Paired-Flash Identification of Rod and Cone Dysfunction in the Diabetic Rat

Abstract

purpose. To investigate the onset of retinal neural dysfunction in the streptozotocin (STZ)-induced diatebic rat.

methods. A cohort of 20 Sprague-Dawley rats were randomly assigned to treatment (STZ 50 mg/kg, n = 10) and control (citrate buffer, n = 10) groups and observed for 12 weeks. Diabetes was confirmed by blood glucose (>15 mmol/L) and HBA_1c (>7.0%). Treated animals received 2 to 3 U insulin daily. Retinal function was monitored using paired-flash electroretinograms (ERGs) at baseline and various time points between 2 days and 12 weeks after treatment, to allow isolation of rod and cone components. Protocols compared photoreceptor and inner retinal responses (rod and cone) at each time point.

results. Losses in the function of rod photoreceptors and the inner retina were seen 2 days after STZ injection, with recovery in some components by 4 weeks and a secondary loss of function at 12 weeks. Some inner retinal responses (cone response and rod oscillatory potentials (OPs) remained consistently depressed over the entire 12 weeks.

conclusions. Retinal neural dysfunction was observed as early as 2 days after STZ injection. These acute changes reflect either STZ toxicity or hyperglycemia as a result of pancreatic compromise. Consistent loss over the 12 weeks of the cone response and OPs suggests a vulnerability of the inner retina to STZ-related effects. The 12-week losses in function of retinal neurons are consistent with a generalized diabetic neuropathy, since impaired function developed simultaneously in both inner and outer retinal neurons.

Diabetic eye disease is the leading cause of blindness among people of working age in the Western World, ¹ yet a thorough understanding of its pathogenesis remains elusive. Many studies have focused on the nature of the retinal vasculopathy accompanying diabetes—in particular, the level of vascular endothelial growth factor (VEGF), ^{2 3} the permeability of the inner blood–retina barrier, ⁴ and the angiotensin pathways. ^{5 6} However, apart from these vascular changes, an increasing body of evidence suggests that retinal neuropathy can occur either before, or in conjunction with, the vascular disorder. ^{7 8 9 10 11} It was the purpose of this study to consider the nature and time course of the retinal dysfunction (using the electroretinogram; ERG) associated with STZ-induced diabetes. 

Losses of retinal function have been reported in both diabetic patients and animal models of type I diabetes, with the most common anomaly being a reduction in the amplitude and frequency of the oscillatory potentials (OPs). ^{9 12 13 14 15 16 17 18} These oscillations are thought to reflect inner retinal function, ^{19 20} which raises the prospect of greater vulnerability of the inner retina in diabetic eye disease. However, there are several arguments for outer retinal dysfunction in diabetes. The retinal rods are known to be particularly energy demanding, ²¹ and the metabolic changes associated with diabetes ^{4 22 23 24} are likely to compromise rod function. That rods should be affected by metabolic disorders is not surprising, given that they are the most metabolically demanding neurons of the central nervous system, ²¹ due to their dark currents. ²⁵ These dark currents are maximum in the absence of light stimulation and arise from ionic fluxes between the outer and inner segments of the photoreceptor. They are sustained by the Na⁺,K⁺-ATPase located in the inner segment of the photoreceptor ²⁶ making these neurons susceptible to disease that alters the available metabolic substrates. 

The ERG and, in particular, the leading edge of the a-wave, can be used to study the effects of altered metabolism on the photoreceptor. ^{27 28} The magnitude of the photoreceptoral response to light can be measured by the amplitude of the ERG a-wave (Fig. 1A) , whereas the gain or slope of the a-wave reflects the sensitivity of the phototransduction G-protein cascade. ^{29 30} As a consequence, the a-wave provides a useful means of assessing retinal metabolic disorders. 

The light response of the photoreceptor is related to the activity of the Na⁺,K⁺-ATPase, and, as diabetes affects this enzyme, ^{22 23} a reduction in a-wave amplitude should be present as an early sign of the disease, leaving the G-protein cascade unaltered, an observation made at 12 weeks in a previous study of the unisolated photoreceptor response. ¹¹ However, this is not a common finding in the literature, which may reflect the manner of data collection and extraction. In this study, we consider the prospect for receptoral loss in diabetes by isolating rod responses and inner retinal components and also examine the timing of the onset of functional changes in the STZ rat. 

Materials and Methods

Animals

Experimental protocols were approved by our institutional ethics committee and complied with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. A cohort of 20 8-week-old Sprague-Dawley rats (male) were fasted overnight (>12 hours). Ten were treated with STZ (50 mg/kg, pH 4.5 in citrate buffer) by injection into the tail vein. The remaining 10 rats acted as control subjects and were given an injection of an equivalent volume of citrate buffer (1 mL/kg; pH 4.5). All diabetic animals had persistent hyperglycemia (>15 mmol/L) over the 12 weeks and returned abnormal glycosylated hemoglobin (HBA_1c >7.0%) at the 12-week time point. Diabetic animals received either 2 or 3 U of long-acting insulin (human protophane) daily, commencing on day 8 after injection: the higher dose was used whenever blood glucose exceeded 30 mmol/L. 

Electroretinography

Animals had their retinal function measured at baseline, 2 days and 1, 2, 4, 6, and 12 weeks after STZ injection. Data collection occurred after general anesthesia with a mixture of ketamine and xylazine (60:5 mg/kg; Therapon Pty. Ltd., Burwood, Victoria, Australia) and corneal anesthesia with proxymetacaine (Ophthetic 0.5%; Allergan, Frenchs Forest, New South Wales, Australia). The ERG procedure is a modification of that described previously ¹¹ with waveforms being collected at high luminous exposures (1.7, 2.1 log cd · s · m⁻² known to saturate the dark current and isolate photoreceptor activity. All manipulation was performed under dim red illumination (λ_max = 650 nm) and measurements were taken after overnight dark adaptation (>12 hours) to maximize retinal sensitivity. ³¹ Flash ERGs were recorded with two photographic flash units (285 V; Vivitar Photographics, Newbury Park, CA) presented with a Ganzfeld sphere. The light sources were attenuated with calibrated neutral-density filters (Wratten; Eastman Kodak, Rochester, NY). ERG waveforms were recorded after pupillary dilation with tropicamide (Mydriacyl 0.5%; Allergan) using custom made silver-silver chloride electrodes referenced to a stainless steel ground inserted in the tail. Responses were amplified (gain ×1000; −3 dB at 0.1 and 3000 Hz; P55; Grass-Telefactor, West Warwick, RI) and digitized at 2 kHz. 

Rod and Cone Isolation

Cone responses were isolated with a paired-flash paradigm. ³² In this protocol an initial bright flash (2.1 log cd · s · m⁻²) elicits a mixed (rod–cone) response that saturates rods, whereas the second flash is presented during the period in which rods have not yet recovered (<2 seconds), to elicit cone responses. ^{33 34} The rod response can be extracted by digital subtraction of the cone component from the mixed (rod-cone) waveform as reported elsewhere ³² and shown in Figure 1 . Our paired flashes were presented using a 1-second interstimulus interval with the average of two signals used in the study. 

The principle components of an ERG waveform are shown in Figure 1A . These include a negative-going a-wave followed by a positive-going b-wave on which can be seen some three to four small oscillations, called the oscillatory potentials (OPs). The leading edge of the saturating a-wave reflects receptor activity ^{35 36} and can be successfully modeled by the biochemical mechanisms mediating phototransduction (Fig. 3A) . This component of the electroretinogram has been described by Granit as the PIII. ³⁷  

The model used to fit the PIII to the a-wave is described by equation 1 , ³⁰  

\[\mathrm{PIII}(i,t)\ {=}\ {\{}1\ {-}\ \mathrm{exp}{[}{-}i\ {\cdot}\ S\ {\cdot}\ (t\ {-}\ t_{\mathrm{d}})^{2}{]}{\}}\ {\cdot}\ R_{\mathrm{max}},\]

where PIII gives the summed photocurrent as a function of luminous exposure, i (cd · s · m⁻²) and time, t (in seconds). R _max (microvolts) is the saturated amplitude of the PIII, S is its sensitivity (m² · cd⁻¹ · s⁻³), and t _d (seconds) is a brief delay that accounts for biochemical and other recording latencies. Parameter optimization was achieved by fitting the model to the raw data as an ensemble over the two luminous exposures (1.7, 2.1 log cd · s · m⁻²). For this purpose, t _d was fixed to 2.65 × 10⁻³ seconds (the average delay of the control group) and the sum-of-square error term was minimized with the solver module of a spreadsheet (Excel; Microsoft, Redmond, WA). Raw data and model outcomes for two representative animals are shown in Figure 3A . 

Inner Retinal Function

The inner retinal components can be visualized by digital subtraction of the receptoral response (PIII) from the raw waveform (see Fig. 4A ). This yields the b-wave generator and the OPs, both of which are thought to reflect inner retinal processing. ^{19 20 38 39} The b-wave generator has been called the PII by Granit, ³⁷ and we conform to this nomenclature herein. After filtering the PII waveform with a 7-ms running average (3 dB down; 59-Hz filter) to remove oscillations from its peak, we describe the PII by its maximum amplitude (in microvolts) and time to peak (implicit time in milliseconds). 

OPs appear on the rising slope of the PII and can be exposed using the method of Bui et al. ⁴⁰ This requires digital subtraction of both the PIII and a modified PII from the raw waveform and filtering (55–280 Hz, 512-tap FIR filter, Blackman window) to expose the oscillations. The outcome of this process is shown in Figure 5B . Bui et al. ⁴⁰ propose that these oscillations can be modeled in the time domain by a Gabor function (see Fig. 5A ), which results when a sine-wave carrier is windowed in a Gaussian time envelope (equation 2) .  

\[\mathrm{Gabor}\ (x)\ {=}\ a\ {\cdot}\ e^{{-}\ \frac{1}{2}\left[\left(\frac{x\ {-}\ m}{s}\right)^{2}\right]}\ {\cdot}\ \mathrm{sin}2\ {\cdot}\ h\ {\cdot}\ x\ {\cdot}\ {\pi}\ {+}\ p\]

 

The Gabor describes the OPs in the time (x, in seconds) domain by their maximum amplitude (a, microvolts), time to peak (m, milliseconds), spread (s, milliseconds), frequency (h, Hz), and phase (p). Data extraction and fitting were achieved by floating all parameters and minimizing the sum-of-squares error term on a computer spreadsheet (Excel; Microsoft). As can be observed from the waveforms of Figures 1 and 2 , four peaks were superimposed on the leading edge of the b-wave at high-luminosity exposures, with the first wavelet occurring in the a-wave trough. As previous studies on diabetic rats have analyzed only those OPs on the rising slope of the b-wave (>15 ms), ^{14 41} we chose to model the later three oscillations in our Gabor analysis, filtering out the early peak, to be consistent with the literature. 

The rat cone waveform (Fig. 6A) was too small for reliable extraction of receptoral and OP components, and so the cone response was analyzed in terms of the maximum amplitude (microvolts) and time to peak (implicit time; milliseconds) of the cone b-wave, which we termed the cone PII. 

Statistics

Because we found significant age-related trends in our parameters, we chose to consider the results from diabetic animals as departures from the average age-matched control value at each time point. These were evaluated using a repeated-measures analysis of variance (RM-ANOVA) and a Geiser-Greenhouse corrected F ratio, with the main effects of group (Control, STZ) and time point (in weeks) and group–time interaction. However, as the variability in the control group did not differ over time, the 95% confidence limits of the control group were considered as the average across the 12 weeks of the experiment. 

In the case of significant interactions, an analysis of simple comparisons was performed at each time point to identify the cause of the significant departure. A corrected α of 0.025 was adopted in cases of multiple comparisons, to protect against type I errors. ⁴²  

Normality was tested with a Kolmogorov-Smirnov test (StatView, version 5.0.1; SAS Institute, Cary, NC), and homogeneity was evaluated with a variance ratio (ς² max/ς² min). The α level (0.025) protects against type I errors incurred with nonhomogenous data. ⁴²  

Intragroup variability was evaluated by comparing group variance at early (average of 2 days, 1 week, and 2 weeks) and late (12 weeks) time points with a variance ratio. This was evaluated with an F statistic using α = 0.05. 

Comparisons across parameters with different magnitudes were achieved with a standardized deviate or z-score ([control mean-individual]/SD). Performance beyond the 5% confidence limit (1.96 SD) was considered significantly different from control. A Pearson product moment correlation was used to evaluate the relationships between early (average z-score 2-day to 2-week) and late (12-week) time points. 

Results

Average blood glucose levels and weights for diabetic and STZ animals are shown in Table 1 . The average HBA_1c was significantly elevated (P < 0.025) in the STZ group (12.85 ± 0.44%) compared with the control (3.58 ± 0.09%). All diabetic animals showed significantly greater blood glucose (P < 0.025) and minimal weight gain over the 12 weeks of experimentation, as indicated in Table 1 . 

Diabetic Animals Show Rod Dysfunction as Early as 2 Days after STZ Injection

Figure 2 shows a representative waveform stack returned by the 2.1 log cd · s · m⁻² exposure from the same control and diabetic animals at each time point. As can be seen in Figure 2 , an immediate decrease in the ERG waveform was observed at 2 days after STZ injection. This decrease was maintained for at least 2 weeks, with recovery at 4 to 6 weeks and a secondary deficit becoming apparent at 12 weeks after STZ treatment. 

Figure 3 shows the rod receptoral (PIII) parameters at each time point for both individual animals and the average departure (±SEM) from age-matched control values. The figure confirms the trends in the waveform stack and indicates significant interindividual variability. On average, there was a significant decrease (P < 0.01) in rod PIII amplitude at 2 days that was sustained for 2 weeks. During the same period, significantly greater variability was found between animals at the early time points (unfilled symbols, F_28,9 = 3.02, P < 0.05) than at 12 weeks. Recovery of the photoreceptoral amplitude was noted at 4 weeks and was maintained to 6 weeks, from which time variability was normal. At 12 weeks, the rod PIII of STZ animals again declined below control limits (P < 0.01), but this time it had a similar variability from the control (F_9,9 = 1.17, P > 0.05). Figure 3B shows the change in sensitivity of the photoreceptoral response, an index of the G-protein cascade. There was no significant change in the average sensitivity parameter from control values (Fig. 3C) over the 12 weeks, although the early time points returned significantly larger variability between animals (F_28,9 = 18.6, P < 0.01). 

An extracted rod PII response for the diabetic animals is shown in Figure 4A . A similar pattern of loss across time was evident in this postreceptoral component, as noted in the previous paragraph for the amplitude of the receptoral (PIII) response. Significant reductions in the average rod PII amplitude were found at the early (P < 0.025) and late (P < 0.01) time points, with normalization between. This reduction was associated with greater variability at the early time points (F_28,9 = 6.45, P < 0.01). The PII implicit time, however, showed a different trend, in which a significant and sustained decrease in the time to peak was recorded after 2 weeks. The variability in the time to peak was also significantly larger at early time points than at 12 weeks (F_28,9 = 18.62, P < 0.01). 

Recovery of OPs and Cone PII

Representative 12-week diabetic and control OPs modeled in the time domain with the Gabor (Fig. 5A) are shown in Figure 5B . Overall, the OPs returned by the diabetic group at 12 weeks had smaller amplitudes (control, 106.9 ± 6.52 μV versus diabetic, 76.0 ± 7.02 μV; P < 0.01), a slower time to peak (control, 27.9± 0.49 ms versus diabetic, 32.7± 0.86 ms; P < 0.001), and an increased spread (control, 4.95 ± 0.12 ms versus diabetic, 6.90 ± 0.14 ms; P < 0.001) and resonated at a reduced frequency (control, 123.6 ± 1.69 Hz versus diabetic, 111.7 ± 2.27 Hz; P < 0.0001). In contrast to the PIII and PII components, the amplitude of the OPs was significantly reduced over most of the 12 weeks (P < 0.025), with the exception of the 1-week (P = 0.30) and 6-week (P = 0.07) data (Fig. 5C) . Similar trends were found for the time to peak (Fig. 5D) , with significant delays at 2 days and 4 and 12 weeks (P < 0.025), and the spread (Fig. 5F) , which was significantly larger at all time points (P < 0.01). In contrast, the frequency of the diabetic OPs (Fig. 5E) only became significantly reduced at 12 weeks (P < 0.001). Consistent with the previous data, at early time points there was significantly greater variability in amplitude (F_28,9 = 3.53, P < 0.05) and spread (F_28,9 = 26.29, P < 0.01) than at 12 weeks. 

Figure 6A shows representative control and diabetic waveforms of the isolated cone response at 12 weeks. In contrast to the diabetic rod response, the cone PII showed a significant reduction in amplitude at all time points measured over the 12 weeks (P < 0.025, Fig. 6B ). The variability in the data was consistently large and not significantly different between the early and late time points (F = 2.02, P > 0.05). The large and persistent intersubject variability in cone amplitude suggests the possibility that some animals display a permanent loss early, a question to be considered next. Although a significant delay was found in the cone PII implicit time at 2 days and 1 week, it returned to normal over the remaining 11 weeks (Fig. 6C) . 

Relationship between Early and Late Losses

To evaluate the relationship between the losses at early and late time points, we performed correlations of rod PIII and PII amplitude changes at these times. These components were chosen because recovery was evident at 4 to 6 weeks. To facilitate comparisons across components with different magnitudes, we represented the data as z-scores (see the Methods section). This correlation is shown in Figure 7 , where it is evident that no diabetic animal had significant losses at both the early and late time points (PIII, r = 0.63; PII, r = 0.52, P > 0.05). This suggests the possibility that the early and late losses are mediated by different mechanisms. 

Discussion

Using the paired-flash paradigm, we isolated the rod and cone components of the ERG waveform and found that both receptoral and postreceptoral neural dysfunction occurred in the diabetic rat as early as 2 days after injection of STZ, at a time when no vascular changes were evident. ⁴³ In general, the pattern of rod dysfunction shows two phases of loss: an early loss from 2 days to 2 weeks that is associated with high levels of variability, followed by recovery at 4 to 6 weeks, and a secondary deficit at 12 weeks. In contrast, the cone and OP components show an immediate decrease in inner retinal function with no recovery, suggesting that inner retinal neurons are vulnerable to STZ-induced diabetes. 

Diabetes and Early Rod Dysfunction

Rod dysfunction began 2 days after STZ injection, suggesting that neural dysfunction precedes the onset of diabetic retinopathy. ⁴³ These changes are consistent with the 27% loss of PIII amplitude reported by Bui et al. ¹¹ at 12 weeks. However, the cause of these losses is not clear. 

Data in previous studies reporting functional loss in diabetes assessed with ERG are not easy to interpret in terms of specific neural losses. Estimates of the b-wave derived from trough-to-peak measurements without prior extraction of the receptoral component result in confounding photoreceptor losses with inner retinal changes. ³⁵ We modeled the leading edge of the rod a-wave, thus isolating the photoreceptoral response, which allows its interpretation in terms of energy-dependent processes that determine the maximum rod response and energy-independent processes that determine the sensitivity of photoreceptor activation (G-protein cascade). ²⁹ As predicted for a metabolic disorder, the energy-dependent process to be affected is the amplitude of the rod PIII (Fig. 3) . This amplitude change could arise from several factors, including the number or length of rod outer segments, ⁴⁴ the number of channels within the outer segment membranes, the lipid profile of the outer segment membranes, ⁴⁵ and an altered transmembrane hyperpolarization associated with dysfunction in Na⁺,K⁺-ATPase. ²⁵ Although a reduction in the size ⁴⁶ of the outer nuclear layer or the lengths ⁹ of outer segments has been reported in diabetes, these changes occur only later in the disease (4–24 weeks), inconsistent with the timing of the losses in our study. However, a reduction in Na⁺,K⁺-ATPase activity is a common finding early in diabetes ⁴⁷ and has been shown to occur in the retina within the first weeks of the disease. ^{22 23} As Na⁺,K⁺-ATPase maintains the amplitude of the photoreceptor response, ^{21 25} any change in the activity of this enzyme should result in a reduced a-wave amplitude. Recent work from our laboratory using in vivo inhibition of Na⁺,K⁺-ATPase has produced a-wave losses consistent with our findings (Phipps JA, et al. IOVS 2004;45:ARVO E-Abstract 3233; Weymouth AE, et al. IOVS 2004;45:ARVO E-Abstract 1349), and we believe that a reduction in retinal Na⁺,K⁺-ATPase activity may be partially responsible for the photoreceptoral losses in the present study. 

Inner Retinal Vulnerability in Diabetes

We observed inner retinal neural dysfunction early after the onset of diabetes, consistent with previous reports. ^{9 12 13 14 15 16 17 18} The pattern of loss in amplitude of the inner retinal responses in this study showed an immediate (2 days) and sustained decrease over the 12 weeks. The OPs are thought to arise from interactions between ON- and OFF-bipolar cells and other inhibitory neurons within the inner retina, ²⁰ most probably amacrine cells. ¹⁹ Previous studies in humans have shown that OPs are predictive of retinopathy, ⁴⁸ which reflects the sensitivity of these components to the hypoxic stress caused by diabetes. ^{28 49 50} Because previous studies on diabetic rats have shown the amplitude and kinetics of the OPs to be inherently related, ¹⁷ modeling the OP response with the Gabor fit of Bui et al., ⁴⁰ is useful, in that it considers the OPs as an ensemble of inner retinal responses, rather than individual peaks arising from different sources. Furthermore, the Gabor fit allows the characterization of the kinetics of the OPs in terms of a delayed timing (the time to peak parameter [m, in milliseconds]) and an expansion in time between the first and last OPs (the spread [s, in milliseconds]). Both these parameters were significantly affected at 12 weeks of diabetes. Considering the OPs in terms of individual times to peak and amplitude may fail to pick up a global slowing of the OP response. Moreover, extraction of the cone response from the mixed rod-cone ERG allows the unmasking of the rod OPs, which are larger than in the mixed response (see Fig. 1 ) and allows differentiation between cone changes and losses in the neurons responsible for the OPs. Considering the potential usefulness of the OPs in monitoring the development of diabetic retinopathy, ⁴⁸ a thorough and standardized classification of the OP response is warranted. 

Similar to the OPs, the cone PII is known to involve postreceptoral neuronal processing, ^{38 51} so the early and sustained losses of both components suggests a particular sensitivity of the inner retina to the changes caused by STZ-induced diabetes. The vulnerability of the cone PII in diabetes is interesting and may be explained partially by a reduction in cone photoreceptor outer segments as early as 1 month after the onset of diabetes. ⁹ As the ERG is a serial waveform, the cone bipolar cell loss may reflect this photoreceptoral change. However, as cone photoreceptors only represent ∼1% of the total photoreceptor population, ⁵² it is more likely that a specific vulnerability of the inner retinal neurons underlies the cone PII change. 

As most diabetic abnormalities take at least 1 week to develop in the retina, ^{3 43 53} it is unlikely that the 2-day losses in all components of the ERG are due to metabolic changes. We believe that these early losses may arise as complications of the STZ treatment, a question to be considered in the following section. 

Effect of STZ Injection on the Retina

There are several explanations for the immediate functional loss, one being that the STZ injection causes the early changes. STZ is not a benign substance, and its use as a toxic agent for diseases other than diabetes suggests that the drug is not specific to pancreatic β cells. ^{54 55} However, it is unknown at this stage whether STZ crosses the blood–retina barrier, and anatomic studies on STZ-treated rats do not support the prospect of early toxic damage. A notable exception is the recent report by Park et al. ⁴⁶ who noted a temporary increase in inner retinal thickness at 1 week, possibly reflecting swelling caused by neurotoxicity that recovers after STZ is cleared from the circulation. This timing is similar to our functional losses, and the changes in the ERG reported at early time points by others. ^{14 18}  

An alternative cause of these early losses may be an increased level of circulating glucose after pancreatic β-cell destruction with STZ. ⁵⁶ Changes in the ERG and electrooculogram have been reported in humans preloaded with glucose (Schneck ME, et al. IOVS 1999;40:ARVO Abstract 3766), ⁵⁷ and injections of both glucose and glucose inhibitors have been shown to result in variations in the ERGs of rats. ^{28 49} Certainly, the variability between animals at the early time points may be consistent with this cause. Further work is needed to investigate the exact nature of these early losses. 

Mechanisms of Early and Late Losses

A correlation between the rod PIII and PII amplitude changes at early and late time points (Fig. 7) found that no diabetic animal had significant losses at both times, suggesting that early and late losses are being mediated by different mechanisms. The cause of the early loss is not clear, as detailed previously, and needs further investigation. However, we believe that the late losses (12 week) arise from diabetes-related neuropathy. In support of this proposal, we note that the rod receptoral change (reduced Rm_PIII and normal sensitivity of the PIII) is likely to be produced by a metabolic disorder, as discussed previously. Moreover, the reduced OP frequency corresponds to a similar change reported in humans with diabetes, ⁵⁸ and the functional changes are consistent with reports of biochemical and anatomic manifestations of diabetic eye disease at this later time. ^{18 22 23 46}  

In summary, we found both rod and cone abnormalities in diabetes, with functional losses in inner and outer retinal neurons from 2 days after the injection of STZ and before the onset of patent vasculopathy. A sustained loss occurred over 12 weeks in some postreceptoral components, suggesting that the inner retina is more vulnerable to STZ-induced diabetes. We believe the STZ rat is a good model for diabetic neuronal change, particularly if the early time points are avoided, as large variability is evident in the functional measurements at this stage and the cause of the early loss is not clear. The use of a paired-flash paradigm allows identification of rod and cone components that can be considered in terms of energy-dependent and -independent processes and should provide an invaluable tool for assessing diabetic change and treatment modalities in the future. 

 Supported in part by Grant 208950 from the National Health and Medical Research Council of Australia and Grant LP0211474 from the Australian Research Council Linkage Project.

 Submitted for publication July 19, 2004; revised August 13, 2004; accepted August 27, 2004.

 Disclosure: J.A. Phipps, None; E.L. Fletcher, None; A.J. Vingrys, 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.

 

Schematic of the paired-flash paradigm and components of ERG waveforms. (A) A mixed response was isolated with the initial saturating flash (2.1 log cd · s · m⁻²) of the paired flash paradigm. (B) The probe, shown 1 second later, elicited a cone response. (C) The putative rod response was achieved by digital subtraction of the cone response from the mixed rod-cone ERG (A, B).

Figure 3.

 

Rod photoreceptor function was affected from 2 days to 2 weeks and at 12 weeks after STZ injection. (A) The leading edge of the first negative-going waveform (a-wave) of the putative rod response was analyzed as the PIII (photoreceptoral response) in terms of its maximum response (Rm_PIII) and slope (Sensitivity). Diabetic animals at 12 weeks (○, dotted lines) showed a reduction in the Rm_PIII, with normal sensitivity compared with control subjects (•, solid lines). (B) The diabetic group (•) showed a reduction in the average Rm_PIII at 2 days and at 1 week, 2 weeks, and 12 weeks after STZ injection. (○) Individual diabetic animals at each time point. (□) Average diabetic value at baseline. Data are shown as the change from the control group average (diabetic individual − control group average). Error bars ± SEM. Shaded area: the 95% confidence limits for the control group mean over the 12 weeks of experimentation (SEM). Dashed lines: ±2 SD in individual control animals. Zero line: the average for the control group over the 12 weeks of the experiment, with the absolute value shown at right. (C) Phototransduction sensitivity of the rod photoreceptoral response was unaffected by STZ treatment and showed no change over the 12 weeks of experimentation. Symbols are as described in (B).

Figure 4.

 

Rod inner retinal (bipolar cell) function was affected from 2 days to 2 weeks and at 12 weeks after STZ injection. (A) Representative rod PII waveforms from a control animal (solid line) and a diabetic animal (dashed line). The postreceptoral PII rod response is derived after subtraction of the rod PIII response from the raw rod waveform (Fig. 1C) . (B) The diabetic group showed a loss in the average rod postreceptoral (PII) response from 2 days to 2 weeks after injection and again at 12 weeks after diabetogenesis. Data are shown as the change from the control group average (diabetic individual − control group average). Symbols, lines, and shading are as in Figure 3B . (B) The STZ-treated animals showed faster PII implicit times at 2, 4, and 6 weeks after injection.

Figure 5.

 

The oscillatory potential response was affected in diabetic rats. (A) After conditioning and filtering, extracted OPs were modeled in a Gabor (Gaussian envelope with a sine wave carrier) and analyzed in terms of their amplitude (microvolts), frequency (Hz), peak time (milliseconds), and spread (milliseconds). (B) Extracted OPs of representative control (•, solid line) and diabetic animals (○, dashed line) at 12 weeks after STZ injection. (C) The STZ group (•) showed a reduction in OP amplitude for the entire 12 weeks of experimentation. Symbols, lines, and shading are as described in Figure 3B . (D) The time for the OP peak increased at 2 days, 4 weeks, and 12 weeks. (E) OP frequency was significantly reduced only at 12 weeks (•) in the diabetic group. (F) The spread of the OPs was significantly larger in the diabetic animals at all time points except 1 and 2 weeks.

Figure 2.

 

A representative waveform stack over the 12 weeks of experimentation on the same control and diabetic animals. The diabetic animal (dashed lines) showed a reduction in electroretinogram responses at 2 days (2d), 1 week (1w), 2 weeks (2w), and 12 weeks (12w) compared with the control animal (solid lines). Note that the oscillatory potentials of this diabetic animal increased between 2 days and 1 week, before decreasing again at 2 weeks, displaying large variability typical of the diabetic group at the early time points.

Figure 6.

 

Cone function was reduced in STZ-treated animals. (A) Representative waveforms at 12 weeks in a control animal (solid line) and an STZ-treated (dashed line) animal. (B) The average diabetic values (filled symbols) showed a sustained reduction in cone PII amplitude from 2 days to 12 weeks. Symbols, lines, and shading are as described in Figure 3B . (C) Cone PII implicit times were increased from the first week after STZ injection and then stayed within normal limits over the remaining period.

Table 1.

 

Average Weight and Blood Glucose Levels (±SEM) for all Animals over the 12 Weeks of Experimentation

Time	Weight (g)			Blood Glucose (mmol/L)
		Control	Diabetes	Control	Diabetes
Baseline	260.3 ± 4.6	247.1 ± 6.33	5.15 ± 0.15	5.03 ± 0.18
2 days	260.3 ± 4.6	230.2 ± 5.16, †	—	26.4 ± 0.75*
1 week	283.3 ± 8.0*	233.0 ± 9.58, †	—	26.4 ± 0.84*
2 weeks	317.5 ± 3.9*	234.5 ± 9.65, †	—	28.4 ± 0.44*
4 weeks	363.8 ± 4.9*	244.5 ± 11.74, †	—	26.9 ± 0.43*
6 weeks	394.9 ± 5.4*	248.1 ± 12.53, †	5.22 ± 0.13	26.4 ± 0.49* , †
12 weeks	468.8 ± 5.3*	283.6 ± 15.50* , †	5.11 ± 0.19	23.9 ± 0.75* , †

 Control blood glucose was measured at baseline, six and 12 weeks, and pair-wise comparisons between control and diabetic blood glucose levels were restricted to these time points.

*  Significantly removed (P < 0.025) from own baseline.

†  Significantly removed (P < 0.025) from controls.

Figure 7.

 

Correlations for PIII (Control ○; STZ •) and PII amplitudes (Control ▵; STZ ▴) early (average of 2 days, 1 week, and 2 weeks) and late (12 weeks). Data are represented as z-scores. Dotted area: 95% control limits. Unshaded area: zone of significant loss at both early and late time points.