Degeneration of retinal ganglion cells in diabetic dogs and mice: Relationship to glycemic control and retinal capillary degeneration

Animals

Experimental animals were treated in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research.

Dogs

Young adult dogs (1.5–2.5 years) were randomly assigned to be made diabetic with alloxan or to remain as untreated control animals. Diabetes was induced in fasted dogs by an intravenous injection of alloxan monohydrate (50–60 mg/kg), which were randomly assigned to one of three prospectively identified diabetic groups (poor glycemic control, moderate glycemic control, good glycemic control) or to remain as nondiabetic controls. Animals in poor control were given only enough insulin to minimize bodyweight loss but not enough to prevent chronic hyperglycemia, whereas animals assigned to good control were given enough insulin (twice per day) to normalize blood glucose. Levels of hyperglycemia in the moderate glycemic control group intentionally were kept between poor and good glycemic control. These animals were part of studies on diabetic retinopathy previously published in collaboration with R. L. Engerman [15,30,31]. Methods by which different levels of glycemia were achieved and maintained were published in those prior reports, but involved food and insulin (dogs in the good glycemic control group got only the amount of food that nondiabetic dogs ate plus adequate insulin to normalize blood sugar; dogs in the poor glycemic control group got ad libitum food and subnormal amounts of insulin; dogs in the moderate glycemic control group got ad libitum food plus the same amount of insulin that dogs in good glycemic control got). Hemoglobin A₁c is summarized at the long-term average of glycemic control. Vascular pathology in these animals (reported previously) is reproduced here for comparison with the new data on retinal neurodegeneration. The present report includes four dogs per experimental group, and all dogs were killed (anesthetized with 1 ml/kg body weight intravenous Nembutal followed by removal of the heart) after five years of study.

Mice

Two rodent models of diabetes were studied (streptozotocin-induced diabetes and spontaneous diabetes in the Ins2Akita mouse). Diabetes was induced in 8-week-old male C57BL/6J mice (strain number 000,664; Jackson Laboratory, Bar Harbor, ME; n=5) with five serial, intraperitoneal injections of streptozotocin (STZ; 55 mg per kg bodyweight, dissolved in citrate buffer). One to two weeks later, diabetes was confirmed by three fasting blood glucose measurements of >275 mg/dL. For the Ins2Akita mice (C57BL/6-Ins2Akita/J (strain number 003,548); Jackson Laboratory), diabetes developed spontaneously in the male mice (n=11) due to a mutated insulin 2 gene (JAX Communication, No. 5, Mar 9, 2000). The STZ model pertains to type 1 diabetes, whereas the Ins2Akita model shows insulin resistance characteristic of type 2 diabetes [32]. Bodyweight was monitored twice per week, and mice were administered insulin if any weight loss was detected. Glycated hemoglobin (GHb) and blood glucose were measured regularly throughout the duration of study. All animals were fed Harlan TekLad diet 7004, except a group of diabetics and their controls that were fed a diet (# 8604) that had been used by other investigators [33] who detected substantial RGC death in diabetic C57Bl/6J mice. Non-diabetic C57BL/6J mice (n=9) served as controls. Some animals were killed (anesthetized with 1 ml/kg body weight intravenous Nembutal followed by removal of the heart) after 10±1 months of diabetes, and others (for diet studies) after 2 months of diabetes.

Optic nerve isolation and preparation

Once removed, the canine and mouse optic nerves were fixed overnight in an isotonic solution of 20% paraformaldehyde and 2.5% glutaraldehyde. After rinsing, processing continued using 3% potassium ferrocyanide and 2% aqueous osmium tetroxide solution for 2 h. Following dehydration in a graded series of alcohols and propylene oxide, nerves were embedded in EPON. One-micron-thick sections were then cut on an ultramicrotome Ultracut E microtome (Reichert-Jung, Depew, NY) about 4 mm in back of the globe, and stained with 1% phenylenediamine in methanol solution to stain the myelin sheathes of the nerves, allowing for their subsequent visualization. Thin sections were generated for electron microscopic analysis.

Optical microscopy and image analysis

Slides were viewed on an Olympus BX-60 microscope (Olympus, Tokyo, Japan; using 40–100×NA 0.75 objectives and DIC optics to increase contrast). Calibrated images were collected using a SPOT RT Slider camera (Diagnostic Instruments, Sterling Heights, MI). Approximately 100 images were needed to cover the area of an entire dog optic nerve. Individual images were stitched together in Adobe Photoshop using the automerge function to produce a composite image of the entire optic nerve cross section. This composite image was then analyzed using MetaMorph imaging software (Molecular Devices, Downington, PA). First, the area of the entire optic nerve was calculated. Because of the large number of axons in the dog optic nerve, sampling was used, and 35%–45% of the total number of axons in each nerve were systematically collected and quantitated. For mice, all axons were counted. Axon diameters and density were determined using software-enhanced analysis of the phenylenediamine-stained nerves. The software identified axons, calculated axon areas, and counted the number and density of the axons.

Electron microscopy and image analysis

Electron micrographs were captured on a JEOL-JEM 1200 TEM (JEOL, Tokyo, Japan) EX transmission electron microscope at 1200×. The resulting negatives were scanned on a Techtronic Flatbed scanner to digitize them. Once digitized, the images were treated similarly to the optical images above.

Terminal transferase deoxyuridine triphosphate nick-end labeling in vivo

Retinal cross-sections (fixed in 10% buffered formalin, embedded in paraffin, and sectioned) were stained for the terminal transferase deoxyuridine triphosphate nick-end labeling (TUNEL) reaction (In Situ Cell Death Detection kit, fluorescein; Roche, Mannheim, Germany). For each assay, one section was treated with DNase (50 U/100 µl) for 10 min to fragment DNA as a positive control. The number of TUNEL-positive nuclei was counted per retinal section. The number of TUNEL-positive cells in the diabetic groups was compared to that in the age-matched nondiabetic controls. The TUNEL assay was performed on two different occasions (each yielded similar conclusions), and at the posterior retina and mid-retina.

Annexin-based cell death detection

This technique is based on a method reported previously [34]. Anesthetized mice were injected intravitreally with 1 µl of annexin V conjugated to Alexa fluor 488 in both eyes. One hour after the injection, the mice were euthanized as described above, and eyes were collected for analysis. For a positive control, ocular ischemia (1 h) followed by reperfusion (24 h) was performed to induce apoptosis in the retinal ganglion cells [35]. Retinas were imaged in three-dimensional plots generated using spinning disk confocal microscopy.

Capillary degeneration

The retinal vasculature from mice diabetic for 10 months was isolated from formalin-fixed tissue using the trypsin digest technique [27,29,36]. Acellular capillaries were quantitated in four to seven field areas in the mid-retina (200X magnification) in a masked manner. Acellular capillaries were identified as capillary-sized vessel tubes having no nuclei anywhere along their length, and were reported per square millimeter of the retinal area.

Statistical analysis

All results are expressed as the mean±standard deviation (SD). Groups were compared by analysis of variance (ANOVA) followed by Fisher’s post-hoc test. Differences were considered statistically significant at p<0.05.

Dogs

As reported previously, diabetic dogs were maintained in prospectively assigned levels of glycemic control for 5 years. Capillary histopathology (microaneurysms, capillary degeneration, pericyte loss) was proportional to glycemic control, with animals in poor glycemic control showing appreciably more pathology that the other groups. In contrast, good glycemic control had a clear beneficial effect on the severity of the diabetes-induced retinal vascular disease. Table 1 summarizes the previously published clinical data and parameters of the vascular disease (pericytes ghosts and microaneurysms) in these dogs, and Figure 1 (left) summarizes the capillary degeneration in the same animals. These previously published data on glycemia and vascular pathology have been pooled from several publications [15,30,31], reanalyzed together, and are presented here to allow easy comparison with the data on neurodegeneration.

The effect of glycemic control on diabetes-induced degeneration of RGCs has not been examined previously. The major parameter examined in this study was mean axon density. Five years of poor glycemic control resulted in a significant loss of axon in the optic nerve (and thus presumably also retinal ganglion cells; p<0.01 compared to nondiabetic) in the diabetic dogs (Table 2 right and Figure 2). The data indicates that 5 years of poor glycemic control resulted in loss of about 40% of all RGCs compared to the age-matched nondiabetic dogs. In contrast, good glycemic control over this 5-year study led to total preservation of the axons (p, not significant compared to nondiabetic animals). Five years of moderate glycemic control, which nevertheless was severe enough to cause a significant increase in vascular pathology (Table 1 and Figure 1), did not cause significant loss of the retinal axons in the optic nerve. Axon density in the moderate glycemic control group was not significantly different from nondiabetic or good glycemic control (p=0.99 and 0.37, respectively), but was significantly different from poor glycemic control (p=0.012). No significant differences were noted in the sizes of individual axons (data not shown).

Mice

Using the axon-counting morphometric technique validated in the dog studies, we attempted to determine if optic nerve axons (and ganglion cells) likewise were lost in diabetic mice. Two strains of mice were studied: C57Bl/6J mice in which diabetes had been induced using the B cell toxin, streptozotocin (STZ), and spontaneously diabetic Ins2Akita mice. Both groups of diabetic animals were comparably hyperglycemic (GHb=11.0%±0.6 and 10.9% ±0.6 for STZ diabetes and Ins2Akita, respectively, compared to 3.4%±0.3 for nondiabetic controls). The duration of diabetes in both models was 10 months. Diabetes of 10-month duration did not result in loss of optic nerve axons (as estimated from axon density) in the STZ model (Table 3 and Figure 3) consistent with our previous studies of diabetic C57Bl/6J mice [26-28]. Likewise, direct counting of axons by electron microscopy did not reveal axon loss in C57Bl/6J mice diabetic for 10 months. The size of individual axons was examined, but no significant differences were noted (not shown).

Since others had reported evidence of diabetes-induced RGC apoptosis in C57Bl/6 mice even at much shorter durations of diabetes [19,23,24], we further assessed RGC death using two additional methods. The first was by injecting fluorescein isothiocyanate–labeled annexin V into the vitreous of nondiabetic and diabetic C57Bl/6J mice [34], and the second was by counting TUNEL-positive cells in retinal cross-sections [26]. Even though both techniques showed that ischemia/reperfusion injury (a positive control) caused RGC death (indicated by the demonstration of numerous positively stained RGCs in mice [not shown]), neither technique demonstrated any differences in the number of apoptotic ganglion cells between long-term diabetic mice and nondiabetic mice. The number of annexin V–positive cells in the RGC layer was 12.2±11.2 and 10.0±5.4 in the diabetic and nondiabetic groups, respectively, and 0.0±0 and 0.1±0.1 TUNEL-positive cells in the RGC layer, respectively (7004 diet). To assess the possible effect of diet on neurodegeneration, we also looked at a different diet with lower levels of antioxidants. Data for mice fed the 8664 diet for the 2-month study likewise showed no increased cell death in the diabetic animals (not shown). This apparent lack of neurodegeneration in both strains of diabetic mice is in contrast to the diabetes-induced significant increase in degenerated capillaries observed in the retinas of the animals (Figure 4).