Dendrite Remodeling and Other Abnormalities in the Retinal Ganglion Cells of Ins2^Akita Diabetic Mice

Abstract

purpose. To determine the extent of retinal ganglion cell loss and morphologic abnormalities in surviving ganglion cells in Ins2^Akita/+ diabetic mice.

methods. Mice that expressed cyan fluorescent protein (CFP) or yellow fluorescent protein (YFP) reporter genes under the transcriptional control of the Thy1 promoter were crossed with Ins2^Akita/+ mice. After 3 months of diabetes, the number and morphology of retinal ganglion cells was analyzed by confocal microscopy. The number of CFP-positive retinal ganglion cells was quantified in retinas of Ins2^Akita/+ Thy1-CFP mice. The morphology of surviving cells was examined, and dendritic density was quantified in Ins2^Akita/+ Thy1-YFP mice by using the Sholl analysis.

results. Thy1-CFP expression was limited to retinal ganglion cell bodies. There was a 16.4% reduction in the density of CFP-positive ganglion cells in the peripheral retina of Ins2^Akita/+ mice compared with wild-type control retinas (P < 0.017), but no significant change in the central retina. Thy1-YFP expression occurred throughout the entire structure of a smaller number of cells, including their soma, axons, and dendrites. Six different morphologic clusters of cells were identified in the mouse retinas. The structure of dendrites of ON-type retinal ganglion cells was affected by diabetes, having 32.4% more dendritic terminals (P < 0.05), 18.6% increase in total dendrite length (P < 0.05), and 15.3% greater dendritic density compared with control retinas, measured by Scholl analysis. Abnormal swelling on somas, axons, and dendrites were noted in all subtypes of ganglion cells including those expressing melanopsin.

conclusions. The data show that retinal ganglion cells are lost from the peripheral retina of mice within the first 3 months of diabetes and that the dendrites of surviving large ON-type cells undergo morphologic changes. These abnormalities may explain some of the early anomalies in visual function induced by diabetes.

Diabetic retinopathy is the leading cause of new cases of legal blindness in working-age adults in the United States. ¹ One of the earliest clinical signs of retinopathy is an increase in vascular permeability, leading to macular edema, which has been closely linked to loss of visual acuity. ² Experimental evidence shows that diabetes also causes degeneration of a variety of neural cells, concomitant with the earliest changes in vascular function, ^{3 4} suggesting that the disease has broad impact on the retina, affecting both the vascular and neural components. ⁵  

The cause of neurodegeneration in the retina during diabetes is not understood, but it involves apoptosis of retinal neurons, leading to a subsequent thinning of the inner plexiform and inner nuclear layers. ^{3 6} Cell death is well documented and has been shown to affect several different populations of neurons in humans and in animal models of diabetes. ^{3 7 8 9 10 11} Studies in humans have also measured decreases in the thickness of the nerve fiber layer, suggesting loss of ganglion cell axons in diabetes. ^{12 13 14} Experimental studies of other disease models show that the morphology of surviving retinal ganglion cells can change in response to glaucoma and axotomy, ^{15 16 17 18 19} but the effect of diabetes on retinal ganglion cell morphology and dendrite structure are not well established. 

In this study, we used the Ins2^Akita mouse model to determine the effect of diabetes on retinal ganglion cell survival and morphology. ^{20 21} Within the first three months of hyperglycemia the retinas of these mice develop increased vascular permeability accompanied by microglial cell reactivity, increased leukostasis, elevated caspase-3 activity, and reduced insulin receptor kinase activity. These changes are accompanied by neurovascular degeneration, indicated by reduced thickness of the inner plexiform layer (IPL) after 5.5 months and an increase in acellular capillaries after 9 months of diabetes. ⁶ A further study identified a significant increase in the number of apoptotic cells and a reduction in the total number of cholinergic and dopaminergic amacrine cells in Ins2^Akita/+ mice. ⁷  

To investigate the impact of diabetes on the morphology of retinal ganglion cells, Ins2^Akita mice were crossed with transgenic mice expressing cyan fluorescent protein (CFP) or yellow fluorescent protein (YFP) regulated by the Thy1 promoter. ²² CFP is expressed in the cell bodies of retinal ganglion cells, whereas YFP is expressed in fewer ganglion cells but appears throughout the entire dendritic structure. Use of these transgenic mice enabled both quantification and detailed assessment of morphology of retinal ganglion cells. The data demonstrate that three months of uncontrolled hyperglycemia results in a significant reduction in the number of retinal ganglion cells, induces frequent morphologic abnormalities in dendrites and axons, and increases the extent and density of dendritic branches in ON-type retinal ganglion cells, suggesting that the dendrites of surviving ganglion cells undergo morphologic changes in diabetes. 

Materials and Methods

Animals

Ins2^Akita mice (no. 003548; The Jackson Laboratory, Bar Harbor, ME) were maintained in a pathogen-free colony by continuous mating with C57BL/6J (no. 003548; The Jackson Laboratory) in the Juvenile Diabetes Research Foundation Diabetic Retinopathy Center Animal Core. The Akita allele of the Insulin 2 gene carries a point mutation that results in misfolding of the protein, leading to its accumulation in the endoplasmic reticulum of β-cells and their subsequent loss. Ins2^Akita/+ male mice become significantly hyperglycemic compared with age-matched wild-type (Ins2^+/+) littermates by 4 weeks after birth. ^{20 21} The hyperglycemic phenotype is 100% penetrant in Ins2^Akita/+ male mice by 6 weeks of age and is irreversible. The determination of blood glucose phenotyping was performed twice, at 4.5 weeks when the penetrance is greater than 95%, to allow experimental groups to be identified, and again at death, ensuring accurate genotypic assignment. In rare cases a mouse originally typed as nondiabetic becomes diabetic before or at harvest, but these were excluded from the experiments. 

Two transgenic lines of mice were purchased and maintained on the C57BL/6J background: (1) carrying a CFP reporter gene under direction of the mouse Thy1 gene promoter (Tg(Thy1-CFP)23Jrs) (Thy1-CFP) (no. 003710; The Jackson Laboratory); and (2) carrying a YFP reporter gene under direction of the mouse Thy1 gene promoter (Tg(Thy1-YFP)23Jrs) (Thy1-YFP; no. 003782; The Jackson Laboratory). ²² Thy1-CFP is expressed in the cell bodies of many retinal ganglion cells, whereas Thy1-YFP is expressed throughout the entire neuronal structure in smaller numbers of all morphologic types of retinal ganglion cells. ²³ Experimental animals were generated by breeding Ins2^Akita/+ males to Ins2^+/+ wild-type Thy1-CFP or -YFP transgenic females. 

Heterozygous female Thy1-CFP and -YFP mice were bred with male Ins2^Akita/+ mice. Only the male progeny carrying the fluorescent reporters were used in the study, because female Ins2^Akita/+ diabetic mice tend not to maintain significant hyperglycemia. ⁶ In all studies, the male Ins2^+/+ Thy1-CFP and -YFP nondiabetic littermates were used as the control. Diabetes was defined as blood glucose greater than 250 mg/dL (One-Touch Lifescan Meter, Milpitas, CA; Table 1 ). All mice were housed with a standard 12-hour light–dark schedule with ad libitum food and water in the Penn State College of Medicine animal facility, in accordance with the Institutional Animal Care and Use Committee guidelines, the NIH Guidelines for the Care and Use of Laboratory Animals, and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Diabetic mice were not given routine insulin injections. The retinas from nondiabetic littermates were used as control specimens. At 4 months of age (3 months of untreated diabetes) the mice were anesthetized with sodium pentobarbital (100 mg/kg, IP) and decapitated. 

Immunohistochemistry

Eyecups were fixed in 2% paraformaldehyde in phosphate buffered saline for 2 to 4 hours at room temperature and blocked in 10% donkey serum with 0.3% Triton X-100. Retinas from Thy1-CFP and -YFP mice (groups 1 and 2, respectively; Table 1 ) were incubated in goat anti-choline acetyltransferase (ChAT; 1:100, AB134; Chemicon, Temecula, CA) and a rabbit anti-green fluorescent protein (GFP) conjugated to AlexaFluor 488 (1:500, A21311; Invitrogen-Molecular Probes, Eugene, OR) antibodies for 5 days at 4°C, followed by affinity-purified donkey anti-goat Cy5 secondary antibody overnight (Jackson ImmunoResearch Laboratories, West Grove, PA). Retinas from Ins2^Akita/+ and wild-type mice (Table 1 ; group 3) were incubated in rabbit anti-mouse melanopsin (1:5000), ²⁴ a generous gift from Ignacio Provencio (University of Virginia, Charlottesville, VA), followed by affinity-purified donkey anti-rabbit Cy3 (Jackson ImmunoResearch Laboratories). The retinas were counterstained with Hoechst (0.5 μg/mL; Sigma-Aldrich, St. Louis, MO) for 2 hours at room temperature. All retinas were flat mounted and coverslipped, ganglion cell side up (Aqua poly/mount; Polysciences Inc., Warrington, PA). 

Confocal Imaging

Images were acquired with a confocal microscope (TCS SP2 AOBS; Leica Microsystems, Exton, PA), with a 488-nm laser for anti-GFP-immunolabeled CFP and YFP (bandpass, 490–520 nm), a 543-nm laser for Cy3 (bandpass, 555–590 nm), and a 633 laser for Cy5 (bandpass, 650–680 nm). Images were acquired with a sequential line scan at 512 × 512- or 1024 × 1024-pixel resolution. For each ganglion cell, a series of optical sections were collected from the soma through the entire dendritic field. The optical sections were reconstructed using the microscope software (Leica Microsystems, Deerfield, IL), by an average projection for cell counting and a maximum projection for neuron tracing. The brightness and contrast of images were optimized with image-enhancement software (Photoshop ver. 8.0; Adobe Systems Inc., San Jose, CA). 

Wholemount Thy1-CFP retinas (Table 1 ; group 1) were imaged using a 20× oil-immersion objective. Images were acquired along the horizontal and vertical meridians: four in the central and four in the peripheral retina. The central retina was defined by a radius of 1050 μm centered on the optic disc. Each image sampled an area of 0.562 mm². Serial optical sections were acquired through the ganglion cell layer (GCL) with a 1.0-μm step size. Confocal image stacks were imported into Image J ²⁵ and an average projection was generated. The gain was adjusted to eliminate background for the images analyzed. The number of CFP-positive somas was counted in each field, in a masked fashion, using the automated Nucleus Counter plug-in program for Image J. 

To measure total retinal area, we imaged flatmounted retinas on a research stereo microscope (SZH10; Olympus, Lake Success, NY) with a charge-coupled device (CCD) video camera (DXC-960MD; Sony Corp., New York, NY). Total retinal area was measured by using the Polygon tool in Image J. The area of the central retina was estimated to be 3.46 mm², whereas the peripheral retinal area was determined by subtracting the central area from the total retinal area. The total number of CFP-positive ganglion cells/retina was estimated by multiplying the mean ganglion cell density by the area in the peripheral and central regions and summing these two numbers. 

Morphometric Analysis

Using the confocal microscope software (Leica), we estimated the dendritic stratification for each cell relative to the two bands of ChAT-immunoreactive processes at 30% and 70% within the IPL. The mean stratification depth was determined from four regions of interest, ranging in size from 0.05 to 1.5 mm². A maximum projection was generated and the cross-sectional soma area was measured. Each neuron was then traced (Photoshop; Adobe) to create a two-dimensional line tracing of the dendrites for each cell. The line tracing was analyzed using Image J. ²⁵ Dendritic field (DF) area was measured as a polygon around the distal tips of the dendrites. The total length of the ganglion cell dendrites was measured with Neuron J. ²⁶ The dendritic branching was examined by using a Sholl analysis, which measures the degree of branching as a function of distance from the soma. ²⁷ Briefly, a series of concentric circles were centered on the soma at 12.5 μm intervals, and the number of dendrites intersecting each circle was counted. 

Statistical Determination of Cell Clusters

For the purpose of grouping and analyzing the characteristics of retinal ganglion cells, each cell was considered independent, regardless of the fact that multiple cells came from the same animal. A dissimilarity matrix was constructed using the cell parameters as described by Kong et al. ²⁸ : stratification depth, dendritic field area, and dendrite density. The cells were then clustered using the method of partitioning around the medoids. ²⁹ In this clustering technique, k clusters are constructed by first selecting k representative cells, such that these cells are centrally located within the clusters they define. The average dissimilarity of a representative cell to all other cells in the same cluster is then minimized. Clusters from size 2 to 30 were formed, and silhouette plots were generated to determine the optimal number of clusters. The cells best fit into six clusters with an average silhouette width of 0.45 (Table 2) . Morphometric characteristics of cells were compared between diabetic and nondiabetic control animals within each cluster by using analysis of variance models. 

A linear mixed-effects model, which accounts for the repeated measurements per cell (i.e., the annuli per cell), was fit to the data, to compare the number of dendrite intersections per annulus between cells within each cluster in retinas from Ins2^Akita/+ and wild-type mice. ³⁰ No correction for multiple comparisons testing was used in these exploratory analyses. All analyses were performed with one of two software programs (S-plus, ver. 7.0; Insightful Corp., Seattle, WA, or SAS, ver. 9.1; SAS Institute Inc., Cary, NC). 

Results

Animals

Diabetes was confirmed by blood glucose greater than 250 mg/dL at the time of death. In all groups, blood glucose was significantly elevated in the Ins2^Akita/+ mice, compared with that in the nondiabetic wild-type littermates (Table 1) , whereas the body weight of diabetic animals was significantly less than that of control subjects. There was no significant difference in total retinal area of Ins2^Akita/+ and wild-type mice. 

Thy1-CFP-Labeled Retinal Ganglion Cells

Retinas from Thy1-CFP mice had CFP-positive cell bodies in the GCL (Fig. 1) . Somas of all sizes were CFP-positive, from 8 to 14 μm in diameter. In the central retina, ganglion cell axon bundles radiating from the optic nerve head were also CFP-positive. Immunohistochemistry for ChAT was used to rule out the possibility that any of these cells were displaced cholinergic amacrine cells. ³¹ None of the ChAT-immunoreactivity colocalized with CFP-positive cells (Fig. 1C) , indicating that none of the CFP-positive cells were displaced cholinergic amacrine cells. 

Quantification of Retinal Ganglion Cells in Diabetes

The number of CFP-positive cells in the GCL was counted in the central and peripheral retinas of 3-month diabetic Ins2^Akita/+ Thy1-CFP mice (n = 7) and nondiabetic wild-type mice (n = 5; Fig. 2 ). In the peripheral retina, the mean density of CFP-positive cells from wild-type and Ins2^Akita/+ mice was 1222.8 ± 46.34 and 1022.2 ± 49.24 cells/mm², respectively (mean ± SEM, P < 0.017), representing a 16.4% difference. There was no significant difference in the density of CFP-positive cells in the central retina of Ins2^Akita/+ mice compared with wild-type control subjects (1145.9 ± 55.64 and 1282.0 ± 84.18, respectively). The total number of CFP-positive cells estimated in wild-type retinas was 27,463 ± 12,282 and that of Ins2^Akita/+ mice was 23,042 ± 1,266. 

Classification of Retinal Ganglion Cells

In Thy1-YFP mice, an anti-GFP antibody was used to enhance YFP-positive cells and labeled the entire structure of individual retinal ganglion cells. Seventy-seven ganglion cells in nondiabetic controls (six mice, eight retinas), and 69 cells in Ins2^Akita/+diabetic mice were analyzed by confocal microscopy after 3 months of hyperglycemia (four mice, six retinas). An average of 10 ± 4 (mean ± SD) YFP-positive cells were analyzed per retina. Retinal ganglion cells were categorized into six distinct morphologic subtypes based on stratification depth, dendritic field area, and dendrite density, using cluster analysis (Table 2) . 

Morphologic Abnormalities of Retinal Ganglion Cells in Diabetes

YFP-positive ganglion cells in retinas of 3-month-old Ins2^Akita/+ mice were analyzed for gross morphologic changes of the soma, dendrites, and axons. In control mice, ganglion cell somas were typically round or oval (Fig. 3A) . In Ins2^Akita/+ mice, retinal ganglion cell somas had morphologic abnormalities, such as large blebs (Fig. 3B) . The cross-sectional area of the soma was quantified in each cell cluster. Clusters 3 and 5 (ON-type medium and small cells) had significantly larger somas in the retinas of Ins2^Akita/+ mice, by 20% (P = 0.03) and 25% (P = 0.002), respectively, compared with wild-type (Fig. 3C , P < 0.05). The soma size in control mice ranged from 102 to 348 μm², where only 20% of the cells sampled were larger than 300 μm² in area. The soma size in Ins2^Akita/+ mice ranged from 121 to 452 μm², and 43% were larger than 300 μm². The soma size of the other cell clusters was not significantly different in Ins2^Akita/+ mice compared with that in wild-type (data not shown). 

Every retinal ganglion cell cluster in Ins2^Akita/+ mice contained individual cells with morphologic abnormalities. Most common was swelling of the primary dendrite near the soma. Approximately 30% of axons also had extensive swelling, 6 to 8 μm in diameter, typically within 50 to 100 μm of the soma, associated with thinning of the axon (Fig. 4) . 

Dendrite morphology was measured in each cell cluster in Ins2^Akita/+ after 3 months of hyperglycemia (Table 3) . Cluster 6 cells had significantly longer total dendrite length (18.6% longer) and significantly more terminal dendrites (32.4%) than did wild-type cells (P < 0.05, Fig. 5 ). Terminal dendrites often projected beyond the normal perimeter of the dendritic arbor of the cell. These cells resemble ON-α ganglion cells, identified by a large DF, large soma, small number of primary dendrites, and relatively sparse but symmetric dendritic arbor in the ON sublamina of the IPL ³² (Table 2) . 

Dendrite morphology was further analyzed by using a Sholl analysis that measures the density of dendrites as a function of distance from the soma (Fig. 5) . In cluster 6 cells (large ON-type), 15.3% more Sholl intersections were measured in Ins2^Akita/+ mice compared with wild-type (P < 0.05; Table 3 ). An analysis of the number of dendrites at individual Sholl circles for cluster 6 cells revealed 18.6% higher dendritic density between 125 and 150 μm from the soma, in the retinas from Ins2^Akita/+ mice compared with wild-type (P < 0.05). Cluster 3 cells (medium ON-type) had 17.0% more dendrites between 75 and 87.5 μm from the soma in retinas from Ins2^Akita/+ mice compared with wild-type (P < 0.05). The dendrite density of the other cell clusters was not significantly different in Ins2^Akita/+ compared with wild-type (data not shown) mice. 

A separate study examined the morphology of melanopsin-containing retinal ganglion cells. These cells are a sparse population with large dendritic fields that are intrinsically responsive to light. ²⁴ In Thy1-YFP mice, melanopsin ganglion cells are only rarely found ²³ and are separate from the other six clusters of cells, and so an antibody was used to label all melanopsin-containing ganglion cells. After 6 months of diabetes in Ins2^Akita/+ mice, primary dendrite swelling and multiple varicosities on axons close to the soma were observed in melanopsin-immunoreactive cells (Fig. 6) . None of these abnormalities was observed in wild-type mice, suggesting that the dystrophic effects of diabetes are not limited to the Thy1-YFP-positive cells. 

Discussion

Neurodegeneration of the retina is an important early consequence of diabetes, occurring in concert with the more established vascular lesions. ^{4 5} This study establishes that morphologic abnormalities are increased in retinal ganglion cells after 3 months of diabetes in the Ins2^Akita/+ mouse. The depletion of retinal ganglion cells occurs within the same time frame as the previously established loss of cholinergic and dopaminergic amacrine cells, also in the Ins2^Akita/+ mouse. ⁷ The results of the present study expand our understanding of the effect of diabetes on ganglion cells in several ways: (1) retinal ganglion cells are preferentially lost in the peripheral regions of the mouse retina; (2) abnormal swellings develop on axons and dendrites; (3) somas of small and medium ON-type cells are enlarged; and (4) dendritic branching is increased in medium and large ON-type cells. Together, these findings provide a new level of specificity and resolution to studies of retinal neurodegeneration in diabetes. 

The 16.4% loss of CFP-positive retinal ganglion cell bodies from the peripheral retina in Ins2^Akita/+ agrees with previous studies reporting loss of cell bodies in the retinal ganglion cell layer of cross-sections of diabetic mouse eyes, although a direct quantitative comparison between these two techniques is difficult. ^{3 10} The average population of CFP-positive retinal ganglion cells was estimated at 27,463 in control mice, and probably represents approximately 50% of the total number of ganglion cells in the mouse retina. ³³ The significant reduction in retinal ganglion cell density in the peripheral but not central retina reported herein also agrees with the more dramatic reduction in thickness of the IPL of the peripheral retina compared with the central retina of Ins2^Akita/+ mice. ⁶ Taken together, these data suggest that retinal ganglion cells are preferentially lost from the peripheral retina of Ins2^Akita/+ mice during the first three months of hyperglycemia. 

Diabetes induced significant morphologic changes in the dendrites of large and medium retinal ganglion cells, identified by a large dendritic arbor, a large soma, a small number of primary dendrites, and a relatively sparse but symmetric dendritic arbor in the ON sublamina of the IPL. ³² In a recent study of diabetic rat retinas, it was noted that large ganglion cells had enlarged dendritic fields. ³⁴ The results presented herein expand on this finding by showing that the dendrite morphology of large ON-α ganglion cells are also altered in Ins2^Akita/+ mice. In mammalian retinas, α-like ganglion cells that survived optic nerve axotomy and experimental glaucoma also expanded their dendritic fields. ^{16 35 36} Conversely, a study of a primate glaucoma model using Sholl analysis found a decrease in dendritic branching of parasol cells and linked this to a reduced responsiveness of individual neurons to temporal and spatial stimuli. ^{18 19} The proliferation of ganglion cell dendrites in diabetes may be a compensatory response to a loss of presynaptic input from cholinergic amacrine cells ³⁷ previously established as being depleted by diabetes, ⁷ although it is uncertain whether the additional dendrites make new synapses. 

The data presented herein confirm and extend similar findings reported recently for rat and human retinas. ^{34 38} The previous studies used a DiI labeling technique that resulted in a random sampling of all morphologic subtypes of retinal ganglion cells. In the present study the Thy1-CFP and -YFP transgenic mice provided excellent noninvasive tools to investigate retinal ganglion cells. Thy1-YFP also enabled categorization of ganglion cells into six separate morphologic clusters using well-established morphometric methods for mouse retina. ^{23 28} Although Thy1 is expressed by Müller cells during periods of ischemia, there was no evidence of this in the present study. ³⁹ The most likely reason for the different expression patterns of Thy1-CFP and -YFP is position–effect variegation, in which the degree of gene expression is influenced by neighboring genes close to the insertion point of the transgenic DNA. ²²  

The electrophysiological response of retinal ganglion cells can be determined by the pattern electroretinogram. ^{40 41} Diabetes induces a delayed latency and reduced amplitude pattern electroretinogram, ^{42 43 44 45} suggesting that a loss of presynaptic input to retinal ganglion cells may occur. These electrophysiological deficits may be explained by the combination of cell loss or dendritic abnormalities described. An electrophysiological study using a low-spatial-frequency pattern electroretinogram further confirms that retinal ganglion cells with large dendritic fields are preferentially affected in diabetes. ⁴⁶ ON-α ganglion cells may also play a role in the development of contrast thresholds. ^{47 48} The loss of contrast sensitivity in diabetic retinopathy ⁴⁹ may be attributable to the preferential changes in ON-α ganglion cells, in combination with a lack of change in OFF-ganglion cells. 

The dystrophic swellings on retinal ganglion cell axons and somas noted in this study are similar to those observed in postmortem retinas from human donors with diabetes. ³⁸ Dystrophic axons of retinal ganglion cells in the nerve fiber layer of retinas from humans with diabetes have neurofilament accumulation and electron dense bodies. ⁵⁰ Similar swellings containing neurofilaments were also reported on retinopetal axons in retinas of streptozotocin diabetic rats. ⁵¹ These structural abnormalities may be related to slowed axonal transport ⁵² and compromised retrograde transport in large and medium-sized ganglion cell axons in the optic nerve. ^{53 54} The enlarged soma size noted in this study may be related to apoptosis. An initial response of ganglion cells to the early stages of experimental glaucoma is an enlarged soma. ¹⁹ Surviving ganglion cells after optic nerve section also develop blebs on the soma surface. ⁵⁵ Alternatively, soma swelling may be an adaptive response to axonal damage and not necessarily a predictor of apoptosis. ⁵⁶  

It is unlikely that the retinal ganglion cell abnormalities and cell loss were artifacts induced by Thy1-CFP and -YFP expression. Thy-1 mRNA has been used as an index of retinal ganglion cell loss in other models of retinal damage, ⁵⁷ and a short period of diabetes did not alter this expression. ⁵⁸ Furthermore, there were no obvious cytotoxic effects of YFP and CFP expression in the control mice in this study or in other studies using Thy1-YFP transgenic mice. ⁵⁹ Expression of very high amounts of green fluorescent protein may be cytotoxic to neurons, but these very high doses may only be achievable by viral transfection. ⁶⁰ No basal or phototoxicity was detected when neuromuscular junctions of Thy1-YFP mice were imaged repeatedly over a 9-month period. ²² Finally, the degree of ganglion cell loss agreed with that predicted by previous measures of apoptosis in the Ins2^Akita/+ mice. ⁶  

In conclusion, there is an increase in dendritic branching patterns in large and medium ON-type retinal ganglion cells of Ins2^Akita/+ mice. There is also a significant loss of retinal ganglion cells accompanied by abnormal axonal and dendritic swelling and increased soma size in surviving cells. The increase in dendritic branching in ON-α-like ganglion cells may be a compensatory response for loss of neighboring cells or presynaptic input. A progressive reduction in the abundance of retinal ganglion cells, accompanied by significant alterations to the morphology of surviving cells could be a basis for some of the early visual deficits associated with diabetes. 

 Supported by a Juvenile Diabetes Research Foundation (JDRF) Postdoctoral Fellowship (MJG), Grant 1-05-RA-09 from the American Diabetes Association (AJB), the Pennsylvania Lions Sight Conservation and Eye Research Foundation (AJB), and the JDRF Diabetic Retinopathy Center.

 Submitted for publication June 6, 2007; revised November 14, 2007, and January 23, 2008; accepted March 31, 2008.

 Disclosure: M.J. Gastinger, None; A.R. Kunselman, None; E.E. Conboy, None; S.K. Bronson, None; A.J. Barber, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Table 1.

 

Weight and Blood Glucose of Animals at Time of Death

Group	Age (wk)			n	Weight (g)	Blood Glucose (mg/dL)	Retinal Area (mm2)
1	15	Thy1-CFP	Control	5	28.0 ± 0.83	198.8 ± 9.83	22.3 ± 0.23
			Diabetic	7	25.3 ± 0.35*	474.1 ± 43.0*	22.1 ± 0.60
2	15	Thy1-YFP	Control	6	29.2 ± 1.09	165.6 ± 12.2	N/A
			Diabetic	4	23.6 ± 0.19*	487.5 ± 42.9*	N/A
3	28	Ins2Akita	Control	6	38.9 ± 3.24	202.2 ± 11.1	20.3 ± 0.61
			Diabetic	6	24.2 ± 0.95*	494.8 ± 21.4*	21.6 ± 0.65

*   P < 0.01, unpaired one-tailed t-test compared with corresponding control animals.

Table 2.

 

Cluster Analysis

Cluster	Control/Ins2Akita Diabetic	IPL Depth (%)	DF Area (μm2)	Dendritic Density (μm/μm2)	Morphological Features	Coombs et al. 23
1	10/7	5.3	61855 ± 15043	0.068 ± 0.008	OFF-type, large DF, long terminal branches	M7 Off
2	10/7	10.4	32223 ± 6483	0.099 ± 0.01	OFF-type, small DF, overlapping dendrites, asymmetric	M3 Off, M5, M4
3	11/13	84.9	63144 ± 6693	0.060 ± 0.007	ON-type, medium sized DF, low-density dendritic branching	M3 On
4	14/10	85.7	24241 ± 6625	0.096 ± 0.008	ON-type, smallest DF, highest dendritic density	M1, M2
5	24/23	82.2	38759 ± 7501	0.072 ± 0.007	ON-type, small DF, rare overlapping branches	M8
6	8/9	85.1	99187 ± 17146	0.047 ± 0.007	ON-type, largest DF, symmetric, sparse, nonoverlapping branches	M10

 GCL = 100%, INL = 0%.

 All data are expressed as the mean ± SD.

Figure 1.

 

CFP-positive neurons in the GCL. A Thy1-CFP mouse retina was double labeled with antibodies to GFP (green) and ChAT (red). (A) CFP-positive cells (arrowheads) were labeled with the anti-GFP antibody (image samples from mid-peripheral retina). Ganglion cell axon bundles (arrow) were also apparent throughout the retina. (B) ChAT-immunoreactive displaced amacrine cells (red) in the GCL. (C) Merged image, demonstrating that none of the CFP-positive cells colocalized with the ChAT immunoreactive amacrine cells. Scale bar, 20 μm.

Figure 2.

 

Diabetes decreased the Thy1-CFP cell density in mouse retinas. Retinas from Ins2^Akita/+ Thy1-CFP mice were flatmounted with the GCL facing up, and imaged by confocal microscopy. (A) Scale diagram of a retina illustrating the 750 × 750-μm regions where confocal images were acquired. Dashed circle: the area of the central retina (radius, 1050 μm). (B, C) CFP-positive cells in the GCL of the central (B) and peripheral (C) retina were easily identified in maximum projections of z-stacks. (D) The average density (± SEM) of CFP-positive cells was calculated for the central and peripheral retina of nondiabetic control (white) and Ins2^Akita/+ (black) mice. There was no significant difference in retinal ganglion cell density in the central region. The cell density was significantly less in the peripheral retina of Ins2^Akita/+ mice compared to controls (*P < 0.05). n = 5 wild-type, n = 7 Ins2^Akita/+ diabetic. Scale bar: (A) 1 mm; (B, C) 100 μm.

Figure 3.

 

Diabetes increases the soma size of ON-type retinal ganglion cells. The soma size of YFP-positive retinal ganglion cells was measured by confocal microscopy. (A) A typical medium ON-type cell from control retina, with regular radiating dendrites and oval soma; (B) a similar cell from the Ins2^Akita/+ mouse with enlarged soma. (C) Soma size was quantified by measuring the cross-sectional area. The mean (±SEM) soma size of cells in cluster 3 and 5 (medium and small ON-type) was significantly larger in Ins2^Akita/+ mice (black) compared to the controls (white; *P < 0.05). Scale bar: (A, B) 20 μm.

Figure 4.

 

Axonal swellings of Thy1-YFP-positive retinal ganglion cells in diabetes. Axonal swellings were noted on several retinal ganglion cells of Ins2^Akita/+ Thy1-YFP diabetic mice. (A) The entire dendritic arbor of a retinal ganglion cell after 3 months of diabetes. (B) Enlarged image of boxed region in (A), showing axon swelling (arrowhead), approximately 60 μm from the soma, preceded by a prominent thinning of the axon (arrow). Scale bars: (A) 50 μm; (B) 20 μm.

Table 3.

 

Dendrite Analysis

Cluster	Dendrite Length (μm)			Terminal Dendrites (n)			Sholl Dendrites (n)
		Control	Ins2Akita Diabetic	Control	Ins2Akita Diabetic	Control	Ins2Akita Diabetic
1	4228 ± 197	3959 ± 236	53 ± 4	54 ± 5	258 ± 12	238 ± 14
2	3245 ± 197	2994 ± 236	75 ± 4	68 ± 5	182 ± 12	165 ± 14
3	3525 ± 188	3998 ± 173	47 ± 4	54 ± 4	209 ± 11	233 ± 11
4	2334 ± 167	2245 ± 197	58 ± 3	56 ± 4	125 ± 10	120 ± 12
5	2772 ± 127	2840 ± 130	57 ± 3	55 ± 3	157 ± 8	158 ± 8
6	4199 ± 220	4980 ± 208*	37 ± 5	49 ± 4*	268 ± 13	309 ± 13*

*   P < 0.05 compared to cells in nondiabetic control retinas.

Figure 5.

 

Sholl analysis of dendrites in Thy1-YFP ON-type retinal ganglion cells in Ins2^Akita/+ diabetic mice. Line tracings were made of YFP-positive ganglion cells in Ins2^Akita/+ and wild-type control mice. Sholl analysis calculated the number of dendrites intersecting with a series of concentric circles centered on the soma of each cell. (A) A cluster 6 cell from a control mouse. (B) A cluster 6 cell from a 3-month-diabetic Ins2^Akita/+ mouse retina. The dendrites of these cells had many small terminals (arrowheads), not typically observed in the control retinas. Long, unbranched dendrites often extended beyond the normal dendritic arbor (arrow). (C) Sholl dendrite analysis, measured dendrite density by placing a series of concentric circles, spaced at 12.5-μm intervals centered on the soma (*). The number of dendritic intersections with each circle was counted. (D) Mean (±SEM) number of intersections with each circle in cluster 6 ganglion cells. (E) Mean (±SEM) number of intersections in cluster 3 cells (*P < 0.05). Scale bar: (A–C) 50 μm.

Figure 6.

 

Morphologic abnormalities in melanopsin-immunoreactive ganglion cells of Ins2^Akita/+ mice. Flatmount mouse retinas were labeled with an antibody to melanopsin and imaged by confocal microscopy. (A) Melanopsin-immunoreactive ganglion cell in a control retina had normal dendrites radiating from the soma. (B) In retinas from Ins2^Akita/+ mice, the primary dendrite was swollen to 8 to 10 μm in diameter as it emerged from the soma (arrowhead). (C) Axons (arrow) in the peripheral retina had multiple swellings (arrowheads). Scale bar: (A–C) 50 μm.

The authors thank Wendy Holtry of the JDRF (Juvenile Diabetes Research Foundation) Animal Core Facility for generating and maintaining the mouse colonies, Heather D. VanGuilder, PhD, for help with figures and Thomas W. Gardner, MD, MS, for critical assessment of the final manuscript.