Ординатура / Офтальмология / Английские материалы / Artificial Sight Basic Research, Biomedical Engineering, and Clinical Advances_Humayun, Weiland, Chader_2007
.pdf370 Hetling
using two RSA devices positioned relative to the retina; e.g. one in the subretinal space, one in the vitreous.
The stimuli used were constant current, biphasic square-wave pulses, although any arbitrary stimulus kinetics can be achieved. Square-wave pulses are convenient to generate, and are commonly employed. The surface potential of the stimulating electrode is often orders of magnitude greater than the response amplitude sensed by the recording electrode. Because the stimulating and recording electrodes are relatively close together (both in contact with ocular tissue), a substantial stimulus artifact is recorded. This artifact can be either computationally subtracted, or minimized by temporarily disconnecting the amplifier during stimulus delivery.
Non-Invasive Recording
As described above, non-invasive recording has the advantages of simplicity and the potential for clinical application. The response of the retina to electrical stimuli can be measured at the cornea using the same recording technology as is employed with standard ERG techniques. For clarity, we will refer to the response of the retina to electrical stimulation as the eERG. Both the fERG and the pfERG have straightforward correlates when measuring the eERG; each will be described below.
Adaptating the Focal ERG
Adaptation of the fERG to the eERG is essentially unavoidable when using small electrodes to stimulate the retina. By virtue of design, a small area of the retina is activated, and the resulting signal can be recorded using a corneal recording electrode identical to that used to measure the ERG. However, there are important differences. A single micron-scale electrode in contact with the retina will stimulate only a tiny fraction of the retinal neurons, even less than typically targeted by the fERG. Consider that the maximum ERG amplitude recorded in a dark-adapted wild-type rat in response to a single full field flash of light is approximately 1 mV. The surface area of a rat retina can be conservatively approximated by assuming that it lines the posterior hemisphere of an eye of 7 mm diameter. This yields an area of 77 mm2. A 100- m-diameter electrode has a surface area of 0 0079 mm2, and thus subtends 0.01% of the retina. If the electrode only stimulated the neurons directly above it, the recorded response would be expected to be no larger than 0.01% of the response to the full-field light stimulus, or 0 1 V. To record this response with a minimum acceptable SNR of 5, the noise level would need to be 0 02 V. This low noise level is rarely achieved in real-time recording, and necessitates the use of averaging (the root mean square (RMS) noise level is inversely proportional to the number of responses averaged).
Figure 20.5 shows representative eERG responses elicited by relatively large (500 m diameter) bipolar electrodes placed in the subretinal space; the good
20. Electrophysiology of Natural and Artificial Vision |
371 |
|
50 |
|
|
|
|
|
|
|
|
|
40 |
|
|
|
|
|
|
|
|
( V) |
30 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Amplitude |
20 |
|
|
|
|
|
|
|
|
10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
0 |
|
|
|
|
|
|
|
|
|
–10 |
|
|
|
|
|
|
|
|
|
–200 |
–100 |
0 |
100 |
200 |
300 |
400 |
500 |
600 |
Time (ms)
Figure 20.5. Representative eERG waveforms recorded in three animals. Each waveform is the average of 100–300 responses to a biphasic square pulse stimulus (100 A each phase) delivered at approximately 3-second intervals. Animals were pigmented P23H rats, 6–12 weeks of age. Retinal stimulating array electrodes were 500 m diameter gold electrodes (cf. Figure 20.4). Major response components (see Figure 20.2) are consistently present across animals.
SNR was achieved by averaging several hundred responses to the same stimulus. The peak amplitude of about 10 V is consistent with the area of the retina subtended by the electrode pair. The lower trace in Figure 20.2 plots the mean eERG response recorded with these electrodes in six animals (different from those in Figure 20.5), and represents the typical waveform. Comparison with the light-induced ERG above reveals obvious differences between the responses to light and the electrical stimulation. The cellular basis for these differences is explored below, but first a number of technical considerations should be noted.
When delivering a large number of repetitive stimuli, several questions must be answered. At what frequency should the stimuli be delivered? Does the retina adapt or fatigue in response to the frequent stimuli? A typical interval between light flashes when performing a typical single-flash ERG experiment is 1–2 minutes, to allow the animal to dark-adapt between stimuli. Light adaptation occurs at several levels in the visual pathway, and so it may be assumed that adaptation to electrical stimulation may also occur. Are there any other physiological changes that might occur during the tens of minutes required to deliver hundreds of repeated stimuli? All of these factors must be considered within the constraint of the practical duration for an in vivo experiment, during which an animal must be kept anesthetized, or a patient able to maintain concentration and avoid excessive discomfort.
Two other issues require special consideration when recording the response to electrical stimulation. These are the determination of threshold and saturation,
372 Hetling
two fundamental characteristics of any response. In a stochastic, binary system, such as a spiking neuron, threshold is typically defined as the stimulus strength required to elicit a just-measurable response (i.e. a single action potential) 50% of the time. For a signal that is continuous, such as the eERG amplitude, threshold can be defined as a signal that is some criterion amount above the baseline noise level (e.g. 3× the RMS noise, or 3 standard deviations above the mean noise level). However, for small signals, the noise level is inversely proportional to the square root of the number of responses averaged, and the threshold is thus a function of the recording protocol, and becomes difficult to assess.
Saturation is defined as the maximum response of the system being measured. A plot of response amplitude vs. stimulus intensity for neural systems often takes the form of a saturating exponential function. As the stimulus strength increases, eventually a maximum response is elicited, and further increases in stimulus strength result in no further increase in response amplitude. In the healthy retina, the saturated ERG response occurs when the number of photons delivered to the eye is sufficient to result in the closing of all of the cGMP-gated cation channels that mediate the photoreceptor dark current. This happens at about 1000 photons absorbed per photoreceptor, and doubling the number of photons delivered to the eye does not further increase the response.
The special consideration when dealing with electrical stimuli is that of stimulus field containment and recruitment of cells. When the potential applied to a stimulating electrode increases, the field potentials around the electrode also increase. If you assume that there is a threshold value of field potential that results in activation of a retinal neuron, then as the field potential at a given point in space increases, it becomes more likely that a neuron in that location will reach threshold and contribute to the recorded response. The region of the retina that is brought above threshold by direct effect of the delivered electrical stimulus (i.e. not via synaptic connections) is termed the stimulus field. The number of neurons contributing to the response (due to both direct stimulation and synaptic connections) then becomes a function of the stimulus strength, which is only limited by the potentials that cause electrode or tissue damage. The plot of eERG amplitude vs. stimulus strength is not well fit by a simple saturating exponential, because as neurons near the electrode become saturated at high stimulus strengths, neurons farther away from the electrode are just reaching threshold. This issue of cell recruitment also illustrates the need for electrode designs that create consistent stimulus fields over a range of stimulus strengths, to avoid recruitment of neurons outside of the desired stimulus field when using stronger stimuli.
Adaptating the Paired-flash ERG
Adapting the pfERG technique to electrical stimulation depends on the ability to establish two critical stimulus amplitudes. The lower stimulus amplitude is defined as that which elicits a response just above threshold (analogous to the first flash in the pfERG protocol). The upper stimulus amplitude is one in which the stimulus field is the same as that for the low stimulus (attained through
20. Electrophysiology of Natural and Artificial Vision |
373 |
judicious electrode design), and which drives the first-order neurons in the stimulus field to saturation (analogous to the second, bright flash in the pfERG protocol). In practice, electrical stimuli would be delivered as paired pulses, and the response of the first-order neurons to the first pulse would be titrated at arbitrary times by the second pulse. Just as the pfERG allows reconstruction of the full time course of the photoreceptor response, the paired-pulse eERG allows the reconstruction of the full-time course of the first-order neurons stimulated by electrodes. The utility of this approach lies in its ability to measure the kinetics of the response. This information can then be used to identify the type(s) of “first-order” cells in the stimulus field and gain insight into their physiology and response characteristics relative to stimuli of arbitrary design.
Adapting Invasive Recording Techniques
Invasive recording, using microelectrodes targeting single cells or small populations of cells, provides spatial information that is inaccessible with the noninvasive techniques. The extent of the electrical stimulus field cannot be visualized directly, so the population of neurons targeted is uncertain. Further, due to the complex and atypical network connectivity in a degenerate retina, the extent of the response field (the population of neurons responding to the stimulus due to both direct and synaptically mediated stimulation) is also unknown when recording at the cornea. If, however, a microelectrode is brought into contact with the ganglion cell layer of the retina, making single-unit recordings, the response field at this level can be determined by scanning the electrode across the retina during repeated delivery of the stimulus. A map of the lateral extent of neurons whose activity is correlated with stimulus delivery is created. The technical considerations are identical to those used to measure the response to light stimuli, with the added minor challenges of moving the recording electrode in carefully defined increments within the eye, and stimulus artifact suppression (discussed above). The former challenge is easily met using a commercially available robotic micromanipulator combined with both visual feedback through the pupil and by inferring the location of the electrode by the recorded response.
A second protocol which can be employed with invasive recording is that of modifying the parameters of the electrical stimulus until the resultant spike train recorded from a single ganglion cell mimics that elicited by a defined light stimulus. Here, a light stimulus is delivered to an area of the retina that is within the stimulus field of an electrode, and the ganglion cell response is recorded. Then the same area of retina is stimulated with the electrode, and the stimulus parameters altered until the ganglion cell response is similar to that obtained in response to the photic input. Key features of the stochastic spike train to be mimicked are latency to first spike, peak (or minimum for an OFF-type cell) firing rate, and latency to peak (or minimum) firing rate, each of which are stereotypical for individual types of ganglion cells. Obviously, this protocol can only be carried out in a retina that still responds to light, but even so it will likely
374 Hetling
provide many insights into the relationship between the input–output functions of the degenerate retina for light and the electrical stimuli.
Pharmacological Dissection of the eERG
The strategy of pharmacological dissection has been one of the most powerful in learning to interpret the ERG, and so it will prove very useful in identifying the cellular origins of the various components of the eERG. Figure 20.6 illustrates the relationship between the retina and a non-uniform electric field generated by an active-reference pair of subretinal electrodes. Also indicated are the cell types in the retina that are incapacitated by three common glutamate agonists. The properties of these agonists are summarized in Table 20.2.
The procedure for studying the response to electrical stimulation is identical to that for studying response to light stimulation, but there is an important distinction that must be made when analyzing the results. In a healthy retina responding to light stimuli, the photoreceptors become active first, and the visual signal propagates through the retina in a reasonably well-understood manner. The relative order of neurons in the retinal network, i.e. presynaptic vs. postsynaptic for each cell type, determines the order in which activity propagates. This is somewhat complicated by lateral pathways in the retina and feedback, but the principle flow of information is from photoreceptors to bipolar cells to ganglion cells.
However, an electrical stimulus might directly stimulate neurons at several levels in the retina. That is, the stimulus field may subtend the cells closest to the electrode as well as those post-synaptic to these neurons. (In the case of epiretinal electrodes, the stimulus field may subtend ganglion cells as well as cells further from the electrode that are pre-synaptic to the ganglion cells). If the stimulus field extends across synaptic layers in the retina (the inner and outer plexiform layers), the effects of agonists and antagonists targeting the same receptor type become quite different.
An agonist will fully depolarize or hyperpolarize the target cell. This will positively block synaptic transmission to this cell type. It may or may not prevent the membrane potential of the target cell from changing in response to an applied electric field, and so this cell may or may not contribute to the recorded response. This is an important question that has yet to be approached experimentally or via computational modeling. Unpublished work by our group suggests that the passive properties of the neural membrane, the morphology of the cell, and the intersection of the cell with an inhomogeneous electric field largely determine the response to applied electric fields, and so a cell may still respond even if the receptor-mediated biophysics are inoperative. A further question is then whether this response (i.e. change in membrane potential) is within a range that will modulate transmitter release and thus elicit additional response contributions from downstream neurons.
An antagonist blocks synaptic transmission, but does not alter the membrane potential of the target neuron significantly from resting level. Therefore, if this
20. Electrophysiology of Natural and Artificial Vision |
375 |
ASP
L-AP4
NMDA
Figure 20.6. Interaction between a complex electric field and the retina, and the targets of common glutamate agonists. A cartoon of a moderately degenerated retina is superimposed on a color map of potential gradient (second spatial derivative of the field potential) generated by a finite element model of an eye containing a pair of subretinal electrodes. For the non-uniform electric field created by these bipolar electrodes (lower left and lower right in each panel), the super-threshold stimulus field subtends only subpopulations of retinal neurons, which may be identified by pharmacological dissection of the response. Boxes indicate the cells that are saturated by the indicated glutamate agonist (see Table 20.2). Aspartate (ASP) binds to receptors of all post-receptor neurons. 2-amino-4-phosphonobutyrate (L-AP4) binds selectively to the bipolar cells of the ON-pathway. N-methyl-D-aspartic acid (NMDA) selectively binds to third-order neurons.
376 Hetling
Table 20.2. Glutamate agonists useful in pharmacological dissection of the summed response of the retina to photic or electrical stimulation.
Compound |
Receptor |
Effect |
Target cell types |
References |
|
|
|
|
|
|
|
1) |
APB (L-AP4) |
mGluR4, 6–8 |
Agonist |
Photoreceptors |
[24, 25] |
|
2-amino-4- |
|
|
(?), Amacrine & |
|
|
phosphonobutyrate |
|
|
On Bipolar Cells |
|
2) ASP Aspartate |
All GluR |
Agonist |
All Retinal Cells |
[26] |
|
3) |
PDA cis-2,3- |
All iGluR |
Weak NMDA |
Off Bipolar, |
[27–29] |
|
piperidinedicarboxylate |
|
Agonist, KA |
Horizontal & |
|
|
|
|
& AMPA |
Third-Order |
|
|
|
|
Antagonist |
Cells |
|
4) |
NMDA |
NMDA |
Agonist |
Amacrine & |
[30] |
|
N-methyl-D-aspartate |
|
|
Ganglion Cells |
|
|
|
|
|
|
|
cell is within the stimulus field of the electrode, it can readily respond to the applied electric field, and presumably pass the visual signal to post-synaptic neurons in a natural manner.
Pharmacological dissection of the eERG response is performed with two objectives: First, the cellular origins of each component in the response need to be identified. Once this first objective is met, this knowledge can be used to identify the direct cellular targets of the stimulus (i.e. the cell types within the stimulus field).
The use of agonists, which are historically more frequently employed in vision science, have been used to meet the first objective. The compounds chosen for this series of experiments – aspartate, L-AP4, and NMDA – are described in Figure 20.6 and Table 20.2. Figure 20.7 illustrates the experimental results obtained in one animal using NMDA to block activity in third-order neurons (amacrine and ganglion cells). The protocol involves using ERG responses to light stimuli to verify the effect of the agonist, which aids in interpreting the eERG responses.
The results of the experiments using all three glutamate agonists are summarized in Figure 20.8. The interpretation of these results reveals the cellular origin of all three main components of the eERG response. In the presence of L-AP4, which blocks activity in the ON-pathway bipolar cells, the positive phase of the eERG is abolished; therefore, P80 is attributed to this population of neurons. This is consistent with the contribution of this cell type to the ERG, the corneal-positive b-wave (see Figure 20.2). The early negative component of the eERG, N35, is preserved in the presence of both L-AP4 and NMDA. The primary population of retinal cells not affected by either of those compounds is the OFF-pathway bipolar cells, so N35 is attributed to this cell type. In the presence of NMDA, and in the presence of L-AP4, the late negative phase of the eERG is suppressed. Third-order neurons in the ON-pathway are blocked by both of these compounds, and therefore N135 is attributed to this group of cells. With knowledge of the cellular origin of each major response component, changes in the eERG response due to progressive retinal degen-
20. Electrophysiology of Natural and Artificial Vision |
377 |
|
|
|
ERG |
|
|
|
|
|
|
|
|
1500 |
|
|
|
|
|
|
|
|
|
V) |
1000 |
|
|
|
|
|
|
Pre NMDA |
|
|
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
( |
|
|
|
|
|
|
|
|
|
|
Amplitude |
500 |
|
|
|
|
|
|
|
|
|
0 |
|
|
|
|
|
|
Post NMDA |
|||
|
|
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
Pre - Post |
|
||
|
–500 |
|
|
|
|
|
|
|
||
|
|
|
|
|
|
|
|
|
|
|
|
–1000 |
0 |
100 |
200 |
300 |
400 |
500 |
600 |
700 |
800 |
|
–100 |
|||||||||
Time (ms)
|
40 |
|
eERG |
|
|
|
|
Pre NMDA |
|
|
|
|
|
|
|
|
|
|
|
||
|
30 |
|
|
|
|
|
|
|
|
|
V) |
20 |
|
|
|
|
|
|
Post NMDA |
||
( |
|
|
|
|
|
|
|
|
|
|
Amplitude |
10 |
|
|
|
|
|
|
Pre - Post |
|
|
|
|
|
|
|
|
|
|
|||
0 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
–10 |
|
|
|
|
|
|
|
|
|
|
–20 |
0 |
|
200 |
|
400 |
|
600 |
|
800 |
|
–100 |
100 |
300 |
500 |
700 |
|||||
Time (ms)
Figure 20.7. ERG and eERG responses recorded in one experiment employing pharmacological dissection. ERG responses to light stimuli are plotted on the upper axes; eERG responses to electrical stimulation are plotted on the lower axes. Waveforms are shifted vertically for clarity; all pre-stimulus baselines have average values of zero. The top trace in each panel was recorded under baseline conditions, and are typical ERG and eERG response waveforms. The middle trace in each panel was recorded approximately 20 minutes following intravitreal injection of NMDA, which binds to third-order neurons. Removing the corneal-negative contribution of third-order neurons enhances the b-wave in the ERG (consistent with previous reports), and eliminates the N135 component of the eERG. This strongly suggests that third-order neurons are responsible for the N135 eERG response component. The bottom trace in each panel plots the difference waveform (preNMDA minus post-NMDA), and isolates the portion of the baseline response removed by the presence of the drug.
eration or to changes in electrode design or stimulus parameters can now be studied.
The second objective of using pharmacological dissection described above (identifying the direct targets of electrical stimulation) requires particular care
378 Hetling
|
35 |
|
|
|
Baseline |
|
30 |
|
|
|
|
|
|
|
|
|
|
|
25 |
|
|
|
|
V) |
20 |
|
|
|
Aspartate |
( |
15 |
|
|
|
|
Amplitude |
|
|
|
L-AP4 |
|
10 |
|
|
|
||
|
|
|
|
||
5 |
|
|
|
NMDA |
|
0 |
|
|
|
||
|
|
|
|
||
|
|
|
|
|
|
|
–5 |
|
|
|
|
|
–10 |
0 |
200 |
400 |
600 |
|
–200 |
Time (ms)
Figure 20.8. Summary of pharmacological dissection results. Each waveform is the average response recorded in two animals, except the Baseline response, which is the average across all six animals. The lower three waveforms were recorded in the presence of the glutamate agonist indicated. See text for interpretation.
in interpreting results. For example, our results show that in the presence of aspartate (analog/agonist which binds to all post-receptor glutamate receptors), the eERG response to subretinal stimulation using bipolar electrodes is abolished. One interpretation of this result is that the stimulus field subtends only the photoreceptor layer, and the photoreceptors themselves do not contribute significantly to the recorded response. Thus, the recorded response is comprised of contributions of cells post-synaptic to the photoreceptors. An alternate interpretation is that the stimulus field extends deep into the retina and subtends all of the retinal cell types above the electrode. We still conclude that the photoreceptors do not contribute strongly to the eERG, but we cannot rule out the possibility that the secondand third-order neurons, which are all maximally hyperpolarized or depolarized, are unable to respond to the stimulus in the presence of aspartate. Agonists can be useful in identifying cellular origins of eERG components, but antagonists are required to determine the depth of the stimulus field.
Summary
Invasive and non-invasive recording, and pharmacological dissection, have been used for decades to teach us a great deal about the response of the retina, both healthy and diseased, to natural light stimuli. We have described a number of these historical techniques and suggested ways that they may be adapted to teach us about the response of the retina to electrical stimulation. Early results demonstrating eERG recording, and the use of pharmacological dissection to understand this response, have been presented. While the example of eERG waveforms provided represent typical responses, these were obtained under a
20. Electrophysiology of Natural and Artificial Vision |
379 |
very specific set of conditions: one electrode geometry, size, material, configurationand placement relative to the retina, delivering one amplitude and one pulse shape of current to one disease state in one animal model. If any of these parameters were changed, the response would also likely change, suggesting a veritable universe of experiments to be performed. We are directing our experimental efforts by paying attention to the electrode designs and stimulus parameters employed by groups developing prototype prostheses, with the goals of aiding in the interpretation of the response of the retina, at the cellular level, to electrical stimulation, and to the optimization of electrode and stimulus design. A major objective is to learn enough about the input–output relationship of the prosthesisdiseased retina system to support detailed computational models, which will ease the experimental burden, and facilitate rapid prosthesis design. If we do our job correctly, it may be possible some day to determine the physiological state of the patient’s retina, enter this information into a simulation environment, and then create a prosthesis design optimized for the individual. However, I suspect that readers of this book will play a larger role in attaining that goal than the authors.
Acknowledgments. The author would like to thank Dr. Monica Baig-Silva, Dr. Casey Hathcock, Safa Rahmani, and Patrick Axtell for providing many of the figures in this chapter. Funding provided by a Bioengineering Research Grant from The Whitaker Foundation.
References
1.Lewicki MS (1998). A review of methods for spike sorting: the detection and classification of neural action potentials. Network-Computation in Neural Systems. 9(4):R53–78.
2.Buzsaki G (2004). Large-scale recording of neuronal ensembles. Nature Neuroscience 7(5):446–51.
3.Granit R (1933). The components of the retinal action potential in mammals and their relation to the discharge in the optic nerve. J Physiology 77:207–239.
4.Marmor MF 1989). An international standard for electroretinography. Documenta Ophthalmologica. 73(4):299–302.
5.Marmor MF (1995). An updated standard for clinical electroretinography. Archives of Ophthalmology. 113(11):1375–6.
6.Marmor MF & Zrenner E (1999). Standard for clinical electroretinography (1999 update). International Society for Clinical Electrophysiology of Vision. Documenta Ophthalmologica 97(2):143–56.
7.Fishman GA & Sokol S (1990). Electrophysiologic Testing in Disorders of the Retina, Optic Nerve, and Visual Pathway. American Academy of Ophthalmology (Pub).
8.Lam BL (2005). Electrophysiology of Vision: Clinical Testing and Applications. Taylor and Francis Group, Boca Raton (Pub)
9.Pepperberg DR, Birch DG & Hood DC (1997). Photoresponses of human rods in vivo derived from paired-flash electroretinograms. Visual Neuroscience 14:73–82.
10.Hetling JR & Pepperberg DR (1999). Sensitivity and kinetics of mouse rod flash responses determined in vivo from paired-flash electroretinograms. Journal of Physiology (London) 516:593–609.
