the IPL. Scale bar, 20 μm. E, magnified view of the micrograph in D. Scale bar, 10 μm. F, dendrites (arrowheads) of polyaxonal amacrine cells then descend to sublamina-b and costratify with dendrites of ON DS ganglion cell (arrow) and dendrites of starburst-b amacrine cells (green). Scale bar, 20 μm.

To confirm that the cofasciculation of stained profiles was indicative of synaptic contacts and not simply due to chance, we analysed the confocal images in which one channel was rotated by 90 deg using Photoshop software. The rationale for this approach was that if the cofascilulation of stained profiles indicated synaptic contacts, then rotation of one channel would produce a significant reduction in the juxapositions. For the image in Fig. 9E showing colocalization of amacrine cell dendrites and ChAT label, we found that rotation of the green (ChAT) channel resulted in an 85% reduction in the number of pixels showing colocalization of label. For the image in Fig. 9C showing the cofasciculation of amacrine cell dendrites and Ribeye protein, we rotated the latter blue channel by 90 deg. There was a 30% reduction in pixels showing both red (amacrine cell dendrite) and blue (Ribeye) label after rotation. Moreover, after rotation there was a 54% reduction in the number of adjacent pixels in which one showed red label and the other blue label. Overall, these analyses support the idea that cofasciculation of labels seen in the confocal images was not due to chance and therefore was likely to be indicative of synaptic contacts.

Discussion

Since the work of Hartline (1938), the segregation of visual information into two separate and parallel ON and OFF streams has been accepted as a principal organizing feature of the vertebrate visual system. It was thus surprising to find that DS ganglion cells with classic ON receptive field physiology also maintain robust OFF responses when released from GABAergic inhibition. This finding indicates not only a mixing of excitatory ON and OFF signals in ON DS cells, but that the latter is masked by inhibitory circuitry, at least under our experimental conditions, so that it is not expressed in the spike code sent to the brain. Interestingly, mixing of ON and OFF signals at the ganglion cell level in which one signal is masked has recently been described in a number of retinas. Geffen et al. (2007) showed that inhibitory circuitry provides a switch to control which of the two synaptic pathways provides the input signal to ganglion cells in salamander. Dependent on background stimulus conditions, these cells switched their responses from OFF to ON. In rabbit, Roska & Werblin (2001) reported that an OFF excitation was revealed in two ON ganglion cell subtypes after GABAergic feedback in the IPL was blocked. Similarly, recent work in our lab has shown that OFF alpha ganglion cells display a robust ON response when GABAergic inhibition is removed (Farajian & Bloomfield, 2008).

Recently, a late, secondary response to the offset of a moving slit of light has been reported in ON DS cells in the mouse retina (Sun et al. 2006). Interestingly, Ariel & Daw (1982) also illustrated an emerging OFF response in rabbit ON DS cells following application of PTX, but they saw this effect in only one of five cells tested and so they attributed it to technical artifact. However, since they applied drugs arterially, the concentration at the retina could not be determined or controlled. Our finding that the unmasking of the robust OFF response by PTX was concentration dependent may explain why Ariel & Daw did not see this robust response in more cells.

Taken together with these previous studies, our results indicate that excitatory ON and OFF signals are not as segregated at the ganglion cell level as widely believed. At first glance, this mixing of ON and OFF signals would appear to conflict with the sublamination scheme, in which excitatory inputs are derived from ON and OFF bipolar cell axon terminals that lay at different levels of the IPL (Famiglietti & Kolb, 1976; Bloomfield & Miller, 1986). That is, how can the ON DS cells whose dendrites are restricted to sublamina-b of the IPL receive excitatory OFF signals that are segregated to sublamina-a? Our finding that 18β-GA reversibly blocked the unmasked OFF response suggests that gap junctions form the conduit for OFF signal transmission.

Like other gap junction blockers, however, 18β-GA has been shown to affect the conductances of a number of ion channels, including non-selective cation and voltage-dependent potassium conductances (Matchkov et al. 2004; Guan et al. 2007). These findings raise the possibility that the effects of 18β-GA on the OFF response of ON DS cells could, at least in part, reflect non-specific actions on upstream ionic channels. However, several findings argue against this possibility. First, we found that whereas 18β-GA blocked the OFF responses of ON DS cells, at the same time it had no affect on the ON responses. If these actions reflected non-specific effects, they would need to be on the OFF pathway, but not the ON pathway. However, we showed recently that PTX unmasks an ON response in OFF α ganglion cells and that this ON response, but not the OFF response, can be blocked by 18β-GA (Farajian and Bloomfield, 2008). In both cases, 18β-GA blocked the unmasked response and not the ‘normal’ light-evoked response, irrespective of whether they were ON or OFF. These data are thus inconsistent with 18β-GA having non-specific effects limited to the ON (or OFF) pathway. Second, we found that it took approximately 20 min of 18β-GA exposure to both completely uncouple ON DS cells and block the OFF response. In contrast, we found that drugs that affected receptors (e.g. picrotoxin and l-AP4) showed maximal effects within 5 min on the same retinas. Likewise, a recent report of the non-specific effects of 18β-GA on vascular tissue showed that actions on ionic conductances occurred within 3–5 min of application (Guan et al. 2007), much quicker than the exposure time we found necessary to block the OFF response. Taken together, these data strongly suggest that the blockade of the unmasked OFF response of ON DS cells by 18β-GA reflected actions directed specifically at uncoupling gap junctions.

We posit that the gap junctions formed between ON DS ganglion cells and polyaxonal amacrine cells form the conduit for OFF signal transmission. Consistent with this idea, we found that dendrites of these coupled amacrine cells cofasciculate with the synaptic terminals of OFF bipolar cells and dendrites of OFF starburst amacrine cells, providing a morphological substrate for the reception of excitatory OFF signals. In this scheme, the multistratified polyaxonal cells receive excitatory OFF signals through chemical synapses with OFF bipolar cell axons and starburst cell dendrites within sublamina-a of the IPL and then deliver them to ON DS cells via gap junctions in sublamina-b.

The finding that the three subtypes of ON DS cells also showed OFF responses with different directional preference suggests that there is selectivity in the polyaxonal amacrine cells to which they are coupled. That is, ON DS cells with different directional preferences, for both the ON and OFF responses, may form gap junctions with different cohorts of polyaxonal cells. Our previous finding that ON DS cells with the same directional preference tend to cluster into neighbouring groups forms a morphological substrate for this segregated coupling pattern (Ackert et al. 2006).

A new role for gap junctions in retina

A number of functional roles have been suggested for gap junctions in the retina (reviewed by Bloomfield & Völgyi, 2008) and our results suggest a novel one, namely to act as conduits for the mixing of ON and OFF signals in ganglion cell subtypes. In this scheme, electrical synapses provide for the transfer of excitatory signals across the sublaminae of the IPL. Interestingly, the ON ganglion cell subtypes observed by Roska & Werblin (2001) to show masked OFF responses maintained dendrites that stratified at the border of sublaminae-a and -b and thus could potentially receive inputs from both ON and OFF bipolar cells. Taken together, these findings suggests that mixing of excitatory ON and OFF signals in ganglion cells may utilize chemical or electrical synapses as conduits, depending on the level at which their dendrites stratify in the IPL.

We recently showed that gap junctions between ON DS ganglion cells and polyaxonal amacrine cells also provide for synchronized activity of the former (Ackert et al. 2006). This synchrony is controlled by stimulus motion whereby null movement activates GABAergic inputs that effectively shunt intercellular currents that move through the gap junctions. A similar mechanism in which inhibitory chemical synapses ‘uncouple’ cells by short-circuiting the spread of intercellular current has been reported in the glomeruli of the inferior olive (Llinás et al. 1974; Sotelo et al. 1974) and the pharyngeal expansion motoneurons in the mollusk Navanax (Spira et al. 1980). We posit that this mechanism is also employed to mask the OFF inputs to ON DS cells, whereby GABAergic inhibition shunts the gap junctional currents presynaptically and/or postsynaptically. However, the GABAergic null inhibition that controls synchrony must be different from the inhibition that masks the OFF responses, since null stimulus movement never evoked an OFF response under control conditions. Whether these two GABAergic inputs are juxtaposed to and control different subsets of gap junctions between ON DS cells and polyaxonal amacrine cells is at present unclear.

Direction selectivity of OFF responses

An intriguing finding was that the unmasked OFF response was direction selective, but with preference opposite to that of the ON response. Several previous studies reported data consistent with this opposite preference. Iwakabe et al. (1997) showed that optokinetic nystagmus (OKN) is depressed in the mGluR6 knockout mouse retina, in which ON responses are abolished, but the surviving nystagmus, presumably activated by the OFF system, displayed an increased preference for reversed directional movement. In chicken, interactions between the ON and OFF channels have been described where each of the pathways generates an OKN with opposite directional preference (Bonaventure et al. 1992). These results suggest that the directional OFF response has the capability to influence optokinetic signals, at least when the ON response component is blocked. Studies of rabbit and turtle retinas indicate that after application of PTX, up to 50% of ON–OFF DS ganglion cells still show direction selective responses, but often with reverse preference (Smith et al. 1996; Grzywacz et al. 1997). PTX-induced reversals in the directional preference of neurons have also been reported in invertebrate eyes (Bülthoff & Bülthoff, 1987; Öğmen, 1991). Curiously, Pan & Slaughter (1991) showed that activation of GABA_B receptors could bring out direction selective responses in a number of amacrine and ganglion cells in the salamander retina. Thus, masked DS inputs, some with directional preference opposite to that of the main synaptic drive, appear to be present in cells within a wide range of species.

A role for acetylcholine in direction selectivity

Over 40 years ago, Barlow & Levick (1965) theorized that an asymmetric inhibition dependent on stimulus motion underlies the direction selectivity of DS cells. There is now compelling evidence that this mechanism reflects asymmetric GABA release from presynaptic starburst amacrine cells (Wyatt & Daw, 1975; Caldwell & Daw, 1978; Kittila & Massey, 1995; Ackert et al. 2006). In addition, a spatially asymmetric excitatory mechanism has been suggested in which acetylcholine release from starburst amacrine cells imparts direction selectivity to DS cells (Vaney, 1990; Borg-Graham & Grzywacz, 1991; Grzywacz et al. 1997; 1998; Fried et al. 2005). However, there are contradictory experimental data as to the role of acetylcholine, if any, in generating direction selective signals in the retina (reviewed by Vaney & Taylor, 2002; Demb, 2007).

The present results provide further evidence that cholinergic circuitry may be important for the generation of direction selective responses. Clearly, the direction selectivity of the OFF response of ON DS cells cannot be due to a GABAergic mechanism in that it was not unmasked until the GABAergic inhibition was abolished. Instead we found that the direction selectivity of the OFF response was sensitive to the nicotinic receptor blocker hexamethonium and the cholinesterase inhibitor neostigmine. It would be expected that the drugs have opposite effects in that hexamethonium blocks the actions of acetylcholine at the receptor site, whereas neostigmine increases the transmitter concentration and thereby enhances its actions. However, while we did find that the drugs had opposite effects on spontaneous activity, they both effectively abolished the direction selectivity of the OFF response. We posit that both these drugs, though with opposite actions, produced an imbalance of acetylcholine activity from some basal and/or light-evoked level, either by increasing transmitter concentration or blocking actions at the receptor level. Both light-dependent and -independent acetylcholine release has been measured in the rabbit retina (Masland et al. 1984).

Interestingly, Ariel & Daw (1982) reported that application of the cholinesterase inhibitor physostigmine could unmask an OFF response in ON DS cells that showed some direction selectivity, either with the same or opposite preference of the ON response. We were unable to replicate this result in the present study using neostigmine. One possible explanation for this discrepancy, other than the use of different inhibitors, is that the concentration of drugs was different. As aforementioned, Ariel & Daw (1982) applied drugs arterially in their in vivo experiments as a single, concentrated bolus and so the concentration at the retina was initially high and slowly declined over time. The concentration of drug at the retina, which could not be determined (M. Ariel, personal communication), may have been higher than the 5 μm neostigmine we applied. In any event, we found that at a concentration sufficient to block the direction selectivity of the unmasked OFF response, neostigmine had no significant effect on the ON response. These data suggest that the circuitry generating the direction selectivity of the OFF response is different from that for the ON response. This may explain why the preferred directions for the ON and OFF responses of individual cells were always opposite. Interestingly, Fried et al. (2005) reported a direct cholinergic effect on the OFF response component, but not the ON response of ON–OFF DS ganglion cells, suggesting a similarity in the cholinergic circuits innervating the different subtypes of DS cells.

Function of the masked OFF response

An important question related to the function of the OFF response is why is it masked by a GABAergic inhibition? Considerable work in the CNS indicates that synaptic inputs can be masked under certain conditions and it has been posited that rapid unmasking of inputs reflects the normal dynamics of signalling in the brain (Meltzler & Marks, 1979; Merzenich et al. 1983a,b; Jacobs & Donoghue, 1991; Gilbert & Wiesel, 1992). As mentioned, Geffen et al. (2007) showed that changes in background stimulus conditions can unmask ON responses in certain OFF ganglion cells in the salamander retina. We posit that, under certain stimulus conditions, the GABAergic inhibition is relieved so that the OFF signals are incorporated in the spike code of ON DS cells to affect optokinetic signalling. In support of this idea, Oyster (1968) reported that while ON DS cells responded vigorously to movement in the preferred direction for most stimuli, they did display a ‘paradoxical response’ in which robust activity was evoked by null direction movement of stimuli with high spatial frequencies. Clearly, OFF signals with opposite direction sensitivity could underlie this paradoxical response evoked by certain stimuli.

Glossary

Abbreviations

DS

direction selective

ECA

effective collecting area; 18β-GA; 18β-glycyrrhetinic acid

GCL

ganglion cell layer

INL

inner nuclear layer

PTX

picrotoxin

Author contributions

J.A.: conception, design, analysis and interpretation of data; drafting and revising of article; final approval of version to be published. R.F.: design, analysis and interpretation of data; revision of the article; final approval of version to be published. B.V.: design, analysis and interpretation of data; revision of the article; final approval of version to be published. S.B.: conception, design, analysis and interpretation of data; drafting and revising of article; final approval of version to be published.