Light-evoked current responses in rod bipolar cells, cone depolarizing bipolar cells and all amacrine cells in dark-adapted mouse retina

Abstract

Light-evoked excitatory cation current (ΔI_C) and inhibitory chloride current (ΔI_Cl) of rod and cone depolarizing bipolar cells (DBC_Rs and DBC_Cs) and AII amacrine cells (AIIACs) in dark-adapted mouse retinal slices were studied by whole-cell voltage-clamp recording techniques, and the cell morphology was revealed by Lucifer yellow fluorescence with a confocal microscope. ΔI_C of all DBC_Rs exhibited similar high sensitivity to 500 nm light, but two patterns of ΔI_Cl were observed in DBC_Rs with slightly different axon morphology. At least two types of DBC_Cs were identified: one with axon terminals ramified in 70–85% of the depth of the inner plexiform layer (IPL) and DBC_R-like ΔI_C sensitivity, whereas the other with axon terminals ramified in 55–75% of IPL depth and much lower ΔI_C sensitivity. The relative rod/cone inputs to DBCs and AIIACs were analysed by comparing the ΔI_C and ΔI_Cl thresholds and dynamic ranges with the corresponding values of rods and cones. On average, the sensitivity of a DBC_R to the 500 nm light is about 20 times higher than that of a rod. The sensitivity of an AIIAC is more than 1000 times higher than that of a rod, suggesting that AIIAC responses are pooled through a coupled network of about 40 AIIACs. Interactions of rod and cone signals in dark-adapted mouse retina appear asymmetrical: rod signals spread into the cone system more efficiently than cone signals into the rod system. The mouse synaptic circuitry allows small rod signals to be highly amplified, and effectively transmitted to the cone system via rod–cone and AIIAC–DBC_C coupling.

In the visual system, rod photoreceptors register dim light signals, and cone photoreceptors encode brighter light signals (Hecht et al. 1942; Dowling, 1987). Anatomical studies have shown that mammalian rods make output synaptic contacts with only one type of bipolar cell, the depolarizing rod bipolar cells (DBC_Rs), whereas cones make synaptic contacts with 8–9 types of cone bipolar cells, some of which depolarize while others hyperpolarize to light stimuli (DBC_Cs or HBC_Cs) (Wassle et al. 1991; Wassle & Boycott, 1991). Additionally, rods and cones are electrically coupled to each other (Raviola & Gilula, 1973; Tsukamoto et al. 2001; Deans et al. 2002) and thus there is a certain degree of signal mixing at the photoreceptor level. It is unclear how much such coupling contributes to the light responses of different types of bipolar cells.

Anatomical evidence indicates that DBC_Rs in most mammals do not send synaptic outputs directly to ganglion cells, the output neurones of the retina, but to the AII amacrine cells (AIIACs) which make electrical synapses with DBC_Cs (which send signals to ‘on’ ganglion cells) and inhibitory chemical synapses with HBC_Cs and ‘off’ ganglion cells. Therefore, the rod bipolar cell signals are transmitted to ganglion cells by virtue of the AIIACs that ‘piggyback’ on the cone bipolar cell synaptic circuitry (Kolb & Famiglietti, 1974; Strettoi et al. 1990, 1992, 1994; Tsukamoto et al. 2001). For this reason, the AIIACs serve as communication hubs of the rod and cone signals in the inner mammalian retina. It is of great interest to understand how rod and cone bipolar cell signals are transmitted to AIIACs, and how various electrical and chemical synapses contribute to light-evoked responses of these cells.

Despite the vast amount of anatomical data on synaptic connections between rods, cones, rod and cone bipolar cells, and AIIACs, detailed physiological information on the light responses of these cells in the mammalian retina is still fragmentary. For example, although microelectrode recordings from rabbit and cat retina reveal the voltage response and receptive field properties of DBC_Rs, DBC_Cs and AIIACs in rabbits and cats (Nelson, 1982; Bloomfield et al. 1997; Bloomfield & Xin, 2000), it is not clear about the relative contributions of the excitatory and inhibitory synaptic currents to these voltage responses. Additionally, it is uncertain how individual electrical and chemical synapses identified by anatomical studies affect the light-evoked signals.

In this study, we used the whole-cell voltage-clamp technique to record light-evoked responses from the DBC_Rs, DBC_Cs and AIIACs in the dark-adapted mouse retinal slices. This technique allows us to separate light-evoked excitatory cation current and inhibitory chloride current and to reveal the cell morphology by Lucifer yellow fluorescence. We chose the mouse retina because physiological results obtained in wild-type mice can later be correlated with findings in genetically manipulated mice. Moreover, the cellular and synaptic inputs to these cells have been well characterized at the ultrastructural level (Carter-Dawson & Lavail, 1979; Tsukamoto et al. 2001) and thus physiological findings can be readily correlated with anatomical observations.

Methods

Experimental approach

Our study constitutes a systematic voltage-clamp analysis of rod and cone bipolar and amacrine cell light responses in the dark-adapted mouse retina. The major advantage of this approach is that excitatory and inhibitory current responses can be separated by holding the membrane potential near chloride and cation reversal potentials, respectively. Thus the photoreceptor and amacine cell contributions to the bipolar cells' light responses and the bipolar cell and amacrine cell contributions to the amacrine cells' light responses (i.e. cation (photoreceptor or bipolar cell inputs) and chloride currents (amacrine cell inputs), or ΔI_C and ΔI_Cl, respectively) can be differentially recorded. Since the sensitivity of mouse rod pigment to 500 nm light is about 2 log units higher than that of the M-cone pigment, and about 4 log units higher than that of the S-cone pigment (Lyubarsky et al. 1999), we use the response threshold to 500 nm lights to estimate the relative rod/cone contributions to ΔI_C and ΔI_Cl in each ‘on’ bipolar cell and amacrine cell under dark-adapted conditions (the relative sensitivities of rods, and M- and S-cone pigments to 500 nm lights under light-adapted conditions, on the other hand, are different, and thus the present study is limited to rod/cone contributions to ‘on’ bipolar cells and AIIACs under dark-adapted conditions). Another advantage of this approach is that the cell's morphology can be easily revealed by Lucifer yellow filling with the recording electrode (assisted by a confocal microscope). This allows us to characterize the morphology of each recorded cell and to compare it with results of earlier anatomical studies.

Preparations and light stimulation

The mouse strain used in this study was C57Black6J from Jackson Laboratory (Bar Harbour, ME, USA). All animals were handled in accordance with Baylor College of Medicine's policies on the treatment of laboratory animals. Mice were dark-adapted for 1–2 h prior to the experiment. To maintain the retina in the fully dark-adapted state, all further procedures were performed under infrared illumination with dual-unit Nitemare (BE Meyers, Redmond, WA, USA) infrared scopes. Animals were killed by an i.p. injection of ketamine + xylazine + acepromazine (0.1 ml, 100 mg ml⁻¹) and the eyes were immediately enucleated and placed in oxygenated Ames' medium (Sigma, MO, USA) at room temperature. The dissection and preparation of living retinal slices followed essentially the procedures described in previous publications (Werblin, 1978; Wu, 1987). Oxygenated Ames' solution (adjusted to pH 7.3) was introduced continuously to the recording chamber, and the medium was maintained at 35°C by a temperature control unit (TC 324B, Warner Instruments, CT, USA). All pharmacological agents were dissolved in Ames' medium.

A photostimulator was used to deliver light spots (of diameter 600–1200 μm) to the retina via the epi-illuminator of the microscope. The intensity of unattenuated (logI = 0) 500 nm light was 1.4 × 10⁶ photons μm⁻² s⁻¹. The number of photoisomerizations per rod per second (Rh* rod⁻¹ s⁻¹) was estimated by methods described in a previous publication (Pang et al. 2003). Since we delivered an uncollimated stimulus light beam through an objective lens with large numerical aperture (Zeiss, ×40, 0.75 NA, water immersion), the incident light enters the retinal slice in many directions, and thus the effect of photoreceptor self-screening is minor (Field & Rieke, 2002a). The peak amplitude of light-evoked current responses was plotted against light stimulus intensity, and data points were fitted by the Hill equation:

where R is the current response amplitude, R_max is the maximum response amplitude, σ is the light intensity that elicits a half-maximal response, N is the Hill coefficient, tanh is the hyperbolic tangent function and log is the logarithmic function of base 10. In this article, we use the R–logI plot for our analysis (the right-hand term of the above equation), and for such plots the light intensity span (dynamic range: range of intensity that elicits responses between 5 and 95% of R_max) of a cell equals 2.56/N (Thibos & Werblin, 1978).

Voltage-clamp recordings

Voltage-clamp recordings were made with an Axopatch 200A amplifier connected to a DigiData 1200 interface and pCLAMP 6.1 software (Axon Instruments, Foster City, CA, USA). Whole cell voltage-clamp recordings were made with patch electrodes made with Narishige or Sutter patch electrode pullers that were of 5–7 MΩ tip resistance when filled with internal solution containing 118 mm caesium methanesulphonate, 12 mm CsCl, 5 mm EGTA, 0.5 mm CaCl₂, 4 mm ATP, 0.3 mm GTP, 10 mm Tris, 0.8 mm Lucifer yellow, adjusted to pH 7.2 with CsOH. The chloride equilibrium potential, E_Cl, with this internal solution was about −60 mV. Estimates of the liquid junction potential at the tip of the patch electrode prior to seal formation varied from −9.2 to −9.6 mV (Pang et al. 2002). For simplicity, we corrected all holding potentials by 10 mV. In order to determine the dark membrane potentials of ganglion cells, we measured the zero-current potentials of 12 bipolar cells and 8 amacrine cells (including all cell types described in this paper) with patch electrodes filled with caesium internal solution (above) and with potassium internal solution (Berntson & Taylor, 2000), and found that the zero-current potentials with K⁺ were consistently 12–17 mV more hyperpolarized that with Cs⁺. For cells recorded with only Cs⁺, we corrected the zero-current potential measured in darkness (dark membrane potential) by 15 mV. Since the DBC_Rs and DBC_Cs have relatively narrow dendritic and axonal fields, it is likely that a large portion of the cell membrane was space-clamped. In order to maximize good space-clamping, we selected cells with higher input resistance (> 500 MΩ) when whole cell recording was made. To further verify that DBC processes are space-clamped in our study, we examined the effects of GABA and glycine receptor antagonists (picrotoxin and strychnine) on ΔI_C (recorded at E_Cl) of eight DBC_Rs and nine DBC_Cs and found that these compounds did not alter the amplitude of ΔI_C (but suppressed ΔI_Cl), suggesting that DBC processes (where inhibitory synaptic inputs are located) were clamped at the command potential by the patch electrodes in the soma. In order to select cells with minimum slicing damage, we recorded from cells that were 2–3 cell layers below the surface of the retinal slice. AIIACs are electrically coupled with one another and it is impossible to space-clamp the coupled network. Since the dark membrane potential of AIIACs is close to E_Cl (see Results), we studied the light responses of these cell by measuring ΔI_C near E_Cl.

Visualization of cell morphology

Cell morphology was visualized in retinal slices using Lucifer yellow fluorescence with a confocal microscope (Zeiss 510). Images were acquired with a ×40 water immersion objective (NA = 1.20), using the 458 nm excitation line of an argon laser, and a long pass 505 nm emission filter. Consecutive optical sections were superimposed to form a single image using Zeiss LSM-PC software, and these compressed image stacks were further processed in Adobe Photoshop 6.0 to improve the signal to noise ratio. Since signal intensity values were typically enhanced during processing to improve the visibility of smaller processes, the cell bodies and larger processes of some cells appear saturated due to their larger volume of fluorophore. The level at which dendritic processes stratified in the IPL was characterized in retinal vertical sections by the distance from the processes to the distal margin (0%) of the IPL.

Results

Light-evoked current response characteristics of morphologically identified rod ‘on’ bipolar cells (DBC_Rs)

Under infrared visual guidance in dark-adapted mouse retinal slices, we recorded rod ‘on’ bipolar cells (DBC_Rs) with patch electrodes filled with internal solution and Lucifer yellow in the whole-cell voltage-clamp configuration. DBC_Rs were identified by their characteristic morphology (with globula (large knob-shaped swellings) axon terminals ramifying between 70 and 100% of the IPL depth; Euler & Wassle, 1995; Hartveit, 1997) revealed by Lucifer yellow fluorescent images in retinal slices and by their inward light-evoked cation current (ΔI_C) recorded at E_Cl. Figure 1A shows the stacked confocal fluorescent image of a DBC_R in the retinal slice; it exhibited typical DBC_R morphology with the dendritic field extending about 20 μm horizontally in the outer plexiform layer (OPL), a soma located in the upper half of the inner nuclear layer (INL), an axon extending into the IPL and globular axon terminals stratifying in the inner half of the IPL. The morphology of all 18 DBC_Rs was very similar with the exception that two-thirds (12/18) of them had axon terminal globules distributed in 75–100% of the IPL depth (similar to Fig. 1A, group 1 DBC_R) whereas the remaining one-third (6/18) had axon terminal globules distributed more proximally (70–85% of the IPL depth, Fig. 2A, group 2 DBC_R). The light-evoked current responses of these two groups of DBC_Rs to a 0.5 s 500 nm light step recorded under dark-adapted conditions at various holding potentials are shown in Figs 1B and 2B, and the current–voltage relations in Figs 1C and 2C. The light-evoked current responses of the two groups of DBC_Rs are very similar, with apparent reversal potentials near −40 mV. Both groups exhibited inward ΔI_C (at E_Cl) and outward ΔI_Cl (at E_C). The mean zero-current potential in darkness of these two groups of cells was −59 ± 5 mV and −57 ± 6 mV, respectively (see Methods). Since DBC_Rs in the mammalian retina receive glutamatergic inputs from rods (Qin & Pourcho, 1996; Brandstatter et al. 1998) and GABAergic/glycinergic inputs from the reciprocal rod amacrine cells (Freed et al. 1987; Sandell et al. 1989; Fletcher & Wassle, 1999; Zhang et al. 2002), ΔI_C is likely to be mediated by rod inputs that gate a cation conductance and the ΔI_Cl is probably mediated by reciprocal rod amacrine cell inputs that gate a chloride conductance (Cohen & Miller, 1994). To verify that ΔI_C and ΔI_Cl in DBC_Rs are mediated by different synaptic inputs, we examined the effects of GABA and glycine receptor antagonists (picrotoxin (PTX) and strychnine (STR)) on five group 1 DBC_Rs and three group 2 DBC_Rs and found that these compounds did not significantly alter the amplitude of ΔI_C (peak ΔI_C in 100 μm PTX + 1 μm STR was 1.11 ± 0.23 (n = 6) of peak ΔI_C in control solution, and peak ΔI_C in 100 μm PTX + 1 μm STR was 0.92 ± 0.18 (n = 2) of peak ΔI_C in control solution). This suggests that the rod inputs to DBC_Rs are not significantly (P = 0.68, paired t test) influenced by GABAergic and glycinergic interneurones in the mouse retina.

Figure 1. Group 1 rod depolarizing bipolar cell (DBC_R).

A, stacked confocal fluorescent image of a group 1 DBC_R in a mouse retinal slice. PRL, photoreceptor layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer (0–1: 0–100% of IPL depth); GCL, ganglion cell layer; calibration bar: 20 μm. B, light-evoked current responses to a 0.5 s light step (500 nm, −4.5 = 22 Rh* rod⁻¹ s⁻¹) at various holding potentials (V_H). C, current–voltage relations of the peak light responses in B; the mean reversal potential of 10 cells was −46 ± 6 mV. D, light-evoked excitatory cation current (ΔI_C) recorded at E_Cl to 500 nm light steps (0.5 s) of various intensities. E, light-evoked inhibitory chloride current (ΔI_Cl) recorded at E_C to 500 nm light steps (0.5 s) of various intensities. F, response–intensity relations of the light-evoked cation and chloride currents: ○, ΔI_C–logI; •, ΔI_Cl–logI; error bars represent ± s.d. The mean threshold of ΔI_C was −6.8 (0.1 Rh* rod⁻¹ s⁻¹) and of ΔI_Cl was −7.2 (0.044 Rh* rod⁻¹ s⁻¹), and the mean dynamic range of ΔI_C was 3.12 log units, and of ΔI_Cl was 2.28 log units.

Figure 2. Group 2 DBC_R.

A, stacked confocal fluorescent image of a group 2 DBC_R in a mouse retinal slice. Abbreviations as for Fig. 1. B, light-evoked current responses to a 0.5 s light step (500 nm, −4.5 = 22 Rh* rod⁻¹ s⁻¹) at various holding potentials. C, current–voltage relations of the peak light responses in B; the mean reversal potential of 5 cells was −48 ± 7 mV. D, light-evoked excitatory cation current (ΔI_C) recorded at E_Cl to 500 nm light steps (0.5 s) of various intensities. E, light-evoked inhibitory chloride current (ΔI_Cl) recorded at E_C to 500 nm light steps (0.5 s) of various intensities. F, response–intensity relations of the light-evoked cation and chloride currents: ○, ΔI_C–logI; •, ΔI_Cl–logI; error bars represent ± s.d.. The mean threshold of ΔI_C was −6.8 (0.1 Rh* rod⁻¹ s⁻¹) and of ΔI_Cl was −7.6 (0.018 Rh* rod⁻¹ s⁻¹), and the mean dynamic range of ΔI_C was 3.10 log units, and of ΔI_Cl was 1.22 log units.

In order to elucidate how rods and amacrine cells mediate DBC_R light responses, we examined ΔI_C and ΔI_Cl elicited by 500 nm light (as discussed in Methods, 500 nm light allows us to separate the rod, M-cone and S-cone contributions to ΔI_C and ΔI_Cl) of various intensities (Figs 1D and E, and 2D and E). In both groups of DBC_Rs, the −7 light step elicited a small ΔI_C, whereas the −6 (0.7 Rh* rod⁻¹ s⁻¹) light step gave rise to an inward ΔI_C of about 25 pA. As the light step became brighter, the inward ΔI_C became larger. We measured the light sensitivity of ΔI_C to 500 nm light in all 12 group 1 DBC_Rs and 6 group 2 DBC_Rs, and the mean (± s.d.) response–intensity (ΔI_C–logI) relations are plotted in Figs 1F and 2F. The continuous curve was fitted by the Hill equation (see Methods). The mean thresholds (defined as eliciting 5% of the maximum response) were the same for the two groups of DBC_Rs (−6.8 = 0.1 Rh* rod⁻¹ s⁻¹), and the dynamic ranges were 3.1 and 3.4 log units, about 1.1–1.4 log units wider (extended to the left along the intensity axis) than that of the rod photocurrent (Field & Rieke, 2002b). Possible mechanisms underlying the dynamic range widening will be discussed later.

Our results show that all DBC_Rs exhibited light-evoked chloride current (ΔI_Cl) at E_C, which was not observed by Berntson & Taylor (2000). The mean threshold of light-evoked inhibitory current ΔI_Cl of the 12 group 1 DBC_Rs was −7.2 (0.044 Rh* rod⁻¹ s⁻¹) and that of the 6 group 2 DBC_Rs was −7.6 (0.018 Rh*rod⁻¹ s⁻¹), suggesting that the amacrine cells mediating ΔI_Cl of DBC_Rs have rod-like threshold. The mean (± s.d.) response–intensity (ΔI_Cl–logI) relations (shown in Figs 1F and 2F) had mean dynamic ranges of 2.3 and 1.2 log units, suggesting that the amacrine cell inputs to the two groups of cells are different (in the 8 DBC_Rs in which we examined the effects of GABA and glycine receptor antagonists (see above), we found that ΔI_Cl of group 1 and group 2 DBC_Rs exhibited different sensitivities to strychnine and picrotoxin, suggesting that these two groups of cells receive different GABAergic and glycinergic amacrine cell inputs; data not shown, J.-J. Pang, F. Gao & S. M. Wu, manuscript in preparation). Additionally, there are more spontaneous inhibitory postsynaptic currents (sIPSCs) in group 2 DBC_Rs (Figs 1E and 2E). Mean response thresholds, dynamic range saturation intensities as well as the dark membrane potentials of the two groups of DBC_Rs are listed in Table 1.

Table 1.

Response threshold, dynamic range and saturation values of various mouse retinal neurones elicited by 500 nm light steps

A.	Threshold (log unit) [Rh* rod−1 s−1]	Dynamic range (log units)	Saturation (log units)	Dark Vm (mV)
Rod	−5.57 [1.9]a	2.0a	−3.54	—
ConeM	−4.17 to −6.57 [48–190]b	2.0	−2.17 to −1.57	—
ConeS	∼−2.55 [∼2000]3	2.0	∼−0.55	—
B.	ΔIC	ΔICl	ΔIC	ΔICl	ΔIC	ΔICl
DBCR (Gr 1)	−6.8 [0.1]	−7.2 [0.044]	3.12	2.28	−3.88	−4.92	−59 ± 5
DBCR (Gr 2)	−6.8 [0.1]	−7.6 [0.018]	3.10	1.22	−3.90	−6.38	−57 ± 6
DBCC1	−6.9 [0.088]	−6.9 [0.088]	4.70	1.80	−2.20	−5.10	−56 ± 4
DBCC2	−4.9 [8.8]	−5.3 [3.5]	2.68	2.10	−2.22	−3.20	−56 ± 4
AIIAC	−9.0 [0.0007]	n.a.	4.92	n.a.	−4.08	n.a.	−65 ± 5
ONαGCd	− 7.2 [0.044]	−6.8 [0.1]	4.90	5.30	−2.30	−1.50	−63 ± 6

A: values of response threshold (intensity for generating 0.05 of maximum response), dynamic range (intensity span from 0.05 to 0.95 of maximum response) and saturation (intensity for generating 0.95 of maximum response) of mouse rods, M-pigment-dominated cones (Cone_M) and S-pigment-dominated cones (Cone_S) derived from previous publications. Light intensity required to elicit a threshold step response was calculated from the flash response threshold by a rod integration time of 0.4 s:

^aField & Rieke (2002b)

^bBaylor (1987)

^cLyubarsky et al. (1999). B: summary of mean response thresholds and dynamic ranges of ΔI_C and ΔI_Cl in group 1 and group 2 rod depolarizing bipolar cells (DBC_R Gr 1 and Gr 2), cone depolarizing bipolar cells DBC_C1 and DBC_C2 and all amacrine cells (AIIAC) obtained in this study (n.a., not applicable).

^dONαGC data are derived from Pang et al. (2003). Values are given in log units of attenuation and in Rh* rod⁻¹ s⁻¹ (in square brackets). The mean ±s.d. dark membrane potentials (V_m) of each cell type are listed in the right column.

Light-evoked response characteristics of AII amacrine cells, the primary postsynaptic neurones of the rod bipolar cells

Anatomical evidence has suggested DBC_Rs in the mammalian retina do not make output synapses on ganglion cells, but primarily on the AII amacrine cells (AIIACs) (Kolb & Famiglietti, 1974; Freed et al. 1987; Veruki & Hartveit, 2002a,b). In order to study the output of the rod bipolar cells, we examined the light response characteristics of morphologically identified AIIACs. We selected cells with the soma located near the inner half the INL and made whole cell voltage-clamp recordings with patch electrodes filled with internal solution and Lucifer yellow. AIIACs were identified by their characteristic morphology (thick globular dendrites in the distal half and pyramidally branching dendrites in the proximal half of the IPL, and with dendritic width less than 30 μm; Famiglietti & Kolb, 1975; Strettoi et al. 1992) revealed by Lucifer yellow fluorescent images in retinal slices and the inward light-evoked cation current (ΔI_C) recorded at E_Cl. Figure 3A shows the stacked confocal fluorescent image of an AIIAC in the retinal slice with the typical AIIAC morphology. We recorded from a total of 15 cells with very similar morphology in retinal slices, and the dendritic width varied between 20 and 35 μm. The light-evoked current responses of an AIIAC to a 2.5 s 500 nm light step at various holding potentials are shown in Fig. 3B, and the current–voltage (I–V) relations in Fig. 3C. We found that all 15 cells with AIIAC morphology had very similar I–V relations, and that the light-evoked current response is inward at all holding potentials (i.e. it did not have a reversal potential at least between −100 mV and +40 mV). This may be explained by the anatomical findings that AIIACs are extensively coupled with adjacent AIIACs and cone ‘on’ bipolar cells (Bloomfield & Xin, 2000; Veruki & Hartveit, 2002a), and thus light-evoked inward current from unclamped neighbouring cells may dominate the cells' response at all potentials. The mean zero-current potential in darkness of the 15AII ACs was −65 ± 5 mV.

Figure 3. All amacrine cells.

A, stacked confocal fluorescent image of a AIIAC in a mouse retinal slice. Abbreviations as for Fig. 1. B, light-evoked current responses to a 2.5 s light step (500 nm, −3.0 = 700 Rh* rod⁻¹ s⁻¹) at various holding potentials. C, current–voltage relations of the peak light responses in B. D, light-evoked excitatory cation current (ΔI_C) recorded at E_Cl to 500 nm light steps (2.5 s) of various intensities. E, response–intensity relation of the light-evoked cation current (○, ΔI_C–logI; error bars represent ± s.d.). The mean threshold of ΔI_C was −9.0 (0.0007 Rh* rod⁻¹ s⁻¹) and the mean dynamic range of ΔI_C was 4.90 log units.

Figure 3D shows the current responses of the same AIIAC at E_Cl to 2.5 s 500 nm light steps of various intensities, and Fig. 3E shows the mean response–intensity relation of all 15 AIIACs. The response–intensity relations of AIIACs are homogeneous as reflected by the relatively small s.d. in Fig. 3E. A striking feature of the AIIAC response–intensity relation is that the dynamic range is very wide (nearly 5 log units, extending 2 log units towards the left beyond the dynamic range of the DBC_RΔI_C, Figs 1F and 2F). This makes the AIIAC response threshold very low (near −9.0, or 0.0007 Rh* rod⁻¹ s⁻¹), about 100 times lower than that of the DBC_Rs, suggesting a large signal gain at the DBC_R–AIIAC synapses.

Table 2 lists the mean peak ΔI_C responses of 10 DBC_Rs and 11 AIIACs to the 9, −8, −7 and −6 (500 nm) lights. The mean peak ΔI_C responses of DBC_Rs to −6 and AIIACs to −8, −7 and −6 were obtained from original response records; responses of DBC_Rs to −7 and AIIACs to −9 were obtained from signal-averaged records (10 repeated responses each for increasing signal/noise ratio). DBC_R responses to −8 and −9 were extrapolated from the mean response–intensity relation because they were too small (note: the small DBC_RΔI_C in Table 2 (0.065 pA) is the theoretical mean response that represents a threshold response of about 9 pA in one out of 143 DBC_Rs that converge to an AIIAC; see Discussion below). By using this method, we estimated the current gain, defined as the ratio [peak ΔI_C(AIIAC)]/[peak ΔI_C(DBC_R)], elicited by a given light, of the DBC_R–AIIAC synapse, and the values are listed in Table 2. It is evident that the current gain is higher at smaller ΔI_C values (elicited by the −9 and −8 lights), and became increasingly lower when ΔI_C were larger (elicited by the −7 and −6 lights). Since the coupled AIIAC network receives synaptic inputs from many DBC_Rs, our data suggest that signal convergence from DBC_Rs to AIIACs is amplitude dependent. Possible mechanisms underlying such non-linearity will be discussed later.

Table 2.

Mean (±s.d.) peak ΔI_C responses of DBC_Rs and AIIACs and gain of the DBC_R–AIIAC synapse

	ΔIC−9 (pA)	ΔIC−8 (pA)	ΔIC−7 (pA)	ΔIC−6 (pA)
DBCR (n = 10)	0.065	0.60	3.4 ± 1.85	24.95 ± 8.65
AIIAC (n = 11)	4.62 ± 2.73	12.81 ± 6.37	30.66 ± 5.74	51.03 ± 10.99
Gain	70.08	21.35	9.02	2.05

Mean peak ΔI_C responses to 500 nm light steps of DBC_Rs to −6 and AIIACs to −8, −7 and −6 (logI) were obtained from original response records, those of DBC_Rs to −7 and AIIACs to −9 were from signal-averaged records (10 repeated responses each for increasing signal/noise ratio). DBC_R responses to −8 and −9 were extrapolated from the mean response–intensity relation because they were too small. The chord current gains, defined as the ratio [peak ΔI_C(AIIAC)]/[peak ΔI_C(DBC_R)], are listed in the bottom row.

Light-evoked current response characteristics of morphologically identified cone depolarizing bipolar cells (DBC_Cs)

Cone ‘on’ bipolar cells (DBC_Cs) were identified by their axonal morphology (ramified in the proximal half of the IPL, with relatively fine axon terminal branches, wider lateral extension and small vesicular expansions along the terminal branches, instead of the globular, large knob-shaped terminals characteristic of the DBC_Rs (Boycott & Wassle, 1999) revealed by Lucifer yellow fluorescent images in retinal slices and the inward light-evoked cation current (ΔI_C) recorded at E_Cl. Anatomical evidence shows that the morphology of DBC_Cs is more heterogeneous than the DBC_Rs, and the axon terminals of different types of DBC_Cs ramified at different depths in the IPL (Euler & Wassle, 1995). In addition, there are two types of cone pigments in the mouse retina, the M and S (or UV) pigments, and different cones contain M and S pigments of various proportions (Applebury et al. 2000). Furthermore, cones are electrical coupled with rods (Tsukamoto et al. 2001), and thus the light response of some cones may be influenced by the rod pigment. These results suggest that the light response of DBC_Cs may be more heterogeneous than that of the DBC_Rs.

Figure 4A shows the stacked confocal fluorescent image of a DBC_C in the retinal slice, and it exhibited typical DBC_C morphology with branching axon terminals stratifying between 70% and 85% of the IPL depth, and Fig. 5A shows the stacked confocal fluorescent image of another DBC_C with branching axon terminals stratifying between 55% and 75% of the IPL depth. These two types of DBC_C had similar current–voltage relations (Figs 4B and C, and 5B and C), the same mean dark membrane potential (−56 ± 4 mV), but very different response thresholds and dynamic ranges for ΔI_C and ΔI_Cl elicited by the 500 nm light (Figs 4D, E and F, and 5D, E and F). ΔI_C in Fig. 4 exhibited a threshold near −6.9 (0.088 Rh* rod⁻¹ s⁻¹) and a dynamic range of 4.6, and ΔI_C in Fig. 5 showed a threshold near −4.9 (8.8 Rh* rod⁻¹ s⁻¹) and a dynamic range of 2.1. Therefore the first types of DBC_C (named DBC_C1) receive substantial input from rods in addition to the cones (judged by the dynamic range span for the 500 nm light), and the second type of DBC_C (DBC_C2) receive input predominantly from the cones with much weaker rod inputs (thus lower sensitivity to 500 nm light and narrower dynamic range).

Figure 4. Type 1 cone depolarizing bipolar cell (DBC_C1).

A, stacked confocal fluorescent image of a DBC_C1 in a mouse retinal slice. Abbreviations as for Fig. 1. B, light-evoked current responses to a 0.5 s light step (500 nm, −4.0 = 70 Rh* rod⁻¹ s⁻¹) at various holding potentials. C, current–voltage relations of the peak light responses in B; the mean reversal potential of 11 cells was −41 ± 5.5 mV. D, light-evoked excitatory cation current (ΔI_C) recorded at E_Cl to 500 nm light steps (0.5 s) of various intensities. E, light-evoked inhibitory chloride current (ΔI_Cl) recorded at E_C to 500 nm light steps (0.5 s) of various intensities. F, response–intensity relations of the light-evoked cation and chloride currents: ○, ΔI_C–logI; •, ΔI_Cl–logI; error bars represent ± s.d.). The mean threshold of ΔI_C was −6.9 (0.088 Rh* rod⁻¹ s⁻¹) and of ΔI_Cl was −6.9 (0.088 Rh* rod⁻¹ s⁻¹); the mean dynamic range of ΔI_C was 4.70 log units, and of ΔI_Cl was 1.80 log units.

Figure 5. Type 2 cone depolarizing bipolar cell (DBC_C2).

A, stacked confocal fluorescent image of a DBC_C2 in a mouse retinal slice. Abbreviations as for Fig. 1. B, light-evoked current responses to a 0.5 s light step (500 nm, −3.0 = 700 Rh* rod⁻¹ s⁻¹) at various holding potentials. C, current–voltage relations of the peak light responses in B; the mean reversal potential of 4 cells was −51 ± 7.5 mV. D, light-evoked excitatory cation current (ΔI_C) recorded at E_Cl to 500 nm light steps (0.5 s) of various intensities. E, light-evoked inhibitory chloride current (ΔI_Cl) recorded at E_C to 500 nm light steps (0.5 s) of various intensities. F, response–intensity relations of the light-evoked cation and chloride currents: ○, ΔI_C–logI; •, ΔI_Cl–logI; error bars represent ± s.d.). The mean threshold of ΔI_C was −4.9 (8.8 Rh* rod⁻¹ s⁻¹) and of ΔI_Cl was −5.3 (3.5 Rh* rod⁻¹ s⁻¹); the mean dynamic range of ΔI_C was 2.68 log units, and of ΔI_Cl was 2.10 log units.

As with DBC_Rs, we verified that ΔI_C and ΔI_Cl in DBC_Cs are mediated by different synaptic inputs by examining the effects of GABA and glycine receptor antagonists on five DBC_C1s and four DBC_C2s and found that these compounds did not significantly alter the amplitude of ΔI_C (peak ΔI_C in 100 μm PTX + 1 μm STR was 1.15 ± 0.19 (n = 9) of peak ΔI_C in control solution). This suggests that GABAergic and glycinergic interneurones exert little influence on photoreceptor inputs to DBC_Cs in the mouse retina (the small increase in peak ΔI_C in PTX and STR may suggest, however, GABAergic/glycinergic feedback from horizontal cells to cone photoreceptors; detailed analysis will be presented in a future publication, J.-J. Pang, F. Gao & S. M. Wu, in preparation).

ΔI_Cl in Fig. 4 exhibited a threshold near −6.9 (0.088 Rh* rod⁻¹ s⁻¹) and a dynamic range of 1.8, and that in Fig. 5 had a threshold of 5.3 and a dynamic range of 2.10. Similar to the two groups of DBC_Rs, DBC_C1 and DBC_C2 also differed in spontaneous inhibitory postsynaptic currents (sIPSCs). sIPSCs in DBC_C2 were larger and of higher frequency. These results suggest that DBC_C1s and DBC_C2s receive inhibitory inputs from different populations of amacrine cells (in the 5 DBC_C1s and 4 DBC_C2s in which we examined the effects of GABA and glycine receptor antagonists, we found that ΔI_Cl of DBC_C1s and DBC_C2s exhibited different sensitivities to strychnine and picrotoxin, suggesting that these two types of DBC_Cs receive different GABAergic and glycinergic amacrine cell inputs (data not shown, J.-J. Pang, F. Gao & S. M. Wu, in preparation)). The mean response thresholds and dynamic ranges as well as the dark membrane potentials of the two groups of DBC_Cs are listed in Table 1.

Among 22 DBC_Cs, 12 had DBC_C1-like morphology and light responses, 4 had DBC_C2-like morphology and light responses, and the remaining 6 had axon terminals branching within 60–90% of the IPL depth, and exhibited response thresholds and dynamic ranges between the corresponding values of the DBC_C1s and DBC_C2s. These results suggest that DBC_Cs in the mouse retina are heterogeneous, 55% of them are DBC_C1-like, 18% are DBC_C2-like, and the remainder have intermediate responses.

Mean response–intensity spans of depolarizing bipolar cells and AII amacrine cells in dark-adapted mouse retina

In order to determine the relative contributions of rod and cone inputs to various types of DBCs and AIIACs, we plot in Fig. 6 the intensity spans of various cells in response to 500 nm light stimuli. Each horizontal bar marks the mean dynamic range (light intensity from threshold (5% of the maximum response) to saturation (95% of the maximum response) of a cell type. It is evident that the dynamic range of rod-mediated responses (ΔI_C recorded at E_Cl) in both groups of DBC_Rs (from −7.0 to −3.88, and from −7.0 to −3.90) is about 1.1–1.4 log units wider than that of the rod photocurrent (from −5.57 to −3.57) (Howes et al. 2002; Field & Rieke, 2002b). The extension of the dynamic range is at the low intensity side (towards left along the intensity axis), while the saturation intensity is close to the saturation intensity for the rod photocurrent.Similarly, the dynamic range of the AIIACs (from −9.0 to −4.08) is about 1.8 log units wider that that of the DBC_Rs, the widening is at the low intensity side, and the saturation intensity is close to the saturation intensity for the rod and DBC_R responses. These plots clearly suggest amplification of small rod signals by the rod–DBC_R synapses and small DBC_R signals by the DBC_R–AIIAC synapses. The observation that both DBC_R and AIIAC responses saturate near the saturation intensity for the rod photocurrent (but not extending into the cone response dynamic range) indicate that cone inputs to DBC_Rs and AIIACs under dark-adapted conditions are relatively weak.

Figure 6. Mean cell dynamic ranges.

Mean ΔI_C and ΔI_Cl dynamic ranges of rods, M-pigment-dominated cones (Cone_M), S-pigment-dominated cones (Cone_S), groups 1 and 2 DBC_Rs, DBC_C1s, DBC_C2s, AIIACs and ONαGCs (Pang et al. 2003) in response to 500 nm light stimuli (plotted from values in Table 1). Each horizontal bar marks the light intensity range from threshold (5% of the maximum response) to saturation (95% of the maximum response). Red lines, ΔI_C; black lines, ΔI_Cl.

The mean dynamic range of DBC_C1ΔI_C is 4.7 log units (ranging from −6.9 to −2.2), and that of DBC_C2ΔI_C is 2.68 log units (ranging from −4.9 to −2.22). Both types of DBC_Cs saturate near the saturation intensity of the M-pigment-dominated cones, but the thresholds to the 500 nm lights are very different. The mean threshold of DBC_C1s is very close to that of the DBC_Rs, suggesting that these cells receive strong rod inputs in their ΔI_C. The mean threshold of DBC_C2s is two log units higher, indicative of much weaker rod inputs in their ΔI_C.

We also plot in Fig. 6 the mean dynamic ranges of ΔI_Cl of DBC_Rs and DBC_Cs. It is evident that ΔI_Cl of different types of DBCs are very different, suggesting that the amacrine cell inputs to these cells are very complex. In order to compare the DBC and AIIAC responses with their output neurones, we added the mean dynamic ranges of ΔI_C and ΔI_Cl of the mouse 'on' alpha ganglion cells (ONαGCs) derived from a previous publication (Pang et al. 2003) at the bottom of Fig. 6. The mean dynamic range of the ONαGC ΔI_C is very close to that of the DBC_C1 and overlaps with the dynamic range of DBC_C2ΔI_C, consistent with the notion that ONαGC ΔI_C is primarily mediated by the DBC_C output synapses. The dynamic range of ONαGC ΔI_Cl is wider than the ΔI_Cl ranges of the DBC_Rs and DBC_Cs, and it is located at an intensity range two log units less sensitive than the dynamic range of the AIIAC ΔI_C, suggesting that the inhibitory inputs to ONαGCs (black arrow 5 in Fig. 7) may involve more than one type of amacrine cell. It is unlikely that the AIIACs make a significant contribution to the ONαGC ΔI_Cl (they contribute ONαGC ΔI_C through DBC_C1s via gap junctions) because the ONαGC ΔI_Cl threshold is much lower than the AIIAC responses. AIIACs may contribute more significantly to the light responses of other ganglion cells, such as the sustained ‘off’ αGCs described in a previous report (Pang et al. 2003).

Figure 7. Synaptic circuit diagram of the DBC_Rs, DBC_Cs, AIIACs and ONGCs in the mouse retina.

R, rods; C, cones; DBC_R, rod depolarizing bipolar cell; DBC_C1 and DBC_C2, types 1 and 2 cone depolarizing bipolar cells; all, AII amacrine cells; ONGC, 'on' ganglion cells. Arrows: chemical synapses (red, glutamatergic; black, GABAergic/glycinergic; +, sign-preserving; −, sign-inverting). Red zig-zag lines, electrical synapses; PRL, photoreceptor layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer (a, sublamina a; b, sublamina b); GCL, ganglion cell layer. This circuit diagram only includes a minimum number of cell types while our data may also be explained by more complex schemes. In order to save space, reciprocal AC inputs are not drawn, but are represented by black arrows (1–5).

Discussion

Synaptic circuitries of rod and cone DBCs and AII amacrine cells in the mouse retina

Based on our analysis of light-evoked cation and chloride currents (ΔI_C and ΔI_Cl), we propose a functional synaptic circuitry diagram of the rod, cone and AC inputs to DBC_Rs, DBC_Cs and AIIACs in the mouse retina (Fig. 7). The outlines of this diagram are consistent with the general plan set forth by anatomical data, but with several new and more detailed findings revealing how synapses in the mouse retina function.

We found that the light-evoked cation current, ΔI_C, of DBC_Rs are homogeneous, although two groups of cells can be distinguished morphologically by the slight difference in axon terminal stratification in the IPL, and more importantly, by their inhibitory inputs. All DBC_Rs are highly sensitive to 500 nm lights, consistent with anatomical data, suggesting that bipolar cells with globular axon terminals at 70–100% of the IPL depth contact only rods (Euler & Wassle, 1995; Hartveit, 1997; Tsukamoto et al. 2001). The light-evoked chloride current (ΔI_Cl) of DBC_Rs are mediated by amacrine cells with rod-like threshold, consistent with anatomical observations that DBC_Rs receive reciprocal feedback synapses from DBC_R-driven amacrine cells (Nelson & Kolb, 1985; Freed et al. 1987; Sandell et al. 1989; Hartveit, 1999). We propose that ΔI_Cl in the mouse DBC_Rs are mediated by amacrine cells in the inner retina because physiological evidence has shown that chloride currents were observed when inhibitory neurotransmitters were applied to the axon terminal regions in the IPL (McCall et al. 2002). The dynamic ranges of ΔI_Cl in DBC_Rs are substantially narrower than those of ΔI_C, indicating that synaptic clipping (Attwell et al. 1987) may occur in the DBC_R→reciprocal rod AC→DBC_R feedback loop. Our observation that the two groups of DBC_Rs differ in ΔI_Cl dynamic range and sIPSCs suggests that the reciprocal rod AC inputs to DBC_Rs with axon terminals at two levels of IPL may not be identical. However, since the dark membrane potential of all DBC_Rs are very close to E_Cl (Table 1), light-evoked voltage responses of the two groups of DBC_Rs under dark-adapted conditions may be similar (the different ΔI_Cl may contribute small differences in the voltage responses by shunting). For this reason, we do not feel it is justified to state that there are two types of DBC_R, but merely propose that DBC_Rs receive reciprocal inputs from different amacrine cells at two strata of the IPL.

In all 15 AIIACs, we were unable to reverse the light-evoked current. This is consistent with anatomical findings that these cells are strongly coupled with one another, and with DBC_Cs (Famiglietti & Kolb, 1975; Kolb & Nelson, 1993; Vaney, 2002; Veruki & Hartveit, 2002a). As the light-evoked current in a given AIIAC is pooled from a network of cells, it was impossible to depolarize all coupled neurones beyond their reversal potentials.

We show that there are at least two types of DBC_C in the mouse retina with distinguishable morphology and light responses. DBC_C1s have axon terminals ramified in 70–85% of IPL depth, a ΔI_C threshold near that of the DBC_Rs and a saturation intensity near that of the M-pigment-dominated cones. It is likely that these cells receive strong rod signals through gap junctions between rods and cones, and/or between AIIACs and their own axon terminals (Deans et al. 2002; Veruki & Hartveit, 2002a) (Fig. 7). DBC_C2s have axon terminals ramified in 55–75% of the IPL depth and a much less sensitive ΔI_C, indicating that their rod-mediated inputs (rod–cone and AIIAC coupling) are much weaker. The ΔI_Cl threshold of DBC_C1s is about 1.5 log units lower than that of the DBC_C2s, suggesting that amacrine cells making reciprocal inhibitory synapses on the former DBCs are more rod-dominated than the amacrine cells synapsing on the latter. In addition to DBC_C1s and DBC_C2s, we also recorded six DBC_Cs with mixed DBC_C1/DBC_C2 light response and morphological characteristics. Although axon terminals of mouse DBCs are rarely as narrowly monostratified in the IPL as the salamander bipolar cells (Wu et al. 2000), the general rod/cone dominance rule set forth by the salamander bipolar cells is obeyed: axon terminals of DBCs with stronger rod inputs ramified closer to the ganglion cell layer and those with stronger cone inputs ramified closer to the centre of the IPL.

It is important to note that we may not have included all morphological types of depolarizing bipolar cells in this study. For example, immunocytochemical evidence has shown that there are four types of ‘on’ cone bipolar cells in the rat retina (Euler & Wassle, 1995). It is possible that some of these cells are less accessible to our patch electrodes and/or they exhibit smaller light responses than DBC_C1 and DBC_C2 in dark-adapted retinal slices.

Signal convergence and amplification of the DBC_R–AIIAC synapse

Anatomical studies show that 22 rods converge to a DBC_R in the mouse retina (Tsukamoto et al. 2001), and our data reveal that the threshold of the DBC_Rs is about 19 ((1.9 Rh* rod⁻¹ s⁻¹)/(0.1 Rh* rod⁻¹ s⁻¹), Table 1) times lower than that of the rods. This suggests that signal convergence plays a major role in sensitivity amplification at the rod–DBC_R synapse: a threshold signal (5% of maximum response) can be elicited in a DBC_R by an incident light that is just bright enough to generate a threshold signal in any one of its presynaptic rods. We named this type of signal convergence the one-to-one threshold scheme (threshold signal in one presynaptic cell is adequate for generating a threshold response in the postsynaptic cell). As the incident light becomes brighter, threshold signals are elicited in more rods and the DBC_R response is larger, and when the light reaches the intensity about 20 times brighter than that for the DBC_R threshold, threshold signals are elicited in every rod (defined as rod threshold) and the DBC_R sums signals from all presynaptic rods.

Anatomical data revealed that an AIIAC in the mouse retina collects synaptic inputs from several (3–4) DBC_Rs (Tsukamoto et al. 2001), and they are electrically coupled to one another (Famiglietti & Kolb, 1975; Deans et al. 2002; Veruki & Hartveit, 2002a). It is not clear how strongly AIIACs are coupled together in the mouse retina, and how many DBC_Rs contribute to the light response of a single AIIAC through the coupled AIIAC network. Our data show that the mean response threshold of the AIIACs to 500 nm light is about 143 ((0.1 Rh* rod⁻¹ s⁻¹)/(0.0007 Rh* rod⁻¹ s⁻¹)) times lower than that of the DBC_Rs. If the one-to-one threshold scheme also applies to the DBC_R–AIIAC synapses, then about 143 DBC_Rs should contribute to the light response of a single AIIAC, or the light response of an AIIAC is pooled through a coupled network consisting of about 40 (143/(3–4)) AIIACs. This estimate is of the same order of magnitude as the number of coupled AIIACs in the cat retina (Smith & Vardi, 1995). On the other hand, if the one-to-one threshold scheme does not hold for the DBC_R–AIIAC synapse (e.g. threshold signals in more than one presynaptic cell are needed to elicit a threshold response in the postsynaptic cell), then the number of AIIACs that share signals would be lower, or less than 40 AIIACs are effectively coupled. We plan to examine the effect of AIIAC coupling on AIIAC light sensitivity by studying light responses of DBC_Rs and AIIACs in the connexin-36 knockout mice in the future.

The data listed in Table 2 suggest that the current gain of the DBC_R–AIIAC synapse is higher for smaller ΔI_C values and became progressively lower for larger ΔI_C. Similar non-linear input–output relations have been observed in other glutamatergic synapses in the retina (e.g. the rod–horizontal cell synapses in amphibian retinas; Attwell et al. 1987). It has been proposed that such non-linearity is caused by voltage dependence of presynaptic calcium current and glutamate receptor saturation (Falk, 1988; Attwell, 1990). Further investigation is needed to clarify this issue in the mouse retina.

Asymmetric interaction of rod and cone signals in dark-adapted mouse DBCs and AIIACs

In this study, we analysed the dynamic ranges of ΔI_C generated by 500 nm light in DBC_Rs, AIIACs and DBC_Cs, and compared them with the dynamic ranges of rods and cones (Fig. 6). Our data indicate that under dark-adapted conditions, the cone's influence on the rod system (DBC_Rs and AIIACs) is weaker than the rod's influence on the cone system (DBC_Cs). For example, ΔI_C in DBC_Rs and AIIACs saturate near the saturation intensity of the rods (−3.5), and do not extend substantially into the cone dynamic range. On the other hand, ΔI_C in DBC_C1s saturate near the saturation intensity of the M cones (as they receive inputs exclusively from cones; Tsukamoto et al. 2001), but their dynamic range extends leftward into the DBC_R response range (threshold near −7.0, Fig. 6). This nearly 3 log unit leftward extension of DBC_C1 dynamic range is unlikely to be mediated by signal convergence (as in rod–DBC_R and DBC_R–AIIAC convergence discussed above), because on average each cone DBC only contacts four cones in the mouse (Tsukamoto et al. 2001). Therefore DBC_C1s must receive strong inputs from the rods.

There are at least two possible explanations for the asymmetric rod–cone signal mixing in DBCs. First, it has been shown that rod–cone contacts in the mouse retina are highly asymmetrical: on average 32 rods contact one cone and only 1.2 cones contact one rod (Tsukamoto et al. 2001). Therefore, rods influence cones much more than cones influence rods. The second reason is that postsynaptic receptors in DBC_Rs and AIIACs may approach saturation when stimulated by 500 nm light near −4.0 (Xin & Bloomfield, 1999). At higher light intensities when cone inputs join in, ΔI_C in DBC_Rs and AIIACs cannot grow larger.

In summary, by analysing light response thresholds and dynamic ranges of dark-adapted mouse DBCs and AIIACs, we found that small rod signals are greatly amplified by the rod–DBC_R–AIIAC synaptic pathway, and they are effectively transmitted to the cone system (DBC_Cs) via gap junctions between rods and cones and between AIIACs and DBC_Cs. This allows ‘on’ ganglion cells, which receive synaptic inputs exclusively from DBC_Cs, to reliably collect information from both the rod and cone signalling systems. Although there is no evidence that the rod–cone and AIIAC–DBC_C gap junctions are rectified, these electrical synapses are functionally asymmetrical under dark-adapted conditions: cone-mediated signals contribute much less to the light responses of cells in the rod system (DBC_Rs and AIIACs) than the rod-mediated signals in cells in the cone system (DBC_Cs and ‘on’ ganglion cells). Under different physiological conditions, such as in the presence of background light, signal interaction between the rod and cone systems can be very different, and cone-mediated responses may dominate all retinal neurones.