Identifying cell class specific losses from serially generated electroretinogram components

Abstract

Purpose: Processing of information through the cellular layers of the retina occurs in a serial manner. In the electroretinogram (ERG), this complicates interpretation of inner retinal changes as dysfunction may arise from "upstream" neurons or may indicate a direct loss to that neural generator. We propose an approach that addresses this issue by defining ERG gain relationships.

Methods: Regression analyses between two serial ERG parameters in a control cohort of rats are used to define gain relationships. These gains are then applied to two models of retinal disease.

Results: The PIII(amp) to PII(amp) gain is unity whereas the PII(amp) to pSTR(amp) and PII(amp) to nSTR(amp) gains are greater than unity, indicating "amplification" (P < 0.05). Timing relationships show amplification between PIII(it) to PII(it) and compression for PII(it) to pSTR(it) and PII(it) to nSTR(it), (P < 0.05). Application of these gains to ω-3-deficiency indicates that all timing changes are downstream of photoreceptor changes, but a direct pSTR amplitude loss occurs (P < 0.05). Application to diabetes indicates widespread inner retinal dysfunction which cannot be attributed to outer retinal changes (P < 0.05).

Conclusions: This simple approach aids in the interpretation of inner retinal ERG changes by taking into account gain characteristics found between successive ERG components of normal animals.

Figure 1

ERG analysis: saturated ERG components arising from the rod through pathway. (a) Schematic of retinal cytoarchitecture showing rod photoreceptors, bipolar, and ganglion cells with relevant interneurons (AII amacrine and cone bipolar (CBC)). (b) a-wave was modelled with a delayed Gaussian over an ensemble of two luminous energies (1.22 and 1.52 log cd·s·m⁻²) to derive Rm_PIII (PIII_amp). The time to reach 80% trough amplitude was taken as the implicit time (PIII_it). (c) To isolate rod responses, a paired flash paradigm was implemented, and the rod waveform was derived following subtraction of cone from mixed waveforms. (d) Rod PII waveform was derived by subtracting the modelled PIII (Panel (b)) from rod isolated waveforms (Panel (c)). Rod PII implicit time was taken at 80% of maximal amplitude (PII_it). (e) STR amplitude was measured at the peak (pSTR_amp) and trough (nSTR_amp). STR implicit time was taken at 80% of maximal amplitude (pSTR_it, nSTR_it). (f) Analysis undertaken on components at their saturated response. Energy-response functions illustrate that, for photoreceptor (PIII, white circles) and bipolar cell (PII, grey circles), this occurs at 1.52 log cd·s·m⁻² and for ganglion cell (pSTR, black circles) at −5.26 log cd·s·m⁻².

Figure 2

Normal gain between ERG components. The gain relationship between two ERG components is established by linear regression (thick line) of normalised upstream to the downstream ERG parameter for a group of control animals (unfilled circles) measured on 5 separate occasions. A slope of unity is represented by the thin diagonal lines. Statistics for this analysis are represented in Table 1. Correlations of (a) PIII_amp and PII_amp give slopes which are near unity. Correlations that produce slopes steeper than unity include (b) PII_amp and pSTR_amp (c)  PII_amp and nSTR_amp, and (d)  PIII_it  and PII_it. Slopes shallower than unity were observed for correlations between (e) PII_itand pSTR_it and (f) PII_it and nSTR_it.

Figure 3

Application of ERG gain analysis in ω-3 deficiency. Averaged group ERG waveforms showing the effect of ω-3 dietary deficiency on the (a) rod-isolated a-b wave complex and (b) STR (reproduced with permission from Nguyen et al. [28]). In Panels (c)–(h) an ERG gain of unity is represented by the thin diagonal lines. The thick black lines represent the ERG gain determined in Figure 2 and Table 1. Grey bars represent the 95% confidence interval for the ω-3 deficiency treatment group. The horizontal arrows represent the predicted downstream loss given the measured upstream change. Asterisks (∗) indicate statistically significant direct loss at the respective component (P < 0.05). Thus ω-3 deficiency produces (c) PII amplitude reductions that can be expected from the PIII_amp decline, and (d) pSTR_amp  losses are greater than predicted by ERG gain. The predicted pSTR_amp loss (arrow) falls outside the 95% confidence limits (grey bar) for the measured pSTR_amp loss in the treated group (filled square). (e) nSTR change can be accounted for by the reduction in PII_amp. In terms of timing ((f), (g), and (h)) the predicted delays (arrows) fall within the 95% confidence interval of the treated groups. Thus the delays in the PII, pSTR, and nSTR can be accounted for by the delay in the outer retinal PIII.

Figure 4

Previous approaches to determine “downstream effects”: ratio analysis ((a) and (b)) and percentage change relative to a normal cohort ((c) and (d)) of ω-3 dietary deficiency. Panels (c) and (d) are reproduced with permission from Nguyen et al. [28]. Asterisks (∗) indicate a statistically significant direct loss at the respective cell (P < 0.05). (a) Ratio analysis of amplitudes indicates no dietary change (ω-3⁺ white bars, ω-3⁻ black bars) in the PII_amp/PIII_amp nor nSTR_amp/PII_amp ratio but a significant decrease in the pSTR_amp/PII_amp ratio. (b) Ratio analysis of implicit times indicates an increase in the PII_it/PIII_it ratio and a decrease in the pSTR_it/PII_it ratio and nSTR_it/PII_it ratio. (c) Percentage change analysis relative to the average control value (±95% CI, grey shaded area) indicates that ω-3 deficiency (average ± SEM, filled circle) exhibits greater dysfunction in the pSTR than the PIII, PII, or nSTR (d) Percentage analysis also indicates a greater delay in timing in the PII than the PIII, pSTR, or nSTR. Thus when comparing the ratio and percentage change analysis with the gain analysis (Figure 3) differences can be noted. Ratio and percentage change analyses indicate that all inner retinal timing changes are direct effects (PII_it, pSTR_it, and nSTR_it), whereas by taking into account gain between ERG components the timing delays of inner retinal components are attributable to the initial photoreceptoral change. Coincidentally, the amplitude changes are in agreement between the 3 analyses techniques.

Figure 5

Application of ERG gain analysis to ERGs recorded from STZ-diabetic rats. Averaged group ERG waveforms showing the effect of diabetes on the (a) rod-isolated a-b wave complex and the (b) STR. (c) The thick line shows the ERG gain slope defined from the control group. Thin diagonal line shows the unity relationship. For a given reduction in input (along the x-axis) the predicted downstream change is given by the arrow (along the y-axis). This prediction can be compared to the measured change induced by the treatment (filled symbol) along with the 95% confidence limits (grey bar). This shows that PII loss cannot be due to photoreceptoral dysfunction. (d) Likewise, pSTR loss is greater than the PII deficits. (e) nSTR increased amplitude is not downstream of PII. (f) The delay in the PII is greater than that predicted from the PIII. (g) Delay in the pSTR can be accounted for by the PII delay. (h) nSTR delay is less than that expected from the PII delay. Asterisks (∗) denote significance (P < 0.05).