Electrical synapses in retinal ON cone bipolar cells: Subtype-specific expression of connexins

Abstract

Retinal bipolar cells are known to form a complex, interconnecting network through electrical synapses that are either heterologous (with amacrine cells) or homologous (with other bipolar cells). These electrical synapses can be functionally as important as chemical synapses because their distinct properties provide a different character for the network. Much less is known, however, about electrical synapses in retinal bipolar cells than about chemical synapses. Here we report the molecular basis for electrical synapses in retinal bipolar cells, particularly ON cone bipolar cells. We have found variable connexin 36 (cx36) expression in different types of ON cone bipolar cells: cx36 message was found in some, but not all, ON cone bipolar cells (4 of 14 cells). In one specific type of ON cone bipolar cell (BP_Gus-GFP), however, cx36 was detected in 17 of 19 cells. Moreover, we have located cx36 puncta at the axonal terminals of BP_Gus-GFP cells, and we have found that these BP_Gus-GFP-associated cx36 puncta always colocalized with AII amacrine cell processes. Molecular and immunocytochemical evidence obtained in this study also shows that connexin 45 (cx45) is not present in BP_Gus-GFP cells. Taken together, our results suggest that connexins are expressed in bipolar cells in a neuronal subtype-specific manner and that cx36/cx36 gap junctions form the heterologous electrical synapses between AII amacrine cells and BP_Gus-GFP cells. Our findings imply that visual information can be differently processed by distinct subtypes of ON cone bipolar cells via electrical synapses.

Bipolar cells, the interneurons that relay information from photoreceptors to ganglion cells in the retina, are essential for visual information processing. The flow of visual information over separate parallel pathways, such as the ON/OFF and transient/sustained pathways, depends on various subtypes of bipolar cells that employ both chemical and electrical synapses to communicate with other neurons in the retina. In the past, most research on the role of bipolar cells focused on their chemical synapses, but it has been known for more than two decades that electrical synapses are also used for signaling by these cells (1-6). In fact, these electrical synapses seem to be an essential component of certain retinal circuits, such as the rod pathway (7). To fully understand information processing in retinal bipolar cells, then, the properties of electrical synapses must be elucidated.

Electrical synapses, also called “gap junctions,” are formed by proteins called connexins that belong to a family composed of >20 members (20 connexins in mouse and 21 in human have been cloned so far). Different connexins have been shown to endow gap junctions with different properties, such as conductance, pH modulation, calcium modulation, etc. Because these functional properties depend on the molecular components that make up an electrical synapse, it is important to know which connexins are used by which retinal neurons. Thus, discovering the possible candidate connexins expressed in the bipolar cells has recently been an active area of research.

Precisely localizing connexins to bipolar cells, however, turns out to be very difficult. First, the immunolabeled connexins appear as puncta in the inner plexiform layer (IPL), and it is impossible to identify retinal neurons from just these puncta. Second, there are at least 10 types of bipolar cells in the mouse retina (8, 9), and the presence of so many subtypes makes sorting out the expression pattern of connexins even more difficult. Third, connexins are located at the junctions between two adjacent cells, and the resolution of light microscopy (even confocal microscopy) is close to the limit for determining which cellular partner contributes connexin immunostaining at those junctions.

Here we examine the expression pattern of connexins in ON cone bipolar cells. Our goal is to overcome the three problems noted above by using single-cell RT-PCR and immunocytochemistry and also by making use of a line of transgenic mice (GUS-GFP) in which a specific type of ON cone bipolar cell expresses GFP (12), referred to hereafter as BP_Gus-GFP. Although we find connexin 36 (cx36) message in only a fraction of mixed types of ON cone bipolar cells, transcripts of cx36 are found in almost all BP_Gus-GFP cells. We conclude that connexin expression occurs in a subtype-specific manner in retinal bipolar cells. Furthermore, the cx36 immunopuncta have been found in BP_Gus-GFP axonal processes, and these BP_Gus-GFP-associated cx36 puncta are always colocalized with Dab-1-positive AII amacrine cell process. Taking these findings together with the presence of cx36 message in BP_Gus-GFP cells, we conclude that cx36 is a constituent of both hemichannels in the heterologous gap junctions between BP_Gus-GFP cells and AII amacrine cells. Additionally, we show evidence that connexin 45 (cx45) is not expressed in BP_Gus-GFP cells.

Materials and Methods

Animal Preparation. All experimental procedures involving animals were approved by the University of Texas Health Science Center at Houston Animal Care and Use Committee. Animals were killed by a lethal injection with ketamine plus xylazine plus acepromazine (0.1 ml, 100 mg/ml) and were immediately decapitated.

Our protocol for dissociating retinal neurons was modified from that described in our previous work (13): 40 units/ml papain (lyophilized, Worthington), 5 mM l-cysteine, 5 mM EGTA, and 400 units/ml DNase I (Worthington) were used for a 50-min incubation at room temperature to dissociate bipolar cells. At the end of the incubation, the retina was rinsed five times with Hanks' solution and gently shaken until the tissue dissociated. The cells were placed on a 35-mm cell culture dish and were used within 2 h of dissociation.

Retinal slices were prepared as described by Werblin (14). The morphologies of Lucifer Yellow-labeled cells were viewed with epifluorescence and captured with a SenSys charge-coupled device camera (Photometrics, Tucson, AZ).

Immunocytochemistry. After fixation, the tissues were washed extensively with 0.1 M phosphate buffer (pH 7.4) and blocked with 3% donkey serum_. The antibodies (Abs) were diluted in 1% donkey serum. The retinal whole-mount sections were incubated with primary Abs for 5 days and with secondary Abs overnight at 4°C. The retinal vertical sections were incubated with primary Abs overnight at 4°C and with secondary Abs for 2 h at room temperature. The primary Abs used in this study included mouse mAbs to cx36 (1:1,000; MAB3045, Chemicon), rabbit polyclonal Abs to cx36 (1:250; 51-6200, Zymed), rabbit polyclonal Abs to Dab-1 (1:500; gift from Brian Howell, National Institutes of Health, Bethesda), and mouse mAbs to cx45 (1:500; MAB3101, Chemicon). The secondary Abs, conjugated with Cy3 or Cy5 (1:200), were obtained from Jackson ImmunoResearch.

Digital images (1,024 × 1,024 or 2,048 × 2,048 pixels) were acquired by using a Zeiss LSM 510 confocal microscope and were processed in photoshop (Adobe Systems, San Jose, CA). The colocalization analysis was done with image-pro plus and custom colocalization software (15). Boxes (18 × 18 or 36 × 36 pixels) were clipped from the image by centering a sampling box around the connexin puncta. Alignment and averaging of these boxes produced a plot of color intensity against the location of pixels in the x-y plane. This method determines the average distribution of each labeled channel around a repeated neuronal structure, in this case an immunolabeled gap junction. Controls were performed by rotating one channel by 90° or transposing one channel from left to right. The ratio of average peak intensity to intensity of GFP in the bipolar cell terminals gave the colocalization rate between connexin puncta and bipolar terminals.

Single-Cell cDNA Amplification. cDNAs were synthesized and amplified by a single-cell RT-PCR procedure as described in refs. 16 and 17. Briefly, individual cells were seeded into thin-walled PCR tubes containing 4 μl of ice-cold cell lysis buffer [1× reverse transcriptase buffer (Invitrogen), 0.5% Nonidet P-40 (United States Biochemical) containing 80 ng/ml pd(T)_19-24 (Amersham Pharmacia), 5 units/ml Prime RNase inhibitor (Eppendorf), 324 units/ml RNAguard (Amersham Pharmacia), and 10 μM each of dNTPs (zcomRoche Applied Science)]. Lysis was subsequently performed at 65°C for 1 min. First-strand cDNA synthesis was performed by adding 50 units of MMLV and 0.5 units of AMV (Invitrogen) at 37°C for 15 min, and then samples were heat-inactivated at 65°C for 10 min. Poly(A) was added to the first-strand cDNA product by using 10 units of terminal transferase (Roche Diagnostics) at 37°C for 15 min, and samples were heat-inactivated at 65°C for 10 min. Amplification of tailed cDNA was done by PCR with primer AL1 (5′-ATT GGA TCC AGG CCG CTC TGG ACA AAA TAT GAA TTC (T)24-3′) (16, 17).

PCR Analyses. Specific primers for the PCR analyses were designed with primer premier 5.0 (PREMIER Biosoft International, Palo Alto, CA) based on known mouse cDNA. Primers were located within 600 bp of the poly(A) addition site and had melting temperatures close to 52°C. (Primers were as follows: PKCα sense, GCCATCAGTAATCATGCCACT; PKCα anti-sense, GGAACCCAAACTATGCTCTT; mGluR6 sense, CCAGAATTTAAGGTACAGAACTC; mGluR6 anti-sense, GGACTCAAACAGGACAGAAG; cx36 outer sense, TGGAGGGTATCTACTCAAGCC; cx36 outer anti-sense, CAATGCTACTCTTGCCTAGTGC; cx36 inner sense, CCGTGTCAATCCCAACTTATTGTG; cx36 inner anti-sense, TGCTACTCTTGCCTAGTGCTTCAG; cx45 outer sense, CTAGCAATCCAGGCCTAC; cx45 outer anti-sense, TCTGGAAGACACAACCTG; cx45 inner sense CATCACCAAAACAACCC; and cx45 inner anti-sense, CTCCACCTTCAGAGTCCC). PCRs were performed with an initial denaturing step of 5 min at 94°C, then 35 cycles at 94°C for 30 s, 52°C for 1 min, and 72°C for 1 min, and a final elongation step of 7 min at 72°C.

Results

Expression of cx36 in ON Cone Bipolar Cells. Because of the great variety of retinal cell types, studies that seek to understand connexin properties of a single neuron type are always difficult. To study the connexin expression pattern in ON cone bipolar cells without contamination by other neuronal types, we examined the expression profile at the single-cell level. We first used isolated ON cone bipolar cells by dissociating the retina; Fig. 1A shows representative micrographs of retinal bipolar cells isolated in this way from the mouse retina. Under optimal dissociation conditions, retinal bipolar cells maintain recognizable morphologies, and rod bipolar cells (Fig. 1A Left) can be easily distinguished from cone bipolar cells (Fig. 1A Center and Right) by their knob-shaped axonal terminal and tapered dendritic terminal. The identities of the bipolar cells were further confirmed by determining the expressions of bipolar cell marker genes PKCα and mGluR6 via a single-cell RT-PCR approach. After the isolated retinal bipolar cells were collected by manual microcapture, single-cell cDNA amplifications were performed (see Materials and Methods). Then, the amplified single-cell cDNA was used for PCR with specific primers for PKCα and mGluR6. As shown by gel electrophoresis (Fig. 1B), cDNA from rod bipolar cells gave PCR products for both PKCα and mGluR6, whereas cDNA from ON cone bipolar cells gave PCR products for mGluR6 only. We successfully amplified single-cell cDNA samples from 30 bipolar cells, including 10 rod bipolar, 6 OFF cone bipolar, and 14 ON cone bipolar cells.

Fig. 1.

Single-cell gene expression analysis of retinal bipolar cells. (A) Microphotographs of dissociated retinal bipolar cells. (B) Single-cell RT-PCR results from representative bipolar cells samples show the expression of marker genes in different types of bipolar cells. (C) Gene expression profile in ON cone bipolar cells (ON CB#) by single-cell RT-PCR. Although the bipolar cell marker gene mGluR6 is detected in all of these neurons, cx36 is detected in only a fraction of ON cone bipolar cells. (D Left) Morphology of a cx36-positive ON cone bipolar cell revealed by the Lucifer Yellow-filled whole-cell patch electrode. (Right) The single-cell RT-PCR results from the same bipolar cell. ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer.

To examine the expression of cx36 in ON cone bipolar cells, PCRs with specific primers to cx36 were performed by using amplified single-cell cDNA from the ON cone bipolar cells identified above. Surprisingly, only a fraction (4 of 14) of ON cone bipolar cells was found to have cx36 transcript. On the other hand, mGluR6 was found in all these samples, as expected (Fig. 1C). Such an expression pattern could arise in two different ways: random and subtype-specific expression. The examination of morphological details of cx36-positive ON cone bipolar cells should distinguish between random and subtype-specific expression.

To examine the detailed morphology of the cx36-positive bipolar cells, retinal slice preparations were used. We combined whole-cell patch clamp recording with the single-cell RT-PCR technique. We used patch electrodes filled with a fluorescence dye (Lucifer Yellow) so that the dye would diffuse into the cell while cytoplasmic mRNA was diffusing into the recording electrode. With this strategy, we had morphological information about the cells whose mRNA we used for RT-PCR to discover which connexin(s) were expressing. In total, we successfully collected seven ON cone bipolar cells samples from retinal slice preparations. We found cx36 expressed in four of these seven ON cone bipolar cells. Among these four cx36-positive bipolar cells, three shared similar morphology as revealed by Lucifer Yellow (Fig. 1D): their axonal terminals stratified in stratum IV of the IPL with relatively narrow axonal arbors (≈20 μm), and their dendritic processes had a gradually tapering shape. These cells resemble type 7 bipolar cells (8) and GFP-expressing ON cone bipolar cells in the transgenic line GUS-GFP [BP_Gus-GFP (12)]. Moreover, the percentage of ON cone bipolar cells that are thought to be type 7 [≈25% (8)] is comparable with the fraction of ON cone bipolar cells that we found to express cx36 (29%, 4 of 14 cells).

Taken together, our observations suggest that cx36 expression in ON cone bipolar cells may be subtype-specific. Below, we further characterize cx36 expression in a uniform subtype of ON cone bipolar cells, BP_Gus-GFP neurons.

cx36 in a Subtype of ON Cone Bipolar Cells: BP_Gus-GFP. To test the hypothesis that cx36 expression in ON cone bipolar cells is cell subtype-specific, we have taken advantage of a transgenic mouse line, GUS-GFP (12). Fig. 2A presents a vertical section of the retina from a GUS-GFP mouse in which strong GFP signals are seen in the narrow-field bipolar cells whose terminals ramified in stratum IV of the IPL. Fig. 2B shows an example of an isolated BP_Gus-GFP cell prepared from the retina of GUS-GFP mice. Because BP_Gus-GFP cells can be easily identified by the fluorescence of GFP, we are able to collect cell content from a homogenous subtype of ON cone bipolar cells. Single-cell cDNA has been successfully amplified from 19 BP_Gus-GFP cells, and cx36 has been identified in 17 of 19 cells. The electrophoresis gel (Fig. 2C Upper) shows the results of PCR, with specific primers to cx36, using amplified single-cell cDNA from representative BP_Gus-GFP cells. To further confirm the results of our PCR analysis, we performed a restriction enzyme digestion. The PCR fragments of cx36 contain a unique cutting site for SspI. Gel electrophoresis (Fig. 2C Lower) shows that the patterns of digestion fragments are as predicted by the cx36 sequence. In addition, we also confirmed the PCR products in some samples by sequence. Together, then, we are confident that it is the cx36 transcript that gives rise to the PCR products we have identified. In contrast to our finding on cx36 expression in all types of ON cone bipolar cells, almost all (17 of 19) of the BP_Gus-GFP cells express cx36.

Fig. 2.

Gene expression of cx36 in BP_Gus-GFP cells from GUS-GFP mice. (A) BP_Gus-GFP is a specific subtype of ON cone bipolar cell. GCL, ganglion cell layer. Other abbreviations are as in Fig. 1. (Scale bar, 20 μm.) (B) A representative dissociated BP_Gus-GFP cell. (C Upper) Single-cell RT-PCR analysis shows that cx36 is expressed in all BP_Gus-GFP cells. ON CB#, ON cone bipolar cell. (Lower) The identities of PCR products are confirmed by restriction (RE) digestion with SspI.

To further examine the subcellular location of cx36 in BP_Gus-GFP cells, immunocytochemical studies were performed. A whole-mount retina from GUS-GFP mice labeled with cx36 Ab is shown in Fig. 3A. Two selected areas are displayed at higher magnification in Fig. 3 B and C. Colocalizations of cx36 immunopuncta (shown in red) and BP_Gus-GFP axonal terminals (shown in green) were first analyzed by counting coincidences. We found that 13% (28 of 215) of cx36 puncta were associated with GFP bipolar terminals. We also signal-averaged all of the pixels surrounding the cx36 puncta. Sampling boxes, centered on cx36 puncta, were selected with the GFP signals turned off to avoid any sampling bias. The averaged intensities were plotted against the location of pixels in the x-y plane, revealing the spatial distribution for GFP around the cx36 puncta (Fig. 3G). A sharp central peak was observed for cx36 because these puncta were chosen as the center of the sampling boxes. A small, broader central peak was observed for GFP signals. This broad peak reflects the random orientation and the large size of the bipolar terminals; the smallness of the peak was due to the existence of cx36 puncta not associated with bipolar terminals. Calculation from the intensity measurements suggests (see Materials and Methods) that ≈12% of the cx36 puncta are associated with bipolar cell terminals, in agreement with the coincidence rate (13%) calculated earlier. As a control, the GFP signals were rotated 90° relative to the cx36 signals (Fig. 3 D-F). Such manipulation destroyed the spatial relationships in the original image but provided a way to assess chance overlap in a dense image. The coincidence rate between cx36 puncta and GFP-labeled bipolar terminals decreased to only 3% (6 of 215) in this control and was mainly due to random overlaps, as indicated by the loss of the central peak for GFP signal in signal-average analysis with the control (Fig. 3H).

Fig. 3.

Immunostaining of cx36 in the whole-mount retina of GUS-GFP mice. (A) cx36 puncta (red) and BP_Gus-GFP axonal arbors (green) are superimposed. (B and C) The overlapping puncta are labeled with numbers. The two boxed areas in A are shown at higher magnification. (G) Sampling boxes (n = 134) around cx36 puncta were randomly selected from those shown in A, and the results of signal-averaging analysis are illustrated. The plot of the red channel, representing the cx36 signals, shows a sharp peak at the center, as expected because sampling boxes were aligned with cx36 puncta in the center. The plot of the green channel, representing the GFP signals, shows a small peak in the center. (D) As a control, the green image in A was rotated 90° relative to cx36 signals, and the results are shown. (E and F) The boxed areas in D are displayed at higher magnification. (D-F and H) The colocalization analysis was performed again with 134 sampling boxes selected from D. Few overlapping puncta (labeled by numbers) were observed (D-F), and no correlated peak for GFP signal was seen (H).

Similar observations were also made in retinal cross sections. In the low-power micrograph (Fig. 4A), cx36 immunostaining appears to overlap with BP_Gus-GFP axonal terminals in stratum IV of the IPL. A single confocal optical image at a higher magnification is shown in Fig. 4B. The GFP signals in Fig. 4B were transposed from left to right for the negative control (Fig. 4C). Because BP_Gus-GFP terminals are narrowly stratified in the IPL, simply rotating the image by 90° would automatically mean the structures in the two channels would no longer overlap. Transposing the image from left to right maintained the same stratification level and made a better control. By counting the coincidences, we found that 10% (17 of 165) of cx36 puncta were associated with GFP bipolar terminals, whereas the coincidence rate was only 1% (2 of 165) in the negative control (Fig. 4C). In summary, we have demonstrated quantitatively that cx36 occurs at the axonal terminals of BP_Gus-GFP cells.

Fig. 4.

Immunostaining of cx36 in the retinal vertical section of GUS-GFP mice. (A) cx36 puncta (red) are found in the GFP-labeled bipolar cell axonal terminals. (Scale bar, 20 μm.) (B) A confocal image at high magnification. The overlapping puncta are labeled with numbers. (C) As a control, the GFP image in B was transposed from left to right; few overlapping puncta were observed. (D) Dab-1 immunostaining of the inner mouse retina. (E-H) Double-labeling of the boxed area in D with cx36 and Dab-1. GFP-labeled bipolar terminals (green), cx36 immunoreactive puncta (blue), and Dab-1 immunostaining (red) are shown. The square boxes highlight the BP_Gus-GFP-associated cx36 puncta. (I) The colocalization analysis of the BP_Gus-GFP terminals and cx36 and Dab-1 immunostains. Sampling boxes (n = 77) with BP_Gus-GFP-associated cx36 puncta at the center were selected. As indicated by the surface plots, there is a high probability of finding an AII dendrite at the site of cx36 plaques on BP_Gus-GFP cells. Abbreviations are as in Figs. 1 and 2.

Because it is known that ON cone bipolar cells form gap junctions with AII amacrine cells, our observations raise an interesting question: Is cx36 in the bipolar cells used to form the gap junctions between BP_Gus-GFP and AII amacrine cells? To address this question, AII amacrine cells from GUS-GFP mice were labeled with Abs against disable-1 (Dab-1) (Fig. 4D). The same section was also immunostained for cx36, and the boxed area in Fig. 4D is shown at higher magnification in Fig. 4 E-H. As indicated by the square boxes, BP_Gus-GFP-associated cx36 puncta always overlap with Dab-1 immunostains. To further confirm this, a colocalization analysis was done. In this analysis, we turned off Dab-1 immunostaining signals and selected the sampling boxes around the BP_Gus-GFP-associated cx36 puncta. In other words, only the cx36 puncta colocalized with green bipolar terminals were selected. Of 77 BP_Gus-GFP-associated cx36 puncta we counted, 73 colocalized with Dab-1-positive AII amacrine cell processes. A sharp peak was also observed for Dab-1 signal in a signal-averaging analysis where sampling boxes were centered at the BP_Gus-GFP-associated cx36 puncta (Fig. 4I). Thus, AII amacrine cells provide the partners for the cx36 puncta on bipolar cells. Taken together with the observation that the cx36 transcript is present in the BP_Gus-GFP cells, we conclude that the cx36/cx36 gap junctions form the heterologous electrical synapses between AII amacrine and BP_Gus-GFP cells.

cx45 in ON Cone Bipolar Cells. A recent study (11) suggests that cx45 is present in multiple types of ON cone bipolar cells and that heterotypic gap junctions are formed by cx36 supplied by AII amacrine cells and cx45 provided by bipolar cells. We have found, however, that BP_Gus-GFP cells express cx36 and that they use cx36 to form gap junctions with AII amacrine cells. It is thus interesting to ask whether cx45 is also expressed in this particular subtype of ON cone bipolar cell. Accordingly, we have examined the cx45 expression in BP_Gus-GFP cells by single-cell RT-PCR and immunocytochemistry. Fig. 5A shows a retinal vertical section labeled with cx45 Ab and many red puncta, which are found throughout the IPL. We also double-labeled sections with anti-cx36 and anti-cx45 (Fig. 5B). Three selected fields (Fig. 5B Upper) from the IPL are shown at higher magnification (Fig. 5B Lower). The cx45 immunopuncta never colocalize with cx36 puncta. Occasionally, these two kinds of puncta are found near each other, but these puncta are not associated with BP_Gus-GFP terminals. In fact, we find no cx45 immunostain on BP_Gus-GFP cells as shown in whole-mount retina from GUS-GFP mice (Fig. 5C). The few colocalized puncta observed are due to random overlap because there was no central peak for GFP signal when cx45 puncta were sampled in our signal-averaging analysis (Fig. 5I). Single-cell RT-PCR performed with cx45-specific primers further confirmed our immunostaining results: BP_Gus-GFP cells do not express cx45 significantly (Fig. 5J). Overall, mGluR6 was found in 19 of 19 samples, and cx36 was present in 17 of 19, but cx45 appeared in only 3 of 19 samples. Furthermore, cx45 puncta were not present at the junctions between AII amacrine and BP_Gus-GFP cells as revealed by double-labeling the retina vertical sections from GUS-GFP mice with Abs to Dab-1 and cx45 (arrows in Fig. 5 D-H). On the other hand, we did observe that cx45 puncta next to Dab-1-positive AII amacrine cell processes (arrowheads in Fig. 5 D-H). This is consistent with our observation that cx45 puncta sometimes are located next to cx36 puncta.

Fig. 5.

Expression of cx45 in the retina. (A) Immunostainings of cx45 in retinal vertical section. (B Upper) Confocal micrographs of vertical sections through the IPL of mouse retina double-labeled for cx36 (rabbit polyclonal) and cx45 (mouse monoclonal). (Lower) The boxed areas (1, 2, and 3) in B Upper are shown at higher magnification. (C) A whole-mount section from GUS-GFP retina was labeled with cx45 Abs. (D) Confocal micrographs of vertical sections of the inner retina from GUS-GFP mice double-labeled for cx45 and Dab-1. (E-H) The boxed area in D shown at higher magnification. (I) Colocalization analysis of cx45 puncta and BP_Gus-GFP axonal terminals. (J) Gene expression profile of BP_Gus-GFP cells shows that this subtype of the ON cone bipolar cell uses cx36 rather than cx45. Abbreviations are as in Figs. 1 and 2.

We conclude, therefore, that cx45 is, at most, very infrequently used by the BP_Gus-GFP cells to form gap junctions with AII amacrine cells.

Discussion

The present study used single-cell RT-PCR and immunocytochemical approaches to investigate the molecular basis for electrical synapses made by ON cone bipolar cells in the mouse retina. We present direct evidence that connexin expression in ON cone bipolar cells is cell subtype-dependent. This finding is important because the absolute sensitivity of mammalian vision depends on the transmission of rod signals from AII amacrine cells to ON cone bipolar cells via gap junctions. Our finding implies that visual information can be differently processed by distinct subtypes of ON cone bipolar cells via electrical synapses.

Despite the well known fact that bipolar cells are coupled either heterologously (with amacrine cells) or homologously (with other bipolar cells), the role of electrical synapses in information processing by bipolar cells is not well understood. To date, the best characterized electrical synapses made by bipolar cells are those between AII amacrine cells and ON cone bipolar cells. These gap junctions play an important role in the rod primary pathway in the mammalian retina: rod signals are transmitted through rod bipolar cells, then to AII amacrine cells that couple to ON cone bipolar cells, and finally to ganglion cells. Recently, the molecular basis for these gap junctions has been under intensive investigation. cx36 puncta have been located at junctions between AII and AII amacrine cells in several species including rabbit, mouse, and rat (7, 10, 18). Also, as predicted, cx36 knockout mice have a defect in scotopic light responses (19). It has, however, been debated whether cx36 is used by bipolar cells (7, 10, 11). One view is that homotypic gap junctions formed by cx36 mediate the communication between AII amacrine cells and ON cone bipolar cells (7). Another quite different view holds that the gap junctions between AII amacrine cells and bipolar cells are heterotypic: according to this hypothesis, cx36 is used only by AII amacrine cell contribution to the bipolar/amacrine electrical synapses, and some different connexin is contributed by the bipolar cell half of the gap junction (10, 11). It is worth pointing out that both views have only taken into consideration the homomeric configuration for the hexameric connexin hemichannels. Thus, if cx36 is present in the bipolar cells, the cx36/cx36 bipolar/amacrine gap junction would be homotypic. Heteromeric gap junctions occur only between connexins of the same group (20), so one might expect cx36 to be unable to form any heteromeric hemichannels with other connexins because it is in a group by itself. However, this issue requires further study and is beyond the scope of the present work.

The first hypothesis, the homotypic hypothesis, is mainly based on the expression of a reporter gene in bipolar cells of cx36 knockout mice (in which cx36 was replaced by a reporter gene) (7). The second hypothesis, the heterotypic hypothesis, is based on the following two lines of evidence. First, different permeabilities were found for gap junctions between homologous AII/AII and heterologous AII/ON cone bipolar cells by injecting tracers of different sizes (21). In other words, these two types have different properties. Second, in the rat retina, recoverin selectively labels a specific type of ON cone bipolar cell and, in retinal dissociates, none of the recoverin-positive isolated bipolar cells were immunolabeled by cx36 Abs (10).

The evidence for the homotypic hypothesis does not completely rule out the possible existence of heterotypic gap junctions between AII amacrine and bipolar cells: it is not clear what subtypes of bipolar cells expressed the reporter gene in the cx36 knockout mice, and it is also not clear whether cx36 is the only connexin expressed in these bipolar cells. By the same token, the two lines of evidence that support the heterotypic hypothesis cannot completely rule out the possibility of homotypic channels. If the two halves of a homotypic gap junction are identical but are modulated differently in bipolar cells and AII amacrine cells, then different permeabilities might be observed between AII/AII and AII/ON cone bipolar cells even though both have cx36 homotypic gap junctions. The question also remains about the extent to which connexin expression in one subtype of bipolar cell can be generalized to other subtypes. In other words, it is possible that heterotypic gap junctions are present between AII and recoverin-positive ON cone bipolar cells and homotypic gap junctions are used by other bipolar cells.

Our finding that cx36 expression in bipolar cells is cell subtype-specific provides a reasonable explanation for the apparently conflicting results and for how the two competing hypotheses about bipolar cell gap junctions could have arisen.

Our results show that the hypothesis that AII/ON cone bipolar cell gap junctions are formed by cx36 in both hemichannels is true at least for the subtype ON cone bipolar cell, BP_Gus-GFP. We have found that cx36 message is present in BP_Gus-GFP cells by single-cell RT-PCR, and we have found cx36 puncta in axonal terminals of BP_Gus-GFP cells. Moreover, all BP_Gus-GFP-associated cx36 puncta are at the junctions between BP_Gus-GFP axonal terminals and AII amacrine cells. Because cx36 has not been found anywhere in BP_Gus-GFP cells but the junctions between BP_Gus-GFP and AII amacrine cells, our conclusion is a reasonable, logical one. The only exception would be that cx36 is never translated into protein in BP_Gus-GFP cells, and such regulation by translation has not been reported.

Our results also show evidence for the possible existence of heterotypic gap junctions between AII amacrine cells and bipolar subtypes other than BP_Gus-GFP. We have found, by single-cell RT-PCR analysis, that cx36 transcripts are not present in all ON cone bipolar cells. These observations cannot be explained simply by experimental artifact. The single-cell cDNA amplification protocol used in our studies has been shown to be a powerful tool in studying gene expression and to give an accurate representation of relative transcript abundances (17, 22). Because variability among samples cannot be completely eliminated, we have carefully characterized false-positive and false-negative events in our experimental procedure with known marker genes (unpublished work). Cell-specific markers were observed mainly in the “correct” cells and were missing from those cells with a false detection rate of ≈12%. Furthermore, the detection of cx36 in a uniform population of BP_Gus-GFP cells is close to 100%. We therefore conclude that cx36 is not used by all of the ON cone bipolar cell types and that the connexin expression in these neurons is cell subtype-dependent. For the bipolar subtypes that lack cx36, communications with AII amacrine cells must use connexins other than cx36 and must occur as heterotypic gap junctions. cx45 has recently been shown to be present in ON cone bipolar cells and might be used to form cx36/cx45 heterotypic gap junctions between bipolar and AII amacrine cells (11). However, we have shown by single-cell RT-PCR and immunocytochemistry that BP_Gus-GFP cells do not have cx45. We did occasionally observe cx45 message in unidentified subtypes of ON cone bipolar cells (data not shown). The double-labeling experiments with cx36 and cx45 show that cx36 puncta are sometimes adjacent to cx45 puncta (Fig. 5C), suggesting the possible existence of cx36/cx45 heterotypic gap junctions. Their cellular identities are, however, not yet clear. It is quite possible that those heterotypic gap junctions are used by AII amacrine cells because the double-labeling experiments with cx45 and Dab-1 show that cx45 is sometimes adjacent to AII amacrine cell processes.

Note. While this work was under review, Lin et al. (23) reported similar observations that are consistent with the results presented here.