Synaptic Plasticity in CNGA3 / Mice: Cone Bipolar Cells React on the Missing Cone Input and Form Ectopic Synapses with Rods

Silke Haverkamp,1 Stylianos Michalakis,2 Ellen Claes,1 Mathias W. Seeliger,3 Peter Humphries,4 Martin Biel,2 and Andreas Feigenspan5 1Department of Neuroanatomy, Max-Planck-Institute for Brain Research, D-60528 Frankfurt/Main, Germany, 2Department Pharmazie–Zentrum für Pharmaforschung, Ludwig Maximilians Universität, D-81377 Munich, Germany, 3Retinal Diagnostics Research Group, Department of Ophthalmology II, Eberhard-Karls University, D-72076 Tuebingen, Germany, 4Department of Genetics, Trinity College, Dublin 2, United Kingdom, and 5Department of Neurobiology, University of Oldenburg, D-26111 Oldenburg, Germany


Introduction
The mammalian retina contains rod and cone photoreceptors, precisely tuned to respond to low (scotopic) and high (photopic) light levels. Rod and cone pathways are separated at their first synapse in the outer retina (Wässle, 2004). Rods connect to a single type of rod bipolar cell, whereas cones are presynaptic to at least nine different types of cone bipolar cells (Ghosh et al. 2004). OFF cone bipolar cells make flat contacts with cone pedicles and express a conventional sign conserving glutamate receptor of the AMPA or kainate subtype (DeVries, 2000). ON cone and rod bipolar cells make invaginating contacts close to the synaptic ribbons and express the sign-inverting glutamate receptor, metabotropic glutamate receptor type 6 (mGluR6) (Nakajima et al., 1993;Nomura et al., 1994;Vardi et al., 2000).
In the mouse retina, one type of OFF cone bipolar cell makes direct synaptic contacts with rods [B2 (Tsukamoto et al., 2001); this cell type is comparable with type 3 in Ghosh et al. (2004) and CB2 in Pignatelli and Strettoi (2004)]. Direct rod to OFF cone bipolar cell connections have also been described in rat, cat, and rabbit retina (Hack et al., 1999;Fyk-Kolodziej et al., 2003;Li et al., 2004;Protti et al., 2005).
Connections between cones and rod bipolar cells have only exceptionally been reported in a normal retina (Dacheux and Raviola, 1986). However, cones make ectopic synapses with rod bipolar cells in diverse photoreceptor degeneration models [P347L pigs, retinal degeneration mice, Royal College of Surgeons (RCS) rats] (Peng et al., 2000(Peng et al., , 2003 as well as in neural retina leucine zipper (Nrl) knock-out mice in which rods fail to form and only cones exist Daniele et al., 2005).
Photoreceptors respond to light by the closure of a cyclic nucleotide-gated (CNG) cation channel, causing hyperpolarization and decrease of the synaptic glutamate release. To investigate whether cone bipolar cells form ectopic synapses in the absence of cone input, we studied the connectivity of cone bipolar cell dendrites in mice lacking CNGA3 (Biel et al., 1999), the A subunit of the cone CNG channel. It was demonstrated previously that these mice lack any cone-mediated photoresponses but have a normal rod photosystem (Biel et al., 1999).
Rhodopsin knock-out (Rho Ϫ/Ϫ ) mice, on the other hand, appear to have normal cone but no rod function, at least until postnatal week 6 (pw6) when cone degeneration is not yet substantial. These animals fail to form rod outer segments, leading to a sequence of primary rod and secondary cone photoreceptor loss that is complete at ϳ3 months postnatally (Humphries et al., 1997;Jaissle et al., 2001) We used these mice, as well as mice deficient for both CNGA3 and Rhodopsin (CNGA3 Ϫ/Ϫ Rho Ϫ/Ϫ ) (Claes et al., 2004), to test our hypothesis that the formation of ectopic bipolar cell synapses in the outer plexiform layer (OPL) requires a functional presynaptic photoreceptor.
Tissue preparation. Mice were anesthetized deeply with halothane and decapitated. The procedures were approved by the local animal care committees and were in accordance with the law of animal experimentation issued by the German Government (Tierschutzgesetz). The eyes were enucleated, the anterior segments removed, and the posterior eyecups immersion fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (PB), pH 7.4, for 30 min (LM) or 60 min (EM). After fixation, the retina was dissected from the eyecup and embedded in 2-4% agar. The agar block was mounted on a vibratome, and vertical sections of 60 m thickness were cut. For electron microscopy, the retina was cryoprotected in graded sucrose solutions (10, 20, and 30%) and frozen and thawed before vibratome sectioning. For retinal GUS-GFP whole mounts, the tissue was cryoprotected and frozen and thawed several times before applying the antibodies. Vibratome sections and whole mounts were processed free-floating.
Antibodies were diluted in PBS, pH 7.4, containing 5% Chemiblocker (Chemicon, Temecula, CA), 0.5% Triton X-100, and 0.05% sodium azide. Immunocytochemical labeling was performed using the indirect fluorescence method. Vibratome sections were incubated overnight in a mixture of primary antibodies, followed by incubation (1 h) in a mixture of the secondary antibodies, which were conjugated to either Alexa TM 488 (green fluorescence; Invitrogen) or to indocarbocyanin 3 (Cy3; red fluorescence; Dianova, Hamburg, Germany). Whole mounts were incubated for 2 d in the primary and for 2 h in the secondary antibody solution.
Confocal micrographs were taken using a Zeiss (Oberkochen, Germany) LSM5 Pascal confocal microscope equipped with an argon and a helium-neon laser. High-resolution scanning was performed with a 63ϫ/1.4 Plan-Apochromat objective (z-axis step size, 0.8 m). Brightness and contrast of the final images were adjusted using Adobe Photoshop 5.5 (Adobe Systems, San Jose, CA).
Preembedding immunoelectron microscopy. Vibratome sections were incubated for 4 d at 4°C in a primary antibody solution (anti-NK3R, 1:1000; anti-PKC␣, 1:20,000) containing 3% normal goat serum (NGS), 1% bovine serum albumin, and 0.05% sodium azide in PBS. Thereafter, the sections were rinsed in PBS and immunolabeling was detected with a biotinylated goat anti-rabbit IgG (1:100; Vector Laboratories, Burlingame, CA) and a peroxidase-based enzymatic detection system (Vectastain Elite ABC kit; Vector Laboratories). After rinses in PBS and in 0.05 M Tris-HCl, pH 7.6, the sections were reacted in 3,3Ј-diaminobenzidine (DAB; 0.05% in Tris-HCl) with 0.01% H 2 O 2 for 5-10 min. Subsequently, the sections were rinsed in Tris-HCl and then in 0.1 M cacodylate buffer, pH 7.4, postfixed in 2.5% glutaraldehyde in cacodylate buffer for 2 h at 4°C, and washed in cacodylate buffer overnight at 4°C. After several washes in distilled water, the DAB reaction product was silver-intensified by incubating the sections in a solution containing 2.6% hexamethylene tetramine, 0.2% silver nitrate, and 0.2% disodium tetraborate for 10 min at 60°C. The sections were then rinsed in distilled water and treated for 2 min with gold chloride (0.05% in distilled water). Finally, the sections were rinsed in distilled water and incubated for 2 min in sodium thiosulfate (2.5% in distilled water). The sections were then postfixed with 0.5% OsO 4 in cacodylate buffer for 30 min, dehydrated in a graded series of ethanol (30 -100%) followed by propylene oxide, and flat-embedded in Epon 812 (Serva, Heidelberg, Germany). Serial ultrathin sections were cut, stained with uranyl acetate and lead citrate, and examined with a Zeiss EM10 electron microscope.
Quantitative analysis. To quantify the amount of flat contacts of NK3R-labeled OFF bipolar cell dendrites at the rod spherules, single ultrathin sections of three CNGA3 Ϫ/Ϫ and three wild-type retinas were analyzed. Photoreceptor somas and synaptic terminals stack in multiple tiers in the mouse retina (Tsukamoto et al., 2001) (see Fig. 5). The densely packed somas form ϳ10 tiers; the underlying synaptic terminals form three to four tiers. The NK3R-labeled dendrites reach the innermost tier of synaptic terminals, where cone pedicles and rod spherules intermingle (see Fig. 2 B, C). Because the dendrites mostly stop there and barely penetrate the remaining tiers of rod spherules, we counted only the rod spherules in this innermost tier.
We did not count the number of PKC␣-labeled rod bipolar cell dendrites at cone pedicles in the Rho Ϫ/Ϫ mouse, because the penetration of the PKC␣-antibody was not optimal and we only found labeling at the surface, not enough to find a sufficient number of examples for quantification.

Intracellular injection of ON cone bipolar cells.
For the preparation of vertical sections, the retinas of C57BL/6 wild-type and CNGA3 Ϫ/Ϫ mice were removed from the eyecup and embedded in 2% agarose in Ames medium (Sigma, Deisenhofen, Germany). Agar blocks were mounted on a vibratome (Leica, Nussloch, Germany) and cut into slices of 200 m thickness. After cutting, the slices were fixed in 4% paraformaldehyde for 10 min, washed in PB, and subsequently transferred to the stage of an upright microscope (Leica) for intracellular injections.
Primary antibodies were diluted in 3% NGS and 0.3% Triton X-100 in PB and applied overnight at 4°C. After several washes in PB, secondary antibodies, dissolved in 1% NGS and 0.3% Triton X-100 in PB, were applied for 2 h at room temperature. Secondary antibodies were conjugated to Alexa Fluor 568 (Invitrogen) and Cy5 (Dianova). In all experi-ments, sections were incubated in a mixture of primary antibodies, followed by a mixture of secondary antibodies.
Confocal micrographs of fluorescent specimens were taken with a Leica TCS SL confocal microscope equipped with an argon and a helium-neon laser. Scanning was performed with a 40ϫ/1.25 Plan-Apochromat and a 63ϫ/1.32 Plan-Apochromat objective at a resolution of 1024 ϫ 1024 pixels (z-axis step size, 0.5 m). Different wavelength scans were performed sequentially to rule out cross-talk between red, green, and blue channels.

Results
The retina of CNGA3 Ϫ/Ϫ mice displays a normal general morphology and lamination (Biel et al., 1999). Only a degeneration of cones is evident over a time course of several months (Michalakis et al., 2005). To investigate the connectivity of OFF cone bipolar cells in the CNGA3 Ϫ/Ϫ retina, we used a specific antibody directed against the neurokinin-3 receptor (Ding et al., 1996).). NK3R labels type 1 and type 2 bipolar cells in the mouse retina but not the rod contacting type 3 cell, which can be labeled by antibodies against the calciumbinding protein 5 (CaB5) (Haverkamp et al., 2003;Ghosh et al., 2004).
NK3R immunoreactivity was present in bipolar cells with axons terminating in the outermost part of the inner plexiform layer (IPL) (Fig. 1 A). The processes in the inner IPL in Figure 1 A belong to NK3R-immmunoreactive amacrine cells (Haverkamp et al., 2003). There was no difference between wild-type and CNGA3 Ϫ/Ϫ retinas except for small NK3R-labeled dendrites extending into the outer part of the outer plexiform layer ( Fig. 1 A,  arrows). To determine potential contact sites of these dendrites, we double labeled vibratome sections with antibodies against NK3R and the cytomatrix protein bassoon. Bassoon labels the photoreceptor ribbons in both cone pedicles and rod spherules (Brandstätter et al., 1999). They show a horseshoe-shaped structure in rod spherules ( Fig. 1 B-F ), whereas they are clustered in a row in cone pedicles ( Fig. 1 B, frame). We found in all CNGA3 Ϫ/Ϫ mice tested (10 mice between pw3.2 and postnatal month 12) clear examples of dendrites extending into the outer OPL ( Fig. 1C-F ). In contrast, this was almost never the case in age-matched wild-type mice (Fig. 1 B). Figure 1C-F shows examples in which the tips of outgrowing bipolar cell dendrites are in close relationship to bassoon-labeled ribbons, indicating that they are in contact with rod spherules.

OFF cone bipolar cells in the CNGA3 ؊/؊ mouse contact rod spherules
At the ultrastructural level, ON and OFF bipolar cells can be clearly distinguished with respect to their contacts with cone pedicles. OFF cone bipolar cells make flat contacts at the cone pedicle base, whereas ON cone bipolar cells make invaginating contacts at the ribbons (Kolb and Nelson, 1995). In the case of the NK3R-immunolabeled OFF cone bipolar cells, we found many examples of flat contacts at the cone pedicle base of CNGA3 Ϫ/Ϫ mice (Fig. 2 A). In addition, labeled dendrites contacting rod spherules were seen frequently (Fig. 2 B, C). To demonstrate that these contacts were flat, noninvaginating synaptic contacts, we performed serial sectioning and followed the labeled dendrites in 8 -10 sections (Fig. 2C,D). Altogether, eight CNGA3 Ϫ/Ϫ mice of different age have been investigated, and in all cases, numerous ectopic synapses at rod spherules have been found. In contrast, almost no ectopic contacts were found in wild-type mouse retinas. For quantification, we counted the number of flat contacts at rod spherules from single sections of three CNGA3 Ϫ/Ϫ and three wild-type animals (see Materials and Methods). In the case of the CNGA3 Ϫ/Ϫ tissue, 207 of 1282 rod spherules made ectopic synapses with NK3R-labeled dendrites (16%); in the case of the wildtype tissue, only 6 of 867 rod spherules were possibly contacted by NK3R-labeled dendrites (0.67%).
We now asked whether the presence of these ectopic synapses depends on the activity of rod photoreceptors. To this end, we looked for ectopic synapses in retinas of CNGA3 Ϫ/Ϫ Rho Ϫ/Ϫ mice, in which both cones and rods are nonfunctional. In all double-mutant mice, we investigated (four mice between postnatal week 3.3 and 6.6), the NK3R-labeled OFF cone bipolar cells did not contact rod spherules. Apparently, cone bipolar cells form ectopic synapses with functional rods but not with rods sending no light-driven output onto second-order neurons, suggesting that the formation of ectopic synapses in the OPL requires a functional presynaptic photoreceptor.
GluR5 expression at ectopic synapses in the CNGA3 ؊/؊ mouse Connections between rods and OFF cone bipolar cells have been described in rodent and rabbit retina, and the synaptic nature of theses contacts was confirmed by the presence of the GluR1 and GluR2 receptors (Hack et al., 1999;Li et al., 2004). To proof the functional relevance of the ectopic bipolar cell synapses in the CNGA3 Ϫ/Ϫ retina, we studied the localization of the glutamate receptor subunit GluR5. GluR5 has been shown to be expressed by OFF cone bipolar cell dendrites at their contacts with cone pedicles (Haverkamp et al., 2001); however, it was not expressed at the dendritic tips of NK3R-immunoreactive bipolar cells (Haverkamp et al., 2003). We double labeled sections for GluR5 and bassoon and found a clear difference between wild-type and CNGA3 Ϫ/Ϫ retina (Fig. 3). In the wildtype retina, GluR5 was exclusively aggregated in postsynaptic clusters at the cone pedicle base (Fig. 3A) [Haverkamp et al. (2003), their Fig. 8 B-D]. In contrast, in the CNGA3 Ϫ/Ϫ retina, clear examples of GluR5 puncta were found at the positions where OFF bipolar cell dendrites contact rod spherules (compare Figs. 3 B, C, 1C-F ). The localization of GluR5 puncta at ectopic contact sides strongly indicates that the rod-cone bipolar cell synapses are functional. In addition, it shows that OFF cone bipolar cells other than NK3Rpositive cells are engaged in ectopic synapses with rod spherules.

Plasticity of ON cone bipolar cells in the CNGA3 ؊/؊ mouse
Direct synaptic contacts of ON cone bipolar cells with rods have not been described so far. This could be because of a technical problem, because immunocytochemical markers for specific ON cone bipolar cell types are not available, and chances of analyzing all ON types in a sufficient number by intracellular dye injection are low. Therefore, we concentrated on a transgenic mouse line with strong GFP expression in a single type of ON cone bipolar cell. GFP expression was also present in rod bipolar cells but was significantly weaker than in ON cone bipolar cells (GUS-GFP mouse) (Huang et al., 2003). The labeled ON cone bipolar cells resemble those termed type 7 by Ghosh et al. (2004). Figure 4 shows the dendritic trees of two type 7 cells double labeled with antibodies against GluR5 and GFP. It can be seen that all the bipolar dendrites terminate at clusters of GluR5 puncta, which represent individual cone pedicles; none of the dendrites terminates at a rod spherule (Fig. 4 B, D). This was the case for all type 7 cells we examined (n ϭ 36). Their dendrites contacted all the cone pedicles (between 6 and 10 per cell) within their dendritic field. None of the cells made a potential contact with a rod spherule.
Because ON cone bipolar cells resembling type 7 in the GUS-GFP mouse are exclusively connected to cones, we injected this bipolar cell type in the CNGA3 Ϫ/Ϫ mouse retina. In addition, we also studied the presynaptic contacts of type 5 ON cone bipolar cells. Figure 5A shows a projection of a type 5 ON cone bipolar cell with its axon terminating in stratum 3 of the IPL. When double-labeled with antibodies against kinesin and mGluR6, the presynaptic and postsynaptic components of the photoreceptor to bipolar cell synapse can be visualized. In single optical sections, cone pedicles can be easily identified according to the row-like clustering of the synaptic ribbon marker kinesin. As expected, all type 5 cells examined (n ϭ 6) contacted cone pedicles (Fig. 5B). However, this cell type also extended dendrites into close proximity of rod spherules, suggesting potential synaptic contact sites (Fig. 5C). We observed putative ectopic synapses between type 5 ON cone bipolar cells and rod spherules in all cells injected (n ϭ 6).
Injection and double-labeling of type 7 ON cone bipolar cells in the CNGA3 Ϫ/Ϫ mouse retina (Fig. 5D) also provided evidence for ectopic contacts with rod spherules. Again, the clustering of  kinesin and mGluR6 immunoreactivity is indicative for cone pedicles, and dendrites of type 7 ON cone bipolar cells clearly make synapses with cones (Fig. 5E). However, individual contact sites (in contrast to the cone pedicle clusters) suggest also putative synapses with rod spherules (Fig. 5F ). Dendrites extending from type 7 ON cone bipolar cells to rod spherules were observed in six of seven injected cells (86%).

Cone contacts of rod bipolar cells in the Rho ؊/؊ mouse
We have shown so far that cone bipolar cells react to the loss of light-driven output by establishing synaptic contacts to rod spherules. We now asked whether rod bipolar cells do also show a corresponding behavior when the rod photoreceptor output is missing. It has been shown that rod bipolar cells form ectopic synapses with cones in diverse photoreceptor degeneration models (Peng et al., 2000(Peng et al., , 2003, and one would expect to find the same in the Rho Ϫ/Ϫ mouse. To analyze their synaptic contacts at the light microscopic level, rod bipolar cells were labeled with antibodies against PKC␣ (Haverkamp and Wässle, 2000) and photoreceptor ribbon synapses with antibodies against CtBP2 (tom Dieck et al., 2005) (Fig.  6). We did not find any PKC␣-labeled dendrites closely associated with cone pedicles in the wild-type retina (Fig. 6 B), and we found hardly any examples of rod bipolar cell dendrites making potential contact with cone pedicles in the Rho Ϫ/Ϫ retina (Fig. 6C).
Analyzing the contacts of rod bipolar cell dendrites at the ultrastructural level was much more promising. We found a number of examples of PKC␣-labeled dendrites at cone pedicles in the Rho Ϫ/Ϫ mouse (Fig. 7). In the wild-type retina, rod bipolar cells make invaginating synaptic connections with rod spherules (Fig. 7A) but not with cone pedicles (Peng et al., 2000). In the Rho Ϫ/Ϫ mouse, most of the rod bipolar cell dendrites, which were followed through a series of sections at the cone pedicle base, made invaginating synaptic contact (Fig. 7B-E). This result is in contrast to the findings in rhodopsin transgenic pigs and RCS rats, in which nearly all of the ectopic cone-rod bipolar cell synapses had flat, noninvaginating characteristics (Peng et al., 2000(Peng et al., , 2003.
When looking for ectopic rod bipolar cell synapses in the CNGA3 Ϫ/Ϫ Rho Ϫ/Ϫ mouse, the success rate, again, was extremely low. We found only one positive example in three animals, indicating that rod bipolar cells form ectopic synapses with functional cones but not with those lacking cone function.

Discussion
In the wild-type mouse retina, most of the cone bipolar cells do not form connections with rods. We analyzed the dendritic connections made by cone bipolar cells in the retina of CNGA3deficient mice, in which the cone light input is specifically switched off by genetic deletion of the A subunit of the cone CNG channel. Interestingly, we found that most of the cone bipolar cells in these mice show synaptic plasticity and form ectopic synapses with rods. The demonstration of ectopic synapses between rod bipolar cells and cones in the OPL of RCS rats has been interpreted as a process of synaptic rewiring during retinal degeneration (Peng et al., 2003). Our results demonstrate that also cone bipolar cell dendrites have the capability to make alternative connections when the preferred contacts are out of function. Hence, the rules that govern synaptic partnering between rods and rod bipolar cells and between cones and cone bipolar cells are not absolute.
The molecular events that mediate the formation of normal rod and cone synapses during retinal development are poorly understood; however, an inherent molecular flexibility for forming synaptic connections may provide an adaptive advantage for the visual system. Molecular flexibility in forming synaptic contacts at the photoreceptor terminals have been explored in retinas in which the cones are genetically eliminated (Soucy et al., 1998) or in retinas in which the rods fail to form and all photoreceptors are cone-like (Nrl Ϫ/Ϫ mouse) Daniele et al., 2005). The direct connections between rods and cone bipolar cells in the "coneless" mouse can be interpreted as a reaction caused by the genetic manipulation, such that cone bipolar cells search for and make inappropriate contacts with rods in the absence of the normal synaptic target. The same holds true for the Nrl Ϫ/Ϫ mouse, in which rod bipolar cells form synaptic connections with cone-like cells that presumably were supposed to further develop into rods.
Although a large number of cones in CNGA3 Ϫ/Ϫ mice degenerate in the first postnatal months, the ultrastructure of the remaining ones appears normal (Michalakis et al., 2005). OFF cone bipolar cells make the usual flat contacts at the cone pedicle base (Fig. 2 A, B); ON cone bipolar cells and horizontal cells make invaginating contacts at the ribbon synapses [Michalakis et al. (2005), their Fig. 7]. In this study, we found several clear examples of OFF cone bipolar cells making flat contacts with rod spherules. ON cone bipolar cells also made putative contact with rod spherules, but EM reconstruction will be necessary to show whether these contacts are flat or invaginating ones. Most of the ectopic rod bipolar cell contacts we found in the Rho Ϫ/Ϫ mouse were invaginating contacts, which shows that bipolar cells keep their synaptic features at ectopic synapses in the CNGA3 Ϫ/Ϫ and Rho Ϫ/Ϫ mouse.
We have no explanation for the contrary findings of his colleagues (2000, 2003). They showed in different retinal degeneration models that most of the ectopic cone-rod bipolar cell synapses had flat, noninvaginating characteristics.
Interestingly, we found no evidence of cone bipolar cell sprouting into the ONL (NK3R immunostaining in CNGA3 Ϫ/Ϫ and CNGA3 Ϫ/Ϫ Rho Ϫ/Ϫ mice; data not shown), whereas rod bipolar cell and horizontal cell sprouting has been demonstrated in several animal models. Horizontal and rod bipolar cell processes grow into the ONL and form ectopic synapses with photoreceptors as a result of photoreceptor degeneration (Claes et al., 2004), after retinal detachment (Lewis et al., 1998), or in mutant mice deficient for bassoon (Dick et al., 2003) or CaBP4 (Haeseleer et al., 2004), which are important proteins for ribbon synapse formation or transmitter release. The molecular mechanisms mediating neurite outgrowth and the formation of ectopic synapses in these diverse animal models are unknown.
We found ectopic bipolar synapses in CNGA3 Ϫ/Ϫ and Rho Ϫ/Ϫ single knock-out mice but not in CNGA3 Ϫ/Ϫ Rho Ϫ/Ϫ double knock-out mice, which clearly shows that the formation of ectopic synapses between rod and cone bipolar cells requires functional rods and between cone and rod bipolar cells requires functional cones. Obviously, heterologous gap junctions between rods and cones (Smith et al., 1986) cannot substitute for the endogenous light-driven input mediated by CNG channels or activation of rhodopsin. This is in line with the observed degeneration of rods and cones in the respective transgenic models. In addition, we found no altered expression of connexin36 (Cx36) in transgenic animals when compared with wildtype mice (data not shown). Therefore, it seems unlikely that upregulation of Cx36 would compensate for defects in the visual transduction cascade. It is possible that the lateral spread of the electrical signal from the network of functional rods into the inoperable cone system (and vice versa) is not sufficient, because the signal is too small to mimic by itself the endogenous photoreceptor response. Although it seems reasonable to assume, we currently do not know whether or not Cx36mediated gap junctions are functional in the transgenic animals. Because gap junctional channels respond in a very sensitive manner to changes in the concentration of various intracellular metabolites, the entire loss of the signal transduction cascade could also have profound functional implications for the electrical coupling between rods and cones.
The expression of GluR5 at locations where OFF bipolar cells contact rod spherules ( Fig. 3 B, C) indicates that the ectopic rod-cone bipolar cell synapses are functional. If this is the case, the rods could provide input to both the rod-mediated and cone-mediated signaling pathways. A study of the functional properties of the ectopic synapses will require a detailed physiological analysis at the cellular level.
Three different pathways responsible for the transmission of rod signals have been postulated (Völgyi et al., 2004) (for review, see Wässle, 2004), and the gap junction protein Cx36 is essential for two of them (Deans et al., 2002). In Cx36 Ϫ/Ϫ retinas, the coupling between AII amacrine and ON cone bipolar cells (primary rod pathway with the highest sensitivity) and between rods and cones (secondary rod pathway with intermediate sensitivity) is lost (Güldenagel et al., 2001). Responses of lowintermediate sensitivity OFF ganglion cells survive in the Cx36 Ϫ/Ϫ mouse retina (Völgyi et al., 2004), and they are carried by the tertiary pathway (direct contact of OFF bipolar cells with rods). Although the rod-mediated electroretinogram of  CNGA3 Ϫ/Ϫ mice shows no anomalous features (Biel et al., 1999), the generation of CNGA3 Ϫ/Ϫ Cx36 Ϫ/Ϫ double knock-out mice combined with single unit (Deans et al., 2002) or multineuron (Meister et al., 1994) recordings might be a way to study the functional properties of the ectopic rod-cone bipolar cell synapses in more detail. Given that functional ectopic synapses between ON cone bipolar cells and rods do form in CNGA3 Ϫ/Ϫ Cx36 Ϫ/Ϫ mice, we would expect to find lowintermediate sensitivity ON ganglion cells, which do not exist in the normal ON system (Völgyi et al., 2004). Furthermore, the data of multielectrode recordings from coneless mice strongly suggest the existence of functional rod-cone bipolar cell contacts (Soucy et al., 1998). In wild-type retina, the APB-resistant OFF pathway relies mainly on electrical coupling between rods and cones, whereas in the coneless mouse retina, the APB-resistant responses can be explained if rods connect directly to OFF bipolar cells. The fact that all ganglion cells with OFF responses were ABP resistant indicates that not only the type 3 OFF bipolar cell makes functional synapses with rods (Tsukamoto et al., 2001) in the coneless mouse, but also the other OFF bipolar cell types. This would be comparable with our model in which the cone bipolar cells react after the missing cone input and form ectopic and most likely functional synapses with rods.