Refining Flash Visual Evoked Potential Analysis in Rats: A Novel Approach Using Bilateral Epidural Electrodes

Abstract

Purpose: Visual evoked potentials (VEPs) are electrical signals generated at the visual cortex following visual stimulation. Flash VEPs (fVEPs) are produced by global retinal stimulation and are considered an objective measure of the integrity of the entire visual pathway. However, fVEP measurements are highly sensitive to external variables, making relative comparisons of the fVEP waveforms between the two eyes in the same individual challenging.

Methods: We used the rodent non-arteritic anterior ischemic optic neuropathy (rNAION) model to induce unilateral ischemic optic neuropathy. The severity of optic disc edema was measured with spectral-domain optical coherence tomography, and visual acuity was measured using a virtual optokinetic system. We developed a procedure utilizing implanted bilateral epidural electrodes and derived a mathematical formula to accurately estimate functional differences between the optic nerves. Immunohistology was performed to quantify retinal ganglion cell (RGC) survival using stereology.

Results: Compared to subcutaneous methods, the new approach significantly improves the signal-to-noise ratio and is more repeatable when comparing the two eyes. The derived formula accounts for asymmetry in the afferent inputs to the visual cortex. Visual function calculated using the formula correlates strongly with other recognized metrics of visual function, including RGC survival and visual acuity.

Conclusions: We have developed a repeatable and accurate method to calculate the relative visual function of diseased optic nerves compared with a contralateral control eye.

Translational Relevance: Our novel method improves fVEP measurement sensitivity and accuracy in rodent preclinical trials, reducing the number of animals needed to achieve statistical significance.

Introduction

Visual evoked potentials (VEPs) measure the electrical signaling at the visual cortex in response to a visual stimulus.¹ VEPs provide a relatively objective assessment of the functional integrity of the visual pathway, from the retina to the visual cortex.² VEPs are useful in measuring visual function in individuals unable to communicate verbally,³ and it is extensively used in animals to assess diseases affecting the retina and optic nerve. VEP analysis is particularly useful in studying neuroprotective and neuroregenerative treatments for the optic nerve, as it evaluates the summation of neural activity rather than simple structural and histological changes.^4,5 

VEPs are useful in both clinical cases and the animal models of non-arteritic ischemic optic neuropathy (NAION). These models include the rodent model (rNAION)⁶ and the primate model (pNAION).⁷ In NAION, it is believed that capillary leakage results in a compartment syndrome with localized ischemia in the proximal portion of the optic nerve, which is followed by subsequent axon degeneration.⁸ In the current animal models, laser-induced superoxide radical induction results in capillary leakage within the optic nerve head.⁶ Both models closely mimic the human disease.⁹ 

Flash VEPs (fVEPs) are extensively used to measure visual function in rats.¹⁰ The early components of fVEPs (P1, N1, and P2) are good candidates for assessing pre-cortical input to the striate cortex.¹¹ Although the origins of P1 are not well understood, N1 and P2 correspond to primary excitatory and inhibitory peaks generated by the postsynaptic potentials of the geniculocortical afferents to the pyramidal cells in the striate cortex.^12,13 These early components are relatively independent of the rat's sleep–wake state compared with the late VEP components.¹² However, several challenges exist in acquiring precise quantitative data from VEP waveforms, especially when comparing waveform patterns between the two eyes of the same animal. 

A number of approaches are used to measure the fVEP, each with its unique set of challenges. Subdermal needles¹⁴ or epidermal cup electrodes^15,16 are convenient non-surgical approaches but are highly dependent on the electrode positioning and have poor signal-to-noise ratios (SNRs).¹¹ Additionally, variability in VEP amplitudes can result from differences in cortical anatomy,¹⁷ basal cortical activity,¹⁸ body temperature,¹⁹ and the type of anesthetic used.^16,20 Thus, these approaches lack the precision necessary to adequately measure visual function, often requiring larger groups of test subjects to obtain meaningful conclusions. 

To improve the SNR, transcranial electrodes, including epidural^13,21 and subdural electrodes, can be used.²⁰ However, species-dependent variability in optic nerve decussation^22,23 results in significant differences in VEP waveforms over the visual cortex depending on which eye is stimulated.^24,25 Mapping the topography of VEP responses at the visual cortex to ipsilateral and contralateral stimuli is complex, and the divide among ipsilateral, contralateral, and binocular zones in the cortex is not clearly delinated.²⁵ The asymmetric response to ipsilateral versus contralateral stimulation is masked when using surface electrodes, as the cutaneous, subcutaneous, and skeletal layers provide considerable impedance, leading to the suppression and averaging of the electrical responses of both visual cortices.²⁶ However, transcranial electrodes significantly increase the SNR by reducing tissue impedance and reveal the asymmetry of afferent inputs to each visual cortex. Improved localization through transcranial electrodes has been extensively explored in electroencephalograms, where subdural electrodes demonstrate significantly greater specificity in localizing epileptogenic areas compared to scalp electrodes.^27,28 

In this report, we describe a novel method for acquiring fVEP responses using two bilateral epidural screw electrodes, combined with a mathematical formula to estimate the visual function of a diseased eye relative to a healthy eye in the same animal. We also compare the visual function obtained from this method with that calculated from fVEPs using subcutaneous electrodes. This approach using bilateral epidural screws not only benefits from increased SNRs but also accounts for the asymmetry in optic nerve innervation to the dorsal lateral geniculate nuclei (dLGNs) to provide a more precise and repeatable measure of visual function. 

Methods

Animals

All animal protocols were approved by the University of Maryland Institutional Animal Care and Utilization Committee (IACUC) and adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. We used adult male pigmented Long-Evans rats (Charles River Laboratories, Wilmington, MA), with body weights ranging from 300 to 500 g. The timeline of the procedures is outlined in Figure 1. 

Figure 1.

 

Schematic diagram showing the timeline of procedures. One day after rNAION induction, rats underwent OCT to assess the severity of rNAION. Visual function was assessed at least 3 weeks after rNAION induction. Visual acuity was first measured with the OptoMotry system over 3 consecutive days. Following this, epidural screws were implanted, and VEPs were measured at least 3 days after surgery. Repeat VEP measurements were taken at least 3 days after the previous procedure. Rats were euthanized following the final VEP measurement, and tissues were collected for RGC stereology. *Exact VEP measurement dates varied, as long as they were performed at least 3 days after screw implantation or the previous VEP measurement. ^†Rats were euthanized after the final VEP measurement.

rNAION Induction

rNAION was induced as previously described.⁶ Briefly, animals were anesthetized with ketamine (80 mg/kg) and xylazine (4 mg/kg) (K/X). Corneas were topically anesthetized using 0.5% proparacaine, and pupils were dilated using 1% tropicamide and 2.5% phenylephrine. A plano-convex contact lens (Micro-R; Cantor & Nissel, Northamptonshire, UK) was placed on the cornea. Following intravenous injection of Rose Bengal (2.5 mM, 1 mL/kg), the optic disc was illuminated for 11 seconds using a 532-nm laser with a 500-µm spot size and 60-mW power. Retinas were evaluated 1 day after lesion induction using both direct slit-lamp examination and spectral-domain optical coherence tomography (SD-OCT) to assess lesion severity. For this study, all rNAION inductions were performed on the right eye, which is interchangeably referred to here as either “right eye” or the “rNAION-induced eye.” 

SD-OCT Examination

A clinically approved SD-OCT instrument (SPECTRALIS; Heidelberg Engineering, Heidelberg, Germany) was used to quantify the severity of optic nerve head (ONH) edema following induction.^29,30 We generated SD-OCT–based retinal cross-sectional images spanning the ONH. ONH edema was assessed as previously described, using the mean distance of the three largest contiguous values from either side of the inner nuclear layer at the disc.³⁰ Example images of the control ONH and progressive ONH edema following rNAION induction are shown in Figure 2. 

Figure 2.

 

(A–F) Representative SD-OCT B-scans of rat retinas of control (A, B) and rNAION-induced (C–F) eyes. (A) En face view of control retina. The ON is relatively dark and small. (B) Cross-section scan of the control retina. The intraretinal portion (indicated by the white bar) of the uninduced ONH is 318 ± 61 µm. (C) En face view of an rNAION-induced eye (cross-section shown in panel F). The ON is pale and enlarged, and the retinal veins are moderately dilated compared to the retinal veins in the control retina in panel A. (D–F) Progressively increased ONH edema in rNAION-induced eyes: mean ONH edema 634 ± 65 µm (D); mean ONH edema 713 ± 21 µm (E); mean ONH edema 822 ± 36 µm (F). The ONH diameters derived from the three largest contiguous values between the inner nuclear layers of either side of the eye were used to measure lesion severity. Ra, retinal artery.

Surgery

At least 3 weeks after rNAION induction, animals were anesthetized with an intraperitoneal (IP) injection of K/X (80/4 mg/kg) and given preemptive analgesia with subcutaneous (SC) buprenorphine. Following local lidocaine infiltration, a midline incision was made from the forehead to the occipital prominence to expose the skull. The periosteum was reflected, and the two craniotomy sites were identified at the approximate locations of the V1 visual cortices, 7 mm caudal to the bregma and 3 mm lateral to either side of the midline. Burr hole craniostomies were made with a 0.5-mm drill bit, with care taken not to penetrate the dura. Two clean, sterile stainless-steel screws (00-90 × 3/16 inch; Minitaps, Seattle, WA) were implanted through the skull up to the level of the dura. Screws were initially fixed in place with cyanoacrylate glue and then stabilized with dental cement. The skin incision was closed around the electrodes with 6-0 nylon suture. Animals were allowed to recover from the surgery for at least 3 days before recordings were performed. 

fVEP Measurements

All VEP measurements were conducted at least 3 days after epidural screw implantation, more than 3 weeks after rNAION induction. Animals were dark adapted for at least 4 hours prior to testing. Following anesthesia with IP K/X injection, the corneas were topically anesthetized with 0.5% proparacaine, and pupils were dilated with tropicamide and phenylephrine eye drops as described above. A stainless-steel ground electrode was inserted at the base of the tail. The reference electrode was placed at the tip of the snout. For VEP measurements using subcutaneous needle electrodes, the positive stainless-steel electrode was inserted at the midline above the occipital groove. For VEP measurements using implanted epidural screw electrodes, the electrodes were connected to the amplifier with a micro alligator clip. 

We used a Celeris device (Diagnosys LLC, Lowell, MA), which incorporates a light-emitting diode in a contact lens to stimulate the retinas and an integrated recording electrode. This approach allows for near-simultaneous electroretinography and VEP measurements. A flash visual stimulus of 1 cd·s/m² was delivered 50 times to each eye. Data were acquired at a sample rate of 2 kHz and for a duration of 300 ms after the flash stimulus. The resultant VEP waveforms were averaged for each test. 

In measurements using subcutaneous electrodes, two waveforms were obtained for each measurement, one per stimulation of each eye. In measurements with epidural screw electrodes, four waveforms were obtained for each measurement. These included waveforms obtained from the right epidural screw in response to right eye flash (RF_RS), from the right screw in response to left eye flash (LF_RS), from the left screw in response to right eye flash (RF_LS), and from the left screw in response to left eye flash (LF_LS). For each waveform, we identified P1, N1, P2, and absolute maximum (AM), which were used to calculate visual function. The early components, P1, N1, and P2, are defined as previously described.³¹ 

Calculating Visual Function Using the Amplitudes from the VEP Waveforms Obtained Via the Epidural Screw Electrodes

Previous reports show that the magnitude, latency, and waveform of VEPs differ between ipsilateral and contralateral stimuli in rats.^24,32 There are several reasons for these differences. The dLGNs of rodents receive fewer synaptic inputs from the optic tract originating from the ipsilateral eye than from the contralateral eye.³³ Additionally, the axons from the dLGN project to different areas in the visual cortex depending on whether the signal was from the ipsilateral or contralateral eye. Contralateral dLGN afferents evoke greatest responses in the posterior–medial cortex, but the ipsilateral dLGN afferents evoke greatest responses more mesially.²⁴ To overcome these challenges, we utilized bilateral epidural screws. Epidural screw electrodes sit above the dura with relatively higher impedance compared to the cortex, allowing summation of volume-conducted signals throughout the visual cortex. Usage of the bilateral screws additionally provides four distinct measurable variables described above (RF_LS, RF_RS, LF_LS, and LF_RS), which can be combined to account for the asymmetry in cortical responses to ipsilateral versus contralateral stimuli. 

Using Gauss's law, the evoked potential measured at the epidural screw electrodes can be represented as a function of the summation of current dipoles generated at the visual cortex multiplied by a constant that represents the surface area and the conductivity of the medium³⁴:  

\begin{eqnarray}V = \sum {{k}_{cortex}}I\end{eqnarray}

(1)

where k_cortex is a constant that is inversely proportional to the conductivity of the medium and the distance from the current source I to the recording location. Meeren et al.¹² demonstrated that cortical N1 is generated by excitatory postsynaptic potentials at layers 5 and 6 of the neocortex. This suggests that the mean transmembrane potential that generates N1 can be modeled as directly proportional to the number of axons in the optic radiation carrying action potentials from the dLGN (N_dLGN).³⁵ 

\begin{eqnarray}\bar{I} = qP{{N}_{dLGN}}\end{eqnarray}

(2)

where q is a quantal amplitude generated by the release of neurotransmitters by the presynaptic neuron, and P is the probability that each axon will release neurotransmitters when it receives an action potential. Ultimately, k_cortex, q, and P do not depend on stimulation of the eye, which means that the N1 amplitude measured at the epidural screw electrode is approximately proportional to the number of dLGN neurons that have generated an action potential.  

\begin{eqnarray}V = k{{N}_{dLGN}}\end{eqnarray}

(3)

where k is a constant that depends on k_cortex, q, and P and remains roughly invariant within a single experimental session lasting less than 20 minutes. The relationship between the number of retinal ganglion cells (RGCs) carrying the action potential to the number of dLGN neurons generating an action potential is far from simple. The exact physics of synaptic connectivity is complex, with asymmetry in the ipsilateral and contralateral projections to the dLGN. We make two careful assumptions to model this scenario. First, the number of dLGN neurons generating action potentials is monotonically dependent on the number of RGCs, meaning that a greater number of RGCs will generate the same or a greater number of post-synaptic action potentials.  

\begin{eqnarray}{{N}_{dLGN}} = f( {{{N}_{RGC}}} )\end{eqnarray}

(4)

where N_RGC is number of axons in the optic nerve carrying the action potential from the RGCs to the dLGN, and f(x) is a monotonic function that describes the relationship between the number of RGCs and dLGN that carry action potentials. Second, we introduce a variable (x), defined as the proportion of RGCs that synapses to contralateral dLGNs compared to total RGCs that synapse to either ipsilateral or contralateral dLGNs. Using this, we generated four equations:  

\begin{eqnarray}{{V}_{R{{F}_{LS}}}} = {{k}_{contralateral}}*x*f( {{{N}_{RG{{C}_{OD}}}}} )\end{eqnarray}

(5)

 

\begin{eqnarray}{{V}_{R{{F}_{RS}}}} = {{k}_{ipsilateral}}* ( {1 - x} )*f ( {{{N}_{RG{{C}_{OD}}}}} )\end{eqnarray}

(6)

 

\begin{eqnarray}{{V}_{L{{F}_{LS}}}} = {{k}_{ipsilateral}}* ( {1 - x} )*f ( {{{N}_{RG{{C}_{OS}}}}} )\end{eqnarray}

(7)

 

\begin{eqnarray}{{V}_{L{{F}_{RS}}}} = {{k}_{contralateral}}*x*f ( {{{N}_{RG{{C}_{OS}}}}} )\end{eqnarray}

(8)

where k is the constant of proportionality of evoked potential to N_dLGN, and \({{N}_{RG{{C}_{OD}}}}\) and \({{N}_{RG{{C}_{OS}}}}\) represent the number of RGCs carrying the action potential in the right and left optic nerves, respectively. Also, k was divided into k_{contralateral} and k_ipsilateral because of the aforementioned topographical differences between cortical responses to contralateral and ipsilateral eye stimulations which result in different synapse kinetics and different conductance values between the ipsilateral and contralateral zones of responses to the screw electrode. Rearranging the above equations, we obtain:  

\begin{eqnarray} \frac{{f( {{{N}_{RG{{C}_{OD}}}}} )}}{{f( {{{N}_{RG{{C}_{OS}}}}} )}} = \sqrt {\frac{{{{V}_{R{{F}_{LS}}}}*{{V}_{R{{F}_{RS}}}}}}{{{{V}_{L{{F}_{LS}}}}*{{V}_{L{{F}_{RS}}}}}}} \end{eqnarray}

(9)

 

In this paper, we define the visual function of the right eye in respective to the left eye as \(\frac{{f( {{{N}_{RG{{C}_{OD}}}}} )}}{{f( {{{N}_{RG{{C}_{OS}}}}} )}}\), which can be calculated from the four variables that can be measured using the bilateral epidural screw electrodes. 

There are several limitations to this method. First, the variable k, which is comprised of conductivity of the cortex and the synaptic properties at the visual cortex, and the variable x may not be fully symmetric due to individual anatomical variances in the rat. However, this can be averaged out by using multiple animals. Second, it is difficult to assign specific physiological significance to the function f(N_RGC). Beyond the asymmetric decussation, the relationship between RGCs and dLGNs is not 1:1 or even a function that can be modeled. However, as long as the function is monotonic, it allows us to estimate the visual function of one eye relative to the other in a non-parametric way. Therefore, our equation provides a meaningful assessment of the health of a diseased optic nerve even if it lacks a direct physiological correlate. For VEP waveforms obtained with needle electrodes, we defined visual function as the ratio of the component amplitudes of the right eye (rNAION-induced) to those of the left eye (control). 

Measuring Visual Acuity

Three weeks after rNAION induction, the rats were placed in a virtual optokinetic system (OptoMotry; Cerebral Mechanics, Toronto, CA) as described in Douglas et al.³⁶ Briefly, the rats were placed on a stable elevated platform in the center of a square. A video camera provided a bird's-eye view of a virtual-reality chamber created with four computer monitors facing inward. We used proprietary software to project a continuously moving sine-wave black-and-white grating on the monitors. Visual thresholds were determined using a staircase procedure, where the step size of the grating was halved after each reversal. The spatial frequencies at which the rats consistently exhibited optokinetic nystagmus were recorded for both control and rNAION-induced eyes in cycles per degree (c/d). 

RGC Stereology

Animals were euthanized after the last fVEP measurements, which were performed at least 30 days after rNAION induction. It previously has been shown that at least 95% of post–rNAION-associated RGC loss occurs by this time.³⁷ Following perfusion with 2% paraformaldehyde in phosphate-buffered saline (PF-PBS) solution, eyes with adjacent optic nerve (ON) were enucleated. The corneas then were punctured with a 26-gauge needle, and the eyes were post-fixed for 24 hours in 4% PF-PBS. Retinas were then dissected and reacted with 1:500 anti-rat Brn3a antibody (Synaptic Systems, Goettingen, Germany). Retinas were immunolabeled with Cy3-donkey anti-rat secondary antibody (Jackson ImmunoResearch, West Grove, PA) and flatmounted, and Brn3a(+) RGCs quantified using stereology via the Stereo Investigator program (MBF Bioscience, Williston, VT) linked to a motorized stage on a FluoView FV300 fluorescent microscope (Olympus, Tokyo, Japan). RGC loss was estimated by random sampling across the retina.³⁸ 

Statistical Calculations

We used an unpaired Student's t-test to assess the descriptive statistics of latencies and amplitudes of the VEP components and to determine if the dependent variables were significantly different from each other. The Mann–Whitney test was used to account for non-normal distributions of the variables when comparing R₉₅ and intrasession and intersession coefficients of variances. Pearson's correlation coefficient was used to assess correlation between the visual function calculated using the above method (Equation 9) and other metrics of rNAION induction. Linear mixed-model regression using restricted maximum likelihood method was used to analyze the effect of time on the fVEP parameters, with data points grouped by each rat. Python and Prism (GraphPad, Boston, MA) were used to perform statistical analysis. Values in this paper are presented as mean ± SD unless otherwise specified. 

Results

Rat Statistics

A total of 64 animals were prepared for VEPs. Sixteen animals underwent VEPs using needle electrodes: 13 with rNAION (right eye only) and three naïve controls. Forty-four animals underwent transcranial VEPs: 38 with rNAION (right eye only) and six naïve controls. Four rats (8% of total) died before any VEPs could be obtained. Post-rNAION ONH edema was evaluated at 1 day after induction using SD-OCT. The mean ONH edema of the rNAION-induced rats was 647 ± 113 µm compared to 288 ± 45 µm in naïve controls. We obtained the visual acuity from all rNAION-induced surviving rats (n = 44). Mean residual RGC survival for all induced eyes was 35.2% ± 24.1%. 

Representative fVEP Waveforms Obtained From Subcutaneous Needle and Epidural Screw Electrodes

Figure 3 shows a comparison of representative fVEP waveforms for both control and rNAION-induced animals as measured by a single subcutaneous midline needle electrode versus two epidural screw electrodes. Needle electrodes generated two different waveforms within a single session: (1) generated by a flash stimulus to the right eye (Figs. 3A, 3C), and (2) generated by a flash stimulus to the left eye (Figs. 3B, 3D). Measurements using epidural screw electrodes generated four waveforms from the combination of the two epidural screw electrodes, one on the dura over each visual cortex, and the two eyes to which the flash stimuli can be delivered. Waveforms are denoted as RF_LS, RF_RS, LF_LS, and LF_RS, where RF_LS refers to fVEP generated by the left epidural screw in response to flash in the right eye, RF_RS refers to fVEP generated by the right epidural screw in response to flash in the right eye, LF_LS refers to fVEP generated by the left epidural screw in response to flash in the left eye, and LF_RS refers to fVEP generated by the right epidural screw in response to flash in the left eye. Epidural fVEPs elicited by the contralateral flash (Figs. 3E, 3I, RF_LS; Figs. 3H, 3L, LF_RS) have approximately threefold greater early component amplitudes than those elicited by the ipsilateral flash (Figs. 3F, 3J, LF_LS; Figs. 3G, 3K, RF_RS). 

Figure 3.

 

Comparisons of representative fVEP waveforms measured with subcutaneous needle electrodes versus epidural electrodes. (A–D) Subcutaneous electrode measurements were obtained from a control rat (A, B) and rNAION-induced rat (C, D). (E–L) These responses were compared against results from obtained from epidural screw electrodes in a control rat (E–H) and an rNAION-induced rat (I–L). Early components, P1, N1, and P2, as well as the absolute maxima are labeled in each waveform. With epidural screw electrodes, waveforms generated by the contralateral flash stimuli produced significantly greater amplitudes than those generated by the ipsilateral flash stimuli. Waveforms generated with epidural electrode had much greater amplitudes than those generated with subcutaneous electrodes. As expected, rNAION induction to the right eye resulted in a decrease in the evoked potential generated by the right-eye flashes relative to those generated by the left-eye flashes.

In animals with rNAION induction in the right eye, both subcutaneous and epidural electrodes show an expected response to rNAION induction. Waveforms obtained via needle electrodes exhibited significantly decreased relative amplitudes of P1, P2, and AM relative to N1 in the induced eye (Fig. 3C) compared to the control eye (Fig. 3D). Similarly, in animals in which epidural electrodes were used, rNAION induction resulted in decreased amplitudes in waveforms obtained from the flash stimuli to the affected eyes (RF_LS, RF_RS) (Figs. 3I, 3K) compared with the corresponding waveforms from the flash stimuli to the control eyes (LF_RS, LF_LS) (Figs. 3J, 3L). 

Epidural Electrodes Enabled More Sensitive Measures Than Those Obtained By Subcutaneous Electrodes: Analysis of rNAION-Induced Changes in fVEP Early Components

Figures 4A to 4D shows the amplitudes and the latencies of the early VEP components, P1, N1, and P2, in control and rNAION-induced rats using subcutaneous electrodes. Although there was statistically significant decrease in absolute amplitudes of the P1 (P < 0.05)and P2 (P < 0.01) components between the rNAION rats and the control rats, there were no statistically significant differences in absolute amplitudes of the early fVEP components between the rNAION-induced and the control eyes (Fig. 4A). However, the rNAION-induced eyes showed a statistically significant decrease in the relative amplitudes of P1, P2, and AM (offset by N1) compared to the control eye (P < 0.05) (Fig. 4B) and a statistically significant decrease compared to the rats in the control group. This confirms that relative amplitudes, calculated as the difference between the positive component and a referenced negative component (N1 in this case) are a much more sensitive measure than the absolute amplitudes of each component. Unlike the amplitudes, no statistically significant differences were seen in either the absolute or the interval latencies between the adjacent components (Figs. 4C, 4D). 

Figure 4.

 

Descriptive statistics of the absolute and relative amplitudes and latencies of the early components in fVEP waveforms obtained from subcutaneous needle electrodes. Values are presented as mean ± SD. (A) Comparisons of absolute amplitude between rNAION-induced rats and control rats. Values were obtained from light stimulation of the uninduced left (OS, blue) and rNAION-induced right (OD, red) eyes. (B) Comparisons of relative amplitudes between rNAION-induced and control rats. (C) Comparison of latencies among different early components (P1, N1, N2). (D) Comparisons of interval latencies for rNAION-induced and control rats. Statistical significance was calculated using unpaired Student's t-test, with comparisons made between the right-eye flashes and left-eye flashes. Comparisons were also made between rNAION-induced rats and control rats. The notation P1N1 represents the absolute difference in amplitude or latency between P1 and N1 components. *P < 0.05, **P < 0.01.

In comparison, the panels in Figure 5 show the descriptive statistics of the amplitudes and latencies of early VEP components of rNAION-induced rats using the epidural screw electrodes. Epidural fVEPs showed statistically significant differences even in the absolute amplitudes of P1 (P < 0.01), N1 (P < 0.001), and P2 (P < 0.001), when comparing the contralateral evoked potentials between the rNAION-induced eye and the control eye. Absolute and relative amplitudes of all early components of contralateral evoked potentials of the right eye (rNAION-induced) in rNAION-induced rats were significantly lower than those of control rats. Additionally, there were statistically significant differences in the relative amplitudes of P1, P2, and AM in both contralateral and ipsilateral evoked potentials (P < 0.01). 

Figure 5.

 

Descriptive statistics of the absolute and relative amplitudes and latencies of the early components in fVEP waveforms obtained from epidural electrodes. Values are presented as mean ± SD. (A) Comparisons of absolute amplitudes between rNAION-induced and control rats. Values were obtained from light stimulation of the uninduced left (OS, blue) and rNAION-induced right (OD, red) eyes. Bilateral electrode implantations enabled individual values to be obtained from both contralateral and ipsilateral visual cortices relative to the ocular stimulus. (B) Comparison of relative amplitudes between rNAION-induced and control rats. (C) Comparison of latencies among different early components (P1, N1, N2). (D) Comparisons of interval latencies for rNAION-induced and control rats. Statistical significance was calculated using unpaired Student's t-test, with comparisons made between the right-eye flash and left-eye flash, preserving the laterality of the stimulus in the epidural fVEPs. For example, P1 from LF_RS was compared with P1 from RF_LS, as both were waveforms obtained from contralateral stimuli. Comparisons were also made between rNAION-induced rats and control rats. The notation P1N1 represents the absolute difference in amplitude or latency between the P1 and N1 components. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

In contrast to results obtained from subcutaneous needle fVEPs, the average interval amplitudes measured with epidural fVEPs in response to contralateral stimuli were 8 to 10 times greater, and average interval amplitudes measured in response to ipsilateral stimuli were 1.5 to 2.5 times greater. There were statistically significant differences in the fVEP latencies measured with epidural screw electrodes between the induced and control eyes, whereas no significant differences were found in the latencies measured with subcutaneous electrodes. P1 and N1 latencies in the contralateral evoked potentials from the rNAION-induced eyes were significantly longer than those obtained from the control eye. In contrast, P2–N1 interval latencies in the contralateral evoked potentials from the rNAION-induced eye were significantly shorter than those from the control eye. These changes highlight the ability of epidural electrodes to detect more subtle changes that might otherwise be missed with subcutaneous electrodes. 

Interestingly, statistical differences in the P2 component were observed even in naïve, non–rNAION-induced, animals. The P2 amplitudes in the contralateral evoked potentials were significantly greater in the right eye compared to the left eye. Similarly, both the P2 absolute latency and the P2–N1 interval latency were significantly longer in the right eye compared to the left. Given the lack of such a pattern in the ipsilateral evoked potentials, these differences may suggest an asymmetry in the higher level visual processing between the two hemispheres. 

VEP Measures Obtained by Epidural Electrodes Enabled Reduced Intra- and Intersession Variability Compared With Those Obtained by Subcutaneous Electrodes

fVEP waveforms and measurements performed on rNAION-induced rats using subcutaneous electrodes showed significant intrasession (Figs. 6I, 6J) and intersession (Figs. 6K, 6L) amplitude variations in response to flashes in both the NAION-induced (Figs. 6I, 6K) and control eyes (Figs. 6J, 6L). However, the latencies of the major early components (P1, N1, and P2) remained consistent across different measurements. In comparison, fVEPs obtained using epidural electrodes exhibited significantly reduced intrasession and intersession variations (Figs. 6A–D, intrasession; Figs. 6E–H, intersession). The overall shapes of all four generated waveforms also were conserved both within and across multiple sessions. 

Figure 6.

 

(A–L) Representative measurement sessions showing the intrasession and intersession variations of fVEP waveforms obtained with either epidural screw electrodes (A–H) or subcutaneous needle electrodes (I–L). The plots in A to D, I, and J show fVEP measurements performed within the same session; Red indicates a first measurement, blue indicates a second measurement, and black represents a third measurement. Due to time constraints, only two measurements per session were obtained with epidural screw fVEPs, as each “measurement” required data acquisition from both the right and left screws. The plots in E to H, K, and L show an example of intersession variation across three fVEP measurements performed across multiple days. Red indicates the first day, blue indicates the second day, and black indicates the third day. There was less intersession and intrasession variation in the fVEP waveforms obtained via epidural screw electrodes, both before and after the rNAION induction, compared with waveforms obtained from subcutaneous needle electrodes. (M–O) Graphs show a statistical comparison of repeatability coefficients (R₉₅) (M), intrasession CVs (N), and intersession CVs (O) of visual function calculated using the intervals P1–N1, P2–N1, or AM–N1 obtained from subcutaneous or epidural fVEPs. Values are presented as mean ± SEM. Statistical significance was calculated using Mann–Whitney test. *P < 0.05, **P < 0.01, ***P< 0.001, ****P < 0.0001.

The conserved patterns of fVEP waveforms obtained using epidural electrodes allowed us to use the relative amplitudes (P1–N1, P2–N1, or AM–N1) to estimate the visual function of an rNAION-induced eye relative to the control eye. This was done by using either the ratio of two amplitudes for subcutaneous fVEPs or Equation 9 for epidural fVEPs. Intrasession and intersession variations in visual function were then measured for the two different metrics by calculating the intrasession and intersession coefficients of variance (CVs) and the repeatability coefficient (R₉₅) (Figs. 6M–O). 

For intrasession variability, the visual function calculated from P1–N1 or AM–N1 using epidural electrodes had significantly lower R₉₅ values (Fig. 6M) and intrasession CVs (Fig. 6N) compared to subcutaneous needle electrodes. Within epidural fVEPs, visual function calculated using P2–N1 had significantly greater intrasession CVs than that from P1–N1 or AM–N1. For intersession variability, visual functions calculated from fVEPs using P1–N1 via epidural electrodes had significantly lower intersession CVs (Fig. 6O) compared to those obtained using subcutaneous needle electrodes. However, no significant differences were seen in intersession CVs among the different metrics of P1–N1, P2–N1, and AM–N1. 

Visual Functions Calculated From Either P1–N1 or AM–N1 Using Epidural Electrodes Strongly Correlated With Visual Acuity, RGC Survival, and ONH Edema

Visual functions calculated using P1 and AM were compared against other metrics of rNAION induction, including ONH edema measured with SD-OCT, relative visual acuity measured with the virtual optokinetic system, and the percentage of surviving RGCs calculated with stereology (Fig. 7). All comparisons showed statistically significant (P < 0.0001) and strong correlations (|r| > 0.7). Visual functions calculated with AM–N1 showed stronger correlation against each individual metric of rNAION than visual function calculated with P1–N1. Of the three other metrics, visual function correlated most strongly with the metric of RGC quantification via stereology. On the other hand, visual functions obtained with subcutaneous electrodes using P1 and AM had weaker correlation than visual functions obtained with epidural electrodes. Visual function with subcutaneous electrodes using P1–N1 had very weak correlation with ONH edema (|r| > 0.19, P = 0.53) and weak correlation with relative visual acuity (|r| > 0.46, P = 0.11). Visual function with subcutaneous electrodes using AM–N1 had moderate correlations with both ONH edema (|r| > 0.54, P = 0.058) and relative visual acuity (|r| > 0.61, P = 0.028). 

Figure 7.

 

(A–K) Scatterplots showing visual function calculated with epidural electrodes using P1–N1 (A–C, first row) and AM–N1 (D–F, second row) and with subcutaneous electrodes using P1–N1 (H, I, third row) and AM–N1 (J, K, fourth row) against different metrics of visual function. Panels A, D, H, and J show ONH edema as measured with SD-OCT. Panels B, E, I, and K show relative visual acuity of OD with respect to OS as measured with a virtual optokinetic system (OptoMotry). Panels C and F show the percentage of RGCs remaining as measured with stereology. Each dot represents a single rat. Pearson's correlation coefficients are included.

Visual Function Measures Remained Stable for ≥30 Days After Epidural Electrode Implantation

fVEP waveforms are known to change over time following transcranial electrode implantation.³⁹ To determine whether this affected our calculation of visual function, we conducted multiple fVEP sessions following implantation of epidural screw electrodes. All epidural screws were noted to remain firmly in place for the first 30 days after surgery. 

We compared the evolution of the P1–N1 amplitude, the corresponding visual function, and the intrasession CV for each animal for at least 1 month post-implantation (Fig. 8). We used the P1–N1 amplitude of the right epidural screw in response to left flash (LF_RS) as it had a greater amplitude as a contralateral evoked potential and was unaffected by rNAION induction. In the initial 20 days post-implantation, the P1–N1 amplitude trended upward, then plateaued in subsequent measurements. The amplitude showed a significant positive trend with time (P < 0.001; 95% confidence interval [Cl], 0.620–1.893). The calculated visual function using P1–N1 did not follow the trend seen with amplitude and no significant trend over time (P = 0.68; 95% Cl, –0.003 to 0.002). On the other hand, the intrasession CV had a negative trend with time (P = 0.001; 95% Cl, –2.256 to –0.549), indicating greater intrasession repeatability of the measurements as the screws remained in place longer. There was a significant positive correlation between the percentage change in amplitude and time (P < 0.001; 95% Cl, 0.620–1.893). There was no significant correlation between changes in visual function and time (P = 0.68; 95% Cl, –0.003 to 0.002). There was a negative correlation between intrasession CV and time (P = 0.001; 95% Cl, –2.256 to –0.549). Statistical significance was assessed using linear mixed model regression with the restricted maximum likelihood method. 

Figure 8.

 

Scatterplots showing the time evolution of fVEP metrics from the date of surgery. (A) Visual function, (B) P1–N1 amplitude of LF_RS. (C) Intrasession CVs are plotted against the number of days elapsed since screw electrode implantation. The change in visual function (A) is shown as the absolute difference from the first session. Changes in amplitude (B) and CV (C) are presented as relative percentage changes from their respective values in the first session. Each line connecting the dots represents an individual rat across multiple sessions.

Discussion

fVEP recordings are widely used in animal studies to quantify visual function, but a review of the literature reveals substantial variations in rat fVEP waveforms due to differing experimental conditions, potentially resulting in confusing interpretations. In this report, we have proposed a new approach that can be applied to interpret results generated by different laboratories. The use of inexpensive stainless-steel epidural screw electrodes not only improves SNR but also enhances repeatability of rat fVEPs. Although unilateral epidural screw electrodes have been utilized in previous studies,^13,21 ours is the first study, to the best of our knowledge, to analyze fVEP waveforms obtained with bilateral epidural screws, with one electrode positioned over each striate cortex. 

We used bilateral epidural electrodes to compare the magnitude of evoked potentials generated by afferent inputs received by each visual cortex in a rat using Equation 9. As expected, the evoked potential waveforms measured with subcutaneous electrodes had much greater amplitudes and SNRs compared to the waveforms measured with subcutaneous electrodes. In addition, we demonstrated that calculating the relative visual function of the two eyes using epidural-screw fVEPs and Equation 9 is significantly more repeatable than calculating visual function as a ratio of two amplitudes measured with subcutaneous needles. The visual function obtained with our method also correlated strongly with other metrics of visual function, including visual acuity and the approximate percentage of surviving RGCs. Finally, our results show that the electrode implantation durations of up to a month do not significantly affect the calculation of visual function. Our approach thus makes the method of calculating visual function using bilateral epidural electrodes more repeatable and reliable, significantly reducing the number of animals required to achieve statistical significance in any individual analysis. 

The epidural positioning of electrodes provides several advantages for comparing relative visual function between the two visual cortices. First and foremost, the low impedance of the epidural screws allows much greater SNRs. Second, each rat has a natural anatomical asymmetry between the two visual cortices, and the topography of fVEP responses to ipsilateral and contralateral stimuli is complex.²⁴ However, epidural screws help minimize this issue by averaging the electrical responses over the overall cortical area, as the electrical impedance of the dura is greater than that of the underlying brain.²⁶ Third, the epidural positioning also provides stability for VEP signals over time. Previous studies have shown that superficial lesions associated with local inflammation and cortical destruction preserve P1–N1 components, suggesting that the source and the sink of the initial VEP components are not significantly affected by the superficial inflammatory process.¹⁰ Finally, our Equation 9 is invariant to external factors that might affect the screw quality, such as screw rusting, as these factors will affect both the denominator and numerator equally, cancelling out in our calculation of the visual function. 

There are limitations to our epidural approach. First, low impedance in a transcranial recording allows the electrodes to pick up volume-conducted signals from distance sources in the cortex, including contralateral cortex and the superior colliculus, which also receives significant afferent input from the optic nerve.^40,41 However, a greater than sevenfold difference in N1 components between the ipsilateral and contralateral epidural screw in response to a flash stimulus in the same eye suggests that the contribution from the contralateral cortex is minimal. This may be due to high impedance of the midline structures and the relatively longer distance between the epidural screw to the contralateral cortex.^26,41 As for the superior colliculus, Barnes et al.⁴² demonstrated that the lesion of superior colliculus did not significantly affect the early components (P1, N1, and P2) in transcranial recordings. A second limitation of our approach is that our derivation of Equation 9 simplifies the visual pathway and does not fully encapsulate the complex physiology of synaptic transmission between the RGC and the dLGN, nor the nontrivial mapping of ipsilateral and contralateral stimulation in the visual cortex. However, empirical data show that the visual function calculated from our equation correlates strongly with other metrics of visual function. Furthermore, the monotonicity of the RGC signal relative to the evoked potential generated by the visual cortex allows for statistical analysis, even if there is no precise physiological correlate to the visual function obtained with Equation 9. 

There are three additional noteworthy findings in our report that warrant further exploration. First, we observed a statistically significant asymmetry in the P2 amplitude and latency of fVEPs in non–rNAION-induced animals using epidural electrodes. Contralateral evoked P2 potential amplitudes were greater and P2 latency was longer when the light stimulus was presented to the right eye compared with the left eye (Fig. 5C). This difference may be due to mechanical bias, as epidural electrode implantation in control animals was performed by a single investigator. However, it also may reflect inherent asymmetry in rats, specifically between the response of the left striate cortex to right eye stimuli and the right striate cortex to left eye stimuli. The asymmetry may not be isolated to P2, which is thought to be generated by inhibitory thalamocortical input on pyramidal cells.¹² Instead, it may arise from the superposition of later components with higher amplitude on the left striate cortex, explaining both the increased amplitude and the increased latency. This hypothesis could also help explain the finding that when the right eyes were rNAION induced, the P2–N1 interval latency in the contralateral fVEP became significantly shorter in the left striate cortex. The superposition of P2 with subsequent VEP components likely contributed to why P2–N1 was the least repeatable metric compared to P1–N1 and AM–N1 in the fVEP measurements with epidural electrodes. Further experiments with larger sample sizes using naïve animals could help elucidate if this asymmetry in P2 and later VEP components exists between the two striate cortices. 

Second, our results show that, when using epidural electrodes, the AM and P1, relative to N1, are both repeatable metrics (Fig. 6). The visual functions calculated from AM–N1 and P1–N1 are both meaningful, as they correlate strongly with visual acuity and the percentage of RGC survival (Fig. 7). The choice of when to use P1–N1 versus AM–N1 depends on the specific goals of investigating visual function. N1 has been previously shown to correlate with the primary excitatory postsynaptic potentials of the geniculocortical afferents on layers 5 and 6 of pyramidal cells and has the advantage of being more resistant to the brain state and anesthetics compared to other later components.¹² Using the P1–N1 amplitude difference is an established method of assessing the health of the eye, optic nerve, and optic radiations. In contrast, AM is a mathematical construct with a less direct physiological basis. The AM broadly represents the highest potential measured at the striate cortex after all higher level visual processing has occurred. Thus, it is not surprising that AM–N1 had stronger correlations with ONH edema, visual acuity, and the percentage of RGC survival than P1–N1. Our findings demonstrate that both metrics, P1–N1 and AM–N1, are reliable and repeatable for analyzing visual function in one eye relative to the contralateral eye. The choice of which metric to use depends on the specific goals of the research. 

Finally, our results also show that, although there were chronic effects of epidural screw implantation on the fVEP signal strength, these ultimately did not affect the calculation of relative visual functions (Fig. 8). A positive correlation between P1–N1 amplitude and the days elapsed from the surgery aligns with the results of previous findings showing that focal lesions on the superficial layers of cortex can augment N1 amplitude.¹⁰ However, this increase in amplitude affected all four waveforms, RF_LS, RF_RS, LF_LS, and LF_RS, by similar proportions. This common factor from the epidural screw implantation is canceled out during visual function calculations using Equation 9. It is also important to note that fVEP measurement repeatability improved after screw implantation surgery, possibly indicating that the local inflammatory reaction resolved within a few days and that the resulting scar tissue surrounding the screws is accounted for by our equation. This allows researchers flexibility in performing multiple measurements across many days without being constrained by time. 

Additional improvements could be suggested for the current configuration. First, previous studies have demonstrated improvement in VEP waveforms by further normalizing the electroencephalogram power spectrum, using a reference screw electrode at the cranial midline, 3 mm rostral to the bregma.⁴³ The addition of a reference epidural screw could potentially cancel out miscellaneous cortical activity, but this must be balanced against the risk of increased morbidity and mortality of the procedure. Second, we employed multiple heuristic approximations to create the current protocol to systematically label P1, N1, and P2 without operator bias, but it may be possible to utilize deep learning models to enhance the protocol and improve the interpretation of noisy or low-amplitude fVEP waveforms. Deep learning models have previously been applied in human VEPs to detect and classify components.^44,45 Finally, the Celeris software system allows only unipolar fVEP recordings. Although Equation 9 helps accommodate for external sources of variations between the first and second electrode measurements, simultaneous bipolar recording from both epidural screws would reduce the time required for the experiment and could further decrease intrasession variation by minimizing external sources of variations such as the animal's anesthetic state. 

In conclusion, we have developed a repeatable and accurate methodology to calculate the relative visual function of a diseased optic nerve compared with a control eye using fVEP measures obtained using bilateral epidural electrodes. We have demonstrated that the visual function calculated from this approach correlates very strongly with other recognized metrics of visual function. Our approach to evaluating visual function in rats will significantly reduce the number of animals needed to achieve statistical significance and power. 

Acknowledgments

Supported by a grant from the National Eye Institute, National Institutes of Health (R01 EY032519-01A1 to SLB) and by the Hillside Foundation. 

Disclosure: K. Woo, None; Y. Guo, None; Z. Mehrabian, None; N.R. Miller, None; S.L. Bernstein, None