|Home | About | Journals | Submit | Contact Us | Français|
To establish dendritic arbors that integrate properly into a neural circuit, neurons must rely on cues from the local environment. The neurons presynaptic to these arbors, the afferents, are one potential source of these cues, but the particular dendritic features they regulate remain unclear. Retinal bipolar cells can be classified by the type of photoreceptor, cone or rod, forming synaptic contacts with their dendrites, suggesting a potential role of these afferents in shaping the bipolar cell dendritic arbor. In the present investigation, the role of photoreceptors in directing the differentiation of bipolar cells has been studied using two genetically modified “coneless” and “conefull” mice. Single cone (Type 7) and rod bipolar cells were labeled with DiI to reveal the entire dendritic arbor, and subsequently analyzed for several morphological features. For both cone and rod bipolar cells, the dendritic field area, number of dendritic terminals, and stratification of terminals in the outer plexiform layer (OPL) were comparable among coneless, conefull, and wildtype retinas, and the overall morphological appearance of each type of cell was largely conserved, indicating an independence from afferent specification. The presence of normal afferents was, however, found to be critical for the proper spatial distribution of dendritic terminals, exhibiting a clustered distribution for the cone bipolar cells and a dispersed distribution for the rod bipolar cells. These results demonstrate a selectivity in the afferent dependency of bipolar cell differentiation, their basic morphogenetic plan commanded cell-intrinsically, and their fine terminal connectivity directed by the afferents themselves.
Dendritic morphogenesis is facilitated by a combination of cell intrinsic and environmental factors, yet how exactly these factors affect individual aspects of the dendritic arbor remains unknown. The retina provides a model system to address this problem, as many characteristics of these arbors can be parsed and measured, including dendritic spread, stratification and branching, allowing for the factors responsible for establishing each of these features to be elucidated. Regulation of dendritic spread, for example, is known to be dependent upon homotypic interactions for some cell types (e.g. the retinal horizontal cell--Reese et al., 2005; Poché et al., 2008), although other types appear to be immune to this form of regulation (e.g. the cholinergic amacrine cell--Farajian et al., 2004; Keeley et al., 2007; the dopaminergic amacrine cell—Keeley and Reese, 2010). Homotypic interactions are not the only environmental cues that shape dendritic arbors; afferents also may play a role, through the expression of cell recognition molecules, release of trophic factors, transmission of visual activity, or a combination of such mechanisms. For instance, the stratification of certain ganglion cell arbors in the inner plexiform layer (IPL) relies on visual activity (Tian and Copenhagen, 2003), modulated through BDNF-TrkB signaling (Liu et al., 2007). Cell adhesion molecules in the retina have also been suggested to play a role (Yamagata and Sanes, 2008), promoting homophilic binding between afferents and targets. Distinguishing the source of these cues, however, is complicated, because of the complex and poorly understood connectivity between many different subtypes of ganglion cells, bipolar cells and amacrine cells in the IPL. Alternatively, exploring the afferent control of dendritic morphogenesis in the outer plexiform layer (OPL) is straightforward, by comparison, as it contains only two sources of afferents, the cone and rod photoreceptors. In this study, we have investigated the role of these afferents upon bipolar cell morphogenesis.
Retinal bipolar cells can be divided into two major classes: nine types of cone bipolar cell and one type of rod bipolar cell (Ghosh et al., 2004, Pignatelli and Strettoi, 2004), being postsynaptic to the cones and rods, respectively (but see Tsukamoto et al., 2001; Mataruga et al., 2007; Haverkamp et al., 2008). During development, the photoreceptor terminals arrive in the OPL in advance of bipolar cell differentiation (Sherry et al., 2003; Morgan et al., 2006), thereby being positioned, both spatially and temporally, to influence the growth of these bipolar processes. The dendrites of bipolar cells have also been shown to remodel following photoreceptor degeneration (Strettoi et al., 2002; Haverkamp et al., 2006), consistent with the idea that dendritic morphology is dependent upon afferent input. Those latter studies, however, demonstrate only an afferent dependency for the maintenance of dendritic morphology, not the initial acquisition of that morphology. To assess this possible role for afferent specification of dendritic morphogenesis, we have analyzed the dendritic arbors of bipolar cells in two genetically engineered mouse lines in which the composition of the ONL is altered during early development, being the “coneless” mutant mouse and the “conefull” Nrl−/− mouse.
Breeder pairs of the Gustducin-GFP mouse line, in which an 8.4kb segment upstream of the α-gustducin gene drives the expression of GFP in retinal bipolar cells (Huang et al., 2003), were obtained and bred in the Animal Resource Center at the University of California at Santa Barbara (UCSB). Offspring were crossed with two other genetically engineered mouse lines, the coneless and the conefull mice. In coneless mice, a human L-cone opsin promoter sequence drives an attenuated diphtheria toxin transgene which subsequently kills off 95% of the cones in early development (Soucy et al., 1998; Raven and Reese, 2003). In the conefull mouse, the neural retina leucine zipper (Nrl) gene is knocked out, causing all photoreceptors to adopt the cone fate (Mears et al., 2001). Since both copies of the Nrl gene must be knocked out to achieve the desired phenotype, GFP-positive, Nrl heterozygotes from the F1 progeny were crossed to produce GFP-positive, Nrl knockouts. GFP-positive littermates that did not have the genotype of interest were used as wildtype (WT) controls for each condition. Both engineered lines are on a C57BL/6 background, but the coneless line is on a mixed background of C57BL/6J and C57BL/6Ncrl. As these two lines are essentially substrains now known to differ in the density of other retinal cell types (e.g. Whitney et al., 2009), we have included WT littermate controls for each comparison, and analyzed them separately. Adult mice were used in all analyses, having an average postnatal age of 72 days (± 14 days s.d.). All experiments were conducted under authorization by the Institutional Animal Care and Use Committee at UCSB and in accord with the NIH Guide for the Care and Use of Laboratory Animals.
Mice were euthanized with sodium pentobarbital (120 mg/kg, i.p.) and eyes were immersion-fixed in 4% paraformaldehyde for 30 minutes. Whole retinas were dissected immediately, rinsed in sodium phosphate buffer, and transferred to an injection-well mounted on a fixed stage Nikon Eclipse E600 microscope. A single GFP-positive bipolar cell axon terminal in one of the two distinct strata in the IPL (figure 1b) was visualized and impaled with a micropipette filled with 0.5% DiI in 100% EtOH solution. A small amount of DiI was deposited by passing positive current through the pipette for approximately 10 seconds, at which point the dye could be seen spreading locally within the membrane of the injected terminal in the IPL (and subsequently left overnight, this lipophilic dye continues to diffuse throughout the entire membrane of the cell). After injecting multiple cells in this fashion across the retina, some retinas were subsequently labeled with peanut agglutinin (PNA) conjugated with the fluorochrome AlexaFluor647 (PNA-AF647; 1:500; #L-32460; Invitrogen; Carlsbad, CA) in PBS overnight to label the active sites of cone pedicles (Haverkamp et al., 2001). GFP-positive, DiI-labeled somata and dendritic fields were imaged within 48 hours using an Olympus Fluoview 500 laser scanning confocal microscope with a 60X oil-immersion objective at 0.5 µm intervals through the Z axis. Captured image stacks were deconvoluted and analyzed using AutoQuant vX1.4.1 (Media Cybernetics; Bethesda, MD) and Metamorph v18.104.22.168 (MDS Analytical Technologies; Downington, PA) software, respectively. Some retinas that had been labeled for DiI as above were subsequently sectioned at 150 µm on a Vibratome (The Vibratome Company; St. Louis, MO), in order to examine their dendritic morphology in the radial axis.
GFP-positive retinas were also prepared for immunofluorescence. Retinas were embedded in 5% agarose and cut into radial sections at a thickness of 150 µm on a Vibratome. Sections were stained with PNA-AF647 (1:1,000), mouse monoclonal antibodies to C-terminal binding protein 2 (1:250; #612044; BD Biosciences; San Jose, CA), and either rabbit polyclonal antibodies to Protein Kinase C (PKC; 1:10,000; #CA-1042; Cambio; Cambridge, UK) or rabbit polyclonal antibodies to Green Fluorescent Protein (GFP) conjugated to AlexaFluor488 (GFP-AF488; 1:1,000; #A-21311; Invitrogen; Carlsbad, CA) by using the following protocol: sections were rinsed in phosphate buffer and preincubated in 5% normal donkey serum and 1% Triton X-100 in PBS for 3 hours and then rinsed with PBS. Sections were then incubated with agitation for 3 days at 4°C in the primary antibodies and PNA-AF647 plus 1% Triton X-100 in PBS. Retinas were rinsed in PBS and stained overnight with donkey anti-mouse secondary IgG conjugated to Cy3 (1:200; #715-165-150; Jackson ImmunoResearch Laboratories; West Grove, PA) and donkey anti-rabbit secondary IgG conjugated to AF488 (1:200; #A21206; Invitrogen; Carlsbad, CA) to detect CtBP2 and PKC primary antibodies, respectively. Sections were imaged using an Olympus Fluoview 1000 laser scanning confocal microscope with a 40X water-immersion objective, sampling images at 1 µm intervals.
Wholemount image stacks of DiI-labeled bipolar cells were measured for the following characteristics: dendritic field area, soma area, terminal number, and terminal location in the x–y plane. Dendritic field area was determined by finding the area of the convex polygon that surrounded the dendrites of each cell; likewise, soma area was calculated by tracing the cell body at its widest extent. Terminals were defined as the tip of any dendritic branch. To analyze the spatial patterning of dendritic terminals in the plane of the retina, the x–y coordinates of each terminal was used to compute the Delaunay triangulation of the dendritic field by using the spatstat package (www.spatstat.org) in the programming language R (http://www.r-project.org/). The coefficient of variation (CV) was determined for each cell by dividing the standard deviation by the mean Delaunay triangle area. Student’s t-tests were used for each comparison with a p-value of < 0.01 determining significance.
Sections of GFP-positive cone bipolar cells were analyzed for the depth at which their branches terminated in the OPL. The boundary of the OPL, defined using CtBP2-labeled ribbon synapses, and the locations of each discernable terminal were recorded independently. Once the terminal locations were overlaid on the OPL boundary, depth was measured as the distance of the terminal relative to the distance of a line drawn radially through the terminal from the inner to outer boundary of the OPL.
As previously described by others, we found GFP expression to be limited to one class of cone bipolar cell and the single class of rod bipolar cell, each expressing at a different intensity (figure 1a) that allowed for the cells to be discriminated consistently (Huang et al., 2003; Lin and Masland, 2005): every brightly labeled GFP-positive bipolar cell extended a terminal into S4 of the IPL (arrowheads in figure 1b), and when subsequently labeled with DiI had a cone bipolar morphology (figure 2a), while every faintly labeled GFP-positive cell extended a terminal into S5 (arrows in figure 1b) and had a rod bipolar morphology (figure 7a). Based on the level of terminal stratification in the IPL, we agree with previous reports (Huang et al., 2003; Lin and Masland, 2005; Wässle et al., 2009) that the gustducin-positive cone bipolar cell most resembles the Type 7/CB4a bipolar cell as classified by Ghosh et al. (2004) and Pignatelli and Strettoi (2004), respectively.
The entire morphology of this Type 7/CB4a cone bipolar cell is particularly well-illustrated in a fortuitous section in which attempts to label the cell from the soma had proven successful (this approach frequently labeled other nearby bipolar cells as well, obscuring the detailed dendritic morphology of the targeted cell): dendritic branches emerge from the soma and splay out as they ascend to the OPL, where each gives rise to a cluster of fine terminal-like extensions (figure 2a). A single basally directed process descends radially through the INL and IPL to reach S4, where a terminal arbor extends laterally approximately 30 µm, being comparable to the width of the dendritic arbor. Bipolar axon arbors are generally thought to produce a tiling of their respective strata rather than exhibiting any overlap (Young and Vaney, 1991; Chan et al., 2001; Telkes et al., 2008), and consistent with this, we rarely obtained more than a single labeled Type 7/CB4a cone bipolar cell when targeting the terminal arbor in wholemount preparations (e.g. figure 2b).
The clusters of dendritic terminals produced by these bipolar cells are restricted to the innermost portion of the OPL (figure 2c), where cone pedicles are known to stratify. By labeling these same retinas to reveal the presence of cone pedicle active sites using PNA (figure 2d), we confirmed that every dendrite yielding clusters of terminals was associated with a cone pedicle (green circles in figure 2e), being the invaginating processes of this type of ON-cone bipolar cell (Mariani, 1981; Hopkins and Boycott, 1996). Occasionally, single terminal-like extensions were positioned away from any pedicles (arrow in figure 2c–figure 2e), which may be associated with rod spherules (Tsukamoto et al., 2007). Notice that individual dendrites do not always grow out to innervate the pedicles directly from the soma; some take conspicuous turns toward the pedicle, while other pedicles are innervated by branches from other clusters (figure 2c, figure 2d). Note as well that not all pedicles near the soma are innervated. The cell in figure 2d contacts seven different pedicles, though sparing one nearby the soma. The average number of pedicles contacted by single Type 7/CB4a cone bipolar cells was 8.7 (± 2.06 s.d.), based on a sample of 15 single-labeled cells from wildtype retinas that were also labeled for PNA. This estimate confirms that from another recent study reporting 8.4 pedicles contacted by Type 7/CB4a cells (Wässle et al., 2009).
The individual number of terminals per pedicle can also be counted in such specimens. The average was 3.7 (± 1.68 s.d), but this number varied conspicuously depending on the proximity of the pedicle to the main dendritic stalk, when evidenced by pooling the data from these 15 labeled cells (supplemental figure 1), and is consistent with evidence from the primate retina (Boycott and Wässle, 1991; Hopkins and Boycott, 1997). Near the stalk, some pedicles had as many as seven or eight terminals, while those farther removed from the stalk had as few as one or two. Those latter pedicles should be innervating a second Type 7/CB4a cell, as Wässle et al. (2009) have shown that pedicles on the fringe of the dendritic field are often connecting to two cone bipolar cells of the same type.
Figure 3 (a and c) illustrates ten different examples of cone bipolar cells from a larger sample of 95 labeled cells from wildtype retinas, and shows, for comparison, five examples from coneless retinas (b) and five examples from conefull retinas (d). What is immediately striking about these bipolar cells is how relatively normal they appear: like those in the wildtype retinas, bipolar cells in both the coneless and conefull retinas exhibit a comparable divergence of primary dendrites within the OPL, establishing a dendritic field that is similar in size to that observed in the wildtype retinas (figure 4a). These cone bipolar cells show no gross morphological atrophy of the dendritic arbor in the absence of nearly all cone pedicles (figure 3b), nor any evidence of conspicuous hypertrophy in the presence of excess numbers of cones (figure 3d). Soma size was also unaltered (figure 4b).
A closer examination of individual cone bipolar cells, however, reveals pronounced differences in the detailed connectivity associated with these cells. While cone bipolar cells in both the coneless and conefull retina established a similar number of terminals as those in wildtype retinas (figure 4c), they generally failed to exhibit discrete clusters of terminals (figures 5a, e; 5d, h; compare with 2e; see also supplemental figure 2, for the plots of individual terminals for each cell illustrated in figure 3). This relative lack of clustering is readily apparent when comparing the Delaunay triangulation associated with the positioning of these terminals within the dendritic field (figure 4d). The clustered presence of terminals increases the variability in triangle size, and so the coefficient of variation is significantly greater in the wildtype controls relative to either the coneless or conefull retinas (figure 4e).
The role of pedicles upon the positioning of dendritic terminal endings was particularly striking in coneless retinas in which single DiI-labeled, GFP-positive bipolar cells were found near a surviving cone pedicle. Such bipolar cells always showed a discrete cluster of terminals associated with single PNA-labeled active sites, to be contrasted with the general absence of this clustering phenotype at the termination sites of other primary dendrites (figure 5b, f). Note in some cases, multiple primary dendrites appear to converge on the same remaining pedicle (figure 5c, g), but in general, these “partially coneless” cells had dendritic fields that were comparable in size: these cells had dendritic fields that averaged 369 µm2 (± 132 µm2 s.d.) while "fully coneless" cells averaged 348 µm2 (± 136 µm2 s.d.), based on samples of 12 and 20 cells, respectively.
In the conefull retina, PNA labeling no longer reveals a distribution of cone pedicle active sites at the inner margin of the OPL; rather, the entire OPL is now filled with PNA, similar to the widespread presence of cone arrestin (Raven et al., 2007), as would be expected if all of the rods had now differentiated as cones (figure 6f, compare with d and e). Yet, these cones do not produce structures that appear as discrete pedicles with a stratified collection of ribbons at their basal surface (Strettoi et al., 2004), not even at the inner limit of the OPL (Raven et al., 2007). Therefore, the PNA labeling no longer provides a clear landmark for identifying individual pedicles (figure 6f), and consequently, it is difficult to associate the dendritic terminals to single pedicles. Nevertheless, it is apparent that bipolar cells in these retinas do not concentrate their terminations at specific locations as do those in wildtype retinas (figure 3d, figure 4d); instead, they appear to sprout sporadically across the dendritic field, with fewer dendrites elaborating into terminal clusters (figure 5d, h; supplemental figure 2d). Interestingly, while these cone bipolar cells extend slightly farther than they normally would (fig 6c, f, i, l), they fail to invade the more outer reaches of the OPL, that portion of the OPL normally supplied by rod spherules but now colonized by respecified cone terminals (supplemental figure 3). Indeed, single DiI-labeled cone bipolar cells in radial sections confirm this general impression, that their arbors are only slightly more widespread across the depth of the OPL relative to those in wildtype retinas (supplemental figure 4).
While we can be certain that those clustered terminals in the wildtype retina truly reflect the postsynaptic sites associated with cone innervation, we are less confident about such terminations in the coneless and conefull retina. Given the limitations of our technique, we could not combine DiI-labeling with most immunolabeling protocols in order to juxtapose, at high-resolution, the dendritic terminals with other elements of the OPL; however, when sections of GFP-positive cone bipolar cells were labeled for the presence of photoreceptor ribbon synapses, dendritic terminals were found in close proximity to these ribbons in all conditions (figure 6j-l), suggesting that in the coneless retina (figure 6k), these cone bipolar cells may be making contacts with the remaining rod spherules.
Rod bipolar cells, by contrast with these cone bipolar cells, produce a strikingly distinctive morphology. The rod bipolar cell produces a stout primary dendritic stalk that climbs to the OPL, where it quickly divides into multiple branches that continue a primarily radial trajectory (figure 7a), giving rise to a large collection of terminal endings that has limited lateral spread but extends through the full thickness of the OPL (figure 7b–e), bypassing the population of cone pedicles (figure 7c) at the inner limit of the OPL. Figure 8 (a and c) shows ten examples of such single rod bipolar cell dendritic arbors from wholemounted wildtype retinas presented as Z-stack reconstructions through the full depth of the OPL. Rod bipolar cells in the coneless retina are, not surprisingly, comparable to those in the wildtype retinas (figure 8b, compare with a). In the conefull retinas, however, their basic morphology was not conspicuously altered in these preparations, certainly not being atrophic in the absence of the rods (figure 8d, compare with c). Neither dendritic field area nor soma size was significantly different in the two experimental conditions (figure 9a and b), and nor was the total number of individual terminals throughout the full depth of the dendritic field altered (figure 9c). These morphometric analyses confirm the observations of others using the Nrl−/− mouse (Strettoi et al., 2004), on the basis of PKC immunoreactivity.
We did notice, however, that dendritic tips in wildtype and coneless rod bipolar cells end in bulbous terminations, interpreted as the expansion of these processes after they pass through the neck of the invagination at the basal surface of the spherule (Grünert and Martin, 1991). Labeled cells in the conefull retina failed to display such conspicuous puncta at the ends of the dendritic terminals (figure 8d, compare with a, b, c). Such terminals are located adjacent to PNA-labeled active sites and ribbon synapses in the conefull retina (figure 10c, f, i, and l), and are believed to invaginate the respecified cones in these retinas, since the terminals of the latter have been shown to exhibit invaginating elements thought to come from the rod bipolar cells (Strettoi et al., 2004).
Closer examination revealed that the primary dendritic stalks of the rod bipolar cells in conefull retinas were notably coarser and extended for a greater distance from the soma into the OPL (figure 10c). This was confirmed by comparing single DiI-labeled rod bipolar cells in sectioned conefull versus wildtype retinas (,supplemental figure 5). These images, like those in the PKC-labeled specimens in figure 10c, suggest that the distribution of terminal endings might be less evenly distributed within the OPL. To assess this, we identified the x,y positions of each dendritic terminal for multiple labeled cells in wholemount preparations, and conducted a comparable analysis of clustering (figure 9d). While this approach ignores all z axis variation, it shows rather clearly that the distribution of terminations for single rod bipolar cells is perturbed in the conefull retina relative to the wildtype and coneless retinas (figure 9e), despite the lack of change in the total number of terminations made by these bipolar cells (figure 9c). We infer this to be a product of larger pedicle-like structures receiving multiple invaginations, as has been shown ultrastructurally for such respecified photoreceptor terminals in the Nrl−/− retina (Strettoi et al., 2004).
The present results demonstrate that two different classes of bipolar cell, innervated by the rods or by the cones, differentiate their characteristic dendritic morphologies even when their normal afferent innervation is eliminated, or, in the case of the cone bipolar cells, significantly increased. In the conefull retinas, rod bipolar cells build small dendritic arbors that branch repeatedly, stratifying across all depths of the OPL amongst the terminals of respecified cones. Likewise, the cone bipolar cell retains its characteristic morphology, with several large dendrites emerging from a single stalk and sending smaller terminal branches to the inner edge of the OPL, despite losing its normal afferents in the coneless retina and being presented with an almost 20-fold increase of afferents in the conefull retina. It therefore appears that the gross morphology of retinal bipolar cells is likely directed by cell intrinsic factors that control their stratification and branching. The extent of the dendritic fields also did not change, but we cannot rule out a role for homotypic interactions restricting this growth (Wässle et al., 2009). Indeed, occasional cone bipolar cells avoided colonizing pedicles in close proximity to the soma, while innervating those farther away (e.g. figure 2c–e). This may reflect the presence of neighboring homotypic cells that colonize such pedicles; alternatively, it may be that such pedicles are associated with the population of true-blue cones, and are not innervated by the Type 7/CB4a bipolar cell (Haverkamp et al., 2005).
Not all features of the bipolar cells remain unchanged, however, as a closer examination of the small terminal branches of the dendritic arbor reveals a marked response to variations in the presence of the photoreceptors, particularly the cone bipolar cells. In wildtype retinas, the cone bipolar cell forms clusters of terminal branches only at loci where cone pedicle active sites reside. When presented with an OPL that lacks these discrete locations, achieved by either removing all of the cones or providing ubiquitous coverage of cone terminals, the cone bipolar cell consistently fails to form such discrete clusters of terminals. The rod bipolar cell also undergoes a slight remodeling of its terminal branches when deprived of its normal afferents, no longer producing characteristic finger-like terminals with bulbous endings that extend vertically, but instead forming thick dendrites that give off terminals with smoother, tapered endings that show a tendency for clustering. These results show that bipolar cells depend on their normal afferents in order to pattern the fine dendritic branches that underlie their connectivity. In each case, however, it is surprising that the total number of terminal endings per dendritic field is conserved, suggesting that this feature of the fine terminal differentiation is also intrinsically defined.
The cone bipolar cells are particularly intriguing, because they normally colonize discrete loci in the OPL where pedicles are positioned, whereas the rod bipolar cells extend fine branchlets to invade a collection of spherules distributed both across and through much of the thickness of the OPL. The implication for the cone bipolar cell is that it must produce dendrites that form independent of the afferents (presumably dictated cell-intrinsically). In the wildtype retina, these dendrites subsequently target the pedicles based on proximity. This is consistent with the sharp turns exhibited by some of the dendrites in the wildtype retina, as if they detect the location of individual pedicles only as they extend through the local microenvironment. In other cases, they appear to colonize pedicles from adjoining pedicles. This suggests that the presence of pedicles does not induce the formation or initial trajectory of the outgrowing dendrites, and the dendritic fields in the coneless retina would tend to support this; they produce a roughly comparable dendritic field area, yet, failing to detect nearby pedicles, they form single sparse terminals that may be connecting with a subset of rods by default. In the conefull retina, on the other hand, their relative immunity to the presence of excess cone numbers may be explained by a temporal dependency, with the later-generated (respecified) cones differentiating their terminals in the more outer parts of the OPL too late to influence cone bipolar cell differentiation. Indeed, ON bipolar cell dendrites have been shown to exhibit exploratory behavior in postnatal day 5 mouse retinas (Morgan et al., 2006), well before rod spherules invade the OPL, around postnatal day 8 (Sherry et al., 2003). Alternatively, the respecified cones may fail to express genes that play a role in promoting bipolar cell recognition of their pedicles.
The failure of these cells to differentiate normal clusters of terminals in the conefull retina is still surprising. This may indicate that even the earlier generated cones (those that would normally have become cones) are not quite normal in this respecified OPL, consistent with their failure to form a stratum of large pedicles along the inner border of the OPL. However, the retinal horizontal cell, which also receives cone innervation, displays a marked hypertrophy of its dendritic arbor in the conefull retina (Raven et al., 2007), suggesting that the respecified cones in these retinas are competent to increase the growth and elaboration of dendrites. It may be that the cone bipolar cells and the horizontal cells respond differentially to these afferent-derived signals by virtue of their own intrinsic constraints: cone bipolar cells exhibit an intrinsic specification of terminal endings, and appear to avoid extending into the outermost portion of the OPL (present results), while the horizontal cells are conspicuously hypertrophic, extending processes throughout the entire dendritic field rather than periodically as clusters, as well as across the full depth of the OPL (Raven et al., 2007).
The rod bipolar cells, by contrast, undergo a change in their morphology in the conefull retina that may simply reflect the different physical constraints imposed therein: the OPL is roughly twice its normal thickness, judging by the distribution of synaptic ribbons (e.g. figure 6 and figure 10i, compare with g in each case), and so their longer dendritic course may demand a greater girth of these processes. Within the OPL, the differentiation of terminations associated with these re-specified cones may yield differences in the terminal structure of the dendrites, producing less of a bulbous expansion within the photoreceptor terminal, and some degree of clustering by virtue of making multiple connections within individual re-specified cone pedicle endings.
In summary, the present analysis has shown that both cone and rod bipolar cells are competent to establish and maintain most features of their dendritic arbors in the absence of their normal afferents. The finer terminal extensions that characterize each of these two bipolar cell types, by contrast, are patterned by the presence of a normal population of afferents in each case: cone bipolar dendritic terminals cluster only the presence of normal cone pedicles, while rod bipolar dendritic terminals are uniformly distributed in the presence of a normal population of rod spherules. The fact that each type of bipolar cell differentiates largely independent of the presence of its specific afferent population may enable these cells to become connected to the alternative afferent population, given the above cell-intrinsic constraints. While the molecular mechanisms that regulate afferent-dependent dendritic patterning remain to be determined, the present study makes clear that the control of such patterning is independent from the morphogenetic instructions governing general dendritic growth and branching, themselves likely to be the downstream consequences of cell-intrinsic differentiation programs (Kim et al., 2008).
This research was supported by the NIH (EY-11087 and RR-22585). We would like to thank Dr. Mary Raven for her assistance with the spatstat software package, Drs. Richard Masland and Bin Lin for providing the Gustducin-GFP transgenic mice, Dr. Anand Swaroop for providing the Nrl −/− mice, and Dr. Jeremy Nathans for providing the coneless transgenic mice.