Utility of Light-Adapted Full-Field Electroretinogram ON and OFF Responses for Detecting Glaucomatous Functional Damage

Abstract

Purpose: To compare parameters of electroretinogram (ERG) responses for their ability to detect functional loss in early stages of nonhuman primate (NHP) experimental glaucoma (EG), including photopic negative responses (PhNR) to a standard brief red flash on a blue background (R/B) and 200-ms-long R/B and white-on-white (W/W) flashes, to W/W flicker stimuli (5–50 Hz), and to a dark-adapted intensity series.

Methods: Light-adapted ERGs were recorded in 12 anesthetized monkeys with unilateral EG. Amplitudes and implicit times of the a-wave, b-wave, and d-wave were measured, as well as amplitudes of PhNRs and oscillatory potentials for flash onset and offset. Flicker ERGs were measured using peak–trough and fundamental frequency analyses. Dark-adapted ERG parameters were modeled by Naka–Rushton relationships.

Results: Only PhNR amplitudes were significantly reduced in EG eyes compared to fellow control (FC) eyes. The d-wave implicit time was delayed in EG versus FC eyes only for the W/W long flash, but in all eyes it was 10 to 20 ms slower for R/B versus the W/W condition. Flicker ERGs were <0.5 ms delayed in EG versus FC overall, but amplitudes were affected only at 5 Hz. The brief R/B PhNR amplitude had the highest sensitivity to detect EG and strongest correlation to parameters of structural damage.

Conclusions: The PhNR to the standard brief R/B stimulus was best for detecting and following early-stage functional loss in NHP EG.

Translational Relevance: These results suggest that there would be no benefit in using longer duration flashes to separate onset and offset responses for clinical management of glaucoma.

Introduction

It is well established that the photopic negative response (PhNR) of the light-adapted full-field electroretinogram (ERG), as well as the pattern electroretinogram (PERG), both reflect function of the retinal ganglion cells (RGCs) and thus can be informative for clinical diagnosis and monitoring of glaucoma and other optic neuropathies.^1–10 A potential advantage of using the PhNR for clinical testing is that the PhNR depends less than the PERG on uncorrected refractive error, clear optical media, and accurate fixation.^6,9,11 Another benefit of full-field photopic ERG recordings is that, along with the PhNR measurement of RGC function, they also provide direct simultaneous assessment of distal retinal function (cone and cone bipolar cell responses via a-wave and b-wave measurements). 

Most studies have found that the PhNR is best elicited by narrowband spectral stimuli that have strong chromatic contrast to the rod-suppressing background.^12–15 The standard recommended by the International Society for Clinical Electrophysiology of Vision (ISCEV) is a red flash on a blue background, for example.⁴ Our own previous work found that the PhNR for a brief (4.5-ms) red flash presented on a blue background provided superior diagnostic accuracy and generally stronger correlation to structural outcome measures than the PERG in a nonhuman primate (NHP) model of experimental glaucoma.¹¹ It may be possible to further extend the diagnostic potential of photopic ERG recordings for glaucoma by using a longer duration flash stimulus to separate the increment and decrement responses.¹⁶ This serves to emphasize the “ON” and “OFF” pathway responses, respectively, which are parallel streams of the visual pathway arising at the cone-to-bipolar cell synapse. Their opposing responses to a brief-duration (<5 ms) flash stimulus normally combine to shape the b-wave and other aspects of the photopic ERG.^17,18 Similarly, the relative amplitude and timing of ON and OFF pathway responses vary with the frequency and temporal profile of flicker ERG stimuli,^19–21 and possibly also exhibit differential sensitivity to glaucoma and other optic neuropathies.^22–24 Indeed, there is developing interest in probing the ON and OFF streams of the visual pathway separately because it may reveal early functional damage in glaucoma and provide insights about its pathophysiology.^{16,22,24–26} 

Mounting evidence, emerging both from experimental models and from humans, suggests that specific types of RGCs may be more vulnerable or susceptible to glaucomatous injury.²⁷ In particular, recent studies have found that the RGCs exhibiting earliest alterations are those whose dendritic arbor stratifies primarily within the distal sublamina of the inner plexiform layer (OFF sublamina), where RGCs make synaptic contacts with OFF bipolar cells and preferentially respond to light decrements within their receptive field center.²⁷ Although much of the recent evidence from rodent models points to vulnerability of OFF RGCs, especially OFF transient alpha-RGCs, some of those studies have reported that ON RGCs exhibit abnormalities prior to OFF RGCs.²⁷ Similarly, studies in human glaucoma have provided conflicting evidence, with some suggesting there may be preferential loss of the OFF pathway and others finding either no differential effect or a greater deficit of the ON pathway.^{16,22,24–26,28,29} 

Given the potential to enhance the clinical diagnostic utility of the full-field ERG and to probe the pathophysiology of glaucoma by separating ON and OFF responses, we conducted a systematic study to compare ERG responses to brief (4.5-ms) red-on-blue (R/B) flashes, longer duration (200 ms) R/B and white-on-white (W/W) flashes, and varied W/W flicker rates for their ability to detect loss of RGC function in NHP experimental glaucoma (EG). 

Methods

Subjects

Twelve adult rhesus macaque monkeys (Macaca mulatta) were included in this study (Table 1). All aspects of the study were carried out in strict accordance with the recommendations in the National Institutes of Health Guide for the Care and Use of Laboratory Animals, as well as with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. This study was approved and monitored by the Institutional Animal Care and Use Committee at Legacy Health (USDA license 92-R-0002; OLAW Assurance A3234-01; maintaining full accreditation by AAALAC International). 

Table 1.

 

Demographic and IOP Information for NHP Study Subjects

 

Study Design

Each of the 12 animals had unilateral EG induced by laser photocoagulation to the trabecular meshwork resulting in aqueous humor outflow infacility and chronic elevated intraocular pressure (IOP) in the lasered eye (Table 1). The contralateral eye of each animal served as the paired, fellow control.^11,30 Beginning at baseline (prior to any laser, when both eyes were healthy) and on an approximately weekly basis thereafter, all subjects underwent imaging by spectral-domain optical coherence tomography (OCT; SPECTRALIS OCT2; Heidelberg Engineering, Heidelberg, Germany) to monitor the status of glaucomatous structural damage.³⁰ Approximately every fourth week, multifocal ERG (mfERG) testing (VERIS; Electro-Diagnostic Imaging, Redwood City, CA) was performed instead of imaging to monitor the status of glaucomatous functional damage, as previously described.^11,31 The complete ERG test protocol for this study (see details below) was conducted when the EG eye of each animal reached the onset of optic nerve head rim tissue thinning, which was defined as the first of two consecutive measurements of the OCT parameter minimum rim width (MRW) being below the 95% confidence interval (CI) of test–retest measurement variability.³² The change in MRW was the sole inclusion criterion for this study. The status of the mfERG test results was not a determinant for inclusion; rather, they are reported here for comprehensive characterization of EG status at the time of cross-sectional assessment of the full-field flash and flicker ERG. Similarly, during all OCT imaging sessions, laser speckle flowgraphy (LSFG) scans (Softcare Co., Ltd., Fukuoka, Japan) were used to measure optic nerve head tissue (capillary) blood flow, as previously described.³³ Retinal nerve fiber layer (RNFL) retardance was measured using scanning laser polarimetry (GDx-VCC; Carl Zeiss Meditec, Dublin, CA), as previously described.^34,35 IOP measurements were obtained using a Tono-Pen XL (Reichert Technologies, Buffalo, NY) and recorded at the start of each weekly session as an average of three consecutive measurements in each eye. 

ERG Recordings

Anesthesia

Anesthesia for ERG testing was induced with an intramuscular (IM) injection of ketamine (10–40 mg/kg, N = 9 animals) or telazol (2.4–8.0 mg/kg IM, N = 3 ketamine-resistant animals), xylazine (0.5–2.0 mg/kg IM) or dexmedetomidine (0.01–0.015 mg/kg IM), atropine (0.05 mg/kg IM), and midazolam (0.05–2.0 mg/kg IM as needed). Note that there were no significant differences found for any ERG response parameter in control eyes of the nine NHPs that received ketamine versus the three NHPs that received telazol. Animals breathed pure oxygen (2.0 L/min) through an endotracheal tube for the duration of the ERG to maintain oxyhemoglobin saturation between 95% and 100%. Other vitals monitored and recorded every 10 to 15 minutes included end-tidal CO₂, respiratory rate, pulse rate, and temperature (kept at 37°C with a heated blanket). Anesthesia was maintained throughout all light-adapted and flicker ERG recordings with an intravenous (IV) continuous infusion of ketamine (0.6–2.4 mg/kg/hr) or pentobarbital (2.0–10 mg/kg/hr IV in animals resistant to ketamine or prone to eye movements) and, as needed, bolus injection of xylazine (0.25–1.0 mg/kg IM) and/or pentobarbital (10 mg/kg IV). Only after completion of all light-adapted and flicker ERG recordings, inhaled isoflurane (0.75–1.25%) was mixed with pure oxygen (2.0 L/min) at the start of dark adaptation and continued throughout the dark-adapted flash series. After recording, a broad-spectrum antibiotic ointment (neomycin and polymyxin B sulfates and bacitracin zinc ophthalmic ointment) was applied topically to both eyes, and animals were dosed with 0.29 ± 0.04 mg/kg atipamezole IM as needed for anesthesia reversal. 

Light-Adapted ERG Recordings

All ERGs were obtained after mydriasis was achieved using topical 1% tropicamide. Following completion of mfERG testing, full-field flash ERG recordings, which are the focus of this study, were acquired using a UTAS BigShot system and its integrated Ganzfeld stimulator (LKC Technologies, Gaithersburg, MD). Custom Burian–Allen contact lens ERG electrodes (Hansen Ophthalmic Development Lab, Bellingham, WA) were used in a monopolar recording configuration with <5 kΩ impedance: The active electrode was the ring at the periphery of the corneal contact lens resting on the test eye, and the reference electrode was the corneal ring resting on the contralateral eye while completely covered with a dark patch. The ground electrode was a subdermal platinum needle placed in a rear limb. A rod-suppressing 30 cd/m² blue background light (CIE coordinates: x = 0.134, y = 0.062) was presented to each eye for 10 minutes prior to recording the light-adapted (photopic) ERGs. Luminance calibrations of all flash stimuli and backgrounds produced by the instrument with light-emitting diodes were confirmed using a spot spectroradiometer (SpectraScan PR-650; Photo Research, Chatsworth, CA). Two responses to each stimulus type were acquired and then exported for offline analysis. No notch filter was used during acquisition; however, during offline processing a mains noise filter (60 Hz, Hamming window with −3 dB points at 54 and 66 Hz) and/or baseline offset of drift correction (based on the 20 ms of pre-stimulus baseline) were applied to any individual record that was obviously contaminated by such artifact. The average of the two traces was then used for measurement of response feature parameters. Details of stimulus/recording conditions and response parameterization are provided in the next sections for each type of ERG (see also example in Fig. 1). 

Figure 1.

 

Example of ERG responses to each stimulus type from an eye with experimental glaucoma (EG, red traces) and control eye (blue traces) of NHP1. (A) Light-adapted (photopic) single-flash ERG responses are shown in the left, middle, and right panels, respectively, for the brief 4.5-ms R/B stimulus, the long-duration 200-ms R/B stimulus, and the long-duration 200-ms W/W stimulus. (B) Light-adapted flicker ERG responses are shown for each flicker rate. (C) Dark-adapted (scotopic) single-flash ERG responses are shown for the complete range of stimulus strength. See Methods for details.

R/B Brief Flash

The brief R/B flash stimulus (560-cd/m² luminance and 4.5-ms duration) had a strength of 2.5 cd·s/m² and CIE coordinates of x = 0.692 and y = 0.305, and it was presented with a 2-second interstimulus interval (ISI; 0.5 Hz) on a 30 cd/m² blue background. Each of the two stored records was an average of 10 sweeps acquired with bandpass filter settings of 0.3 to 500 Hz at a 2000-Hz sampling rate (512 samples total length). As shown in Figure 1, the a-wave amplitude and implicit time were measured at the minimum voltage between 0 and 25 ms after stimulus onset. The b-wave implicit time was measured at the maximum voltage between 0 and 50 ms, and the amplitude was the difference between that voltage and the minima defining the a-wave. PhNR amplitude was measured as the voltage at a criterion time of 65 ms after stimulus onset (i.e., as the voltage difference from zeroed pre-stimulus baseline). Oscillatory potentials (OPs) were extracted using a Blackman filter with center frequency of 152.5 Hz and −3 dB points at 93 and 212 Hz. Their amplitudes were measured as root mean square (RMS) over the post-stimulus epoch of 15 to 55 ms and were referenced to a noise window between 100 and 140 ms after stimulus onset. 

R/B Long-Duration Flash

The longer duration R/B flash stimulus (560-cd/m² luminance and 200-ms duration) had a strength of 112 cd·s/m² and CIE coordinates of x = 0.692 and y = 0.305, and it was presented with a 2-second ISI on a 30-cd/m² blue background. Each of the two stored records was an average of 10 sweeps acquired with bandpass filter settings of 0.3 to 300 Hz at 1000-Hz sampling rate (1024 samples total length). The a-wave, b-wave, and ON PhNR were measured in the same way as for the R/B brief flash. The d-wave latency was measured as the time from stimulus offset to the start of the steep rise, and the d-wave implicit time was measured at the voltage maximum between 200 and 300 ms. The d-wave amplitude was measured as the voltage difference between its peak and the voltage value at its onset. The OFF PhNR was measured at a criterion time of 280 ms after stimulus onset (i.e., 80 ms after stimulus offset). OP RMS amplitudes were measured between 20 and 60 ms for those associated with the b-wave (ON OPs) and between 230 and 270 ms for those associated with the d-wave (OFF OPs), each referenced to a noise window measured between 100 and 140 ms. 

W/W Long-Duration Flash

The longer duration W/W flash stimulus (560-cd/m² luminance and 200-ms duration) had a strength of 112 cd·s/m² and CIE coordinates of x = 0.293 and y = 0.320, and it was presented with a 2-second ISI on a 30-cd/m² white background (CIE: x = 0.462, y = 0.406). Each of the two stored records was an average of 10 sweeps acquired with bandpass filter settings of 0.3 to 300 Hz at a 1000-Hz sampling rate (1024 samples total length). All response features were measured in the same way as for the R/B long flash, except that the OP RMS amplitudes were measured between 15 and 55 ms for those associated with the b-wave and between 215 and 255 ms for those associated with the d-wave. 

Flicker Series

All individual flashes within the W/W flicker series had a strength of 2.5 cd·s/m² and CIE coordinates of x = 0.293 and y = 0.320, and they were presented on the same 30-cd/m² white background as the long-duration W/W stimulus. All flicker responses were acquired using bandpass filter settings of 0.3 to 300 Hz and sampled at 1000 Hz. Flickering stimuli were presented at 5, 10, 20, 30.3, 40, and 50 Hz, in ascending order, and two records, each the average of 10 sweeps, were stored for the response at each frequency, then averaged for subsequent analysis. The amplitude of each flicker response was measured in two different ways: (1) as the average trough-to-peak voltage difference (i.e., in the time domain), and (2) using the fast Fourier transform (FFT) analysis available in the LKC Technologies software to measure the amplitude of the fundamental frequency response (i.e., in the frequency domain). 

Dark-Adapted Intensity Series

After all light-adapted ERG recordings were completed, isoflurane inhalation anesthesia commenced (see above), along with a period of 20 minutes of dark adaptation. Dark-adapted (scotopic) ERG responses were recorded simultaneously from both eyes using the Burian–Allen electrodes in their typical configuration for bipolar signal derivation and differential amplification (i.e., the active corneal ring electrodes were each referenced to their ipsilateral speculum). All stimuli for dark-adapted ERGs were broad-spectrum (“white”) flashes ranging from −3.6 to 2.4 log cd·s/m² (increasing in 4-dB steps until 0.4 log cd·s/m², then increasing in 5-dB steps thereafter). Signals were acquired at a sampling rate of 1000 Hz with bandpass filter settings of 0.3 to 300 Hz (−3.6 to −1.6 log cd·s/m²) or 0.3 to 500 Hz (−1.2 to 2.4 log cd·s/m²). A single record was stored for each eye at each flash strength, which was the average of either three sweeps (for stimuli ranging from −3.6 to 2.5 cd·s/m²) or two sweeps (for 0.4 log cd·s/m²), or it was a single sweep (for stimuli ranging from 0.9 to 2.4 log cd·s/m²), with ISI increasing from 2 seconds up to 2 minutes as stimulus strength increased. The scotopic a-wave amplitude was measured as the minimum voltage up to 50 ms after the stimulus (with a criterion minimum amplitude of −2 µV), and the a-wave implicit time was measured as the time from the flash stimulus to that trough. The scotopic b-wave amplitude was measured trough-to-peak as the voltage difference between the a-wave trough and the peak observed up to 150 ms after the stimulus with a criterion minimum amplitude of 2 µV; the scotopic b-wave implicit time was the time from stimulus to the b-wave peak. Scotopic OP RMS amplitudes were measured over the epoch 20 to 60 ms and referenced to a noise window between 130 and 170 ms after the stimulus. Naka–Rushton functions^36,37 were fit (least-squares regression) to intensity–response data of all 12 EG eyes as a group and to all 12 control eyes as a group for further analysis of the a-wave, b-wave, and OP amplitude results. 

Statistical Analysis

Statistical analysis was done using Prism 9 (GraphPad Software, San Diego, CA). Two-way, repeated-measures analysis of variance (RM-ANOVA) was used to test the significance of observed differences between the EG and control groups with matching for animal and eye as either stimulus strength (in the case of dark-adapted/scotopic ERG parameters) or flicker rate varied. In a few eyes (5/24), the scotopic b-wave amplitude did not achieve the criterion minimum amplitude of 2 µV at the lowest flash strength; thus, a mixed-effects model was used instead of ANOVA to evaluate the amplitude and implicit time. Šídák's multiple-comparisons post hoc test was used to compare EG versus control eyes at each flash strength or at each flicker rate. An extra sum-of-squares F-test was used to compare EG versus control eyes for each of the three free parameters of the Naka–Rushton functions fit to the dark-adapted a-wave, b-wave, and OP amplitude series. Similarly, two-way RM-ANOVA was used to test differences between the EG and control eyes for light-adapted (photopic) single-flash responses while matched for animal and eye, with chromatic contrast (R/B vs. W/W) as a second independent variable. Again, Šídák's multiple-comparisons post hoc test was used to compare EG versus control eyes within and across each condition. Pearson correlation analysis was applied to test the significance of observed structure–function relationships. Last, for each parameter used to characterize the status of EG eyes at the time the ERG study was conducted, a two-way RM-ANOVA matched for animal was used with Dunnett's multiple-comparisons post hoc tests to test differences between the EG and control eyes and differences within each group over time (compared to pre-laser baseline average values). 

Results

Figure 2 shows results for an array of parameters used to characterize the degree of glaucomatous damage at the time the full-field flash ERG study was conducted. On the day of the ERG study, IOP in EG eyes was elevated by 7.1 ± 7.5 mmHg, on average, above the pre-laser baseline measurements from the same group of eyes, representing an increase of 54% ± 58% (Figure 2A and rightmost columns of Table 1), whereas IOP in the group of control eyes had decreased by 3.1 ± 3.2 mmHg, on average, compared with their baseline values. 

Figure 2.

 

Characterization of experimental glaucoma (EG) stage at the time the ERG study was conducted. (A) Each box plot represents the median, interquartile range, and extremes of the distribution of parameter values among the 12 eyes in each group. Checkered boxes indicate baseline average values; unfilled boxes, values at the time of the ERG study. IOP was measured at the start of the ERG study session. (B) Optic nerve head (ONH) rim tissue thickness MRW parameter measured by OCT (^†7.3 ± 2.3 days earlier). (C) Circumpapillary retinal nerve fiber layer thickness (RNFLT) measured by OCT (^†7.3 ± 2.3 days earlier). (D) Circumpapillary RNFLT derived by scanning laser polarimetry (SLP; ^†7.3 ± 2.3 days earlier). (E) ONH tissue blood flow measured by the laser speckle flowgraphy (LSFG) mean blur rate (MBR) parameter (^†7.3 ± 2.3 days earlier). (F) Average amplitude of the multifocal ERG high-frequency component (mfERG HFC), measured during the same session just prior to the full-field flash ERGs. ^*P < 0.05; ^**P < 0.01; ^***P < 0.001; ^****P < 0.0001 (Šídák's multiple-comparisons post hoc test). ns, not significant.

Measurements of glaucomatous structural damage were obtained at the imaging session preceding the ERG study session, which was 7.3 ± 2.3 days (range, 4–14 days) earlier. Onset of significant thinning of the optic nerve head rim tissue, measured using the OCT MRW, was the inclusion criterion for this ERG study. Figure 2B shows that MRW had decreased by an average of 28% ± 10% (range, −14% to −45%) in EG eyes compared to their baseline average values. Consistent with prior reports on separate cohorts, loss of circumpapillary RNFL thickness measured by OCT (−12% ± 7%) (Fig. 2C) was relatively less than the loss of MRW.^32,38 Also similar to previous reports on separate, larger cohorts, RNFL retardance measured by scanning laser polarimetry exhibited slightly worse loss in EG eyes than did RNFL thickness (Fig. 2D),^34,35 but ONH tissue blood flow measured by LSFG did not manifest any decline at this relatively early stage of EG (Fig. 2E).³³ There was also a significant loss of mfERG high-frequency component amplitudes in EG eyes on the day of the full-field flash ERG recordings (Fig. 2F), consistent with previous findings.^11,31,35 Thus, this cohort of 12 NHPs had established EG, spanning a range from early through moderate glaucomatous damage, at the time the full-field ERG study was conducted. 

Light-Adapted Single Flash ERG Responses

Figure 1A shows the results for a single representative subject, and Figure 3 and Tables 2, 3, and 4 show the aggregate results for the light-adapted (photopic) single full-field flash ERG responses to the brief R/B stimulus, the 200-ms-long R/B stimulus, and the 200-ms-long W/W stimulus conditions, respectively. Among the major response features (parameters) measured for the brief R/B stimulus, only the PhNR amplitude was reduced (less negative) in EG eyes compared to control eyes (Table 2, Fig. 3A). Similarly, for the long R/B flash response (Table 3), only the PhNR amplitude was reduced, which was true for both the ON PhNR (following the b-wave) (Fig. 3B) and the OFF PhNR (following the d-wave) (Fig. 3C). For the long W/W flash response, both the ON and OFF PhNR amplitudes were reduced (Table 4, Figs. 3B, 3C), and there was also a significant delay of the d-wave implicit time in EG eyes (Fig. 3F). This latter finding was selective insofar as there was no effect of EG on the long R/B d-wave implicit time nor on the d-wave latency for either stimulus condition (P = 0.31) (Fig. 3E). The altered dynamics of the d-wave in EG are also worth considering in the context of the dramatic differences of d-wave timing between stimulus conditions, with the R/B decrement response being delayed substantially (10–20 ms for both latency, P < 0.0001, and implicit time, P < 0.0001) relative to the W/W decrement response, irrespective of EG. In contrast, the long R/B b-wave implicit time was only 1 to 2 ms delayed relative to the W/W b-wave (P = 0.06, data not shown). 

Figure 3.

 

Light-adapted (photopic) full-field flash ERG results. Bars indicate group mean (±SEM) for EG eyes (red bars, n = 12) and control eyes (blue bars, n = 12). (A) PhNR amplitude for the brief R/B stimulus condition. ^**P < 0.01 (paired t-test). (B, C) ON PhNR amplitude (B) and OFF PhNR amplitude (C) for the long-duration R/B and W/W stimulus conditions. ^*P < 0.05; ^**P < 0.01, Šídák's multiple comparisons post-hoc test. (D–F) The d-wave amplitude (D), d-wave latency (E), and d-wave implicit time (F) for long-duration R/B and W/W stimulus conditions. ^*P < 0.05 (Šídák's test). (G) Oscillatory potential (OP) root mean square (RMS) amplitude for the brief R/B stimulus condition. (H, I) ON OP RMS amplitude (H) and OFF OP RMS amplitude (I) for the long-duration R/B and W/W stimulus conditions. ns, not significant.

Table 2.

 

Light-Adapted (Photopic) Brief R/B Flash ERG Results (N = 12)

 

Table 3.

 

Light-Adapted (Photopic) Long R/B Flash ERG Results (N = 12)

 

Table 4.

 

Light-Adapted (Photopic) Long W/W Flash ERG Results (N = 12)

 

There were other notable differences between the two long-duration flash conditions (R/B vs. W/W), including a significant effect on both the ON PhNR amplitude (P < 0.0001) and the OFF PhNR amplitude (P = 0.008), as well as the ON OPs (P = 0.0004) (Fig. 3H) and OFF OPs (P = 0.0002) (Fig. 3I). However, unlike the ON and OFF PhNR amplitudes for both R/B and W/W conditions and the W/W d-wave implicit time, EG had no significant effect on either the ON OPs or the OFF OPs for either R/B or W/W conditions. 

Light-Adapted Flicker ERG Responses

Although frequency had a very strong influence, as expected, on the peak-to-trough amplitude (P < 0.0001) (Fig. 4A), implicit time (P < 0.0001) (Fig. 4B), and fundamental frequency amplitude (P < 0.0001) (Fig. 4C) of the flicker ERG response, EG had a significant overall effect only on the implicit time (P = 0.006), which was delayed by 0.44 ms, on average, compared to control eyes (Fig. 4B). However, this short implicit time delay for the flicker ERG response was not significant for any of the individual stimulus frequencies when evaluated by Šídák's multiple-comparisons post hoc test (Fig. 4B, Table 5). EG had no significant effect, overall, on the flicker response peak-to-trough (P = 0.44) or fundamental response (P = 0.39) amplitudes. There was also no significant effect of EG on flicker response harmonics (2f, 3f, 4f, 5f, and 6f) at any of the other stimulus frequencies tested (P = 0.95, P = 0.77, P = 0.40, P = 0.75 and P = 0.75, respectively, for 10 Hz, 20 Hz, 30.3 Hz, 40 Hz, and 50 Hz). However, Šídák's post hoc tests revealed that, for the 5-Hz condition only, peak-to-trough amplitude was significantly larger (P = 0.02) (Fig. 4A), and the fundamental frequency component amplitude (P = 0.0008) (Fig. 4C) was significantly smaller in EG eyes. 

Figure 4.

 

Light-adapted flicker ERG results. Symbols indicate group mean (±SEM, hashmarks) for EG eyes (red circles, n = 12) and control eyes (blue circles, n = 12); hashmarks are not shown if the SEM was smaller than the size of the symbol. (A, B) Peak-to-trough amplitude (A) and implicit time (B) of the ERG flicker response to each frequency tested. (C) Amplitude of the ERG response fundamental frequency versus stimulus frequency. ^*P < 0.05, ^***P < 0.001 (Šídák's multiple-comparisons post hoc test); all other pairwise differences were not significant.

Table 5.

 

Light-Adapted (Photopic) Flicker ERG Results (N = 12)

 

Dark-Adapted ERG Responses

As expected, we found a very strong influence of stimulus strength on dark-adapted (scotopic) ERG responses (P < 0.0001 for a-wave, b-wave, and OP amplitudes); however, there was no significant effect of EG on any of these scotopic ERG parameters (a-wave, P = 0.29, Fig. 5A; b-wave, P = 0.82, Fig. 5B; OPs, P = 0.75, Fig. 5C). There were also no significant differences found by Šídák's post hoc test between EG andcontrol eyes for any of the individual stimulus intensities. Similarly, there were no significant differences found for the maximum response amplitude (R_max) or the sensitivity (k) parameters of the Naka–Rushton functions fit to the scotopic a-wave, b-wave, and OP intensity-response data (Table 6, Fig. 5). 

Figure 5.

 

Dark-adapted (scotopic) ERG results for response amplitude versus stimulus strength. Symbols indicate group mean (±SEM, hashmarks) for EG eyes (red circles, n = 12) and control eyes (blue circles, n = 12). Curves represent the best fit of the Naka–Rushton intensity–response function to the aggregate data for each group. (A) a-Wave amplitude. (B) b-Wave amplitude. (C) OP amplitude.

Table 6.

 

Dark-Adapted (Scotopic) ERG Results: Parameters of Naka–Rushton Functions (N = 12)

 

Structure–Function Relationships and Diagnostic Sensitivity

The PhNR amplitude for each stimulus condition was compared to measures of glaucomatous structural damage as shown in Figure 6. For both ONH rim tissue thinning (measured by change from baseline average of the MRW parameter) and circumpapillary RNFL thickness, the strongest structure–function correlation was found for the brief R/B PhNR (Table 7, Fig. 6). There was no significant correlation between the PhNR amplitude of any stimulus condition and the IOP measured at the start of the ERG test session (Supplemental Table S1). 

Figure 6.

 

Structure–function relationships. Scatterplots show relationship between PhNR amplitude and the ONH rim tissue thickness parameter MRW (top row) or the circumpapillary RNFL thickness parameter (bottom row), expressed as the percent change from their baseline average values (N = 24 eyes total, 12 EG and 12 control eyes). The solid line in each plot represents the best fit of ordinary least squares linear regression, and dashed curves represent the 95% confidence intervals. Regression results and equations are shown below each panel. The horizontal dotted line represents the lower limit or normal amplitude (least negative value of control eye group).

Table 7.

 

Structure–Function Correlations (N = 12)

 

Diagnostic sensitivity (the proportion of EG eyes flagged as having an abnormal PhNR amplitude), assessed at 100% specificity (i.e., for a criterion amplitude set at the lower limit or least negative value of the control eye group), was 75% (n = 9 EG eyes flagged) for the brief R/B PhNR, 75% (n = 9) for the long R/B ON PhNR, 17% (n = 2) for the long R/B OFF PhNR, 50% (n = 6) for the long W/W ON PhNR, and 25% (n = 3) for the long W/W OFF PhNR, as shown in Figure 6. 

Discussion

The results of this study indicate there would be no clear benefit of using a longer duration flash stimulus to separately measure the ON and OFF PhNR for clinical diagnosis or monitoring of early-to-moderate glaucoma. The PhNR of the brief R/B stimulus had the highest diagnostic sensitivity and the strongest correlation to common structural measures of glaucomatous damage (albeit 75% sensitivity was equal to that of the ON PhNR of the long R/B flash). It is also simpler to measure one rather than two parameters, so the diagnostic utility of a long flash approach should clearly outweigh that of the brief R/B flash before supplanting the latter for standard clinical testing. Further, the results of this study do not show any advantage of the OFF PhNR over the ON PhNR for either the R/B or W/W long flash stimulus conditions. The effect size of the ON PhNR was larger than that of the OFF PhNR under both conditions, and all four of those long flash PhNR measures had a smaller effect size than that of the brief R/B flash PhNR amplitude. Similarly, the diagnostic sensitivity of the OFF PhNR was lower than that of the ON PhNR for both R/B and W/W long flash conditions. 

A previous study of human glaucoma by Horn and colleagues¹⁶ found that the OFF PhNR to a long W/W flash had higher sensitivity and a stronger correlation to circumpapillary RNFL thickness than the ON PhNR. The discrepancy between their findings and ours may well be due to pathophysiological differences between human glaucoma and this NHP model of EG. One such difference could be due to the IOP at time of testing, considering that their clinical patients were under IOP control (all had IOP below 23 mmHg at time of testing), whereas the NHPs with EG in our study had elevated IOP chronically, including the time of ERG testing. However, in our study, there were no significant effects of IOP on the day of the ERG test. Another subtle difference between the two studies is that we used a fixed criterion time for measurement of PhNR amplitude, while Horn et al. used a combination approach wherein PhNR amplitudes were measured at the trough within a criterion time window. In our study, we report results for the best performing criterion times, but we also found that the performance of these selected criterion times was superior to amplitude measurements of the PhNR based on peak to trough and baseline to trough for all stimulus conditions. Another potential explanation for differing results could be related to the severity stage or range of glaucomatous damage included in each cohort. The range of glaucomatous damage in our cohort of NHPs with EG was early to moderate; that is, loss of RNFL thickness from baseline values ranged from 4% to 26%, which would likely correspond to a range of 20% to 40% loss of RGC axons from the optic nerve based on previous histological studies using the same model.^32,39 The OCT measurement of RNFL thickness in our study had similar sensitivity (83%) to the brief R/B flash PhNR and the long R/B flash ON PhNR for discriminating between EG eyes and control eyes over this range. In contrast, the RNFL thickness measure in the study by Horn et al.¹⁶ was vastly more sensitive than any of the ERG parameters and showed more extensive loss in their study cohort (see their Fig. 4) compared to ours. These findings together suggest that the asymmetry between OFF and ON pathway dysfunction may depend on the stage of damage, manifesting perhaps only as a later-stage phenomenon and thus representing relative resilience of the ON pathway rather than greater susceptibility of the OFF pathway to early glaucoma per se. The important study by Horn et al.¹⁶ did not include a brief flash stimulus for comparison, so it is not clear whether that might have provided superior performance to the OFF PhNR in their study cohort. 

The same group of investigators conducted a separate study using full-field flicker stimuli that had a sawtooth temporal profile to emphasize either the ON (increment) or the OFF (decrement) response.²² In that study, they found that the late negative component (analogous to the PhNR) of the ON sawtooth flicker response was not effective for detecting glaucoma, whereas the late negative component of the OFF response was, and yet even better performance was obtained when the ON and OFF responses were added together.²² This suggests that the superior performance of the brief flash PhNR is due to the combination of ON and OFF late negative components providing the most informative signal for detecting glaucomatous functional damage, consistent with our findings in this study. 

A recent clinical study by Kong et al.²⁴ used a handheld Ganzfeld ERG system to evaluate a wide array of photopic stimulus conditions for their ability to detect loss of function in glaucoma, including brief W/W and R/B flashes, long flashes with strong contrast (red flashes on green background), 30-Hz brief-pulse flicker, and sinusoidal flicker from 50 to 0.3 Hz. Kong et al.²⁴ reported significant differences between the glaucoma and healthy control eye groups only for sinusoidal flicker at rates above 10 Hz. This finding was therefore interpreted as being among the earliest detectable ERG abnormalities in glaucoma and also as being consistent with selective vulnerability of the OFF pathway in glaucoma.²⁴ The latter inference was drawn using support from previous work by Kondo and Sieving^20,21 in which postsynaptic responses (beyond cone photoreceptors) were blocked using glutamate analogs to separate cone, ON pathway, and OFF pathway contributions to the flicker ERG. However, that work by Kondo and Sieving^20,21 demonstrated that the ON and OFF pathways contribute equally to the amplitude of the 32-Hz flicker ERG regardless of the temporal profile of flicker (e.g., sinusoidal vs. square wave).²¹ In addition, the authors showed that the OFF pathway has a greater contribution than the ON pathway to the sinusoidal flicker ERG fundamental frequency component at frequencies below 32 Hz, with the ON pathway contributing more than the OFF only at 50 Hz and higher.²⁰ Thus, the observation by Kong et al.²⁴ that amplitudes of the fundamental frequency response to sinusoidal flicker was reduced in glaucoma only above 10 Hz is not consistent with the selective loss of the OFF pathway. However, vector modeling of phase relationships, as done by Kondo and Sieving,^20,21 might still reveal differences between ON and OFF pathway contributions to the flicker ERG in glaucoma. Using brief-pulse flicker stimuli in our study, we did not find clinically meaningful differences between glaucomatous and control eyes at any frequency other than 5 Hz. The peak-to-trough amplitude of the 5-Hz flicker response was larger in EG eyes than controls, on average, which may be consistent with the tendency toward EG eyes having slightly larger a-wave and b-wave amplitudes for single brief and long flash stimuli (Tables 2–4). In contrast, the fundamental frequency response was reduced in EG eyes compared to control eyes, which may be more reflective of reduced PhNR-like contributions at this relatively slow flicker rate. At higher frequencies, it is unlikely that the slow negative components underlying the PhNR manifest to influence the flicker ERG response, which is likely dominated by ON and OFF bipolar cells, particularly at 30 Hz and higher.^20,21 Furthermore, as in other studies, lower flicker rates produced responses with relatively large contributions from higher frequency harmonics and more rapid flicker responses were dominated by the fundamental component. 

In the context of interpreting the flicker ERG results of our study, it is also worth noting that we found no abnormalities in EG eyes among the dark-adapted (scotopic) full-field flash ERG responses. This suggests that, during early-to-moderate NHP EG, the rod photoreceptors (a-wave), rod bipolar cells (b-wave, perhaps also Müller glia), and inner plexiform layer circuitry of the rod pathway (OPs) remain intact. The conditions under which our dark-adapted ERG recordings were obtained—in particular, under isoflurane anesthesia and using bipolar signal derivation—are not well suited for detecting loss of RGC function. Thus, our results should not be interpreted as an indication that the dark-adapted (scotopic) ERG is completely normal in glaucoma; indeed, it is well documented that the scotopic threshold response (which we did not record in this study) is altered, often selectively, by loss of RGC function and optic nerve injury.^40–43 It is also important to consider an additional aspect of the anesthesia regimen used for this study, which both differs from clinical testing (when sedation is used only rarely) and may also impact results, as three of the 12 animals received telazol instead of ketamine for the light-adapted (photopic) portion of testing (see Methods). However, no significant differences were found for any ERG response parameter in control eyes of animals that received ketamine versus those that received telazol. 

Another recent study, by Norcia et al.,²⁸ used an elegant approach for analysis of steady-state visual evoked cortical responses to sawtooth increment and decrement stimuli favoring ON and OFF visual pathways, respectively. Their results confirmed that cortical responses to decrements are normally larger and faster than responses to increments in healthy persons, but in the subgroup of eyes with moderate-to-advanced glaucoma (worse than −6 dB visual field mean deviation), decrement responses were statistically significantly reduced and delayed and the normally strong ON–OFF asymmetry was diminished.²⁸ Although these results were interpreted as representing preferential damage of the OFF pathway in glaucoma, there was a trend evident in the subgroup of eyes with mild glaucoma for greater loss of increment (ON pathway) responses than OFF responses. Thus, again, it may be that relative functional loss of ON and OFF pathways depends on the stage of glaucoma severity with relatively worse OFF pathway function occurring only at later stages. 

The only evidence of worse OFF pathway function we found in our study was a delay of the d-wave implicit time in EG eyes, limited to only the W/W stimulus condition and without any delay of the d-wave latency for either R/B or W/W long flash conditions. This finding represents changes in the response dynamics around the d-wave peak, rather than during its more abrupt initial rise from baseline, and may implicate the OFF OPs, which likely include the OFF i-wave as we used greater flash strength and background luminance relative to some other studies. It should be noted that the flash strength used was the nominal standard recommended by ISCEV and used in the default protocols of the LKC Technologies BigShot system; other studies may use lower strength flashes for increased patient comfort and reduced interference by blink and photomyoclonic response artifacts, which are much less intrusive when recorded under anesthesia as in this study. However, when the OFF OPs were isolated, we did not find any difference in their ensemble amplitude between EG and control eyes for either the R/B or the W/W stimulus, nor did we find any difference between EG and control eyes for the amplitude of the ON OPs for either stimulus. Thus, it may be that glaucomatous damage causes altered timing or amplitude of individual OFF OPs (and/or a larger than normal OFF i-wave)¹⁶ in response to the offset of an achromatic long-duration flash, resulting in a delayed d-wave peak. Further evaluation of this finding might be useful for questions about glaucoma pathophysiology but remains beyond the scope of this study, as the d-wave implicit time flagged only 50% of the EG eyes as being abnormal and, therefore, has lower clinical diagnostic utility than the best PhNR measures. 

The apparent delay of the W/W d-wave implicit time in glaucomatous eyes is also interesting in the context of the profoundly slower d-wave timing, both latency and implicit time, for the R/B stimulus condition compared to the W/W achromatic condition. The latter effect has been demonstrated qualitatively in previous reports, for example, in Figures 6A and 8B of Rangaswamy et al.¹² and in Figure 8 of Kremers et al.⁴⁴ It likely arises in the circuitry of the outer plexiform layer as differential adaptation states may influence the balance between strongly color-opponent and luminance signaling and their relative contributions to the ERG, particularly at stimulus offset/decrement,^44,45 consistent with other previous observations showing a strong influence of chromatic contrast (flash vs. background color) on the form of the d-wave peak and offset/decrement OPs.⁴⁶ For example, ERG responses to long-duration flashes shown in Figure 8 of the study by Sustar et al.¹⁸ do not show any delay of the d-wave for red flashes relative to white, blue, or green flashes of equal photopic strength when all were presented on a white background. Nor is there a clear indication from their study that less intense background adaptation results in profoundly slower d-waves under achromatic conditions (see their Fig. 6).¹⁸ It should be noted, however, that the study by Sustar et al.¹⁸ was with human participants but the effect may be more salient in macaques and depend on the L:M cone ratio.^12,44 Future studies should help determine the specific basis of this phenomenon, as well as its potential clinical significance. It is possible that the same underlying mechanisms may explain why ERG stimulus conditions with strong chromatic contrast (flash vs. background) produce a more robust PhNR and provide greater sensitivity to glaucomatous damage. 

In summary, the results of this study demonstrate no benefit of using longer duration flashes to separate onset and offset responses, as the PhNR to the standard brief R/B stimulus performed best for detecting and following functional loss in early-to-moderate stages of NHP EG. The d-wave was altered in glaucomatous eyes only for the W/W long flash and only as a brief implicit time delay, but this was a less sensitive parameter than the standard PhNR for the R/B brief flash or the ON PhNR for the R/B long flash. The marked prolongation of d-wave latency and implicit time for R/B compared to W/W long flashes warrants further exploration and explanation. 

Acknowledgments

The authors thank the Department of Comparative Medicine at Legacy Research Institute for providing helpful support and veterinary care during these experiments. 

Supported by grants from the National Institutes of Health (R01-EY030590, R01-EY030838) and by the Legacy Good Samaritan Foundation. 

Disclosure: M. Dunn, None; G. Cull: None; J. Reynaud, None; D. Jennings, None; T. Holthausen, None; A. Di Polo, None; B. Fortune, Heidelberg Engineering (F), Perfuse Therapeutics (F, C), Perceive Biotherapeutics (C), Amydis (C), Stoke Therapeutics (F, C)