Pathologic Alterations of the Outer Retina in Streptozotocin-Induced Diabetes

Abstract

Purpose.: Neurodegeneration as an early event of diabetic retinopathy preceding clinically detectable vascular alterations is a widely proven issue today. While there is evidence for the impairment of color vision and contrast sensitivity in early diabetes, suggesting deteriorated photoreceptor function, the underlying neuropathology of these functional alterations is still unknown. The aim of the present study was to investigate the effects of early diabetes on the outer retinal cells.

Methods.: The retinal pigment epithelium, photopigment expression, and density and morphology of photoreceptors were studied using immunocytochemistry in streptozotocin-induced diabetes in two rat strains. The fine structure of photoreceptors and pigment epithelium was also investigated with transmission electron microscopy.

Results.: Here we found that retinal thickness was unchanged in diabetic animals and that no significant increase in the number of apoptotic cells was present. Although the density of cones expressing middle (M)- and shortwave (S)-sensitive opsins was similar in diabetic and control retinas, we detected remarkable morphologic signs of degeneration in the outer segments of diabetic rods, most M-cones, and some S-cones. A decrease in thickness and RPE65 protein immunoreactivity of the pigment epithelium were evident. Furthermore, an increased number of dual cones, coexpressing both M- and S-opsins, was detected at the peripheral retina of diabetic rats.

Conclusions.: Degenerative changes of photoreceptors and pigment epithelium shown here prior to apoptotic loss of photoreceptors may contribute to functional alterations reported in diabetic human patients and different animal models, thus may serve as a potential model for testing the efficacy of neuroprotective agents in diabetes.

Introduction

Diabetic retinopathy, a common complication of diabetes mellitus, is clinically characterized by progressive exudative, ischemic, and proliferative vascular alterations. Current treatments of diabetic retinopathy are established for proliferative lesions and clinically significant macular edema. 

It is becoming clear, however, that early in the course of the disease, preceding clinically observable vasculopathy, retinal microvasculature and neural retina are already compromised. ¹ In morphologic studies, several types of inner retinal elements, including astrocytes, ² amacrine cells, ^3,4 and ganglion cells, ^5,6 were shown to be affected by diabetes. 

Neuroretinal functions are also impaired in early diabetes, even without manifest vascular lesions, as confirmed by many electrophysiologic and clinical studies from the past decades. In human patients, impaired rod and contrast sensitivity ⁷ and abnormal electroretinogram recordings ^8–11 have been reported as features of early diabetes. Acquired color vision impairment may occur in diabetes with a prevalence as high as 30% to 80%, ¹² and diabetic patients even without vascular retinopathy have an increased risk for color vision impairment. ¹³ Both red–green and yellow–blue chromatic sensitivities are reported to be affected, ^7,14,15 showing a strong correlation with care of the disease. ¹⁶  

Also, according to a recent study on Long Evans rats, visual deficits such as impaired visual acuity and contrast sensitivity, as well as scotopic and photopic electroretinogram abnormalities, are already present as early as 4 weeks after streptozotocin (STZ) induction of diabetes. ¹⁷  

These neural retinal dysfunctions detected with clinical and electrophysiologic tests raise the possible involvement of the outer retinal layers in early diabetic neurodegeneration. The underlying pathophysiology may include the apoptotic death of photoreceptors. Apoptosis in diabetes was observed by several groups assessing postmortem human retinas and animal models as well. ^5,6,18,19 In STZ-induced diabetes, with the progression of the disease, a remarkable reduction of the thickness of inner ⁵ and outer ^19,20 retinal layers occurs due to apoptosis 6 months after the induction of diabetes. 

Early neuropathologic changes of photoreceptors preceding apoptosis in diabetes have not gained much attention in the past. The cone matrix sheath was reported to show alterations in length and number as early as 3 weeks after induction of diabetes. ^21,22 However, morphology and photopigment expression of rods and cone subpopulations were not assessed in these studies. 

The aim of the present study was to examine how photoreceptors respond to STZ-induced early diabetic conditions in the rat retina. We describe here for the first time definite pathologic alterations in the morphology of outer segments of rods and cones without detectable loss of photoreceptors, as well as changes in opsin expression with an increased number of dual cones (coexpressing cone opsins), especially at the periphery of the retina. We show evidence that retinal pigment epithelium (RPE) is also affected by diabetes. The alterations reported here may potentially contribute to the functional retinal deficits seen in early diabetes. 

Materials and Methods

Animal Handling and Induction of Diabetes

Based on previous reports on the effects of diabetes on the inner retina, strain dependency of neurodegenerative alterations should be considered. For example, 12 weeks after STZ induction of diabetes the immunoreactivity of parvalbumin, a calcium-binding protein of inner retinal amacrine cells, was reported to increase in Wistar rats, ³ while it was reported to decrease in Sprague-Dawley rats. ⁴ Therefore, in order to minimize the effects caused by the diversity of strains, we chose in this study to apply the same treatment on two different albino rat strains. 

Male Wistar (n = 14) and Sprague-Dawley (n = 24) rats (obtained from Charles River Laboratories Ltd., Isaszeg, Hungary), aged 12 weeks, were used in this study. Diabetes was induced with a single intraperitoneal injection of STZ (Sigma-Aldrich Kft., Budapest, Hungary) at a dose of 70 mg/kg body weight dissolved in 0.1 M sodium citrate buffer, pH 4.5. Control rats were injected with buffer alone. Diabetes was confirmed and monitored by measurements of blood glucose levels using a glucometer (Dcont Personal; 77 Elektronika Ltd., Budapest, Hungary). Rats were considered diabetic and enrolled in the study if fasting blood glucose concentration exceeded 20 mmol/L 1 day after STZ injection. Insulin was not administered to the diabetic rats. 

Animals had unlimited access to standard chow and water and were maintained under a 12-h/12-h alternating light and dark cycle in an air-conditioned room. Diabetic rats were housed alongside age-matched control animals. All protocols involving rats were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the local ethical committee (Semmelweis University, approval no. 22.1/2194/3/2010). 

Tissue Preparation

Diabetic (Wistar n = 7, Sprague-Dawley n = 14) and age-matched control rats (Wistar n = 7, Sprague-Dawley n = 14) were euthanized with an overdose of ketamine 12 weeks after the induction of diabetes. Eyes were oriented and enucleated; then the cornea, lens, and vitreous body were carefully removed. 

For light microscopic analysis, posterior eye cups were fixed in 4% paraformaldehyde (PFA) diluted in 0.1 M phosphate buffer, pH 7.4 (PB), for 2 hours at room temperature. 

For cone density measurements, after repeated rinses in PB, retinas were detached and processed further as whole mounts. For frozen sectioning, posterior eye cups were cryoprotected in 30% sucrose diluted in PB overnight at 4°C and embedded in tissue freezing medium (Shandon Cryomatrix; Thermo Scientific, Inc., Waltham, MA, USA). Tissue blocks were stored at −80°C until further processing. Vertical 10-μm-thick sections were cryosectioned, mounted onto gelatinized glass slides, and stored at −20°C until use. 

For electron microscopy, eye cups were fixed as described later. 

Measurement of Retinal Thickness

To assess the possible changes in the thickness of the retina, we measured the distance between the outer and inner limiting membranes and the thickness of the outer nuclear layer at six locations in at least three sections per retina in Wistar and Sprague-Dawley rats (n = 4 diabetic and control retinas). Vertical cryosections at the level of the optic nerve were analyzed on a Zeiss Axiophot (Carl Zeiss, Oberkochen, Germany) microscope using a 40× lens at given distances from the optic nerve. Measurements were carried out at distances of 250 and 500 μm (central values) as well as 4000 μm (peripheral values) from the optic nerve in superior and inferior directions. 

TUNEL Assay

To evaluate the rate of apoptosis, terminal deoxynucleotidyltransferase-mediated biotinylated UTP nick-end labeling (TUNEL) staining was performed ²³ using an in situ cell death detection kit from Roche (Hungary) Ltd., Budaörs, Hungary. 

Negative control sections were incubated omitting the terminal transferase; positive control sections were preincubated with DNase prior to TUNEL reaction. Retinal sections from young postnatal rats (postnatal day 2) known to contain apoptotic bodies ²⁴ were also used as positive controls. 

In order to quantify the number of TUNEL-positive elements, the total number of stained nuclei was counted on vertical cryosections at the level of the optic nerve from Sprague-Dawley rat retinas (n = 4 diabetic and control retinas, four sections per animal). All positive elements were counted from the optic nerve head to the ora serrata both superiorly and inferiorly. 

Lectin Cytochemistry and Immunohistochemistry

For the selective evaluation of photoreceptor types and subtypes, lectin cytochemistry and immunohistochemistry were performed both on whole-mounted retinas and on cryosections. Lectin from Arachis hypogaea (peanut agglutinin lectin, PNA, 5 μg/mL; Sigma-Aldrich Kft.) was used for cone matrix labeling. Peanut agglutinin lectin binds to the galactosyl (b-1,3) N-acetyl-D-galactosamine disaccharide linkages, thus labeling the interphotoreceptor matrix of all cones. ²⁵ Cone subpopulations were stained selectively using antibodies against the short (S) and middle (M) wavelength-sensitive opsins. 

For staining rods, we applied either lectin from Triticum vulgaris (wheat germ agglutinin lectin, WGA, 1:1000; Sigma-Aldrich Kft.), or a polyclonal antiserum against rhodopsin molecule (AO) produced in our laboratory. ²⁶  

To study the morphology and staining characteristics of the RPE, immunohistochemistry against retinal pigment epithelium-specific protein (RPE65, an enzyme involved in the reisomerization of 11-cis retinal and visual pigment regeneration) was carried out. 

Details of the primary antibodies applied are listed in Table 1. 

Table 1

 

Details of the Primary Antibodies Used

Antibody	Source	Reference	Dilution	Host and Type	Epitope Specificity or Labeling Pattern in Rat Retina
Anti-rhodopsin, AO	Produced in our laboratory	26	1:1000	Rat polyclonal	Primarily rhodopsin, N-terminal
OS-2	Produced in collaboration with the Biological Research Center, Szeged, Hungary	26, 27	1:5000	Mouse monoclonal	S-cone opsin, C-terminal
COS-1	Produced in collaboration with the Biological Research Center	26	1:50	Mouse monoclonal	M-cone opsin, C-terminal
AB5407	Merck Kft., Budapest, Hungary	28	1:1000	Rabbit polyclonal	S-cone opsin
AB5405	Merck Kft.	29	1:1000	Rabbit polyclonal	M-cone opsin
Antibody to RPE65	Merck Kft.	30	1:500	Mouse monoclonal	Isomerohydrolase, retinal pigment epithelium

To reduce nonspecific staining, prior to immunoreactions all sections and whole-mounted retinas were blocked for 2 hours with 1% bovine serum albumin (BSA) diluted in 0.1 M phosphate-buffered saline (PBS) with 0.4% Triton X-100 (Sigma-Aldrich Kft.). All primary and secondary antibodies were diluted in the same solution. Incubation time lasted overnight in the case of sections and for 48 hours in the case of whole-mounted retinas at 4°C. After repeated rinses, the primary antibodies bound were visualized by incubating with the appropriate species-specific fluorescent-conjugated secondary antibodies (Alexa Fluor 488, 594 conjugates, 1:200, 2 hours, at room temperature; Life Technologies Magyarország Kft., Budapest, Hungary). Lectins were applied in a biotinylated form and visualized using streptavidin-conjugated Alexa Fluor 594 conjugate (1:200, 2 hours, at room temperature; Life Technologies Magyarország Kft.). Slides and whole-mounted retinas were then rinsed in PBS at room temperature; and in the case of sections, cell nuclei were counterstained with 4,6-diamidino-2-phenylindole (Sigma-Aldrich Kft.) for 10 minutes. Negative immunohistochemistry controls were conducted in parallel by omission of the primary antibodies. 

Imaging and Image Processing

Immunoreactions were visualized and recorded with a Bio-Rad Radiance 2100 Rainbow Confocal Laser Scanning System (Carl Zeiss) coupled to a Nikon Eclipse E800 microscope (Nikon, Tokyo, Japan). The LaserSharp 2000 6.0 software (Carl Zeiss) was used for image acquisition, while the final figures were composed with Adobe Photoshop 7.0 (San Diego, CA, USA). Only minor adjustments with respect to brightness and contrast were made, which in no case altered the appearance of the original material. 

Analysis of the Density and Distribution of Cones

To compare the density and distribution pattern of cone types expressing S- and M-pigment, whole-mounted retinas were immunostained with opsin-specific antibodies. On average, 83.2 ± 4.7 fields of view per retina were analyzed in a systematic manner in diabetic and control retinas from Wistar rats (n = 4 for each group) with a Neurolucida work station consisting of a modified light microscope (BX50; Olympus, Tokyo, Japan), motorized specimen stage for automatic sampling (Märzhäuser, Wetzlar, Germany), charge-coupled device color video camera (HV-C20AMP; Hitachi, Tokyo, Japan), and a PC with a frame grabber board (Flashpoint Intrigue light; Integral Technologies, Indianapolis, IN, USA). Details of the analysis parameters are listed in Table 2. 

Table 2

 

Details of the Analysis for Estimating the Density of S- and M-Cones and Dual Cones

	Central S-Cones	Central M-Cones	Peripheral S-Cones	Peripheral M-Cones	Peripheral Dual Cones
Obj.	20×	40×	20×	20×	20×
A, μm2	62,500	16,900	2500	2500	2500
Dx, μm	630	750	75	75	75
Dy	630	750	150	150	150
ΣUVCF	83	83	330	330	330

 Obj., objective used for the analysis; A, area of the unbiased virtual counting frames; D, distance between the unbiased virtual counting frames in directions x and y; ΣUVCF, average sum of unbiased virtual counting frames used.

S- and M-cone outer segments were counted manually, in 250- × 250-μm squares in the case of S-cones and 130- × 130-μm squares in the case of M-cones. To approximate cone numbers in uncounted regions from measured data and to construct isodensity maps, estimation was created using Adaptive Neuro-Fuzzy Inference System (ANFIS). Detailed information on ANFIS can be found in Supplementary Material S1 and the references. ^31–34  

Our previous observations, ³⁵ in agreement with findings by others, ^36–38 suggest that in healthy control animals a thin S-cone–rich retinal rim with some dual cones (coexpressing both cone opsins) may exist on the extreme peripheral retina (75–100 μm from the ora serrata). Therefore, peripheral retina was separately investigated using a more thorough approach. Peripheral retinas were analyzed all around the retinal circumference in a 200-μm-wide ring using 50- × 50-μm counting frames, placed evenly in a centripetal manner. The density of peripheral S-cones and dual cones was plotted using circles indicating the actual position of the counting frames. On average, 330 ± 32 sites per retina were analyzed on the periphery. 

To estimate possible changes in the number of cone types in Sprague-Dawley rats, a similar measurement was carried out in the superior-nasal quadrants of three diabetic and three control specimens. 

Electron Microscopy

In Wistar rats, after PFA fixation, 1- × 1-mm pieces of the posterior eye cups were cut, superior to the optic nerve, and were postfixed with 1.5% glutaraldehyde in 0.1 M PB (90 minutes, room temperature). In Sprague-Dawley rats, after euthanasia and enucleation, posterior eye cups were fixed with a solution containing 2.4% PFA and 1.65% glutaraldehyde diluted in 0.1 M PB (30 minutes, 37°C). Four diabetic and four control animals were used from both strains. After fixation, samples were repeatedly rinsed with 0.1 M cacodylate buffer (CB) at room temperature, and further postfixation was carried out with 1% osmium tetroxide in 0.1 M CB for 1 hour at 4°C. After repeated rinsing, retinas were dehydrated in graded alcohol series and embedded in araldite (Durcupan ACM Fluka; Sigma-Aldrich Kft.). Semithin and ultrathin sections were prepared on a Reichert-Jung UltraCut E ultramicrotome (Reichert, Inc., Buffalo, NY, USA) equipped with a diamond knife. Semithin sections stained with toluidine blue were used for primary evaluation. Ultrathin sections were contrasted with uranyl acetate and lead citrate and examined using a transmission electron microscope (Hitachi H 7500). The fine structure of the photoreceptor outer segments and pigment epithelium was photographed and the thickness of the pigment epithelial layer measured on electron micrographs derived from identical locations of diabetic and control retinas (n = 4 specimens/group, three measurements per section). 

Statistical Analysis

Data are expressed as mean ± standard deviation (SD). Blood glucose, body weight, and thickness of the retina and RPE cells, as well as the number of TUNEL-positive elements, were analyzed using paired and unpaired Student's t-test. In all cases, a P value less than 0.05 was considered statistically significant. 

Results

General Clinical Features: The Two Strains Reacted Similarly to Diabetic Conditions

Streptozotocin-induced diabetic rats had significantly lower body weight on day 7 (1 week after the induction of diabetes) compared to vehicle-treated controls (Table 3). Under diabetic conditions average body weight was reduced significantly by week 12 in both strains, while control rats were gaining weight gradually. Blood glucose levels were significantly elevated from day 1 throughout the observation period of 12 weeks compared to those in animals in the vehicle-treated groups. 

Table 3

 

Body Weight and Serum Glucose Levels in Diabetic Rats Compared to Normal Values (Day 0 = Day of Diabetes Induction)

Group	n	Body Weight ± SD, g			Blood Glucose ± SD, mmol/L
		Day 0	Day 7	Week 12	Day 0	Day 1	Week 12
Wistar control	7	383.8 ± 12.6	377.6 ± 12.1	457 ± 29.5	6.1 ± 0.3	5.7 ± 0.4	6.3 ± 0.2
Wistar diabetic	7	395.6 ± 9.1	333.3 ± 23.3*	287.4 ± 21.1*	5.8 ± 0.3	>25.0*	>25.0*
Sprague-Dawley control	12	441.7 ± 24.2	458.8 ± 23.4	609.8 ± 40.2	6.0 ± 0.4	6.1 ± 0.3	7.0 ± 0.6
Sprague-Dawley diabetic	12	440.8 ± 18.1	393.8 ± 20.5*	295.8 ± 47.3*	6.15 ± 0.3	21.5 ± 4.3*	>25.0*

 All data are presented as mean ± SD.

*   P < 0.05.

No Significant Cell Loss Was Detected in Early Diabetes

To determine if there is a significant loss of neuronal cells causing detectable atrophy in the retina after 12 weeks of diabetes, the thickness of the retina between the outer and inner limiting membranes was measured in corresponding locations in diabetic and control groups. 

Retinal thickness did not vary significantly at any retinal locations analyzed in either of the two strains. The thickness of the outer nuclear layer alone was also measured and showed no difference between control and diabetic groups (Fig. 1). 

Figure 1

 

The distance between the outer and inner limiting membranes and the thickness of the outer nuclear layer alone were measured at given distances from the optic nerve head (ON) in diabetic and control Wistar (A) and Sprague-Dawley (B) rats from n = 4 specimens/group, using three sections per specimen. No significant difference (P > 0.05) in thickness was detected compared to that in age-matched controls at any retinal location, in either strain, after 12 weeks of diabetes. Data are given in micrometers, expressed as mean ± SD.

In both strains in control as well as in diabetic retinas, sparse TUNEL-positive elements could be seen in the inner and outer nuclear layers and also in the ganglion cell layer. 

For quantitative analysis, TUNEL-positive nuclei were counted on cryosections from Sprague-Dawley rat retinas. Average number of TUNEL-positive elements was 6.5 ± 2.1 vs. 9.3 ± 3.5 in control and diabetic sections, respectively. Approximately one-third of the positive elements (3.0 ± 1.4 in control versus 3.3 ± 2.1 in diabetic sections) were located in the outer nuclear layer (Fig. 2). The difference between control and diabetic retinas was not significant. Our data on unchanged retinal thickness and sparsely distributed apoptotic cells demonstrate consistently that overt retinal atrophy is not present in this model of early diabetes. 

Figure 2

 

Representative photomicrographs illustrating TUNEL labeling in control (A) and diabetic (B) Sprague-Dawley rat retinas. TUNEL-positive elements were detectable only occasionally (arrow in [B]). (C) The number of TUNEL-positive elements counted on complete vertical cryosections through the optic nerve in control and diabetic Sprague-Dawley rats (n = 4 diabetic and control retinas, four sections per animal analyzed). No statistically significant difference was detectable, whether counting was in all retinal layers or the outer nuclear layer alone (P = 0.34, P = 0.84, respectively). Results are expressed as mean ± SD. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar: 30 μm.

Photoreceptor Cells in Control and Diabetic Rats

Generally, photoreceptors of Wistar and Sprague-Dawley rats showed similar changes under diabetic conditions. However, we detected differences in the severity of photoreceptor degeneration between the two strains. 

Cones.

Double-labeling fluorescent immunohistochemistry against S- and M-opsins was carried out on retinal whole mounts to examine and compare the density and distribution pattern of cone types in diabetic and age-matched control animals. Isodensity maps of Wistar rat retinas are shown in Figure 3. In control retinas, a typical centroperipheral gradient was present for M-cones, with central peak densities reaching approximately 5000 cones/mm². Density values dropped significantly toward the periphery. For S-cones there was a detectable difference between the superior and the inferior retinal halves, with average values of 100 to 200 cones/mm² superiorly and 800 to 900 cones/mm² inferiorly. We detected a sharp increase in the density of S-cones at the peripheral rim of the retina, reaching extremely high numbers (>1200 cones/mm²) in the superior retinal regions. The density values shown here are in close agreement with those reported by others. ^38–40 In line with the literature, great interindividual variability was detected even between healthy controls. Comparing the results with those of diabetic retinas, similar density values and distribution patterns were revealed in diabetic animals, and no changes for either cone type were evident (Fig. 3). Quantitatively, the average number of cones counted in a single retina with 83 counting frames did not show any statistically significant difference. A total of 5621 ± 754 vs. 5467 ± 660 M-cones and 3048 ± 776 vs. 2725 ± 259 S-cones were counted in control and diabetic retinas, respectively. However, an interesting feature was noted. A significant population of cones coexpressed both opsins in the diabetic retinas, reaching especially high local densities in peripheral regions (dual cones, Figs. 3, 4). Such dual cones were rarely seen in adult control specimens. 

Figure 3

 

Representative isodensity maps of the distribution of M-opsin (A, B) and S-opsin (C, D) expression as well as dual (E, F) cones in control (A, C, E) and diabetic (B, D, F) retinas of Wistar rats. Four specimens were analyzed from each group. Density values are represented by color codes. At the S-cone–rich rim of the retinas, counting frames were placed more frequently. These peripheral regions are indicated by light gray shading on the maps in (C, D). In these regions, counting was performed in 50- × 50-μm frames placed evenly around the circumference. Density values of the peripheral S-cones are represented by circles indicating the actual position of the counting frames. In the case of dual cones, counting was performed at identical positions, but only the regions containing at least one dual element are represented. For better interpretation of the images, the peripheral regions were enlarged (1.6×). Note that no major difference could be observed in the distribution patterns of S- and M-cones in diabetes compared to age-matched controls either centrally or in the far periphery. However, the number of dual cones coexpressing both S- and M-opsins is explicitly elevated in diabetes (F) compared to healthy controls (E). S, superior; T, temporal; I, inferior; N, nasal. Scale bar: 1 mm.

Figure 4

 

Representative whole-mounted retina samples (from n = 4 specimens analyzed per group) demonstrating morphologic alterations of M-opsin–expressing cones in diabetes in all retinal positions, and an increase in the number of dual-labeled cones at the periphery. S-cones (OS-2 in red) and M-cones (AB5405 in green) were stained in control (A–C) and diabetic (D–F) retinas of Wistar rats. Samples were taken from distinct locations: central (close to the optic nerve head [A, D]), midperipheral (approximately halfway between the optic nerve head and the superior ora serrata [B, E]), and far peripheral (superior retina, immediately at the ora serrata [C, F]) retinal regions. Note the highly deformed cone outer segments in almost all M-cones and the increased number of double-labeled cones in the retinal periphery in diabetic rats compared to controls. Some of the dual elements are indicated by arrows. Scale bar: 50 μm.

Similarly, no major difference was detected regarding the S- and M-cone densities in the superior-nasal quadrant of Sprague-Dawley rat retinas analyzed, but an increase in the number of dual cones was highly visible (not shown). 

When the morphology of cones was studied, a remarkable difference was found between diabetic and control retinas, especially regarding the outer segments. A vast majority of M-cones showed clear signs of outer segment degeneration. The outer segments seemed to be fragmented, but thorough investigation showed that all the pieces were connected with a thin stalk (Figs. 4D–F, 5B, 5C), suggesting that they belonged to the same deformed outer segment. This particular feature was confirmed using both M-opsin–specific antibodies in both rat strains. Interestingly, in diabetic rats, the PNA-labeled cone matrix sheath was preserved in its original form and did not seem to follow deformations of the outer segments (Figs. 5A–C). 

Figure 5

 

Representative photomicrographs of vertical retinal sections of n = 8 animals per group, showing altered morphology and staining characteristics of photoreceptor cells. Control Wistar retinas showed normal labeling of M-cones (AB5405 in green) along with the surrounding interphotoreceptor matrix labeled with PNA lectin (in red) (A). In diabetic rats, abnormal morphology of M-opsin–expressing cone outer segments is seen in both Wistar (B) and Sprague-Dawley rats (C). The broadened endings are linked by a narrow stalk-like connection (arrowheads). However, the interphotoreceptor matrix around cone outer segments shows normal morphology. In diabetic Sprague-Dawley rats, PNA also labeled rod outer segments (C), indicating a possible defect in protein glycosylation in rods. In agreement with this, WGA (wheat germ agglutinin lectin in red) staining of rods is drastically reduced in diabetic Sprague-Dawley rats (E) compared to controls (D). Some S-cones (OS-2 in red) also show signs of degeneration, characterized by the staining of the complete cell membrane with S-opsin–specific antibodies and a missing or unstained outer segment (F). S-cone abnormality was detected only in Wistar rats. Scale bars: 20 μm (A–E); 5 μm (F).

We also detected morphologic changes in a small population of S-cones. These cells had a reduced or invisible outer segment, while the complete cell body was labeled with S-opsin–specific antibodies (Fig. 5F). While these cells appeared regularly in Wistar rats (one or two cell bodies per section), interestingly, none were detected in retinas of Sprague-Dawley rats. 

Rods.

In diabetic and control groups, rhodopsin immunoreactivity was present in the outer segment of rods and there was no major change in their staining intensity. However, the appearance of the rhodopsin-positive outer segments was characteristically blurred in diabetes (Figs. 6B, 6D) in contrast to the fine rod-like appearance seen in control retinas in both strains (Figs. 6A, 6C). This change was more prominent in Sprague-Dawley rats (Fig. 6D), where the layer of outer segments seemed to be almost completely disorganized, indicating significant outer segment degeneration. 

Figure 6

 

Decreased immunoreactivity of RPE65 (in red) in the retinal pigment epithelium is accompanied by morphologic degeneration of rod outer segments in diabetes. Rod outer segments were labeled with a polyclonal anti-rhodopsin antibody (AO, in green). (A, C) Control; (B, D) diabetic retinas. Upper row shows retinas from Wistar, lower row from Sprague-Dawley rats. Analysis included n = 8 animals per group. Note that in diabetes, the immunoreactivity of RPE65 is substantially weaker in Wistar rats (B) compared to control (A), while the change is less pronounced in Sprague-Dawley rats (control [C], diabetes [D]). However, in both strains, diabetic degeneration of rod outer segments is evident, manifested by the characteristically blurred appearance of the outer segments. Scale bar: 30 μm.

Lectin staining of photoreceptor cell types revealed another interesting phenomenon. In Sprague-Dawley rats, where rod outer segment degeneration was more prominent, PNA labeling did not only and selectively bind to cone cells, but also showed a discrete but visible staining of rods (Fig. 5C). Parallel to the appearance of PNA labeling in rods, WGA labeling was diminished in these cells (Fig. 5E) when compared to controls (Fig. 5D). 

Electron microscopic investigation confirmed a remarkable disorganization of the outer and inner segments of photoreceptor cells in the diabetic groups. While retinas from the control group showed the well-known pattern of a normal retina with a clear separation of the outer and inner segment layers, densely packed photoreceptors of uniform thickness, and well-ordered discs in the outer segments (Figs. 7A, 7C), severe pathologic alterations could be observed in the diabetic rats (Figs. 7B, 7D). These were manifested in the swelling of inner segments to such an extent that they were often pressed high between the outer segments, containing rounded and swollen mitochondria. Inner segments also frequently exhibited vacuoles of various sizes. Most outer segments in the diabetic group showed a strongly disorganized inner structure with clear areas from which discs were missing or contained stacks of incomplete and misoriented discs. Presence of mitochondria in the vicinity of rod discs without intervening plasma membrane gave the impression of a possible fusion of outer and inner segments. Severity of these pathologic changes varied from cell to cell in the same retina or between animals in the same experimental group. 

Figure 7

 

Representative electron micrographs from n = 4 animals per group of the photoreceptor layer in control (A, C) and diabetic (B, D) Sprague-Dawley rat retinas. Note the swelling of the inner segments and the disorganization or lack of outer segments in diabetes, with only a few stacks of intact discs present. Detailed description is given in the text. Scale bars: 5 μm (A, B); 1 μm (C, D).

Retinal Pigment Epithelial Cells in Control and Diabetic Rats

As RPE is required for proper photoreceptor functions, morphologic changes of the photoreceptors may be related to possible RPE defects. Retinal pigment epithelial cells were studied using immunocytochemical labeling for RPE65 protein and electron microscopic morphometric analysis. 

At the ultrastructural level, a significant reduction of the thickness of RPE cells could be seen in the diabetic Wistar rats (Figs. 8, 9). In contrast to those in the control group, the RPE cells were thinner, ovoid nuclei could be observed regularly, and the smooth endoplasmic reticulum (sER) content of the cytoplasm was reduced. In agreement with the altered morphology, RPE65 immunoreactivity was also drastically weaker in diabetic retinas when compared to controls (Fig. 6). The remarkable volume decrease of the cytoplasm, the strongly reduced sER, and also the drastically diminished amount of the isomerohydrolase enzyme (RPE65) suggest severely impaired function of the RPE. 

Figure 8

 

Representative electron microscopic images of the RPE cells in control (Wistar [A], Sprague-Dawley [C]) and diabetic rats (Wistar [B], Sprague-Dawley [D]). Four specimens from each group were inspected. In agreement with the immunocytochemical data, the morphologic changes of RPE cells are more obvious in Wistar than in Sprague-Dawley rats. Note the reduction of the thickness, the decreased smooth endoplasmic reticulum content, and the flattened nuclei in diabetes. Arrow indicates a fragment from an outer segment phagocytosed by the RPE cell. N, nucleus; OS, rod outer segment. Scale bar: 2 μm.

Figure 9

 

Average thickness of an RPE cell (means in μm ± SD) in diabetic and control retinas (*P = 0.001). Data were obtained from n = 4 specimens per group, with three measurements per section. In Wistar rats, the thickness of RPE cells decreased significantly in diabetes; however, in Sprague-Dawley rats, no statistically significant difference was found compared to controls (P = 0.262).

All the above-mentioned changes of the pigment epithelium were less pronounced in Sprague-Dawley rats. The reduction of the thickness of RPE cells was not significant, and there was only a moderate decrease in the immunoreactivity of RPE65 (Figs. 8, 9). 

Discussion

In the present study in a rat model of diabetes mellitus, signs of outer segment degeneration and changes in the pattern of opsin expression of retinal photoreceptors were shown prior to apoptotic reduction of retinal thickness. These changes were accompanied by alterations in the morphology of the pigment epithelium as well. Since data may be controversial when different species and strains are used as diabetic models (e.g., as in Refs. 3, 4), in our study experimental diabetes was induced in two different albino rat strains. The two strains studied responded similarly to diabetic conditions, both clinically and histologically. The same cell types were shown to be affected in the outer retinal layers, although variations in the sensitivity of rods, S-cones, and RPE cells were detected. These findings may provide evidence for the development of neurodegeneration as early events of diabetic retinopathy and thus may be part of the underlying neuropathology of early functional deficits reported by others. ¹⁷  

No Significant Apoptotic Cell Loss in Early Diabetes

Since the extensive degeneration of photoreceptors without appreciable apoptotic loss was surprising, three different methods to search for apoptosis in photoreceptor cells were used in our study. First, the thickness of the whole retina and the outer nuclear layer was separately measured, but no change between control and diabetic retina was detectable. Second, the number of different types of cones in whole-mounted retinas or retinal pieces was counted; again, no major change was found in the number of S- or M-/L-cones, except for the increase in the number of dual-labeled elements. We must emphasize that at this stage, almost all M-/L-cones showed prominent signs of outer segment degeneration, yet their number was not changed significantly, as shown by near-identical isodensity maps of cone subtypes. Third, we labeled a number of slides with TUNEL assays and found no statistically significant increase in the number of TUNEL-positive cells. A set of positive and negative control slides was also included to rule out false-positive or -negative results. All these data taken together indicate that the morphologic changes detected in our study most probably preceded apoptosis. This is in agreement with data in the literature indicating that significant apoptosis is to be expected after 6 months of diabetes. ^5,19,20,41 Further studies are needed to elucidate the fate of degenerating photoreceptor cells in later time periods. 

Degenerative Changes in Outer Segment Morphology

Outer segment degeneration of some degree was observed by us in all three types of photoreceptors in diabetic animals; however, no difference was found in the overall density and spatial distribution pattern of S- and M-opsin–expressing cones between control and diabetic retinas. 

Signs of outer segment degeneration were evident in all diabetic specimens studied on both the light and electron microscopic levels, in all retinal positions and in both strains. The changes seen here could not be attributed to poor fixation, as they were not present in any of the control retinas processed using the same protocol. 

Generally, S-cones are supposed to be more vulnerable to diabetes than M-cones. ⁴² In line with this, a significant decrease of S- but not M- and L-cone densities was reported in human patients with diabetic retinopathy. ⁴³ Our data in rats demonstrated changes in the morphology of outer segments in the majority of M-cones, whereas only few S-cones were affected. In most M-cones, deformed, seemingly fragmented outer segments were visible, with the “segments” connected only by a thin stalk. Similar or even more robust changes indicating cone degeneration have already been reported in several other pathologic conditions, for example, age-related macular degeneration ⁴⁴ and RPE65^−/− dogs, ⁴⁵ suggesting that M-cones may respond similarly to all noxious stimuli, possibly including diabetes. In these diseases, it is generally assumed that the photoreceptors affected will eventually die. Therefore, these degenerative changes may be considered a preapoptotic event, but the final fate of M-cones in diabetic rats and also in human patients requires further investigation. Interestingly, in the present study only some S-cones in Wistar but not Sprague-Dawley rats seemed to be affected. Some S-cones appeared to have lost their outer segments completely, with the opsin accumulating throughout the whole cell membrane. Either these cells truly lack outer segments; or the transport mechanism itself, which is responsible for trafficking opsin to the outer segments, became defective due to the disease. Similar accumulation of opsin in the cell body of cones has also been reported in age-related macular degeneration. ⁴⁴  

Previous studies using PNA lectin labeling in diabetic rat retinas reported shortening of the labeled interphotoreceptor matrix of the cones. ^21,22 However, PNA lectin does not label the cone outer segment itself; therefore it does not give decisive information regarding outer segment morphology or length. In our studies, we directly labeled cone outer segments using opsin-specific antibodies. Despite the evident degeneration and incomplete fragmentation detected in the majority of M-cones, shorter or missing outer segments were seen in only a small population of S-cones. 

By electron microscopy we detected morphologic changes also in rods; these were manifested by severe disorganization of the inner and outer segments. Although the majority of rods seemed to be degenerated, the total number of rod cells most probably did not change in our experimental setup. Although no precise counting was performed due to overlapping of the slender rod outer segments, the fact that the thickness of the outer nuclear layer was similar in the two groups and that apoptotic bodies were seldom detected strongly argues against significant loss of rod cells. 

We also observed changes in the intensity of the lectin staining of the rod interphotoreceptor matrix; this was more pronounced in Sprague-Dawley rats, coinciding with the more advanced level of outer segment degeneration. Decrease of the WGA staining intensity in rods along with the parallel appearance of a faint PNA staining may possibly reflect dysfunctions of protein glycosylation in these cells. 

Alteration of the RPE in Diabetes

In diabetic Wistar retinas, a significant involution of the RPE was detected, with a drastically decreased amount of the sER within the cells. Also, staining intensity with RPE65 immunohistochemistry was significantly weaker in diabetic rats. These data raise the possibility that dysfunction of the pigment epithelium may be primarily responsible for the alterations detected in the photoreceptor layer. 

One of the many ways in which RPE influences photoreceptor functions is continuous phagocytosis of the apical portions of outer segments. In certain retinal degenerative diseases this retinal turnover fails; toxic photooxidative products accumulate in the outer segment of photoreceptors, resulting in subsequent photoreceptor degeneration. ^46,47 Phagocytic activity of macrophages and monocytes has been shown to be impaired in diabetic mice ⁴⁸ and humans. ⁴⁹ Furthermore, Simó et al. ⁵⁰ suggest in their review that under diabetic conditions, pigment epithelial phagocytosis might be affected as well. 

Retinal pigment epithelium cells also participate in retinal reisomerization. This process is performed by the enzyme isomerohydrolase associated with the sER. Not surprisingly, parallel with the involution of the RPE cells, the immunoreactivity of RPE65 protein (isomerohydrolase enzyme) was also found to be reduced. The lower expression level of RPE65 might indicate a disturbed visual cycle under diabetic conditions, which might be an underlying mechanism of the degenerative changes observed in outer segments. The importance of RPE65 is further underlined by studies showing retinal degeneration when RPE65 was missing due to genetic failures. ⁴⁵  

Since RPE cells are also known to have a vital trophic function, secreting a variety of growth factors including pigment epithelium-derived factor, ciliary neurotrophic factor, and TGF-β, ⁴⁷ one cannot exclude the possibility that this function of the RPE is compromised in diabetic rats, leading to photoreceptor degeneration. 

Our present findings fail to prove involution of RPE as the primary cause of outer segment pathology in experimentally induced diabetes in rats. Since RPE damage and the degree of outer segment degeneration showed a reverse relationship in the rat strains studied, with involuted RPE and less prominent outer segment alterations in Wistar rats and less affected RPE accompanied with severe outer segment damage in Sprague-Dawley rats, it is possible that RPE defects and outer segment degeneration occur independently. 

Other Possible Causes of Degeneration in the Neural Retina

As an alternative to RPE damage as the primary cause of photoreceptor degeneration, the vasculopathy of the choriocapillary layer nourishing photoreceptor cells may be mentioned. Subclinical vascular alterations, like changes in choroidal blood flow and impaired autoregulation, are present as early as 10 weeks after the induction of diabetes in mice. ⁵¹ Moreover, in human diabetic patients, a decrease in the thickness of the choroid can be detected with optical coherence tomography, ⁵² accompanied by pathologic changes like degenerative capillaries and capillary dropouts. ⁵³ Therefore, it is possible that microvascular alterations of the choroid, rather than RPE defects, may contribute to the photoreceptor degeneration described in this study. 

A further possible cause of photoreceptor degeneration may be the direct effects of hyperglycemia and hypoinsulinemia. Diabetes alters certain elements in the insulin signaling pathway in the photoreceptors, thus impairing this important survival and neuroprotective signal. ^54,55  

Changes in Opsin Expression—Dual Cones

Previous results from our laboratory have shown that in the early postnatal life of rats, M-cones develop via transdifferentiation from originally S-opsin–expressing elements. For a limited time interval, most presumptive M-cones transitionally coexpress both S- and M-opsins ^29,40 (for review of dual cones see Ref. 35). Such dual cones are reported to be present also in adult rats, probably in an age-dependent manner ^35,38 (Lukáts Á, Énzsöly A, Szabó A, Szabó K, Szél Á, unpublished observations, 2012). Furthermore, under certain pathologic conditions, such as hypothyroidism, cones may change their phenotype, ⁵⁶ and dual elements or S-opsin–expressing cones may again become the dominating cone elements of the adult rat retina. 

Here, for the first time we have shown that in diabetic animals, a high number of cones coexpressed both the S- and the M-opsins. Such dual cones were abundant especially in peripheral regions of diabetic retinas with density values exceeding those of the controls. 

What mechanisms may lead to the accumulation of dual cones in the diabetic retina? We offer two hypotheses for explanation. First, disturbed hormonal regulation, for example, thyroid dysfunction, may account for changes in cone opsin expression. Indeed, Glaschke et al. ⁵⁶ reported an increased number of dual cones in hypothyroidism, a hormonal disorder often associated with ⁵⁷ or assumed to be pathophysiologically related to diabetes. ⁵⁸ Although we did not directly measure hormone levels here, it is possible that opsin expression disturbed by diabetes reflects changes in thyroid hormone homeostasis. Secondly, an increase in the number of dual cones may also be explained if one assumes cone regeneration at this stage of diabetes. Since the overall cone density and retinal thickness did not change even after 12 weeks, it is not unreasonable to assume that degenerated cones had been eliminated and replaced by new postmitotic cells that may have undergone the same differentiation process seen during normal development, with coexpression of both cone opsins in the same cell. ⁴⁰ Photoreceptor regeneration is a well-known process in submammalian species, ⁵⁹ and it has been suggested to take place also in mammals to a limited extent under physiologic and also pathologic conditions. ^35,60  

In conclusion, the results reported here provide a new component to the pathology of diabetic retinopathy. We detected morphologic evidence for the early degeneration of photoreceptors and RPE, as well as a shift in the expression pattern of cone opsins before significant apoptotic loss and expected vascular alterations. We lack evidence on whether the functions of photoreceptors are altered or not in this experimental setup. However, it seems reasonable to assume that the degeneration of the outer segments we observed may be functional and may contribute to changes of color vision, dark adaptometry, and contrast sensitivity reported in diabetic patients ^7,12–16 and also in rats. ^17,61 The remarkable damage of rods shown here may also go clinically unnoticed for a long time in human patients, where daylight vision depends mainly on the integrity of the cone mosaic. 

Neurophysiologic tests as well as functional ophthalmologic tests are not a part of routine diabetes care today. However, our results indicate that morphologic degeneration is present indeed in early diabetes, well before the expected vasculopathy. Electroretinography and psychophysical tests may provide positive results earlier and therefore might be more sensitive indicators of the onset of diabetic retinopathy than monitoring vascular lesions alone. ¹  

Based on our results, we propose that neuroprotective therapy should be considered in early diabetes. Streptozotocin-induced diabetes in rats may serve as a valuable model system for testing future neuroprotective agents. Considering the lack of macular region and rod-dominant retinas of rats as limitations of this experiment, we believe that further studies involving human patients and retinas are required to draw final functional consequences. 

Supplementary Materials

pdf S1 

Acknowledgments

We thank Pál Röhlich for critically reading the manuscript and for help with electron microscopic analyses. The assistance of Lászlóné Szemere, Éva Kovácsné Dobozi, and Györgyné Vidra is appreciated. 

Supported by a grant from the Hungarian Scientific Research Fund (OTKA 73000) and TÁMOP-4.2.1.B-09/1KMRB2010-0001. 

Disclosure: A. Énzsöly, None; A. Szabó, None; O. Kántor, None; C. Dávid, None; P. Szalay, None; K. Szabó, None; Á. Szél, None; J. Németh, None; Á. Lukáts, None