A sodium afterdepolarization in rat superior colliculus neurons and its contribution to population activity

Nearly all neurons in the rat superior colliculus produce an afterdepolarization immediately following an action potential. This afterdepolarization, which results from the opening of Na⁺ channels, has properties that suggest it plays a role in delay-period or persistent activity seen during shifts in attention.

Abstract

The mammalian superior colliculus (SC) is a midbrain structure that integrates multimodal sensory inputs and computes commands to initiate rapid eye movements. SC neurons burst with the sudden onset of a visual stimulus, followed by persistent activity that may underlie shifts of attention and decision making. Experiments in vitro suggest that circuit reverberations play a role in the burst activity in the SC, but the origin of persistent activity is unclear. In the present study we characterized an afterdepolarization (ADP) that follows action potentials in slices of rat SC. Population responses seen with voltage-sensitive dye imaging consisted of rapid spikes followed immediately by a second distinct depolarization of lower amplitude and longer duration. Patch-clamp recordings showed qualitatively similar behavior: in nearly all neurons throughout the SC, rapid spikes were followed by an ADP. Ionic and pharmacological manipulations along with experiments with current and voltage steps indicated that the ADP of SC neurons arises from Na⁺ current that either persists or resurges following Na⁺ channel inactivation at the end of an action potential. Comparisons of pharmacological properties and frequency dependence revealed a clear parallel between patch-clamp recordings and voltage imaging experiments, indicating a common underlying membrane mechanism for the ADP in both single neurons and populations. The ADP can initiate repetitive spiking at intervals consistent with the frequency of persistent activity in the SC. These results indicate that SC neurons have intrinsic membrane properties that can contribute to electrical activity that underlies shifts of attention and decision making.

NEW & NOTEWORTHY

Nearly all neurons in the rat superior colliculus produce an afterdepolarization immediately following an action potential. This afterdepolarization, which results from the opening of Na⁺ channels, has properties that suggest it plays a role in delay-period or persistent activity seen during shifts in attention.

persistent activity in neural circuits has been proposed to serve as a mechanism of short-term information storage in the brain (Wang 2001) and is linked to shifts of attention and decision making in the superior colliculus (SC) of primates (Basso and Wurtz 1998; Horwitz and Newsome 1999; Ignashchenkova et al. 2004; Krauzlis and Dill 2002) and rodents (Felsen and Mainen 2008; Ngan et al. 2015). The high-frequency discharges of neurons in the SC, referred to as bursts, underlie the initiation of saccades in monkeys and orienting movements in rodents. Studies in rat SC slices suggest that this activity arises from reverberations within local circuits arising from the activation of NMDA receptors and enhanced by reduced GABAergic input (Isa and Hall 2009; Moschovakis 1996; Saito and Isa 2003). However, little is known about how persistent or delay-period activity arises in the SC. Many neurons have the capacity for sustained electrical activity and repetitive discharge, independent of network activity, and this property depends critically on a neuron's intrinsic excitability. SC neurons exhibit complex discharge properties such as spike adaptation (Ghitani et al. 2014; Saito and Isa 1999), suggesting that these neurons possess a rich assortment of voltage-gated conductances.

Voltage-gated conductances generate the classical biphasic action potential as Na⁺ and K⁺ channels open and close in sequence, but neurons can exhibit very complex firing behavior depending on modifications of the classical kinetic properties of these channels and by the addition of other types of channels. The classical voltage-gated conductances in axons are tailored for propagating electrical signals over long distances, but in other cellular compartments voltage-gated conductances integrate diverse signals and generate complex firing patterns. In many neurons an action potential can be followed by a coordinated series of membrane potential changes. These complex properties enable neurons to integrate inputs, encode information in the form of patterns of activity, and thereby enrich the computational capacity of neuronal circuits (Bean 2007; Llinas 1988).

Voltage-gated Na⁺ channels belong to a diverse family of proteins, and this diversity enables these proteins to play a major role in determining the complex firing properties of neurons. The canonical Na⁺ channel inactivates rapidly and completely during a depolarization. However, in many types of cells a small residual Na⁺ current persists at depolarized potentials. Persistent Na⁺ current can play important roles in synaptic integration and in the generation of rhythmic activity and bursts (Azouz et al. 1996; Crill 1996; Raman et al. 1997). Na⁺ channels can also reopen during or immediately after the action potential down stroke (Lewis and Raman 2014). This reactivation arises from a modification of inactivation kinetics by an endogenous cytosolic blocking protein, which plugs the open channel even more rapidly than classical Na⁺ channel inactivation and is ejected during repolarization to initiate resurgence (Raman and Bean 2001). Persistent and resurgent Na⁺ currents can enhance a neuron's excitability, enrich its discharge properties, and contribute to its complex electrical behavior.

In this article we report an investigation of an afterdepolarization (ADP) that follows action potentials in neurons of the rat SC. Experiments in SC slices revealed this ADP at the population level in voltage imaging recordings and at the single-cell level in patch-clamp recordings. This ADP arises from a Na⁺ current, and its kinetics roughly resemble that of resurgent Na⁺ currents in other types of cells. Furthermore, the ADP observed in voltage imaging mirrors the ADP in single SC neurons, suggesting that this Na⁺ current makes a major contribution to population-level activity. This intrinsic property of SC neurons has the capacity to contribute to the persistent activity seen in vivo in the SC during decision making and shifts of attention.

METHODS

Slice preparation.

Fresh coronal slices from the SC were prepared as described previously (Vokoun et al. 2010) from 2- to 4-wk-old Sprague-Dawley rats of either sex. All procedures were approved by the Animal Care and Use Committee of the University of Wisconsin-Madison. Briefly, animals were rendered unconscious by inhalation of vaporized isoflurane or CO₂ and immediately killed by decapitation. Brains were carefully removed and placed in ice-chilled cutting solution (in mM: 124 NaCl, 3.2 KCl, 26 NaHCO₃, 1.25 NaH₂PO₄, 1 CaCl₂, 6 MgSO₄, and 10 glucose) and bubbled with carbogen (95% O₂–5% CO₂). Coronal slices through the SC were cut from caudal to rostral at 400 μm in chilled cutting solution using a Leica 1200S vibratome. Slices were recovered in artificial cerebral spinal fluid (ACSF; same as cutting solution but with 2.5 mM CaCl₂ and 1.2 mM MgCl₂) at room temperature for at least 45 min before the start of recordings. All experiments were performed while slices were continuously perfused with ACSF bubbled with 95% O₂–5% CO₂.

Whole cell patch-clamp recordings.

Slices were transferred to a recording chamber mounted on the stage of a Zeiss Axioskop microscope and viewed with infrared-differential interference contrast optics through a ×63 objective. Patch electrodes were fabricated from borosilicate glass capillaries using a Narishige PC-10 puller. Electrodes were filled with internal solution (in mM: 130 K-gluconate, 10 HEPES, 7 KCl, 1 EGTA, 2 Na₂ATP, and 2 MgATP, pH 7.2) and had resistances of 3–5 MΩ. After whole cell recordings were established, voltage or current was measured using an Axopatch 200B amplifier (Molecular Devices) and digitized through a Digidata 1440A interface to a computer running the acquisition software pClamp 10.2 (Molecular Devices). For electrical stimulation of afferents, borosilicate electrodes with fire-polished tips (opening 5–10 μm) filled with ACSF were positioned in the slice. Stimulus current was applied by a WPI A385 stimulus isolator controlled by pClamp 10.2.

Voltage imaging.

After incubation in ACSF for 30–60 min, slices were stained with the voltage-sensitive dye RH482 (synthesized in-house) at 0.05 mg/ml for 45 min at room temperature and returned to ACSF. Slices were transferred to a specially designed recording chamber for imaging with a Reichert Jung Diastar microscope. The instrumentation used for voltage imaging is similar to that commercially available from RedShirt Imaging. This instrumentation was described previously (Chang 2006) and has been used to image voltage in SC slices (Bayguinov et al. 2015; Vokoun et al. 2010). Transmitted light was recorded with a 464-channel fiber optic-coupled photodiode camera with hexagonal geometry. The center-to-center distance between fields of neighboring photodiodes in the array was 67 μm. Light was collected, amplified to 5 V/nA of photocurrent, and digitized at a frame rate of 5 kHz using a DAP5200 acquisition board (Microstar Laboratories) on a computer running image acquisition software developed in-house (Chang 2006). Slices were illuminated by a 100-W tungsten-halogen bulb powered by a Kepco ATE 36-30 power supply. Light was bandpass filtered (700 ± 25 nm) and collected by a ×10 Olympus objective (NA 0.4). Trials were collected at 10-s intervals and averaged. Electrical stimulation of slices in voltage imaging experiments was as described above for patch-clamp recordings.

Data analysis.

Measurements of amplitude, 10% to 90% rise time, and decay time of ADPs recorded under current clamp were performed using the event detection function of Clampfit 10.2 (Molecular Devices). Measurements of spike and ADP amplitudes recorded by voltage imaging were also made with Clampfit 10.2. All statistical analyses were performed using Origin 8.6 (OriginLab). Statistical significance in comparing means was evaluated using Student's t-test.

RESULTS

Afterdepolarizations in SC neurons.

The SC has a laminar structure, as illustrated in a coronal section (Fig. 1A). The superficial layers of the SC, the stratum griseum superficiale (SGS) and stratum opticum (SO), receive visual input, primarily from the retina and visual cortex. The intermediate layer, the stratum griseum intermediale (SGI), serves as the “output” layer and generates premotor commands to initiate orienting movements in rodents. Our experiments focused on SGS, SO, and SGI (SGIa and SGIb), all of which contain neurons from which persistent activity has been recorded in rodents in vivo (Felsen and Mainen 2008; Ngan et al. 2015).

Fig. 1.

Afterdepolarization in superior colliculus neurons. A: brightfield image of a coronal slice through the superior colliculus (SC) with stratum griseum superficiale (SGS), stratum opticum (SO), and stratum griseum intermediale (SGI) demarcated. B: electrical stimulation of the SGI evoked an initial spike and afterdepolarization (ADP) recorded at the population level by voltage imaging in the superficial layers (top trace) and at the single-neuron level by whole cell current-clamp (bottom trace). C: representative traces of spikes and ADPs evoked by current injection in neurons from the SGS, SO, and SGI. D: mean amplitude (i), mean 10%–90% rise time (ii), and mean 10%–90% decay time (iii) of ADPs recorded from neurons in the SGS (n = 48), SO (n = 159), and SGI (n = 43). E: spikes evoked at depolarized membrane potentials maintained by holding current reveal 3 phases of the intrinsic voltage change following a spike: a fast afterhyperpolarization (fAHP), an ADP, and a slow afterhyperpolarization (sAHP). F: superimposed recordings from a neuron at membrane potentials of −55 and −60 mV maintained by holding current revealed a larger ADP at more negative membrane potentials.

Responses of coronal SC slices to extracellular stimulation exhibited a characteristic biphasic behavior at the population level in voltage imaging recordings (Fig. 1B, top trace), as described previously (Bayguinov et al. 2015; Vokoun et al. 2010). At the single-neuron level, whole cell patch-clamp recordings under current clamp also revealed biphasic responses to stimulation (Fig. 1B, bottom trace). With both techniques we observed that electrical stimulation evoked a rapid initial spike followed by an ADP. The same characteristic sequence was observed in neurons from all three layers of the SC (Vokoun et al. 2010), and patch-clamp recordings of responses triggered by direct current injection confirmed this at the single-cell level (Fig. 1C). Initial spikes were reliably followed by an ADP in nearly all neurons from which recordings were made. In a total of 250 neurons from which we recorded in SC slices under whole cell patch clamp, only 3 of 48 cells in the SGS, 4 of 159 in the SO, and 2 of 43 in the SGI failed to produce an ADP following an initial spike.

In each layer we evaluated the basic characteristics of all ADPs recorded at a holding potential of −60 mV (Fig. 1D). The mean ADP amplitude measured from the spike afterhyperpolarization to the peak of the ADP was 10.6 ± 1.3 mV for SGS neurons, 14.4 ± 0.7 mV for SO neurons, and 11.5 ± 1.6 mV for SGI neurons. ADP amplitudes in SO neurons were significantly greater than in SGS neurons (P < 0.05). The mean 10%–90% rise times were indistinguishable in neurons from the SGS, SO, and SGI (Fig. 1Dii), but the mean 10%–90% decay times varied from 66.2 ± 17.3 ms in SGS neurons to 33.3 ± 3.0 ms in SO neurons to 29.1 ± 6.6 ms in SGI neurons. These results suggest that the population ADP seen in voltage imaging reflects an intrinsic and general property of SC neurons in all layers.

The characteristic voltage changes that follow the rapid initial spike in SC neurons have three phases that become clear in neurons held at membrane potentials depolarized above −60 mV. In neurons maintained at −45 mV by holding current, action potentials terminate with a fast afterhyperpolarization, followed by an ADP and a slow afterhyperpolarization (Fig. 1E). The amplitude of the ADP decreased when the holding potential was raised to a more depolarized value, as illustrated by comparison of recordings of events from neurons held at −55 and −60 mV (Fig. 1F). In 19 neurons with the membrane potential adjusted to −60 mV, the ADP amplitude was 16.0 ± 1.3 mV. Allowing these neurons to go to their resting potentials (−50 to −55 mV) reduced the ADP amplitude to 14.5 ± 1.4 mV (P = 0.0014).

ADP mediation by Na⁺ current.

To identify the ionic current giving rise to the ADP, we carried out pharmacological experiments in patch-clamped SC neurons. We first tested the role of a voltage-activated Ca²⁺ channels, since R-type Ca²⁺ channels give rise to ADPs in pyramidal neurons of the hippocampus (Metz et al. 2005) and T-type Ca²⁺ channels contribute to ADPs in cerebellar Purkinje neurons (Swensen and Bean 2003). We found that changing to a Ca²⁺-free bathing solution while a recording was underway did not inhibit the ADP in any of the neurons tested but unexpectedly produced an increase in the ADP amplitude and duration for the majority of neurons tested (9 of 12 neurons; Fig. 2Ai). For neurons in which Ca²⁺-free bathing solution enhanced the ADP, prolonged incubation for more than 20 min under Ca²⁺-free conditions brought the ADP above threshold and produced bursts of spikes in 8 of 9 cells. Figure 2Aii presents an example. Similar results have been reported previously in the cerebellum (Swensen and Bean 2003). Ca²⁺ channel blockade by Ni²⁺ (1 mM) produced no change in the ADP (data not shown). These experiments indicate that a voltage-gated Ca²⁺ channel cannot account for the ADP in SC neurons.

Fig. 2.

Effects of Ca²⁺-free ACSF and Na⁺ channel blockers on the ADP. Ai: superimposed recordings in control ACSF (black trace) and 5 min after the change to Ca²⁺ free ACSF (gray trace). Aii: after ∼30 min in Ca²⁺-free ACSF, the ADP rose above threshold (gray traces with spikes). B: recordings of spikes and ADPs in control ACSF (black trace) and a spike-like depolarization after the addition of 1 μM TTX (gray trace). C: ADP amplitude was diminished 20 min after the addition of 200 μM lidocaine. Action potentials could still be triggered (gray trace) while the ADP amplitude was reduced.

We next tested the role of voltage-activated Na⁺ conductance in the ADP. Previous voltage imaging experiments in SC slices showed that the Na⁺ channel blocker tetrodotoxin (TTX; 1 μM) eliminated both the initial spike and the ADP (Vokoun et al. 2010). However, since TTX blocks the initial spike, this experiment does not rule out the possibility that the ADP is triggered by the voltage change during the initial spike, which could then activate other voltage-gated channels that are responsible for the ADP. This result with TTX therefore does not assess the dependence of the ADP on Na⁺ channels. We addressed this issue using patch-clamp recordings by applying large, brief current pulses to produce a spike-like depolarization in the presence of TTX. We found that this artificial spike, which mimics an action potential without relying on Na⁺ channels, was not followed by an ADP in any of the neurons tested (Fig. 2B, n = 8). Thus a rapid voltage change similar to that seen during an action potential did not trigger the gating of other TTX-insensitive channels to produce an ADP. This experiment with TTX provides much stronger support for the conclusion that an ADP is generated by voltage-gated Na⁺ channels. These Na⁺ channels open immediately after the initial action potential has repolarized.

We next sought to inhibit the ADP by partially blocking voltage-gated Na⁺ channels under conditions where action potentials could still be triggered. The local anesthetic lidocaine blocks Na⁺ channels in a use-dependent manner, so blockade is accelerated while channels are open (Bant et al. 2013). On the basis of this use dependence, we expect that blockade will begin with the initial spike and that more Na⁺ channels will be blocked by the time the ADP starts. As a result, lidocaine will reduce the ADP more than the initial spike. We added 200–500 μM lidocaine and found we could still trigger action potentials but with slightly reduced amplitudes and higher thresholds. The amplitude of the initial spike was reduced to 82 ± 3.6% of its control value, but the amplitude of the ADP was reduced to 67 ± 4.5% of its control value (from 17.1 ± 1.2 to 11.0 ± 0.6 mV; n = 6, P = 0.001; Fig. 2C and Table 1). The greater effect on the ADP than the spike is consistent with blockade of open Na⁺ channels during the spike, leaving fewer Na⁺ channels available for the ADP. However, this difference could also reflect the balance between Na⁺ and other ion channels, as well as the large difference in Na⁺ driving force at the different voltages. Nevertheless, lidocaine reduced both events, and the greater effect on the ADP than the initial spike provides a criterion for assessing the contribution of Na⁺ channels to the population ADP seen with voltage imaging, as presented below.

Table 1.

Lidocaine actions

	Current Clamp, mV		Population (ΔI/I × 103)
	Control	200–500 μM	Control	100 μM
Initial spike	132 ± 3.6	116 ± 3.5	0.121 ± 0.0035	0.0743 ± 0.0026
ADP	17.1 ± 1.2	11.0 ± 0.6	0.068 ± 0.0027	0.0381 ± 0.0024

Data are comparisons of lidocaine effects on the initial spike and ADP for single cells under current clamp evoked by current injection and for population responses seen with voltage imaging evoked by extracellular stimulation.

Activity-dependent attenuation of the ADP.

High-frequency firing alters spike characteristics in rat SC neurons (Ghitani et al. 2014; Saito and Isa 1999). With repetitive stimulation of action potentials in patch-clamped neurons by current injection, the ADP amplitude was attenuated in many, but not all, cells. As can be seen in superimposed recordings of 5 spikes in a train evoked by brief current pulses at 500-ms intervals (2 Hz) in an SO neuron, the ADP amplitude declined as the train progressed (Fig. 3Ai). The degree of attenuation was quite variable. In 11 of 26 neurons tested, the ADP amplitude decreased by <3.5% with a second spike 500 ms after the first. In the remainder of neurons, attenuation ranged from 6.7% to 50% (mean 19 ± 4%, n = 14). Despite this variation between neurons, each neuron exhibited reproducible attenuation from trial to trial. In some cases, the ADP of the first spike exceeded threshold and triggered another spike, but the ADP of later spikes did not (Fig. 3Aii). Interestingly, the action potentials in trains had identical waveforms; only the ADP attenuated (Fig. 3B).

Fig. 3.

Attenuation of the ADP during repetitive spiking. Ai: 5 spikes evoked by current injection at 500-ms intervals are superimposed to show a reduction in the ADP after the first event. Aii: with the membrane potential lowered to −65 mV by negative holding current, only the first ADP in the train exceeded threshold and triggered a second spike (indicated by an arrowhead). B: the first 2 events in Ai displayed at an expanded timescale show attenuation of the ADP, with no change in the initial action potential. C: attenuation of the ADP was graded depending on the frequency of repetitive spiking (from left to right: events at 2, 1, and 0.5 Hz superimposed). Di: discharge characteristics of the same neuron presented in A, with spikes induced by a long 1-s current pulse (starting where the voltage first changes 40 ms after the start of the trace). The initial burst of spikes was followed by marked frequency adaptation (gray trace, injected current = 1 nA; black trace, 0.6 nA). Dii: aligned superimposed spikes illustrate rapid attenuation of the ADP in this neuron during the train. Ei: in another neuron, spiking occurs in a regular pattern without adaptation (gray trace, injected current = 3 nA; black trace, 6 nA). Eii: attenuation of the ADP in this neuron during discharge of action potentials at 2 Hz.

The ADP attenuated more with higher frequency trains (Fig. 3C). We found that activity-dependent attenuation of the ADP correlated with the bursting properties of SC neurons. This can be seen in the comparison between a neuron that exhibited frequency adaptation (Fig. 3D) and a regular spiking neuron (Fig. 3E). In long trains of spikes triggered by sustained current injection, we measured the ratio of the interspike interval between the first two spikes to the average interspike interval of the train. This ratio varied between attenuating and nonattenuating neurons. Of a total of 24 neurons tested for ADP attenuation with repetitive stimulation, neurons that displayed ADP attenuation (≥6.7%) had a mean ratio of first interspike interval to average interspike interval of 0.23 ± 0.05 (n = 11), whereas neurons that did not display ADP attenuation (≤3.5%) had a significantly larger mean interspike interval ratio (0.62 ± 0.11; n = 13, P = 0.005). These results suggest that although nearly every SC neuron exhibits an ADP, there is some variation in the extent to which it attenuates with frequency, and this variation may impact the way in which the ADP contributes to circuit function.

Late Na⁺ current.

Some voltage-gated Na⁺ channels have distinct kinetic properties that enable a neuron to generate an ADP. Such Na⁺ channels exhibit a second, late phase of opening that can take the form of persistent (Azouz et al. 1996; D'Ascenzo et al. 2009; Yue et al. 2005) or resurgent current (Bean 2007; Lewis and Raman 2014), which appears immediately after ending a brief depolarizing pulse used to evoke the large transient phase of Na⁺ current. The distinction between persistent and resurgent Na⁺ current is readily seen in voltage-clamp recordings, but such experiments are generally conducted in dissociated neurons rather than neurons in brain slices because of the difficulty in brain slices of obtaining the spatial control of voltage needed for accurate measurement of Na⁺ current. In our attempts to voltage clamp neurons in SC slices, depolarizing voltage steps elicited Na⁺ current, but unclamped Na⁺ spikes, presumably originating from remote unclamped processes, appeared with delays of tens of milliseconds. This reflects poor control of voltage in neuronal processes. However, before the onset of these spikes, brief 2- to 20-ms depolarizing steps to positive potentials triggered rapidly inactivating, voltage-activated current, and this inward current was blocked by 1 μM TTX (data not shown). When a brief step to a positive voltage was followed by return steps to various potentials ranging from −40 to −70 mV, we observed a second phase of Na⁺ current during the return step. Like the first transient component, the second component of inward current was also blocked by TTX (data not shown), As the TTX washed in, both the large transient current during the positive step and the second component during the return step were blocked in parallel, indicating that like the first phase, the second component represents current through voltage-gated Na⁺ channels. A representative current trace from an SC neuron shows that the first step to +40 mV elicits a large, conventional, rapidly inactivating Na⁺ current. The return step to −40 mV Na⁺ elicits a second component of Na⁺ current (Fig. 4A). The current during the second step resembles the resurgent form of Na⁺ current described in many other neurons (Lewis and Raman 2014), but slow changes in voltage within poorly clamped processes could also account for this component of current.

Fig. 4.

Voltage dependence of late Na⁺ current. Ai: under voltage clamp, a voltage step from −70 to +40 mV evoked a transient Na⁺ current (I_Na − peak), and a subsequent repolarizing step to −40 mV elicited a second phase of late Na⁺ current (dashed box). Aii: action potential with ADP evoked in the same neuron under current clamp. Bi: comparison of the ADP (top trace) and late Na⁺ current (bottom trace) at the same timescale. Bii: repetitive stimulation at 2 Hz attenuated the ADP under current clamp (top) and late Na⁺ current under voltage clamp (bottom). C: current traces from a voltage-clamped SO neuron showing voltage dependence of late Na⁺ current with voltage steps to +40 mV followed by repolarizing steps to −30, −40, −50, −55, −60, and −70 mV. The gray arrow indicates the current at the start of the repolarizing step in the −30-mV trace. D: amplitude of the late Na⁺ current during the return voltage plotted against voltage. Late Na⁺ current was measured from the start of the return voltage step to the peak indicated by arrows. Current plotted was an average from 21 neurons pooled from the SGS, SO, and SGI.

In comparing the time course of the ADP with the time course of the second component of Na⁺ current (seen under voltage clamp), we saw that this current reaches a peak roughly at the time when the ADP (seen under current clamp) is rising most steeply, and is very small by the time the ADP reaches peak. In fact, the current appears to behave like the derivative of the voltage, in a manner consistent with this current driving the voltage change (Fig. 4Bi). As with attenuation of the ADP following repetitive stimulation (Fig. 3), we saw a similar reduction in amplitude of the second component of inward current with repetitive voltage steps at the same frequency (Fig. 4Bii).

The voltage dependence of the second component of Na⁺ current, recorded from a neuron in the SO, is illustrated with current traces measured during return steps ranging from −70 to −30 mV (Fig. 4C). The amplitude of this component of Na⁺ current, measured as the increase in current from the start of the second voltage level (indicated by the gray arrow in the −30 mV trace) to the peak a few milliseconds later (solid arrows in Fig. 4C), increases as the voltage step increases from −70 to −40 mV (Fig. 4D). A similar voltage dependence was reported in the calyx of Held (Kim et al. 2010). The results of our voltage-clamp experiments indicate that the Na⁺ channels in SC neurons have a kinetically distinct component of current that emerges during repolarization. The time course, frequency, and voltage dependence of the second phase of inward Na⁺ current are consistent with the hypothesis that this second component of resurgent inward current generates the ADP.

ADP in population signals.

Figure 1 showed that the ADP in single neurons recorded with current clamp bears a qualitative resemblance to the ADP in populations of neurons recorded with voltage imaging. We therefore explored the links between the population and single-neuron ADPs further by comparing voltage imaging experiments with patch-clamp experiments. During repetitive stimulation at 2 Hz, the ADP seen in voltage imaging attenuated (Fig. 5Ai), as observed in single neurons (Figs. 3A and 5Aii). The mean ratio of ADP to spike amplitudes decreased from 0.42 ± 0.012 to 0.36 ± 0.055 in voltage imaging experiments (P <0.0001, n = 9; Fig. 5B), and this also paralleled the attenuation of the ADP in single neurons. Because lidocaine was shown to reduce the amplitude of the ADP in single neurons (Fig. 2C), we tested the action of lidocaine on the population ADP. Within 10 min of application of 100 μM lidocaine to the bathing solution, the ADP and spike amplitudes both fell in parallel as the block of Na⁺ channels progressed, but the reduction in the ADP was greater (Fig. 5Ci). (We used a lower concentration of lidocaine than with single-cell experiments described above because we needed to see the drug effect more rapidly while maintaining a whole cell patch-clamp recording through the solution change.) This action resembled that seen in patch-clamp recordings (Fig. 2C) despite the lower concentration used and the fact that lidocaine should reduce the number of neurons reaching threshold (Fig. 5Cii). Lidocaine reduced the spike to 62.9 ± 0.01% of its control value but reduced the ADP to 53.7 ± 0.02% of its control amplitude (Table 1 and Fig. 5E; P < 0.0001). The mean ADP/spike ratio was significantly lower in lidocaine (0.57 ± 0.02 in control to 0.52 ± 0.03 in lidocaine; P = 0.0011, n = 6; Fig. 5D). Thus voltage imaging measurements of the population ADP paralleled the patch-clamp measurements from single neurons in demonstrating a greater impact of lidocaine on the ADP than on the initial spike. The greater reduction in amplitude of the ADP vs. the initial spike with both repetitive stimulation and lidocaine suggests that the population ADP seen in voltage imaging operates by the same Na⁺ channel-dependent mechanism as the ADP in single neurons.

Fig. 5.

Population ADP attenuation by repetitive stimulation and lidocaine. Ai: voltage imaging shows 2 successive evoked optical responses at 2 Hz superimposed, revealing attenuation of the ADP in the second event, similar to the attenuation of an ADP measured in single neurons (Aii and Fig. 3A). B: mean ratio of the ADP amplitude to spike amplitude for the first and second events shows a significant decrease in the second event (P < 0.0001). Ci: voltage imaging shows reduction of ADP amplitude shortly after lidocaine addition before complete block of activity by prolonged exposure. These results parallel ADP inhibition by lidocaine in single-neuron recordings (Cii and Fig. 2C). D: mean ratio of ADP to spike amplitudes in control ACSF and after addition of 100 μM lidocaine (P = 0.0011). E: ratio of the initial spike amplitude in lidocaine to that in control illustrates the inhibition of the initial spike by lidocaine. The ratio of the ADP amplitude in lidocaine to that in control was significantly lower (P < 0.0001), indicate a greater effect of lidocaine on the ADP.

DISCUSSION

Electrical stimulation of SC brain slices evokes a highly characteristic response consisting of an initial spike followed by an ADP (Bayguinov et al. 2015; Vokoun et al. 2010). The initial spike had the temporal attributes and pharmacological sensitivity of a Na⁺-dependent population action potential, but these earlier studies left the origin of the ADP unclear. The present study characterized the ADP of SC neurons. Both population activity and single neurons in visuosensory layers (SGS and SO) and premotor layers (SGI) of the SC display a prominent ADP in their response to electrical stimulation. Recordings from single neurons indicate that the ADP results from TTX-sensitive voltage-gated Na⁺ channels that open following the repolarization of the initial spike. The population ADP observed with voltage imaging corresponds well with the ADP observed with patch-clamp recordings. The parallel reduction of the ADP with both lidocaine and repetitive activity provides a strong link between the population and single-neuron behavior.

Neurons capable of generating an ADP were found in every layer of the SC, consistent with the ubiquitous distribution of the ADP throughout SC slices revealed by voltage imaging. The ADP amplitude in single neurons was smaller in the SGS compared with the SO and SGI. It is intriguing that a small fraction of neurons in every layer (∼5%) appeared otherwise healthy but had no ADP. This may indicate the existence of a distinct class of neurons with specialized functions.

Whole cell patch-clamp recording provided evidence that after an action potential Na⁺ channels open to produce a second wave of depolarization. This ADP could result from resurgent Na⁺ channels, as shown previously in cerebellar Purkinje (Raman and Bean 1997), granule cells (Magistretti et al. 2006), and auditory brain stem nerve terminals (Kim et al. 2010), or from persistent Na⁺ channels, as shown previously for CA1 pyramidal cells (Azouz et al. 1996; Yue et al. 2005) and medium spiny neurons of the nucleus accumbens (D'Ascenzo et al. 2009). The second component of Na⁺ current seen under voltage clamp resembles the resurgent Na⁺ current more than the persistent Na⁺ current, but distinguishing between these two alternatives will require voltage-clamp recordings in dissociated neurons under conditions of better space clamp. The present focus on slices makes it more difficult to answer such questions but has the advantage of providing results with greater relevance to persistent activity and circuit function in the SC.

Using TTX to block voltage-gated Na⁺ channels completely, we observed that rapid, depolarizing current injections that recapitulate the rapid voltage changes of action potentials failed to elicit an ADP. Thus a rapid spike-like depolarization in itself did not activate other channels to trigger the ADP. The ADP depends on flux through TTX-sensitive Na⁺ channels. Furthermore, the Na⁺ channel blocker lidocaine had a greater effect on the ADP than on the initial spike in both voltage imaging and patch-clamp recordings. The greater reduction of the ADP likely reflects lidocaine blockade of channels as they open during the initial spike. This use-dependent block during the initial spike reduces the availability of Na⁺ channels for a second round of channel activity that can give rise to the ADP. In this way the same population of voltage-gated Na⁺ channels activated during the rising phase of the action potential can produce a second component of current and a second wave of depolarization.

The ADP recorded from single neurons under current clamp was qualitatively similar to, but generally briefer than, the ADP seen with voltage imaging. The longer duration of the population ADP could reflect asynchronous rebound spikes or circuit activity. Current-clamped neurons occasionally generated second, rebound action potentials within the time window of the voltage-imaging ADP. The low incidence of these rebound spikes under control conditions suggests that the ADP will not ordinarily make a major contribution to persistent activity. However, the ADP could amplify a second synaptic input that occurs in the ADP timeframe to enhance SC responses to inputs with frequencies in the range of roughly 20–100 Hz. Furthermore, because the ADP comes close to the spike threshold, small changes in excitability could convert the ADP into a potent drive for persistent activity. Metabotropic glutamate receptor-mediated enhancement of the ADP through persistent Na⁺ channel modulation enhances spiking in nucleus accumbens medium spiny neurons (D'Ascenzo et al. 2009). Enhancement of persistent Na⁺ current by nitric oxide (Ahern et al. 2000) could also enhance the ADP. Ca²⁺ removal increased the ADP amplitude and duration in the majority of neurons tested, with the amplitude growing large enough to elicit bursts of action potentials (Fig. 2Aii). This enhancement likely resulted from reduced activity of Ca²⁺-activated K⁺ channels, which may underlie the prominent slow spike afterhyperpolarization (Fig. 1E). The powerful influence of this conductance on the ADP and the sensitivity of neuronal excitability to small changes in ADP amplitude suggest a possible function in modulating the excitability of SC neurons. Reducing the activity of the channel responsible for the afterhyperpolarization, either directly by modulating a Ca²⁺-activated K⁺ channel or indirectly by modulating Ca²⁺ channels, could enhance the ADP, raising its amplitude above the spike threshold to produce more rebound spikes and persistent activity. Any channel modulation that enhances the ADP could induce a transition in SC neurons from brief transient responses to sustained activity. The attenuation of the ADP with frequency seen in slightly more than half of SC neurons tested could counter or reverse such a transition and limit the duration of sustained activity to play a role in triggering or preventing an ensuing movement. Elucidating the mechanisms by which signaling pathways modulate the conductances of SC neurons could thus shed light on how the SC generates attention and decision-related signals.

Many neurons in the motor layers of the SC discharge with phasic bursts in response to the onset of visual stimuli, and these transient bursts reach frequencies of up to ∼200 spikes/s. Bursts are often followed by a lower frequency, persistent activity referred to as delay period activity that appears while monkeys wait for a cue to make a saccade (Basso and Wurtz 1998; Glimcher and Sparks 1992; Munoz and Wurtz 1995). The persistent activity in SC neurons averages between 20 and 60 spikes/s and, in many neurons, is followed by a second transient burst of action potentials with frequencies as high as 800 spikes/s. It is this second burst that is thought to trigger the onset of a saccade (Sparks 1975; Wurtz and Optican 1994). The timing of the ADP and its rebound spikes correspond well with persistent or delay-period activity but are clearly too slow for burst activity. In our experiments, we generally found that single brief pulses did not elicit persistent activity in slices. Whether the ADP actually contributes to persistent activity in vivo during the relevant behaviors is an important question with clear implications for how the SC processes inputs in preparation for a saccade or orienting movement. If an in vivo role for the ADP can be established, then it will be important to ascertain the relative contributions of intrinsic membrane properties and excitatory synaptic circuitry. In rodents, neurons throughout the SC exhibit persistent activity, and our finding that nearly all neurons in SGS, SO, and SGI produce an ADP represents a striking parallel with the observation of persistent activity in most neurons of these same layers in the rat SC in vivo (Felsen and Mainen 2008; Ngan et al. 2015). However, in the monkey, persistent activity has been reported in the SO and SGI and is relatively rare in the SGS (Goldberg and Wurtz 1972; Li and Basso 2008; Mays and Sparks 1980; McPeek and Keller 2002).

It is significant that intrinsic membrane properties can generate the ADP in the absence of synaptic circuitry. Thus, in contrast to the burst activity, which spreads robustly through its dependence on excitatory synapses (Isa and Hall 2009; Moschovakis 1996; Saito and Isa 2003), the intrinsic nature of the ADP can confine persistent activity to a subset of neurons. These neurons could then be selected by a slight enhancement of their ADP. A signal that reduces Ca²⁺ or K⁺ channel activity could enable the ADP to drive sustained action potential discharges in SC neurons on delay-period timescales. Signals that modulate the ADP may arise from descending inputs in vivo and provide a stimulus that enables repetitive spiking in the SC. Modulation of the ADP would then serve as a cellular mechanism for the triggering of persistent activity seen in vivo that correlates with the delay-period activity seen in tasks involving decision making and shifts of attention. Further modulation of the ADP from descending inputs would resolve the competition among pools of neurons exhibiting delay-period activity leading to saccade-related bursts in monkeys and orientation-related bursts in rodents in a subset of those neurons, culminating with the initiation of the relevant movement.

GRANTS

This work was funded by National Institutes of Health Grants R01EY1019963 (to M. A. Basso and M. B. Jackson), R21EY024153 (to M. A. Basso), and R21NS078301 (to M. B. Jackson).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

N.G., P.O.B., M.A.B., and M.B.J. conception and design of research; N.G. and P.O.B. performed experiments; N.G., P.O.B., and M.B.J. analyzed data; N.G., P.O.B., and M.B.J. prepared figures; N.G., P.O.B., M.A.B., and M.B.J. drafted manuscript; N.G., P.O.B., M.A.B., and M.B.J. edited and revised manuscript; N.G., P.O.B., M.A.B., and M.B.J. approved final version of manuscript; N.G., P.O.B., M.A.B., and M.B.J. interpreted results of experiments.