Delayed-Rectifier K Channels Contribute to Contrast Adaptation in Mammalian Retinal Ganglion Cells

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Neuron Article

Delayed-Rectiﬁer K Channels Contribute to Contrast Adaptation in Mammalian Retinal Ganglion Cells

Michael Weick1 and Jonathan B. Demb1,2,3,* 1Department of Ophthalmology and Visual Sciences 2Department of Molecular, Cellular and Developmental Biology Kellogg Eye Center, University of Michigan, 1000 Wall Street, Ann Arbor, MI 48105, USA 3Present address: Department of Ophthalmology and Visual Science, Yale University School of Medicine, 300 George Street, Suite 8100, New Haven, CT 06511, USA *Correspondence: [email protected] DOI 10.1016/j.neuron.2011.04.033

SUMMARY tion when the input is strong (Shapley and Victor, 1978; Victor, 1987; Chander and Chichilnisky, 2001). Furthermore, contrast Retinal ganglion cells adapt by reducing their sensi- adaptation enhances information transmission at low contrast tivity during periods of high contrast. Contrast (Gaudry and Reinagel, 2007a). In the retina, a major goal is to adaptation in the firing response depends on both understand how contrast adaptation arises in the circuitry at the presynaptic and intrinsic mechanisms. Here, we level of synapses and intrinsic membrane properties. investigated intrinsic mechanisms for contrast Contrast adaptation has been studied in several cell types of adaptation in OFF Alpha ganglion cells in the in vitro salamander retina, including cone photoreceptors and two of their postsynaptic targets: horizontal and bipolar cells. Neither guinea pig retina. Using either visual stimulation or cones nor horizontal cells adapt to contrast, and thus contrast current injection, we show that brief depolarization adaptation first appears beyond the point of cone glutamate evoked spiking and suppressed firing during subse- release (Baccus and Meister, 2002; Rieke, 2001). Bipolar cells, quent depolarization. The suppression could be the excitatory interneurons that transmit cone signals to ganglion explained by Na channel inactivation, as shown in cells, do adapt to contrast (Baccus and Meister, 2002; Rieke, salamander cells. However, brief hyperpolarization 2001). The bipolar cell’s contrast adaptation is reflected in the in the physiological range (5–10 mV) also suppressed excitatory membrane currents and membrane potential (Vm)of firing during subsequent depolarization. This sup- ganglion cells (salamander: Baccus and Meister, 2002; Kim pression was selectively sensitive to blockers of and Rieke, 2001 and mammal: Beaudoin et al., 2007; 2008; Manookin and Demb, 2006; Zaghloul et al., 2005). However, delayed-rectifier K channels (KDR). In somatic membrane patches, we observed tetraethylammonium- this presynaptic mechanism for contrast adaptation explains only a portion of the adaptation in the ganglion cell’s firing rate sensitive K currents that activated near 25 mV. DR (Kim and Rieke, 2001; Zaghloul et al., 2005; Manookin and Recovery from inactivation occurred at potentials Demb, 2006; Beaudoin et al., 2007; 2008). Thus, the presynaptic hyperpolarized to Vrest. Brief periods of hyperpolar- mechanism combines with intrinsic mechanisms within the ization apparently remove KDR inactivation and ganglion cell to reduce sensitivity during periods of high contrast. thereby increase the channel pool available to sup- In dim light, where signaling depends on rods and rod bipolar press excitability during subsequent depolarization. cells, contrast adaptation depends predominantly on the ganglion cell’s intrinsic mechanism (Beaudoin et al., 2008). In theory, an intrinsic mechanism for contrast adaptation

INTRODUCTION should sense changes in Vm during high-contrast exposure. During high contrast, a ganglion cell’s Vm spans a wide range The visual system adjusts its sensitivity depending on the history of and includes periods of both hyperpolarization (up to 10 mV) light stimulation, a property known as adaptation. In the retina, and depolarization (up to 20 mV) from the resting potential

cellular responses adapt to several statistics of the visual input, (Vrest); the depolarizations are accompanied by increased firing. including the mean light level, variation around the mean The durations of hyperpolarizations and depolarizations are (or contrast), and higher correlations over space and time determineds by the temporal filtering of retinal circuitry, which (Demb, 2008; Rieke and Rudd, 2009; Gollisch and Meister, under light-adapted conditions shows band-pass tuning with 2010). Retinal ganglion cells, the output neurons, adapt to contrast peak sensitivity near 8 Hz; this tuning results in brief periods presented within both well-controlled laboratory stimuli and more of depolarization and firing (50–100 msec) that are themselves natural stimuli (Lesica et al., 2007; Mante et al., 2008). Contrast separated by 100–200 msec (Berry et al., 1997; Zaghloul et al., adaptation improves signal processing because it enables high 2005; Beaudoin et al., 2007). Therefore, an intrinsic mechanism sensitivity when the input is weak and prevents response satura- that suppresses firing at high contrast should recover with

166 Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. Neuron K Channels Mediate Contrast Adaptation

a time course longer than the interval between periods of ﬁring; in across cells, each of which had slightly different Vrest and input this way, ﬁring in one period could activate a suppressive mech- resistance (Rin), we replotted the data as a function of Vm during anism that would affect the subsequent period. the prepulse (Figure 1C). The number of spikes evoked by the

One intrinsic mechanism for adaptation was discovered in iso- test pulse peaked when the prepulse was near the average Vrest lated salamander ganglion cells. A period of depolarization and and was suppressed by prepulses that evoked hyperpolariza- spiking caused Na channel inactivation and resulted in a smaller tions and depolarizations within the physiological range pool of available channels during subsequent periods of excita- (10 mV to +20 mV relative to Vrest)(Figure 1D). tion (Kim and Rieke, 2001; 2003). Channel inactivation recovered with a time constant of 200 msec, and thus one period of depo- Depolarization and Hyperpolarization Suppress larization could influence the next period. During prolonged high- Subsequent Firing to Visual Contrast variance current injection (i.e., a substitute for high-contrast To test the physiological relevance of the prepulses in the above stimulation), a steady pool of inactive channels accumulated, current-injection experiment, we substituted the test pulse with resulting in a tonic suppression of excitability. a visual stimulus: a spot (0.4 mm diameter) that decreased Here, we investigated this Na channel mechanism and also contrast by 100% (i.e., the mean luminance switched to black). investigated additional intrinsic mechanisms for contrast adap- The number of spikes evoked by the contrast stimulus peaked tation in intact mammalian ganglion cells. We focused on when the prepulse was near Vrest and was suppressed by pre- a well-characterized cell type, the OFF Alpha cell, which shows pulses that evoked hyperpolarizations or depolarizations both presynaptic and intrinsic mechanisms for contrast adapta- (Figures 1E, 1G, and 1H). Thus, prepulses evoked by current tion (Shapley and Victor, 1978; Zaghloul et al., 2005; Beaudoin injection at the soma suppress subsequent visually-evoked, et al., 2007; 2008). We studied intact cells in light-sensitive synaptically-driven responses originating at the dendrites. tissue, where channels in both the soma and dendrites could We tested whether the prepulse and associated change in Vm contribute, and where the cell type could be targeted and and firing rate could have fed back through the circuitry (i.e., confirmed based on its soma size, physiological properties, through gap junctions with inhibitory amacrine cells) to suppress and dendritic morphology (Demb et al., 2001; Manookin et al., the visual response. We injected either hyperpolarizing or depo- 2008). In addition to Na channel inactivation, we found a second larizing prepulses in current clamp, as above, and then switched mechanism that contributes to contrast adaptation. This mech- to voltage clamp to record contrast-evoked synaptic currents anism involves a common voltage-gated K channel, the delayed (Vhold near Vrest, 65 mV). Under these conditions, the pre- rectifier (KDR). Brief periods of hyperpolarization in the physiolog- pulses had essentially no effect on the synaptic input (Figure 1F), ical range (10 mV negative to Vrest) suppressed subsequent suggesting that the effect of prepulses on the firing response to excitability during a depolarizing test pulse or contrast stimulus. contrast depends on intrinsic properties of the ganglion cell and The suppressive effect of hyperpolarization lasted for does not involve feedback onto presynaptic neurons. 300 msec. Pharmacological experiments and somatic patch recordings linked the mechanism to KDR channels. Depolarization and Hyperpolarization Suppress Subsequent Excitability for Hundreds of Milliseconds RESULTS Depolarization and hyperpolarization typically stimulate different sets of voltage-gated channels, and so it seemed likely that the Excitability Depends on the History of Both Membrane suppressive effects observed by depolarizing and hyperpolariz- Depolarization and Hyperpolarization ing prepulses depended on separate mechanisms. Indeed, the We first studied intrinsic mechanisms for contrast adaptation in time course of suppression differed after the two classes of pre- OFF Alpha cells by using a paired-pulse current-injection para- pulse. In most experiments below, prepulse current injections digm. The retinal circuit filters the visual input to emphasize were designed to change Vm within the physiological temporal frequencies in the range of 5–10 Hz (Zaghloul et al., range: +400 pA versus 280 pA, which typically evoked +15 mV

2005), and thus the relevant time scale for direct stimulation in versus 10 mV changes in Vm. A depolarizing prepulse sup- our experiments is in the range of 50–100 msec (i.e., a half- pressed firing to the test pulse across the entire test-pulse dura- period of 5–10 Hz). In the basic experiment, a cell was tion, whereas a hyperpolarizing prepulse suppressed only the recorded in current clamp in the whole-mount retina in the pres- late firing to the test pulse (Figure 2A). The time course of the ence of a background luminance and intact synaptic input suppressive effects can be visualized in cumulative firing-rate (see Experimental Procedures). A hyperpolarizing or depolariz- plots for each stimulus (Figure 2A, inset). We performed a similar ing current was injected during a prepulse (100 msec). The analysis in the case where visual contrast replaced the test membrane was allowed to return to Vrest (65 mV) during pulse. In this case, the depolarizing prepulse suppressed firing a 25 msec interpulse interval, and then depolarizing current to contrast across the duration of the response, whereas a hyper- was injected during a test-pulse (+400 pA, 100 msec). Firing to polarizing prepulse suppressed primarily the late firing the test pulse was suppressed by both depolarizing and hyper- (Figure 2B). polarizing prepulses (Figure 1A). Plotting the spike number, We next varied the duration of the prepulse to determine how measured over the duration of the test pulse, as a function of much time was required to generate a suppressive effect. Depo- prepulse current showed a peak in firing around the baseline larizing prepulses suppressed firing even after short prepulse condition (prepulse = 0 pA) and suppression for both negative durations (<5 msec) that evoked only a single spike, whereas and positive prepulse amplitudes (Figure 1B). In order to average hyperpolarizing prepulses suppressed firing only after longer

Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. 167 Neuron K Channels Mediate Contrast Adaptation

AB10 EF Figure 1. Depolarizing and Hyperpolarizing +480 pA V-clamp +480 pA +400 pA +400 pA Prepulses Suppress Firing during a Depola-

Vhold = Vrest rizing Stimulus -160 pA (A) OFF Alpha ganglion cell response to current -280 pA 5 0 0 -280 pA -100% -100% injection. Control shows the response to the test spikes pulse alone (+400 pA). Lower traces show the test control control n=6 pulses preceded by either a depolarizing or 0 0 hyperpolarizing prepulse. Horizontal line before −200 0 200 400 each trace indicates 60 mV. pre−pulse current [pA] (B) Across four trials, the number of spikes during C 100 ms 500 pA 10 the 100-msec test pulse is plotted against the G current injected during the prepulse. Colored +400 pA +400 pA symbols indicate the conditions shown in A. 15 (C) The firing rates from (B) are plotted against the 5 median Vm during the prepulse. spikes spikes 10 (D) Population plot of test-pulse firing rate versus n=8 20 mV prepulse Vm, normalized to the firing rate when the 20 mV −200 0 200 400 0 prepulse equaled the average Vrest (65 mV; pre−pulse current [pA] −80 −70 −60 −50 n = 69 cells). Error bars represent ± 1 standard D H −160 pA −160 pA error of the mean (SEM) across cells (y dimension) 1 1 or ± 1 SD of binned Vm (x dimension). (E) Same format as (A), except the depolarizing 0.8 0.8 test pulse was replaced by a 100% contrast step of a spot stimulus (0.4-mm diameter, centered on 0.6 0.6 the cell body; background equaled the mean 0.4 0.4 luminance). Stimulus timing allowed V to return to n=69 n=8 m normalized spike rate normalized spike rate Vrest prior to the visual response (see Results). 0 100 200 −80 −70 −60 −50 0 100 200 300 −80 −70 −60 −50 Colored bars indicate the period of analysis in (G). time [ms] Vm in pre-pulse [mV] time [ms] Vm in pre-pulse [mV] and (H). (F) Voltage-clamp recordings of the response to the spot stimulus in (E) A cell was stimulated with prepulses in current clamp and then switched to voltage clamp to record synaptic input during the spot. Traces show the average across 6 cells. (G) The firing rate during the contrast response in (E) is plotted against the current injection of the prepulse (n = 8 cells; error bars indicate ± 1 SEM).

(H) Population plot of the contrast response versus prepulse Vm, after the conventions in (D). (n = 8 cells). prepulse durations (>20 msec) (Figure 2B). Thus, the differences (+400, 160 pA). The I-F relationships were relatively linear in the time dependence on prepulse duration suggest that depo- and could therefore be characterized by a slope and an offset larizing and hyperpolarizing prepulses act by different (Figure 3A). Both types of prepulse suppressed the firing by mechanisms. reducing the slope, indicating a reduction in gain (Figure 3B). As discussed above (see Introduction), an intrinsic mechanism However, there were different effects on the offset (Figure 3C). that suppresses firing at high contrast should recover with a time The depolarizing prepulse increased the offset, so that a larger course longer than the interval between periods of firing; in this test-pulse was required to evoke spiking. The hyperpolarizing way, firing in one period could activate a suppressive mecha- prepulse decreased the offset, so that in most cases the firing nism that would affect the subsequent period (i.e., >100 msec near threshold was slightly enhanced by the prepulse, and the for recovery). We therefore examined the recovery of suppres- suppression of firing occurred primarily for the largest test sion after depolarizing or hyperpolarizing prepulses. Both types pulses. Thus, hyperpolarizing prepulses suppress subsequent of prepulse suppressed firing and required >300 msec for firing primarily for strong stimuli, whereas depolarizing prepulses complete recovery (Figure 2C). The fitted half-maximum time suppress subsequent firing for all stimuli. constants for recovery were 182 msec and 195 msec for depo- We repeated the above experiment substituting different larizing and hyperpolarizing prepulses, respectively. Thus, both contrast levels for the test pulse: a spot (1 mm diameter) that hyperpolarization and depolarization can suppress subsequent decreased contrast by variable amounts (9%–100%). We varied excitability and have the appropriate recovery time to contribute the timing of stimulus onset so that lower contrast stimuli to contrast adaptation to physiological stimuli. occurred earlier in time; this ensured that firing at all contrast levels would begin 25 msec after prepulse offset (see Experi- Both Depolarization and Hyperpolarization Reduce mental Procedures; Figure 3D). The contrast response function the Gain of Excitatory Responses measured under control conditions (prepulse, 0 pA) was sup- We tested the influence of depolarizing and hyperpolarizing pre- pressed when the visual stimulus was preceded by either depo- pulses on the complete input-output function of the test-pulse larizing or hyperpolarizing prepulses (Figure 3D). However, the response. We varied test-pulse amplitude to mimic different two forms of suppression differed. The depolarizing prepulse contrast levels (up to +480 pA). The current-firing (I-F) relation- shifted the contrast response function rightward on the log- ship during the test pulse was measured under control condi- contrast axis and thereby suppressed the response to all tions (prepulse, 0 pA) and in the two prepulse conditions contrasts, whereas the hyperpolarizing prepulse suppressed

168 Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. Neuron K Channels Mediate Contrast Adaptation

A B A 100 ms D 100 ms 100 400 pA 0 400 pA 1 400 pA

50 ms 400 pA -100% 50 ms −160 pA 10 400 15 0

spikes −160 pA 200 spikes 50 0.5 200 -100% n=69 0 n=8 0 125 150 200 spikes/s spikes/s 125 150 300

0 0 n=18 normalized spike rate n=10 0 100 200 0 100 200 300 0 0 time [ms] 0 200 400 10 25 50 100 injected current [pA] negative contrast % E C τ D τ B C 400 pA 50 ms 400 pA 50 ms 200 -100% 1 1 0.2 200

0.8 0.8 0 0.1 0 200 0.6 0.6 -35% 0 100 0.4 n=5/5 0.4 n=5/5 offset [pA] spikes/s spikes / [s * pA] normalized spike rate 0 1 10 100 1000 1 10 100 1000 0 τ τ −0.05 100 duration [ms] gap [ms] -12%

Figure 2. The Suppressive Actions of Hyperpolarizing and Depola- control control −160 400pA pA −160 400pA pA rizing Prepulses Show Different Time Courses 0 0 50 100 (A) The firing rate to a test pulse preceded by a depolarizing prepulse (+400 pA, time [ms] red), a hyperpolarizing prepulse (280 pA, blue) or no prepulse (control condition, green). Poststimulus time histograms (PSTHs) show the average of Figure 3. Depolarizing and Hyperpolarizing Prepulses Reduce 69 cells (2-4 trials/cell). Inset: the cumulative firing to the test pulse; early Response Gain response is shown at a finer time scale than the later response (tick marks (A) Firing rate during test pulses of variable current amplitude (+80 to +480 pA) every 25 msec). were suppressed by either depolarizing (+400 pA, red) or hyperpolarizing (B) Same format as (A), except the test pulse was replaced by a 100% prepulses (160 pA, blue). Responses could be fit by a line, defined by an contrast step (stimulus described in Figure 1E). offset on the x axis and a slope. (C) Suppressive effects of various prepulse durations (n = 5 cells). Suppression (B) Upper: shows the average slope across cells in each condition (error bars mediated by depolarizing (red) or hyperpolarizing prepulses (blue) is plotted indicate SD across cells, n = 18). Lower: shows the difference between each relative to the firing in the control condition (no prepulse). Error bars show SEM prepulse condition and the control condition (error bars indicate SEM across across cells (n = 5). Smooth lines show exponential fits to the two activation cells). rates with half-maximal effects at 44 ms (depolarization) or 71 ms (hyperpo- (C) Same format as (B) for the effect of prepulses on the offset of the I-F larization). function. (D) Recovery of the suppressive effects of prepulses for various interpulse (D) Average, normalized firing rate to stimuli of various negative contrasts intervals (t). Smooth lines show exponential fits to the two recovery rates (see (9%–100%) under control conditions (no prepulse, green) and after either Results). Error bars show SEM across cells (n = 5). a depolarizing (red) or hyperpolarizing prepulse (blue). Stimulus timing was designed to evoke firing at approximately the same time after the prepulses mostly the response to high contrasts (Figure 3D). Furthermore, (i.e., lower contrasts presented earlier). Responses are normalized to the maximum firing rate under control conditions. Error bars indicate SEM across the time course of suppression differed for the two prepulses, as cells (n = 10). Stimulus was a 1-mm diameter spot, centered on the cell body is illustrated most clearly at high contrast (Figure 3E). The depo- (dark background). larizing prepulse suppressed the spike rate during the entire (E) Average PSTH to spot stimuli at three contrast levels under control responses, whereas the hyperpolarizing prepulse suppressed conditions (no prepulse, green) and after either a depolarizing (red) or hyper- the spike rate during the late phase of the response. polarizing prepulse (blue) (n = 10).

Visually-Evoked Hyperpolarization Suppresses Firing 2 s and then switched to a clamped state in which dynamic to Subsequent Visually-Evoked Depolarization current injection prevented stimulus-evoked hyperpolarization To demonstrate further the physiological relevance of the (Figure 4A). In a second condition, the cell started in the clamped suppressive effect of hyperpolarization, we used a purely visual state and then switched to the unclamped state. At certain paradigm to generate periods of hyperpolarization and depolar- stimulus frequencies, the response was suppressed in the ization. Sinusoidal contrast modulation of a spot was presented unclamped state, suggesting that visually-evoked hyperpolar- for 4 s. In one condition, the cell responded naturally for the ﬁrst ization normally suppresses ﬁring during subsequent periods

Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. 169 Neuron K Channels Mediate Contrast Adaptation

A 1 of depolarization. The level of response crossed over after 2 s, 0 when the recording state switched on each trial (Figure 4B, −1

contrast S gray line). We quantified the suppressive effect of contrast-evoked hyperpolarization on the firing rate as a function of temporal U U C frequency. For the initial stimulus period, the response was sup- 1 S-1 S+1 pressed across a wide frequency range (Figure 4C). There was a significant decrease in firing in the unclamped state (expressed as a percentage difference from firing in the clamped state) between 2 and 10 Hz (Figure 4E, p < 0.01 at each frequency). Thus, at the switch from mean luminance (i.e., 0% contrast) to C C U 20 mV 1 S-1 S+1 high-contrast modulation, hyperpolarization preceding the initial depolarization was generally suppressive. After 2 s of stimulation, the initial firing rate adapted to a steady rate (illustrated 1 for the 3 Hz stimulus; Figure 4B). At this point, the hyperpolariza- 0.5

I [nA] 0 tions had a smaller suppressive effect on subsequent depolar- 0.2 0.4 1.8 2 2.2 2.4 ization (Figure 4D) and depended more on the temporal time [s] frequency of modulation; suppression was observed in the B 12 2–5 Hz range (Figure 4E; p < 0.01 for 2–3 Hz; p < 0.05 for 5 Hz; 10 n = 10). Thus, the suppressive effect of hyperpolarization on 8 subsequent ﬁring could be evoked by visual contrast stimuli spikes 6 but was frequency dependent.

2 4 6 8 10 12 period Depolarizing PrePulses Suppress Subsequent C D E 20 Excitability by Inactivating Na Channels

60 We next turned to the mechanisms for the suppressive effects of 10 depolarizing and hyperpolarizing prepulses. Based on an earlier 40 0 study in isolated salamander ganglion cells, we hypothesized

spikes/s that Na channel inactivation would explain the suppression after 20 % change −10 depolarizing prepulses and the accompanying spiking (Kim and 0 −20 Rieke, 2001; 2003). We could not measure Na currents directly 0 5 10 0 5 10 0510from intact ganglion cells, and we were not successful in frequency [Hz] frequency [Hz] frequency [Hz] preparing nucleated patches. However, we could gauge Na channel availability in the intact cell by measuring the spike slope Figure 4. Firing Rate Is Suppressed after Periods of Hyperpolariza- (Colbert et al., 1997). tion Evoked by Visual Contrast In response to depolarizing-current injection, the maximum (A) Ganglion cell response to 3-Hz sinusoidal contrast modulation of a spot (0.4-mm diameter, background equal to the mean luminance). Black trace spike slope declined during the burst, presumably because shows the response to the first cycle and to the sixth and seventh cycles. fewer Na channels were available on each subsequent spike Between the sixth and seventh cycles (at 2 s; vertical gray line), the cell (Figure 5A). Furthermore, the initial spike slope during the test switched (‘S’) from an unclamped state (‘U’) to a clamped state (‘C’), where pulse was suppressed after depolarizing prepulses (+400 pA; dynamic current injection prevented stimulus-evoked hyperpolarization. Red Figure 5A). Across cells, the firing rate during the depolarizing trace shows the same cell, but the recording started in the clamped state and prepulse increased roughly linearly with current amplitude (Fig- then switched to the unclamped state. Lowest panel shows injected current. Horizontal line before each voltage trace indicates 60 mV. ure 5B). In the same recordings, the slope of the first action (B) The average number of spikes during each cycle at 3 Hz is plotted over time potential during the test pulse decreased linearly (Figure 5C). (n = 10 cells; error bars here and in C. and D. indicate SEM across cells). Thus, there was an approximately linear relationship between Recordings started in either the unclamped state (black) or clamped state (red) the spike number during the prepulse and the apparent number and then switched states after two sec (gray line). of available Na channels at the beginning of the subsequent test (C) Average firing rate during the first stimulus cycle in the clamped (red, C1)or pulse (i.e., as reflected by the maximum slope of the first action unclamped states (black, U1) over a range of temporal frequencies (n = 10 cells). potential). Consistent with this interpretation, the spike latency (D) Average firing rate during the final stimulus cycle preceding the switch (red, during the test pulse increased with the current amplitude of

CS-1; black, US-1). the prepulse (Figure 5D). However, hyperpolarizing prepulses (E) Points show the percentage decrease in firing during the unclamped state had no consistent effect on spike slope during the test pulse (Fig- compared to the clamped state. Blue points show this comparison for the first ure 5C); comparing within cells, there was only a trend toward stimulus cycle: (U1-C1)/C1; green points show this comparison for the two a higher spike slope after a hyperpolarizing prepulse (p < 0.11), cycles surrounding the switch between states: [(U -C )+(U -C )]/ S-1 S+1 S+1 S-1 which may indicate a small increase in the availability of Na chan- (CS-1 +CS+1). Error bars indicate ± 1 SEM across cells (n = 10). nels. There was a small but significant decrease (p < 0.001, compared within cells) in the spike latency after a strong hyperpolarizing prepulse (Figure 5D, inset); this may be explained by

170 Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. Neuron K Channels Mediate Contrast Adaptation

ABsingle cell (4 repeats) population (n=69) Figure 5. Evidence that Depolarizing Prepulses Suppress control Subsequent Firing Because of Na Channel Inactivation 500 10 (A) The derivative of the voltage response in the control and two pre- 250 pulse conditions. The maximum slope of the initial action potential during the test pulse is suppressed by a preceding depolarizing pre- 0 5 pulse. Data are from the recordings shown in Figure 1A. -250 (B) Firing rate to each level of the prepulse for one cell (left, four trials)

pre-pulse spikes and for the population (right; error bars in B-D show ± 1 SEM across -500 0 C cells, n = 69). (C) Same format as (B) for the maximum slope of the ﬁrst spike during +400 pA 500 600 the test pulse. (D) Same format as (B) for the latency between test-pulse onset and 250 500 the ﬁrst spike. Inset: An expanded scale, a decreased spike latency 0 after the most negative current injections. 400

dVm/dt [V/s] -250

-500 300 max. spike slope [V/s]

D 2 ment for large currents may have been caused by the -160 pA 500 10 block of T-type channels or by a shift in activation 250 threshold for KDR channels (Mayer and Sugiyama, 1988; 1 −200 0 Margolis and Detwiler, 2007). Using a larger range of pre- 0 5 pulse amplitudes (560 to +800 pA) and a larger test

-250 latency [ms] pulse (+800 pA), we still observed both forms of suppres-

-500 0 sion on the test pulse, indicating that Ca channels are not 0 100 200 -200 0 200 400 −200 0 200 400 mediating either effect. Nevertheless, because the ﬁring time [ms] pre-pulse current [pA] properties were altered so dramatically by cobalt, we tested several other speciﬁc blockers.

Several types of Ca or KCa channels were blocked selectively to test for a role in the suppression by hyperpo- an increased availability of Na and voltage-gated Ca channels, larizing prepulses. T-type Ca channels were blocked by mibefra- but we did not investigate this further. dil (10 mM). This blocker is not entirely speciﬁc to Ca channels (Eller et al., 2000) because it also lowers the activation and inac- Suppressive Effect of Hyperpolarizing Prepulses Does tivation threshold for voltage-gated K channels (Perchenet and not Depend on Ih or Ca Channels Cle´ ment-Chomienne, 2000; Chouabe et al., 1998; Yoo et al., We next turned to the mechanism for the suppressive effect of 2008). Nevertheless, there was essentially no impact on the hyperpolarizing prepulses. A potential mechanism could be the suppressive effect of the prepulses (Figures 6AII and BII). As hyperpolarization-activated current Ih, associated with CNG a positive control, an apparent T-type ICa observed under channels (Lee and Ishida, 2007; Gasparini and DiFrancesco, voltage clamp in control conditions was blocked by mibefradil

1997). To test for the involvement of Ih, we measured the effect (Figure 6BII, inset). of prepulses in the presence of the channel blocker ZD7288 We tested KCa channel blockers, including two BK blockers (25 mM). For this drug, and all others described below, we (charybdotoxin, 20 nM; paxilline, 200 nM) and an SK blocker show the drug’s effect on basic physiological properties (apamin; 2.5 mM). Charybdotoxin increased the spike rate compared to the control state for the same sample of cells and evoked by a +400 pA test pulse (Figure 6C). However, both compared to the larger sample of control recordings (n = 69 cells; hyperpolarizing and depolarizing prepulses suppressed firing Figure 6C). ZD7288 had little effect on basic physiological prop- during the test pulse under all conditions (Figures 6BIII–6BV). erties. Furthermore, both hyperpolarizing and depolarizing pre- In each case, we compared drug versus control conditions for pulses continued to suppress subsequent firing to a test pulse normalized firing rates after hyperpolarizations to 75 ± 5 mV; in the presence of the drug (Figure 6BI). These results suggest of all the drugs tested, the largest effect on the hyperpolarizing that Ih does not mediate the suppressive effect of hyperpolariz- prepulse was observed in the presence of charybdotoxin ing prepulses. Indeed, we did not observe a prominent sag (p < 0.12; n = 5; 6BIII). However, because paxilline, a more selec- during the 10 mV hyperpolarizations evoked by the prepulse. tive antagonist of BK channels (Rauer et al., 2000; Grissmer

As a positive control, we found that an apparent Ih observed et al., 1994), had no effect (Figure 6BIV), we questioned the spec- under voltage clamp in control conditions was blocked by ificity of charybdotoxin and proceeded to test the involvement of ZD7288 (Figure 6BI, inset). other K channels. We next tested whether Ca channels were involved in the suppression caused by hyperpolarizing prepulses, either directly Suppressive Effect of Hyperpolarizing Prepulses or indirectly via activation of Ca-activated K (KCa) channels. We Depends on a Tetraethylammonium- and first blocked all voltage-gated Ca channels with cobalt (2 mM). a-Dendrotoxin-Sensitive Delayed-Rectifier K Channel This condition hyperpolarized the membrane (5mV), and rela- We next asked whether the mechanism for the suppressive tively large currents were required to evoke spiking. This require- effect of hyperpolarizing prepulses could be explained by

Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. 171 Neuron K Channels Mediate Contrast Adaptation

A I B I a voltage-gated K (KV) channel. A useful tool for determining KV ZD 7288 n=6 1 channel involvement is the general blocker Tetraethylammonium (TEA). However, adding TEA to the bath caused oscillations in Vm presumably because of altered synaptic release from presyn- 0.6 aptic bipolar and amacrine cells (data not shown). Therefore, 1nA we applied TEA intracellularly by adding it to the pipette solution 25ms (and increased the osmolarity of the extracellular solution by the −80 −70 −60 −50 same amount with glucose). Intracellular TEA caused little II II Mibefradil change in basic properties aside from an increase in spike width 1 n=4 at higher concentrations (Figure 7E). At 20 mM, internal TEA had no effect on the action of the depolarizing prepulse but 0.6 completely suppressed the action of the hyperpolarizing pre- 1nA pulse (Figure 7B). We measured TEA’s suppression of the hyper- 25ms polarizing prepulse effect at six levels of intracellular TEA (i.e., six −80 −70 −60 −50 different pipette solutions tested in different cell groups). A half- III III maximum effect was achieved at 7.4 mM TEA (Figure 7B). This Charybdotoxin n=5 1 result suggests that the suppression of ﬁring by a hyperpolarizing

prepulse is mediated by a KV channel. We explored further the Kv channel involved in hyperpolariza- 0.6 tion-mediated spike suppression through additional pharmaco-

logical experiments. We tested for a role of fast inactivating KV channels (e.g., A type and D type) by adding the blocker 4-AP

normalized spike rate −80 −70 −60 −50 IV IV (Storm, 1993). Initially, we added 4-AP to the extracellular solu- Paxilline n=4 1 tion (1–2 mM), but this produced large oscillations of Vm, presumably mediated by presynaptic effects. We therefore added 0.2 mM 4-AP to the pipette solution. This concentration 0.6 blocked the after-hyperpolarization (AHP) that followed a spike and dramatically increased the spike width (Figure 7E) but had little effect on the suppressive effects of hyperpolarizing or depo- −80 −70 −60 −50 V V larizing prepulses (Figure 7D). Higher concentrations of 4-AP Apamin (2 mM) led to dramatically altered spiking and oscillatory depo- 1 n=4/5 larizations that precluded the main analysis (data not shown). 20 mV Thus, fast-inactivating Kv channels are responsible for the 0.6 after-hyperpolarization that followed a spike but not the suppressive effect of hyperpolarizing prepulses. Consistent with this interpretation, the hyperpolarizing prepulses, under 0 50 100 0 50 100 −80 −70 −60 −50 control conditions, seemed to have no effect on the after-hyper- time [ms] Vm in pre−pulse [mV] polarization that followed a spike. C We tested the involvement of KDR channels by applying the K -specific blocker stromatotoxin (1 mM; Escoubas et al., SW v2 0.4 2002). This drug had no effect on basic physiological properties [ms] 0.3 (Figures 7FI and 7H) and did not block the suppressive effect of V −60 rest −65 [mV] −70 7288, 25 mM), T-type Ca channel (Mibefradil, 10 mM), BK channel (charybdotoxin, 20 nM; Paxilline, 200 nM) and SK channel (Apamin, 2.5 mM). Hori- 60 Rin zontal line before each trace indicates 60 mV. [MΩ] 40 (B) Normalized firing rate versus prepulse Vm in each drug condition (red) versus control condition in the same cells (black; see Figure 1D). Error bars 150 show ± 1 SEM across cells (n = 4–6 per condition). Inset in first row: ZD 7288 FR 100 blocks a hyperpolarization-activated inward current (arrow) under voltage [sp/s] 50 clamp (voltage step from 70 mV to 110 mV and back to 70 mV). Inset in second row: Shortly after wash in, Mibefradil blocks an inward current (arrow) under voltage clamp (activated by a voltage step to 60 mV after a hyper- Paxilline Apamin ZD 7288 Mibefradil polarizing step to 90 mV). Charybdotoxin (C) Basic membrane and firing properties in each drug condition. For each

group of cells, the spike width (SW), resting potential (Vrest), input resistance

Figure 6. The Suppressive Effect of Hyperpolarizing Prepulses Does (Rin) and maximum ﬁring rate to +400 pA current injection (FR) are plotted for Not Depend on Ih,CaV or KCa Channels the control (black) and drug condition (red). Error bars indicate SD across cells. (A) Example responses to +400 pA current injection (0–100 msec) under For each measure, the gray horizontal bar indicates the mean ± 1 SD range control conditions (black) and in the presence of a channel blocker (red): Ih (ZD across a large sample of cells under control conditions (n = 69).

172 Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. Neuron K Channels Mediate Contrast Adaptation

AB FI Stromatotoxin G I 0.2 mM TEA 1 1

0.2 20 mV 1 0.6 20 mV 0.6 2 5 10 n=5

TEA [mM] 20 II −DTX II −80 −70 −60 −50 Vm in pre−pulse [mV] 1

0.6 20 mM TEA 1 n=5 III III 0.6 SB nomalized spike rate spike nomalized 1

0.1 1 10 100 0.6 TEA [mM] normalized spike rate C D n=10 4−AP IV SB + −DTX* IV 1 1

0.6 0.6

n=6/69 n=4

0 50 100 −80 −70 −60 −50 0 50 100 0 50 100 −80 −70 −60 −50 time [ms] Vm in pre−pulse [mV] time [ms] Vm in pre−pulse [mV]

E TEA [mM] H 0.8 SW SW 0.5 0.6 0.4 [ms] 0.4 [ms] 0.3

V −60 V −60 rest −65 rest −65 [mV] −70 [mV] −70

60 100 Rin 50 Rin [MΩ] 40 [MΩ] 50 30 150 200 FR 100 FR 100 [sp/s] 50 [sp/s]

0.2 1 2 5 10 20 4-AP Stromatotoxin −DTX SB SB + −DTX* (n=6) (n=4) (n=7) (n=5) (n=6) (n=4)

Figure 7. The Suppressive Effect of Hyperpolarizing Prepulses Depends on a TEA- and a-Dendrotoxin-Sensitive K Channel (A) Example responses to +400 pA current injection (0-100 msec) with recording solution that contained either 0.2 or 20 mM TEA. Horizontal line in A., C. and F. indicates 60 mV.

(B) upper panel: Normalized ﬁring rate versus prepulse Vm with various levels of TEA added to the pipette solution (0.2 to 20 mM; see Figure 1D). lower panel: Normalized spike rate for conditioning prepulses that depolarized the cell to 55 mV (triangles) or hyperpolarized the cell to 75 mV (circles) plotted against the TEA concentration. The line shows an exponential ﬁt to the spike rates after hyperpolarization (circles; error bars show ± 1 SEM across cells, n = 4-7 per condition). The suppression of spike rate after depolarizing prepulses was not affected by TEA concentration (triangles). (C) Example responses to +400 pA current injection (0-100 msec) with 4-AP added to the pipette solution (200 mM).

(D) Normalized ﬁring rate versus prepulse Vm in the presence of 4-AP. Data from control cells (n = 69) are shown for comparison. Error bars show ±1 SEM across cells. (E) Basic membrane and ﬁring properties for recordings with various levels of either TEA or 4-AP in the pipette solution. Same format as Figure 6C.

(F) Example responses to +400 pA current injection (0–100 msec) under control conditions (black) and in the presence of a Kv2 channel blocker (stromatotoxin,

1 mM), a Kv1 channel blocker (a-dendrotoxin, 70 nM) and synaptic blockers (SB; strychnine, 100 mM; gabazine, 10 mM; CNQX, 100 mM; D-AP-5, 100 mM; L-AP-4, 100 mM). In the fourth row, the SB condition served as the control and a-dendrotoxin (70 nM) was the drug condition (*, the current injection was reduced to +200 pA in the SB+ a-dendrotoxin condition).

(G) Normalized firing rate versus prepulse Vm in each drug condition (red) versus control (black). Error bars show ± 1 SEM across cells (4–10 per condition). (H) Basic membrane and firing properties for each condition in F. and G. Same format as part E. either type of prepulse (Figure 7GI). We next tested the involve- 2001), which have been found in ganglion cells (Pinto and ment of Kv1 channels by applying the specific blocker a-dendro- Klumpp 1998; Ho¨ ltje et al., 2007). This drug increased the toxin (70 nM; Harvey, 2001; Scott et al., 1994), a pore blocker of maximum firing rate (p < 0.001), tended to increase spike width channels that contain Kv1.1, Kv1.2 or Kv1.6 subunits (Harvey slightly (p < 0.11, n = 5) (Figures 7FII and 7H), and lead to mild

Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. 173 Neuron K Channels Mediate Contrast Adaptation

Vm oscillations but did not increase Rin (Figure 7H). In addition, by a test pulse to +30 mV, to activate all available channels (Fig- the drug blocked the action of the hyperpolarizing prepulse ure 8E; n = 3 patches). The voltage dependence of inactivation (p < 0.005 n = 5; Figure 7GII). showed two components, consistent with the presence of at We tested whether the positive effect of a-dendrotoxin could least two channel types (Figure 8F, green line). The rate of be explained by an action at a presynaptic site in the retinal inactivation was high between 120 and 60 mV and again network. Thus, we repeated the experiment after ﬁrst blocking between 35 and 10 mV. Thus, a component of inactivation

ON bipolar cell pathways (L-AP-4), AMPA- and NMDA-type can be removed by hyperpolarization from Vrest. The collected glutamate receptors (CNQX, D-AP-5), and GABA-A (gabazine) pharmacology and somatic patch recordings suggest that a and glycine receptors (strychnine; see Figure 7 legend for drug Kv1-family KDR channel mediates the suppressive effect of concentrations). These blockers, on their own, increased Rin hyperpolarization on subsequent depolarization and firing in (p < 0.005) and tended to increase the spike width (p < 0.07, retinal ganglion cells and thereby contributes to an intrinsic n = 10; Figure 7H) but did not block the effect of either hyperpo- mechanism for contrast adaptation. larizing or depolarizing prepulses (Figure 7GIII). The blockers also induced mild oscillations of Vm. Adding a-dendrotoxin to DISCUSSION the blockers increased the oscillations dramatically, making it difficult to obtain isolated responses to injected current. Further- The contrast adaptation observed in ganglion cell firing exceeds more, the excitability was increased under this condition, neces- that present in the subthreshold Vm or excitatory membrane sitating a lower level of test-pulse current (+200 pA instead currents (Kim and Rieke, 2001; Zaghloul et al., 2005; Beaudoin of +400 pA). Nevertheless, the average of four cells showed et al., 2007; 2008). This discrepancy implicates intrinsic mecha- that a-dendrotoxin blocked the suppressive effect of hyperpola- nisms for adaptation within ganglion cells (Gaudry and Reinagel, rizing prepulses under conditions with most synapses blocked 2007b). Here, we demonstrate two distinct intrinsic mechanisms (p < 0.005, n = 4; Figure 7GIV). for contrast adaptation in the OFF Alpha ganglion cell: Na channel inactivation and removal of delayed-rectifier K channel

Somatic Membrane Patches Show a KDR Current that (KDR) inactivation. Importantly, both mechanisms act within the Could Explain the Suppressive Effect of Membrane physiological range of Vm, and both mechanisms show the Hyperpolarization on Subsequent Firing appropriate time course to suppress visually-evoked ﬁring We recorded currents in somatic membrane patches to charac- during periods of high contrast. Below, we consider the evidence terize the properties of KDR channels (see Experimental Proce- for these two mechanisms, their key properties for evoking dures). Patches were held at 70 mV and then stepped to adaptation, their interaction with each other and with synaptic a series of potentials (110 to +35 mV). From a population of inputs, and their presence in other retinal cell types and neural patches, we recorded outward currents activated at Vm positive circuits. to 35 mV (Figures 8A and 8C). These currents were enhanced after a 20 mV hyperpolarizing prepulse (Figure 8A). The differ- Evidence for Two Distinct Intrinsic Mechanisms ence between the conditions showed an outward current acti- for Contrast Adaptation within an Intact Mammalian vated at Vm positive to 25 mV (Figure 8C). This difference Ganglion Cell current activated rapidly (<0.5 msec at Vholds between 10 One intrinsic mechanism for contrast adaptation, Na channel and +35 mV) and inactivated to half the maximum level in inactivation, was identiﬁed originally in studies of isolated sala- <100 msec (Figure 8A). We tested the effect of TEA in a subset mander ganglion cells of unknown type (Kim and Rieke, 2001; of patches (n = 7) and found that all currents were suppressed. 2003). In these cells, the Na current could be studied directly In most cases, we were not able to record long enough to reverse to characterize activation and inactivation properties. Slow the effect of TEA, although in one case we measured partial recovery from inactivation (>200 msec) explained low-output recovery (30%). Thus, these currents showed characteristic gain at high contrast because of the reduced pool of available activation properties and TEA-sensitivity of KDR channels Na channels, and there was little or no apparent involvement of (Storm, 1993; Baranauskas, 2007). Ca or K channels (Kim and Rieke, 2001; 2003). Our results

For the KDR current activated by preceding hyperpolarization, show a similar Na channel mechanism in the intact OFF Alpha we measured the time course of deactivation. Patches (n = 5) ganglion cell. The maximum slope of the action potential, a proxy were stepped from 70 mV to +30 mV for 10 msec (Figure 8D) measure of Na current, suggested reduced channel availability and then to a series of potentials in the physiological range after periods of depolarization and firing (Figure 5). Furthermore, (85 to 40 mV) without or with a preceding hyperpolarizing the suppressed firing persisted in the presence of multiple step (to 90 mV, 200 msec). The difference in the tail currents blockers of K and Ca channels, consistent with a Na channel between the two conditions show deactivation of those KDR mechanism (Figures 6 and 7). currents activated by the hyperpolarizing step. Deactivation We also identified an intrinsic mechanism for adaptation medi-

(averaged over Vholds between 60 to 40 mV) was complete ated by KDR channels. In intact cells, brief hyperpolarization in 100 msec; the time constant was 35 msec (Figure 8D, inset). within the physiological range (10 mV negative to Vrest) reduced The difference current in Figure 8A suggests that hyperpolar- subsequent firing to a depolarizing test pulse or contrast ization from Vrest increases the availability of KDR channels. To stimulus (Figures 1–3). During contrast stimulation, blocking explore further the voltage dependence of inactivation, we hyperpolarization by dynamic current injection removed the made a prestep to several potentials (120 to +30 mV) followed suppression of subsequent firing and enhanced the firing rate

174 Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. Neuron K Channels Mediate Contrast Adaptation

A D I II II - I I II II - I

+35mV +35mV +30mV +30mV 1 100 ms 100 ms -40mV -40mV -70mV -70mV

-90mV -90mV pA/pF -85mV -85mV -110mV -110mV 0 0100 pA pA 10 1 n=24 pF pF n=5

B −10 0 50 100 −10 0 50 100 −10 0 50 100 time [ms]

n=7 E +30mV +30mV F

-90mV -90mV 100 -120mV + 10mM TEA pA 2 pA pF 100 ms pA/pF 50 20 pA pF 25 pF n=7 0 n=3 −75 −25 0 25

0 50 100 150 0 50 100 150 0 50 100 150 Vm [mV] C time [ms] 60 I 8 II - I

40 6

pA/pF 4

20 pA/pF +TEA 2 +TEA 0 0

−75 −25 0 25 −75 −25 0 25 Vm [mV] Vm [mV]

Figure 8. Hyperpolarization from Rest Activates Delayed-Rectiﬁer Currents in Somatic Patch Recordings (A) Somatic patch recordings of currents activated by voltage steps from 70 mV to potentials ranging from 110 to +35 mV. The step was presented alone (I) or was preceded by a conditioning voltage step to 90 mV (II). The difference traces (II – I) show the currents sensitive to the hyperpolarizing step. Traces show the average across patches (n = 24) normalized by the capacitance (pA/pF). (B) Same format as A. for a subset of patches (n = 7) before (blue) and after (red) adding TEA (10 mM) to the bath medium. (C) The voltage-current relationship for series I and the difference curve, II – I. Points show the average during the times indicated by the gray bars in A. and B. The current evoked by positive steps from 70 mV activated near 35 mV and was TEA sensitive. The current from the difference trace (II – I) activated near 25 mV and was TEA sensitive. (D) Voltage dependence of deactivation (n = 5 patches). Patches were stepped from 70 mV to +30 mV and then to potentials ranging from 85 to 40 mV; the steps to these extreme values are illustrated. The initial step to +30 mV was presented alone (I) or was preceded by a conditioning voltage step to 90 mV (II). The difference traces (II-I) show the currents sensitive to the hyperpolarizing step. Inset: the average tail current at postpulse potentials between 60 and 40 mV. Red line shows a ﬁtted exponential with a time constant of 35 msec. (E) Steps to +30 mV were preceded by prolonged steps to potentials ranging from 120 to +30 mV. Both black traces show steps from 90 to +30 mV, indicating the stability of the recording over time (averaged over 3 patches). (F) The voltage-dependence of activation and inactivation of K currents measured in E. Activation (black line) and inactivation (green line) were measured over the time windows indicated in E. Inactivation of the K current can be removed at potentials negative to the typical 65 mV resting potential (left of the gray vertical line).

(Figure 4). The hyperpolarization-induced suppression of subse- The study of intrinsic mechanisms for adaptation in intact cells quent firing was blocked by internal TEA and by external a-den- and circuits is challenging and requires complementary experi- drotoxin (Figures 6 and 7). Furthermore, somatic membrane mental approaches. Some of these approaches yielded unex- patches showed characteristic properties of a KDR current (acti- pected results worth mentioning. For example, external a-den- vation at 25 mV) and steady-state inactivation at Vrest could drotoxin increased firing significantly, whereas internal TEA be removed by hyperpolarization (Figure 8). The collected results (20 mM) did not (Figures 6 and 7), even though TEA is a more support a KDR mechanism whereby hyperpolarizations from Vrest general blocker of K channels. However, TEA increased the increase the available KDR channel pool and suppress firing spike width substantially. The increased spike width increases during subsequent depolarization. the time that unblocked K channels could be activated and

Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. 175 Neuron K Channels Mediate Contrast Adaptation

also possibly leads to increased Na channel inactivation. Thus, input is conveyed primarily by the AII amacrine cell (Manookin unintended effects on K and Na channels may have counter- et al., 2008; Murphy and Rieke, 2006; Mu¨ nch et al., 2009; van acted any increase in firing rate caused by specifically blocking Wyk et al., 2009). Suppressing bipolar cell glutamate release a-dendrotoxin-sensitive KDR channels. Another unexpected cannot generate substantial hyperpolarization, because the result was that hyperpolarizing current had no effect on subse- release is rectified (Demb et al., 2001; Liang and Freed, 2010; quent firing to weak visual stimulation but enhanced slightly Werblin, 2010). Thus, direct synaptic inhibition serves not only the firing to weak current injection (Figure 3). The distinct effects to hyperpolarize Vm and counteract simultaneous depolarizing may be explained by the different time courses of the stimuli: the inputs (Mu¨ nch et al., 2009) but also leads to a short-term memory current injection was a square-pulse, whereas the low-contrast of synaptic activity that influences excitability on a physiologi- synaptic input was necessarily more sluggish due to the filtering cally-relevant time scale. by retinal circuitry. However, in general, similar results were eli- Contrast adaptation in the ganglion cell firing rate is routinely cited by protocols that used either current injection or synaptic quantified with a linear-nonlinear (LN) cascade model, in which stimulation as the test stimulus (Figures 1, 2, and 4). the adaptation of an underlying linear filter is separated from the nonlinearity imposed by the firing threshold (Chander and

Critical Properties of KDR Channels for Contrast Chichilnisky, 2001; Kim and Rieke, 2001; Zaghloul et al., 2005). Adaptation While this model is useful for quantifying adaptation and explains Somatic membrane patch recordings showed several properties much of the variance in the ﬁring response (Beaudoin et al., of KDR currents that could explain their role in contrast adapta- 2007), it clearly confounds several underlying mechanisms. For tion. First, the channels activate at voltages traversed during local contrast stimulation, there are two major inputs to the an action potential (activation at 25 mV). Second, channel inac- OFF Alpha cell, bipolar input and AII amacrine cell input. The tivation at Vrest could be removed by hyperpolarization (Figure 8). adaptation in these inputs is distinct; both inputs show reduced Thus, a period of hyperpolarization would increase the number of gain at high contrast, but the excitatory inputs exhibit a relatively available channels, which could then be activated during larger speeding of response kinetics (Beaudoin et al., 2008). The a subsequent burst of ﬁring. The initial spikes in the burst would two intrinsic mechanisms also show distinct tuning: the hyperpo- be largely unaffected, but channels opened during these initial larization-induced suppression contributes primarily at low spikes would suppress subsequent spikes (Figure 2) because temporal frequencies (Figure 4). Each synaptic or intrinsic mech-

KDR deactivation is relatively slow (Figure 8D) compared to the anism for contrast adaptation contributes a 10%–20% gain typical interspike interval during the initial spike burst (Figure 1); reduction, so that the total reduction at high contrast is 40%–

Kv1 channels also contribute to the interspike interval in neo- 50% (Kim and Rieke, 2001; Beaudoin et al., 2007). Exploring cortical pyramidal cells (Guan et al., 2007). Furthermore, our detailed interactions between each of the mechanisms requires whole-cell recordings showed that hyperpolarization for periods a more sophisticated biophysical model of the ganglion cell. of >100 msec were most effective at suppressing subsequent firing (Figure 2). Therefore, the suppression after hyperpolariza- Comparison to Other Systems and Cell Types tion should be tuned to temporal frequencies that both drive To our knowledge, the paired-pulse current-injection paradigm hyperpolarization for 100 msec periods or longer (i.e., 5 Hz or used here has not been explored extensively with hyperpolariz- lower) and drive a strong burst of firing during subsequent depo- ing prepulses in the physiological range. However, certain para- larization (i.e., above 1 Hz). This tuning was confirmed in contrast digms were similar and could be compared to ours. For example, stimulation experiments in which hyperpolarization-induced in parasympathetic neurons (Fukami and Bradley, 2005), neo- suppression was maximal in the 2–5 Hz range (Figure 4). striatal neurons (Nisenbaum et al., 1994) and striatal neurons (Mahon et al., 2000) a period of hyperpolarization lead to Interaction between Synaptic and Intrinsic Mechanisms decreased sensitivity and reduced spiking to subsequent depo- for Contrast Adaptation larization. In these cases the suppression was explained by the

Under physiological conditions, there are opportunities for the activation of KV currents. However, the mechanism was appar- two intrinsic mechanisms to interact. For example, hyperpolar- ently different from the one demonstrated in ganglion cells, ization from Vrest could remove both KDR and Na channel inacti- because the previous effects were blocked by low concentra- vation. These two actions could have opposing effects on ﬁring tions of 4-AP (<0.2 mM; compare to Figures 7C and 7D). Exper- during subsequent depolarization. However, the increased Na iments in pyramidal neurons of sensorimotor cortex (Spain et al., channel availability induced by a brief 10 mV hyperpolarization 1991) and the Hippocampus (Nistri and Cherubini, 1992) also seemed to be minor: the spike slope was barely enhanced by showed a suppressive effect of hyperpolarization on subsequent prior hyperpolarization, although the spike latency was excitability. These effects may have been similar to the one decreased somewhat (Figure 5). Thus, physiological levels of shown here, because they were blocked by relatively high hyperpolarization studied here appear to affect primarily the concentrations of 4-AP (>1mM) and TEA (20mM). However, in

KDR channels. Furthermore, the AHP after each spike seemed most cases, the previous experiments hyperpolarized cells insufﬁcient for substantially removing inactivation of KDR beyond the physiological range (to 90 mV). Further experi- currents that are inactivated at rest. Rather, inhibitory synaptic ments could determine whether milder hyperpolarization has input to the ganglion cell would be necessary for prolonged a suppressive effect similar to the one shown here.

(>100 msec) hyperpolarization of sufﬁcient magnitude (5– The suppressive actions of Na and KDR channels studied here 10 mV; Figure 4). For the OFF Alpha ganglion cell, such inhibitory can be distinguished from other mechanisms for adaptation of

176 Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. Neuron K Channels Mediate Contrast Adaptation

firing rate studied in cortical cells. For example, an adaptation of (The Mathworks; Natick, MA) to generate current-injection protocols and to the firing mechanism was observed under conditions of visual analyze responses. Results are from 177 OFF Alpha cells (Vrest, 65.1 ± 2.1 mV, n = 69). During current-clamp recordings, the bridge (10–20 MU) stimulation that lead to tonic depolarization and decreased Rin was checked continuously (every 1–5 min.) and balanced. The recording (Cardin et al., 2008). However, this and related effects (Chance was terminated if the bridge exceeded 25 MU. In a few cases, drug actions et al., 2002) show reduced gain of the firing mechanism and visually-evoked currents were evaluated in voltage clamp (Figure 6). measured in the presence of increased synaptic conductance. Series resistance (Rs) was 10–15 MU and compensated by 60%. The mechanism we describe is apparently distinct: a change in the gain of the firing mechanism after a brief period of depolariza- Visual Stimuli tion and hyperpolarization that would typically be evoked by The retina was continuously illuminated at 2 3 103 isomerizations a transient synaptic input. Both mechanisms may typically M-cone 1 sec 1 by either a monochrome 1 inch computer monitor (Lucivid combine in intact cells under physiological conditions. MR 1-103; Microbrightfield; Colchester, VT), an RGB OLED Display (SVGA+ Each of the 15–20 ganglion cell types probably expresses Rev. 2, eMagin, Bellevue, WA), or the green channel of an RGB LED (NSTM515AS, Nichia America Co., Wixom, MI). LED intensity was controlled a unique combination of ion channels (Kaneda and Kaneko, by pClamp 9 software via a custom noninverting voltage-to-current converter 1991; Ishida, 2000; O’Brien et al., 2002; Margolis and Detwiler., with operational amplifiers (TCA0372, ON Semiconductor, Phoenix, AZ). For all 2007). Thus, it is possible that certain cell types lack one or stimulation devices, the Gamma curve was corrected in software. Responses both of the mechanisms for intrinsic adaptation shown here. were measured to spots, annuli, and gratings to confirm the OFF-center and For example, parvocellular-pathway (midget) ganglion cells in nonlinear properties of OFF Alpha cells (Demb et al., 2001; Hochstein and the monkey apparently show little or no contrast adaptation, Shapley, 1976). In one experiment, we combined current injection with visual stimulation. In and therefore they may have Na and K channel properties that this case, the timing of the contrast stimulus was adjusted so that spiking do not produce adaptation (Benardete et al., 1992). Preliminary would occur 25 msec after the offset of a current step. Preliminary experi- experiments on two other cell types in the guinea pig suggested ments with loose-patch recordings (n = 5 cells) suggested that such timing that one (the OFF Delta cell) adapted to hyperpolarizing pre- could be achieved if a 100% contrast stimulus was displayed 70 msec prior pulses, whereas a second (the ON Alpha cell) did not. Future to the desired onset time. For lower contrast stimuli, where there is a longer studies will be required to relate channel subunit expression to delay to the first spike, stimulus onset was advanced by 55 msec/ (log (contrast)), so the first spike was evoked at roughly the same time at the two intrinsic mechanisms for adaptation demonstrated 10 each contrast level. here. KDR channels may play additional roles in adaptive behavior upstream of the ganglion cell. For example, in isolated Dynamic Current Injection salamander bipolar cells, these channels mediated adaptation to In some current-clamp recordings, we dynamically compensated visually- the mean membrane potential (Mao et al., 1998; 2002). At depo- evoked hyperpolarization with current injection. We employed a small circuit larized levels, the bipolar cells showed reduced gain and devel- with a dual operational amplifier TCA0372 (ON Semiconductor) and an oped band-pass tuning to temporal inputs. Thus, within the Attiny85 microcontroller (Atmel, San Jose, CA). In order to prevent unintended compensation of spike AHPs, the time constant of current injection was retina KDR channels could play a role in adaptation to both the mean and the contrast of the visual input. voltage and time dependent: small hyperpolarizations from rest were compensated slowly. Dynamic current injection, I(t), was calculated from a simplified Hodgkin-Huxley equation:

EXPERIMENTAL PROCEDURES 2 IðtÞ = nðVmðtÞÞ Imax; (Equation 1) 2 Tissue Preparation and Electrophysiology where Imax is the maximum possible current injection of the setup (2 nA) and n The experimental procedures have been described in detail previously is the voltage-dependent proportion of that current. Changes in n over time (Beaudoin et al., 2008; Manookin et al., 2008). In each experiment, a Hartley were computed as follows: guinea pig was dark adapted for >1 hr and then anesthetized with ketamine dn 1 1 = ðn ðV ÞnÞtðV Þ; (100 mg kg ) and xylazine (10 mg kg ) and decapitated, and both eyes dt N m m (Equation 2) were removed. All procedures conformed to National Institute of Health and n (V ) University of Michigan guidelines for the use and care of animals in research. where N m is the steady-state activation: The eye cup (retina, pigment epithelium, choroid, and sclera) was mounted ﬂat 1 1 n ðV Þ = : N m V V (Equation 3) in a chamber on a microscope stage and superfused ( 6 ml min ) with 1 + e m 1=2 oxygenated (95% O2 and 5% CO2) Ames medium (Sigma, St. Louis, MO) at 33 C. The retina and electrode were visualized with a cooled CCD camera V1/2 is the voltage that generates a half-maximal value of steady-state acti-

(Retiga 1300, Qcapture software; Qimaging, Burnaby, British Columbia). Large vation and was set in each case by measuring Vrest at the beginning of each cell bodies in the ganglion cell layer (diameter: 20–25 mm) were targeted for experiment and subtracting 7 mV; this value ensured that voltage was clamped recording. A glass electrode (tip resistance, 3–6 MU) was filled with Ames at 2 mV from rest. The time constant in Equation 2, t(Vm), is defined as: medium for loose-patch extracellular recordings. Once the cell type was tðV Þ = t + t 1 ; confirmed by responses to visual stimulation, the pipette was withdrawn and m min max 1 V V (Equation 4) 1 + e 1=2 m a second pipette was used for whole-cell recording. The intracellular recording solution contained (in mM): K-methanesulfonate, 120; 4-(2-hydroxyethyl)-1-pi- where tmin =52ms (the sample rate) and tmax = 4 ms; this latter value was deter- perazineethanesulfonic acid (HEPES), 10; NaCl, 5; EGTA, 0.1; ATP-Mg2+,2; mined empirically to cancel synaptic current but not affect the spike AHP. + GTP-Na , 0.3; and Lucifer Yellow, 0.10%; titrated to pH = 7.3. All chemicals Vm was measured with 0.15 mV resolution (i.e., 10-bit resolution, range = were purchased from Sigma-Aldrich (St. Louis, MO) except mibefradil and 100 to +50 mV) at 19.3 kHz from the output of the Multiclamp 700B amplifier.

ZD7822 (Tocris; Ellisville, MO). Calculations used look-up tables for the voltage dependence of t(Vm) and

The Vm was recorded at 20 kHz and stored on a computer with a MultiClamp nN(Vm) and were completed in 40 ms. I(t) was then updated with 8 pA resolu- 700B amplifier and pClamp 9 software (Axon Instruments; Foster City, CA). tion, low-pass filtered at 10 kHz and injected into the cell via the Multiclamp The junction potential (9 mV) was corrected. We wrote programs in Matlab 700B amplifier. Improper bridge balance (e.g., >20 MU or changed

Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. 177 Neuron K Channels Mediate Contrast Adaptation

by >2MU) caused strong oscillations that in some cases even triggered attenuation of dendritic action potentials in hippocampal CA1 pyramidal spikes. Only recordings without such oscillations were analyzed. neurons. J. Neurosci. 17, 6512–6521. Demb, J.B. (2008). Functional circuitry of visual adaptation in the retina. Somatic Patch Recordings J. Physiol. 586, 4377–4384. Outside-out patches were pulled from identified OFF Alpha ganglion cells in Demb, J.B., Zaghloul, K., Haarsma, L., and Sterling, P. (2001). Bipolar cells U order to study voltage-gated currents. After establishing a seal of >5 G on contribute to nonlinear spatial summation in the brisk-transient (Y) ganglion the soma and correcting for the pipette capacitance, the cell membrane was cell in mammalian retina. J. Neurosci. 21, 7447–7454. disrupted to establish a whole-cell configuration with V = 60mV. The hold Eller, P., Berjukov, S., Wanner, S., Huber, I., Hering, S., Knaus, H.G., Toth, G., pipette was slowly removed from the cell using the manipulator’s piezo Kimball, S.D., and Striessnig, J. (2000). High affinity interaction of mibefradil drives, while constantly checking R ,R , and capacitance. After reaching s in with voltage-gated calcium and sodium channels. Br. J. Pharmacol. 130, >100 MU of R (from originally 10–20 MU), the pipette was quickly pulled s 669–677. away from the cell by several hundred micrometers. Initial membrane capac- Escoubas, P., Diochot, S., Ce´ le´ rier, M.L., Nakajima, T., and Lazdunski, M. itance and Rin were recorded from the membrane patch. Cases where Vm was (2002). Novel tarantula toxins for subtypes of voltage-dependent potassium positive to 30 mV or when the ratio of Rin to Rs was <10 were not studied channels in the Kv2 and Kv4 subfamilies. Mol. Pharmacol. 62, 48–57. further. Voltage-clamp recordings were performed without Rs compensation at 10 kHz with a 4 kHz Bessel filter. Capacitance artifacts and leak currents Fukami, H., and Bradley, R.M. (2005). Biophysical and morphological proper- were measured during the voltage-clamp recordings with 5 mV steps from ties of parasympathetic neurons controlling the parotid and von Ebner salivary 93 Vholds and used to record changes in membrane parameters. Recordings glands in rats. J. Neurophysiol. , 678–686. with roughly constant leak current were used for analysis. Capacitance arti- Gasparini, S., and DiFrancesco, D. (1997). Action of the hyperpolarization-acti- facts were fitted with a double exponential function and together with the vated current (Ih) blocker ZD 7288 in hippocampal CA1 neurons. Pflugers leak current subtracted from the current traces. Because of imperfect fits of Arch. 435, 99–106. the first two recorded points in the capacitance artifact, the first 0.2 ms after Gaudry, K.S., and Reinagel, P. (2007a). Benefits of contrast normalization a voltage step were omitted. demonstrated in neurons and model cells. J. Neurosci. 27, 8071–8079. Gaudry, K.S., and Reinagel, P. (2007b). Contrast adaptation in a nonadapting ACKNOWLEDGMENTS LGN model. J. Neurophysiol. 98, 1287–1296. Gollisch, T., and Meister, M. (2010). Eye smarter than scientists believed: We thank Mania Kupershtok for technical assistance and Dr. Josh Singer for neural computations in circuits of the retina. Neuron 65, 150–164. comments on the manuscript. Supported by a Research to Prevent Blindness Career Development award, an Alfred P. Sloan Foundation fellowship and the Grissmer, S., Nguyen, A.N., Aiyar, J., Hanson, D.C., Mather, R.J., Gutman, National Institutes of Health (EY14454; EY14454-S1; core grant EY07003). G.A., Karmilowicz, M.J., Auperin, D.D., and Chandy, K.G. (1994). Pharmacological characterization of five cloned voltage-gated K+ channels, Accepted: April 27, 2011 types Kv1.1, 1.2, 1.3, 1.5, and 3.1, stably expressed in mammalian cell lines. 45 Published online: July 13, 2011 Mol. Pharmacol. , 1227–1234. Guan, D., Lee, J.C., Higgs, M.H., Spain, W.J., and Foehring, R.C. (2007). REFERENCES Functional roles of Kv1 channels in neocortical pyramidal neurons. J. Neurophysiol. 97, 1931–1940. Baccus, S.A., and Meister, M. (2002). Fast and slow contrast adaptation in Harvey, A.L. (2001). Twenty years of dendrotoxins. Toxicon 39, 15–26. 36 retinal circuitry. Neuron , 909–919. Hochstein, S., and Shapley, R.M. (1976). Linear and nonlinear spatial subunits Baranauskas, G. (2007). Ionic channel function in action potential generation: in Y cat retinal ganglion cells. J. Physiol. 262, 265–284. 35 current perspective. Mol. Neurobiol. , 129–150. Ho¨ ltje, M., Brunk, I., Grosse, J., Beyer, E., Veh, R.W., Bergmann, M., Grosse, Beaudoin, D.L., Borghuis, B.G., and Demb, J.B. (2007). Cellular basis for G., and Ahnert-Hilger, G. (2007). Differential distribution of voltage-gated contrast gain control over the receptive field center of mammalian retinal potassium channels Kv 1.1-Kv1.6 in the rat retina during development. ganglion cells. J. Neurosci. 27, 2636–2645. J. Neurosci. Res. 85, 19–33. Beaudoin, D.L., Manookin, M.B., and Demb, J.B. (2008). Distinct expressions Ishida, A.T. (2000). Ion channel mutations in retinal neurons. Arch. Soc. Esp. of contrast gain control in parallel synaptic pathways converging on a retinal Oftalmol. 75, 69–70. ganglion cell. J. Physiol. 586, 5487–5502. Kaneda, M., and Kaneko, A. (1991). Voltage-gated sodium currents in isolated Benardete, E.A., Kaplan, E., and Knight, B.W. (1992). Contrast gain control in retinal ganglion cells of the cat: relation between the inactivation kinetics and the primate retina: P cells are not X-like, some M cells are. Vis. Neurosci. 8, the cell type. Neurosci. Res. 11, 261–275. 483–486. Kim, K.J., and Rieke, F. (2001). Temporal contrast adaptation in the input and Berry, M.J., Warland, D.K., and Meister, M. (1997). The structure and precision output signals of salamander retinal ganglion cells. J. Neurosci. 21, 287–299. of retinal spike trains. Proc. Natl. Acad. Sci. USA 94, 5411–5416. Kim, K.J., and Rieke, F. (2003). Slow Na+ inactivation and variance adaptation Cardin, J.A., Palmer, L.A., and Contreras, D. (2008). Cellular mechanisms in salamander retinal ganglion cells. J. Neurosci. 23, 1506–1516. underlying stimulus-dependent gain modulation in primary visual cortex Lee, S.C., and Ishida, A.T. (2007). Ih without Kir in adult rat retinal ganglion neurons in vivo. Neuron 59, 150–160. cells. J. Neurophysiol. 97, 3790–3799. Chance, F.S., Abbott, L.F., and Reyes, A.D. (2002). Gain modulation from Lesica, N.A., Jin, J., Weng, C., Yeh, C.I., Butts, D.A., Stanley, G.B., and Alonso, background synaptic input. Neuron 35, 773–782. J.M. (2007). Adaptation to stimulus contrast and correlations during natural Chander, D., and Chichilnisky, E.J. (2001). Adaptation to temporal contrast in visual stimulation. Neuron 55, 479–491. primate and salamander retina. J. Neurosci. 21, 9904–9916. Liang, Z., and Freed, M.A. (2010). The ON pathway rectifies the OFF pathway Chouabe, C., Drici, M.D., Romey, G., Barhanin, J., and Lazdunski, M. (1998). of the mammalian retina. J. Neurosci. 30, 5533–5543. HERG and KvLQT1/IsK, the cardiac K+ channels involved in long QT Mahon, S., Delord, B., Deniau, J.M., and Charpier, S. (2000). Intrinsic proper- syndromes, are targets for calcium channel blockers. Mol. Pharmacol. 54, ties of rat striatal output neurones and time-dependent facilitation of cortical 695–703. inputs in vivo. J. Physiol. 527, 345–354. Colbert, C.M., Magee, J.C., Hoffman, D.A., and Johnston, D. (1997). Slow Manookin, M.B., and Demb, J.B. (2006). Presynaptic mechanism for slow recovery from inactivation of Na+ channels underlies the activity-dependent contrast adaptation in mammalian retinal ganglion cells. Neuron 50, 453–464.

178 Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. Neuron K Channels Mediate Contrast Adaptation

Manookin, M.B., Beaudoin, D.L., Ernst, Z.R., Flagel, L.J., and Demb, J.B. Pinto, L.H., and Klumpp, D.J. (1998). Localization of potassium channels in the (2008). Disinhibition combines with excitation to extend the operating range retina. Prog. Retin. Eye Res. 17, 207–230. 28 of the OFF visual pathway in daylight. J. Neurosci. , 4136–4150. Rauer, H., Lanigan, M.D., Pennington, M.W., Aiyar, J., Ghanshani, S., Cahalan, Mante, V., Bonin, V., and Carandini, M. (2008). Functional mechanisms M.D., Norton, R.S., and Chandy, K.G. (2000). Structure-guided transformation shaping lateral geniculate responses to artificial and natural stimuli. Neuron of charybdotoxin yields an analog that selectively targets Ca(2+)-activated 58, 625–638. over voltage-gated K(+) channels. J. Biol. Chem. 275, 1201–1208. Mao, B.-Q., MacLeish, P.R., and Victor, J.D. (1998). The intrinsic dynamics of Rieke, F. (2001). Temporal contrast adaptation in salamander bipolar cells. retinal bipolar cells isolated from tiger salamander. Vis. Neurosci. 15, 425–438. J. Neurosci. 21, 9445–9454. Mao, B.-Q., MacLeish, P.R., and Victor, J.D. (2002). Relation between potas- Rieke, F., and Rudd, M.E. (2009). The challenges natural images pose for visual sium-channel kinetics and the intrinsic dynamics in isolated retinal bipolar adaptation. Neuron 64, 605–616. 12 cells. J. Comput. Neurosci. , 147–163. Scott, V.E., Muniz, Z.M., Sewing, S., Lichtinghagen, R., Parcej, D.N., Pongs, Margolis, D.J., and Detwiler, P.B. (2007). Different mechanisms generate O., and Dolly, J.O. (1994). Antibodies specific for distinct Kv subunits unveil maintained activity in ON and OFF retinal ganglion cells. J. Neurosci. 27, a heterooligomeric basis for subtypes of alpha-dendrotoxin-sensitive K+ 5994–6005. channels in bovine brain. Biochemistry 33, 1617–1623. Mayer, M.L., and Sugiyama, K. (1988). A modulatory action of divalent cations Shapley, R.M., and Victor, J.D. (1978). The effect of contrast on the transfer on transient outward current in cultured rat sensory neurones. J. Physiol. 396, properties of cat retinal ganglion cells. J. Physiol. 285, 275–298. 417–433. Spain, W.J., Schwindt, P.C., and Crill, W.E. (1991). Post-inhibitory excitation Mu¨ nch, T.A., da Silveira, R.A., Siegert, S., Viney, T.J., Awatramani, G.B., and and inhibition in layer V pyramidal neurones from cat sensorimotor cortex. Roska, B. (2009). Approach sensitivity in the retina processed by a multifunc- J. Physiol. 434, 609–626. tional neural circuit. Nat. Neurosci. 12, 1308–1316. Storm, J.F. (1993). Functional diversity of K+ currents in hippocampal pyra- Murphy, G.J., and Rieke, F. (2006). Network variability limits stimulus-evoked midal neurons. Semin. Neurosci. 5, 79–92. spike timing precision in retinal ganglion cells. Neuron 52, 511–524. van Wyk, M., Wa¨ ssle, H., and Taylor, W.R. (2009). Receptive field properties of Nisenbaum, E.S., Xu, Z.C., and Wilson, C.J. (1994). Contribution of a slowly in- ON- and OFF-ganglion cells in the mouse retina. Vis. Neurosci. 26, 297–308. activating potassium current to the transition to firing of neostriatal spiny Victor, J.D. (1987). The dynamics of the cat retinal X cell centre. J. Physiol. 386, projection neurons. J. Neurophysiol. 71, 1174–1189. 219–246. Nistri, A., and Cherubini, E. (1992). Inactivation characteristics of a sustained, Werblin, F.S. (2010). Six different roles for crossover inhibition in the retina: Ca(2+)-independent K+ current of rat hippocampal neurones in vitro. correcting the nonlinearities of synaptic transmission. Vis. Neurosci. 27, 1–8. 457 J. Physiol. , 575–590. Yoo, H.Y., Zheng, H., Nam, J.H., Nguyen, Y.H., Kang, T.M., Earm, Y.E., and O’Brien, B.J., Isayama, T., Richardson, R., and Berson, D.M. (2002). Intrinsic Kim, S.J. (2008). Facilitation of Ca2+-activated K+ channels (IKCa1) by mibe- physiological properties of cat retinal ganglion cells. J. Physiol. 538, 787–802. fradil in B lymphocytes. Pflugers Arch. 456, 549–560. Perchenet, L., and Cle´ ment-Chomienne, O. (2000). Characterization of mibe- Zaghloul, K.A., Boahen, K., and Demb, J.B. (2005). Contrast adaptation in fradil block of the human heart delayed rectifier hKv1.5. J. Pharmacol. Exp. subthreshold and spiking responses of mammalian Y-type retinal ganglion Ther. 295, 771–778. cells. J. Neurosci. 25, 860–868.

Neuron 71, 166–179, July 14, 2011 ª2011 Elsevier Inc. 179