Characterization of Retinal Injury Using ERG Measures Obtained with both Conventional and Multifocal Methods in Chronic Ocular Hypertensive Primates

Abstract

purpose. To characterize, using both conventional and multifocal electroretinogram (ERG) recordings as well as histologic measures, retinal injury in the chronic ocular hypertensive primate model for experimental glaucoma.

methods. Ocular hypertension was induced in the right eye of 7 cynomolgous monkeys, Macaca fascicularis, using laser injury to the aqueous outflow tissue at the anterior chamber angle. At 16 months after IOP elevation, ERG recordings were made from both eyes of all animals using both conventional and multifocal methods. After electrophysiological recording, animals were killed and retinal samples were radially sectioned for histologic analysis.

results. Histologic measures showed that ocular hypertensive injury was largely or completely limited to a loss of retinal ganglion cells (RGCs). The degree of RGC loss was similar in central and peripheral retina. Amplitudes of conventional ERG responses were mostly unaffected in eyes having severe loss of RGCs, a finding that is consistent with limited injury to photoreceptors, bipolar cells, and amacrine cells. Peaks in both the first- and second-order multifocal ERG responses were attenuated in ocular hypertensive eyes, and amplitude of these peaks was highly correlated with the density of surviving RGCs.

conclusions. The results are consistent with a conclusion that both first- and second-order components of the multifocal ERG response from the monkey reflect a significant contribution from activity in RGCs and may provide a useful measure for the clinical diagnosis and management of glaucoma.

The chronic ocular hypertensive (COHT) primate, originally described by Gaasterland and Kupfer, ¹ has been studied in a number of different laboratories as a model for glaucomatous injury to the retina and optic nerve. In this model, chronic elevation of intraocular pressure (IOP) results in retinopathy and neuropathy that reflect an apparently selective loss of retinal ganglion cells (RGCs) and their axons. ^{2 3} This pattern of injury to the retina and optic nerve, combined with the fact that the insult consists of an elevation of IOP, has led many investigators to refer to this model as“ experimental glaucoma.” Results of studies using electroretinogram (ERG) recordings for functional characterization of retinal injury have shown that pattern reversal, ^{4 5 6} scotopic negative, ³ and photopic negative ⁷ responses provide a measure of RGC injury in this model, whereas flash a-wave, b-wave, and oscillatory potential (OP) responses are relatively unaffected by hypertensive injury. ^{3 4 5 7}  

Multifocal ERG (mfERG) recordings from humans show evidence for contributions from activity of inner retinal cells including RGCs. ^{8 9 10 11 12} Glaucoma has also been associated with changes in the mfERG response, ^{13 14 15 16} though a recent report showed that measures of the local mfERG response were not well correlated with local glaucomatous sensitivity losses. ¹⁷ In rhesus macaque, evidence for a contribution from RGCs to mfERG responses has also been demonstrated. ¹⁸ These findings support the notion that mfERG recordings may have utility in the diagnosis and management of glaucomatous injury to RGCs. We report here our results from histologic measures of retinal injury as well as measures of ERG responses obtained using both conventional and multifocal techniques in the COHT primate. Some preliminary findings from this study have been published previously in a brief report. ¹⁹  

Methods

Animal Subjects

Seven young adult female cynomolgous monkeys, Macaca fascicularis, were used for this study. All experimental procedures as well as animal care and handling adhered to the guidelines outlined in the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. All experimental procedures were also reviewed and approved by an internal institutional review committee for animal use. 

Ocular Hypertension

IOP in the right eye of each animal was elevated using a procedure similar to that originally described by Gaasterland and Kupfer. ¹ Briefly, animals were anesthetized with an intramuscular injection of ketamine (15 mg/kg) and topical application of proparacaine (0.5%) in the right eye. Pupillary miosis was induced with instillation of 2.0% pilocarpine. After placement of a goniolens, energy from an argon laser (488 nm + 519 nm, model Novus 2000; Coherent, Inc., Palo Alto, CA) was directed through the anterior chamber to the trabecular meshwork. Individual burns were produced using focused laser spots of 1-W power, 50-μm diameter, and 0.5-seconds duration. An initial treatment consisted of 30 to 40 burns applied over the superior 180° of the chamber angle. Two weeks later, the inferior 180° of the trabecular meshwork was similarly treated. IOP was measured in both eyes of each animal at regular intervals under light anesthesia (intramuscular ketamine, 5 mg/kg) using a pneumotonometer (Digilab, Norwell, MA). 

Electrophysiology

ERG recordings were made from both eyes of each animal at approximately 16 months after induction of ocular hypertension. Animals were anesthetized with an intramuscular injection of ketamine (10 mg/kg) in combination with acepromazine (0.5 mg/kg) before placement of an endotracheal tube and intravenous catheter. Instillation of one drop of 1% tropicamide maintained pupil diameter at approximately 6 mm during the recording session. Animals were then positioned in a holder that used soft pressure points and a bite bar to stabilize the head. Neuromuscular block was induced with an intravenous bolus of norcuronium bromide (0.06 mg/kg) and maintained for the duration of the session with continuous infusion (0.04 mg/kg-hr.). Anesthesia was maintained with periodic intramuscular injection of ketamine (10 mg/kg). Animals were mechanically ventilated with 100% oxygen. Heart rate, rectal temperature, and expired Pco ₂ were continuously monitored. 

Corneal voltage was recorded using a bipolar contact lens electrode (Hansen Ophthalmic Laboratories, Iowa City, IA) configured such that the cornea was active and the speculum was reference. The indifferent electrode consisted of a subcutaneous needle located at the glabella. Corneal voltage signals were amplified using a Grass model 12 amplifier (Astro Med, West Warwick, RI) and digitized online at 2 kHz using a model DAS 1200 converter (Keithley Metrabyte/Asyst, Taunton, MA). Conventional ERG responses were elicited with stimuli of 10-μsec duration and an intensity of 0.09 J/flash (intensity setting of“ 1”) generated by a Grass model PS33 xenon flash tube. The stimulus was positioned at 10 cm from the cornea on the visual axis and subtended approximately 50° of visual angle centered on the fovea as shown in Figure 1A . Ambient background room illumination was approximately 0.05 footcandles during recording. For flash and OP responses, stimuli were delivered every 10 seconds. For flicker responses, 30-Hz stimulus trains of 512-msec duration were delivered at 1-second intervals. Averages of 10, 25, or 30 responses were used for analysis of flash, OP, and flicker responses, respectively. OP responses were isolated using analog bandpass filtering from 100 to 1000 Hz, whereas all other conventional responses were bandpass filtered from 3 to 1000 Hz. Line frequency noise was minimized with an analog 60-Hz notch filter. 

For multifocal recordings, stimuli were generated on a 21-inch Radius Intelicolor monitor (Radius, Inc., San Jose, CA) using VERIS 1 Scientific Software and video driver board (Electro Diagnostic Imaging, Inc., San Mateo, CA) and consisted of an array of 61 hexagonal elements of equal size as shown in Figure 1B . At the test distance of 30 cm, the entire array subtended approximately 50° of visual angle, each element having an angular subtense of approximately 5.5°. The intensity of each element was temporally modulated in a stepwise fashion at a frame rate of 67 Hz between 95 cd/m² (white) and 5 cd/m² (black) according to a binary m-sequence. ^{20 21} A steady background of 45 cd/m² surrounded the stimulus field to minimize contributions from light scatter. Responses to 2¹⁵ stimulus frames were recorded (m-sequence = 15), resulting in records of approximately 8-minutes duration for each eye. The corneal voltage signal was digitized at approximately 1000 Hz and bandpass filtered from 3 to 300 Hz in conjunction with 60-Hz notch filtering. 

Experimental Procedure for Electrophysiological Recording

The same recording sequence was used for all animals: (1) multifocal ERG OD, (2) conventional ERG OD, (3) multifocal ERG OS, and (4) conventional ERG OS. During recording, the contralateral eye was always occluded. After placement of the contact lens electrode, retinoscopy was performed to determine the best spherical equivalent lens power to make the retina optically conjugate to the stimulus monitor. A lens of this power (typically +3 to +5 diopter) was then placed at 1 cm anterior to the cornea. The stimulus monitor was then positioned at 30 cm anterior to the cornea such that the estimated visual axis projected to the center of the stimulus field. A series of multifocal recordings of approximately 2-minutes duration (m sequence = 13) was then used to adjust the stimulus position such that the fovea projected to the center of the central stimulus element and a clear amplitude maximum was elicited by the central stimulus element for both first- and second-order responses. The precision of this method for stimulus alignment was verified in several eyes by optically projecting the fundus onto the stimulus monitor and noting the location of the optic nerve head and macula. Stimulus alignment was also verified for each recording by observing the location of the optic nerve head projection (response minimum) in the first-order response trace array. After stimulus alignment, a multifocal recording of 8 minutes duration was made. The stimulus monitor was then covered with a light-tight shield, the corrective lens was removed, and the xenon flash stimulator was positioned. After a 5-minute period of dark adaptation, conventional recordings of the flash, OP, and flicker response were made in that order. 

Histology

Within several days after ERG recording, animals were deeply anesthetized with intramuscular injection of a combination of ketamine (15 mg/kg), xylazine (1 mg/kg), and acepromazine (0.2 mg/kg) and transcardially perfused with heparinized saline (37°C) followed by a mixture of paraformaldehyde (4%) and glutaraldehyde (0.1%) in phosphate buffer (pH 7.0–7.2, 37°C). Sutures were used to mark the 12 o’clock position on each eye before enucleation. After removal of the anterior segment and vitreous, the retina/choroid was removed and flat-mounted on a glass slide. Using a trephine and a template made from transparent film, 3 × 3-mm samples of the retina/choroid were obtained from eight regions, including one sample centered on the fovea (PF), three perimacular samples (1–3), and four samples from the far periphery (samples 4–7) as shown in Figure 2 . Tissue samples were then dehydrated and paraffin-embedded for sectioning. 

For samples 1 to 7, radial sections of 7-μm thickness were cut as indicated in Figure 2 . Six sections obtained at 50-μm intervals from each sample were then stained with hematoxylin & eosin (H&E). In each of these sample regions, the normal ganglion cell layer is comprised of a single layer of cells, and ganglion cell counts were obtained by manually counting and averaging cells in the ganglion cell layer of all six sections. The perifoveal ganglion cell layer is six to eight cells thick, and ganglion cell counts for the perifoveal sample (PF) were obtained differently. Beginning at the inferior border, horizontally oriented radial sections were made until the foveal pit was located. Three sections were then selected from the region of highest ganglion cell density between 0.5 and 0.7 mm from the center of the foveal pit. Ganglion cell counts in these three sections were made using a BIOQUANT imaging system (R&M Biometrics, Inc., Nashville, TN) and stereology software. 

Results

Ocular Hypertension

Mean IOP of all seven laser-treated (OD) eyes is summarized for the duration of the study in Figure 3 . Following soon after the second laser treatment (inferior 180° of chamber angle), there was a dramatic rise in IOP. The magnitude of this initial elevation, as well as the pressure over the remainder of the study, varied considerably among eyes, as indicated by the large SDs from the group means. IOPs of some eyes were maintained at relatively high levels, whereas in other eyes the IOP gradually dropped to some lower level at a rate that also varied among those eyes. When mean pressure elevation was expressed for each eye as the integral of IOP elevation above baseline (∼20 mm Hg) over the duration of the study, there was a strong correlation between mean IOP elevation and histologically measured loss of retinal ganglion cells (data not shown). That is, higher mean pressures were associated with greater loss of retinal ganglion cells. Pressure in the contralateral (normotensive OS) eyes showed only the normal degree of variation over the course of the study (data not shown). The relationship between the level of IOP elevation and the degree of retinal ganglion cell loss is beyond the scope of this report and will described in detail in a later article. 

Ocular hypertension resulted in optic neuropathy that was most severe in eyes with the highest mean IOP elevation. Fundus photographs from one animal, obtained approximately 2 months before killing, are shown in Figure 4 . Each panel is actually one of the two images from a stereo pair. The IOP for the hypertensive (OD) eye of this animal remained relatively high over the entire course of the study. The atrophic appearance of the nerve head in the hypertensive eye is readily apparent from a comparison with the normotensive (OS) eye. 

Histology

PF sections from each eye of the same animal whose fundus photographs are shown in Figure 4 are shown in Figure 5 . Comparison of the two sections shows that the RGC layer is approximately six to seven cells thick in the normotensive eye but is reduced to a single layer in the hypertensive eye. This is in contrast to the appearance and thickness of the inner nuclear layer, outer plexiform layer, and outer nuclear layer, which show no evidence of injury in the hypertensive eye. Measures of inner and outer nuclear layer thickness in sections from all sample regions of all eyes showed no significant difference between normotensive and hypertensive eyes (data not shown), evidence that supports the notion that hypertensive injury is largely or completely limited to a loss of RGCs. 

Figure 6 summarizes results of ganglion cell counts from all seven hypertensive eyes. For each sample region, ganglion cell counts from each hypertensive eye were normalized with respect to counts obtained from the same region in the contralateral eye. Mean normalized values for all eyes are shown with SE bars. When compared with the perifoveal sample region, no other region showed a significantly different (P < 0.05) level of RGC loss. 

Conventional ERG Responses

Conventional ERG responses obtained from the normotensive eye of one animal are illustrated in Figure 7 . For flash responses, amplitude of the a-wave and b-wave peak voltage was measured (Fig. 7A) . OP response amplitude was measured as the RMS voltage over a window extending from 10 to 75 msec after the stimulus (Fig. 7B) according to the equation:  

\[V_{\mathrm{RMS}}{=}\sqrt{(({\sum}v_{i}^{\mathrm{2}}\mathrm{){/}}\mathit{n})}\]

where V _RMS is RMS voltage, v _i is voltage at each timepoint over the range 10 to 75 msec, and n is number of voltage measures (n = 130). 

Amplitude of the 30-Hz flicker response was measured as the average peak-to-peak voltage of the last three cycles in the response train (Fig. 7C) . 

For each animal, amplitude measures from the hypertensive eye were normalized with respect to measures obtained from the contralateral normotensive eye (OD/OS). Results of previous recordings from normal monkeys indicated that comparing responses from both eyes recorded during the same recording sessions might yield less variability than comparing responses obtained from the same eye during different recording sessions (unpublished observations). These normalized amplitudes are plotted for each response measure as a function of normalized PF ganglion cell counts (OD/OS) in Figure 8 . In each panel, a single data point thus represents the amplitude measure and ganglion cell number obtained from an individual animal. Results from linear regression analysis of these plots show that a-wave amplitude was not correlated with histologic measures of ganglion cell loss/survival (Fig. 8A) , whereas amplitude measures for the b-wave (Fig. 8B) , OPs (Fig. 8C) , and flicker (Fig. 8D) responses showed a weak correlation. That is, a reduction in the number of surviving ganglion cells was associated with a reduction in b-wave, OP, and flicker response amplitude. However, the slopes for the linear regression plots were all less than 0.30, and eyes which lost most of their RGCs showed only a modest reduction in ERG response amplitude. Similar results were obtained when response amplitudes were plotted as a function of either total RGC counts (sum of counts in all eight sample regions), perimacular counts (sum of samples 1–3), or peripheral counts (sum of samples 4–7; data not shown). 

Multifocal ERG Responses

Amplitude measures were made for both first- and second-order multifocal ERG responses. Each of the 61 traces in Figure 9A represents the local first-order retinal response from a normotensive (OS) eye to the stimulus element in that location of the stimulus field. The central seven highlighted traces correspond to responses from the macular region of the retina (0° → 8° eccentricity), whereas the more peripheral traces correspond to responses from perimacular retina (8° → 25° eccentricity). Summation of traces in either of these groupings yields the macular and perimacular responses that are shown in Figure 9B . Note that central and peripheral response waveforms differ most markedly in the amplitude and kinetics of the later peaks. The macular response consists of five transients: an early negativity (N1) having a time-to-peak of approximately 17 msec, an early positivity (P1) having a time-to-peak of approximately 33 msec, a late negativity (N2) having a time-to-peak of approximately 50 msec, a late positivity (P2) having a time-to-peak of approximately 70 msec, and a late negativity (N3) having a time-to-peak of approximately 95 msec. Amplitude measures for the five peaks of the macular response were made as indicated. The first-order trace array for the hypertensive (OD) eye from the same animal is plotted in Figure 9C , whereas macular and perimacular responses from this eye are shown in Figure 9D . 

Second-order responses obtained from the same recording that generated the responses of Figures 9A and 9B are shown in Figures 10A and 10B . The macular response contains a prominent biphasic waveform: an early positivity (P, time-to-peak of approximately 45 msec) and a later negativity (N, time-to-peak of approximately 90 msec). This biphasic component is relatively much smaller in the perimacular response. A single peak-to-peak amplitude measure (PN) of the second-order macular response was made as indicated. Second-order responses from the hypertensive (OD) eye of the same animal are shown in Figures 10C and 10D . 

Responses from Figures 9B and 9D and 10B and 10D are replotted in Figure 11 , where heavy traces represent responses from the normotensive (OS) eye and light traces are from the hypertensive (OD) eye. Histologic analysis showed that few ganglion cells remained in the hypertensive eye of this animal (∼10% of the density measured in the normotensive eye), whereas there was no apparent loss of cells in other retinal layers. A comparison of first-order macular responses from the two eyes shows that although the amplitude of N1, N1P1, and P1N2 is greater for the hypertensive eye, the amplitude of N2P2 is smaller. Comparison PN amplitudes for the second-order responses shows an even greater relative reduction in the hypertensive eye. Hypertensive injury was associated with relatively less effect on either first- or second-order perimacular responses. Similar effects on macular mfERG responses were seen in all hypertensive eyes that had severe RGC loss. 

First- and second-order macular responses from Figure 11 are replotted in Figure 12 , trace A, whereas responses from both eyes of the other six animals are plotted as traces B through G. The traces in this figure are ordered from greatest (A) to lowest (G) RGC loss in the hypertensive eye (expressed as normalized RGC density measures, OD/OS) with values equal to A, 0.11; B, 0.15; C, 0.21; D, 0.69; E, 0.92; F, 0.96; and G, 1.41. Note that, for animals with severe hypertensive injury (A through C), later peaks of the first-order response (left column) are smaller in the hypertensive eye, whereas, for all animals, amplitude of peaks N1 and N1P1 is greater in the hypertensive eye. Furthermore, amplitudes of all first-order peaks are larger in hypertensive eyes that had little or no RGC loss (E through G). This may be explained if recordings from normotensive (OS) eyes produced generally smaller first-order responses, whereas hypertensive injury is associated with a selective attenuation of later response peaks. Inspection of second-order response traces (right column) shows that, when compared with the normotensive eye, PN amplitude is greatly reduced in eyes with severe hypertensive RGC loss (A through C), whereas there is a tendency for PN amplitude to be somewhat greater in hypertensive eyes with little or no RGC loss (F, G). These observations are illustrated in Figure 13 , where normalized macular mfERG response amplitude is plotted as a function of normalized perifoveal ganglion cell density for each response measure in all animals. For the first-order response, amplitude of the three late peaks (P1N2, N2P2, P2N3) is strongly correlated with histologic measures of ganglion cell survival (Figs. 13C 13D 13E) . Amplitude of these late peaks decreases with decreasing numbers of surviving ganglion cells in the hypertensive eyes. Amplitude of the early peaks (N1, N1P1) was unaffected in even the most severely injured eyes (Figs. 13A 3B) . Second-order response amplitude (PN) is also strongly correlated with the number of surviving ganglion cells (Fig. 13F) . Similar results were obtained for correlations with either total RGC counts (sum of all 8 retinal sample regions), perimacular counts (sum of three perimacular regions), or peripheral counts (sum of four peripheral regions; data not shown). 

Discussion

Ocular Hypertensive Retinal Injury

Earlier studies of the chronic ocular hypertensive monkey model have shown that retinal injury appears to be limited to a loss of RGCs. ^{2 3} Our own results from histologic analysis show that, even in eyes that had the most severe ganglion cell loss, there was no evidence for loss of any other retinal cell type. Furthermore, amplitudes of conventional ERG responses were relatively unaffected by hypertensive injury. Because these responses are believed to reflect primarily the activity of retinal cell types other than RGCs, ^{22 23} this finding is consistent with a conclusion that hypertensive injury has relatively little effect on the function of photoreceptors, bipolar cells, or amacrine cells. 

Our histologic analysis, however, could detect only gross manifestations of cellular injury such as loss of somas and neuropil. More subtle structural changes would presumably not be detected, and functional integrity may be compromised in cells that otherwise have a normal histologic appearance. It is also possible that some functionally distinct subset of non-RGC retinal cell was lost in hypertensive eyes but that this loss was not obvious because it makes up such a small fraction of all cells in the retinal layer. In addition, any loss of displaced amacrine cells, which comprise approximately 10% to 15% of cells in the macular RGC layer, ²⁴ would have been counted as lost RGCs in our analysis. The combination of histologic and electrophysiological results, however, are consistent with a conclusion that hypertensive injury is at least largely limited to a loss of RGCs. 

Results of ganglion cell counts from both central and peripheral retina showed that the degree of ganglion cell loss in hypertensive eyes was similar in all retinal regions. This finding is in agreement with earlier reports of histologic studies in ocular hypertensive monkeys ^{3 25} and is also consistent with the fact that macular mfERG response amplitude was similarly correlated with RGC counts from either central or peripheral retina. It is possible that regional differences in RGC loss might have been seen if the animals had been killed at an earlier time. That is, RGC loss may have occurred earlier or at a greater rate in some regions, but this difference was not apparent at 16 months after IOP elevation. Also, our analysis did not include a discrimination between morphologically distinct RGC subtypes and thus does not permit any conclusion with regard to regional subtype sensitivity to hypertensive injury. Furthermore, the variability inherent in our method for quantifying RGC loss does not allow us to exclude the presence of small regional differences. 

Electrophysiological Measures of RGC Injury

Flash, OP, and 30-Hz flicker ERG responses are thought to reflect primarily the activity of retinal cells other than RGCs. ^{22 23 26 27 28 29} These responses were chosen to provide a functional measure of injury to non-RGC retinal cells. The weak correlation that exists between amplitudes of these responses and histologic measures of RGC injury may be taken as evidence for two very different interpretations: (1) Ocular hypertension results in injury to retinal cells other than RGCs. The extent of this injury is much less than that seen for RGCs but is somewhat correlated with the degree of RGC loss. (2) These ERG responses reflect a small but significant contribution from RGC activity. A detailed discussion of these two possibilities is beyond the scope of this article. It is sufficient to note that the results of histologic analysis showed no evidence for injury to any retinal cell type other than RGCs and that either of these two interpretations for the conventional ERG results is consistent with this conclusion. 

The first-order mfERG responses obtained from normal eyes in this study are similar in waveform and kinetics to those obtained from humans under similar conditions. ^{21 30} These responses, however, are rather different in comparison to published responses obtained from rhesus macaque, ¹⁸ which are much more oscillatory in nature. Although it is possible that this may reflect a species difference, it is more likely the consequence of different recording conditions. First, the rhesus recordings were made using the contralateral cornea as the voltage reference. This method has been reported to enhance contributions of an “optic nerve head component” (ONHC), which gives the response a more oscillatory appearance. ⁹ Second, the responses in the present study were made using analog 60-Hz notch filtering to remove line noise. A comparison of responses recorded from a normal eye with (heavy traces) and without (light traces) notch filtering is shown in Figure 14 . Note that notch filtering eliminates late oscillatory components in the first-order macular response, whereas there is a moderate attenuation of N1, N1P1, and P1N2. Notch filtering also attenuates an early oscillation in the second-order response but has relatively less effect on peak PN. Thus, it may be assumed that notch filtering had a moderate effect to attenuate the amplitude of peaks that were measured in both first- and second-order responses of the present study. The attenuation of oscillatory components is more pronounced and, because these components have a presumed origin in activity of inner retinal cells, ^{9 18} it is possible that inner retinal contributions to the responses of the present study were reduced by notch filtering. 

Ocular hypertensive injury was associated with a relative reduction in amplitude for peaks in both the first- and second-order multifocal ERG response (see Figs. 11 12 ). These same response components are prominent in normal macular retina where RGC density is high and relatively smaller outside the macula where RGC density is low. Amplitude of these response components is highly correlated with the number (density) of surviving ganglion cells, with slopes for this relationship ranging from approximately 0.6 to 1.0 (see Fig. 13 ). Also, RGC loss is the only obvious histologic consequence of hypertensive retinal injury. Taken together, these observations are consistent with a conclusion that activity in RGCs makes a contribution to both the first- and second-order multifocal ERG responses in monkey. 

For the conventional flash response, a-wave amplitude was uncorrelated with RGC counts in hypertensive eyes, whereas the b-wave showed a very weak correlation (see Fig. 8 ). The early negative (N1) and positive (N1P1) peaks of the first-order multifocal ERG response were also uncorrelated with RGC counts. These results are consistent with a previous report that concluded that the early negative and positive peaks in the first-order mfERG response from humans appear to be driven by activity in the same retinal cells that drive the flash a-wave and b-wave. ³¹ Although N1 and N1P1 amplitudes were not correlated with RGC loss (or IOP history), there was a tendency for the amplitude of these peaks to be relatively greater in the hypertensive (OD) eye (see Figs. 12 13 ). This observation may be a consequence of the order for recording responses from the two eyes. Because OD was always recorded first and OS was recorded approximately 45 minutes later, the additional time under anesthesia might have resulted in a generalized decrease in OS response amplitude that is similar for all animals. For this reason, OS responses from Figure 12 were linearly scaled by a factor that made peak N1 amplitude equal for both eyes. Scaled OS responses have been replotted with the original OD responses in Figure 15 . Note that, except for linear scaling of OS responses, this figure is identical with Figure 12 . Also note that second-order OS responses have been scaled by the same factor used for the first-order response. A comparison of Figures 12 and 15 shows that scaling the OS response in this manner makes OD and OS responses more similar in animals that had little or no RGC loss and enhances the difference in animals having severe RGC loss. When amplitude measures were made using scaled OS responses for normalization and results replotted as in Figure 13 , the slopes for linear regression were unchanged but the correlation coefficients for P1N2, N2P2, P2N3, and PN increased (not shown). 

Multifocal ERG as a Measure for RGC Function

Perhaps the best evidence for an RGC-specific component in the mfERG is found in nasotemporal response asymmetries, which have properties consistent with their generation by impulse conduction in axons of the retinal nerve fiber layer. ^{8 9} This ONHC has been shown to be blocked by both TTX and NMDA (agents that block voltage-gated sodium channels and glutamatergic ligand-gated channels, respectively) in monkey ^{14 18} and has also been shown to be reduced in both glaucoma patients and glaucoma suspects. ¹³ Because TTX- and NMDA-sensitive mechanisms are relatively localized to inner retinal cells, this provides further support for an inner retinal origin of the ONHC. As discussed earlier, the recording conditions for the present study likely minimized any ONHC contribution. Using recording conditions similar to those used in the present study, we have previously identified a small OHNC in responses from normal monkey eyes (unpublished observations). We also found that it could be blocked by intravitreal application of TTX but that TTX had a much greater effect on other components of the first- and second-order response, including the peaks that were most affected by RGC loss in this study. 

It was recently reported ¹⁵ that patients with glaucoma show evidence for amplitude reduction in peaks of first- and second-order macular responses, which are similar to the findings of the present study. It has also been reported that branch retinal artery occlusion results in relatively selective reduction of late components in the first-order response and complete elimination of the second-order response in that retinal region corresponding to ischemic insult. ¹¹ Because branch artery occlusion is thought to result in ischemic insult, which is relatively localized to inner retinal elements, this finding is consistent with an inner retinal origin for these response components. More recently, however, changes in implicit times but not amplitude for peaks of the first-order response were seen in patients with glaucoma, ¹⁶ whereas, in another study, measures of either first- or second-order responses showed no correlation with visual sensitivity loss in subjects with glaucoma. ¹⁷  

Early studies using the COHT monkey model of glaucoma showed that pattern ERG (PERG) and pattern VEP (PVEP) amplitude were reduced in eyes having normal conventional flash ERG responses, ^{4 5 6} and similar results have been obtained in clinical glaucoma studies. Results of the present study as well as results from other work with animal models and human subjects are consistent with the notion that mfERG responses contain significant contributions from inner retinal cells including RGCs. The development of optimal stimulus and recording conditions as well as effective strategies for response component analysis can be expected to enhance the utility of the mfERG for assessment of function in the visual pathways. Evaluation of these methods in animal models is especially useful in this regard. 

 Submitted for publication December 20, 1999; revised August 10, 2000; accepted September 20, 2000.

 Commercial relationships policy: E.

 

Figure 1.

 

Stimulus and recording configuration for conventional (A) and multifocal (B) ERG are shown here for a left (OS) eye. In each case, the stimulus field subtends approximately 50° of visual angle centered on the visual axis. Relative position and projection of the fovea (*) and optic nerve head (**) are indicated. The shaded central seven stimulus elements in the multifocal stimulus field (B) correspond to macular retina (approximately 0° → 8°) while the surrounding 54 unshaded stimulus elements correspond to perimacular retina (approximately 8° → 25°).

Figure 2.

 

Location of retinal sample regions for histologic analysis of ocular hypertensive injury is shown here for a right (OD) eye. The retina/choroid was flat-mounted on a glass slide (RGC layer up) and samples were cut from eight locations with the orientation indicated. Each sample region was 3 mm × 3 mm. The perifoveal sample was centered on the fovea. Samples 1 to 3 were located on the horizontal and vertical meridians from 3.5 to 6.5 mm from the fovea. Samples 4 to 7 were located on the oblique meridians from 8.5 to 11.5 mm from the fovea. Sections were cut parallel to the heavy border as indicated for each sample.

Figure 3.

 

IOP history of all animals for the duration of the study. Each data point represents the mean IOP for the laser-treated (OD) eyes of all seven animals at that timepoint. Three baseline IOP measures were made before the first laser treatment. The zero timepoint corresponds to the IOP measure made 1 day before the second laser treatment. Note that the IOP response to laser treatment varied considerably for individual eyes as indicated by the large SDs (vertical bars) for mean pressure.

Figure 4.

 

Fundus photographs of the hypertensive OD (A) and normotensive OS (B) eyes from one animal. Each panel is one frame of a stereo pair. The photographs were taken approximately 2 months before sacrifice (approximately 13 months after IOP elevation). The pressure in the hypertensive eye of this animal remained relatively high over the duration of the study. The atrophic appearance of the hypertensive optic nerve head is readily apparent from a comparison with the contralateral normotensive nerve head.

Figure 5.

 

Perifoveal H&E stained sections from the normotensive OS (A) and hypertensive OD (B) eyes of the same animal whose fundus photographs are shown in Figure 4 . Each of the sections was cut orthogonal to the vertical meridian at 500 to 700 μm inferior to the center of the foveal pit as described in the text (see also Fig. 3 ). The field of the micrographs is centered on the vertical meridian. Note that the ganglion cell layer (GCL) in the normotensive eye is approximately six to seven cells thick, whereas the hypertensive eye has only a single sparsely populated row of cells. The thickness and appearance of the inner nuclear layer (INL), outer plexiform layer (OPL), and outer nuclear layer (ONL) are similar in the two eyes. The calibration bar in (B) is equal to 100 μm and applies to (A) and (B).

Figure 6.

 

Summary of RGC counts from the eight retinal sample regions in all seven animals. Counts of cells in the ganglion cell layer of each sample region were made as described in the methods section. For each animal, RGC counts in the hypertensive eye were normalized with respect to counts obtained from the same sample region in the contralateral normotensive eye (OD/OS). For each sample region, mean normalized counts for all animals are shown with SE bars. Note that there is no significant difference (P < 0.05) in the degree of RGC loss/survival for any sample region.

Figure 7.

 

Conventional ERG response measures. Traces represent averaged responses recorded from a normotensive eye as described in the methods. (A) Flash response. Stimulus delivered at 0 msec. Peak amplitude of a-wave (a) and b-wave (b) measured as indicated. Trace is average of 10 responses. (B) Oscillatory potential. Stimulus delivered at 0 msec. Response amplitude measured as RMS voltage from 10→ 75 msec after the stimulus. Trace is average of 25 responses. (C) Flicker response. Stimulus consists of 30-Hz train of 512 msec duration beginning at 0 msec. Average peak-to-peak amplitude of last three response cycles measured as indicated. Trace is average of 30 responses.

Figure 8.

 

Summary of correlation between conventional ERG response amplitude and counts for the number of surviving cells in the perifoveal (PF) RGC layer. Methods for measures of response amplitude and cell counts as described in text. ERG response amplitude and cell counts for the hypertensive eye are normalized, for each animal, with respect to values obtained for the normotensive eye (OD/OS). In each panel, the straight line represents the linear regression plot. (A) Correlation of a-wave amplitude with RGC counts: regression slope = +0.07, r = 0.18. (B) Correlation of b-wave amplitude with RGC counts: regression slope = +0.18, r = 0.71. (C) Correlation of OP RMS voltage with RGC counts: regression slope = +0.28, r = 0.90. (D) Correlation of flicker response amplitude with RGC counts: regression slope = +0.28, r = 0.76. Note that (C) has only six data points since OP responses were not recorded from one of the animals with moderate RGC loss.

Figure 9.

 

First order mfERG responses from both eyes of an animal with severe hypertensive injury. Recording made as described in Methods. (A) Normotensive (OS) eye. Trace array of 61 responses, each of which represents the local retinal response corresponding to the stimulus element at that location in the stimulus field. The central 7 traces (shaded) represent responses from macular retina extending to approximately 8° retinal eccentricity. The surrounding 54 traces (unshaded) represent responses from perimacular retina extending from approximately 8° → 25° retinal eccentricity. Calibration bars, 200 nV, 100 msec. Note that amplitude of individual traces is expressed in units of volts since each trace represents the response from retinal areas of the same size. (B) Macular and perimacular responses obtained by summation of either the central 7 or surrounding 54 response traces, respectively, from (A). Amplitude measures for the 5 peaks of the macular response were made as indicated. Calibration bars, 5 nV/deg² and 25 msec. Note that macular and perimacular response amplitude is expressed as response density (volts/unit retinal area) since these responses reflect retinal stimulus areas of different size. (C) Sixty-one response array obtained from the hypertensive (OD) eye of the same animal whose responses are shown in (A) and (B). Calibration as for (A). (D) Macular (top trace) and perimacular (bottom trace) responses obtained from (C). Calibration as for (B).

Figure 10.

 

Second order mfERG responses from the same recordings which produced the first-order responses shown in Figure 9 . (A) Normotensive (OS) eye. Sixty-one response trace array with responses from macular (shaded) and perimacular (unshaded) retina as indicated. Calibration bars, 200 nV, 100 msec. (B) Macular and perimacular response obtained by summation of responses from the macular and perimacular retina, respectively, in (A). A single peak-to-peak measure of second-order macular response amplitude was made as indicated. Calibration bars, 2.5 nV/deg² and 25 msec. (C) Sixty-one response array from hypertensive (OD) eye. Calibration as for (A). (D) Macular (top trace) and perimacular (bottom trace) responses obtained from (C). Calibration as for (B).

Figure 11.

 

Comparison of macular and perimacular responses from the hypertensive (OD, light traces) and normotensive (OS, heavy traces) eyes shown in Figures 9 and 10 . Histologic analysis showed that normalized (OD/OS) perifoveal RGC density in the hypertensive eye was 0.11. See text for description of responses. First order response calibration bars, 5 nV/deg² and 25 msec; second-order calibration bars, 2.5 nV/deg² and 25 msec.

Figure 12.

 

Comparison of first-order (left column) and second-order (right column) macular responses from normotensive (OS, heavy traces) and hypertensive (OD, light traces) eyes of all seven animals. RGC loss in hypertensive eyes was expressed as normalized (OD/OS) histologic measures of perifoveal RGC density and yielded values for each animal equal to: A, 0.11; B, 0.15; C, 0.21; D, 0.69; E, 0.92; F, 0.96; G, 1.41. Calibrations for all first-order responses, 5 nV/deg² and 50 msec. Calibrations for all second-order responses, 5.0 nV/deg² and 50 msec.

Figure 13.

 

Summary of the correlation between amplitude for both first- and second-order mfERG responses and counts for the number of surviving cells in the perifoveal (PF) RGC layer. Measures of response amplitude and cell counts as described in Methods. In each panel, response amplitude and RGC counts from the hypertensive eye are each normalized, for each animal, with respect to values obtained from the contralateral normotensive eye. Straight-line plots in each panel represent results of linear regression analysis. (A) First order peak N1: regression slope = −0.11, r = 0.27. (B) First order peak N1P1: regression slope = +0.11, r = 0.22. (C) First order peak P1N2: regression slope = +0.77, r = 0.80. (D) First order peak N2P2: regression slope = +1.04, r = 0.93. (E) First order peak P2N3: regression slope = +0.76, r = 0.64. (F) Second order peak PN: regression slope =+ 0.60, r = 0.92.

Figure 14.

 

First- (A) and second-order (B) macular (top traces) and perimacular (bottom traces) responses obtained from a normal eye either with (heavy traces) or without (light traces) use of the 60-Hz notch filter. Recording was first made with the notch filter active and immediately followed by a recording with the notch filter inactivated. Stimulus and recording conditions were, otherwise, identical with those described in the Methods section. Calibrations, 10 nV/deg² and 20 msec (A), and 5 nV/deg² and 20 msec (B).

Figure 15.

 

Responses from Figure 12 are replotted here. For each animal, the first- and second-order response from the normotensive (OS, heavy trace) eye has been linearly scaled by a factor which makes peak N1 amplitude equal to that obtained from the hypertensive (OD, light trace) eye. Responses to the hypertensive eye are identical with those plotted in Figure 12 . OS responses were linearly scaled by the following factors: A, 1.46; B, 1.44; C, 1.34; D, 1.47; E, 1.22; F, 1.34; G, 1.41. Calibrations are 5 nV/deg² and 50 msec (first-order responses), and 5.0 nV/deg² and 50 msec (second-order responses).

The authors are grateful for the expert assistance of James Burke, who performed the laser treatment on all animals in this study, Don Long, who provided assistance with animal preparation and anesthesia, and also for many technical discussions with Erich Sutter.