Loss of Cholinergic and Dopaminergic Amacrine Cells in Streptozotocin-Diabetic Rat and Ins2^Akita-Diabetic Mouse Retinas

Abstract

purpose. To identify amacrine cells that are vulnerable to degeneration during the early stages of diabetes.

methods. Whole retinas from streptozotocin (STZ)-diabetic rats and Ins2^Akita mice were fixed in paraformaldehyde. Apoptotic cells in the retina were quantified using terminal dUTP nick-end labeling (TUNEL) and active caspase-3 (CM-1) immunohistochemistry. Immunohistochemical markers for choline acetyltransferase (ChAT) and tyrosine hyroxylase (TH) were also used to quantify populations of amacrine cells in the Ins2^Akita mouse retinas.

results. The number of TUNEL-positive nuclei increased from 29 ± 4 in controls to 72 ± 9 in the STZ-diabetic rat retinas after only 2 weeks of diabetes. In rats, CM-1-immunoreactive (IR) cells were found primarily in the inner nuclear and ganglion cell layers after 2, 8, and 16 weeks of diabetes. At each end point, the number of CM-1-IR cells in the retina was elevated by diabetes. Approximately 2% to 6% of the CM-1-IR cells in the inner nuclear layer (INL) were double-labeled for TH immunoreactivity. After 6 months of diabetes in the Ins2^Akita mouse, the morphology of the labeled ChAT-IR and TH-IR amacrine cell somas and dendrites appeared normal. A quantitative analysis revealed a 20% decrease in the number of cholinergic and a 16% decrease in dopaminergic amacrine cells in the diabetic mouse retinas, compared with the nondiabetic control.

conclusions. Dopaminergic and cholinergic amacrine cells are lost during the early stages of retinal neuropathy in diabetes. Loss of these neurons may play a critical role in the development of visual deficits in diabetes.

Sight-threatening neurodegenerative diseases of the retina such as glaucoma, retinitis pigmentosa, and diabetic retinopathy involve the progressive loss of retinal neurons. Apoptosis is one possible mechanism of neuronal cell death in these disorders. Retinal ganglion cells undergo apoptosis in glaucoma, ¹ whereas photoreceptors are susceptible to apoptosis in retinitis pigmentosa. ² In diabetic retinopathy, apoptotic cells have been observed in all retinal layers, ^{3 4 5 6} suggesting that multiple types of retinal neurons may be affected by diabetes. Further evidence for neural cell loss in diabetes is the decrease in the thickness of the inner nuclear layer (INL) ^{3 6} and outer nuclear layer (ONL). ^{4 7} It is likely that activation of caspase-3 is part of the mechanism of apoptosis. A previous study demonstrated an increase in the number of caspase-3-positive cells after 4 weeks of diabetes in Ins2^Akita mice. ⁶ Caspase-3 enzyme activity has also been shown to increase in the retinas of STZ-diabetic rats ^{8 9} and Ins2^Akita mice. ¹⁰ Using an antibody to active caspase-3 may be a useful tool to identify the type of neurons undergoing apoptosis. 

Current evidence suggests that amacrine cells are affected by diabetes. Specifically, amacrine cells that use dopamine ^{11 12} and acetylcholine ^{13 14} as neurotransmitters have reduced enzyme activity for tyrosine hydroxylase and acetylcholinesterase in diabetes. However, it is not known whether these amacrine cells undergo apoptosis in diabetes. 

Neurodegeneration plays an important role in the pathology of diabetic retinopathy, but the specific cell types undergoing degeneration have not been identified. The use of animal models of diabetes is essential to identify the onset of cell death as well as the types of neurons undergoing degeneration in the retina. Using TUNEL and active caspase-3 immunoreactivity, apoptotic cells were identified and quantified in wholemounted retinas after 2 weeks of diabetes. Double-label immunofluorescence was used to identify apoptotic cells. After 6 months of diabetes in the Ins2^Akita mouse, populations of amacrine cells were quantified using confocal microscopy. The results show that neurons, including amacrine cells, undergo apoptosis, and thus contribute to the development of retinal neuropathy in rodent models of diabetes. 

Methods

Animals

Male C57BL6J Ins2^Akita mice (Jackson Laboratory, Bar Harbor, ME) were housed in the Penn State Hershey College of Medicine Juvenile Diabetes Research Foundation Diabetic Retinopathy Center Animal Core. Male Sprague-Dawley rats were made diabetic via a single tail vein injection of streptozotocin (STZ; 65 mg/kg dissolved in citrate buffer [pH 7.4]). Food and water were provided ad libitum in a room maintained on a 12-hour light-dark cycle. Diabetes was confirmed with blood glucose higher than 250 mg/dL (One-Touch Meter; Lifescan, Burnaby, BC, Canada) at 4 weeks of age in the mice and 1 week after STZ injection in the rats. Body weight and blood glucose for each experimental group were tested at time of death (Table 1) . Animals were anesthetized with sodium pentobarbital (0.1 mg/g) and decapitated. All methods and care were in accordance with the Penn State Milton S. Hershey College of Medicine Institutional Animal Care and Use Committee guidelines and adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Terminal Transferase dUTP Nick-End Labeling

One retina from each animal in groups 1 and 2 (Table 1)was isolated and immersion-fixed in 4% paraformaldehyde for 10 minutes at room temperature and wholemounted on microscope slides (Superfrost Plus; Fisher Scientific, Pittsburgh, PA). Retinas were processed with an apoptosis-detection kit (ApopTag Peroxidase In Situ kit; Chemicon, Temecula, CA) using a modification of methods previously described. ³ 3-Amino-9-ethylcarbazole (AEC; Sigma-Aldrich, St. Louis, MO) in 50 mM sodium acetate buffer (pH 5.0) was used as the substrate for horseradish peroxidase during development of the TUNEL-positive label. This reaction produced a dark red precipitate that was easily identified by light microscopy. As a positive control, the prostate from a castrated and a normal rat were immersion-fixed in 10% formalin for 1 hour at room temperature and frozen in optimal cutting temperature compound. Prostate tissue from normal and castrated rats was cut into 20-μm sections on a microtome cryostat (HM505E; Microm, Walldorf, Germany) and processed in parallel with the retinas. As a negative control, the terminal transferase enzyme was omitted from sections of the degenerative prostate. TUNEL-positive neurons were counted in whole retina on a research microscope (BH-2; Olympus, Melville, NY) with a 40× objective and standardized to a 0.5-cm² area, after measuring the total retinal area with a dissection microscope and image analysis software (Optimus; Media Cybernetics, Silver Spring, MD). 

Immunohistochemistry

Rat eyes (Table 1 , groups 3–5) were enucleated and the retinas isolated and immersion-fixed in 2% paraformaldehyde in phosphate-buffered saline (PBS [pH 7.4]) for 10 minutes at room temperature. The left eyes of mice from group 6 (Table 1)were enucleated and immersion-fixed in 2% paraformaldehyde in PBS for 2 hours at room temperature. The retina was carefully isolated from the retinal pigment epithelium and a small point mark was made in the superior quadrant using a high-temperature, fine-tip cautery pen (2200°F; Medi-Pak Surgical Cautery; McKesson Medical-Surgical, Richmond, VA) to orient the retina. 

Whole retinas were incubated in 10% donkey serum with 0.3% Triton X-100 in PBS (PBST) for 2 to 4 hours at room temperature. All retinas were incubated for 3 to 5 days at 4°C in primary antibodies diluted in PBST. Rat retinas labeled with a rabbit polyclonal anti-active caspase-3 (CM-1; BD Biosciences, Mountain View, CA) were double labeled with mouse anti-neuronal nuclei (NeuN; 1:1000, mAb377; Chemicon), monoclonal mouse anti-tyrosine hydroxylase (TH; 1:10,000, T2928; Sigma-Aldrich), or mouse anti-agrin (1:1000, clone AGR131; StressGen, Victoria, BC, Canada). Ins2^Akita mouse retinas were double labeled with an affinity-purified polyclonal goat anticholine acetyltransferase (ChAT; 1:100, AB144P; Chemicon) and the TH antibody. Retinas were then incubated overnight at 4°C in F(a,b′)₂ fragments of affinity-purified secondary antibodies (Jackson ImmunoResearch, West Grove, PA) in PBS with 10% donkey serum: donkey anti-rabbit RRX (1:1000), donkey anti-mouse Cy2 (1:1000), donkey anti-goat Cy5 (1:1000), and donkey anti-rabbit Cy3 (1:2000). To stain the nuclei of all cells, bis-benzimide (0.5 μg/mL Hoechst; Sigma-Aldrich) was added to the secondary antibody incubation. All retinas were coverslipped ganglion cell side up in aqueous mounting medium (Aqua poly/mount; Polysciences Inc., Warrington, PA). 

Image Acquisition

Images were acquired with a laser confocal microscope (TCS SP2 AOBS, Leica Microsystems, Manheim, Germany), using a 488-nm laser for Cy2, a 543-nm laser for Cy3, and a 633-nm laser for Cy5 fluorophores. All three fluorophores were imaged with a sequential line scan. Each image was saved at a resolution of either 512 × 512 or 1024 × 1024 pixel image size. The optical sections were reconstructed with a maximum projection using the microscope software (Leica). The brightness and contrast were optimized in image-analysis software (Photoshop ver. 7.01; Adobe Systems, Inc., Mountain View, CA). Whole retinas were viewed on a research stereomicroscope (model SZH10; Olympus, Tokyo, Japan) and imaged with a charge-coupled device (CCD) color video camera (DXC-960MD, Sony Corp., New York, NY). Total retinal area was measured by using the polygon tool in Image J (developed by W. S. Rasband, provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD; http://rsb.info.nih.gov/ij/). 

Analysis

Wholemount retinas (Table 1 , group 6) were analyzed by acquiring images on the confocal microscope at approximately 400-μm intervals along the superior-inferior and nasal-temporal meridians of the retina. With the 40× oil objective, each image covered an area of 0.141 mm². Serial optical sections were acquired first through the ganglion cell layer (GCL), then the inner nuclear layer (INL) with a 1.0-μm step size. Confocal files (Leica) were imported into Image J and projected maximally. For each labeled cell type, the somas in each field were counted using the Cell Counter plug-in for Image J. The density of cells in each field was determined by dividing the area of each region. To estimate the total number of cells in each retina, the mean density for all regions was multiplied by the area of the retina. 

Nearest Neighbor and Autocorrelation Analysis

The pixel coordinates for the displaced and conventional ChAT amacrine cells were determined via the Point Picker plug-in for Image J and converted to micrometers based on the pixel image size of the original confocal stack. The coordinates of each cell were imported into WinDRP (ver. 1.6.4, http://sun0.mpimf-heidelberg.mpg.de/∼teuler/WinDRP/ReadMe.htm/ R. H. Masland, Howard Hughes Medical Institute, Harvard Medical School, Boston, MA). For each cell in the sampled field, the distance to its nearest neighbor (NN) was measured and the mean NN distance for each field calculated. To verify the regularity of the mosaic, a random distribution of the cells (averaged 10 times) was generated and the mean NN was calculated. The conformity ratio (CR) was defined as the ratio of the mean NN to the standard deviation, a classic measure of mosaic regularity. ¹⁵  

To characterize the spatial organization further, an autocorrelation of each field was generated, and a density recovery profile was created using WinDRP. Briefly, the autocorrelation analysis takes each cell in the field and places it in the center, then maps all the other cells relative to that position. A central exclusion zone is associated with each autocorrelogram. The size of this zone was estimated via the WinDRP software and is referred to as the effective radius. ¹⁵ The effective radius represents a measure of the empty space surrounding each cell of the mosaic. 

Statistical Analysis

Statistical analyses were performed on computer (Statistica ver. 7.0; StatSoft Inc., Tulsa, OK). Cell counts were analyzed by one-way ANOVA. Nonparametric Kruskal-Wallis-by-ranks tests were used to verify the significance of the spatial indices. All results are reported as the mean ± SEM. 

Results

The heterozygous Ins2^Akita mice carry a point mutation in the insulin 2 gene that leads to a failure to secrete the insulin gene product, ultimately causing pancreatic β cell loss. In these Ins2^Akita mice elevated blood glucose (>250 mg/dL) develops at 4 weeks of age, ¹⁶ whereas the wild-type mice maintain a normal blood glucose. In rats, the STZ injection selectively kills the pancreatic β cells, ¹⁷ resulting in elevated blood glucose (>250 mg/dL). A summary of the weight and blood glucose of all the animals used in these experiments is presented in Table 1 . In both the rats and mice, the blood glucose was significantly elevated at the time of death (P < 0.01) and the diabetic animals weighed significantly less than the age-matched controls (P < 0.01). 

Effect of Diabetes on Retinal Apoptosis

Whole-mounted retinas from 2-week STZ-diabetic rats (group 1) and 3-month Ins2^Akita diabetic mice (group 2) were processed for TUNEL. The TUNEL-positive nuclei were identified by a dark red reaction product and were found in all regions of the retina. The Ins2^Akita mice had 72% more TUNEL-positive nuclei than the nondiabetic littermate control (Table 2 , P < 0.05). After only 2 weeks, STZ-diabetic rat retinas had 2.5-fold more TUNEL-positive nuclei compared to the control (P < 0.0002). 

Effect of Diabetes on the Number of CM-1-IR Cells in STZ-Diabetic Rat Retinas

Wholemounted retinas from rats after 2, 8, and 16 weeks of STZ diabetes were labeled with an antibody that recognizes active caspase-3, an indicator of apoptosis. ¹⁸ The CM-1-IR cells were typically ∼8 μm in diameter and located sporadically throughout the retinas of both diabetic and age-matched control rats. All CM-1-IR cells were spatially separate from agrin immunoreactivity (Fig. 1) , suggesting they were not endothelial cells or pericytes. To determine which CM-1-IR cells are neurons, retinas were double labeled with the neuronal marker NeuN which labeled nuclei in the INL and GCL. Approximately 6% to 8% of CM-1-IR cells were colocalized with NeuN-IR neurons in the INL (Figs. 2A 2B 2C)and GCL (not illustrated). To further identify CM-1-IR cells, retinas were double labeled with an antibody to TH, which labels dopaminergic amacrine cells (Figs. 2D 2E 2F) . Occasionally, CM-1-IR cells were found within the outer nuclear layer (ONL; Figs. 2G 2H 2I ), suggesting that photoreceptors also undergo apoptosis. 

To quantify the CM-1-IR cells in STZ-diabetic rats and the age-matched controls, all CM-1-IR cells were counted in the wholemounted retina. The highest density of CM-1-IR cells was found in 2-week STZ-diabetic rats, 120 ± 15 cells/0.5 cm² compared with 34 ± 6 in the control (P < 0.0002). The density of CM-1-positive cells was also at least threefold greater than in the control at both 8 and 16 weeks of diabetes (P < 0.05; Table 2 ). Taken together, these data show that nonvascular retinal cells in STZ-diabetic rats undergo apoptosis involving activation of caspase-3 soon after the onset of diabetes. 

Immunofluorescence for Amacrine Cells in the Ins2^Akita Mouse Retina

To quantify the loss of specific types of amacrine cells after 6 months of hyperglycemia, whole retinas from Ins2^Akita-diabetic and littermate control mice were labeled with antibodies against ChAT and TH (Table 2 , group 6). ChAT-IR cholinergic amacrine cells were located in the GCL (Fig. 3A)and INL (Fig. 3B) . The cell bodies of these amacrine cells were small, typically 8 μm in diameter, and distributed in an ordered array throughout the retina. ChAT-IR processes formed two dense plexuses in the inner plexiform layer (IPL; Fig. 3C ). The ChAT-IR amacrine cells in the Ins2^Akita-diabetic mouse retinas showed no gross morphologic changes compared with nondiabetic littermates. 

TH-IR dopaminergic amacrine cells were located in the INL (Fig. 3D) . These cells were large in diameter (10–15 μm) and had large primary dendrites that ramified in stratum 1 of the IPL, as described previously in the mouse retina. ^{19 20 21} Other smaller TH-IR cells were also labeled, which were probably nondopaminergic catecholamine cells. The TH-IR amacrine cells in the Ins2^Akita-diabetic mouse retinas showed no gross morphologic changes compared with nondiabetic littermates. 

Diabetes and the Number of Amacrine Cells

The inner layers of the Ins2^Akita mouse retina are reduced in thickness during diabetes, ⁶ probably due to cell loss or dendritic atrophy. To determine whether cholinergic neurons in the retina are involved in degeneration, retinas were analyzed after 6 months of diabetes (Table 1 , group 6). Each of the two types of cells was quantified in whole retinas by calculating the mean density. An estimate of the number of neurons in each retina was determined by multiplying the mean density by the area of the retina. The number of conventional and displaced ChAT-IR amacrine cells decreased by 12% and 8% (P < 0.05 and P < 0.0005, respectively; Table 3 ). When both types of ChAT-IR amacrine cells were combined, the total number in the diabetic mice was 20% less than in the nondiabetic littermates (P < 0.05). 

The number of dopaminergic neurons was also analyzed in the same retinas (Table 3) . The TH-IR cell bodies and dendrites appeared normal in the retinas of the Ins2^Akita diabetic mice compared with nondiabetic littermates. The dopaminergic amacrine cells (type I) were distinguished from the nondopaminergic (type II) amacrine cells by their larger diameter, oblong shape, and the presence of a primary dendrite emerging from the soma, as previously described. ²¹ The number of dopaminergic amacrine cells in the diabetic Ins2^Akita mice was 16% less than in nondiabetic age-matched littermates (457 ± 34 compared with 546 ± 12; P < 0.05). There was no significant difference in the number of nondopaminergic TH-IR cells in diabetic Ins2^Akita mice compared with the littermate control. 

Distribution of Cholinergic Amacrine Cells

Cholinergic amacrine cells are normally distributed in an ordered spatial array throughout the retina. Changes in the array were measured using three spatial indices of mosaic regularity: NN, CR, and effective radius. ¹⁵ The mean CR was 5% more for both the displaced (P < 0.001) and conventional ChAT-IR cells (P < 0.05) in Ins2^Akita-diabetic mice compared with nondiabetic littermates (Fig. 4) . This result suggests that cholinergic amacrine cells are lost in a random pattern, thus producing an increase in the mean distance between cells. 

Further analysis nested the data by eccentricity, to detect any changes in the ChAT-IR cell mosaics (Table 4)in the central and peripheral retina. This analysis revealed the percentage change for each of the three measured indices at different eccentricities (only the comparisons with significant difference P < 0.05 are shown). The CR and nearest NN of the displaced ChAT-IR cells were significantly changed in the central retina (regions 1 and 2), with the largest difference in CR occurring in region 2, approximately 400 μm from the optic disc. There was no significant difference in the effective radius of displaced ChAT cells in any region. In contrast, the greatest changes in the CR, NN, and effective radius of conventional ChAT cells were in the peripheral retina (regions 4 and 5). 

Discussion

This study shows that neuropathy occurs in the retina and affects specific neurons, including dopaminergic and cholinergic amacrine cells. The increased apoptosis probably occurs via a pathway involving active caspase-3. The results confirmed that the rate of retinal cell apoptosis is elevated by diabetes, similar to findings in other published reports, ^{3 6} and that most of the cells affected were nonvascular. Apoptosis quantified by TUNEL and immunoreactivity for active caspase-3 in whole retinas was elevated as early as 2 weeks after the onset of diabetes, which is earlier than previously reported, suggesting that cell loss may be a direct result of diabetic physiology rather than a later consequence of widespread vascular disease. ²²  

This study identified apoptosis in specific types of neurons soon after the onset of diabetes. Previously, apoptotic cells were quantified by TUNEL in whole retina, which was not compatible with immunofluorescence, because the extensive fixation with alcohol and xylene destroyed protein antigenicity and caused excessive autofluorescence. In this study, apoptotic cells were identified by immunofluorescence with an antibody (CM-1) that recognizes the active form of caspase-3. This provided the opportunity to double label retinas with the blood vessel marker agrin. No apoptotic cells were noted within blood vessels, and therefore the caspase-3-positive cells were unlikely to be endothelial cells or pericytes. ²³ These data suggest that at least during the early weeks of diabetes, apoptosis of vascular endothelial cells is less abundant than that of neural cells. This is in agreement with other studies quantifying TUNEL labeling in trypsin digest retinas from diabetic rats, which show that vascular cell apoptosis is elevated by diabetes in only a small number of vascular cells. ^{24 25} Other cells with positive immunoreactivity for caspase-3 colocalized with neuronal markers. Approximately 6% to 8% of caspase-3-positive cells were also positive for NeuN, whereas between 2% and 6% of cells were positive for TH. One consideration in interpreting these data is that most active caspase-3-positive cells did not localize with the cell-specific markers, possibly because most proteins, including cell-specific antigens, are cleaved by the time active caspase-3 reaches a detectable level in cells undergoing apoptosis, rendering the cell impossible to identify by immunohistochemistry. 

The inner retina contains several types of amacrine cells that may be damaged in diabetes. In this study, we report for the first time that TH-IR amacrine cells undergo apoptosis in STZ-diabetic rats; however, at this point in degeneration, the identification of the cells as either type I or II TH-IR amacrine cells is not possible. Although, based on the proximity of caspase-positive cells to the large type I dopaminergic cells, they are presumed to be dopaminergic. Based on the steady rate of apoptosis in the retinas of diabetic animals, ^{3 6} a significant decline in specific populations of neurons after prolonged diabetes was expected. When we used immunohistochemical quantification methods, we detected a significant loss in the total number of dopaminergic and cholinergic amacrine cells. A loss of dopaminergic amacrine cells was reported in diabetic rats. ¹² Together with the loss of TH activity ^{26 27} in the retinas of diabetic rats, these data suggest that dopaminergic neurotransmission is compromised by diabetes. There is also evidence that cholinergic neurotransmission is altered. ^{14 28} Other cell types, such as nitric oxide-containing amacrine cells are also lost in diabetes. ^{29 30}  

The loss of amacrine cells may play a role in the changes in the oscillatory potentials of the electroretinogram in animals and humans with diabetes. The oscillatory potentials, one component of the electroretinogram, are probably due to inner retinal neurotransmission. ³¹ In diabetes, the oscillatory potentials have prolonged peak latencies ³² and decreased amplitudes. ³³ The exact cause of these deficits is not known, but one possibility is amacrine cell dysfunction. The function of ganglion cells is compromised when there is a loss of dopaminergic ³⁴ or cholinergic ^{35 36} signaling in the retina. Loss of dopaminergic and cholinergic neurons may cause changes to visual processing that play a role in the vision loss associated with diabetes. 

It is likely that amacrine cells are not the only neurons lost in diabetes. Neurons in the GCL are also susceptible to apoptosis ^{3 4 5} and cell loss. ^{6 7} Caspase-3-positive cells in the GCL may either be displaced amacrine cells or retinal ganglion cells. Together with the nerve fiber loss in human subjects with diabetes ³⁷ and ganglion cell axon loss in optic nerve of diabetic rats, ³⁸ these data suggest that retinal ganglion cells also degenerate in diabetes. 

Spatial analysis of the cholinergic cell mosaics in the GCL and INL in this study revealed an interesting difference. In general, the highest density of both displaced and conventional cholinergic cells is found approximately 400 μm from the optic disc and becomes progressively less dense toward the peripheral retina. In the central retina, the spacing among displaced cholinergic cells was larger, indicated by increased CR and NN measures. In contrast, the conventional cholinergic cells had wider spacing in the peripheral retina, indicated by increased CR, NN, and effective radius measurements. These data suggest that diabetes leads to a greater reduction of cholinergic amacrine cell density in the peripheral retina than in the central regions. It is interesting to note that in Ins2^Akita mice a significant decrease in the thickness of the INL was found in the peripheral retina, but not in the central retina. ⁶ The results from our study also suggest that conventional amacrine cells in the peripheral rat retina may be more susceptible to apoptosis. However, the displaced cholinergic cells in the GCL may be more affected in central retina. 

Some apoptotic cells were identified in the ONL of STZ-diabetic rats, suggesting that apoptosis of photoreceptors is another component of diabetic neuropathy. A substantial loss of photoreceptors was previously shown in STZ-diabetic rats, ⁴ but such a large magnitude of degeneration was not noted in this study. The discrepancy may be due to the more severe level of hyperglycemia in the previous study. In patients with diabetes, color vision deficits often develop before vascular retinopathy and can be used as a potential indicator of functional changes in photoreceptors and retinal neurons. ³⁹ Humans with diabetes often have reduced blue-yellow contrast sensitivity, known as tritanopia. ^{40 41} This phenomenon is most likely due to a loss of function in blue cone photoreceptors in the fovea of diabetic patients. ⁴² Of interest, a similar type of deficit is apparent in patients with Parkinson’s disease, suggesting that it may be due to dopamine-deficiency. ^{34 43} It is intriguing that some other visual deficits and retinal abnormalities associated with diabetic retinopathy have been identified in Parkinson’s disease. In both diseases, patients exhibit a thinning of the nerve fiber layer ^{37 44} and an increase in the latency of the pupillary light reflex. ^{45 46} A decrease in retinal dopamine level in Parkinson’s disease ⁴⁷ is similar to the loss of tyrosine hydroxylase activity in experimental diabetes. 

In conclusion, diabetes increases the amount of apoptosis in retinas of rats and mice within 2 weeks after the onset of hyperglycemia. The mechanism of apoptosis includes activation of caspase-3. Most of these cells are nonvascular and data presented herein identify some of the affected cells as dopaminergic and cholinergic amacrine cells. Although small, this gradual loss of neurons may compound over an extended period, leading to a chronic neurodegeneration that could give rise to the vision deficits in diabetes. 

 Supported by The Juvenile Diabetes Research Foundation (JDRF) Diabetic Retinopathy Center, JDRF Career Development Award 2-2001-91, American Diabetes Association Grant 1-05-RA-09, and the Pennsylvania Lions Sight Conservation and Eye Research Foundation.

 Submitted for publication October 20, 2005; revised February 16, 2006; accepted May 8, 2006.

 Disclosure: M.J. Gastinger, None; R.S.J. Singh, 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	Study	Animal/Diabetes	Duration (wk)		n	Weight (g)	Blood Glucose (mg/dL)
1	TUNEL	Rat/STZ	2	Control	12	427.0 ± 11.95	79.3 ± 4.63
				Diabetic	11	318.2 ± 8.11*	354.5 ± 18.48*
2	TUNEL	Mouse/Ins2Akita	12	Control	16	30.7 ± 0.98	154.0 ± 12.18
				Diabetic	17	25.1 ± 0.58*	474.2 ± 16.6*
3	Caspase-3-IR	Rat/STZ	2	Control	9	408.9 ± 11.25	87.7 ± 2.12
				Diabetic	6	306.7 ± 9.60*	481.3 ± 27.71*
4	Caspase-3-IR	Rat/STZ	8	Control	4	463.0 ± 20.0	88.0 ± 12.0
				Diabetic	4	304.0 ± 11.0*	372.5 ± 2.5*
5	Caspase-3-IR	Rat/STZ	16	Control	4	584.3 ± 56.66	81.0 ± 2.12
				Diabetic	4	419.8 ± 51.5*	319.3 ± 72.64*
6	ChAT, TH	Mouse/Ins2Akita	24	Control	6	38.87 ± 3.24	202.17 ± 11.10
				Diabetic	6	24.20 ± 0.95*	494.83 ± 21.38*

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

Table 2.

 

Apoptosis in Diabetic Retinas

Animal/Diabetes	Duration (wk)	Density (cells/0.5 cm2)			Change (%)
				Control		Diabetic
TUNEL labeling
mouse/Ins2Akita	12	25 ± 6	43 ± 6*	72
rat/STZ	2	29 ± 4	72 ± 9, **	144
Caspase-3 labeling
rat/STZ	2	34 ± 6	120 ± 15, **	256
rat/STZ	8	22 ± 4	81 ± 17*	279
rat/STZ	16	29 ± 17	80 ± 9*	175

*   P < 0.05.

**   P < 0.0002, two-tailed t-test compared with the control.

Figure 1.

 

Cells containing immunoreactivity for active caspase-3 are nonvascular. Retinas from STZ-diabetic and control rats were labeled for active caspase-3 (red) and the vascular basement membrane glycoprotein agrin (green) and imaged as wholemounts by confocal microscopy. The image shows a typical CM-1-IR cell spatially distinct from the retinal vasculature. Scale bar, 50 μm.

Figure 2.

 

Active caspase-3 immunoreactivity was localized within neurons in retinas of STZ-diabetic rats. (A, D, G) Apoptotic cells in whole retinas from 1-month STZ-diabetic rats were labeled with an antibody to active caspase-3 (CM-1, red). A CM-1-IR cell (A) and NeuN-IR cells (B, green) were found in the INL. (C) Merged image of (A) and (B) shows that CM-1 IR was colocalized with NeuN. Another CM-1-IR cell (arrow) was found in the proximal part of the INL (D), where TH-IR amacrine cells (E, green) were located. The merged image shows that the CM-1 labeling was colocalized with a TH-IR cell (F). A CM-1-IR cell (G, arrow) was located in the ONL indicated by the Hoechst-labeled nuclei (H, blue). The merged image shows the localization of a CM-1-IR nucleus in the ONL (I). Scale bars: (A–C) 8 μm; (D–F) 40 μm; (G–I) 8 μm.

Figure 3.

 

Typical immunoreactivity for ChAT and TH for amacrine cells in the mouse retina. Whole mouse retinas were labeled with an antibody against ChAT (A, B), imaged by confocal microscopy, and used for quantitative analysis. Each image is a maximum projection of several optical sections. (A) Displaced ChAT-IR amacrine cells were identified in the GCL. (B) Conventional ChAT-IR amacrine cells were also located in the INL. (C) Vertical section of a retina with ChAT-IR showing the formation of the two bands of dendrites in the IPL. (D) An antibody against TH labeled both dopaminergic (arrows) and nondopaminergic amacrine cells (*). Scale bars: (A, B, D) 50 μm, (C) 20 μm.

Table 3.

 

Retinal Neuron Loss after 6 Months of Diabetes

	Wild-Type	Diabetic	% Loss
Conventional ChAT	28812 ± 1256	25356 ± 466*	12
Displaced ChAT	23826 ± 391	21891 ± 326, **	8.1
Type I TH	546 ± 12	457 ± 34*	16.3
Type II TH	366 ± 19	403 ± 29	No change

 Data are the mean number of cells (± SEM).

*   P < 0.05.

**   P < 0.005, one-way ANOVA.

Figure 4.

 

Diabetes altered the spatial organization of ChAT-IR cells in the Ins2^Akita mouse retina. The CR is defined as the mean NN distance divided by the mean standard deviation. In the diabetic Ins2^Akita mice, the CR increased by 5% for the population of displaced and conventional ChAT-IR cells. *P < 0.05; **P < 0.001.

Table 4.

 

Changes in Spatial Indices of Cholinergic Cells in Diabetes

	Region
		1	2	3	4	5
Displaced ChAT cells
Conformity ratio	+6.2%	+13.1%	—	—	—
Nearest neighbor	—	+4.8%	—	—	—
Effective radius	—	—	—	—	—
Conventional ChAT cells
Conformity ratio	+5.7%	—	+6.4%	—	+8.8%
Nearest neighbor	—	—	—	+5.8%	—
Effective radius	—	—	—	+10.3%	—

The authors thank Rhona Ellis for assistance with immunofluorescence and confocal microscopy and Wendy Holtry and Neelam Desai for maintaining animal facilities.