Abstract
To restore meaningful vision
to blind patients, retinal prosthetic devices should elicit spiking patterns in
ganglion (output) cells that match the patterns normally generated by light.
Here we have developed a method to generate high temporal precision spike
trains that can mimic light-elicited patterns. Cell-attached patch clamp
recordings were used to measure spiking responses from individual retinal
ganglion cells in the flat mount rabbit retina. Biphasic electrical stimulus
pulses were delivered epi-retinally by small-tipped platinum-iridium
electrodes. Long duration electrical pulses (≥1 msec) elicited a single
short-latency spike followed by a train of much longer latency spiking which
was variable and depended on pulse amplitude levels. Short duration pulses
(~0.1 msec) also elicited the single short-latency spike but no longer latency
spikes. High frequency trains of short duration pulses also reliably elicited a
single spike per pulse, suggesting that short duration pulse trains can be used
to generate precise temporal patterns of spiking in ganglion cells. The short
pulse stimulus paradigm allows us to mimic physiologically relevant light
evoked responses, e.g., transient or sustained cells, as well as the spiking
patterns that code for changes in intensity and contrast.
1.
INTRODUCTION
Retinitis pigmentosa and age-related macular
degeneration are two of the leading causes of blindness, resulting in more than
100,000 cases per year in the
2. METHODS
We measured spiking responses in ganglion
cells, the output cells of the retina, using cell- attached patch clamp
recordings in the flat mount rabbit retina. The extraction of the retina,
development of the preparation and maintenance of the tissue are described
elsewhere3. Electrical stimulation was delivered by small tipped
platinum-iridium (Pt-Ir) electrodes (Impedance: 10 100 kΩ) and consisted
of biphasic (cathodic first), charge balanced square wave current pulses.
Amplitudes ranged from 1 to 400 µA and pulse durations ranged from 60 µsec to 6
msec. Light responses were measured at the start of each stimulus trial to
ensure stability of the recording setup.
3. RESULTS
A typical spiking response to electrical
stimulation is shown in Figure 1a (solid line). Large transient currents
(horizontal arrows) were recorded which temporally correlate with the onset and
terminations of the individual phases of the stimulus pulse. This electrical artifact that occurred
during the stimulus pulse tended to obscure a neural action potential (spike)
buried in the trace. In TTX, a blocker of neuronal spiking, the response was
different in the region immediately following the onset of the pulse.
Subtracting the TTX response from the control response unmasked a single spike
(Fig. 1b, solid line) which was similar in magnitude and kinetics to light
evoked spikes (Fig. 1b, dotted line). In response to the anodic phase of the
stimulus pulse, there was no difference between the control and TTX responses
(data not shown).
|
|
|
Figure 1: Single spikes are elicited by the leading edge of electrical pulses. (a) Voltage clamp response to a 3 ms 50 µA cathodic pulse (solid trace). Large current transients are seen (horizontal arrows) at pulse onset and offset (timing given by square wave above. The response was sensitive to TTX (dotted trace) but only in the time period immediately following the pulse onset (grey box). (b) Subtracting the TTX record from control extracts the spike which is similar to light-elicited spikes (dotted trace). (c) Voltage clamp response to a 200 µsec 50 µA electrical pulse (solid line). Large transient currents are again seen at pulse onset and offset (horizontal arrows). The response here was also TTX sensitive (dotted trace). (d) Subtracting the TTX record from control extracts the spike. |
The spike was
elicited immediately after the onset of the pulse (Fig. 1b), suggesting that it
was elicited at the leading edge of the pulse.
To test this, we shortened the duration of the stimulus pulse. The
responses to a 0.2 ms pulse under control conditions and in TTX (Fig. 1c, solid
and dotted lines respectively) were subtracted again, revealing the presence of
a spike (Fig. 1d). The response to a short pulse in control conditions was always
triphasic, consisting of an upward, downward and then second upward deflection.
In TTX, the response was always biphasic (no second upward deflection). In
later experiments with short duration pulses the second upward deflection was a
reliable marker to indicate that a spike had been elicited.
|
|
|
Figure 2: Early and late phases of spiking are activated by different mechanisms. (a) Left trace: A second phase of spiking (individual late-phase spikes indicated by asterisks) is elicited after completion of long duration pulses (vertical arrow indicates completion of pulse in a-c). Spikes are eliminated in TTX (right trace). Stimulus pulse waveform and timing is represented at top. (b) Late phase spiking in a different cell (left trace) is completely eliminated by 1 mM CNQX, 10 mM AP7 and 5 mM Curare (right trace). Inset: expanded time scale of the interval between the dotted lines from control (dotted) and excitatory synaptic blocker (solid) traces. The normalized TTX response from a different cell is shown for comparison. The early phase spike remains intact. (c) Late phase spiking initiated by long pulses (left, asterisks) is eliminated by reducing the duration of the pulse to 0.1 ms (right trace). The scale bar in (c) applies to all traces. Cathodic and anodic pulses were 3 ms for long pulses (a-c) and 0.1 ms for short pulse (c). Interval was 5 ms for all pulses. |
Longer duration
pulses elicited longer latency spiking that occurred after completion of the
stimulus pulse spikes (Fig. 2a, left, asterisks). These spikes were also eliminated in the
presence of TTX (Fig. 2a, right). The number and timing of these spikes was
variable and generally increased with increasing pulse amplitude or duration.
Short duration pulses (~0.1 ms) did not elicit any of these spikes at all (Fig.
2c, right). We refer to spiking responses that occur immediately following the
onset of the cathodic pulse as early phase (Figure 1) and spikes that are
elicited following completion of the entire pulse as late phase (Figures 2a-c,
left).
All of the late
phase spiking was eliminated in the presence of a cocktail that blocked
excitatory synaptic input to ganglion cells (Fig. 2b, compare left and right
panels). This indicates that excitatory synaptic input to ganglion cells is
required for the generation of these spikes, and suggests that long-duration
pulses activate presynaptic neurons that release excitatory neurotransmitter.
The early phase spike was not eliminated in the presence of these blockers
(Fig. 2b, inset), indicating that it arises from direct activation of the
ganglion cell.
At higher
stimulation frequencies, short duration pulses continued to elicit one spike
per pulse (Fig. 3a). The responses to 200 individual pulses, applied at 200 Hz
(Fig. 3a, solid lines) are all tri-phasic, indicating the presence of a
neuronal spike in each. Comparison with the TTX response (dotted line) confirms
that every stimulus pulse elicited a spike.
|
|
|
Figure 3. Short electrical pulses reliably elicit one spike per pulse. (a) Response to 200 consecutive 0.1 msec pulses delivered at 200 Hz (overlaid black lines). A second upward deflection is present in each response indicating a single spike is elicited by each pulse. Single TTX record (dotted line) is presented for comparison. (b) Response to increasing amplitude 100 µsec pulses at two different time scales; left traces shows early phase spiking, right traces show late phase spiking. Threshold for early phase spiking is approximately 100-120 µA (shaded region, vertical arrow) and for late phase spiking is 360 µA (shaded region, vertical arrow). |
A single spike per
pulse could be reliably elicited over a wide range of pulse amplitudes as shown
in Figure 3b. For this cell, pulse amplitudes above 120 µA consistently
elicited a single spike. Above 360 µA, late phase spikes were elicited, likely
due to activation of presynaptic neurons.
4. DISCUSSION AND CONCLUSIONS
We have developed
a stimulus paradigm that elicits one neuronal spike per stimulus pulse and is
effective over a wide range of stimulation frequencies and amplitudes. This is
significant because it will allow electrical prosthetic devices to mimic the
spike patterns sent to the brain during normal vision.
Short pulses are
likely to simplify the generation of spatially complex patterns of activity
because they activate only ganglion cells. Longer pulses appear to activate
bipolar cells. Amacrine cells are also activated, either directly or via
synaptic input from bipolar cells4. Activation of these presynaptic
elements elicits long latency excitation as well as long latency inhibition
that can spread over broad spatial regions4. The use of short pulses
avoids the long latency and broad spatial activity.
Short pulses
elicit spiking in ganglion cells using less total charge than long pulses. This
may allow for reduced electrode size thereby generating a more focal response.
References
[1] Friedman, DS, OColmain, BJ et al, The prevalence of age-related macular degeneration in the United States, Arch. Ophthal., 122, 564-572, 2004.
[2] Foundation Fighting Blindness, Pamphlet on Retinitis Pigmentosa, 2004. http://www.blindness.org/RetinitisPigmentosa/
[3] Fried, SI, Münch, TA, Werblin, FS, Mechanisms and circuitry underlying directional selectivity in the retina, Nature, 420(6914), 411-414, 2002.
[4] Wassle, H. Parallel processing in the mammalian retina. Nat Rev Neurosci,, 5(10), 747-757, 2004.
Acknowledgements
We wish to thank Matt McMahon and David Zhou from Second Sight, LLC for technical advice, Tommy Chong for help with data processing. This work was supported by grants from DARPA, NIH and funding from Second Sight, LLC.