Differential Protection of Injured Retinal Ganglion Cell Dendrites by Brimonidine

Abstract

Purpose.: To determine whether brimonidine protects against the retraction and loss of retinal ganglion cell (RGC) dendrites after optic nerve crush (ONC).

Methods.: Fluorescent RGCs of mice expressing yellow fluorescent protein (YFP) under the control of the Thy-1 promoter (Thy1-YFP mice) were imaged in vivo and assigned to one of six groups according to dendrite structure. The mice then received brimonidine every other day starting 2 days before, or 2 or 6 days after, unilateral ONC. Control animals received vehicle every other day starting 2 days before ONC. Control animals received vehicle every other day starting 2 days before ONC. Total dendrite length, dendrite branching complexity, and the time until complete loss of dendrites were assessed weekly for 4 weeks.

Results.: Overall, brimonidine treatment significantly slowed the complete loss of RGC dendrites and significantly slowed the reduction of total dendrite length and branching complexity. Separate analysis of each RGC group showed brimonidine significantly delayed the time until complete loss of dendrites in four of the RGC groups. These delays generally were similar when treatment started either 2 days before or 2 days after ONC, but were smaller or absent when treatment started 6 days after ONC Protection against loss of total dendrite length and loss of branching complexity was observed in three of the RGC groups. In two of these RGC groups, protective effects persisted until the end of the study.

Conclusions.: Brimonidine protects many RGC types against dendrite retraction, loss of branching complexity, and complete loss of dendrites following ONC. However, the pattern and magnitude of this protection differs substantially among different RGC types. These results indicate that requirements for RGC-protective therapies following optic nerve injury may differ among RGC types.

Introduction

The pruning of retinal ganglion cell (RGC) dendrites is a well-known consequence of optic nerve damage that occurs in experimental and clinical glaucoma.^1–5 Several studies indicate that the magnitude of this injury-associated dendritic remodeling varies among different broad categories of RGCs, as well as among certain functional RGC subtypes.^1,5,6 This could be important as different RGC types have different central connections, physiology, and function.^7–9 These and other recent findings suggest diverse RGC types may provide parallel information streams that are essential for visual perception.¹⁰ However, whether different RGC types have differing requirements for neuroprotection following injury is unknown. 

Studying injury-induced changes of RGC dendrite structure within postmortem specimens is difficult because assignment of potential dendrite alteration to specific RGC types is uncertain. In particular, pruning or remodeling of one RGC type may yield a dendritic tree that appears similar to the normal structure of another RGC type. Likewise, assignment of an RGC type made strictly on the basis of expression of a particular marker protein is ambiguous if optic nerve injury causes the expression of this protein to decline prior to cell death, as has been noted for Thy-1.^11–13 These limitations also make it challenging to determine if neuroprotective treatments might preferentially target certain RGC types and whether there are different effects of neuroprotective treatments on dendritic remodeling among different RGC types. 

Recently, we have developed the capability to image the structure of RGC dendrites in vivo within a transgenic mouse strain expressing yellow fluorescent protein (YFP) under control of the promoter for Thy-1 using a modified confocal scanning laser ophthalmoscope (mCSLO).¹⁴ This approach allows the same RGC to be repeatedly imaged during the course of degeneration. Analysis of the two-dimensional shape of the dendritic trees in this model distinguished six different RGC groups that contain most of the RGC types reported in prior histologic studies.^15–17 There are two key strengths of following RGC shape with this technique. First, each RGC can be assigned to a particular RGC group before experimental injury alters its dendrite structure. Second, subsequent changes in dendrite structure can be followed longitudinally. 

The α2-adrenergic agonist brimonidine is well known to protect against RGC death and loss of Thy-1 promoter activation following optic nerve injury.^18–20 Brimonidine also reduces optic nerve axon loss following chronic model of elevated IOP.²¹ Unknown, however, is whether brimonidine protects against the retraction of RGC dendrites, and whether there are different effects of brimonidine on the dendrites of different types of RGCs. The present study directly investigated these issues and shows that the dendrites of many RGC groups are protected by brimonidine. Among the different RGC groups, however, there were substantial differences in the onset, durability, and magnitude of the protective effect. 

Materials and Methods

Animals

Animals used in this study were obtained by breeding pairs of B6.Cg-TgN (Thy195 YFP)16Jrs mice obtained from The Jackson Laboratory (Bar Harbor, ME, USA). All experimental procedures conformed to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the University of California San Diego (La Jolla, CA, USA) Institutional Animal Use and Care Committee. The animals were housed using standard vivarium enclosures, provided standard food and water ad libitum, and maintained with 12 hours light and 12 hours dark daily light cycle. Trained personnel under the supervision of a laboratory research veterinarian checked animal health daily. 

Experimental Design

Ocular biomicroscopic examination was performed to exclude any animals with abnormal ocular appearance. Baseline mCSLO images for both eyes of 40 Thy1-YFP mice of either sex were obtained as described below. One week later, the animals were randomly divided into four groups. The first group received systemic brimonidine (100 mg/kg dissolved in ~200 μL PBS, intraperitoneally) every other day starting 2 days before unilateral optical nerve crush (ONC) according to our established procedure²⁰ and continuing until the end of the study. This dosing regimen was chosen based on a prior study that showed it provides good protection of RGC of Thy-1 promoter activation in RGCs without excessive side effects.²⁰ The second group received the same brimonidine injections starting 2 days after ONC and continuing every other day until the end of the study. The third group received the same brimonidine injections every other day starting 6 days after ONC and continuing until the end of the study. Control animals received PBS vehicle starting 2 days before ONC and continuing every other day until the end of the study. Fluorescent RGCs were imaged weekly as described below using a mCSLO until the end of the study. In addition to daily veterinarian inspections, the body weight of each animal was measured at the time of baseline image collection and then measured weekly for the remainder of the study as a measure of animal health. 

mCSLO Imaging

Fluorescent RGCs were imaged as previously described using the confocal scanning laser microscopy mode of a Spectralis imager (Heidelberg Engineering, Carlsbad, CA, USA) that had modified optics allowing focus on the mouse retina.¹⁴ Animals were gently restrained by hand and no anesthesia, contact lenses, or corneal lubricant was used. A series of 20 scans of a retinal region that had an area of approximately 2 mm² was obtained in less than 5 seconds. The use of active eye tracking allowed for the images to be averaged resulting in improved signal to noise ratio. Three to five areas from each eye were recorded. The short exposure period needed for focusing and image collection was generally well tolerated by the mice. Both eyes were imaged at each imaging session. This allowed monitoring of any fluorescent changes that might have occurred for reasons unrelated to the ONC. 

Baseline images were reviewed for quality, subject RGCs were identified, and the locations of these RGCs was recorded to facilitate subsequent reimaging. Any baseline images that were inadequate were retaken prior to initiating the experimental studies. 

For all postcrush images, comparison of the imaged fields at the time of collection with printed copies of the baseline images facilitated reimaging of the same retinal areas as were imaged at the beginning of the study. Images were reviewed for quality immediately after collection and retaken, if necessary. From time to time, some eyes acquired corneal scratches that transiently limited the ability to obtain high quality images in certain areas of the retina. These usually healed quickly allowing the affected areas to be imaged again 3 to 7 days later. 

Optic Nerve Crush

The optic nerve was crushed according to our established method.²² Briefly, a limbal conjunctival peritomy was performed in the temporal region and was gently peeled back to allow access to the posterior region of the globe. The optic nerve was then exposed through a small window made between the surrounding muscle bundles and fatty tissue by gentle blunt dissection. Care was taken not to damage muscles or the blood vessels. At a site approximately 1-mm posterior to the globe, the optic nerve was firmly clamped using fine forceps (Dumont No. 7; Manufactures D'Outils Dumont SA, Basse-Allaine, Switzerland) for 5 seconds. After the procedure, antibiotic ointment was applied to the surgical site. This procedure for ONC has been shown to reliably avoid compromise of the retinal circulation.²² Nevertheless, we examined the fundus of the animals after ONC with the mCSLO using red-free light²³ and confirmed that retinal circulation was normal in all animals. 

To assess the completeness of RGC axonal injury following ONC, four mice received ONC followed immediately by injection of 0.25% cholera toxin subunit B conjugated to Alexafluor-488 dissolved in PBS. Four days after the injection, the eyes with optic nerve attached were dissected, the optic nerve plus peripapillary optic nerve head tissue were isolated, then this preparation was whole mounted for fluorescence microscopy examination. As shown in Supplementary Figure S1, the optic nerve was strongly fluorescent between the optic nerve head and the crush site, and there was no detectable fluorescence distal to this site. 

Analysis

An average of 13 RGCs were suitable for analysis from each study eye or approximately 130 per experimental treatment group containing 10 mice. Overall, 580 RGCs were studied in 40 mice. Each RGC with clearly defined dendrites at baseline and an axon coursing toward the ONH was designated as belonging to one of six different groups based upon dendrite thickness, length, and arborization pattern in the baseline image, and position in the retina as previously described.¹⁴ Masked trials by two of the authors (JL and DH) yielded 96.4% agreement in assignment of RGCs to the six groups. Two types of quantitative analysis were performed on each analyzed RGC: a kinetic analysis of postcrush changes in the total dendrite length and dendrite branching complexity, and an evaluation of when it was first observed that dendrites were no longer visible using Kaplan-Meier analysis. The full dendritic pattern of each RGC was traced using Filament Tracer within Imaris 7.3.0 (Bitplane AG, Zurich, Switzerland). Measurements of total dendrite length and dendrite branching complexity were determined from these tracings. Each filament was reviewed to eliminate erroneous tracing of fluorescent axons or dendrites from nearby fluorescent RGCs. Length measurements of each dendrite segment of the filament were summed to obtain the total dendrite length measurement. To assess dendrite branching complexity, the total Sholl score was determined for each RGC with Imaris by counting the number of intersections between the corresponding RGC filament and concentric circles centered at the soma and drawn with radii that increased at 10-μm intervals. In addition, the number of intersections with the single concentric circle that was the largest of the set for each RGC was recorded as the maximum Sholl score. Prior reports indicate that this maximum Sholl score can be a more sensitive measure of dendritic remodeling than the total Sholl score.^24,25 For each RGC, total dendrite length, total Sholl scores and maximum Sholl scores collected after ONC were expressed as a percent of the baseline measurements. This compensated for fact that peripheral cells were viewed at an angle greater from orthogonal to the visual axis than were cells closer to the optic nerve head. Calibration of Imaris was performed by comparing the distances between specific fluorescent RGCs in an in vivo image obtained with the Spectralis with the distances between photomicrographs of the same RGCs observed in a postmortem flat-mount preparation observed with a conventional upright fluorescence microscope (E800; Nikon Instruments, Inc., Melville, NY, USA). The photomicrograph magnification was calibrated by photographing a stage micrometer. To evaluate the time of dendrite disappearance, the first week after ONC that dendrites were no longer observed was recorded for each RGC. These results were used to generate Kaplan-Meier plots for each RGC type and experimental group. 

Statistical Evaluation

For the evaluation of the time until there was complete loss of RGC dendrites in each RGC group, Kaplan Meier plots of each of the three brimonidine treatment groups were independently compared with the corresponding results for the vehicle treatment group using the log rank test.²⁶ The results were appropriately adjusted for multiple probing of data sets using the Bonferroni correction. Normal distribution within total dendrite length and maximum Sholl data sets was evaluated using the Anderson-Darling test. Nonparametric comparisons of experimental responses in two data sets were made using the Mann-Whitney U test. Multiple nonparametric pair-wise comparisons of responses from more than two data sets were made using the Steel-Dwass-Critchlow-Fligner test.^27,28 These comparisons were not compromised by the inability to sometimes image an eye due to transient corneal scratches (typically no more than one eye in each group at each time point). Analysis of body weight changes over the course of the study was made using ANOVA. 

Double-Label Immunohistochemistry

Sixteen Thy1-YFP mice received ONC and were allowed to survive for 1 to 4 weeks prior to fixation and isolation of the retinas for whole-mount immunohistochemistry. Anesthetized mice were fixed by transcardial perfusion with PBS followed by 4% freshly-prepared, chilled formaldehyde in PBS, immersion fixation in situ for 30 minutes, and 30 minutes fixation of the isolated retinas. The retinas were then washed in PBS, permeabilized, and blocked for 3 hours with 5% fetal calf serum, 0.3% Triton X-100 in PBS. The retinas were incubated in primary antibodies for goat anti-green fluorescent protein (recognizes YFP, 1:250; Abcam, Cambridge, MA, USA) and mouse anti-nonphosphorylated neurofilament heavy chain (SMI-32, 1:250; Covance, Denver, PA, USA). After washing, the retinas were incubated with corresponding secondary antibodies conjugated to Alexafluor-488 or Alexafluor-568 (1:500; Life Technologies, Carlsbad, CA, USA), washed again, and mounted using antifade mounting medium (Prolong Gold; Life Technologies). 

Results

Distinguishing Features of the Six RGC Groups

For RGC appearance, RGCs in the baseline images collected before ONC were assigned to one of six groups according to differences in their dendritic structure, as previously described.¹⁴ Briefly, Group 1 RGCs had moderately large, symmetric dendritic fields with robust dendrites that had minimal sharp bends and branching (Fig. 1A). Group 2 RGCs had smaller symmetric dendritic fields with thin dendrites that displayed few bends and minimal branching. Group 3 RGCs were similar in size to Group 2 RGCs, but had symmetric dendritic fields with thin dendrites that displayed numerous sharp bends and more branching. Group 4 RGCs had the largest dendritic fields with many robust dendrites and moderate branching. Group 5 RGCs had relatively small asymmetric dendritic fields with thin highly branched dendrites. Finally, Group 6 RGCs were smaller than Group 5 RGCs with thin, tortuous dendrites. Group 4 RGCs were rarely observed since they mostly are too peripherally positioned in the retina to be observed by the mCSLO. Figure 1B shows that the YFP-expressing RGCs usually were well separated with minimal overlap of adjacent dendritic fields. Moreover, fluorescent axons were observed connecting RGC somas to the optic nerve head. In the present study, approximately one-half of the fluorescent RGCs observed were evenly divided between Groups 2 and 5 RGCs; approximately 30% of RGC were evenly divided between Groups 3 and 6 RGCs, and Groups 1 and 4 RGCs constituted 9% and 4% of total observed RGCs, respectively (Fig. 1C). 

Figure 1

 

The wide variety of fluorescent RGCs in Thy1-YFP mice (A) can be assigned to six different groups based in part on their appearance (B). The prevalence of RGCs observed in vivo varied according to RGC group (C). Scale bars: 100 μm.

Following ONC, gradual shortening of the dendrites was accompanied by a reduction in branching complexity and a reduction in dendrite fluorescence intensity (Fig. 2). Although often slower in course, the pattern of dendrite degeneration of RGCs of mice that received brimonidine was generally similar to those that received vehicle. In Group 3 RGCs, dendrite end bulbs often were observed along with dendrite retraction and dimming after ONC. This characteristic was seen in both mice that had received vehicle and mice that had received brimonidine treatments. 

Figure 2

 

Effect of brimonidine treatment started 2 days before ONC on the time course of dendrite changes in Groups 1 (A) and 3 RGCs (B). Scale bars: 100 μm.

Close correlation exists between the Sholl profiles generated from images of the same RGC obtained in vivo using the mCSLO and ex vivo in histologic retinal flat-mounts examined by conventional fluorescence microscopy.²⁹ Likewise, close correlation exists between flat-mount images of the same RGC that demonstrate YFP fluorescence and immunoreactivity for neurofilament heavy chain (NFH), a marker for RGC dendrites.²⁹ The possibility that decreased YFP fluorescence in injured RGCs might lead to underestimations of total dendrite length was further evaluated in Thy1-YFP mice that received ONC 1 to 4 weeks prior to isolation of the retina for dual-label immunostaining using antibodies that recognized YFP and NFH. As shown in Figure 3, dendrites of RGCs expressing YFP also were immunoreactive for NFH. By 4 weeks following ONC, the distribution of YFP along dendrites often became less linear (Figs. 3G–L), however, double labeling of the dendrites persisted. In addition, no NFH-expressing dendrites were observed extending from YFP-expressing cell bodies that did not also contain YFP. Nevertheless, a recent study has shown that although actively remodeling neurite growth cone filopodia contain actin, they do not contain neurofilaments.³⁰ This raises the possibility that neurofilament immunoreactivity may have not always extended out to the tips of remodeling dendrites. Hence, mCSLO images of YFP distribution in RGCs provide a useful index of dendrite degeneration, but not one that can necessarily be mapped to structural dendritic degeneration. 

Figure 3

 

Immunostaining for YFP (anti-GFP, green) and for nonphosphorylated neurofilament heavy chain (SMI-32, red) within RGC dendrites at 1 week after ONC (A–F) and at 4 weeks after ONC (G–L) shown at low magnification (A–C, G–I) and at high magnification (D–F, J–L). Merged images shown in (C, F, I, L). Consistent overlap of YFP and SMI32 immunoreactivities in the dendrites (arrowheads), dendrite endings (asterisk), and axons (arrows) indicates that YFP distribution provides good representation of the extent of RGC structure. Scale bars: 20 μm.

Variable Loss of RGC Dendrites Among the RGC Groups

Within each of the experimental groups, the time after ONC until there was complete loss of visible RGC dendrites was variable. Hence, this parameter was compared among all treatment groups by Kaplan-Meier plots. Group 4 RGCs were not analyzed because too few of these RGCs were observed. This was because Group 4 RGCs mostly were located too peripherally to be observed in vivo. As shown in Figure 4A, the proportion of RGCs retaining dendrites declined to zero by 4 weeks after crush for Groups 1, 2, and 6 RGCs. In contrast, 21% of Group 3 RGCs and 15% of Group 5 RGCs retained dendrites at the end of the 4-week study. Although the decline of Group 3 RGCs was slower than all other RGC groups, this delay was statistically significant and only relative to Groups 2 or 6 RGCs (Fig. 4F, 4J; P < 0.05, Bonferroni log rank test). N for each group and P values for all comparisons are shown in Supplementary Table S1. 

Figure 4

 

Kaplan-Meier plots from vehicle-treated animals comparing the loss of RGCs with dendrites in each the RGC groups. (A) All RGC groups shown together. (B–K) All pairwise differences in the loss of RGCs with dendrites. Asterisk indicates P < 0.05 by the Bonferroni corrected log rank test. N for each group and P values for all comparisons are reported in Supplementary Table S1.

Brimonidine Protection Against Dendrite Loss

Treatment with brimonidine starting 2 days before crush, or 2 or 6 days after crush significantly protected all RGCs against loss of dendrites (P = 0.0000023, 0.048, and 0.048, respectively; Bonferroni corrected log rank test, Fig. 5A). At the end of the study, the magnitude of these differences ranged from 13% to 18% of baseline RGC counts. When each RGC group was evaluated separately, however, Groups 1, 2, and 6 RGCs had up to 59%, 34%, and 47% greater preservation of RGCs with dendrites, respectively (P = 0.0069, 0.000024, and 0.0067, respectively; Figs. 5B, 5C, 5F; Supplementary Table S2). For RGC Groups 1 and 2, significant protection was observed both when brimonidine treatment was started 2 days before ONC and when it was started 2 days after ONC. In contrast, the declines in RGCs with dendrites from RGC Group 3 and/or RGC Group 5 were not significantly altered by any of the brimonidine treatments (Figs. 5D, 5E). P values for all comparisons are shown in Supplementary Table S2. 

Figure 5

 

Kaplan-Meier plots comparing the loss of RGCs with dendrites in mice that received vehicle with mice that received brimonidine treatment every other day starting 2 days before ONC, 2 days after ONC, or 6 days after ONC. The overall analysis (A) compares the results of all RGCs within each experimental treatment group. Asterisks indicate P < 0.05, Bonferroni corrected log-rank test. N for each experimental treatment group is indicated in parentheses. P values for each comparison reported in Supplementary Table S2.

Dendrite Retraction Rate Differs Among the RGC Groups

Plots of the decline in total dendrite length of six example individual RGCs from each RGC group within the vehicle control mice illustrate variability within each RGC group as well as distinct differences (Fig. 6A). For example, total dendrite length of several Groups 2 and 5 RGC examples declined to zero at each of the four time points. In contrast, total dendrite length in Group 1 RGCs typically did not decline to zero until later. To facilitate statistical comparison, the total dendrite length measurements at 1 to 4 weeks after ONC for each RGC were expressed as a percent of the baseline measurements in the same RGC measured prior to ONC. The patterns of dendrite retraction within each experimental group are shown in Figure 4B using stack bar graphs in which each stack bar element indicates the proportion of RGCs for which total dendrite length was 75% to 100% of baseline, 50% to 75% of baseline, 25% to 50% of baseline, 0% to 25% of baseline, or in which dendrites were absent (0% of baseline). Evaluation of all RGCs within each RGC group at each time point using the Anderson Darling test indicated they were not normally distributed (P > 0.05). Hence, the distributions in total dendrite length of all RGCs within each RGC group were plotted as stack bar graphs (Fig. 6B) and the losses of total dendrite length among the different RGC groups were compared using the nonparametric Steel-Dwass-Critchlow-Fligner test. Horizontal lines above certain stack bar pairs indicate significant differences (P < 0.05). The reductions in total dendrite length were generally similar among the different RGC groups at 1 week after crush. The only exception was that the reduction of total dendrite length in Group 1 RGCs was significantly slower than in Group 6 RGCs. Similar declines in total length were generally seen as 2 weeks after crush except the reduction of total dendrite length in Group 3 RGCs was significantly slower than in RGC Groups 2, 5, and 6 (P < 0.05 in each case). At 3 and 4 weeks after crush, differences in the dendrite length distributions among the all the various RGC groups were insignificant. P values for all comparisons are reported in Supplementary Table S3. 

Figure 6

 

Differences in total dendrite length changes after ONC among the RGC groups in control (vehicle) mice. Total dendrite length changes of six example RGCs from each RGC Group illustrate the range of total length at baseline and responses to ONC (A). To compare the RGC groups, total dendrite length changes from all RGCs in each group were expressed as a percent of baseline total dendrite length. The distribution of changes within each RGC group at each time point are presented as stack bar graphs (B). Horizontal lines above the stack bars indicate significant differences among the RGC groups (P < 0.05, all pair-wise analysis, Steel-Dwass-Critchlow-Fligner test). P values for all comparisons are reported in Supplementary Table S3.

Brimonidine Protection of Total Dendrite Length

Brimonidine treatment started 2 weeks before ONC significantly delayed the reduction of dendrite length in all RGCs at 1, 2, and 3 weeks after crush (P = 0.0051, 0.00019, and 0.027, respectively, Mann-Whitney U test; Fig. 7A, stack bar graphs). When the RGC groups were evaluated separately, significant delays in the reduction of dendrite length at one or more time points were observed in Groups 1, 2, 5, and 6 RGCs (Figs. 7B–F). For Group 3 RGCs, there was no significant difference in the course of dendrite length reduction at any time point (P > 0.6 at every time point). For the mice that received brimonidine treatments starting 2 or 6 days after ONC, the protective effects were typically smaller than those described above and often were not significantly different from the vehicle-treated mice (not shown). Similarly to loss of dendrites, these results indicate that brimonidine significantly delayed reduction of dendrite length after ONC in Groups 1, 2, and 6 RGCs, but not in Group 3 RGCs. In contrast to dendrites loss, dendrite retraction in Group 5 RGCs was protected. 

Figure 7

 

Protection of total dendrite length by brimonidine treatment. Total length changes for all RGCs in each experimental group were expressed as a percent of baseline measurements prior to analysis. The distributions of these changes for each experimental group are presented as stack bar graphs. Overall results compare all RGCs within each treatment group (A). Evaluation of the effect of brimonidine for each RGC group separately shows significant protection of dendrite length within RGC Groups 1, 2, 5, and 6, but not within RGC Group 3 (B–F). P values from analysis of the differences in the distributions using the Mann-Whitney U test are shown above each stack bar pair. Significant differences are marked with an asterisk. N for imaged RGCs in each group is indicated at the base of each stack bar.

Branching Simplification After ONC Differs Among the RGC Groups

Sholl plots of dendrite branching complexity for four example RGCs from each RGC group illustrate characteristic branching differences among the RGC groups (Fig. 8A). For example, Group 1 RGCs typically attained maxima of 20 to 30 intersections per Sholl circle at 90 to 150 μm from the soma center, while Group 2 RGCs typically attained 14 to 18 intersections at 60 to 100 μm from the soma center and Group 6 RGCs typically attained only 9 to 12 intersections at 35 to 85 μm from the soma center. Prior reports,^22,23 as well as pilot studies conducted with the present data (not shown), indicate that the maximum Sholl value is a more sensitive measure of dendritic branching complexity changes than the total Sholl score. Hence, the maximum Sholl score for each RGC measured after ONC was expressed as a percent of its baseline maximum Sholl score and the distributions of these maximum Sholl score changes were plotted as stack bar graphs (Fig. 8B). The decline of Sholl maxima after ONC was generally similar among the RGC groups. However, maximum Scholl score decline at 2 weeks after crush was significantly slower in Group 3 RGCs than in Groups 2, 5, or 6 RGCs (horizontal lines above certain stack bar pairs indicate P less than 0.05, Steel-Dwass-Critchlow-Fligner test). P values for all comparisons are shown in Supplementary Table S4. 

Figure 8

 

Dendrite branching complexity changes after ONC. Baseline Sholl plots of four example RGCs from each RGC group illustrate characteristic similarities in branching complexity patterns within each group and the characteristic differences among the RGC groups (A). To compare the group responses to ONC, maximum Sholl scores for each RGC were expressed as a percent of baseline maximum Sholl score. The distributions of maximum Sholl score changes after ONC within each RGC group are presented as stack bar graphs (B). Horizontal lines above the stack bars indicate significant differences between certain RGC group pairs (P < 0.05, all pair-wise analysis, Steel-Dwass-Critchlow-Fligner test). N for imaged RGCs in each RGC group are indicated by the numbers at the base of each stack bar. P values for all comparisons are reported in Supplementary Table S4.

Brimonidine Protection of Branching Differs Among RGC Groups

Sholl plots of example RGCs obtained at weekly intervals after ONC illustrate simultaneous reduction of dendritic branching complexity in both proximal and distal portions of the dendritic arbor. This general pattern was present within each of the analyzed RGC groups from vehicle-treated mice (Fig. 9; left panels). In addition, there was a progressive shift of the Sholl maxima toward smaller radii. Brimonidine treatment did not alter this general pattern (Fig. 9, right panels). However, it slowed the rate at which this process advanced. For statistical analysis, the maximum Sholl score of each RGC at each time point after ONC was expressed as a percent of the baseline RGC score for that RGC and the distributions of these maximum Sholl score changes were plotted as stack bar graphs (Fig. 10). Because the maximum Sholl score changes were not normally distributed (Anderson-Darling test, P > 0.05), differences between the Sholl score changes from vehicle-treated mice and brimonidine-treated mice were evaluated at each time point using the Mann-Whitney U test. Brimonidine significantly protected against maximum Sholl value decline in all RGCs considered together at all time points (Fig. 10A, P = 0.0063, 0.00013, 0.0075, and 0.029 at 1, 2, 3, and 4 weeks after ONC, respectively). When the RGC groups were evaluated independently, brimonidine treatment significantly delayed the decline of maximum Sholl values at weeks 2, 3, and 4 after ONC in Group 2 RGCs. In addition, brimonidine significantly delayed maximum Sholl decline and at 3 weeks after ONC in Group 1 RGCs, and at 2 weeks after ONC in Group 5 RGCs. In contrast, brimonidine treatment did not significantly alter the course of dendrite branching simplification in Group 3 RGCs or in Group 6 RGCs (Figs. 10D, F; P > 0.15 at all time points). Brimonidine-mediated protection of dendrite complexity in Group 1 and Group 2 RGCs is consistent with its protection against both dendrite loss and dendrite retraction in these RGC Groups (Figs. 5, 7). Likewise, brimonidine-mediated protection of dendrite complexity in Group 5 RGCs is consistent with its protection against dendrite retraction. Although brimonidine protected Group 6 RGCs against dendrite loss and dendrite retraction, its effect on dendrite complexity in this RGC group was insignificant. 

Figure 9

 

Sholl plots of example RGCs obtained at weekly intervals after ONC illustrating simultaneous reduction of dendritic branching complexity in both proximal and distal portions of the dendritic arbor. This general pattern was present within each of the analyzed RGC groups from vehicle-treated mice (A, C, E, G, I) as well as from mice that received brimonidine (B, D, F, H, J). Note that in both treatment groups and for each RGC group, there is a graded reduction of branching complexity at all radii accompanied by a shift of the maximum Sholl score toward smaller radii. Evaluation of dendritic complexity changes in all RGCs of each experimental group is shown in Figure 10.

Figure 10

 

Protection of dendrite branching complexity by brimonidine treatment. Each stack bar shows the distribution of RGCs of maximum Sholl scores within each experimental group at each time point. Overall results compare all RGCs within each treatment group (A). Evaluation of the effect of brimonidine for each RGC group separately shows significant protection of dendrite branching complexity within RGC Groups 1, 2 and 6, but not within RGC Groups 3 and 6 (B–F). P values from analysis of the differences in the distributions using the Mann-Whitney U test are shown above each stack bar pair. Significant differences are marked with an asterisk. N for imaged RGCs in each RGC group are indicated by the numbers at the base of each stack bar.

Body Weight

To assess the effect of brimonidine on metabolic health, body weight was measured at baseline and then weekly following ONC. There were no significant changes in body weight in any of the treatment groups over the course of the study (P > 0.05 in each case, ANOVA, detailed results and P values for each comparison presented in Supplementary Table S5). 

Discussion

This study demonstrates that brimonidine treatment provides significant protection of RGC dendrites following optic nerve injury. In addition, it has demonstrated that the various types of RGCs differ in their responses to ONC as well as in the protection against these responses that is achieved with brimonidine treatment. Interestingly, changes in total dendrite length were similar, though not identical, to changes in branching complexity. Of particular note, Group 3 RGC dendrites were significantly slower in their responses to ONC than the dendritic responses in many other RGC groups. This was true for complete loss of dendrites, for reduction of total dendrite length, and for simplification of dendritic branching complexity (reflected by reduction of maximum Sholl values). When all RGCs were considered together, brimonidine significantly slowed loss of dendrites, reduction of total dendrite length, and simplification of branching complexity. When the RGC groups were analyzed independently, brimonidine significantly protected against all three of these degenerative changes in Groups 1 and 2 RGCs. Also, brimonidine protected against dendrite length reduction and dendrite loss in Group 6 RGCs, and against dendrite length reduction and branching simplification in Group 5 RGCs. In Group 3 RGCs, however, none of these dendrite measures was altered by brimonidine treatment. These differences are summarized in further detail within Table 1. Overall, brimonidine protected against dendrite degeneration in Groups 1, 2, 5, and 6 RGCs. These RGC types together represent 79% of the total RGCs analyzed (Fig. 1C). Thus, the present results show brimonidine protects against RGC dendrite degeneration in most, but not all, RGCs. They also reveal complexities in RGC responses to ONC and to brimonidine treatment that have not previously been appreciated. Finally, they demonstrate that the RGCs in different RGC groups have differing neuroprotective requirements following axonal injury. 

Table 1

 

Summary of Different Responses to ONC and to Protection by Brimonidine

RGC Group	Vehicle	Brimonidine Effect on Complete Dendrite Loss:		Brimonidine Effect on Loss of Total Dendrite Length	Brimonidine Effect on Loss of Dendrite Branching Complexity
		During the Study	At the End of Study
Group 1		Markedly slower loss	44%–59% retained some dendrites	Slower loss at 3 weeks after ONC	Slower loss at 3 weeks after ONC
Group 2		Markedly slower loss	21%–34% retained some dendrites	Slower loss at 2 weeks after ONC	Slower loss at 2, 3, and 4 weeks after ONC
Group 3	Delayed complete dendrite loss, slower decrease of dendrite length, slower reduction of dendrite branching complexity	No protection	No protection	No protection	No protection
Group 5		No protection	No protection	Slower loss at 2 weeks after ONC	Slower loss at 2 weeks after ONC
Group 6		Moderately slower loss	17%–47% retained some dendrites	Slower loss at 1 week after ONC	No protection
Summary of protected RGC groups		1, 2, 6		1, 2, 5, 6	1, 2, 5

Several advantages arise from the ability to visualize the same RGC prior to ONC and then repeated after injury. First, it is easy to confirm that all of the cells analyzed in this study are RGCs since in each case an axon coursing to the optic nerve head is observed. This avoids a concern that some of the faintly fluorescent cells that were imaged could be amacrine cells; this concern arose in prior studies using Thy1-CFP23Jrs mice.^12,13,31 Second, assignment of each RGC to a group according to its dendritic arbor structure can be positively determined before injury-associated changes in dendrite structure might make it appear like an RGC of another group. Thus, changes in response to injury or neuroprotective treatment are positively associated with a particular RGC group. Finally, repeated observations of the same cell allowed the determination that most RGCs respond to injury with progressive retraction of dendrites, loss of branching complexity, and eventually total elimination of fluorescent dendrites. 

It is important to consider whether the fluorescent dendrite structures observed faithfully represent the total dendritic arbor. Our double-label immunohistochemistry study found that in both healthy eyes and eyes following ONC, YFP immunoreactivity in dendrites overlapped neurofilament heavy chain immunoreactivity. Sometimes at several weeks after ONC, YFP immunoreactivity became discontinuous along the dendrites of injured RGCs, but nevertheless was present out to close to the end of the dendrites as defined by neurofilament staining. This supports the view that the fluorescence image of RGC dendrites in Thy1-YFP mice provides a good measure of the full extent of the dendritic arbor. Hence, following fluorescent dendrite changes in vivo is a useful approach to evaluate dendrite responses of different RGC groups to axonal injury and to assess whether neuroprotective treatments protect RGC dendrites. 

In the present study, RGCs were divided into six groups based on differences in dendrite length, branching pattern, and asymmetry as seen in two-dimensional in vivo images. Comparison of the mouse RGC types described in three-dimensional histologic studies by Sun et al.¹⁵ and Völgyi et al.¹⁶ reveals that nearly all can be assigned to one of the six RGCs groups of the present study (Table 2, columns 1, 2). A primary reason for the fewer groups in the present study is that, unlike the prior histologic studies, the present in vivo imaging method did not allow differentiation of those RGC types with similar dendrite branching that have different patterns of dendritic terminations in various layers of the inner plexiform layer (IPL). In addition, some fine differences observable in high resolution microscopic images of postmortem retinal flat-mounts may be difficult to appreciate in the lower resolution in vivo images obtain using the mCSLO. Finally, G18 and G19 RGCs identified by Völgyi were not included in the present study because they were too small or not observed, respectively. Likewise, the 18 RGC types that were distinguished by Cook et al.¹⁷ using automated cluster analysis of RGC structural features and IPL termination patterns also can be assigned to the six RGC groups of the present study (Table 2, column 3). It remains possible that some of the different RGC types that would be assigned to the same RGC group in the present study may have responded differently to protection by brimonidine than other RGC types also assigned to the same group. This limitation may better addressed in future studies with increasing availability of high resolution, 3-dimensional, in vivo imaging recently achieved by the use of adaptive optics confocal scanning laser ophthalmoscopy.³² 

Table 2

 

Comparison of mCSLO RGC Groups and Three Histologic RGC-Type Classification Schemes

RGC Group	Sun et al.15	Volgyi et al.16	Coombs et al.17
Group 1	A1	G1	10
Group 2	A2 inner, A2 outer, B1, C5	G2, G3, G4, G14	3 off, 9 on, 9 off, 13
Group 3	B3 inner, B3 outer, D2	G6, G7, G17, G21	7 on, 7 off, 8
Group 4	C1, C2 inner, C2 outer, C3	G10, G11, G12	6 on, 6 off
Group 5	B2, C6, D1	G5, G9, G15, G16, G20, G22	2, 4a, 5a, 5b, 11, 12, 13, 14
Group 6	B4, C4	G8, G13	1, 11

The presence of differences in RGC group responses to brimonidine raises the possibility of corresponding differences in their expression of α2-adrenergic receptors, in adrenergic signal transduction, or in their capacity to regulate dendrite remodeling. Experiments with mice lacking genes for various α2-receptor subtypes and pharmacological investigations indicate that the neuroprotective benefit by α2-agonists is mediated by activation of the α2A-receptor.^33,34 Examination of the figures in two different studies show that although many ganglion cell layer (GCL) neurons express α2A-receptor immunoreactivity, they do so with variable intensity.^16,35 Though it is possible that the weakly stained GCL neurons include displaced amacrine cells, it also is possible this variable staining intensity reflects variability in the expression of α2A receptors on different types of RGCs. Similarly, because activation of RGC α2-receptors inhibits adenylate cyclase (AC) activity,³⁴ observed variability in the immunoreactivity for certain receptor-linked and nonreceptor-linked ACs in GCL neurons³⁶ also may contribute to variable responses of different RGC types to brimonidine. Likewise, variable expression or activation of cAMP targets such as Protein Kinase A, Epac1, or Epac2,^37,38 phophodiesterase-4,^34,39 or nimodipine-sensitive L-type calcium channels⁴⁰ also could contribute to differing responses of RGCs in the different RGC groups. 

The present results suggest a new strategy for the design of neuroprotective treatments for RGCs. This derives from the observation that various RGC subsets respond differently to optic nerve injury and to neuroprotective support by brimonidine. Although Thy-1 promoter activation is undetectable in greater than 95% of RGCs by 4 weeks after the severe optic nerve injury produced by ONC,^11–14,41 this model does have a days-long period between the onset of initial responses optic nerve injury and the elimination of Thy-1 expression during which degenerative events such as dendritic arbor remodeling and axonal beading occur.¹⁴ If the days-long period of dendritic pruning after ONC accurately models important aspects of what happens in RGCs of traumatic optic neuropathy (TON) patients, then brimonidine may provide enhanced recovery in TON. In support of this potential for recovery, epidemiological studies have reported a portion of untreated TON cases attain three or more Snellen lines improvement of visual acuity between presentation and final follow up.^42,43 Similarly, growing experimental evidence suggest neuroprotective treatments may increase preservation of vision in glaucoma patients.^44–46 These results also support further studies to define the neuroprotective requirements of all RGC subtypes. Moreover, they suggest increased potency for neuroprotective treatments preserve or induce recovery of vision in patients with optic neuropathies such as TON or glaucoma may be achieved by designing multitarget therapies that simultaneously provide for the various neuroprotective requirements of different RGC subtypes. 

Acknowledgments

Supported by grants from Allergan (JDL; Irvine, CA, USA), and an unrestricted grant from Research to Prevent Blindness (New York, NY, USA). 

Disclosure: J.D. Lindsey, Allergan (F); K.X. Duong-Polk, None; D. Hammond, None; P. Chindasub, None; C.K.-S. Leung, None; R.N. Weinreb, Allergan (C)