Introduction
There is increasing evidence that ribbon synapse formation in rod and cone photoreceptor terminals is regulated by both presynaptic and postsynaptic activities.
Several approaches using mutant mice with rod and/or cone defects have demonstrated the effects of presynaptic factors on the target exchanges from rods to cones or from cones to rods, on the finer dendritic terminal connectivity of bipolar cells, and on the high-order dendritic patterning of horizontal cells (Strettoi et al., Reference Strettoi, Mears and Swaroop2004; Reese et al., Reference Reese, Raven and Stagg2005; Haverkamp et al., Reference Haverkamp, Michalakis, Claes, Seeliger, Humphries, Biel and Feigenspan2006; Raven et al., Reference Raven, Oh, Swaroop and Reese2007; Keeley & Reese, Reference Keeley and Reese2010; Reese et al., Reference Reese, Keeley, Lee and Whitney2011). In terms of more detailed molecular architecture, the presynaptic pikachurin–dystroglycan complex is required for the formation of invaginating bipolar dendrites (Sato et al., Reference Sato, Omori, Katoh, Kondo, Kanagawa, Miyata, Funabiki, Koyasu, Kajimura, Miyoshi, Sawai, Kobayashi, Tani, Toda, Usukura, Tano, Fujikado and Furukawa2008; Omori et al., Reference Omori, Araki, Chaya, Kajimura, Irie, Terada, Muranishi, Tsujii, Ueno, Koyasu, Tamaki, Kondo, Amano and Furukawa2012).
The most upstream component of the postsynaptic G protein signaling cascade (Anderson et al., Reference Anderson, Posokhova and Martemyanov2009) in ON bipolar cells is mGluR6, which is also called Grm6. Two allelic mutants of the mGluR6 gene, nob4 (Pinto et al., Reference Pinto, Vitaterna, Shimomura, Siepka, Balannik, McDearmon, Omura, Lumayag, Invergo, Glawe, Cantrell, Inayat, Olvera, Vessey, McCall, Maddox, Morgans, Young, Pletcher, Mullins, Troy and Takahashi2007) and nob3 (Maddox et al., Reference Maddox, Vessey, Yarbrough, Invergo, Cantrell, Inayat, Balannik, Hicks, Hawes, Byers, Smith, Hurd, Howell, Gregg, Chang, Naggert, Troy, Pinto, Nishina and McCall2008), have normal a-waves but lack b-waves due to dysfunctional photoreceptor–bipolar synaptic connections. Electron microscopy performed by Cao et al. (Reference Cao, Masuho, Okawa, Xie, Asami, Kammermeier, Maddox, Furukawa, Inoue, Sampath and Martemyanov2009) showed that the occurrence frequency of the complex formed by a ribbon and two horizontal processes with invaginating bipolar dendrites in rod spherules (RS) is much lower in nob4 mutant mice (about 6%) than in control wild-type (WT) mice (about 59%). This almost 10-fold difference suggests that some RS may completely lack invaginating bipolar dendrites, although no one is certain because of the limitation of random section analysis. In contrast, our previous observation of mGluR6-KO mice by electron microscopy using serial section analysis did not find RS with a complete lack of invaginating bipolar dendrites, although this has not been verified quantitatively (Ishii et al., Reference Ishii, Morigiwa, Takao, Nakanishi, Fukuda, Mimura and Tsukamoto2009).
The effects of mGluR6 deficiency on ON bipolar cell dendrites have been characterized to a certain extent. In the rod bipolar (RB) cells of mGluR6-KO mice, the dendrite morphology is grossly normal (Masu et al., Reference Masu, Iwakabe, Tagawa, Miyoshi, Yamashita, Fukuda, Sasaki, Hiroi, Nakamura and Shigemoto1995; Tagawa et al., Reference Tagawa, Sawai, Ueda, Tauchi and Nakanishi1999), but electron microscopy shows that the invaginating dendrites are rounder and larger than in cells from WT mice and frequently contain ectopic ribbons (Ishii et al., Reference Ishii, Morigiwa, Takao, Nakanishi, Fukuda, Mimura and Tsukamoto2009). Simultaneously, the dendrites of type-7 ON cone bipolar cells express ectopic mGluR7 (Tsukamoto et al., Reference Tsukamoto, Morigiwa, Ishii, Takao, Iwatsuki, Nakanishi and Fukuda2007). However, the invagination of bipolar and horizontal cell processes has not been quantitatively elucidated.
In this study, we counted the exact number of invaginating bipolar dendrites in rod and cone terminals using serial-section electron microscopy and compared the findings from WT and mGluR6-KO mouse retinas. We also attempted to clarify the retrograde effects on synapse formation in rods and cones as shown by other researchers in Gβ3-KO (Dhingra et al., Reference Dhingra, Ramakrishnan, Neinstein, Fina, Xu, Li, Chung, Lyubarsky and Vardi2012), Gβ5S-KO (Rao et al., Reference Rao, Dallman, Henderson and Chen2007; Cao et al., Reference Cao, Masuho, Okawa, Xie, Asami, Kammermeier, Maddox, Furukawa, Inoue, Sampath and Martemyanov2009), and RGS7:RGS11 double-KO mice (Cao et al., Reference Cao, Pahlberg, Sarria, Kamasawa, Sampath and Martemyanov2012; Shim et al., Reference Shim, Wang, Chen, Chau, Fu, Yang, McQuiston, Fisher and Chen2012), which have defects in the same G-protein signaling cascade. Our study provides a more definitive morphological characterization than previous studies that used random section analysis. We compared our data to those published previously by replicating their random section strategy using our samples as well as using our own serial section strategy. This comparison helps clarify the seemingly contradictory findings of these earlier analyses.
Materials and methods
Tissue preparation and electron microscopy
A series of 300 radial sections were prepared from the retina of a 30-week-old mGluR6-KO mouse (129/SvJ × C57Bl/6J hybrid; Masu et al., Reference Masu, Iwakabe, Tagawa, Miyoshi, Yamashita, Fukuda, Sasaki, Hiroi, Nakamura and Shigemoto1995). As a control, another series of 366 radial sections was prepared from the retina of a 9-week-old WT mouse (C57BL/6J, Japan SLC, Inc., Shizuoka, Japan; Tsukamoto et al., Reference Tsukamoto, Morigiwa, Ueda and Sterling2001). We observed previously that the basic architecture of rod and cone synapses in mice appears almost constant during adulthood (from age 8 to 90 weeks; Ishii et al., Reference Ishii, Morigiwa, Takao, Nakanishi, Fukuda, Mimura and Tsukamoto2009). Thus, our analysis of the retinas of these adult mice did not address the developmental sequences of synapse formation; rather, the focus was on the methodology used to assess structural differences. The tissues were prepared as described previously (Tsukamoto et al., Reference Tsukamoto, Morigiwa, Ueda and Sterling2001; Ishii et al., Reference Ishii, Morigiwa, Takao, Nakanishi, Fukuda, Mimura and Tsukamoto2009). Briefly, the mice were anesthetized deeply using sodium pentobarbital (45 mg/kg i.p.) before perfusion with 4% paraformaldehyde and 1% glutaraldehyde in cacodylate buffer (pH 7.4; Sigma-Aldrich Japan, Tokyo, Japan). Fragments excised from the posterior retina were immersed in 2% paraformaldehyde and 3% glutaraldehyde, heated in a microwave oven for 10 min, and incubated in fixative for either 3 h at room temperature or overnight at 4°C. The tissue was postfixed with 2% osmium tetroxide for 1 h, stained en bloc with 3% uranyl acetate in 80% methanol, dehydrated with ethanol, and embedded in araldite (Nisshin EM, Tokyo, Japan). This procedure was performed in compliance with the Guide for the Care and Use of Experimental Animals at the Hyogo College of Medicine. The 90-nm serial sections were mounted consecutively on Formvar-coated single-slot grids, stained with uranyl acetate and lead citrate, and photographed at 3000× using JEM1200EX and JEM1220 electron microscopes (Jeol Ltd, Tokyo, Japan) at the Joint-Use Research Facilities, Hyogo College of Medicine.
Serial and random section sampling procedures and statistical analysis
Serial and random section samples were compared based on two series of micrographs of WT and mGluR6-KO mouse retinas. The occurrence frequency by serial section sampling was determined by examining all serial sections from one end to the other of each rod spherule for a total of 30 rods (Fig. 1A). We call this procedure “Serial 28” because the average number of sections in each rod spherule was 28. As illustrated in Fig. 1A, Serial 28 sections included the marginal areas of a spherule at both ends, which weakened the discrimination power. Therefore, we also used a revised procedure called “Serial 22” whereby three marginal sections were omitted from each end of a spherule (i.e., a total of six sections were omitted); thus, an average of 22 sections were used to evaluate the occurrence frequency of specific structures. To replicate the random section sampling procedure, we used computer-generated numbers to collect 10 sections from each of the three laterally nonoverlapping areas, yielding a total of 30 sections. The sections had to be at least 10 sections apart. The frequencies of specific structures were calculated by observing 20 spherules per section (22 × 33 μm, Fig. 1B). The occurrence frequency obtained from 20 spherules per section corresponded to the occurrence frequency obtained from serial sections per spherule (28 or 22 sections on average). Thus, the number of occurrence frequency values was 30 for each dataset. We tried to replicate the random section analysis of cone pedicles (CPs) in the same way as that of RS, but we could not obtain a reliable dataset because of the small sample size.
Illustrations of serial (A) and random (B) section sampling strategies. (A) An entire rod spherule spanned an average of 28 serial sections. We determined the occurrence frequency of specific structures by counting the number of sections with those structures. When all sections were used for the analysis, the sampling method was termed “Serial 28.” When we omitted six sections, three from the marginal areas of the two ends of the spherule, the sampling method was termed “Serial 22.” We obtained 30 frequency values as a dataset from 30 spherules. (B) We observed 20 RS characterized by cytological features (ribbon, bipolar, and horizontal cell processes, nucleus, cytoplasm, etc.) in the rectangular area of a randomly chosen section. This illustration depicts the various profiles, such as spherules with synaptic ribbons (S1–S13), with bipolar and horizontal processes (S1–S8, S14–S15) and with ribbons, bipolar, and horizontal processes (S1–S8). Of the spherules (S1–S20), 65% had synaptic ribbons (13/20) and 50% had bipolar and horizontal processes (10/20). Of the spherules with ribbons and horizontal processes (S1–S12), 67% had ribbons, bipolar, and horizontal processes (8/12). We obtained 30 frequency values as a dataset from 30 sections.
In addition, we performed a serial section analysis of 300 whole RS from each series to classify bipolar invagination types (i.e., 0B, 1B, and 2B) and of 20 whole CPs from each series to count the ribbons and bipolar dendrites. We also measured the ribbon surface area per face of 12 rods in each series using 3D reconstruction. We chose the 12 rods from a collection of 30 rods examined by Serial 28 and 22 procedures; these 12 rods fulfilled the criterion that their ribbon surfaces were at nearly right angles to the cutting plane.
The statistical analysis was performed using Statistica 06J (StatSoft Japan, Tokyo, Japan), and datasets were compared with the Student’s t-test (unpaired, two-tailed), χ 2 test, and the Pearson’s correlation test, as indicated. The data were expressed as mean ± standard deviation (s.d.) in the text or as mean ± standard error (s.e.) in figures. Differences for which P < 0.05 were considered significant.
Results
Ribbon synapses in the RS of the WT mouse retina
There are four known types of connectivity among RB and horizontal cells and RS in the WT mouse retina. In our previous analysis of WT mice (Tsukamoto & Omi, Reference Tsukamoto and Omi2013), the two-bipolar and two-horizontal (2B2H) type connection was most common, making up 94% of all connections. This complex of two bipolar and two horizontal cell processes (Figs. 1A and 2D) was once named a “tetrad” based on the 3D-reconstructed image (Rao-Mirotznik et al., Reference Rao-Mirotznik, Harkins, Buchsbaum and Sterling1995). Later, Migdale et al. (Reference Migdale, Herr, Klug, Ahmad, Linberg, Sterling and Schein2003) recognized that even a “tetrad” had two locally specialized “ribbon synaptic units.” When we observed the tetrads, their two ribbon synaptic units were most frequently in different sections—only occasionally were they together in a single section (Fig. 2A and 2D-2). The postsynaptic side of each unit consisted of one central RB element and two lateral horizontal elements. This postsynaptic complex was conventionally called a “triad” (e.g., Strettoi, Reference Strettoi, Masland and Albright2008) in the context of describing images in single sections. Because the authors of reports with which we tried to compare our data also used the term “triad,” we use this terminology as necessary. Inside the spherule, each invaginating horizontal process bifurcates into lobular elements directed toward two ribbon-associated ridges. Thus, two horizontal processes comprise two lateral elements at both ribbon synaptic units (Fig. 2D). Peichl and Gonzalez-Soriano (Reference Peichl and Gonzalez-Soriano1994) reported that these horizontal processes are derived from the axonal terminal of B-type horizontal cells. These bipolar and horizontal processes demarcated a pouch-like invagination that was formed by the continuous folding of the spherule cell membrane. In an atypical case (3%, 1B2H-type), one invaginating RB dendrite formed a Y-shaped terminal directed at two ridges inside the rod spherule (Fig. 2B), which has also been described in cats (Boycott & Kolb, Reference Boycott and Kolb1973; Linberg et al., Reference Linberg, Cuenca, Ahnelt, Fisher and Kolb2001). In another atypical case (2%), one RB dendrite bifurcated outside the rod spherule and subsequently invaginated the rod spherule. In yet another atypical case (1%), one RB and one type-7 ON cone bipolar cell had two invaginated bipolar dendrites in the rod spherule (Tsukamoto et al., Reference Tsukamoto, Morigiwa, Ishii, Takao, Iwatsuki, Nakanishi and Fukuda2007). The two latter cases were considered 2B2H-type in the context of this particular study because two bipolar and two horizontal processes were observed in a spherule.
Electron micrographs of the synaptic areas of RS: radial (A–C) and tangential (D–F) sections; cells from WT (A, B, D) and mGluR6-KO (C, E, F) mice. (A) A spherule containing two ribbon synaptic units, each comprising a ribbon (arrow), a bipolar dendrite (B1 or B2), and two horizontal processes (H1 and H2). (B) A single Y-shaped bipolar dendrite (B1) invaginating a spherule with two horizontal processes (H1 and H2). (C) A single enlarged bipolar dendrite (B1) forms two ribbon contacts (arrows), each of which is accompanied by a pair of horizontal processes (H1 and H2). (D) The first (D-1) and fifth (D-2) sections in a series show that each of two horizontal processes (H1 and H2 in D-1) bifurcates into two terminal endings to participate in the formation of two ribbon synaptic units (D-2). A long ribbon (arrows in D-1) also extends into two ends (arrows in D-2). (E) An enlarged bipolar dendrite (B1) contains a ribbon-like structure (arrowhead) in the cytoplasm. (F) 1B2H-type (upper large arrow) and 2B2H-type (lower large arrow) bundles are close to the RS. The process extending from a CP forms a focal gap junction with the rod spherule (small arrows). M, lamellar processes of Müller cells.
Invaginating RB dendrites in the mGluR6-KO mouse retina
The dendrites of RB cells were generally larger inside the spherules of the mGluR6-KO mouse than inside the spherules of the WT mouse. As shown in Fig. 2C, the single invaginating bipolar dendrite was so swollen that it occupied the entire space between two ribbon-associated ridges where the active zones were assumed to be. One or more ectopic ribbons were observed frequently in the cytoplasm of these enlarged invaginating dendrites (Fig. 2E). The enlarged dendrites and ectopic ribbons in mGluR6-KO mice have been described in detail previously based on cytological evidence (Ishii et al., Reference Ishii, Morigiwa, Takao, Nakanishi, Fukuda, Mimura and Tsukamoto2009). These single enlarged bipolar dendrites indicate the presence of 1B2H-type spherules. The 1B2H-type bundles of processes were also observed frequently in the invagination aperture at the spherule base (Fig. 2F).
Quantification of invaginating bipolar dendrites
Using serial sections of 300 RS, we compared the occurrence frequencies of 0, 1, or 2 bipolar dendrites (0B, 1B, or 2B) in the RS in WT and mGluR6-KO mouse retinas. We identified only eight 1B-type spherules (2.7%) in the WT mouse versus 86 1B-type spherules (28.7%) in the mGluR6-KO mouse (Fig. 3A; χ 2 value = 76, P* < 0.0001). In contrast, there were almost always two invaginating horizontal processes. The distributions of the 2B2H- and 1B2H-type spherules are shown in Fig. 3B. In the WT and mGluR6-KO mouse retinas, all of the RS had at least one invaginating RB dendrite. Based on our observational data and assuming a Poisson distribution, the occurrence probability of 0B-type RS was less than 1% (P* = 0.0498; n = 300 spherules).
(A) Frequencies of 0B-, 1B-, and 2B-type bipolar dendrite invaginations in 300 RS. There were 10-fold more 1B-type spherules in the mGluR6-KO mouse (28.7%) than in the WT mouse (2.7%; χ 2 test, P* < 0.0001). No 0B-type spherules were observed. (B) Schematic top view of the distributions of 2B2H- and 1B2H-type RS in an area of the retina.
We next compared serial and random section sampling procedures to estimate the occurrence frequencies of ribbon and triad morphological features as follows.
Statistical analysis of ribbon and triad morphological features
First, we analyzed the occurrence frequencies of sections that contained spherules with a ribbon in datasets from Serial 28 and random sections (Fig. 4A). Using Serial 28 section analysis (Fig. 1A), the occurrence frequency of spherules with a ribbon was 4% lower in the mGluR6-KO mouse (mean ± s.d., 38 ± 11%) than in the WT mouse (42 ± 16%); this difference was not statistically significant (t-test, P = 0.29). Using random section analysis (Fig. 1B), the occurrence frequency of spherules with a ribbon was 8% lower in the mGluR6-KO mouse (57 ± 13%) than in the WT mouse (65 ± 10%), and this difference was statistically significant (*P = 0.01). Thus, estimates in WT mice using Serial 28 sampling were 23% lower than those using random sampling (i.e., 42 vs. 65%), and estimates in mGluR6-KO mice were 19% lower (i.e., 38 vs. 57%). The Serial 28 section dataset included the entire marginal area of a rod spherule (e.g., the first, second, and last sections in Fig. 1A). In contrast, this marginal area would have been underrepresented in the random section dataset because it was not discernible due to the small surface contours and a lack of characteristic intracellular structures (Fig. 1B). These findings prompted us to perform Serial 22 analysis as well. The Serial 22 analysis found that the occurrence frequency of spherules with a ribbon was 12% lower in mGluR6-KO mice (43 ± 12%) than in WT mice (55 ± 21%); this difference was statistically significant (*P = 0.009). Thus, random and Serial 22 section analyses effectively detected differences between WT and mGluR6-KO mice. In contrast, no significant difference was detected using Serial 28 section analysis, which contained some redundant sections.
Comparisons of various parameters in retinas from a WT (black bars) and a mGluR6-KO (gray bars) mouse. (A) The frequency of ribbons in spherules as measured by Serial 28, Serial 22, and random section sampling. The difference in ribbon frequencies was statistically significant in the two phenotypes using Serial 22 and random section sampling but not using Serial 28 section sampling. (B) The surface area per face of rod ribbons as measured by serial reconstruction. (C) The frequency of triads (bipolar and horizontal processes) in spherules as measured by Serial 28, Serial 22, and random section sampling. (D) The frequency of bipolar dendrites in complexes of ribbon and horizontal processes as measured by serial and random section sampling. * indicates statistically significant differences. Bar, s.e. (n = 30 in A, C, and D; n = 12 in B).
Second, we measured the surface area of rod ribbons using 3D reconstruction of serial sections (Fig. 4B) because the lower estimates of ribbon occurrence frequency indicated that rod ribbons were smaller in the mGluR6-KO mouse than in the WT mouse. The ribbon surface area per face (one of two surfaces) was about 23% smaller in the mGluR6-KO mouse (mean ± s.d., 0.61 ± 0.19 μm2) than in the WT mouse (0.79 ± 0.12 μm2), which was a statistically significant difference (t-test, *P = 0.011).
Third, we compared the occurrence frequencies of spherules with invaginating processes of bipolar and horizontal cells (triads) in Serial 28, Serial 22, and random section samples (Fig. 4C). Serial 28 analysis indicated that the occurrence frequency of triads in the spherule was 5% lower in the mGluR6-KO mouse (22 ± 7.4%) than in the WT mouse (27 ± 7.1%). Serial 22 analysis indicated that the occurrence frequency of triads in the spherule was 13% lower in the mGluR6-KO mouse (25 ± 2%) than in the WT mouse (38 ± 2%). The differences between WT and mGluR6-KO were statistically significant using both section sampling methods (t-test, *P = 0.006 for Serial 28 and *P < 0.0001 for Serial 22). Similarly, random section analysis showed that the occurrence frequency of triads in the spherule was 10% lower in the mGluR6-KO mouse (33 ± 9.9%) than that in the WT mouse (43 ± 9.2%); this difference was statistically significant (*P < 0.001). Thus, the triad occurrence frequencies in the mGluR6-KO mouse ranged from 66 to 81% (22/27 = 0.81, 25/38 = 0.66, and 33/43 = 0.78) compared with those in the WT mouse.
Forth, we estimated the occurrence frequency of the complex of ribbon and horizontal processes that accompanied bipolar dendrites. The numerator of the ratio for this frequency measurement was the number of sections that contained three structures: a ribbon, horizontal, and bipolar processes. The denominator of the ratio was not the number of sections containing the entire spherule but rather just the number of sections that contained two structures: a ribbon and horizontal processes (i.e., no bipolar dendrites). Therefore, if the denominator remained constant, this ratio would decrease as the number of bipolar dendrites decreased. Fig. 4D shows the estimates of this ratio in percentage. The estimate using serial section analysis was 2% lower in the mGluR6-KO mouse (58 ± 20.4%) than in the WT mouse (60 ± 18.3%), and the estimate using random section analysis was also 2% lower in the mGluR6-KO mouse (60 ± 15.8%) than in the WT mouse (62 ± 13.9%). The differences between the WT and mGluR6-KO mice were not significant for either section sampling method (t-test, P = 0.59 for serial section analysis or P = 0.55 for random section analysis). This result could suggest that relative to the WT mouse, the space occupied by the complex of a ribbon and horizontal processes had also decreased in the mGluR6-KO mouse to an extent similar to the decrease in the space occupied by the complex of a ribbon, horizontal, and bipolar processes.
Reduction in ribbon synapses in CPs in the mGluR6-KO mouse
In CPs, an array of ribbons is more irregularly spaced from the pedicle base in the mGluR6-KO mouse than in the WT mouse (Fig. 5A and 5B). We examined 20 whole CPs in each series and compared pedicle size measurements on two orthogonal axes. The geometrical mean of depth and width was 5.1 ± 0.6 µm in the WT mouse and 5.2 ± 0.6 µm in the mGluR6-KO mouse. Thus, there was no significant difference in pedicle size in the two genotypes (t-test, P = 0.73).
(A, B) Electron micrographs of ribbon synapses in CPs. Each ribbon faces a triad of one bipolar (B) and two horizontal (H) cell processes. (C) The number of ribbons (left) or invaginating bipolar dendrites (right) per whole CP. * indicates statistically significant differences. (D) Pearson’s correlation between the number of ribbons and the number of invaginating bipolar dendrites in CPs (upper, WT; lower, mGluR6-KO).
The serial section analysis revealed the number of ribbons per pedicle to be 9.8 ± 1.6 (mean ± s.d.) in the WT mouse and 7.9 ± 1.5 in the mGluR6-KO mouse (Fig. 5C, left bars). The difference of 1.9 was statistically significant (t-test, *P < 0.001). The number of invaginating dendrites of ON cone bipolar cells per pedicle was 11.8 ± 1.0 in the WT mouse and 9.5 ± 1.5 in the mGluR6-KO mouse (Fig. 5C, right bars). The difference of 2.3 was statistically significant (*P < 0.0001). The invaginating bipolar dendrites slightly outnumbered the ribbons because the ribbons were frequently accompanied by more than one invaginating bipolar dendrite. Two bipolar dendrites had to fully invaginate the pedicle basal pouch and be close to the ribbon-associated ridge in order to be counted as two invaginating bipolar dendrites.
The number of invaginating bipolar dendrites correlated with the number of ribbons in CPs in the WT mouse (Pearson’s correlation coefficient: r = 0.69, *P < 0.001) and in the mGluR6-KO mouse (r = 0.59, *P = 0.006). There were never fewer invaginating bipolar dendrites than ribbons (Fig. 5D).
Discussion
Bipolar dendrite invagination in spherules and pedicles
The percentage of 2B-type spherules was 97% in the WT mouse compared to 71% in the mGluR6-KO mouse, and the percentage of 1B-type spherules was 3% in the WT mouse compared to 29% in the mGluR6-KO mouse. We never observed spherules with no RB dendrites (Fig. 3A). The single, invaginating bipolar dendrites in the mGluR6-KO mouse were usually enlarged at the tip (Fig. 2C), which may suggest that steric hindrance plays a role in the formation of 1B-type spherules. In this scenario, invagination of the second dendrite is prevented by enlargement of the first bipolar dendrite. In addition to this hypothetical explanation of the mechanism underlying the reduction of bipolar invagination, it is also possible that the underdeveloped rod ribbons play a role. The ribbon surface area in the mGluR6-KO mouse was 23% smaller compared to the WT mouse. Because the presynaptic ribbon-associated ridge is the target of the bipolar dendrites as they invaginate, it could be that well-developed presynaptic architecture is critical for normal bipolar invagination. Further work is needed to elucidate the underlying mechanism.
The number of invaginating bipolar dendrites per CP was 19% lower in the mGluR6-KO mouse (9.5) than in the WT mouse (11.8). This difference was equal to the 19% decrease in the number of ribbons per CP from 9.8 in the WT mouse to 7.9 in the mGluR6-KO mouse (Fig. 5C). In other words, the number of invaginating bipolar dendrites per ribbon was 1.2 in both the WT and mGluR6-KO mouse retinas. The number of invaginating bipolar dendrites correlated significantly with the number of ribbons in CPs (Fig. 5D). Therefore, it seems likely that the loss of invaginating bipolar dendrites in the mGluR6-KO mouse is closely linked to the loss of ribbons.
Quantitative differences in the effects of different mutations
To compare our data with data from previous studies by other researchers, we used the relative parameter of change from the WT mouse to the KO mouse, (WT−KO)/WT. There was a 12% decrease, determined using (65−57%)/65%, in the occurrence frequency of a rod ribbon in the mGluR6-KO mouse compared with the WT mouse using random section sampling in this study. In contrast, previous studies of rod ribbon frequencies that used random section sampling reported a 25% decrease in the Gβ5-KO mouse relative to WT (Rao et al., Reference Rao, Dallman, Henderson and Chen2007) and a 23% decrease in the RGS7:RGS11 double-KO mouse compared to WT (Shim et al., Reference Shim, Wang, Chen, Chau, Fu, Yang, McQuiston, Fisher and Chen2012). These decreases of 23 and 25% were approximately twofold greater than the 12% decrease we observed in the mGluR6-KO mouse. In the mGluR6-KO mouse, the rod ribbon remained a single entity, whereas the cone ribbons decreased in number: there was a 19% decrease in the number of cone ribbons per whole pedicle (Fig. 5C). A previous study reported a 26% decrease in the occurrence frequency of cone ribbons per pedicle section in the Gβ5-KO mouse (Rao et al., Reference Rao, Dallman, Henderson and Chen2007). These values differ moderately by a factor of 1.4.
We detected decreases of 19 and 23% in the occurrence frequency of triads in the rod spherule in the mGluR6-KO mouse compared with the WT mouse using serial and random section analyses, respectively. Random section analyses by others showed a 79% decrease in the occurrence frequency of triads in the rod spherule in the Gβ5-KO mouse (Rao et al., Reference Rao, Dallman, Henderson and Chen2007) and a 53% decrease in the RGS7:RGS11 double-KO mouse (Shim et al., Reference Shim, Wang, Chen, Chau, Fu, Yang, McQuiston, Fisher and Chen2012); these decreases were two- to threefold greater than our findings. Our serial section analysis detected a 20% decrease in the number of bipolar dendrites per CP in the mGluR6-KO mouse relative to the WT mouse. Horizontal processes almost always surround bipolar dendrites in the pedicle; therefore, this value corresponds to the occurrence frequency of triads per pedicle. Rao et al. (Reference Rao, Dallman, Henderson and Chen2007) reported a 36% decrease in the occurrence frequency of triads per pedicle in the Gβ5-KO mouse. These estimates differ by a factor of 1.8.
Taken together, these findings suggest that dysfunctional G protein signaling has a significant negative effect on ribbon formation and bipolar dendrite invagination in rod and cone terminals. Although we must be very careful when comparing results obtained under different conditions, the loss of Gβ5 or RGS7:RGS11 is presumed to cause greater changes than the loss of mGluR6.
Methodological considerations for statistical analysis
Previous studies reported contradictory results in terms of the characterization of RGS7:RGS11 double-KO mice. Shim et al. (Reference Shim, Wang, Chen, Chau, Fu, Yang, McQuiston, Fisher and Chen2012) reported significant differences in the triad frequency, whereas Cao et al. (Reference Cao, Pahlberg, Sarria, Kamasawa, Sampath and Martemyanov2012) did not find any significant differences. Cao et al. attributed this discrepancy to differences in the genetic manipulation and/or genetic background. However, it should be noted that these groups assayed different aspects of the same phenotype. Shim et al. measured the occurrence frequency of triads (bipolar dendrites and horizontal processes) in the spherule. Cao et al. first identified RS with ribbons and horizontal cell processes, then counted the normal synaptic triads that additionally contained bipolar dendrites, and finally determined the ratio of these two values.
Our analysis indicated that the spherule was invariable, whereas both the ribbon surface area and the triad of bipolar and horizontal processes were 23% smaller in the mGluR6-KO mouse than in the WT mouse (Fig. 4B and 4C, Random). Thus, the ratio of “the ribbon, horizontal, and bipolar complex” to “the ribbon and horizontal complex” hardly changed (Fig. 4D). This provides a methodological explanation for the differences between the findings of the two studies.
The proportion of the variable structure (ribbon, triad) relative to the invariable frame (spherule) is an effective parameter with good discriminatory power. The random section analysis using such a parameter can provide adequate data to detect significant differences, as demonstrated by Rao et al. (Reference Rao, Dallman, Henderson and Chen2007) and Shim et al. (Reference Shim, Wang, Chen, Chau, Fu, Yang, McQuiston, Fisher and Chen2012). However, serial electron microscopy is indispensable when the absolute numbers of subcellular structures such as ribbons and dendrites and of structural parameters such as surface area need to be measured.
Acknowledgments
We are grateful to our previous collaborators of the studies using mGluR6-KO mice, namely Drs. M. Ishii, K. Morigiwa, M. Takao, S. Nakanishi, Y. Fukuda, and O. Mimura. We also thank two anonymous reviewers for their comments on the manuscript. This work was supported in part by a JSPS Grant-in-Aid for Scientific Research (22500317) to Y.T. Y.T. designed the study and wrote the paper. Y.T. and N.O. conducted the study and analyzed the data.
