 |
 Previous Article | Next Article 
The Journal of Neuroscience, January 1, 2001, 21(1):287-299
Temporal Contrast Adaptation in the Input and Output Signals of
Salamander Retinal Ganglion Cells
Kerry J.
Kim and
Fred
Rieke
Department of Physiology and Biophysics, University of Washington,
Seattle, Washington 98195
 |
ABSTRACT |
We investigated how the light-evoked input and output signals of
salamander retinal ganglion cells adapt to changes in temporal contrast, i.e., changes in the depth of the temporal fluctuations in
the light intensity about the mean. Increasing the temporal contrast
sped the kinetics and reduced the sensitivity of both the light-evoked
input currents measured at the ganglion cell soma and the output spike
trains of the cell. The decline in sensitivity of the input
currents after an increase in contrast had two distinct kinetic
components with fast (<2 sec) and slow (>10 sec) time constants. The
recovery of sensitivity after a decrease in contrast was dominated by a
single component with an intermediate (4-18 sec) time constant.
Contrast adaptation differed for ON and OFF cells, with both the kinetics and amplitude of the light-evoked currents of OFF cells adapting more strongly than those of
ON cells. Contrast adaptation in the input currents of a
ganglion cell, however, was unable to account for the extent of
adaptation in the output spike trains of the cell, indicating that
mechanisms intrinsic to the ganglion cell contributed. Indeed, when
fluctuating currents were injected into a ganglion cell, the
sensitivity of spike generation decreased with increased current
variance. Pharmacological experiments indicated that adaptation of
spike generation to the current variance was attributable to properties
of tetrodotoxin-sensitive Na+ channels.
Key words:
contrast gain control; contrast adaptation; retinal
ganglion cell; adaptation; temporal contrast; retinal signal
processing; nonlinear; contrast response
| |
INTRODUCTION |
A striking property of the visual
system is its ability to operate over an enormous range of lighting
conditions. This flexibility relies on adaptation mechanisms that
permit vision to maintain sensitivity as the light inputs change. The
best appreciated of these is adaptation to the mean light intensity,
although the visual system also adapts to the spatial and temporal
contrast (Blakemore and Campbell, 1969 ; Schieting
and Spillmann, 1987). We focus here on temporal contrast
adaptation: adaptation to the depth of temporal fluctuations in light
intensity about the mean.
Temporal contrast adaptation helps match the range of input signals
that a cell encounters to the range of its outputs. An important source
of temporal contrast under natural conditions is spatial structure in a
visual scene coupled to eye movements. Local eye movements cause cells
with receptive fields that fall in regions of a scene with little
spatial structure to encounter low temporal contrast and cells with
receptive fields that fall in regions with high spatial structure to
encounter high temporal contrast. By matching their sensitivity to the
fluctuations in their inputs, visual neurons can efficiently encode
signals with widely varying temporal structure.
A fast-onset form of temporal contrast adaptation, often called
contrast gain control (Victor, 1987), can directly
influence the computations that visual neurons perform by shaping their response time course. One result is to decrease the response latency of
retinal ganglion cells to moving objects (Berry et al.,
1999). This computational role requires that contrast
adaptation act on a time scale comparable to the integration time of
the cell.
Adaptation to the mean and contrast begin in the retina, although
contrast adaptation has a strong cortical component (Albrecht et
al., 1984; Ohzawa et al., 1985). Adaptation to
the mean includes contributions from the photoreceptors (for review,
see Koutalos and Yau, 1996) and post-receptoral
mechanisms in the retina (for review, see Walraven et al.,
1990). Adaptation to temporal contrast is purely a
post-receptor phenomenon (Sakai et al., 1995;
Smirnakis et al., 1997). Temporal contrast adaptation in
the retina has components that differ in their onset and recovery
kinetics and in their spatial properties (for review, see
Shapley, 1997; Meister and Berry,
1999); this diversity of functional properties suggests a corresponding diversity in the underlying biophysical mechanisms.
Although it is clear that temporal contrast adaptation plays an
important role in visual function and that the retina provides the
first step in this process, little is known about the retinal locations
of contrast adaptation or the mechanisms responsible. We have
investigated the effects of temporal contrast on the inputs and outputs
of salamander retinal ganglion cells and found that (1) contrast
adaptation included contributions from mechanisms acting on the
currents reaching the ganglion cell soma and mechanisms intrinsic to
spike generation in the ganglion cell; (2) the OFF pathway
adapted to changes in contrast more strongly than the ON
pathway; and (3) the different sites of contrast adaptation had
distinct temporal properties.
Some of this work has been published previously in abstract form
(Kim and Rieke, 2000).
| |
MATERIALS AND METHODS |
Experimental procedures
Dissection. All experiments used retinae from larval
tiger salamanders (Ambystoma tigrinum; Charles Sullivan,
Nashville, TN) measuring 10-25 cm in length and maintained at 5°C on
a 12 hour light/dark cycle. Animal procedures followed protocols
approved by the administrative panel on laboratory animal care at the
University of Washington. Experiments were at 20-22°C.
Light-evoked current and voltage responses of retinal ganglion cells
were measured in a flat-mount preparation. After 12-15 h of dark
adaptation, salamanders were rapidly decapitated and pithed, and the
eyes were removed. These initial steps in the dissection were performed
under infrared light (>850 nm) using night-vision goggles (ITT Night
Vision, Roanoke, VA). The remaining steps were performed under a
dissecting microscope equipped with an infrared light source and
infrared-visible image converters (BE Meyers, Redmond, WA). Each eye
was hemisected and placed in bicarbonate Ringer's solution containing
(in mM): 110 NaCl, 2 KCl, 30 NaHCO3, 1.5 CaCl2, 1.6 MgCl2, 10 glucose; pH
was 7.4 when equilibrated with 5% CO2/95%
O2 and osmolarity was 270-275 mOsm. The lens was removed
with tweezers, and the retina was gently peeled away from the pigment
epithelium. A piece of retina ~1 mm in diameter was placed
photoreceptor side down in a recording chamber and held in place with a
coarse nylon mesh glued to a platinum weighting ring. The chamber was
placed on the stage of an upright microscope equipped with an infrared
viewing system. The retina was continuously superfused with bicarbonate
Ringer's solution at a rate of 1-2 ml/min.
The basement membrane often prevented access to the
ganglion cells in the flat-mount preparation. In some cases, pieces of the basement membrane were removed along with the lens during the
dissection; in other preparations we cut a small hole in the basement
membrane using a large suction electrode to hold the membrane and a
sharp quartz microelectrode to tear a hole in it. The area around the
hole was cleaned of debris by suction to expose the cell bodies of
5-10 cells. This procedure provided access to cells in the ganglion
cell layer without excessively stretching the retina.
Immunocytochemistry studies indicate that ~10% of the cells in the
ganglion cell layer are displaced amacrine cells (Watt et al.,
1988). Preliminary experiments indicate that the properties of
contrast adaptation in amacrine cells are similar to those of ganglion
cells. Hence we refer to all cells that were recorded from as ganglion cells.
For experiments on solitary cells, the retina was isolated
in room light, treated with papain (7-14 U/ml; Worthington) for 20 min
and dissociated (Rieke and Schwartz, 1996). Dissociated cells were plated onto glass coverslips pretreated with concanavalin-A. Cells were stored at 4°C for up to 6 h. During recording, the cells were continuously superfused with a HEPES Ringer's solution containing (in mM): 136 NaCl, 2 KCl, 1.5 CaCl2, 10 glucose, 2 NaHCO3, 1.6 MgCl2, 3 HEPES; pH was adjusted to 7.4 with NaOH and osmolarity was 270-275 mOsm. In some experiments, 10-20
mM NaCl was replaced with either
N-methylglucamine (NMG) chloride or pharmacological agents;
this allowed us to keep the Na+ concentration
constant while suppressing K+ currents.
Pharmacological agents were applied by bath superfusion. Isolated
ganglion and amacrine cells were distinguished from other cell types by
their ability to generate Na+ spikes and the
presence of multiple processes extending from their somata. Data were
collected only from cells that were able to generate repetitive action
potentials with a width of <2 msec and an amplitude of at least 40 mV.
The electrical properties of all isolated cells meeting these criteria
were similar, and we did not attempt to distinguish between them.
Light stimuli. Light from three light-emitting diodes (LEDs)
with peak outputs at 470, 570, and 640 nm were combined using a
trifurcated fiber optic and focused on the photoreceptors through the
bottom of the recording chamber. The light stimulus was spatially uniform and illuminated a circular area 1.3 mm in diameter centered on
the cell that was recorded from. During an experiment, each of the LEDs
produced a photon flux at the retina of 1000-3000 photons per
µm2 per sec. The temporal contrast of the light
stimulus was controlled by adding Gaussian fluctuations (bandwidth
0-30 Hz) to the signal controlling the light output of each LED.
Salamanders have five photoreceptor types: ultraviolet,
short-wavelength (S)- and long-wavelength (L)-sensitive cones, and short-wavelength- and long-wavelength-sensitive rods. We used light
stimuli that favored the responses of L cones over the other photoreceptor types. The ultraviolet-sensitive cones account for only
~2% of the cone population (Sherry et al., 1998) and
are insensitive to light of wavelength >450 nm. The
short-wavelength-sensitive rods account for ~7% of the rod
population (Sherry et al., 1998), and their sensitivity
should have been significantly reduced by the high mean light levels
used in our experiments. Finally, the three LEDs were modulated
coherently with relative amplitudes chosen so that the photon
absorption rates in the L cones varied, whereas the absorption rates in
the short-wave-sensitive cones and long-wavelength-sensitive rods
remained constant. We refer to this as an L-cone isolating stimulus.
The temporal contrast, c, of the light stimulus was
defined as the SD of the photoisomerization rate in the L cones,
 L, divided by the mean rate,
mL:
|
(1)
|
L and mL were calculated
from the published spectral sensitivity of salamander L cones
(Makino and Dodd, 1996) and measured spectra of the
light reaching the retina from the three LEDs. Unless stated otherwise,
repetitions of stimuli with the same temporal contrast used different
instantiations of the Gaussian fluctuations and thus were uncorrelated.
Light intensities used in each experiment are given in the Figure
legends in terms of effective photon flux at the wavelength of peak
sensitivity of the L cone.
Patch recording procedure. Voltage and current responses of
ganglion cells were measured using perforated-patch recordings (Rae et al., 1991) and an Axopatch 200B patch-clamp
amplifier (Axon Instruments, Foster City, CA). Patch pipettes were
pulled from borosilicate glass and fire-polished before use. The
pipette tip was filled with an internal solution containing (in
mM): 115 K-aspartate, 20 KCl, 10 HEPES, 1 NMG-EGTA,
0.2 CaCl2; pH was adjusted to 7.2 with NMG-OH and
osmolarity was 260-265 mOsm. In some experiments we replaced 10 mM KCl with K-aspartate; lowering the
Cl concentration did not have a noticeable effect
on contrast adaptation. Pipettes were back-filled with internal
solution with an added 1 mg/ml amphotericin-B (solubilized formulation;
Sigma, St. Louis, MO). Filled pipettes had resistances of 3-6 M ,
and the series resistance during recording was 10-20 M. Ganglion
cells in the intact retina had input resistances between 0.3 and 0.7 G and capacitances between 60 and 90 pF. Solitary spiking cells had resistances between 2.5 and 5 G and capacitances between 15 and 25 pF.
We recorded in both current-clamp and voltage-clamp modes. For
voltage-clamp recordings the cell was held at  50 to 60 mV, resulting in a mean current between 40 and 0 pA. We refer to light-evoked currents measured under voltage clamp at the ganglion cell
soma as "input currents." The input currents reflect synaptic currents with possible shaping by processes in the ganglion cell dendrites. Because salamander ganglion cells are thought to be electrotonically compact for light-driven excitatory input
(Taylor et al., 1996), the measured input currents
should closely resemble the synaptic currents. In current-clamp
recordings, the holding current was set between 0 and 40 pA so that the
cell produced action potentials at a rate of 1-2 Hz. Under these
conditions the mean potential was approximately 50 mV. Junction
potentials for all the solutions used were <5 mV and were not corrected.
Data analysis
In this section we present a model that describes the
transformation between a continuous input signal and the resulting
current or spike response of a ganglion cell. The parameters of the
model were determined by correlating the stimulus with the measured response; these parameters provided a measure of the amplitude and
kinetics of the response of a cell to the stimulus. Comparison of the
parameters measured for two different stimuli, e.g., light stimuli of
different contrasts, allowed us to characterize how retinal ganglion
cells adapted to these stimuli. We used this model in two ways: (1) to
quantify the effect of temporal contrast on the light-evoked currents
and spike trains of a ganglion cell and (2) to quantify the effect of
the variance of current injected into a ganglion cell on the
sensitivity of spike generation. We describe the method in detail below
for adaptation to the contrast of the light input.
Static nonlinearity model. We modeled the dependence of the
response of the cell to a light stimulus using a linear filter followed
by a time-independent or static nonlinearity (Fig.
1A). This was the simplest
model that captured most of the structure in the light responses of
ganglion cells, such as strong rectification of the voltage-clamp
currents (Fig. 1D). A key property of the static
nonlinearity model is that it permitted separation of an instantaneous
nonlinearity in the light response of the cell from a change in
sensitivity produced by contrast adaptation (Sakai et al.,
1995; Chander and Chichilnisky, 1999). The
response of the cell to a continuously varying light input
s(t) was predicted by passing s(t) through a
linear filter f( ) and passing the output of the
filter through a static nonlinearity g:
|
(2)
|
where rpred(t) is the
predicted response (the time-dependent current in voltage clamp or
spiking probability in current clamp). The linear filter
f() provides an estimate of the time course of the
response of the cell to a brief light flash in the presence of the
contrast signal. The static nonlinearity g corrects the output of the filter for the nonlinear relation between the strength of
the light input and the response of the cell. Together
f() and g provide a description of the
steady-state response of the cell to light inputs of a particular
contrast; comparison of these parameters for lights of different
contrasts provides a measure of how the response of the cell adapts to
changes in contrast. Note that Equation 2 describes the steady-state
relationship between the light input and the response of the cell and
does not describe the dynamics of the onset or recovery from contrast
adaptation.
View larger version (27K):
[in this window]
[in a new window]
|
Figure 1.
Static nonlinearity model. A, Schematic
of model. The static nonlinearity model predicts the transformation
between light intensity and ganglion cell response by first linearly
filtering the light intensity and then passing the filter output
through a static nonlinearity. This model was used to analyze the
transformation from light stimulus to input currents in a ganglion cell
(light-to-current). B, Linear filter (light-to-current) for
a voltage-clamped OFF ganglion cell. The filter was
calculated from Equation 3 from six 40 sec epochs during an 8%
contrast light stimulus. The first 20 sec of each epoch were discarded
from the calculation. C, Static nonlinearity corresponding
to filter in B. Each of the dots represents a
single comparison of the measured response and the linear prediction
created by convolving the light input with the filter in B.
The open circles plot the average of 120 adjacent points on
the x-axis. Error bars (SEM) are obscured by the markers.
D, Section of measured and predicted response to an 8%
contrast light stimulus. The thin trace plots the average
current response to 20 repetitions of the same random contrast signal.
The thick trace is the predicted response from the static
nonlinearity model using the linear filter and static nonlinearity
shown in B and C. The responses used to generate
the thin trace were not used in the calculation of the
linear filter and static nonlinearity to guard against overfitting.
Mean light intensity: 4300 photoisomerizations/sec per L cone; holding
potential: 60 mV.
|
|
The linear filter and static nonlinearity were calculated from the
current or spike response of a cell to several minutes of stimulation
at each contrast. Stimuli of different contrasts were interleaved to
prevent mistaking changes in sensitivity over the course of an
experiment for adaptation. The retina does not adapt instantaneously to
contrast changes (see Results and Fig. 5); thus measurements made in
the first 20 sec after a contrast change were discarded to permit
contrast adaptation to approach steady state.
Figure 1, B and C, shows the linear filter and
static nonlinearity from a voltage-clamped OFF ganglion
cell during an 8% contrast light stimulus. The shape of the linear
filter f() was determined by calculating the best linear
predictor of the response of the cell given the light stimulus
(Wiener, 1949). The linear predictor that minimizes the
mean-square error between the measured and predicted response for a
given contrast c is the cross-correlation between the light
stimulus and the response divided by the power spectrum of the
stimulus:
|
(3)
|
where the brackets  ··· indicate an average over multiple
stimuli s from the ensemble described by the power spectrum
S( ), * denotes the complex conjugate,
 () is the Fourier transform of the light stimulus,
and  () is the Fourier transform of the measured
response r(t), defined as () =  dt
exp(it)r(t). The response r(t) is the
measured current for voltage-clamp experiments or a series of impulses
occurring at the spike times for current-clamp experiments. For some
current-clamp experiments, a fluctuating injected current replaced the
light stimulus but the analysis was otherwise identical. Filters
calculated from Equation 3 were restricted to a bandwidth of 0-30 Hz
to match the bandwidth of the light stimuli, unless noted otherwise.
After calculating the linear filter, the static nonlinearity
g was determined by comparing each independent time point of the measured response with the corresponding time point in the linear
prediction obtained by convolving f() with the light
stimulus (Eq. 2). Each small dot in Figure 1C
represents one such comparison, where the y-axis plots the
measured response and the x-axis the corresponding linear
prediction. If the response of the cell were linearly related to the
light stimulus, the points in Figure 1C would be scattered
about a straight line and the extent of scatter would provide a measure
of the noise in the response of the cell. Instead the points clearly
deviate from linearity. To estimate the shape of the nonlinearity,
adjacent points on the x-axis were averaged and plotted as
the open circles in Figure 1C.
The most significant nonlinearity in the input currents of a ganglion
cell was the rectification of positive currents (Fig. 1D);
on average the measured outward currents were much smaller than the
linear prediction. Inward currents were strongly correlated with the
prediction and showed no evidence of saturation, which would cause the
curve to flatten for large negative values of the prediction. The
polarity of the linear filter and shape of the static nonlinearity
indicate that this is an OFF cell. Light decrements lead to
the prediction of inward (excitatory) currents that are not attenuated
by the static nonlinearity. Light increments lead to the prediction of
outward currents that are significantly attenuated.
Figure 1D compares the measured current response with
the prediction calculated from Equation 2 and the linear filter and static nonlinearity shown in Figure 1, B and C.
The measured response is for a section of the experiment that was not
used in calculating the filter or nonlinearity. Noise in the measured
response was reduced by averaging >20 repetitions of an identical
light stimulus. The prediction captured most of the structure in the
measured response, although the correspondence is clearly not exact.
The correlation coefficient between the measured and predicted currents is 66%. In three such experiments the correlation coefficient between
predicted and measured responses ranged from 57 to 66%. These
correlation coefficients are underestimates by at least 15% because of
noise that remained after averaging across repetitions of the light stimulus.
The average correlation coefficient between the predicted and measured
response to a single repetition of the contrast signal was within 6%
of the correlation coefficient between measured responses to different
repetitions of the contrast signal (three cells, each at two
contrasts). Thus the difference between the predicted and measured
response was similar to the difference between independent responses to
the same stimulus. The correlation between predicted and measured
responses did not improve significantly when the linear filter in
Equation 2 was replaced with a second-order Wiener series
(Wiener, 1949; Marmarelis and Marmarelis,
1978). Thus the static nonlinearity model captured most of the
structure in the light-evoked currents of the cell, and more
complicated models did not offer substantial improvements.
We also used the static nonlinearity model to study the ganglion
cell responses in current-clamp experiments (Fig.
2A). In some of these
experiments, we injected current into a ganglion cell to bypass the
retinal circuitry and study spike generation in the ganglion cell
directly. Figure 2B and C, shows the linear filter and static nonlinearity for one such experiment. The filter was
calculated according to Equation 3; it predicts the time course of the
change in firing rate in response to a brief injected current pulse in
the presence of a fluctuating injected current. The static nonlinearity
was determined by convolving the filter with the injected current to
generate a linear prediction of the firing rate and comparing this
prediction with the measured spike train on a point-by-point basis as
in Figure 1C. The shape of the static nonlinearity reflects
the low resting firing rate of the cell and rapid increase in rate as
the cell was depolarized.
View larger version (25K):
[in this window]
[in a new window]
|
Figure 2.
Spike generation in static nonlinearity model.
A, Schematic of transformation of light input to the spike
output of a ganglion cell. We compared two models to predict the
light-to-current transformation. The first (top path) was a
cascade of two static nonlinearity models describing the
light-to-current and current-to-spikes transformations. The second was
a single static nonlinearity model (bottom path). Linear
filter (B) and static nonlinearity (C) describing
the transformation of injected currents into spike outputs for a
current-clamped OFF ganglion cell. The injected current
was Gaussian with a variance of 225 pA2.
The filter and nonlinearity were calculated from two 120 sec epochs;
the first 20 sec of each epoch were discarded. Linear filter
(D) and static nonlinearity (E) describing the
transformation of light inputs into the spike output of a ganglion cell
from the same OFF cell. The filter and static nonlinearity
were calculated from current-clamp responses to five 40 sec epochs of
8% contrast light stimulus. Mean light intensity: 4300 photoisomerizations/sec per L cone; holding current: 40 pA.
|
|
A complete description of the conversion of light inputs into the spike
output of a ganglion cell should include the linear filters and static
nonlinearities for both the light-to-current and current-to-spikes
transformations (Fig. 2A). This somewhat complex model could
be simplified, however, because the kinetics of the filter describing
the conversion of injected current-to-spike outputs was brief compared
with that describing the conversion of light inputs to currents (Figs.
1B, 2B). In this case the current-to-spikes linear filter
can be replaced with a scalar, and the two nonlinearities can be
combined. Thus the conversion of light inputs to ganglion cell spike
trains could be approximated by a linear filter followed by a static
nonlinearity (Fig. 2A). This intuition was confirmed by
comparing predictions generated by the top and bottom cascades of
Figure 2A with the measured light-driven spike train.
Correlations of predictions of either cascade with the measured spike
response were similar to the correlation between independent spike
responses to a repeated light stimulus. Thus we conclude that the two
cascades in Figure 2A are effectively equivalent.
Figure 2, D and E, show the linear filter and
static nonlinearity for 8% contrast light inputs for a current-clamped
OFF cell (bottom cascade of Fig. 2A). The linear
filter describes the time course of the change in firing rate in
response to a brief light flash at time 0 in the presence of the
contrast signal. The shape of the static nonlinearity again reflects
the low resting firing rate of the cell and rapid increase in rate for
depolarizing light inputs.
Contrast adaptation in static nonlinearity model. We used
the static nonlinearity model to study contrast adaptation by comparing the linear filters and static nonlinearities measured for light stimuli
at two or more contrasts (Fig.
3A,B). The model allowed us to
measure the effect of contrast on the kinetics and amplitude of the
light response of a cell. The shape of the static nonlinearity was
contrast independent in the majority of cells except for a scaling of
the x-axis. This is not required by the model, which permits
the static nonlinearities to have arbitrary shapes at different
contrasts; the contrast invariance of the static nonlinearity, however,
greatly simplified our analysis.
View larger version (22K):
[in this window]
[in a new window]
|
Figure 3.
Use of static nonlinearity model to study contrast
adaptation. Light-evoked currents were recorded from an
ON/OFF ganglion cell during interleaved 8 and
24% contrast inputs. A, Linear filters calculated from
Equation 3. B, Static nonlinearities calculated as in Figure
1C. The smooth curve fit to the low-contrast static
nonlinearity is a best fit quintic polynomial. The x-axis
scaling of this curve was varied to fit the high-contrast static
nonlinearity. C, Scaled linear filters. The filter at low
contrast has been multiplied by 1.3, a factor chosen to produce the
best overlap of the static nonlinearities from B. D, Scaled
static nonlinearities. The x-axis of the low-contrast static
nonlinearity has been expanded by a factor of 1.3, corresponding to the
scaling of the linear filters in C. Mean light intensity:
8500 photoisomerizations/sec per L cone; holding potential: 60
mV.
|
|
Changes in the kinetics of the response of the cell with changes
in contrast influenced only the linear filter because the static
nonlinearity is time independent. Kinetic changes attributable to
contrast adaptation were taken as the ratio of the time-to-peak of the
linear filters for the two contrasts.
Changes in the amplitude of the response of the cell with changes in
contrast were taken as the ratio of the maximum amplitudes of the
linear filters for the two contrasts after the filters were corrected
for the static nonlinearity. Contrast-dependent changes in the
amplitude of the response of the cell are potentially shared between
the linear filter and the static nonlinearity. However, the shape of
the static nonlinearity in the majority of cells (28 of 32) was
contrast independent except for a scaling of the x-axis
(Fig. 3D), permitting the effect of contrast to be
restricted to changes in the linear filter. A change in the amplitude
of the linear filter corresponds to a change in the x-axis
scaling of the static nonlinearity. Thus the output of the static
nonlinearity model (Eq. 2) is unchanged if the linear filter and the
x-axis scaling of the static nonlinearity are both multiplied by a factor  (Fig. 3A,B, arrows). By
finding a scale factor that produced an overlap of the static
nonlinearities, contrast adaptation was restricted to changes in the
linear filter (Fig. 3C,D). Cells for which the static
nonlinearities could not be made to overlap (4 of 32) by this procedure
were not analyzed further.
To determine the x-axis scaling producing the best overlap
of the static nonlinearities at two contrasts, a cubic or quintic polynomial was fit to the static nonlinearity measured at low contrast
(Fig. 3B). The x-axis scaling of this curve was
varied to provide the best fit (minimum mean-square error) to the high contrast static nonlinearity (Fig. 3D), and the linear
filter measured at low contrast was multiplied by this scale factor
(Fig. 3C). This fitting procedure assumes that the low and
high contrast static nonlinearities have the same general shape except
for a scaling of the x-axis. Consistent with this
assumption, fitting the high-contrast static nonlinearity with a free
cubic or quintic polynomial rather than a scaled version of the fit to
the low-contrast static nonlinearity decreased the mean-square error
between fit and data by <3%. The error in determining the
x-axis scaling producing the best overlap of the static
nonlinearities and thus the relative amplitudes of the filters was
estimated from the error bars on each point in the static
nonlinearities (Fig. 1C). The errors for the relative filter
amplitudes determined from this procedure were <10% in all of the
cells reported.
| |
RESULTS |
The experiments described below indicate that temporal contrast
adaptation in the retina is mediated in part by mechanisms acting on
the currents reaching the ganglion cell soma (the input currents) and
in part by changes in how the ganglion cell converts its input currents
to action potentials. We begin by describing the properties of contrast
adaptation in the input currents to a ganglion cell and then discuss
how changes in contrast alter spike generation.
Contrast adaptation in the ganglion cell input currents
Contrast adaptation caused the response of a ganglion cell to a
test stimulus to depend on the temporal contrast present when the test
stimulus was delivered. Figure 4 shows an
example. An ON/OFF ganglion cell was
voltage-clamped, and the current-response to a 50 msec light decrement
superimposed on a fluctuating light stimulus of 8 or 24% contrast was
measured. Figure 4A shows the light stimulus itself; Figure
4B shows the average response to the light decrement
delivered during high and low contrast. The response to the decrement
delivered during the low contrast stimulus was larger and took longer
to reach peak than the response to an identical decrement delivered
during the high-contrast stimulus. Similar contrast-dependent changes
in the response to a test stimulus were observed in three cells. Thus
temporal contrast affected the amplitude and kinetics of the input
currents of the ganglion cells.
View larger version (22K):
[in this window]
[in a new window]
|
Figure 4.
Temporal contrast affects the amplitude and
kinetics of the input currents to a ganglion cell. A, Light
stimulus used to test the effect of contrast on the flash response of
the cell. The contrast was alternated between 8 and 24% every 20 sec.
Every 2 sec a 20% L cone-isolating light decrement lasting 50 msec was
superimposed on this alternating contrast signal (inset).
B, Average current responses to the light decrement in the
presence of 8% contrast and 24% contrast for an OFF
ganglion cell. The averages of >20 responses all measured >10 sec
after a change in contrast. Increasing the temporal contrast from 8 to
24% decreased the amplitude and sped the kinetics of the current
response of the ganglion cell. Mean light intensity: 3560 photoisomerizations/sec per L cone; holding potential: 60 mV.
|
|
Time course of onset and recovery of contrast adaptation
The effect of temporal contrast on the light-evoked input currents
of a ganglion cell persisted for several seconds after the light
fluctuations ceased. To measure the recovery of sensitivity after a
period of high contrast, we varied the time between the end of an
adapting contrast signal and a test light step (see stimulus trace in
Fig. 5A). Figure 5A
superimposes responses to a light decrement from an
ON/OFF ganglion cell for several recovery times; the decrement response in the absence of the adapting contrast signal is shown on the far right. The response measured shortly after
the contrast signal ended was approximately half as large as that
without the adapting signal. The response of the cell recovered to its
unadapted value over the course of several seconds after the end of the
contrast signal; this recovery of sensitivity was essentially complete
after 2-3 sec. A similar dependence of the light response on the past
history of the contrast signal was observed in three other experiments.
View larger version (27K):
[in this window]
[in a new window]
|
Figure 5.
Time course of onset and offset of contrast
adaptation. A, Recovery of the sensitivity of an
ON/OFF ganglion cell after an adapting contrast
signal. A 21% contrast stimulus was presented for 10 sec. At time 0, modulation of the light ceased, whereas the mean light intensity
remained constant. After a variable delay, a 15% L cone-isolating
light decrement lasting 400 msec was delivered (see timing
trace). Average responses to this stimulus for each delay are
superimposed, with the response to the light decrement in the absence
of the adapting contrast signal at the far right. Mean light intensity:
2600 photoisomerizations/sec per L cone; holding potential: 60 mV.
B, Onset and recovery of contrast adaptation for continuous
stimuli. The contrast of the light input was switched between 8 and
21% every 20 sec. The time-dependent RMS current response from 14 statistically independent stimuli of this type is plotted; the RMS
current provided a measure of the change in the amplitude of the
response of the cell after a change in contrast (see
Results). RMS values at each time point were calculated across
the 14 stimulus repeats, and the resulting trace was smoothed with a
sliding 200 msec time window. At time 0, the contrast increased from 8 to 21%. After an initial increase, the RMS current gradually declined
during the subsequent 20 sec. The smooth curve fit to the decline in
RMS current after the increase in contrast is a sum of two exponentials
with time constants of 0.6 and 11 sec. At 20 sec the contrast decreased
from 21 to 8% and the RMS current steadily increased. The smooth curve
fit to the increase in RMS current after the decrease in contrast is a
single exponential with a time constant of 4 sec. Mean light intensity:
2900 photoisomerizations/sec per L cone; holding potential: 55
mV.
|
|
The time courses of the onset and offset of contrast adaptation
were measured more quantitatively using the response of the cell to the
fluctuating contrast signal itself. The light stimulus alternated
between 8 and 21% contrast every 20 sec while the input currents to a
ganglion cell were recorded. To determine the time course of the
contrast-induced changes in sensitivity, we measured the time-dependent
root-mean-square (RMS) amplitude of the input currents. The RMS current
included contributions from light-evoked responses and cellular
fluctuations within the retina. Thus the time course of the RMS current
after a change in contrast reflected contrast-induced changes in the
sensitivity of the input currents of a ganglion cell to both light
signals and cellular noise.
Figure 5B shows the time course of the RMS current for an
ON/OFF ganglion cell. When the contrast changed
from 8 to 21% (at time 0), the RMS current was initially large but
declined over time as the sensitivity of the cell decreased. Similarly,
when the contrast changed from 21 to 8% (at 20 sec), the RMS current was initially small but gradually increased as the sensitivity of the
cell recovered from the high-contrast period. The onset and offset of
contrast adaptation had different time courses, as shown by the smooth
exponential fits to the experimental trace. The change in RMS current
after the increase in contrast was fit by a sum of two exponentials,
one with a relatively fast time constant (0.6 sec in Fig.
5B; range 0.5-2.6 sec in seven cells) and the other
relatively slow (11 sec in Fig. 5B; >9.8 sec in seven
cells). On average, the fast component had an amplitude 2.2 ± 0.6 times that of the slow component (mean ± SEM; seven cells) and
thus made a larger contribution to the total contrast adaptation
observed. The change in RMS current after the decrease in contrast at
20 sec was well approximated by a single exponential (time constant 4 sec in Fig. 5B; range 4-18 sec in eight cells). The
contribution of at least two distinct temporal components to the onset
of contrast adaptation suggests that it is mediated by multiple
mechanisms, and the large difference in kinetics suggests distinct
functional roles for these mechanisms (see Discussion).
Adaptation for continuous stimuli
Experiments like those in Figures 4 and 5 show that the input
currents to a ganglion cell adapted to changes in contrast. To
characterize more thoroughly the effect of contrast on the response of
a cell, we investigated how continuously varying stimuli of different
contrasts were encoded by the cell. Figure
6A shows a section of the
current recorded in response to stimuli of 8 and 21% contrast from an
ON/OFF ganglion cell. For each contrast the
transformation from light input to ganglion cell current was characterized as a linear filter (Fig. 6B) followed by a
static nonlinearity (Fig. 6C) (see Materials and Methods for
details). The linear filter provides an estimate of the time course of
the response of the cell to a light flash delivered at time 0 in the presence of the contrast signal. The static nonlinearity is an instantaneous amplitude correction acting on the output of the linear
filter. This model allowed separation of an instantaneous nonlinearity
in the response of the cell from contrast adaptation: i.e., a
contrast-dependent change in how light inputs are transformed to neural
responses (Sakai et al., 1985; Chander and
Chichilnisky, 1999).
View larger version (19K):
[in this window]
[in a new window]
|
Figure 6.
Contrast adaptation in static nonlinearity model.
A, Section of current record from a voltage-clamped
ON/OFF ganglion cell in response to a light
input that alternated between 8 and 21% contrast every 40 sec (see
stimulus trace). B, Linear filters calculated
from the response to 160 sec of record at 8% (thick trace) and 21%
(thin trace) contrast. C, Static nonlinearities
measured by correlating the measured input current with the linear
prediction generated by convolving the filters in B with the
light input at 21% ( ) and 8% ( ) contrast. Error bars (SEM; Fig.
1) are obscured by the data points. The calculation of the linear
filter and static nonlinearity used only the measurements made >20
after a contrast change to allow contrast adaptation to approach steady
state. D, Normalized power spectra of the filters shown in
B. Increasing the contrast reduced the power of the filter
at low temporal frequencies but did not significantly change the
high-frequency power. Mean light intensity: 2900 photoisomerizations/sec per L cone.
|
|
In the majority of the cells that we recorded from (28 of 32), the
static nonlinearities measured at high and low contrast had similar
shapes; Figure 6C shows an example. When the static nonlinearities overlapped, the effect of contrast was restricted to
changes in the linear filter. Thus the transformation from light
input to ganglion cell current could be described as a
contrast-dependent linear filter followed by a contrast-independent
static nonlinearity (Fig. 7). We did not
attempt to characterize contrast adaptation in the four cells for which
the shape of the static nonlinearity changed with contrast.
View larger version (16K):
[in this window]
[in a new window]
|
Figure 7.
Effects of contrast in static nonlinearity model.
The transformation of light input into the input current of a ganglion
cell was described by a linear filter followed by a static
nonlinearity. In 28 of 32 cells, the shape of the static nonlinearity
was independent of the contrast of the light input, and hence the
effect of contrast on the sensitivity was restricted to changes in the
linear filter.
|
|
Figure 6B superimposes linear filters measured at 8 and 21%
contrast, and Figure 6D shows power spectra of the filters
at the two contrasts. The filter measured at high contrast had a smaller amplitude and faster kinetics than that at low contrast, similar to the effects seen for the decrement responses in Figure 4.
These changes were primarily caused by a decrease in the amplitude of
low temporal frequency components of the high-contrast filter (Fig.
6D). The relative amplitudes of the low- and high-contrast filters at temporal frequencies >10 Hz were essentially identical. Similar results were seen in all 28 cells for which the static nonlinearities overlapped. Thus both the amplitude and kinetics of the
input currents of a ganglion cell adapt to temporal contrast.
ON and OFF pathways adapt differently
to contrast
The strength of contrast adaptation in the spike outputs of
ON and OFF salamander ganglion cells differs
(Chander and Chichilnisky, 1999). We found similarly
that the effect of contrast on the amplitude and kinetics of the input
currents differed for different functional types of ganglion cells.
Cells were identified as OFF,
ON/OFF, or ON based on their
response to a 1 sec light increment or decrement; examples are shown in
Figures 8A-C (insets).
OFF cells responded to an increase in light intensity with
an outward (inhibitory) current and to a decrease in intensity with an
inward current (Fig. 8A, inset). ON cells
responded with the opposite polarity (Fig. 8C, inset),
and ON/OFF cells responded to both increases and decreases in light intensity with inward currents (Fig. 8B, inset).
View larger version (25K):
[in this window]
[in a new window]
|
Figure 8.
The strength of contrast adaptation in the
ganglion cell input currents varies with ganglion cell type.
A-C, Linear filters for an OFF (A),
ON/OFF (B), and ON
(C) ganglion cell. The thick traces plot the
linear filters calculated for 8% contrast stimuli, and the thin
traces plot those for 21% contrast. Cells were classified on the
basis of their response to an L cone-isolating 1 sec light increment
superimposed on the mean light intensity used to characterize contrast
adaptation. Examples of responses to a 16% light increment are shown
in the insets. D, Summary of changes in the peak
amplitude of the filter at high contrast relative to that at low
contrast for each cell type. E, Summary of changes in the
time-to-peak at high contrast relative to that at low contrast.
Increasing the contrast had a larger effect on the amplitude and
time-to-peak of the input currents to OFF cells than
ON cells, with ON/OFF cells falling
in between. Error bars are SEM.
|
|
We characterized contrast adaptation in each ganglion cell type using
the static nonlinearity model as in Figure 6. The correlation between
predictions of the model and the measured responses did not differ
systematically among the cell types. Linear filters measured at 8 and
21% contrast for representative cells of each type are shown in Figure
8A-C. In each case, increasing the contrast decreased the
amplitude and the time-to-peak of the linear filter, but these changes
were more pronounced in OFF cells than ON
cells, with ON/OFF cells falling in between.
Figure 8, D and E, summarizes the changes in
amplitude and kinetics for each ganglion cell type. Although the extent
of contrast adaptation varied significantly within a given ganglion
cell type, the OFF cells showed larger changes in amplitude
and kinetics than ON cells. Thus the mechanisms controlling
contrast adaptation in the input currents to ganglion cells differed in
strength according to ganglion cell type. The difference in contrast
adaptation in the input currents of ON and OFF
cells could originate from differences in either the ON and
OFF circuitry before the ganglion cell or in the dendrites of the ganglion cells themselves.
Contrast adaptation in the ganglion cell spike output
As described below, contrast adaptation of the input currents of a
ganglion cell was not sufficient to explain the extent of adaptation in
the output spike trains of the cell, suggesting that intrinsic
properties of the ganglion cell also contributed. By directly injecting
fluctuating currents into a ganglion cell, we found that spike
generation adapted to the current variance and thus contributed to
contrast adaptation.
Spike trains adapt more strongly than input currents
To test for a contribution of intrinsic properties of the ganglion
cell to contrast adaptation, we compared the extent of adaptation in
the input currents of a ganglion cell with that in its output spike
trains. If the ganglion cell itself adapts to contrast, we should see
additional contrast adaptation in its spike output compared with its
input currents. We first delivered 8 and 21% contrast light inputs
under voltage clamp and computed the linear filters (Fig.
9A) and static nonlinearities
relating the light stimulus to the measured input currents. The static nonlinearities overlapped, and hence the effect of contrast on the
sensitivity of the cell was restricted to changes in the linear filter.
We then delivered the same contrast stimuli under current clamp and
studied how the transformation of light inputs into the spike output of
the ganglion cell depended on contrast. In principle, the
light-to-spikes transformation should be described by two successive
static nonlinearity models (Fig. 2A): one for the
light-to-current transformation and another for the current-to-spikes transformation within the ganglion cell itself. However, because the
kinetics of the current-to-spikes transformation was fast (Fig.
9, compare time scale in A and C), this complex
model could be simplified to a single linear filter and static
nonlinearity (see Materials and Methods for a complete description).
Figure 9B shows the linear filters relating the light
stimulus to the spike output of the cell; the static nonlinearities
again overlapped and hence did not contribute to contrast
adaptation.
View larger version (29K):
[in this window]
[in a new window]
|
Figure 9.
Changes in sensitivity of spike
generation contributed to contrast adaptation. All measurements are
from the same OFF ganglion cell. A, Linear
filters describing the transformation of light intensity to ganglion
cell input currents for 8% (thick trace) and 21% contrast
(thin trace). Input currents were measured under voltage
clamp. B, Linear filters describing the transformation of
light intensity to output spikes for 8% (thick trace) and
21% contrast (thin trace). The peak amplitude of the filter
at 21% contrast is 0.52 ± 0.01 times that at 8% contrast. Spike
outputs were measured under current clamp. C, Linear filters
describing the transformation of injected currents to spike outputs for
Gaussian currents with variances of 225 pA2
(thick trace) and 625 pA2 (thin
trace). Current variances were chosen to match those of the input
currents measured for 8 and 21% contrast light signals. The peak
amplitude of the filter at 625 pA2 variance is
0.82 ± 0.01 times that at 225 pA2 variance.
Bandwidth of injected current: 0-50 Hz. D, Predicted linear
filters for 8 and 21% contrast. The filters and static nonlinearities
describing adaptation in the input currents (A) and in spike
generation (C) were used to predict contrast adaptation in
the spike output. The amplitude of the predicted filter at 21%
contrast was 0.55 ± 0.02 times that at 8% contrast, very similar
to the measured effect of contrast shown in B. Holding
potential in voltage clamp: 55 mV; holding current in current clamp:
40 pA. Mean light intensity: 4300 photoisomerizations/sec per L
cone.
|
|
Comparison of the effect of contrast on the input currents of the cell
(Fig. 9A) with that on its output spike trains (Fig. 9B) indicated that the spike outputs adapted more strongly
than the input currents. The contrast-dependent change in the
time-to-peak of the filters for the input currents and spike outputs
was similar, but the change in the amplitude was significantly greater
for the spike output of the cell than for its input currents. In three experiments of this type, the changes in amplitude of the filters describing the light-to-spikes transformation were 2.2 ± 0.1, 1.28 ± 0.04, and 1.33 ± 0.03 times larger than the change
for the light-to-current transformation. Thus contrast adaptation was
significantly greater in the spike output of a cell than in its input
currents. This result indicates that properties intrinsic to the
ganglion cell contributed to contrast adaptation.
Spike generation adapts to the variance of the input current
The contribution of intrinsic properties of the ganglion cell to
contrast adaptation shown in Figure 9 could result either from
voltage-dependent processes in the dendrites that are altered when the
cell is voltage-clamped or from changes in how the cell converts its
input currents to action potentials. Two observations indicate that
changes in the sensitivity of spike generation are responsible.
First, the currents at the cell soma would differ under
current and voltage clamp if voltage-clamping the soma altered
dendritic processing. The experiment of Figure
10 tested for such a difference. We
measured the response of a ganglion cell to a 10 msec light flash under
both voltage clamp (current response in Fig. 10A) and current clamp (post-stimulus time histogram in Fig. 10B,
thick trace). We then injected the current response from
Figure 10A into the cell and measured the evoked spike
response (Fig. 10B, thin trace). If the light-driven
currents reaching the site of spike generation were similar under
current and voltage clamp, the response elicited by the light flash
should be similar to that elicited by the injected current, as was the
case in Figure 10. In three such experiments, the number of spikes
elicited by the light flash and the injected current differed by
<15%. A similar result was obtained in one cell in which we used a
21% contrast light input rather than a flash; >90% of spikes were
consistent between the two conditions. Thus the light-evoked currents
reaching the site of spike generation were similar under current and
voltage clamp, and the additional contrast adaptation in the spike
outputs is unlikely to be mediated by voltage-dependent processing in
the dendrites.
View larger version (12K):
[in this window]
[in a new window]
|
Figure 10.
Light-evoked input currents to a ganglion cell
were similar under current and voltage clamp. A, Average of
18 responses to a 10 msec light flash measured under voltage clamp.
Holding potential: 60 mV. B, Post-stimulus time histograms
of spike responses measured under current clamp. The thick
trace plots the average firing rate to 28 light flashes identical
to those used in A. The thin trace plots the
average firing rate to 26 trials in which the current-response from
A was used as a current-clamp command signal. The similarity
of the light-evoked and current-evoked changes in firing rate indicate
that the light-evoked currents reaching the ganglion cell soma are
similar under current and voltage clamp. Holding current: 5 pA. Flash
strength: 1000 photons µm2 sec.
|
|
Second, direct measurements of the sensitivity of spike generation to
injected current suggested that spike generation contributed to
contrast adaptation. The variance of the input currents to a ganglion
cell increased with increasing contrast (Fig. 6A). To test
whether this change in variance altered how input currents were
converted to spike trains, we injected currents of different variances
into a ganglion cell and computed the linear filter and static
nonlinearity describing the relationship between the injected current
and spike output (see Materials and Methods). Figure 9C
shows the linear filters measured for Gaussian injected currents with
variances of 225 and 625 pA2. The static
nonlinearities overlapped at the two variances and thus the effect of
changes in variance on the sensitivity of spike generation was
restricted to changes in the linear filter. The variances were chosen
to match the variances of the input currents of the cell measured in
response to 8 and 21% contrast light stimuli (288 and 668 pA2). Increasing the variance of the injected
current decreased the amplitude and sped the kinetics of the filter.
Thus the conversion of currents to spike output in the ganglion cell
itself adapted to the variance of the injected current. Changes in the
contrast of the light input and the consequent change in the variance
of the input currents of the ganglion cell should cause a similar change in the sensitivity of spike generation.
Can adaptation of spike generation to the variance of the input current
account for the difference in contrast adaptation in the input currents
of a ganglion cell and its output spike trains? We compared the
measured effect of contrast on the light-to-spikes transformation with
that predicted by combining the effects on the light-to-current and
current-to-spikes transformations. The light stimulus was passed
through the light-to-current linear filter (Fig. 9A) and
associated nonlinearity to predict the light-dependent current. This
predicted current was then passed through the current-to-spikes linear
filter (Fig. 9C) and nonlinearity to generate a prediction of the time-dependent firing rate. The transformation between the light
stimulus and the predicted firing rate was described as a linear filter
followed by a static nonlinearity. Predicted filters for 8 and 21%
contrast are shown in Figure 9D. The predicted ratio of the
amplitudes of the filters is 0.55 ± 0.02 compared with the
measured ratio of 0.52 ± 0.01. Thus the contrast-dependent changes in the spike output of the cell are consistent with the combined effects of contrast on the input currents of the cell and of
the change in current variance on spike generation.
Adaptation of spike generation is mediated by
Na+ channels
To investigate the mechanisms mediating adaptation of the ganglion
cells to the variance of the input current, we studied how injected
currents were transformed into spike trains in isolated cells. These
experiments indicated that adaptation of spike generation to the
current variance is mediated by properties of voltage-activated Na+ channels.
As for cells in the intact retina, increasing the variance of the
current injected into an isolated spiking cell decreased the amplitude
and sped the kinetics of the response of the cell. Figure
11A shows linear filters for
one such experiment; the static nonlinearities for the two current
variances overlapped (data not shown), and thus the effect of the
variance on spike generation was restricted to changes in the linear
filter. Adaptation of spike generation was essentially complete 0.5-1
sec after both increases and decreases in current variance (data not
shown). Thus in the intact retina this mechanism should contribute a
fast onset and offset component of contrast adaptation.
View larger version (17K):
[in this window]
[in a new window]
|
Figure 11.
Na+ channels are required for
adaptation in isolated ganglion cells. A, The variance of
the injected current altered the sensitivity of the cell. Gaussian
noise (bandwidth 0-50 Hz) was injected into a current-clamped isolated
ganglion cell, and the measured voltage response was thresholded to
identify the spike times. The relation between the injected current and
the spiking probability was described using the static nonlinearity
model (Fig. 1). The thick trace plots the filter for
injected currents with a variance of 16 pA2 and the
thin trace for currents with a variance of 144 pA2. The static nonlinearities overlapped, and hence
the effect of changes in variance in the response of the cell were
confined to the linear filter. B, TTX reduced the effects of
current variance on the sensitivity of the cell. The experiment
described in A was repeated in the presence of 100 nM TTX. The relation between the injected current and the
voltage response was described using the static nonlinearity model. The
linear filters were similar for injected currents with variances of 16 pA2 (thick trace) and 144 pA2 (thin trace). C, Collected
results on adaptation to fluctuations in the injected current. Relative
areas of the linear filters are plotted as a function of the relative
variance of the injected current for 16 cells in normal Ringer's
solution () and 7 cells with 100 nM TTX added to the
Ringer's solution (). Filter areas and injected current variances
are normalized to a control variance (usually 16 pA2). Error bars are SEMs. The smooth curve is the
third root of the relative variance.
|
|
The change in firing rate in response to a brief pulse of injected
current had a much shorter duration than the light-dependent current
response of the cell (Fig. 9A,C). Thus the primary
functional effect of the dependence of spike generation on current
variance is to alter the gain with which light-evoked input currents
are converted to spike trains. This gain is determined by the area of
the filter describing the conversion of input currents to spike outputs. To collect results from multiple experiments, we measured the
dependence of the area of the filter on the variance of the injected
current. The closed circles in Figure 11C plot
the relative area of the filter against the relative current variance
for 16 similar experiments. For comparison, increasing the contrast of a light stimulus from 8 to 21% increased the variance of the input currents to a ganglion cell in the intact retina by a factor of 1.8 ± 0.1 (mean ± SEM; 18 cells). The measurements
summarized in Figure 11C indicate that this change in
variance would be expected to decrease the sensitivity of spike
generation by ~20%.
The change in the amplitude and kinetics of the linear filter shown in
Figure 11, A and C, required the activity of
voltage-activated Na+ channels. Figure
11B shows filters for the same cell after the activity of
Na+ channels was suppressed with tetrodotoxin (TTX).
In this case the filters relate the injected current to the voltage
response of the cell because the cell was unable to generate action
potentials. The current-to-voltage transfer in the absence of
Na+ channels did not change substantially with
increasing current variance. The open circles in Figure
11C plot the filter area against the relative current
variance for seven cells in the presence of TTX. The effect of changing
the variance of the injected current was substantially decreased when
Na+ channel activity was suppressed.
The lack of adaptation in the absence of Na+
channels indicates either that the Na+ channels
themselves mediate adaptation or that the voltage excursion produced by
activation of the Na+ channels activates another
conductance that is responsible for adaptation. To distinguish between
these alternatives, we inhibited the other currents that are likely to
be activated during an action potential. Several pharmacological agents
that suppressed Ca2+,
Ca2+-activated, and K+ currents
had little effect on the ability of the cell to adapt to changes in the
current variance. Figure 12A
shows currents measured for a voltage step from 60 to 20 mV before
and after suppressing Ca2+ and K+
currents by adding 200 µM Ba2+, 400 µM Cd2+, and 500 nM apamin
to the external solution. In addition to decreasing the outward current
during depolarization, the action potentials (Fig. 12A,
inset) were wider and had little or no
afterhyperpolarization with Ca2+ and
K+ currents suppressed. Figure 12C shows
the linear filters measured for injected currents of high and low
variance in normal Ringer's solution. Figure 12B
shows filters with Ca2+ and K+
currents suppressed. In both cases the static nonlinearities overlapped, and thus the effect of current variance on the sensitivity of spike generation was restricted to changes in the linear filter. The
change in both the amplitude and kinetics of the filter with the change
in variance of the injected current was indistinguishable with and
without Ca2+ and K+ currents.
View larger version (19K):
[in this window]
[in a new window]
|
Figure 12.
Adaptation to the current variance persisted with
Ca2+ and K+ currents suppressed.
A, Suppression of Ca2+ and
K+ currents. The voltage of an isolated ganglion
cell was stepped from 60 to 20 mV for 200 msec. The
current-responses in HEPES Ringer's solution (thick line)
and in Ringer's solution with 200 µM
Ba2+, 400 µM Cd2+,
and 500 nM apamin added (thin line) are shown.
Suppression of Ca2+ and K+
currents tripled the width at half-height of the average action
potential and reduced the afterhyperpolarization (inset).
B, Adaptation to changes in current variance without
Ca2+ and K+ currents. Gaussian
noise (bandwidth 0-50 Hz) was injected into the cell under current
clamp. Linear filters are shown for injected currents with variances of
16 pA2 (thick trace) and 144 pA2 (thin traces). C,
Adaptation to changes in current variance was similar with and without
Ca2+ and K+ currents. Solid
traces plot filters measured in HEPES Ringer's solution, and the
dashed traces plot those measured with
Ca2+ and K+ currents suppressed.
The filters in the absence of blockers were calculated with identical
stimuli with and without toxins. Holding current was 10 to 16 pA,
chosen so the cell had similar mean firing rates in both
conditions.
|
|
Results from four such experiments are summarized in Figure
13, which plots the relative filter
area with Ca2+ and K+ currents
suppressed against that in control conditions. The points fall close to
the line of unity slope, indicating that adaptation to the variance of
the injected current was similar with and without Ca2+ and K+ currents. Figure 13
also summarizes results from experiments like that in Figure 11 in
which activity of Na+ channels has been suppressed
with TTX. The sensitivity of adaptation to the current variance to
suppression of Na+ channels and the insensitivity to
suppression of Ca2+ and K+
channels suggest that the effect of changes in variance on spike generation is mediated by the Na+ channels
themselves and not another conductance activated during the action
potential.
View larger version (14K):
[in this window]
[in a new window]
|
Figure 13.
Summary of pharmacology experiments on adaptation
to variance of injected current in isolated ganglion cells. The
relative area of the linear filters in the presence of pharmacological
agents is plotted against that in control Ringer's solution for a
collection of cells. In each cell the variance of the injected current
was increased ninefold to probe adaptation. Open symbols are
experiments on cells with various blockers of Ca2+
and K+ currents (Fig. 11): () 5 mM
TEA, 2.5 mM 4-aminopyridine (4-AP), 100 µM
Cd2+, ( ) 20 mM TEA, 10 mM
4-AP, 100 µM Cd2+, and ( ) 200 µM Ba2+, 400 µM
Cd2+, 500 nM apamin. Experiments
were performed with 100 nM TTX () (Fig. 11). In all
cells the static nonlinearities overlapped, and the effect of changing
the variance was restricted to the linear filter. Holding currents were
chosen so that the cell had the same mean firing rate in the presence
and absence of toxin. Error bars were estimated as described in
Materials and Methods.
|
|
| |
DISCUSSION |
We investigated how changes in the temporal contrast of the light
stimulus affected the amplitude and kinetics of the light-evoked currents and spike trains of ganglion cells. These experiments led to
three main conclusions: (1) adaptation to temporal contrast included
contributions from mechanisms acting on the input currents of a
ganglion cell and from spike generation in the ganglion cell itself;
(2) the input currents of OFF cells adapted more strongly than those of ON cells; and (3) the onset of contrast
adaptation had both a fast and a slow temporal component. We discuss
the implications of these results below.
Contrast adaptation in the ganglion cell input currents
Increasing the temporal contrast decreased the amplitude and sped
the kinetics of the light-evoked input currents of a ganglion cell.
Sakai et al. (1995) proposed that contrast adaptation in the inputs to amacrine and ganglion cells could be attributed to a
saturating nonlinearity between the horizontal and amacrine cells. Such
a nonlinearity cannot account for the effect described here because our
analysis specifically removed the effect of such a nonlinearity.
Contrast adaptation in the input currents of a ganglion cell could be
mediated by processes in the retinal circuitry before the ganglion cell
or by processes in the ganglion cell dendrites. The lack of contrast
adaptation in horizontal cells (Sakai et al., 1995) and
the large spatial extent of the signal controlling adaptation
(Donner et al., 1991; Smirnakis et al.,
1997) indicates that the photoreceptors and their output
synapses are not responsible. Thus the mechanisms controlling contrast
adaptation in the input currents must be in the bipolar cell and/or the
bipolar-ganglion cell synapse. Preliminary experiments indicate that
both sites are involved (Kim and Rieke, 2000).
The onset of adaptation in the input currents of a ganglion cell after
an increase in contrast had two distinct temporal components: a fast
component (~1 sec) and a slow component (>11 sec) like that
described by Smirnakis et al. (1997). Both of these
differ from the effectively instantaneous onset of contrast gain
control described by Victor (1987) in cat ganglion
cells. The offset of adaptation after a decrease in contrast was
well-described by a single temporal component. The retinal locations
and mechanisms responsible for these temporal components remain to be determined.
Contrast adaptation in spike generation
Contrast adaptation in the spike outputs of a ganglion cell
included a contribution from spike generation in the ganglion cell
itself. Increasing the contrast of the light input increased the
variance of the input currents of a ganglion cell, which in turn
decreased the sensitivity with which input currents were converted to
spike outputs. Adaptation of spike generation to the current variance
required the activity of voltage-activated Na+
channels but not Ca2+ or K+
channels. Intrinsic cellular mechanisms also contribute to contrast adaptation in visual cortical neurons (Sanchez-Vives et al.,
2000a); however, adaptation in these cells appears to be
largely mediated by K+ channels activated during the
action potential (Sanchez-Vives et al., 2000b). These
distinct biophysical mechanisms may provide a rich repertoire of
adaptation properties in spike generation itself, permitting a cell to
adapt to several distinct properties of its input signals.
A general mechanism that could account for the adaptation of spike
generation in retinal ganglion cells is an increase in the threshold
for spike generation with an increase in the current variance. Larger
current fluctuations, for example, could increase subthreshold
activation of Na+ channels. If the
Na+ channels recover from inactivation relatively
slowly, this increase in subthreshold activation would increase the
fraction of Na+ channels in the inactive state and
raise the threshold for spike generation. The increased threshold would
lower the gain with which input currents were converted into spike outputs.
The ability of retinal ganglion cells to adapt to the variance of their
input currents helps the cells match the range of their input signals
to the range of possible outputs. Many cells face a similar problem of
representing a large range of input signals with a relatively small
dynamic range of outputs. An adaptation mechanism based on properties
of voltage-activated Na+ channels could provide a
simple solution to this general problem.
Differences in contrast adaptation in ON and
OFF pathways
The effect of contrast changes on the input currents
of OFF ganglion cells was two to three times larger than
that on the currents of ON cells. A similar asymmetry
occurs in the spike outputs of ON and OFF
salamander ganglion cells (Chander and Chichilnisky, 1999). Our results indicate that the greater extent of contrast adaptation in the spike outputs of OFF cells is
attributable, at least in part, to a difference in mechanisms in the
ON and OFF pathways before spike generation
rather than differences in spike generation itself. Why do
OFF cells adapt more strongly to contrast? The light-evoked
input currents of OFF ganglion cells were two to three
times larger than those of ON cells, presenting a greater
risk of saturation for high-contrast light inputs in the
OFF pathway. Thus the greater extent of contrast adaptation in OFF cells may be a safeguard against such saturation. In
addition to differences in the strength of contrast adaptation in
ON and OFF cells, the spatial properties in the
two cell types also differ (Smirnakis et al., 1997).
Because the separation of signals into ON and
OFF pathways is a highly conserved aspect of retinal
processing, the ON-OFF asymmetries in contrast
adaptation may be an important determinant of visual sensitivity to
light increments and decrements.
Our results also suggest that the contribution of spike
generation to contrast adaptation relative to that in the input
currents was larger for ON cells than OFF
cells. Changing the contrast of the light inputs from 8 to 21%
decreased the amplitude of the filter between light input and ganglion
cell current by an average of 47% in OFF cells and 17% in
ON cells (Fig. 8). For both ganglion cell types, the total
current variance (noise and light-evoked) during 21% contrast was
approximately twice that during 8% contrast. From the dependence of
the gain of spike generation on the current variance (Fig.
11C), this twofold increase in variance produced a 20%
decrease in the sensitivity of the spike output of a ganglion cell.
Thus in ON cells we estimate that spike generation accounts for approximately half of the total contrast adaptation, whereas in
OFF cells spike generation should make a relatively smaller but still significant contribution to the total contrast adaptation. Because the onset and offset of adaptation in spike generation were
fast, the relative strengths of the fast and slow components of
contrast adaptation in the spike trains of ON and
OFF cells should also differ, with ON cells
showing greater fast contrast adaptation.
Functional roles of contrast adaptation
Adaptation of retinal ganglion cells to changes in
contrast exhibits a diverse set of spatial (Smirnakis et al.,
1997; Benardete and Kaplan, 1997a,
1997b) and temporal
(Shapley and Victor, 1978; Smirnakis et al.,
1997; Benardete and Kaplan, 1999) properties, and these properties differ among ganglion cell types (Shapley and Victor, 1978; Benardete et al., 1992). This
diversity suggests a corresponding diversity in the functional roles of
contrast adaptation.
One role of contrast adaptation is to dynamically adjust
visual sensitivity to match the temporal and spatial structure of the
light inputs (Shapley and Victor, 1978), but contrast
adaptation is not restricted to a scaling of the amplitude of the light
response of a ganglion cell. Instead, changes in contrast affect the
amplitude and kinetics of the light response of the cell
(Shapley and Victor, 1978; Smirnakis et al.,
1997). As contrast increases, the gain and integration time
decrease, protecting the cell from saturation and improving its ability
to encode fast temporal changes in the light inputs. Changes in both
gain and integration time are more pronounced in OFF cells
than ON cells (Fig. 8) (Chander and Chichilnisky, 1999), suggesting that the temporal resolution of the
OFF pathway may be more sensitive to contrast than that of
the ON pathway.
Contrast adaptation has both fast- and slow-onset components (Fig. 5).
Although the slow-onset component seems well suited for matching the
input and output signals of a cell, the fast-onset component may play a
more prominent computational role by shaping the responses of ganglion
cells to single visual objects, such as lowering the latency for
responses to moving objects (Berry et al., 1999). This
computational role of contrast adaptation may predominate in
ON cells because a larger fraction of their contrast
adaptation is of the fast-onset form.
| |
FOOTNOTES |
Received July 19, 2000; revised Oct. 9, 2000; accepted Oct. 16, 2000.
This work was supported by National Institutes of Health (NIH) through
Grant EY-11850 and by the McKnight Foundation. K.J.K. received support
from NIH Training Grant GM07108 and from the University of Washington
Graduate School Fund for Excellence and Innovation Special Fellowship.
We thank Cecilia Armstrong, Divya Chander, E. J. Chichilnisky,
Greg Field, Josh Gold, and Maria McKinley for helpful discussions and
Eric Martinson for excellent technical assistance. We also thank the
anonymous reviewers for helpful suggestions and comments.
Correspondence should be addressed to Kerry Kim, Department of
Physiology and Biophysics, HSB Room G424, Box 357290, University of
Washington, Seattle, WA 98195. E-mail:
kerrykim{at}u.washington.edu.
| |
REFERENCES |
-
Albrecht DG,
Farrar SB,
Hamilton DB
(1984)
Spatial contrast adaptation characteristics of neurones recorded in the cat's visual cortex.
J Physiol (Lond)
347:713-739[Abstract/Free Full Text].
-
Benardete EA,
Kaplan E
(1997a)
The receptive field of the primate P retinal ganglion cell. I: Linear dynamics.
Vis Neurosci
14:169-185[Web of Science][Medline].
-
Benardete EA,
Kaplan E
(1997b)
The receptive field of the primate P retinal ganglion cell. II: Nonlinear dynamics.
Vis Neurosci
14:186-205.
-
Benardete EA,
Kaplan E
(1999)
The dynamics of primate M retinal ganglion cells.
Vis Neurosci
16:355-368[Web of Science][Medline].
-
Benardete EA,
Kaplan E,
Knight BW
(1992)
Contrast gain control in the primate retina: P cells are not X-like, some M cells are.
Vis Neurosci
72:483-586.
-
Berry MJ,
Brivanlou IH,
Jordan TA,
Meister M
(1999)
Anticipation of moving stimuli by the retina.
Nature
398:334-338[Medline].
-
Blakemore C,
Campbell FW
(1969)
On the existence of neurones in the human visual system selectively sensitive to the orientation and size of retinal images.
J Physiol (Lond)
203:237-260[Abstract/Free Full Text].
-
Chander D,
Chichilnisky EJ
(1999)
Contrast adaptation and gain changes in salamander and monkey retina.
Soc Neurosci Abstr
25:1431.
-
Donner K,
Djupsund K,
Reuter T,
Vaisanen I
(1991)
Adaptation to light fluctuations in the frog retina.
Neurosci Res [Suppl]
15:S175-S184[Medline].
-
Kim KJ,
Rieke F
(2000)
Temporal contrast adaptation in the salamander retina (ARVO Abstract).
Invest Ophthalmol Vis Sci
41:4984.
-
Koutalos Y,
Yau KW
(1996)
Regulation of sensitivity in vertebrate rod photoreceptors by calcium.
Trends Neurosci
19:73-81[Web of Science][Medline].
-
Makino CL,
Dodd RL
(1996)
Multiple visual pigments in a photoreceptor of the salamander retina.
J Gen Physiol
108:27-34[Abstract/Free Full Text].
-
Marmarelis PV,
Marmarelis VZ
(1978)
In: Analysis of physiological systems: the white-noise approach. New York: Plenum.
-
Meister M,
Berry MJ
(1999)
The neural code of the retina.
Neuron
22:435-450[Web of Science][Medline].
-
Ohzawa I,
Sclar G,
Freeman RD
(1985)
Contrast gain control in the cat's visual system.
J Neurophysiol
54:651-667[Abstract/Free Full Text].
-
Rae J,
Cooper K,
Gates P,
Watsky M
(1991)
Low access resistance perforated patch recordings using amphotericin B.
J Neurosci Methods
37:15-26[Web of Science][Medline].
-
Rieke F,
Schwartz EA
(1996)
Asynchronous transmitter release: control of exocytosis and endocytosis at the salamander rod synapse.
J Physiol (Lond)
493:1-8[Abstract/Free Full Text].
-
Sakai HM,
Wang JL,
Naka K
(1995)
Contrast gain control in the lower vertebrate retinas.
J Gen Physiol
105:815-835[Abstract/Free Full Text].
-
Sanchez-Vives MV,
Nowak LG,
McCormick DA
(2000a)
Membrane mechanisms underlying contrast adaptation in cat area 17 in vivo.
J Neurosci
20:4267-4285[Abstract/Free Full Text].
-
Sanchez-Vives MV,
Nowak LG,
McCormick DA
(2000b)
Cellular mechanisms of long-lasting adaptation in visual cortical neurons in vitro.
J Neurosci
20:4286-4299[Abstract/Free Full Text].
-
Schieting S,
Spillmann L
(1987)
Flicker adaptation in the peripheral retina.
Vision Res
27:277-284[Web of Science][Medline].
-
Shapley R
(1997)
Retinal physiology: adapting to the changing scene.
Curr Biol
7:R421-423[Web of Science][Medline].
-
Shapley R,
Victor JD
(1978)
The effect of contrast on the transfer properties of cat retinal ganglion cells.
J Physiol (Lond)
285:275-298[Abstract/Free Full Text].
-
Sherry DM,
Bui DD,
Degrip WJ
(1998)
Identification and distribution of photoreceptor subtypes in the neotenic tiger salamander retina.
Vis Neurosci
15:1175-1187[Web of Science][Medline].
-
Smirnakis SM,
Berry MJ,
Warland DK,
Bialek W,
Meister M
(1997)
Retinal processing adapts to image contrast and spatial scale.
Nature
386:69-73[Medline].
-
Taylor WR,
Mittman S,
Copenhagen DR
(1996)
Passive electrical cable properties and synaptic excitation of tiger salamander retinal ganglion cells.
Vis Neurosci
13:979-990[Web of Science][Medline].
-
Victor JD
(1987)
The dynamics of the cat retinal X cell centre.
J Physiol (Lond)
386:219-246[Abstract/Free Full Text].
-
Walraven J,
Enroth-Cugell C,
Hood DC,
MacLeod DIA,
Schnapf JL
(1990)
The control of visual sensitivity.
In: Visual perception: the neurophysiological foundations (Spillmann L,
Werner SJ,
eds), pp 53-101. San Diego: Academic.
-
Watt CB,
Yang SZ,
Lam DM,
Wu SM
(1988)
Localization of tyrosine-hydroxylase-like-immunoreactive amacrine cells in the larval tiger salamander retina.
J Comp Neurol
272:114-126[Web of Science][Medline].
-
Wiener N
(1949)
In: Extrapolation, interpolation and smoothing of stationary time series. Cambridge, MA: MIT.
Copyright © 2001 Society for Neuroscience 0270-6474/01/211287-13$05.00/0
This article has been cited by other articles:
|
|
|
|
 
L. C. Osborne and S. G. Lisberger
Spatial and Temporal Integration of Visual Motion Signals for Smooth Pursuit Eye Movements in Monkeys
J Neurophysiol,
October 1, 2009;
102(4):
2013 - 2025.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
M. VanLeeuwen, I. Fahrenfort, T. Sjoerdsma, R. Numan, and M. Kamermans
Lateral Gain Control in the Outer Retina Leads to Potentiation of Center Responses of Retinal Neurons
J. Neurosci.,
May 13, 2009;
29(19):
6358 - 6366.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
A. J. Camp, C. Tailby, and S. G. Solomon
Adaptable Mechanisms That Regulate the Contrast Response of Neurons in the Primate Lateral Geniculate Nucleus
J. Neurosci.,
April 15, 2009;
29(15):
5009 - 5021.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
D. L. Beaudoin, M. B. Manookin, and J. B. Demb
Distinct expressions of contrast gain control in parallel synaptic pathways converging on a retinal ganglion cell
J. Physiol.,
November 15, 2008;
586(22):
5487 - 5502.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
J. B. Demb
Functional circuitry of visual adaptation in the retina
J. Physiol.,
September 15, 2008;
586(18):
4377 - 4384.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
H. Kondgen, C. Geisler, S. Fusi, X.-J. Wang, H.-R. Luscher, and M. Giugliano
The Dynamical Response Properties of Neocortical Neurons to Temporally Modulated Noisy Inputs In Vitro
Cereb Cortex,
September 1, 2008;
18(9):
2086 - 2097.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
B. Werner, P. B. Cook, and C. L. Passaglia
Complex Temporal Response Patterns With a Simple Retinal Circuit
J Neurophysiol,
August 1, 2008;
100(2):
1087 - 1097.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
D. Kerschensteiner, H. Liu, C. W. Cheng, J. Demas, S. H. Cheng, C.-c. Hui, R. L. Chow, and R. O. L. Wong
Genetic Control of Circuit Function: Vsx1 and Irx5 Transcription Factors Regulate Contrast Adaptation in the Mouse Retina
J. Neurosci.,
March 5, 2008;
28(10):
2342 - 2352.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
M. Diaz-Quesada and M. Maravall
Intrinsic Mechanisms for Adaptive Gain Rescaling in Barrel Cortex
J. Neurosci.,
January 16, 2008;
28(3):
696 - 710.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
E. P. Cook, J. A. Guest, Y. Liang, N. Y. Masse, and C. M. Colbert
Dendrite-to-Soma Input/Output Function of Continuous Time-Varying Signals in Hippocampal CA1 Pyramidal Neurons
J Neurophysiol,
November 1, 2007;
98(5):
2943 - 2955.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
K. S. Gaudry and P. Reinagel
Contrast Adaptation in a Nonadapting LGN Model
J Neurophysiol,
September 1, 2007;
98(3):
1287 - 1296.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
K. S. Gaudry and P. Reinagel
Benefits of Contrast Normalization Demonstrated in Neurons and Model Cells
J. Neurosci.,
July 25, 2007;
27(30):
8071 - 8079.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
D. L. Ringach and B. J. Malone
The Operating Point of the Cortex: Neurons as Large Deviation Detectors
J. Neurosci.,
July 18, 2007;
27(29):
7673 - 7683.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
G.-L. Li, J. Vigh, and H. von Gersdorff
Short-Term Depression at the Reciprocal Synapses between a Retinal Bipolar Cell Terminal and Amacrine Cells
J. Neurosci.,
July 11, 2007;
27(28):
7377 - 7385.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
K. A. Zaghloul, M. B. Manookin, B. G. Borghuis, K. Boahen, and J. B. Demb
Functional Circuitry for Peripheral Suppression in Mammalian Y-Type Retinal Ganglion Cells
J Neurophysiol,
June 1, 2007;
97(6):
4327 - 4340.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
D. J. Margolis and P. B. Detwiler
Different Mechanisms Generate Maintained Activity in ON and OFF Retinal Ganglion Cells
J. Neurosci.,
May 30, 2007;
27(22):
5994 - 6005.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
S. C. Lee and A. T. Ishida
Ih Without Kir in Adult Rat Retinal Ganglion Cells
J Neurophysiol,
May 1, 2007;
97(5):
3790 - 3799.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
A. Kohn
Visual Adaptation: Physiology, Mechanisms, and Functional Benefits
J Neurophysiol,
May 1, 2007;
97(5):
3155 - 3164.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
D. L. Beaudoin, B. G. Borghuis, and J. B. Demb
Cellular Basis for Contrast Gain Control over the Receptive Field Center of Mammalian Retinal Ganglion Cells
J. Neurosci.,
March 7, 2007;
27(10):
2636 - 2645.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
B. N. Lundstrom and A. L. Fairhall
Decoding stimulus variance from a distributional neural code of interspike intervals.
J. Neurosci.,
August 30, 2006;
26(35):
9030 - 9037.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
V. Bonin, V. Mante, and M. Carandini
The statistical computation underlying contrast gain control.
J. Neurosci.,
June 7, 2006;
26(23):
6346 - 6353.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
V. Bonin, V. Mante, and M. Carandini
The Suppressive Field of Neurons in Lateral Geniculate Nucleus
J. Neurosci.,
November 23, 2005;
25(47):
10844 - 10856.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
J. W. Pillow, L. Paninski, V. J. Uzzell, E. P. Simoncelli, and E. J. Chichilnisky
Prediction and Decoding of Retinal Ganglion Cell Responses with a Probabilistic Spiking Model
J. Neurosci.,
November 23, 2005;
25(47):
11003 - 11013.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
M. Carandini, J. B. Demb, V. Mante, D. J. Tolhurst, Y. Dan, B. A. Olshausen, J. L. Gallant, and N. C. Rust
Do We Know What the Early Visual System Does?
J. Neurosci.,
November 16, 2005;
25(46):
10577 - 10597.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
S. J. Slee, M. H. Higgs, A. L. Fairhall, and W. J. Spain
Two-Dimensional Time Coding in the Auditory Brainstem
J. Neurosci.,
October 26, 2005;
25(43):
9978 - 9988.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
M. A. Freed
Quantal Encoding of Information in a Retinal Ganglion Cell
J Neurophysiol,
August 1, 2005;
94(2):
1048 - 1056.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
K. A. Zaghloul, K. Boahen, and J. B. Demb
Contrast Adaptation in Subthreshold and Spiking Responses of Mammalian Y-Type Retinal Ganglion Cells
J. Neurosci.,
January 26, 2005;
25(4):
860 - 868.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
T. Gollisch and A. V. M. Herz
Input-Driven Components of Spike-Frequency Adaptation Can Be Unmasked In Vivo
J. Neurosci.,
August 25, 2004;
24(34):
7435 - 7444.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
C. L. Passaglia and J. B. Troy
Information Transmission Rates of Cat Retinal Ganglion Cells
J Neurophysiol,
March 1, 2004;
91(3):
1217 - 1229.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
X.-J. Wang, Y. Liu, M. V. Sanchez-Vives, and D. A. McCormick
Adaptation and Temporal Decorrelation by Single Neurons in the Primary Visual Cortex
J Neurophysiol,
June 1, 2003;
89(6):
3279 - 3293.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
N. K. Dhingra, Y.-H. Kao, P. Sterling, and R. G. Smith
Contrast Threshold of a Brisk-Transient Ganglion Cell In Vitro
J Neurophysiol,
May 1, 2003;
89(5):
2360 - 2369.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
S. F. Davis and C. L. Linn
Mechanism linking NMDA receptor activation to modulation of voltage-gated sodium current in distal retina
Am J Physiol Cell Physiol,
May 1, 2003;
284(5):
C1193 - C1204.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
K. A. Zaghloul, K. Boahen, and J. B. Demb
Different Circuits for ON and OFF Retinal Ganglion Cells Cause Different Contrast Sensitivities
J. Neurosci.,
April 1, 2003;
23(7):
2645 - 2654.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
K. J. Kim and F. Rieke
Slow Na+ Inactivation and Variance Adaptation in Salamander Retinal Ganglion Cells
J. Neurosci.,
February 15, 2003;
23(4):
1506 - 1516.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
E. J. Chichilnisky and R. S. Kalmar
Functional Asymmetries in ON and OFF Ganglion Cells of Primate Retina
J. Neurosci.,
April 1, 2002;
22(7):
2737 - 2747.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
D. Chander and E. J. Chichilnisky
Adaptation to Temporal Contrast in Primate and Salamander Retina
J. Neurosci.,
December 15, 2001;
21(24):
9904 - 9916.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
F. Rieke
Temporal Contrast Adaptation in Salamander Bipolar Cells
J. Neurosci.,
December 1, 2001;
21(23):
9445 - 9454.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
C. F. Vaquero, A. Pignatelli, G. J. Partida, and A. T. Ishida
A Dopamine- and Protein Kinase A-Dependent Mechanism for Network Adaptation in Retinal Ganglion Cells
J. Neurosci.,
November 1, 2001;
21(21):
8624 - 8635.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
J. B. Demb, K. Zaghloul, L. Haarsma, and P. Sterling
Bipolar Cells Contribute to Nonlinear Spatial Summation in the Brisk-Transient (Y) Ganglion Cell in Mammalian Retina
J. Neurosci.,
October 1, 2001;
21(19):
7447 - 7454.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
TOP
ABSTRACT
INTRODUCTION
