Genetic elimination of rod/cone coupling reveals the contribution of the secondary rod pathway to the retinal output

Abstract

In the retina, signals originating from rod and cone photoreceptors can reach retinal ganglion cells (RGCs)—the output neurons—through different pathways. However, little is known about the exact sensitivities and operating ranges of these pathways. Previously, we created rod- or cone-specific Cx36 knockout (KO) mouse lines. Both lines are deficient in rod/cone electrical coupling and therefore provide a way to selectively remove the secondary rod pathway. We measured the threshold of the primary rod pathway in RGCs of wild-type mice. Under pharmacological blockade of the primary rod pathway, the threshold was elevated. This secondary component was removed in the Cx36 KOs to unmask the threshold of the third rod pathway, still below cone threshold. In turn, the cone threshold was estimated by several independent methods. Our work defines the functionality of the secondary rod pathway and describes an additive contribution of the different pathways to the retinal output.

The rod/cone pathway is the entry to an important functional pathway in the retina.

INTRODUCTION

Every day, the retina is confronted with the challenge of encoding and processing light inputs that vary widely in intensity across the 24-hour light/dark cycle. To cope with this challenge, vertebrate retinal processing relies on the presence of two classes of photoreceptors, rods and cones, that differ in sensitivity to light (1, 2). Rods mediate signaling at low light levels, such as starlight, whereas cones signal at higher light levels, such as daylight. Rods are exquisitely designed for the transmission of single-photon absorption events and saturate at flashes that deliver 100 to 1000 photons (3–5). Yet, unexpectedly, rod signaling at the retinal output extends beyond the operating range of single rods to cover more than 6 log units in intensity. The lower end of the range can be as low as 3 log units below the threshold of individual rods, due to the high degree of convergence within the retinal circuit. The upper end extends well above the cone threshold and the intensity at which rods saturate and reflects adaptive mechanisms in the rod themselves (3, 6). In addition, it has been proposed that rod-mediated signals can flow across the retina through different pathways that have distinct sensitivities. These different rod pathways are thought to combine with the cone pathway to support continuous signaling throughout the entire operating range of the retina, over more than 9 log units (3, 7–10).

Anatomical, functional, and psychophysical evidence supports the presence of at least three rod pathways in the mammalian retina (3, 5, 7, 11). These include the following: a primary rod pathway in which rods signal to dedicated rod bipolar cells (RBCs) and then to a specific type of amacrine cell—AII amacrine cells—and eventually to the cone circuitry, a secondary rod pathway in which rod signals enter cone pathways through rod/cone gap junctions, and a tertiary rod pathway in which rods signal directly to off-center (OFF) cone bipolar cells (CBCs) (Fig. 1A). It is well accepted that the primary pathway processes single-photon absorption events (3–5), but the threshold and range of intensities at which the secondary rod pathway operates have yet to be comprehensively characterized, and the functional importance of the tertiary rod pathway remains largely unexplored. In addition, previous studies reported conflicting results as to whether every retinal ganglion cell (RGC) receives the whole set (12, 13) or a subset of inputs (8–10, 14). Last, little effort has been made to link the degree of convergence (or segregation) of the rod pathways to specific types of RGC, which we now know encompass more than 40 different types in mouse retina (15).

Fig. 1. Schematic representation of the rod and cone pathways of the wildtype mammalian retina and of the rod- or cone-Cx36 KO retinas.

(A) Architecture of rod and cone circuits in the wild-type retina. Rod signals can reach RGCs—the output neurons of the retina— through three different pathways. The primary rod pathway (1), where rod signals are transmitted to RBCs, is of high sensitivity and plays a prominent role in the transmission of scarce light signals under very dim light (low scotopic) conditions near absolute visual threshold. In the secondary rod pathway or rod/cone pathway (2), rod signals may enter directly into the cone pathway through rod/cone gap junctions. In the tertiary rod pathway (3), rod signals may enter directly into the cone pathway through synaptic contacts between rods and some types of OFF CBCs. Cones contact both ON CBCs and OFF CBCs, thereby forming the entry of the cone pathways (C). AII, AII amacrine cell. (B) The secondary rod pathway is missing in the rod-Cx36 KO and cone-Cx36 KO mutant lines. The primary rod pathway can be blocked by APB. Therefore, the light responses of OFF RGCs are driven by the tertiary rod pathway and the cones in the mutant lines in the presence of APB. Because APB blocks the ON cone pathway as well, all light responses in ON RGCs are abolished in the presence of APB. Red lettering indicates circuits sensitive to APB.

We recently developed mouse lines that lack the gap junction–forming protein connexin36 (Cx36) specifically in rods or in cones [rod-Cx36 knockout (KO) and cone-Cx36 KO] (16). We used these lines to establish the connectivity rules between photoreceptors and demonstrated that (i) Cx36 is required and sufficient for photoreceptor coupling and (ii) rod/cone gap junctions are predominant in the photoreceptor network (16). Because the rod/cone gap junction is the entry of the secondary rod pathway, we set out to use these new lines as a tool to eliminate the secondary rod pathway. In this way, we could determine the threshold and functional contribution of the secondary rod pathway to a single type of RGC: the transient OFF alpha RGC (tOFF αRGC) in the fully dark-adapted mouse retina (Fig. 1B). Using complementary genetic and pharmacological approaches, we establish the contribution of all three rod pathways and the cone pathway to the light responses of tOFF αRGCs. Thus, our work establishes both qualitatively and quantitatively the full set of rod inputs to a specific RGC type, the tOFF αRGC. This knowledge will be valuable to further study the modulation of each rod pathway, such as under light adaptation and/or under the influence of circadian signals. Furthermore, the similarity in the results obtained in the rod-Cx36 KO and cone-Cx36 KO lines validates both mutant lines as genetic models that lack a secondary rod pathway.

RESULTS

Targeting tOFF αRGCs in the dark-adapted mouse retina

We made loose-patch recordings from αRGCs in isolated whole-mount living retinas. a cells were targeted under infrared illumination on the basis of their soma size, typically the largest among RGCs [>20 μm; (17, 18)]. We chose to use the tOFF αRGCs for most of these experiments because their primary rod input via RBCs can be selectively blocked with l-(+)-2-amino-4-phosphonobutyric acid (APB) (7, 19). This is because rod/RBC synapses use mGluR6 receptors that are blocked by APB, while the cone-driven responses to OFF RGCs pass directly via cone/OFF CBC synapses, which use ionotropic glutamate receptors of the AMPA/kainate type (20). The identity of tOFF αRGCs was confirmed by their characteristic transient OFF responses and postfixation morphological analysis (Fig. 2).

Fig. 2. Targeting, dendritic morphologies, and light response characteristics of the tOFF αRGC.

(A) αRGCs have the largest soma size among cells in the ganglion cell layer. Scale bar, 25 μm. (B) Dendritic processes and axon (white arrow) are revealed following neurobiotin injection and processing of the tissue. Scale bar, 50 μm. (C) Volume projection of z-stack images that include the inner plexiform layer showing the location of the dendrites of a tOFF αRGC (white arrow) between the two choline acetyltransferase (ChAT) bands. Two ChAT-positive starburst amacrine cell bodies (C1 and C2) also appear in the image. (D) Loose-patch recording of a tOFF αRGC showing a family of 25 consecutive recordings (spike raster sweeps) and the calculated mean firing rate. Note the transient nature of its light responses at lights OFF.

Thresholds of the rod-driven pathways

The light responses of tOFF αRGCs were measured in rod- and cone-specific Cx36 KOs and their respective wild-type littermate controls (ctl) (Fig. 3). Figure 3A shows examples of a series of 20 consecutive light responses (spike raster plots) obtained in cone-Cx36 KO ctl retinas at three different intensities [0.06 [1], 2 [2], and 20 [3] effective isomerizations (R*) per rod per s (R*/rod/s)] before, during, and after 25 μM APB application.

Fig. 3. Contribution of the rod/cone pathway to the retinal output.

(A to F) Loose-patch recordings of the light responses of tOFF αRGCs in wild-type (A to C) and mutant Cx36 retinas (D to F). (A and D) Examples of raster plots that include 20 consecutive light responses and averaged firing rates. Recordings were obtained in the dark-adapted retina. A 0.5-s stimulus was presented at three different light intensities (0.06 R*/rod/s, top row [1]; 2 R*/rod/s, middle row [2]; and 20 R*/rod/s, bottom row [3]), before (left column, control), in the presence of 25 μM APB (middle column, APB), or after APB washout (> 30 min, right column, washout). In the wild type, APB eliminated the responses to the dimmest light stimulus, most of the responses at the intermediate intensity but had limited effect on responses to the brightest stimuli. In the mutant, APB eliminated the responses to both dim and intermediate light stimuli and greatly attenuated responses to the brightest stimuli. (B, C, E, and F) Averaged data from tOFF αRGCs obtained in cone-Cx36 KO ctl (n = 5) (B), rod-Cx36 KO ctl (n = 13) (C), cone-Cx36 KO (n = 8) (E), and rod-Cx36 KO retinas (n = 9) (F). The results are consistent with the primary (APB-sensitive) rod pathway being the most sensitive pathway and with the threshold of the secondary rod pathway being ~1 R*/rod/s and the elimination of this pathway in the mutant lines (also see Table 1 and fig. S2 for details). White arrows in (B) and (E) show the three intensities tested in (A) and (D).

APB application (>15 min, middle traces) totally eliminated the responses to the dimmest stimulus but had less effect on the responses to brighter stimuli. Because APB blocks all ON pathways, its effect on the dimmest stimulus is consistent with the primary rod pathway being the major input to tOFF αRGCs at very low light levels. The increase in RGC intrinsic activity observed in the presence of APB is consistent with the suppression of the inhibitory input from AII amacrine cells to OFF CBCs and OFF RGCs and results from relieved inhibition or disinhibition of the OFF pathway (21). Despite the change in baseline, we measured the OFF response amplitude as the difference between the basal firing level calculated before light onset and the mean peak OFF response value at light offset (see the “Single-cell recording of αRGCs” section in Materials and Methods for details). For dim light responses, blockade of the primary rod pathway inhibitory input to tOFF αRGCs directly translated into the elimination/reduction of the suppression of the firing rate at light onset and the elimination/reduction of the response at light offset, which was now missing the rebound component normally provided at the end of the inhibitory signal. Note that APB had less effect on the OFF responses to brighter stimuli, indicating less contribution of the primary rod pathway and revealing input from the secondary and/or tertiary rod pathways (see Discussion for possible limitations to the use of APB).

Plotting the relative intensity-response profiles of normalized responses of tOFF αRGCs as a function of the stimulus intensity (ranging from 0.001 to 10,000 R*/rod/s) and using a threshold criterion of 5% of the maximum response indicated a threshold under control conditions (i.e., no APB) around 0.05 R*/rod/s (Fig. 3B). tOFF αRGCs in rod-Cx36 KO ctl retinas had a similar threshold (Fig. 3C). Our estimates of the threshold of tOFF αRGCs in wild-type littermate retinas are in close agreement with our measurements in wild-type B6 retinas (Table 1 and fig. S1, A and B) and with previous works (9, 17, 21, 22). In both the cone-specific (Fig. 3, D and E) and the rod-specific (Fig. 3F) Cx36 KO lines, we found that the threshold of tOFF αRGCs is close or similar to their respective control littermates (~0.05 R*/rod/s). We calculated a small but significant increase in threshold for the rod-Cx36 KO line compared to its littermate control (0.172 R*/rod/s versus 0.036 R*/rod/s) (Table 1), but both thresholds were within the low scotopic range. Part of the difference may have come from the somehow shallower slopes of the response-intensity functions in the rod mutant line and the criterion to establish threshold (5% of maximum). We did not pursue this difference any further. Globally, the results indicate that the most sensitive pathway, the primary rod pathway, is functional in the KO lines.

Table 1. Threshold (stimulus intensity to generate 5% of maximal response).

Data are means ± SEM. Units are R*/rod/s. **P < 0.01. ***P < 0.001 [within genotype, compared to control (no drug)]. #P < 0.05. ###P < 0.001 (between genotypes, compared to wild type B6). †††P < 0.001 [compared to respective wild-type control littermates (ctl)]; Tukey post hoc test. APB, 2-amino-4-phosphonobutyric acid; RM-ANOVA, repeated measures analysis of variance.

	n	Control	+APB	Washout	RM-ANOVA
Wild type (B6)	5	0.062 (0.036)	3.12 (0.781)**	0.075 (0.008)	F2,14 = 14.8 (P = 0.002)
Cone-Cx36 KO ctl	5	0.063 (0.006)	2.00 (0.548)**	0.022 (0.006)	F2,14 = 12.8 (P = 0.003)
Cone-Cx36 KO	8	0.055 (0.004)	63.8 (8.44)***, ###, †††	0.029 (0.002)	F2,23 = 57.0 (P < 0.0001)
Rod-Cx36 KO ctl	13	0.036 (0.006)	1.86 (0.435)***	0.037 (0.007)	F2,38 = 17.5 (P < 0.0001)
Rod-Cx36 KO	9	0.172 (0.045)†††	62.2 (8.12)***, ###, †††	0.328 (0.095)#, †††	F2,26 = 59.0 (P < 0.0001)
ANOVA		F4,35 = 5.35 (P = 0.0018)	F4,35 = 35.7 (P < 0.0001)	F4,35 = 7.93 (P = 0.0001)

In the wild-type mouse, blocking the primary rod pathway with APB caused ~1.5 log unit shift to the right in the threshold of tOFF αRGCs, indicating that the primary rod pathway is the sole input to the cells at the lowest intensities (Fig. 3, B and C; Table 1; and fig. S1A). The difference between inflection points (K_1/2) gave a more accurate reading of the shift between the curves (table S1). This revealed the threshold of the other rod and cone pathways, ~1 R*/rod/s (see fig. S1C for a schematic representation). Because this threshold was substantially below the cone threshold (expected to be >100 R*/rod/s), it indicated the threshold of the next most sensitive rod pathway.

When we repeated this experiment in the cone- or rod-Cx36 KO (Fig. 3, D to F), APB caused a rightward shift of ~3 log units (Table 1 and table S1). This showed the absence of both the primary and the secondary rod pathways and revealed the threshold of the tertiary rod pathway (Fig. 3, E and F; Table 1; and fig. S2). This threshold (~60 R*/rod/s) was still rod driven because it was below the expected cone threshold.

Cone threshold

In the absence of selective tools to block the tertiary rod pathway from rods to OFF CBCs, we estimated the cone threshold by three independent methods.

1) By recording from ON αRGCs in pan-Cx36 KO mice (Fig. 4). ON RGCs in pan-Cx36 KO mice have no primary rod pathway input due to the absence of AII/ON CBC gap junctions in the inner plexiform layer and no secondary rod pathway due to the absence of rod/cone coupling (8, 16). Furthermore, there are no connections between rods and ON CBCs (23, 24), so there is no tertiary pathway to ON RGCs. In the absence of the primary, secondary, and tertiary rod pathways, the threshold of ON αRGCs in the pan-Cx36 KO retina yielded the cone threshold at 175 ± 48 R*/rod/s (means ± SEM, n = 7), in close agreement with previous reports [see, for instance, (25)] yet slightly higher than others (8, 9).

Fig. 4. Establishing the contribution of pure cone input to αRGCs – recording from ON αRGCs in the pan-Cx36 KO mouse.

(A) In the pan-Cx36 KO mouse, Cx36 gap junctions between photoreceptors and between AII amacrine cells and ON CBCs are eliminated. Consequently, rod signals cannot reach ON RGCs through the primary or the secondary rod pathways, and, therefore, cone signals only drive ON RGC light responses. Red lettering indicates circuits sensitive to APB. (B) Raw data. A 0.5-s stimulus was presented at three different light intensities (0.06 R*/rod/s, top row [1]; 20 R*/rod/s, middle row [2]; and 1000 R*/rod/s, bottom row [3]), before (left column, control), or in the presence of 25 μM APB (right column, +APB). As expected, the cone ON pathway was sensitive to APB. Inset shows averaged normalized responses of seven cells (one cell per retina) to intensities ranging from 0.002 to 100,000 R*/rod/s. Recording from ON αRGCs in the pan-Cx36 KO retina enabled us to determine the cone threshold in αRGCs (~175 R*/rod/s, gray arrowhead).

2) By recording from tOFF αRGCs in the Gnat1^irdr mutant mouse line (Fig. 5). Gnat1^irdr mice have no rod function, and therefore, tOFF αRGCs lack all three rod-driven pathways (26). This approach gave a cone threshold response at ~172 R*/rod/s (control, 167 ± 58 R*/rod/s; +APB, 200 ± 100 R*/rod/s; washout, 150 ± 30 R*/rod/s; means ± SEM, n = 3).

Fig. 5. Establishing the contribution of pure cone input to αRGCs—recording from tOFF αRGCs in the Gnat1^irdr mouse.

(A) In the GNAT1^irdr mouse, rods are present but do not respond to light. Consequently, cone signals only drive tOFF αRGC light responses. Red lettering indicates circuits sensitive to APB; only the cone ON pathway is sensitive to APB. (B) Raw data. A 0.5-s stimulus was presented at three different light intensities (0.06 R*/rod/s, top row [1]; 20 R*/rod/s, middle row [2]; and 1000 R*/rod/s, bottom row [3]), before (left column, control), in the presence of 25 μM APB (middle column, +APB), or after APB washout (> 30 min, right column, washout). As expected, the cone OFF pathway was insensitive to APB. Inset shows averaged data from three cells (one cell per retina). Recording from tOFF RGCs in the GNAT1^irdr retina enabled us to determine the cone threshold (~172 R*/rod/s, colored arrowheads).

3) Last, we separated rod- and cone-mediated signals on the basis of their selectivity to stimulus temporal frequencies. Specifically, we recorded the sensitivity of ON αRGCs to a 10-Hz stimulus, a frequency above the rod range and thus selective for cones, in the wild-type (B6) retina (Fig. 6). We compared a steady 1-s stimulus to a train of the same intensity at 10 Hz. In low light conditions, in the rod-driven range, the continuous stimulus and the 10-Hz train produced similar responses. These observations are consistent with the slow kinetics of the rod response and its inability to follow a flickering stimulus of frequency greater than ~8 Hz (8). However, at around 83 R*/rod/s, the ON αRGC clearly began to follow the 10-Hz stimulus, reflecting the contribution of the fast cone kinetics. We plotted the goodness of fit (R²) to a 10-Hz sine wave against intensity to derive the intensity to reach a 5% threshold and measured 20 R*/rod/s (see Supplementary Text for discussion on the cone threshold with this method).

Fig. 6. Contribution of the cone signals to the ON αRGC.

(A to C) The responses of an ON αRGC to a stimulus of intensity 25 R*/rod/s (A), 83 R*/rod/s (B), or 250 R*/rod/s (C). The stimulus was delivered for 1 s, twice, first at 1 Hz and 2 s later at 10 Hz (top trace). The bottom trace shows the raw spiking data. (D to F) Raster plots from 25 consecutive recordings under each of the conditions depicted in (A) to (C) (top trace). The bottom trace is a plot of the averaged spike rate (binned over 10 ms). (G to I) Nonlinear regression analysis of the data illustrated in D (G), E (H), and F (I) and collected during the presence of the 1-s 10-Hz stimulus (i.e., from time = 3.0 s to time = 4.0 s). Data were fit to a sine equation of frequency exactly 10.00 Hz. Regression curve and goodness of fit (R²) are shown in red. (J) Goodness of fit as a function of stimulus intensity at 10 Hz. Results are from 12 cells. The red curve was the fit of all the data to the Hill equation. (K) Standardized R² values as a function of stimulus intensity. R² is standardized for each cell to its highest value. The red curve is as in (J). The intensity at threshold (5% maximum response) equals 20 ± 3 R*/rod/s. The data suggest that cones contribute to the light responses of ON αRGCs at intensities greater than 20 R*/rod/s.

These three independent methods to measure cone threshold in RGCs all produced similar results—cone threshold in αRGCs was between 20 and 175 R*/rod/s. The mean of the three independent estimates was 122 ± 51 (SEM) R*/rod/s.

Establishing the contribution of each pathway

Plotting the rod and cone data on the same axes yielded a comparison of the respective contribution of the three rod pathways and the cone pathway to the light responses of tOFF αRGCs (Fig. 7). Using the conditional Cx36 KO lines to dissect this system and defining the operating range as the interval of intensities at which the pathway contributes between 5 and 95% of the response (see example in fig. S1C), we were able to establish the operating ranges for all rod and cone inputs to tOFF αRGCs. Consistent with the literature, the primary rod pathway operates at very low light intensities and its range is ~0.05 to 50 R*/rod/s. We were able to demonstrate that the secondary rod pathway is dependent on rod/cone coupling and contributes to the tOFF αRGC light responses from ~2 to ~5000 R*/rod/s. The similarity in the results between cone-Cx36 KO and rod-Cx36 KO lines further supports the view that either KO eliminated the entry of the secondary rod pathway (rod/cone coupling). The tertiary rod pathway is the least sensitive pathway and has a range of ~50 to 50,000 R*/rod/s. The primary, secondary, and tertiary rod pathways contribute to the output of the retina, measured in αRGCs, in ascending order until the cone threshold is reached.

Fig. 7. Establishing the contribution of all rod and cone inputs to tOFF αRGCs.

(A and B) Combining the various intensity-response curves allows us to establish the threshold (arrowheads) and contribution of the different rod pathways to tOFF αRGCs. From left to right, the intensity-response profiles are from wild-type littermate control (ctl) (control + APB) and from the mutant lines in the presence of APB. The cone intensity-response profile is generated from tOFF αRGCs recordings in the GNAT1^irdr (see Fig. 5). Note the similarity of the results obtained in cone-Cx36 KO mice (A) and in rod-Cx36 KO mice (B). Thresholds calculated as intensity to elicit 5% maximum response (see Table 1 for details).

Because previous work suggested that different threshold sensitivities of RGCs may reflect an inhibitory mechanism supported by γ-aminobutyric acid (GABA) and GABA-C receptors that mask inputs by particular rod pathways (13), we tested whether the application of the nonselective GABA receptor blocker picrotoxin (100 μM, > 15 min) would alter our measurements. Our results indicate that GABA-mediated inhibition does not significantly control the threshold responses of tOFF αRGCs or the effects of APB (see fig. S3 and tables S2 and S3). We are therefore confident that our data demonstrate the true “feedforward” dynamic ranges of the rod pathways.

DISCUSSION

Summary

Using rod- and cone-specific Cx36 mutants, we demonstrate the important contribution of the rod/cone pathway to the light responses of RGCs, the output neurons of the retina, via the secondary rod pathway. Thus, the primary and secondary rod pathways can coexist at the same level of light adaptation. The secondary and tertiary rod pathways overlap to some extent, but the tertiary is less sensitive. Our data exemplify a prime example of the role played by gap junctions in signal processing in neural networks.

Below cone threshold, there are three different rod pathways through the retina, but the threshold of each one is unclear. The primary rod pathway, via RBCs, is the most sensitive, and it is often presumed that the secondary and tertiary pathways cover distinct intensity ranges in an additive way to account for the extended range of rod signaling, well into the mesopic range.

The rod/cone pathway is the entry to an important functional pathway

The absence of rod/cone gap junctions in the rod- or cone-Cx36 KO lines provides a clear and unambiguous method to investigate the secondary rod pathway. Specifically, our measurements establish the contribution of the secondary rod pathway in the dark-adapted mouse retina from ~2 to ~5000 R*/rod/s. This is in close agreement with data from recent studies where the primary and secondary inputs to third-order neurons were separated pharmacologically. The primary rod input from RBCs to AII amacrine cells is carried exclusively by AMPA receptors and can be blocked by the competitive antagonist 6,7-dinitroquinoxaline-2,3-dione (DNQX). In contrast, the secondary input enters AII amacrine cells via gap junctions with ON CBCs. At low light levels, the responses of AII amacrine cells were blocked by DNQX and thus arose from the primary RBC input. DNQX-independent input from the secondary rod pathway had a higher threshold [8 to 16 R*/rod/s (27) or 0.5 R*/rod/s (28)]. In multielectrode array recordings from primate retina, blocking with APB revealed a secondary threshold of 1.4 R*/rod/s (21). However, these studies did not consider the tertiary rod pathway.

The tertiary rod pathway is the least sensitive rod pathway

In addition, the present work provides data on the tertiary rod pathway, supported by direct contacts between rods and certain OFF CBCs. Among these are types 3A and 3B, which stratify at the same depth as the tOFF αRGC (23). In the presence of APB, to block the primary rod pathway and using a rod- or cone-specific Cx36 KO mouse to remove the secondary rod pathway, the ganglion cell threshold, ~60 R*/rod/s is still below the cone threshold, ~120 R*/rod/s, and thus defines the threshold of the tertiary rod pathway. Our estimates of the tertiary rod pathway sensitivity threshold in the two mutant lines are close: 63.8 ± 8.44 R*/rod/s and 62.2 ± 8.12 R*/rod/s in the cone-specific and rod-specific mutants, respectively (means ± SEM, from Table 1). This further supports validation of our lines as genetic models that lack a secondary rod pathway. We did not attempt to isolate the tertiary rod pathway in the pan-Cx36 KO as the sensitivity of the rod pathways may be altered in this line (9). In the original report of the tertiary rod pathway (12), the OFF responses of RGCs in coneless mice were measured in the presence of APB at an intensity between 10 × 10³ and 4 × 10³ R*/rod/s, with the latter representing rod saturation (29). Overall, our data are consistent with previous evidence, suggesting that the tertiary rod pathway is the least sensitive rod pathway and its threshold is ~60 R*/rod/s.

Potential limitations

We used APB to eliminate the primary rod pathway input to tOFF αRGCs. However, because APB is selective for the mGluR6 receptor, it blocks all ON pathways through the retina, including those driven by ON CBCs (19). Under mesopic/photopic conditions, AII amacrine cells are driven by cone-originated signals through their electrical coupling with ON CBCs (30). This pathway constitutes a crossover inhibition pathway that allows ON signals to inhibit OFF RGCs (30). Relief of this inhibition or disinhibition at light offset contributes to the responses of OFF RGCs (31). It follows that APB likely affects our measurements of the sensitivity of the secondary and/or tertiary rod pathways by blocking crossover inhibition. However, we note that, at high contrast, the excitatory input from OFF CBCs dominates [figure 5 of (31)]. Because the stimulus used in our experiments was a full-field flash from a zero background, we would expect the OFF CBCs to provide the major input. Under these conditions, the block of the crossover inhibition pathway by APB may only have a minor effect on the intensity response function.

APB may also affect other light-driven processes, such as dopamine release, which, in turn, would modulate OFF RGCs. Although this issue is inherently present and unavoidable when the goal is to block the ON pathway, for the reasons that we detail below, we believe that these additional pathways have a minor effect on the data and conclusions of our paper.

First, disinhibition of the OFF pathway provides a component of the OFF RGCs response at lights off (31), yet the excitatory input from OFF bipolar cells, which dominates at high contrast, should be unaffected by APB. For the secondary rod pathway, the excitatory input should have a threshold close to that of the rod-originated signals in cones, which, in our hands, is ~0.5 R*/rod/s (16), a value that is very close to the one that we report here (~2 R*/rod/s).

Second, the sensitivity of the scotopic electroretinogram in mutant mice, in which the primary rod pathway has been genetically eliminated, is shifted by ~1.5 log units when compared to wild-type mice (32, 33). The shift reveals the next most sensitive ON rod pathway, i.e., the secondary rod pathway, after elimination of the primary rod pathway. A ~1.5 log unit shift in sensitivity between primary and secondary rod pathways is in agreement with the sensitivity shift that we observed in wild type of tOFF αRGCs between control (no APB) and APB.

Third, it is unlikely that application of APB affected dopamine release and thereby triggered dopamine-dependent processes. Under dark-adapted conditions, dopamine release is low in the B6 mouse, which is deficient in melatonin synthesis and therefore lacks a circadian rhythm in dopamine release (34). The presence of low dopamine levels is consistent with strong photoreceptor electrical coupling (16, 35, 36). Previous work also showed that APB moderately decreased the basal level of dopamine release in the dark while totally preventing the light-induced release of dopamine (37, 38). Notably, the light-triggered release of dopamine has a very high threshold [>10,000 R*/rod/s; (38)]. Thus, although we cannot exclude that dopamine was released and/or that APB blocked dopamine release at the higher end of the range of stimulus intensities that we used, it is unlikely that these effects significantly changed our estimates of the threshold sensitivities of the rod pathways.

As emphasized above, disinhibition of the OFF pathways and other downstream effects is inherently present when using APB, and this is true for all previous experimental use. However, these additional pathways likely had only a minor effect on our data and, therefore, do not significantly alter the conclusions of our study.

The stepwise increase in threshold supports the idea that the different pathways contribute over distinct intensity ranges. However, recent studies in primate retina show that the primary rod pathway extends over a greater range, accounting for nearly all signals of rod origin, while the secondary rod pathway has a much higher threshold (28). This is a clear difference from the mouse data (28). It is still unknown how signals from the three rod pathways are merged seamlessly, but there may also be species differences that emphasize one pathway over another. The dopaminergic modulation of rod/cone coupling, which provides an entry to the secondary rod pathway, suggests that circadian clocks and light adaptation may also alter the balance of the three rod pathways (16, 39). We recently reported that rod/cone coupling can be reduced to virtually 0 pS by application of the dopamine D₂-like receptor agonist quinpirole or increased up to 1200 pS in the presence of the D₂-like receptor antagonist spiperone (16). In addition, recent studies strongly support a role for light adaptation in the modulation of rod signal processing in the retina (27, 28) and a functional contribution of the rods under photopic lighting (6, 10, 40). Thus, the position of the threshold and relative contribution of each rod pathway is expected to change with the time of day. In the present study, we took great care to minimize the impact of potential confounding factors linked to the mouse strain, time of day, and light/dark adaptive conditions. In particular, we used the B6 strain to exclude the influence of circadian modulation (see the “Minimizing the influence of light/dark adaptation and circadian time” section in Materials and Methods). Under these conditions, the resting (dark-adapted) state of rod/cone coupling is ~300 pS (16). Our data therefore establish a reference state that will help define new modalities to study the relative weighting of rod- and cone-derived signals and their routing in retinal circuits under different conditions.

In summary, we have identified the threshold in αRGCs for the three rod pathways and the cone pathway. Our data set the relative position of the three rod pathways in the fully dark-adapted mouse retina and, together with our recent study (16), validate the photoreceptor-specific Cx36 KO models. This new knowledge and new tools will be useful to further study the secondary rod pathway and the influence of light and circadian adaptive mechanisms on signal processing in retinal circuits.

MATERIALS AND METHODS

Animals

All animal procedures were reviewed and approved by the Institutional Animal Care and Use Committees (locally called Animal Welfare Committee) at the University of Texas Health Science Center at Houston. We used mice 2 to 6 months of age of either sex. The photoreceptor cell type–specific Cx36 KO lines have been fully described in a recent paper (16). The pan-Cx36 KO line was created and provided by D. Paul (Harvard University). The different mutant lines were backcrossed for more than five generations to the B6 background. C57Bl/6J mice, referred to as B6 (stock no. 000664) and GNAT1^irdr (stock no. 008811), were purchased from the Jackson ImmunoResearch laboratories. Animals were housed under standard laboratory conditions, including a 12-hour light/12-hour dark cycle.

Minimizing the influence of light/dark adaptation and circadian time

A main objective of this study was to determine the contribution of the secondary rod pathway to the light responses of RGCs. Previous work showed that electrical synapses between retinal cells, and particularly between photoreceptors, are expected to change according to light/dark adaptation and/or the action of circadian clocks (41). To minimize the influence of these confounding factors on our data, we used mice of B6 background, in which photoreceptor coupling has a weak/absent circadian component (16, 35, 42). We performed all the electrophysiology experiments during daytime and in darkness, when rod/cone coupling is maximum [~300 pS (16)]. Before an experiment, animals were dark-adapted overnight. Deep dark-adapted conditions were preserved during the entire duration of the tissue preparation using infrared light and night vision equipment. Similarly, we took great care to maintain all retinas in a deep dark-adapted state during a recording session, which typically lasted a full day (~8 to 10 hours) for each cell. For instance, after presentation of a series of different stimuli, cells were allowed to recover and to fully dark-adapt again for up to 60 min. In addition, the intensity of the light stimulus was increased incrementally with the dimmest stimuli presented at the beginning and the brightest toward the end of a session. Despite these precautions, however, we acknowledge that some cells showed evidence of loss of sensitivity at the brightest intensities tested or toward the end of a recording session. Yet, this had little effect on our threshold estimates or other early kinetics of the intensity-response function because they were typically measured early in a session and/or with dim stimuli. Thus, because the loss in sensitivity occurred after saturation, it had little effect on our measurements and, therefore, does not significantly change the conclusions of our study.

Preparation of the retinal tissue

Mice were anesthetized and euthanized using ketamine/xylazine (100/10 mg/kg, i.p.) followed by cervical dislocation. One eye was enucleated and rapidly placed in Ames’ medium with glutamine buffered with 23 mM NaHCO₃ (Sigma-Aldrich). Freshly collected retinas were isolated from sclera and epithelium and placed on a membrane filter paper (0.45-μm HAWP; MilliporeSigma), RGC side up in a recording chamber, and continuously perfused at 2 ml/min (turnover 1/min) with the bicarbonate-buffered Ames’ solution at 32°C gassed with 5% CO₂/95% O₂ to maintain pH 7.4. Tissue was allowed to recover from surgery for 60 min in the dark before the start of electrical recording. When required, drugs were applied in the superfusate. l-(+)-APB (catalog no. A1910) and picrotoxin (catalog no. P-1675) were purchased from Sigma-Aldrich.

Electrophysiology setup

The perfusion chamber was positioned on a BX51 WI Olympus microscope (Olympus, Center Valley, PA) with fixed stage placed in a light-tight Faraday cage on an isolation table. The preparation and electrode tips were visualized with infrared (>900 nm) differential interference contrast (DIC) microscopy. Recordings (current-clamp configuration with I = 0) were obtained under visual control with a 3900A amplifier (Dagan Corporation, Minneapolis, MN, USA) using Clampex 10.2 software and digitized with a Digidata 1440A interface (Molecular Devices, Sunnyvale, CA). Signals were low-pass–filtered at 1 kHz and high-pass–filtered at 100 Hz with a four-pole Bessel filter and sampled at 10 kHz. Electrodes were fashioned from borosilicate glass capillaries (outer diameter, 1.2 mm; inner diameter, 0.69 mm; Sutter Instruments, Novato, CA) and filled with Ames’ medium. We used positive pressure to give electrode tips clean access to RGC membrane.

Light stimulus

Full-field light stimulation was provided by a 175-W xenon arc lamp (Sutter Instruments). Calibrated neutral density filters and narrowband interference filters were used to control light density and stimulus wavelength, respectively. Retinas were stimulated with full-field unpolarized monochromatic (500 nm, 10 nm half-width) light, and the duration of the stimulus was 500 ms. The intensity of the unattenuated stimulus at 500 nm was 2.18 × 10⁻⁴ W/cm² or 5.49 × 10¹⁴ photons/cm² per second, converted to 2.06 × 10⁶ R*/rod/s) [for details, see (43)]. Intervals between light stimuli were adjusted so that the baseline of activity was able to fully recover to its initial level before the next stimulus. For the experiments depicted in Fig. 6, we used a 500-nm light-emitting diode (LED) as the light source. The LED signal was delivered either at 1 Hz or at 10 Hz (square signal).

Single-cell recording of αRGCs

The light responses of single αRGCs were recorded using a loose-patch technique, as previously described (44). αRGCs were identified on the basis of (i) their morphology under DIC illumination (αRGCs typically have the largest somas), (ii) their light responses, and (iii) stratification following neurobiotin injection and histochemistry. Analyses were performed using the Spike software (Spike 2 version 7.03b, Cambridge Electronic Design, UK). Action potentials were binned into 10-ms bins. The mean rate of generation of action potentials or mean firing rate (in hertz) was computed from the average number of action potentials in a given bin across trials. To calculate the OFF (or ON) response amplitude, we calculated the mean firing rate of a cell across trials before lights ON [t0, t + 1 s] and subtracted it from the mean peak value at lights OFF (or ON) [as in (8, 9, 13)]. Mean peak OFF (or ON) response amplitudes were calculated for individual cells at each of the light intensities tested and standardized to the highest amplitude (see the “Data analysis” section). Standardized data from different cells within an experimental group were then averaged.

Immunohistochemistry

At the end of a recording, the micropipette was replaced with one backfilled with 0.5% neurobiotin (Vector Laboratories, Burlingame, CA, USA) in a standard electrode solution. The new pipette was approached to the same cell, and whole-cell configuration was obtained. Neurobiotin was diffused into the cell for about 10 min. Typically, only one recording per injection was made per retina. Retinas were maintained for at least 30 min before being fixed in a solution of 4% paraformaldehyde + 0.1% glutaraldehyde in 0.01 M phosphate-buffered saline (PBS; 0.8% NaCl, pH 7.3) for 12 min at room temperature and postfixed in 4% paraformaldehyde in PBS overnight at 4°C. The following day, retinas were washed in PBS and blocked in 3% donkey serum/0.3% Triton X-100 (in PBS) overnight. Retinas were then reacted for 2 days in a solution of PBS that also contained Dylight 488–conjugated streptavidin (5 μg/ml; Jackson ImmunoResearch, West Grove, PA) and an antibody against choline acetyl transferase (1:500; Chemicon, AB15282, MilliporeSigma, Burlington, MA). Following incubation with the conjugated streptavidin and the primary antibody, retinas were rinsed in PBS and reacted with a DyLight-conjugated secondary antibody (Jackson ImmunoResearch Laboratories Inc., West Grove, PA) used at 1:600 dilution, overnight at room temperature in the dark. Thereafter, retinas were washed in PBS and flat-mounted on a microscope slide. Labeled RGCs could then be located and observed under a confocal microscope (Zeiss LSM800, Carl Zeiss Microscopy LLC, Thornwood, NY). Last, z-stacked images of the inner retina were acquired. Sections of the z stack revealed the position of the dendrites of the injected cells relative to the two choline acetyltransferase bands. Images were exported as 16-bit Tag Image File Format (TIFF) files.

Data analysis

For the intensity-response functions, the response amplitude (peak of the mean firing rate) of individual cells was plotted versus the logarithm of the stimulus intensity. The values were fitted with the Hill function in the form

$$y=R_{\text{max}}.x^n/({K_{1/2}}^n+x^n)$$

where y is the response amplitude (in hertz), R_max represents the maximal response amplitude (in hertz), x corresponds to the stimulus intensity (in R*/rod/s), n represents the slope of the curve (i.e., the Hill coefficient), and K_1/2 is the intensity evoking a half-maximal response. Values from each cell were then normalized to the calculated maximal response (R_max). For each cell, we calculated x given y = 0.05, which gave the threshold. Normalized data from all cells within an experimental group were fitted with the Hill equation. We used K_1/2 of the averaged normalized intensity-response function as a reference point to quantify shifts between curves (table S1).

To establish the contribution of the cone signals to the light responses of ON αRGCs (Fig. 6), the data were fit to a sine equation of frequency exactly 10.00 Hz in the form

$$y=y_0+A.\text{sin} (\mathrm{\pi }.(x–x_c)/w)$$

where y is the response frequency (in hertz), y₀ represents the offset, x corresponds to time (in seconds), x_c represents the phase shift (in seconds), and w the half-period set to 0.05 s (10.00 Hz). The goodness of fit (R²) was calculated for each cell at each intensity and normalized to the highest R² value. The normalized values were plotted versus the logarithm of the stimulus intensity and fitted with the Hill function in the form as above, with y being the R² value and R_max being the maximal R² value.

Statistical analyses were performed with OriginPro 8.5.1 (OriginLab). We used analysis of variance (ANOVA) or repeated measures (RM)–ANOVA to detect differences between groups. For RM-ANOVA, sphericity was assumed: the number of degrees of freedom between groups = number of groups − 1, and the number of degrees of freedom for the error = number of observations − 1. Data are shown as means ± SEM.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S3

Tables S1 to S3

View/request a protocol for this paper from Bio-protocol.