Early Inner Retinal Dysfunction in Streptozotocin-Induced Diabetic Rats

Abstract

purpose. Diabetes is known to alter retinal function, as measured with the electroretinogram (ERG), which shows a propensity toward inner retinal oscillatory potential (OPs) abnormalities. However, the effect that diabetes has on other ganglion cell–related responses is not known. This study was a systematic evaluation of streptozotocin (STZ) diabetes–related ERG changes in rats for the first 11 weeks after diabetogenesis.

methods. Thirty Sprague-Dawley rats were randomly assigned to treated (50 mg/kg STZ (n = 16) and control groups (1 mL/kg citrate buffer, n = 14) at 6 weeks of age. Two control animals and four STZ animals were excluded because of blood glucose criteria or systemic complications. Diabetic animals were given daily SC injections of 1 to 2 units of long-acting insulin. ERGs were measured at 4, 8, and 11 weeks after treatment. The a-wave was used as an index of outer retinal function, whereas the b-wave, OPs, and the scotopic threshold response (STR) were used as indices of inner retinal function.

results. Photoreceptoral (a-wave) and bipolar cell (b-wave) responses were not significantly reduced by STZ treatment. OPs were significantly reduced by 8 weeks (−25% ± 7%, P < 0.05). The most severely affected component was the ganglion cell–dominated positive STR, which was significantly decreased from the first time point (−51% ± 11% at 4 weeks, P < 0.05), but the negative component was unaffected over the 11-week period.

conclusions. The ganglion cell dominated pSTR showed large losses in STZ treated rats.

Diabetes can damage neurons, ¹ glia, ^{2 3} and the vascular tissues within the retina. ^{4 5} Evidence of neuronal alterations includes the presence of apoptosis ^{6 7} in the inner retinal layers and photoreceptors in diabetic animals. ^{7 8} This diabetes-related cell loss leads to a reduction in retinal thickness, which has been demonstrated with histopathology, ^{1 6 7} optical coherence tomography, ⁹ and scanning laser polarimetry, ¹⁰ in both animals and humans with diabetes. The histologic changes to the inner retina seem to be more severe. ^{1 6}  

Clinical evidence for neuronal damage comes from studies showing that diabetic patients can have reduced visual acuity, ¹¹ visual field sensitivity, ¹² contrast sensitivity, ¹³ color vision, ¹⁴ and flicker sensitivity. ¹² Studies in which objective electrophysiological tests, such as the electroretinogram (ERG), ¹⁵ the pattern ERG (PERG), ¹⁶ the multifocal ERG, ¹⁷ and the visual evoked potential, ¹⁶ were used have shown abnormalities before evidence of vascular change in diabetic eyes. 

In animals and humans with diabetes, the most common ERG finding is for a reduced amplitude and prolonged implicit time in the oscillatory potentials (OPs) and in some cases an altered photoreceptor response. ¹⁵ OPs are small-amplitude, high-frequency wavelets, found on the rising slope of the b-wave and thought to involve amacrine cell activity. ¹⁸ The early OP abnormality suggests that these retinal interneurons are more susceptible to diabetes, especially as the photoreceptor changes have been shown to be related to the omega-3 fatty-acid changes secondary to the diabetes-induced lipid anomaly and not the diabetic hyperglycemic state (Yee P et al. IOVS 2004;45:ARVO E-Abstract 4151). As anatomic studies also find apoptosis in ganglion cells early in diabetes, ^{1 19} it is likely that ERG components reflecting ganglion cell integrity will be affected along with the OPs. 

One ERG component that reflects ganglion cell activity is the scotopic threshold response (STR). Recent work in rats has shown that the positive (p)STR reflects ganglion cell activity, whereas the negative (n)STR reflects both ganglion cell and amacrine cell activity. ²⁰ Conflicting evidence for STR loss has been reported in humans and animals with diabetes. Aylward ²¹ found in humans that the reduction in STR amplitude and increased latency correlate with the stage of diabetic retinopathy. On the other hand, Kaneko et al. ^{22 23} report no STR deficits in diabetic rats or humans. However, as both investigators measured only the nSTR, this may confound contributions from cells other than the ganglion cell. ²⁰ Moreover, the nature of the receptoral and postreceptoral inputs to the ganglion cells have not been well defined in the same diabetic cases. In this study, we considered the nature of ERG changes early in diabetes with an emphasis on inner retinal changes manifest in the OPs and STR. 

Methods

Animals

Procedures adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Thirty, male Sprague-Dawley rats were housed in an air-conditioned environment (21°C) with diurnal light cycling (50 lux, 8 AM–8 PM). Food and water were provided ad libitum. At 6 weeks, the rats were randomly assigned to receive a tail vein injection of either 50 mg/kg streptozotocin (n = 16, STZ; MP Biomedicals, Solon, OH) dissolved in trisodium citrate buffer (1 mL/kg of 0.01 M, pH 4.5; Sigma-Aldrich, Castle Hill, NSW, Australia) or buffer alone (n = 14). 

Blood glucose levels were measured at 1, 3, 9, and 12 weeks after STZ injection (Ascensia Esprit2 and Glucodisc; Bayer HealthCare, Pymble, NSW, Australia), with levels >15 mmol/L indicative of diabetes. One week after injection, animals with STZ-induced diabetes (STZ animals) had daily SC injections of insulin (1 to 2 units Protaphane; Novo Nordisk Pharmaceuticals, Baulkham Hills, NSW, Australia) to sustain body weight and general condition and to better mimic the human condition. All animals received at least 1 unit of long-acting insulin, those showing poor grooming and increased urine output as indicated by the condition of the bedding, received 2 units. Three STZ-injected animals were excluded, as blood glucose levels were not sustained above this criterion (>15 mmol/L). Systemic complications arose in two control and one STZ rat, and these data were excluded. Thus, group data represent the average of 12 control and 12 STZ animals. 

Electroretinographic Procedures

ERGs were recorded at 4, 8, and 11 weeks after diabetogenesis using a Ganzfeld bowl (Photometric Solutions International, Huntingdale, VIC, Australia) containing 20 white LEDs (5-watt, Luxeon; Philips Lumileds Lighting Co., San Jose, CA). Luminous energy was calibrated using an integrating photometer (IL1700) with a filter (Z-CIE; International Light Technologies Inc., Newburyport, MA) and is specified in scotopic units. LED voltage and duration (1–4 ms) were varied to yield luminous energies ranging from −6.08 to 1.92 log scot cd · s · m⁻². LED temporal profile and conversion from scotopic cd · s · m⁻² to photoisomerizations/rod are given elsewhere. ²⁴  

At the dimmest energy, 20 responses were averaged, with fewer being collected at brighter energies. The interstimulus interval was 2 seconds for the dimmest flashes and increased to 120 seconds for bright energies. Responses were filtered 0.1 to 1000 Hz, amplified ×1000 (P511; Grass Technologies, West Warwick, RI) and digitized at 4 kHz (PowerLab, ADInstruments, Bella Vista, NSW, Australia) over a 640-ms epoch. Signals were collected (Scope software, ver. 3.7.6; ADInstruments) for post hoc analysis (Excel; Microsoft, Redmond, WA). Electrodes were silver chloride, one placed on the center of the cornea and the other looped about the equator of the same eye, both referenced to a stainless-steel needle (Grass Technologies) inserted in the tail. 

Responses were collected simultaneously from both eyes after overnight dark-adaptation (>12 hours) and after anesthesia, with an intramuscular injection of a mixture of ketamine hydrochloride (60 mg/kg; Troy Laboratories, Smithfield, NSW, Australia) and xylazine (5 mg/kg; Ilium Xylazil-100; Troy Laboratories). Both corneas were anesthetized with 0.5% proxymetacaine hydrochloride (Alcaine; Alcon Laboratories, Frenchs Forest, NSW, Australia) and lubricated (Celluvisc; Allergan, Irvine, CA). The pupils were dilated (>4 mm) with 0.5% tropicamide (Mydriacyl; Alcon Laboratories), and body temperature was maintained at 37 ± 0.5°C, with a water heating pad. ²⁴ All procedures and electrode placement were performed under dim red illumination (λ_max = 650 nm). 

Photoreceptor Response (P3)

We modeled the leading edge of the a-wave with a delayed Gaussian ²⁵ as given by equation 1 :  

\[\mathrm{P}3(i,t){=}\mathrm{Rm}_{\mathrm{P}3}{\cdot}{[}1{-}\mathrm{exp}({-}i{\cdot}S{\cdot}(t{-}t_{\mathrm{d}})^{2}){]}\ \mathrm{for}\ t{>}t_{\mathrm{d}}\]

This equation describes the P3 response as a function of luminous energy, i (log cd · s · m⁻²), and time, t (s). The response is scaled by the maximum saturated amplitude (Rm_P3, μV). The delay, t _d (s) of the phototransduction cascade and its sensitivity, S (m² · cd⁻¹ · s⁻³), modify this response. Although the t _d is mainly determined by nonphysiological factors, work by our group has shown that this parameter can be altered by diabetes, ²⁶ and so we floated t _d. Parameter optimization was achieved over an ensemble of the three highest energies (1.32–1.92 log cd · m⁻²), using the solver module of a spreadsheet (Excel; Microsoft) by minimizing the sum-of-squares (SS) merit function. Past work has shown that the cone contribution to the a-wave at these luminous energies is less than 8% ²⁷ and thus negligible for our purpose. 

Bipolar Cell Response (P2)

The P2 component underlying the b-wave reflects inner retinal function, particularly that of ON-bipolar cells. ²⁸ Its was isolated by digital subtraction of the P3 model from the raw data to yield the P2–OP complex, ²⁴ which was then low-pass filtered (−3 dB, 50 Hz) to expose the P2. P2 amplitude and implicit time were measured from baseline to peak of this waveform and the relationship between P2 amplitude and luminous energy was described by a Naka-Rushton function ²⁹ :  

\[V(i){=}V_{\mathrm{max}}{\cdot}\frac{i^{n}}{i^{n}{+}K^{n}}\]

The P2 amplitude (microvolts) as a function of luminous energy i (log cd · s · m⁻²) is described by its saturated response, V _max (microvolts), a semisaturation constant K (log cd · s · m⁻²), and a slope n. The slope returns to 1 for a single underlying cellular generator ³⁰ and <1 in the case of multiple generators. 

Inner Retinal Response (OPs and STR)

OPs were extracted from the P2–OP complex with a fifth-order Butterworth band-pass filter (−3 dB; 50–250 Hz). In Sprague-Dawley rats OP3 is the largest oscillation and the wavelet analyzed for intensities at or brighter than −4.2 log cd · s · m⁻². The smooth transition in OP peak time across intensity (see 1 2 3 4 5 Figs. 6G 6H 6I ) is consistent with the isolation of OP3. ³¹ However, OP3 is difficult to identify at dimmer intensities, as the waveform becomes more complex (see Fig. 2C , bottom two traces) with signals that are close to noise levels. As such, we cannot be confident that, at intensities of less than −4.2 log cd · s · m⁻², we have returned OP3. 

To decrease variability, we averaged OP amplitude over the five brightest energies (≥0.72 log cd · s · m⁻²) for each animal (Fig. 6 , shaded regions). This value was then used to calculate the relative OP change between treated and control animals (Fig. 7) . 

The amplitudes of the pSTR and nSTR were returned at fixed times of 120 and 220 ms (A₁₂₀ and A₂₂₀) after stimulus onset. The implicit times of the pSTR and nSTR were taken from the peak and trough of the STR waveform. STR amplitude and implicit time were averaged across four energies (−6.08 to −5.27 log cd · s · m⁻²), to increase signal to noise. These intensities contain minimal intrusion from components other than the STR. 

Statistical Analysis

Group data are specified as the mean ± SEM. Normality was determined with a Kolmogorov-Smirnov test, and variance homogeneity was tested by using a variance ratio. Statistical trends across intensity were determined using repeated-measures ANOVA (Prism, ver. 4.00; GraphPad Software Inc., San Diego, CA) with Bonferroni post hoc comparison between groups. For P3 and P2 analysis age (4, 8, and 11 weeks) was nested in treatment (control versus STZ). For other analysis both age and intensity were nested within treatment. We used an unpaired t-test to evaluate differences in systemic parameters between the control and STZ groups. An α of 0.05 was applied for all statistical purposes. 

Results

Figure 1Ashows that our STZ-treated animals expressed the typical diabetic milieu. Although the body weight of the STZ group increased by 31% (261 ± 6 to 341 ± 9 g) over the 12-week period, control rats showed significantly greater weight gain (49%; 290 ± 5 to 433 ± 11 g; F _1,3 = 8.37, P < 0.001). Post hoc analysis shows that body weights diverged significantly after 4 weeks (P < 0.01, Fig. 1A ). Figure 1Bshows that STZ animals returned significantly higher blood glucose at all time points (STZ: 25.4 ± 0.8 mmol/L versus control 9.7 ± 1.1 mM, P < 0.001). The elevation in control blood glucose at the final time point (12.8 ± 0.6 mmol/L) occurred because it was collected with anesthesia needed for tissue harvest and known to elevate blood glucose. ³²  

Figure 2shows representative ERG waveforms, at selected intensities, for a control (thin traces) and a diabetic rat (bold traces), 11 weeks after treatment. The dimmest three waveforms show the STR response, where the positive lobe of the STR was smaller in STZ rats than in control animals (Fig. 2A) . For brighter energies, the b-wave was reduced in the STZ animal, whereas the a-wave appeared unaffected (Fig. 2B) . The STZ animal returned lower OP amplitudes at brighter intensities (> 0.72 log cd · s · m⁻²), as shown in Figure 2C . 

Photoreceptor Response (P3)

We found a paradoxical increase in Rm_P3 at 8 weeks in the STZ group (P < 0.01) which was confirmed by the significant treatment × time interaction (F _1,2 = 4.36, P = 0.02, Fig. 3A ). However, as the STZ amplitude returned to control levels at 11 weeks (control, 536 ± 20 μV; STZ, 554 ± 36 μV) we interpret this finding as a chance observation. We do not find the reduced Rm_P3 reported by others (−16% to −27%), an issue that will be considered later. The decline in Rm_P3 with age found in the control group is consistent with the age-related change to Rm_P3 reported in rats over similar age ranges. ³³  

Phototransduction sensitivity (S) also significantly declined in both groups at 8 and 11 weeks (P < 0.001). However, the treatment × time interaction and treatment main effect were not significant (F _1,2 = 0.55, P = 0.58; F _1,22 = 2.78, P = 0.11; Fig. 3B ) indicating that diabetes does not affect sensitivity in insulin-treated diabetic rats, as reported by others. ^{26 34 35} The latency of the phototransduction cascade (t _d), was not significantly different between control and diabetic groups (control, 4.54 ± 0.06 ms; STZ, 4.55 ± 0.07 ms; F _1,22 = 2.78, P = 0.11). 

Bipolar Cell Response (P2)

Figure 3shows that the maximum amplitude of the P2 (V _max) decreased with age between 4 and 8 weeks for both control and diabetic groups (F _1,22 = 5.56, P = 0.01). However, neither the treatment × time interaction (F _1,2 = 2.20, P = 0.12) nor the treatment effect (F _1,22 = 1.73, P = 0.20) were significant. Similarly, the semisaturation constant K decreased in both groups between 4 and 8 weeks (significant time effect: F _2,69 = 7.05, P = 0.002; Fig. 3E ) indicating improved P2 sensitivity. However, the treatment × time interaction and treatment main effects were again not significant (F _1,2 = 0.69, P = 0.51: F _1,22 = 0.04, P = 0.84). Finally, the slope n for the STZ group was significantly increased compared with that in the control (treatment effect F _1,22 = 17.75, P < 0.001, Fig. 3F ). However, the lack of a significant treatment × time interaction (F _1,2 = 0.22, P = 0.81) indicated no selective time related changes. 

Scotopic Threshold Response

Figure 4shows the peak amplitude (A–C) and implicit times (G–I) for the pSTR. The shaded area of each panel highlights the dim stimulus energies (−6.08 to −5.27 log cd · s · m⁻²) at which the STR dominates, as evidenced by a stable implicit time in control animals over these energies (∼115 ms, Figs. 4G 4H 4I ). Note that the implicit time shows a transition to a slower peak at approximately 130 ms, presumed to be the rod b-wave, and which in turn transitions to a much faster peak at bright energies (∼85 ms), presumed to be the mixed rod–cone b-wave. 

The best fit Naka-Rushton model shows that the STZ group (bold curve) had a steeper slope compared with the control group (thin curve, Fig. 4 ). Normalizing pSTR amplitudes to the average control value confirms a greater amplitude reduction at lower stimulus energies in STZ-treated rats (Figs. 4D 4E 4F , indicated by the horizontal bar). This accounts for the steeper slope of the Naka-Rushton function in Figures 4A 4B 4C . Not surprisingly, a significant treatment × intensity interaction was found at all weeks (4 weeks: F _1,2 = 5.25, P < 0.001; 8 weeks: F _1,2 = 11.46, P < 0.001; 11 weeks: F _1,2 = 5.67, P < 0.001). 

Figures 4G 4H 4Ishow that STZ affected P2 timing differently across energies with a significant treatment x intensity interaction at all weeks (4 weeks: F _1,2 = 2.48, P < 0.001; 8 weeks: F _1,2 = 2.16, P = 0.001; 11 weeks: F _1,2 = 1.76, P = 0.015). This finding was more evident in the normalized timing change in Figures 4J 4K 4L . At intermediate intensities (−3.79 to −0.48 log cd · s · m⁻²), the STZ group had a slower P2, whereas at high intensities (> 0.41 log cd · s · m⁻²) the P2 was faster than in the control. It is of note that peak time becomes more variable at the dimmest intensities (gray area) and with the increasing duration of diabetes. This variation may reflect the selective loss of the positive STR component, which in the absence of other positive components becomes difficult to define. 

Figure 5shows the amplitude (A–C) and the implicit time (G–I) for the negative component of the STR. When the four lowest intensities were analyzed across all weeks, the STZ animals had significantly larger nSTR amplitudes (F _1,69 = 6.74, P = 0.015). This result shows a significant treatment × intensity interaction (F _1,3 = 3.11, P = 0.034). Post hoc analyses revealed that the nSTR was larger in the STZ group at 4 and 8 weeks only (P < 0.05). 

In terms of nSTR implicit time, there was a statistically significant treatment × intensity interaction at 11 weeks (F _1,2 = 4.70, P = 0.005), but not at 4 and 8 weeks (4 weeks: F _1,2 = 0.78, P = 0.51; 8 weeks: F _1,2 = 0.59, P = 0.62). However, treatment effects were not significant at 4 and 8 weeks (4 weeks: F _1,22 = 4.40, P = 0.05; 8 weeks: F _1,22 = 4.33, P = 0.05). 

Oscillatory Potentials (OPs)

Figure 6shows the amplitude (A–C) and implicit time (G–I) of the largest OP (or OP3 for intensities at or above −4.2 log cd · s · m⁻²), with relative amplitude (D−F) and timing change (J−L) also shown. The shaded area of each panel shows the brighter stimulus energies (0.72–1.92 log cd · s · m⁻²), where OP parameters were averaged to give an overall effect. Figure 6shows two phases in OP amplitude growth: The first growth phase began at −4 log cd · s · m⁻² where it returns its slowest timing. The first plateau was at −2.5 log cd · s · m⁻², and thereafter the OPs sped up and showed amplitude saturation at the highest energies. 

The effect of STZ on OP amplitudes varies as a function of intensity, with a decrease at higher light levels (Figs. 6D 6E 6F)as reflected in the significant treatment × intensity interaction (F ₁₅₂₈ = 4.22, P = 0.025). Overall OP amplitude was reduced in the STZ group (Figs. 6A 6B 6C , F _1,22 = 4.93, P = 0.037), with significant reductions in OP amplitudes at 8 and 11 weeks (P < 0.05) but not 4 weeks. At 8 weeks, post hoc tests showed that STZ OPs were significantly smaller for energies above −1.5 log cd · s · m⁻². Post hoc analyses at other ages were not significant. Figures 6G 6H 6Ishow OP implicit times for all ages. OP peak time in STZ animals was slower at intermediate energy levels, but faster than in control animals at high energies. Although STZ appears to slow OPs most at intermediate intensities, the effect was not significant. 

Relative STZ Effect on ERG Components

Figure 7compares the effect of STZ on the various ERG components normalized to the average control. Figure 7Ashows that both P2 (V _max) and P3 (Rm_P3) amplitudes were little affected by diabetes. Averaged over all ages, only the P2 was significantly smaller in STZ rats (average; −6% ± 3%: F _1,22 = 10.28, P = 0.004). Figure 7Bshows that whereas the P3 timing was slower (at 8 weeks), P2 peak time was faster at this age. Figure 7Cshows that the nSTR was paradoxically increased at 4 and 8 weeks (overall, 22% ± 6%), whereas the pSTR was reduced at all weeks (overall average, −56% ± 7%). In addition, the nSTR was 6% ± 2% slower (Fig. 7D) and was significant at 11 weeks. Figure 7Eshows that OPs were reduced (F _1,2 = 7.20, P = 0.014) at 8 and 11 weeks. However the overall average OP reduction of −19% ± 5% was less than the pSTR. It is also worth noting that at 4 weeks, the only ERG component significantly reduced by STZ treatment was the pSTR. OP and pSTR timing was unchanged (Figure 7F) . 

Discussion

Photoreceptoral a-Wave (P3)

We found that 11 weeks of STZ-diabetes returned photoreceptor amplitude, sensitivity and timing parameters (Rm_P3, S and t _d; Fig. 3 ), that were not significantly different from the control. However, the absence of an a-wave abnormality cannot be ruled out, given that, despite a longitudinal experimental design, we needed an experimental power of 59% to find a 20% a-wave difference between the control and STZ-treated group. Nevertheless, normal a-waves have been reported in STZ-treated rats ^{36 37 38} and human diabetics. ³⁹ On the other hand, studies also report photoreceptor deficits in STZ animals ^{26 34 35 40} and humans. ⁴¹ Differences in the duration of diabetes in the above studies may account for the disagreement as to the degree of a-wave loss. Li et al. ⁴⁰ and Phipps et al. ³⁴ find significant a-wave changes 10 to 12 weeks after STZ induction and not before. Thus, it is possible that a-wave deficits may have become apparent if we had evaluated the STZ effect later. 

Age-Related Changes in Control and STZ-Treated Rats

It is worth noting that Rm_P3 decreased by approximately 15% in both groups between 4 and 11 weeks (age, 10–17 weeks). This decline may be attributed to ageing of the eye. Fulton and Hansen ³³ showed that the rat a-wave reaches its peak amplitude at 5 weeks after birth. Kiyosawa ⁴² showed that Wistar rats undergo a ∼17% decline in a-wave amplitude between 5 and 17 weeks of age. Likewise, Hancock and Kraft ³⁸ reported a 28% reduction in a-wave amplitude from 8 to 20 weeks of age. These changes are similar in magnitude to our age-related changes. 

Age related ERG amplitude attenuation may arise from eye growth. However, Guggenheim et al. ⁴³ show that rat axial length reaches adult dimensions by 10 weeks. Alternatively, Kiyosawa ⁴² demonstrated thinning of outer and inner nuclear layers in rats between postnatal weeks 5 and 13. Our control animals showed the greatest a-wave change between postnatal weeks 10 (4 weeks after STZ) and 14, and only a small reduction between weeks 14 and 17. This functional pattern matches the anatomic trends, ⁴² thus postnatal refinement of retinal layers may account for our amplitude reduction. An interesting finding is that the age-related a-wave decline occurred later at 14 weeks of age (8 weeks after treatment) in STZ rats. However, as we have only a single time point at 8 weeks, the existence of this trend needs further investigation. 

ON-Bipolar Cell b-Wave (P2)

Li et al. ⁴⁰ reported a >30% b-wave loss at 2 weeks, whereas Hancock and Kraft ³⁸ reported a 33% b-wave loss 12 weeks after STZ-injection. Parisi et al. ⁴⁴ found the b-wave reduced in diabetic patients. Although, we found only a small b-wave decrease (−13%) in STZ rats, our finding is similar to those in a previous report of normal b-wave amplitudes at 8 weeks, ³⁴ followed by a b-wave loss of 15% to 18%, 12 weeks after STZ induction. The lack of change in b-wave implicit time in our study is also consistent with results in another study. ³⁴  

An interesting finding is that the slope of the b-wave intensity response function was significantly increased at all times. This increase in slope is consistent with the finding that amplitude change is greater at dimmer intensities. This issue is discussed in the following sections. 

Inner Retinal Oscillatory Potentials (OPs)

A robust outcome in diabetes is that the OPs are either reduced or delayed, as reviewed by Shirao and Kawasaki. ¹⁵ OP deficits in diabetes were first observed in the early 1960s. ⁴⁵ Several groups have since shown that smaller OP amplitudes are associated with greater retinopathy in humans. ^{21 44 46} Other studies have found OPs to be affected early in STZ diabetes (<5 weeks). ^{34 37 38} However, Li et al. ⁴⁰ reported that OP losses occurred only at later ages (≥20 weeks) and after the b-wave was reduced. 

Kizawa et al. ⁴⁷ found that the OPs were more affected by diabetes than was the a- or b-wave, in a large group of human diabetics. Similarly, animal studies, have found OPs to be more attenuated ²⁶ and affected earlier than outer retinal responses. ^{34 36} This is consistent with our finding that OPs were significantly reduced at 8 (−25% ± 7%) and 11 (−22% ± 11%) weeks, before any losses in the a- or b-waves. 

Scotopic Threshold Response

Parisi et al. ⁴⁴ and Aylward ²¹ have reported that OP amplitudes are better indicators of diabetic retinopathy severity than other waveforms arising from the inner retina. They show that the PERG (review by Parisi and Uccioli ¹⁶ ), which is the ganglion cell response to a contrast-reversing checkerboard ⁴⁸ is less sensitive in diabetes than are the OPs. Similarly, Kizawa et al. ⁴⁷ showed in human diabetics that OPs are better indicators of retinopathy progression than is the photopic negative response (PhNR), which receives contributions from ganglion cells in humans. ⁴⁹ However, Arden et al. ⁵⁰ reported the PERG as the best indicator of diabetic retinopathy. Consistent with Arden et al., ⁵⁰ we found an apparent pSTR deficit 4 weeks after STZ treatment, whereas we did not see significantly reduced OPs until 8 weeks. 

The finding that the pSTR loss was greater than the b-wave reduction, together with no loss in the nSTR suggests that ganglion cell dysfunction does not arise from reduced input from distal elements. This contention receives further support from our finding that the slope of the b-wave intensity response function was significantly steeper (Fig. 3F) . A steeper slope has been detected after removal of proximal retinal contributions, by using γ-amino butyric acid application in mouse retina. ³⁰  

Kaneko ²³ reported normal nSTR in human diabetics and in rats 2 weeks after STZ administration. ²² Although, our findings are in agreement with those of Kaneko et al., ²² the absence of an nSTR loss (Fig. 7)is paradoxical, given that the nSTR receives contributions from both ganglion and amacrine cells. ²⁰ It is likely that, in early diabetes, losses in positive components such as the pSTR and the P2 mask any reduction in the corneal negative nSTR. However, nSTR reductions may manifest later in the course of diabetes. Consistent with this possibility, Aylward ²¹ has found the nSTR to be a good predictor of more severe retinopathy in humans. It is also worth noting that the paradoxical increase in the nSTR argues that the loss in inner retinal function reflects gains between the ON-bipolar cells and inner retinal pathways, as this would manifest as a reduction in all inner retinal waveforms. 

It is also worth noting that the glial cell damage known to exist in diabetes ^{2 3} may influence the STR, as glial cell–mediated potassium flux is thought to be involved in pSTR and nSTR generation. ⁵¹ However, our finding for a pSTR loss with a larger nSTR suggests that the deficit is unlikely to be mediated solely by glial cell dysfunction, as this would affect both pSTR and nSTR to a similar magnitude. 

Mechanisms of Retinal Dysfunction

Our finding that the pSTR demonstrated the greatest deficit is consistent with the increased apoptosis observed in the ganglion cell layer at 2 to 4 weeks after STZ treatment in rats, ^{1 52} after 2 months in mice, ⁶ and in human diabetics. ¹⁹ However, increased apoptosis has also been observed in other retinal cell populations—in particular, amacrine cells ⁷ and photoreceptors. ⁸  

Several mechanisms may account for increased ganglion cell dysfunction. First increased activation of the polyol pathway ⁵³ and sorbitol accumulation ⁵⁴ can alter retinal function. Sorbitol accumulation has been shown to be higher in the ganglion cell layer in alloxan-treated rabbits. ⁵⁴ Moreover, ganglion cell loss is decreased with treatment to reduce sorbitol accumulation. ⁵³ Second, vascular abnormalities specific to the inner retinal capillary plexus would be more detrimental to proximal (amacrine and ganglion cells) than to distal retinal neurons. Do Carmo et al. ⁵ reported that the inner blood–retinal barrier in diabetic rats is more compromised than is the outer barrier in STZ rats. Third, glutamate concentration is increased in the retina of diabetic animals ⁵⁵ and in the vitreous of humans with proliferative diabetic retinopathy. ⁵⁶ Ganglion cells are particularly susceptible to a sustained extracellular increase in glutamate, leading to excitotoxicity. ⁵⁷  

The exact mechanism leading to greater ganglion cell dysfunction in STZ rats must be further investigated. Nevertheless, diabetes induced ganglion cell dysfunction may represent a risk factor for glaucoma (reviewed by Toda and Nakanishi-Toda ⁵⁸ ). In glaucoma, it is thought that mechanical compression of the laminar cribrosa ⁵⁹ leads to reduced retrograde transport of neurotrophic factors and ganglion cell loss. ⁶⁰ Abnormal axonal transport has been reported in diabetic ganglion cell axons. ⁵³  

In summary, we showed that in the presence of minor photoreceptor and ON-bipolar cell changes, inner retinal function was significantly reduced in STZ-treated rats. The ganglion cell–dominated pSTR was the most sensitive component to STZ diabetes, manifesting as early as 4 weeks after STZ treatment and before OP loss. 

 Supported by National Health and Medical Research Council Grants 400127 (BVB) and 350224 (AJV).

 Submitted for publication January 2, 2008; revised March 11 and 31, 2008; accepted May 28, 2008.

 Disclosure: K. Kohzaki, None; A.J. Vingrys, None; B.V. Bui, 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.

 

Figure 1.

 

Average (±SEM) body weight and blood glucose for control (n = 12) and STZ animals (n = 12). (A) Body weight was significantly decreased after 4 weeks in STZ compared with control animals. (B) Blood glucose was significantly increased at 3, 9, and 12 weeks after STZ treatment compared with the control (each n = 5, 2, and 12). Dashed line: normal limit of blood glucose. Statistically significant: **P < 0.01, ***P < 0.001

Figure 2.

 

The effect of STZ treatment on the full-field ERG waveforms at 11 weeks after STZ injection for a representative control (thin traces) and STZ-treated rat (bold traces). (A) Response series to dim luminous energies (−6.08 to −3.09 log cd · s · m⁻²). (B) Response series to brighter luminous energies (−1.22 to 1.92 log cd · s · m⁻²). (C) OPs isolated from flash energies in (B). Luminous energies (log cd · s · m⁻²) are given to the left of the waveforms. Scale bars are shown to the top of each group of waveforms to which they apply.

Figure 3.

 

The effect of STZ treatment on P3 and P2 at 4, 8, and 11 weeks. P3 parameters: (A) saturated photoreceptor amplitude (Rm_P3), (B) sensitivity (S), (C); delay (t _d) of phototransduction cascade. P2 parameters: (D) maximum amplitude (V _max); (E) the semisaturated constant (K); (F) the slope (n) of the Naka-Rushton function. Bars, group mean ± SEM. Statistically significant: **P < 0.01, ***P < 0.001.

Figure 4.

 

Effect of STZ on the first positive peak. (A–C) Show the relationship between peak amplitude and stimulus energy modeled with a Naka-Rushton function (lines) at 4, 8, and 11 weeks, respectively. In the each panel, lines fit control (thin line) and STZ (bold line) data. (D–F) Data normalized to the control mean. Asterisk and bar: the range over which post hoc analysis was significant. (G–I) show the relationship between peak implicit time and stimulus energy at 4, 8, and 11 weeks, respectively. (J–L) Normalized data as before. Shaded area gives the region for STR responses recorded from dimmer stimulus energies (−6.08 to −5.27 log cd · s · m⁻²). Each symbol and error bar represents the mean ± SEM in the control (n = 12) and the STZ (n = 12) groups.

Figure 5.

 

Effect of STZ on the average (± SEM) nSTR amplitude and implicit time. Amplitudes were taken at a fixed criterion time of 220 ms after stimulus flash. Data are shown for 4 (A), 8 (B), and 11 (C) weeks after STZ treatment. (D–F) Normalized amplitudes. Implicit times were measured to the trough minimum (G–I). (J–L) Relative implicit times. The shaded region identifies the nSTR response recorded from dimmer stimulus energies (−6.08 to −5.27 log cd · s · m⁻²).

Figure 6.

 

Effect of STZ on the peak OP response. (A–C) Average (± SEM) OP amplitude; (D–F) normalized response at 4, 8, and 11 weeks, respectively. Asterisk and bar: the range over which post hoc analysis was significant. (G–I) Average (± SEM) implicit time; (J–L) normalized implicit time. The shaded region shows OP responses recorded from brighter stimulus energies (0.72–1.92 log cd · s · m⁻²).

Figure 7.

 

STZ data (mean ± SEM) normalized to the average control values at 4, 8, and 11 weeks. The amplitude and the timing of Rm_P3 versus the V _max is shown in (A) and (B), respectively. (C, D) Amplitude and timing of the nSTR versus the pSTR, respectively. (E, F) Amplitude and timing of the OPs versus the pSTR, respectively. In each panel, the shaded area represents the normal range for the filled symbols, whereas the solid lines represent the normal range for open symbols. Dashed line: normal level. Statistically significant: *P < 0.05, **P < 0.01, ***P < 0.001.