Complex autonomous firing patterns of striatal low-threshold spike interneurons

Joseph A. Beatty, Matthew A. Sullivan, Hitoshi Morikawa, Charles J. Wilson


During sensorimotor learning, tonically active neurons (TANs) in the striatum acquire bursts and pauses in their firing based on the salience of the stimulus. Striatal cholinergic interneurons display tonic intrinsic firing, even in the absence of synaptic input, that resembles TAN activity seen in vivo. However, whether there are other striatal neurons among the group identified as TANs is unknown. We used transgenic mice expressing green fluorescent protein under control of neuronal nitric oxide synthase or neuropeptide-Y promoters to aid in identifying low-threshold spike (LTS) interneurons in brain slices. We found that these neurons exhibit autonomous firing consisting of spontaneous transitions between regular, irregular, and burst firing, similar to cholinergic interneurons. As in cholinergic interneurons, these firing patterns arise from interactions between multiple intrinsic oscillatory mechanisms, but the mechanisms responsible differ. Both neurons maintain tonic firing because of persistent sodium currents, but the mechanisms of the subthreshold oscillations responsible for irregular firing are different. In LTS interneurons they rely on depolarization-activated noninactivating calcium currents, whereas those in cholinergic interneurons arise from a hyperpolarization-activated potassium conductance. Sustained membrane hyperpolarizations induce a bursting pattern in LTS interneurons, probably by recruiting a low-threshold, inactivating calcium conductance and by moving the membrane potential out of the activation range of the oscillatory mechanisms responsible for single spiking, in contrast to the bursting driven by noninactivating currents in cholinergic interneurons. The complex intrinsic firing patterns of LTS interneurons may subserve differential release of classic and peptide neurotransmitters as well as nitric oxide.

  • basal ganglia
  • tonically active neurons
  • irregular firing
  • burst firing
  • oscillations

in behaving animals, anesthetized animals, and in brain slices, striatal neurons form two classes based on their firing patterns: phasically active neurons and tonically active neurons (TANs) (Albe-Fessard et al. 1960; Alexander and DeLong 1985; Anderson 1977; Connor 1970; Feltz and Albe-Fessard 1972; Kimura 1986; Kimura et al. 1984; Wilson and Groves 1981). Phasically active neurons have hyperpolarized membrane potentials and their firing is driven by excitatory synaptic inputs from the cerebral cortex and thalamus (Kita 1993; Wilson and Groves 1981). Among the phasically active neurons are spiny projection neurons, which in rats constitute approximately 95% of striatal neurons and are the sole output of the striatum (Graveland and DiFiglia 1985), and the fast-spiking parvalbumin-containing interneurons (Mallet et al. 2005). TANs respond with bursts and pauses in their ongoing tonic activity (Aosaki et al. 1994, 1995; Apicella 2007; Kimura 1986; Kimura et al. 1984). TANs are identified by their long-duration action potential waveforms, and distinctive firing rates and patterns (Aosaki et al. 1994; Kimura 1986; Kimura et al. 1984). Striatal cholinergic interneurons in brain slices maintain firing rates and patterns similar to those of TANs in vivo even when disconnected from synaptic input (Bennett and Wilson 1999; Bennett et al. 2000). The autonomous activity of cholinergic interneurons explains the tonic activity of TANs, which fire even during the down state in anesthetized animals, when other striatal neurons are silent and synaptic excitation is greatly reduced (Wilson et al. 1990). Like TANs in vivo (Aosaki et al. 1994), cholinergic interneurons can fire in three different endogenous firing patterns: regular, irregular, and bursting (Bennett and Wilson 1999). Therefore, TANs in the striatum are often viewed as synonymous with cholinergic interneurons.

However, there are a number of striatal neuron subtypes (Chang et al. 1982; Kawaguchi et al. 1990) whose physiologic properties have not been studied in detail because they were difficult to specifically target for electrophysiologic recording. Transgenic mice expressing green fluorescent protein (GFP) under control of specific promoters facilitate identification of striatal interneurons in brain slices (Gittis et al. 2010; Ibáñez-Sandoval et al. 2010, 2011; Partridge et al. 2009). One striatal interneuron is GABAergic and contains multiple neurotransmitters/neuromodulators including somatostatin, neuropeptide-Y (NPY), and nitric oxide (NO). These neurons have the morphologic characteristics of aspiny type III medium neurons (Chang et al. 1982), with extended sparsely branched dendrites and a large and sparse axonal arborization (Kawaguchi 1993). They are named for low-threshold, inactivating calcium channels that produce a low-threshold spike (LTS). Initial studies on these LTS interneurons did not report spontaneous activity (Centonze et al. 2002; Kawaguchi 1993; Koós and Tepper 1999; Kubota and Kawaguchi 2000). However, recent data suggest that these neurons are spontaneously active in brain slices, at least occasionally (Dehorter et al. 2009; Farries and Perkel 2002; Ibáñez-Sandoval et al. 2011; Partridge et al. 2009; Tepper et al. 2010). Here we report the spontaneous firing patterns of these neurons, and the ionic mechanisms underlying them. We conclude that LTS interneurons are a second type of striatal TANs.


All experimental procedures were carried out according to the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and were approved by the University of Texas at San Antonio and University of Texas at Austin Institutional Animal Care and Use Committees. Initial transgenic mice were obtained from the Mutant Mouse Regional Resource Center at University of California at Davis for neuronal nitric oxide synthase (nNOS)-GFP mice [Tg(Nos1-EGFP)185Gsat/Mmcd] or Jackson Laboratory for NPY-GFP mice [B6.FVB-Tg(Npy-hrGFP)1Lowl/J] and breeding colonies were maintained thereafter. All offspring were genotyped; gene-positive nNOS-GFP mice were bred with wild-type Swiss Webster mice and gene-positive NPY-GFP mice were bred with wild-type C57BL/6. Both nNOS-GFP and NPY-GFP mice (2–5 wk old) of either sex were deeply anesthetized with 5% isoflurane in 100% O2 and intracardially perfused with cold (∼2°C), oxygenated (95% O2-5% CO2) slicing medium containing (in mM): 2.5 KCl, 1.25 NaH2PO4, 10.0 MgSO4, 0.5 CaCl2, 26.0 NaHCO3, 10.0 glucose, and 230.0 sucrose. The brains were quickly removed and striatal slices (300 μm thickness) were cut in the sagittal plane using a vibrating slicer and transferred to a holding chamber containing warmed (∼35°C), oxygenated (95% O2-5% CO2) physiologic solution containing (in mM): 126.0 NaCl, 2.5 KCl, 1.25 NaH2PO4, 2.0 MgCl2, 2.0 CaCl2, 26.0 NaHCO3, and 10.0 glucose for 0.5 h, after which the holding chamber was maintained at room temperature for the remainder of the experiment. One hour after slicing, individual slices were transferred to a submersion-type recording chamber and superfused (2–3 mL/min) with oxygenated physiologic solution maintained between 30 and 32°C.

Recording pipettes were pulled from 1.5-mm outer diameter capillary tubing and had tip resistances of 3–5 MΩ. For perforated-patch recordings, electrodes were filled with an intracellular solution containing (in mM): 140.5 KMeSO4, 7.5 NaCl, 10.0 HEPES, and two different concentrations (100 μg/mL or 0.5 μg/mL) of gramicidin, depending on the filtration of the intracellular solution. If filtration of the intracellular solution occurred after addition of gramicidin, a concentration of 100 μg/mL gramicidin was used; however, if the intracellular solution was filtered prior to adding the gramicidin, we found lower concentrations (0.5 μg/mL) of gramicidin were needed presumably because micelles containing dimethyl sulfoxide and gramicidin were removed by the 0.2-μm filter, reducing the effective gramicidin concentration. Perforated-patch electrodes were front-filled with gramicidin-free solution and then back-filled with the solution containing gramicidin. For whole-cell recordings, pipettes were filled with an intracellular solution containing (in mM): 140.5 KMeSO4, 7.5 NaCl, 0.2 EGTA, 10.0 HEPES, 2.0 Mg-ATP, and 0.2 Na-GTP. In some experiments, 20–40 μM Alexa Fluor 594 biocytin (Molecular Probes, Eugene, OR) was added to the electrode solution to visualize the morphology of recorded neurons. Cell-attached recordings were made with both intracellular solutions and in either current clamp with no holding current, voltage clamp at a holding potential of 0 mV, or voltage clamp at a membrane potential that produced minimal current injection to the neuron.

Recordings were made from GFP-positive neurons and were visualized using an Olympus BX51WI microscope equipped with Ultima scanhead and detectors (Prairie Technologies, Middleton, WI). Fluorescent emissions from GFP and Alexa Fluor 594, both excited at 810 nm, were separated using a dichroic beamsplitter (575) and barrier filters (525/70 and 607/45) from Chroma (Rockingham, VT). Image stacks were deconvolved using AutoQuant X (Media Cybernetics, Bethesda, MD) and maximum projection images were generated using Imaris (Bitplane, Zurich, Switzerland).

Data were acquired using a Multiclamp 700B amplifier (Molecular Devices, Union City, CA) in either current-clamp or voltage-clamp mode. Data were filtered at 10 kHz and digitized at 10 or 20 kHz using a HEKA ITC-18 digitizer, and collected using IGOR Pro 5 (WaveMetrics, Lake Oswego, OR). Mean firing rates of autonomous activity were calculated as the inverse of the mean interspike interval (ISI) over a 60-s recording. Input resistances were calculated from −10-mV steps during voltage-clamp recordings. The frequencies of the subthreshold oscillations were quantified using discrete Fourier transforms of 60-s samples and measuring the power spectral densities (PSDs) over the frequency range of 0.1–100 Hz. The median frequencies were then determined by first-order interpolations of the cumulative probabilities of the PSDs over the range of 0.1–100 Hz. Maximal amplitudes of the subthreshold oscillations were calculated as the median membrane voltages over the full 60-s samples subtracted from the maximal positive voltage deflections observed. Data were analyzed using custom routines programmed using Mathmatica (Wolfram Research, Champaign, IL). All data are presented as mean ± SD unless otherwise noted. Differences between means were considered significant when P < 0.05 using parametric statistical tests when data passed a normality test or nonparametric statistical tests when the data were not normally distributed. The junction potential was determined to be 8 mV and has been corrected for in all whole-cell voltage measures. All drugs were diluted to a final concentration from stock solutions and bath applied. All compounds were purchased from Sigma (St. Louis, MO) or Tocris (Ellisville, MO).


Identification of striatal LTS interneurons.

Electrophysiologic studies of striatal LTS interneurons have been limited due to the difficulty in targeting these neurons because of their low numbers (∼1%) and lack of clearly distinguishable morphologic characteristics under enhanced contrast microscopy used for brain slice electrophysiology (Centonze et al. 2002; Kawaguchi 1993; Kubota and Kawaguchi 2000; Larsson et al. 2001; West et al. 1996). To address this issue, we recorded GFP-positive neurons within the striatum of transgenic mice expressing GFP controlled by either the nNOS or NPY promoters to target LTS interneurons specifically (Fig. 1A), since most LTS interneurons express both nNOS and NPY (Dawson et al. 1991; Rushlow et al. 1995; Vincent and Hope 1992; Vincent et al. 1983). The morphologic features of GFP-positive neurons were revealed by intracellular labeling with 20–40 μM Alexa Fluor 594 and two-photon imaging at the time of the experiment (Fig. 1, B–D). Most nNOS-GFP-positive and NPY-GFP-positive neurons possessed the characteristic morphologic properties of LTS interneurons first described by Kawaguchi (1993), including fusiform somata with few (2–5) primary dendrites that are poorly branched and extended long distances (> ∼150 μm) from the soma (Fig. 1, B and D). They also displayed the same defining electrophysiologic properties of LTS interneurons reported in previous studies (Kawaguchi 1993; Koós and Tepper 1999), including prolonged (> ∼50 ms) afterspike hyperpolarizations, large input resistances (852 ± 272 MΩ, n = 46), and rebound bursts following hyperpolarizing current pulses (Fig. 1, E and F).

Fig. 1.

Identification of green fluorescent protein (GFP)- positive striatal neurons as low-threshold spike (LTS) interneurons. A–C: maximum projection two-photon microscopy images of a neuropeptide-Y (NPY)-GFP-positive striatal neuron (A) filled intracellularly with 40 μM Alexa Fluor 594 (B) and the merged two-channel image (C). D: maximum projection image of an intracellularly labeled (40 μM Alexa Fluor 594) neuronal nitric oxide synthase (nNOS)-GFP-positive striatal neuron depicting the stereotypical morphologic properties of the LTS interneuron. E and F: electrophysiologic properties of an NPY-GFP-positive striatal neuron recorded in the perforated-patch configuration. Average of 50 spontaneous action potential waveforms (E). Voltage response to a hyperpolarizing current pulse reveals electrophysiologic properties characteristic of striatal LTS interneurons, including a large input resistance and LTS burst on removal of hyperpolarizing current (same neuron as E) (F).

Some (29/81) GFP-positive neurons recorded from the nNOS-GFP mice had both morphologic and electrophysiologic properties of spiny projection neurons (hyperpolarized resting potentials, low input resistance, inward rectification, and a delay to the first spike; data not shown). Immunohistochemical staining confirmed that some striatal GFP-positive neurons in the nNOS-GFP animals were negative for nNOS (data not shown). Therefore, in nNOS-GFP mice, only GFP-positive neurons that could be morphologically as well as electrophysiologically identified as LTS interneurons, based on the above-stated characteristics, were included in this study (41/81). Due to the apparently ectopic expression seen in the nNOS-GFP mice we sought another mouse strain that more specifically labels LTS interneurons. This problem was solved using the NPY-GFP mouse. The NPY-GFP mouse strain has previously been shown to contain two electrophysiologically and morphologically distinct NPY-positive interneuron subtypes, one previously described as LTS interneurons and the other a novel non-LTS interneuron subtype that can be distinguished in fluorescent microscopy (Ibáñez-Sandoval et al. 2011). We have encountered both neurons within the striatum but have limited this study to only that of LTS interneurons. All results obtained with LTS interneurons in the nNOS-GFP animals were confirmed in NPY-GFP animals. There were no substantial differences between the two samples of neurons, and the results from the two groups of neurons were pooled except where otherwise noted. A total of 73 mice were used in this study (nNOS-GFP = 32 mice, NPY-GFP = 41 mice) aged 23 ± 5 postnatal days.

Striatal LTS interneurons are autonomously active in brain slices.

The earlier intracellular recording studies of striatal LTS interneurons did not report spontaneous activity (Centonze et al. 2002; Kawaguchi 1993; Koós and Tepper 1999; Kubota and Kawaguchi 2000). More recent studies have reported that LTS interneurons may be spontaneously active in brain slices (Dehorter et al. 2009; Farries and Perkel 2002; Ibáñez-Sandoval et al. 2011; Partridge et al. 2009; Tepper et al. 2010), but not all neurons were firing and the mechanism responsible for spontaneous activity was not determined. We performed extracellular cell-attached recordings to avoid possible changes in firing caused by intracellular recording. Cell-attached recordings were performed on 82 LTS interneurons and spontaneous activity was present in 91% (75/82). LTS interneurons possessed a wide range of firing rates (mean = 8.15 ± 4.92 Hz, range = 0.99 to 22.83 Hz, measured from 60-s samples, n = 75) and had complex patterns of activity. Neurons varied widely in how regularly they fired, as measured by the coefficient of variation (CV) of the ISI (mean = 0.476 ± 0.693, range = 0.065 to 4.391). No difference was seen between the mean firing rates of LTS interneurons recorded from nNOS-GFP mice (7.15 ± 4.79 Hz, n = 23) and NPY-GFP mice (8.59 ± 4.96 Hz, n = 52, P > 0.05, Mann-Whitney U test).

Although our initial data confirm that LTS interneurons are spontaneously active, they do not specify whether the spontaneous activity is dependent on synaptic input or intrinsic to the individual neurons (i.e., autonomous). To address this issue, we examined the effects produced by blockade of major neurotransmitter receptors within the striatum. Neurotransmitter receptor antagonists 6,7-dinitoquinoxaline-2,3-dione (DNQX; α-amino-3-hydroxy-5-methyl-4-isoazolepropionic acid [AMPA] antagonist, 10 μM), d(−-2-amino-5-phosphonopentanoic acid (D-AP5; N-methyl-d-aspartate [NMDA] antagonist, 50 μM), picrotoxin (γ-aminobutyric acid type A [GABAA] antagonist, 100 μM), hexamethonium (nicotinic antagonist, 50 μM), and scopolamine (muscarinic antagonist, 1 μM) were bath applied together (10–15 min, n = 6) and had no effect (P > 0.05, Wilcoxon signed-rank test) on the mean firing rate (before: 6.54 ± 4.32 Hz vs. after: 6.57 ± 4.46 Hz) or mean CV (before: 0.436 ± 0.346 vs. after: 0.424 ± 0.339). These results indicate that the spontaneous activity observed in LTS interneurons is not dependent on common synaptic inputs, but is most likely due to intrinsic factors that are responsible for the sustained activity much like the other autonomously active striatal interneuron, the cholinergic interneuron (Bennett et al. 2000).

LTS interneurons display complex autonomous firing patterns.

Striatal cholinergic interneurons display complex autonomous firing patterns in brain slices (Bennett and Wilson 1999), consisting of regular, irregular, and bursting firing, and spontaneous transitions between them. This strengthens the argument that they are TANs as seen in vivo, because striatal TANs exhibit a similar variety of firing patterns and spontaneous transitions between regular and irregular firing patterns (Aosaki et al. 1994). LTS interneurons in our study similarly fired in a wide range of patterns, and the activity of individual neurons often spontaneously shifted between patterns. In the most extreme cases, these patterns ranged from regular single-spike firing to rhythmic-burst firing, with the majority of neurons exhibiting transitions in pattern along a continuum from regular single spiking to complex irregular firing, punctuated with occasional bursts and pauses. Figure 2A shows examples of these three firing modes (regular, irregular, and irregular-burst) recorded from 10-min samples from three LTS interneurons firing at about the same rate. One neuron (Fig. 2A1) fired in a regular pattern that produced a narrow unimodal Gaussian distribution of ISIs, a relatively low CV, and a periodic autocorrelogram. An example showing irregular firing (Fig. 2A2) exhibits a broad unimodal ISI distribution that is skewed toward longer ISIs and a larger CV than that seen during regular firing. The corresponding autocorrelograms often displayed no clear structure or occasionally one early peak (Bar-Gad et al. 2001). An example showing irregular-burst firing (Fig. 2A3) exhibits bursts and pauses intermixed with irregular firing, an ISI distribution that is bimodal and skewed, with the first peak corresponding to the within-burst ISI distribution and the second peak corresponding to the ISI distribution during interleaved episodes of irregular firing. This mode of firing also resulted in two peaks in the autocorrelogram and led to the highest CVs encountered.

Fig. 2.

LTS interneurons display complex autonomous firing patterns. A: examples of the range of LTS interneuron firing patterns encountered from cell-attached recordings (10-min samples). Columns are (from left to right) sample of inverted voltage-clamp recording of autonomous firing, interspike interval (ISI) histogram, and autocorrelogram (bin width = 1 ms). Firing rates and coefficients of variation (CVs) shown in middle columns are mean values taken from the full 10-min samples. Regular firing (A1) displayed a unimodal Gaussian distribution in the ISI histogram and multiple evenly spaced peaks in the autocorrelogram. Irregular firing (A2) exhibited a unimodal skewed distribution in the ISI histogram and no clear peak in the autocorrelogram. Irregular-burst firing (A3) alternated between an irregular firing pattern and a bursting pattern, which produced a bimodal distribution in the ISI histogram and two distinct peaks in the autocorrelogram. B: population graph displaying mean CV vs. mean firing rate for all 60-s samples of cell-attached activity. Dotted line corresponds to Poisson firing (CV = 1). Green circles correspond to LTS interneuron illustrated in A1. Blue circles correspond to LTS interneuron illustrated in A2. Red circles correspond to the LTS interneuron illustrated in A3. C: graph of mean CV vs. mean firing rate for five LTS interneurons, each labeled by color displaying their entire recorded activity divided into 60-s samples.

LTS interneurons did not consistently fire in one pattern, but showed gradual shifts in rate and regularity, and occasionally abrupt transitions between patterns. To analyze these transitions in firing, we collected spontaneous activity during cell-attached recordings from 75 spontaneously active LTS interneurons, divided the recording times (1 to 41 min) into 1-min samples, and plotted the mean CV against the mean firing rate for each of the 987 1-min samples (Fig. 2B). Completely random (i.e., Poisson) firing (CV = 1) is represented on the graph as a dotted line. Burst firing increases the CV, and the neurons that burst the most had a CV >1. The 30 colored dots represent the 10 min of activity for the three neurons shown in Fig. 2A. Most 1-min samples of data lay along a curve with irregular firing (higher CVs) at low rates and more regular firing (lower CVs) with increasing firing rate. A smaller set of points fell off this curve, having a higher CV than that of the remaining group, regardless of rate. To determine whether this range of rates and patterns corresponded to discrete groups of LTS interneurons with different firing patterns or to variation within a single neuronal subtype, we compared the samples from individual neurons in which multiple 1-min samples of data were collected. When the activity from individual neurons was examined, 69% varied in rate and regularity but remained consistent with the overall trend (regularity varying continuously with rate). Six percent of the neurons produced samples that lay completely off the general trend, possessing bursts or very irregular regardless of rate, and 25% of neurons showed continuous variation of regularity, with rate over some portion of the recording, but also abrupt transitions to burst firing. This mixture of regular, irregular, and burst firing is similar to that previously reported for striatal cholinergic interneurons (Bennett and Wilson 1999; Goldberg et al. 2009). Examples of the activity of five individual neurons are shown in Fig. 2C, coded by color. The green and magenta dots represent two neurons with high rates of firing and low CVs. These neurons fluctuated in rate but maintained regular firing, as evidenced by their low CVs, which changed predictably with rate. The black dots represent a neuron with low CVs at moderate rates but became more irregular and eventually produced bursts of spikes as its firing slowed. The blue and red dots belong to neurons that are transitioning between modes. The blue neuron fired somewhat irregularly but with CVs <1 and spontaneously shifted to burst firing with little change in firing rate. The red neuron displayed periods of regular and moderately irregular firing as it spontaneously changed firing rate, but abruptly shifted to burst firing, which was maintained for several minutes while firing between 10 and 15 Hz. The spontaneous shifts in rate seen in all neurons were usually gradual and did not have any preferred direction, often reversing after a period of time, and so were not due to any general trend during recording. In 3 of the 75 neurons studied in this way, fast synaptic transmission was blocked by treatment with 10 μM DNQX, 50 μM D-AP5, and 100 μM picrotoxin and those neurons did not differ from the others in the pattern, rate, or variability of their firing. Thus the spontaneous variations in firing pattern cannot be attributed to any residual synaptic interactions between neurons in the slice.

These results indicate that the variation in firing rates and patterns seen within the sample of LTS interneurons do not result from the presence of qualitatively different classes of neurons, but from temporal fluctuations in the firing patterns of a single class of LTS interneuron. Individual LTS interneurons have a dynamic range of firing, being capable of both single spiking and burst firing. Our next experiments were designed to determine the mechanisms underlying the generation of regular, irregular, and burst firing in individual neurons.

Mechanisms contributing to the single spiking firing pattern.

Because the neurons are never deeply hyperpolarized, it is unlikely that low-threshold, inactivating (T-type) calcium channels contribute to the single spiking autonomous activity. In many neurons, including the striatal cholinergic interneurons, this firing pattern depends on a tetrodotoxin (TTX)-sensitive persistent sodium current active at subthreshold voltages (e.g., Bennett et al. 2000). To examine if LTS interneurons possess a persistent sodium current that activates in the subthreshold range, perforated-patch voltage-clamp recordings were performed on six LTS interneurons before and after the application of TTX. Access resistances for these recordings were <50 MΩ (mean: 39 ± 5 MΩ) and produced voltage errors <5 mV between the voltages −100 and −30 mV and were not corrected. One-second voltage steps were applied from −50 mV to potentials ranging from −100 to −20 mV by 5-mV increments in control artificial cerebrospinal fluid (ACSF) and following 1 μM TTX application (Fig. 3A). This holding potential was selected to avoid a contribution by low-threshold inactivating calcium currents, but depolarizing voltage pulses could still evoke sodium-dependent action currents. Currents were measured at the end of the 1-s voltage step to obtain steady-state current readings (mean of last 250 ms). In control ACSF (Fig. 3B, blue), LTS interneurons' current vs. voltage (I–V) curve showed no zero-current point in the subthreshold range, indicating there is no stable subthreshold resting membrane potential for these neurons. The same I–V curve also displayed a negative slope conductance in the range of −60 to −40 mV, indicating that the balance of persistent currents is inward and will depolarize the neurons to the action potential threshold. In the presence of 1 μM TTX (Fig. 3B, red), the same voltage steps in the same neurons produced a zero-current crossing at −40 mV and eliminated the negative-slope conductance seen in the control condition. We estimated the TTX-sensitive inward current by subtracting the TTX I–V curve from the ACSF I–V curve and plotting the difference (Fig. 3B, green). This current activated near −60 mV and persisted to −20 mV. The same experiment was conducted in the whole-cell configuration (n = 13) and yielded similar results to those seen with perforated-patch recordings. In ACSF, LTS interneurons' whole-cell I–V curve also exhibited no zero-current point in the subthreshold range and had a negative slope conductance that activated near −60 mV. TTX (1 μM) also eliminated the negative-slope conductance seen in the whole-cell recordings, although the entire I–V curve was shifted in an outward current direction (20–40 pA) and produced a zero-current crossing at −63 mV. This outward drift is likely due to whole-cell dialysis compared with our perforated-patch recordings. However, the persistent sodium current was not greatly affected by whole-cell dialysis. These results indicate that LTS interneurons have a TTX-sensitive persistent sodium current that activates at membrane potentials more depolarized than −60 mV, exceeds the steady-state outward currents active in this voltage range, and depolarizes the neuron to action potential threshold.

Fig. 3.

Intrinsic mechanisms governing the autonomous activity of LTS interneurons. A: in perforated-patch voltage-clamp recordings, voltage steps from −50 to −45 mV elicited an inward current (top traces, artificial cerebrospinal fluid [ACSF]) that was blocked by 1 μM tetrodotoxin (TTX) (bottom traces). Current deflections truncated for ease of viewing. B: current vs. voltage command (Vcmd) graph (n = 6, mean ± SE) shows a negative-slope conductance in the range of −60 to −40 mV in control (ACSF, blue) conditions. That conductance is blocked by 1 μM TTX (red) and is plotted as the difference between TTX and ACSF (green). C: an example of autonomous firing of LTS interneuron recorded in the perforated-patch configuration before (top trace, ACSF) and after application of 1 μM TTX (bottom trace) revealing TTX-insensitive oscillations. To the right of each trace are their discrete Fourier transforms with power spectral density (PSD) analysis.

To confirm that the persistent sodium current was responsible for spontaneous activity, we recorded from LTS interneurons intracellularly in current-clamp mode before and after application of 1 μM TTX to eliminate sodium-dependent action potentials. We reasoned that any TTX-insensitive contribution to spontaneous firing might appear as a subthreshold oscillation after TTX application. Initial intracellular current-clamp recordings were carried out in the whole-cell configuration; however, autonomous activity proved difficult to maintain for extended periods of time using this recording method. A similar phenomenon is seen in striatal cholinergic interneurons (Bennett et al. 2000; Wilson 2005). We thus used perforated-patch recordings for the remainder of the study to examine the origins of the autonomous activity. In all autonomously active LTS interneurons tested (n = 15), 1 μM TTX eliminated the sodium-dependent action potentials and revealed spontaneous TTX-insensitive oscillations of the membrane potential (Fig. 3C). The TTX-insensitive oscillations could not be observed in whole-cell recordings. As seen in Fig. 3C, these oscillations had maximum amplitudes between 3 and 15 mV (mean = 7.7 ± 3.8 mV at 0 pA injected current, n = 15) and had a lower frequency than the sodium-dependent autonomous firing frequency seen in the same neuron before TTX. We quantified the frequencies of the TTX-insensitive oscillations from 60-s episodes of data using the discrete Fourier transform over a range of 0.1–100 Hz. We then determined the median frequencies by first-order interpolations of the cumulative probability of the PSD over the range of 0.1–100 Hz (Figs. 3C and 4B). The frequencies of the spontaneous TTX-insensitive oscillations (2.57 ± 2.24 Hz at 0 pA injected current, n = 15) were significantly lower than the autonomous firing frequencies (8.76 ± 5.09 Hz, n = 15) of the same neurons prior to 1 μM TTX application (P < 0.05, Student's t-test following Kolmogorov-Smirnov [K-S] test for normality). The variance around the peak frequency of the subthreshold oscillation (Fig. 3C) was caused by irregularity of the oscillation as seen in Figs. 3C and 4A. To determine if the TTX-insensitive oscillations were voltage dependent, we injected constant depolarizing or constant hyperpolarizing currents of varying intensities and examined their effects on the oscillation frequencies. Constant depolarizing current injections increased the frequency of the oscillation and constant hyperpolarizing currents decreased the frequency, eventually eliminating the oscillation at sufficiently low membrane potentials (Fig. 4, A and B). The median frequency of the oscillation was plotted against the median membrane voltage during the 60-s sample, binned into 5 mV, and plotted as mean ± SD (Fig. 4C, n = 15). The frequency of the oscillation was dependent on membrane voltage. When membrane voltages rose above −45 mV, clear oscillations were present. These oscillations increased in frequency with increasing depolarization, reaching a maximum of around 6 Hz at a median membrane potential of −30 mV. The increase in frequency implies an increase in the current responsible for the oscillation, suggesting that a noninactivating, depolarization-activated inward current that is insensitive to TTX generates the membrane potential oscillation.

Fig. 4.

The spontaneous TTX-insensitive oscillations of LTS interneurons are voltage dependent. A: the effects of constant intracellular current injections on TTX-insensitive (1 μM) oscillations in perforated-patch current-clamp recordings. B: the discrete Fourier transforms with PSD analysis for three example traces in A (1 μM TTX/20 pA, 1 μM TTX/5 pA, and 1 μM TTX/0 pA). C: median frequency vs. median voltage of spontaneous TTX-insensitive oscillations graph (n = 15, mean ± SD, bin width = 5 mV).

To examine if the TTX-insensitive oscillation is calcium dependent, we attempted to block it with the nonspecific calcium channel blocker cadmium. Application of 400 μM cadmium in the presence of 1 μM TTX eliminated the TTX-insensitive oscillation (Fig. 5, n = 9). In addition, 400 μM cadmium significantly hyperpolarized the median membrane potential to −49.0 ± 5.2 mV, suggesting a resting calcium current is maintained in LTS interneurons (1 μM TTX: −39.1 ± 9.4 mV, n = 9, P < 0.05, Wilcoxon signed-rank test). Although TTX-insensitive membrane potential oscillations were routinely observed at this level of depolarization in the absence of cadmium (see Fig. 4), we controlled for the hyperpolarizing effect of cadmium by adjusting the membrane potential to precadmium levels or higher using constant depolarizing current. The TTX-insensitive membrane potential oscillations were abolished by cadmium at all membrane potentials tested (Fig. 5, bottom trace). These results suggest that these voltage-dependent oscillations are dependent on cadmium-sensitive calcium channels, but not low-threshold, inactivating calcium channels, which should be inactivated at these voltage levels.

Fig. 5.

Cadmium eliminates the spontaneous TTX-insensitive oscillations in LTS interneurons. An example of autonomous firing in control conditions (ACSF) and the TTX-insensitive oscillations following 1 μM TTX. Additional application of cadmium (1 μM TTX/400 μM cadmium) eliminated the oscillations and hyperpolarized the neuron. Even when the neuron was brought back into the voltage activation range of the oscillations by injection of intracellular current, no oscillations were seen, indicating these oscillations were calcium dependent (1 μM TTX/400 μM cadmium/40 pA constant current).

These results suggest that the single-spiking firing pattern is generated by two different oscillatory mechanisms acting simultaneously: one generated by persistent Na+ current and one based on a persistent Ca2+ current. Because these two oscillatory mechanisms had different frequencies but are both voltage dependent, the net membrane potential oscillation reflects their interaction. This interaction may be the origin of the irregular pattern of spontaneous activity, similar to that previously suggested for cholinergic interneurons (Wilson 2005) and dopaminergic neurons (Wilson and Callaway 2000).

Hyperpolarization-induced bursting.

The regular and irregular single-spiking firing patterns did not rely on the occurrence of low-threshold spikes or on the inactivating calcium channels believed to be responsible for them. In anesthetized animals, striatal LTS interneurons have been reported to fire predominantly in the bursting mode (Sharott et al. 2009). LTS-driven bursts in thalamic neurons predominate in anesthetized animals, because the neurons are dominated by synaptic inhibition that hyperpolarizes the membrane and removes inactivation from inactivating calcium channels (Llinás and Steriade 2006). To determine whether hyperpolarization might release the bursting pattern in striatal LTS interneurons, we examined the neurons' responses to brief hyperpolarizing current pulses, and to constant hyperpolarizing current injection. LTS interneurons' firing in the single-spiking pattern transiently changed their firing rates during injection of 1-s current pulses, but their firing patterns were unchanged (Fig. 6A). The offset of hyperpolarizing pulses triggered bursts and subsequent afterhyperpolarizations, but firing returned to the single-spiking pattern within 1 or 2 s. Like cholinergic interneurons, LTS interneurons' firing rates varied linearly from 0 to about 35 Hz, with current varying from −20 to 60 pA (n = 12, Fig. 6B). Small hyperpolarizing currents (−5 to −10 pA) were capable of reducing the mean firing frequencies and stronger hyperpolarizations eventually led to elimination of firing during the 1-s current pulse (Fig. 6A, bottom trace). Short hyperpolarizing current pulses could slow the neurons to arbitrarily low firing rates, suggesting these neurons are class I, as defined originally by Hodgkin (1948). However, if hyperpolarizing currents were maintained for several seconds, the membrane potential would gradually depolarize and many of the neurons would resume burst firing while the current was maintained. Twenty LTS interneurons firing in the single-spiking mode were hyperpolarized with constant current. Of these, 15 (75%) shifted to a burst firing pattern with firing consisting of approximately 1- to 2-s bursts of action potentials, alternating with a slow hyperpolarization that could last approximately 2–5 s before giving way to another burst (Fig. 6C). The remaining five neurons maintained a hyperpolarized membrane potential with no activity present. Two LTS interneurons that were autonomously firing in a consistent bursting pattern were tested with constant depolarizing current. Constant depolarizing current eliminated burst firing and gave rise to irregular single spiking in both LTS interneurons tested (Fig. 6D). Therefore, striatal LTS interneurons undergo a discrete change from single spiking to bursting when subjected to tonic hyperpolarization, as in thalamic neurons (Llinás and Steriade 2006). This occurs not only by removal of inactivation from inactivating low-threshold calcium channels (as in the thalamus) but also because hyperpolarization moves the membrane potential out of the activation range of the oscillatory mechanisms responsible for single spiking.

Fig. 6.

Modulation of LTS interneuron firing rates and patterns by brief or constant intracellular current injections. A: examples of brief (1-s) intracellular depolarizing (top trace) and hyperpolarizing (bottom trace) current steps. B: frequency vs. current step graph for LTS interneurons (n = 12) given 1-s hyperpolarizing and depolarizing intracellular current steps (mean ± SD). C: an example of LTS interneuron (same neuron as A) with 0 pA of intracellular constant current applied exhibiting autonomous single spiking (top trace) and with −15 pA of intracellular constant current applied induced burst firing (bottom trace). D: an example of a different LTS interneuron that displayed burst firing with 0 pA of intracellular constant current applied (top trace) and with 15 pA of intracellular constant current applied produced single spiking (bottom trace).


The identity of TANs.

It is well established that cholinergic interneurons are TANs, but it has not been clear if all TANs are cholinergic interneurons. There are a variety of striatal interneurons that have not been characterized (Chang et al. 1982), some of which may be tonically active. Our results indicate that at least one common striatal interneuron type, the LTS interneuron, is also tonically active. Like the cholinergic interneuron, the LTS interneuron has no stable membrane equilibrium potential (no zero-crossing in the steady-state I–V curve) and this is ensured by a substantial persistent sodium current that guarantees that the neuron will fire continuously, even in the complete absence of synaptic input. When constantly hyperpolarized, the neuron's response is to fire bursts, rather than to go quiet. LTS interneurons exhibit a range of autonomous firing patterns, from regular single spiking to bursting, all of which can be seen in the absence of any similarly patterned synaptic input, as in cholinergic interneurons.

The LTS interneuron and the cholinergic interneuron are primarily modulatory in function in the striatum. The LTS interneuron is GABAergic, but its direct inhibitory effect on neighboring projection neurons may be weak (Gittis et al. 2010; but see also Koós and Tepper 1999). In addition, it releases nitric oxide, somatostatin, and NPY, all of which, similar to acetylcholine from cholinergic interneurons, are key controllers of excitability or synaptic plasticity in the neostriatum. The tonically active network of the striatum, consisting of the cholinergic interneurons and LTS interneurons (and perhaps others), maintains a background of neurotransmitters and neuromodulators within the striatum by continuous release, even in the absence of the specific inputs that drive the firing of phasically active neurons. Synaptic input can trigger both bursts of firing and pauses in the intrinsic firing pattern, thereby producing increases, as well as decreases, in tonic levels of these modulators.

The mechanism of complex autonomous firing patterns.

Many neurons in the basal ganglia exhibit autonomous activity in the form of rhythmic single spiking, mostly via a relatively simple mechanism. Over most of the subthreshold range of membrane potentials visited during the interspike interval, the neurons have a net inward current, because of a persistent sodium current, perhaps assisted by a small contribution from hyperpolarization-activated cyclic nucleotide-gated (HCN) current (Bevan and Wilson 1999; Chan et al. 2004). After each action potential, the membrane potential is reset by spike-triggered afterhyperpolarization currents, which decay after the action potential, leaving the membrane potential in the voltage range dominated by depolarizing inward current. This sort of rhythmic single spiking may be perturbed by intrinsic membrane noise, or by uncorrelated background synaptic input, to produce more irregular firing patterns. Neurons firing spontaneously in this way exhibit a simple relationship between firing rate and regularity, being more irregular at low rates. This occurs because of the accumulation of the effects of perturbations that generate noise in the intervals between spikes (Ermentrout 2010).

In contrast, some basal ganglia neurons, including striatal cholinergic interneurons, produce more complex patterns of activity (Bennett and Wilson 1999; Bennett et al. 2000; Wilson 2005). In these neurons, patterns of activity include regular single spiking, but also very irregular patterns including bursts and pauses. Although these variations from the regular spiking pattern may be triggered by specific patterns of synaptic input, they continue to be seen when synaptic input is blocked. In cholinergic interneurons, the variety of patterns of autonomous activity arise from the interaction between three different cellular mechanisms of oscillation: 1) a regular single-spiking oscillation that depends on persistent sodium current, 2) a slower subthreshold oscillation that depends on hyperpolarization-activated potassium conductance causing irregular firing, and 3) a slow bursting mechanism that depends on noninactivating calcium conductance. All of these are active in the same neurons, and interact through the membrane potential and intracellular calcium concentration (Bennett et al. 2000; Goldberg and Wilson 2005; Goldberg et al. 2009; Wilson 2005; Wilson and Goldberg 2006). Because they have different natural frequencies, their interaction can produce a variety of firing patterns, depending on their relative strengths in control of firing.

The LTS interneurons studied here also exhibited a variety of autonomous firing patterns, including rhythmic and irregular single spiking, and spontaneous bursts and pauses. They could spontaneously change the relative contribution of regular and irregular firing and bursting to their firing pattern over the course of a few minutes. The mechanism of rhythmic single spiking relied on persistent sodium current and spike-triggered afterhyperpolarization currents, as in other neurons. Blockade of sodium currents with TTX revealed a second oscillatory mechanism in the same neurons, operating over the same voltage range, but with a different natural frequency. Unlike cholinergic interneurons, this second oscillatory mechanism in LTS interneurons was not generated by hyperpolarization-activated currents, but rather by TTX-insensitive noninactivating calcium currents. The calcium current responsible for this oscillation was activated in the voltage range normally visited by the membrane potential during the interspike interval. This mechanism would normally expect to exert a perturbing influence on the faster sodium-based oscillation that dominates in single spiking, and may be responsible for the irregularity of firing seen in LTS interneurons. A third oscillation was seen when applied current hyperpolarized the membrane potential. This oscillatory process was not activated during single spiking in most neurons, as indicated by the fact that it could not be expressed by 1-s hyperpolarizing pulses, but required several seconds. This mechanism resembles that observed in thalamocortical neurons, which exhibit similar low-threshold spikes, caused by removal of inactivation from T-type calcium conductance (Jahnsen and Llinás 1984). Voltage-dependent switching to burst firing in thalamic neurons requires removal of inactivation from channels that are inactivated during single spiking, and this is consistent with the activity we have observed in striatal LTS interneurons. It is also consistent with our observation of some LTS interneurons that were spontaneously bursting, and could be switched to a single-spiking pattern by passage of constant depolarizing current.

The function of complex autonomous firing patterns.

Synaptic inputs to LTS interneurons may determine not only their firing rates, but also their pattern of firing. For example, dopamine produces long-lasting depolarizations in LTS interneurons via D1-like dopamine receptors in brain slices (Centonze et al. 2002), and so should position LTS interneurons in a single-spiking firing pattern. Removal of tonic dopamine might hyperpolarize these neurons and shift them to a burst-firing mode. Recent evidence suggests that LTS interneurons shift from a single-spiking pattern to a burst pattern in brain slices following dopamine depletion (Dehorter et al. 2009). Our results show that any sustained hyperpolarization can shift these neurons to a burst-firing mode and, in turn, any depolarizing effect can elicit more regular single-spike firing.

The autonomous single-spiking activity in LTS interneurons maintains a background synaptic release of neurotransmitters whose presence is required for the function of the striatum, and allows for both increases and decreases in their release. Cholinergic interneurons release mainly acetylcholine, which alone can produce several direct modulatory effects on spiny projection neurons in addition to modulating synaptic plasticity (Akins et al. 1990; Calabresi et al. 1998; Centonze et al. 2003; Gabel and Nisenbaum 1999; Shen et al. 2005; Zhou et al. 2001). LTS interneurons contain the largest variety of neurotransmitters (i.e., somatostatin, nitric oxide, NPY, and GABA) of any known interneuron in the striatum and these neurotransmitters may have different functional roles in the striatal circuitry (Kawaguchi et al. 1995). Somatostatin modulates the excitability of striatal projection neurons via effects on calcium currents and calcium-dependent potassium currents (Galarraga et al. 2007; Vilchis et al. 2002). It also regulates the effectiveness of lateral inhibitory synaptic connections among spiny projection neurons (Lopez-Huerta et al. 2008). Nitric oxide regulates striatal cAMP via phosphodiesterase (Lin et al. 2010) and controls excitatory synaptic transmission and synaptic plasticity in striatal projection neurons (Calabresi et al. 1999; Kawaguchi et al. 1995; Sammut et al. 2010). NPY and nitric oxide regulate dopamine release in the striatum (Adewale et al. 2007; Hartung et al. 2011).

Neurons that colocalize classical neurotransmitters and neuropeptides often release these chemicals in a frequency-dependent manner, with only classical neurotransmitter release occurring at low frequencies and the release of both classical neurotransmitters and neuropeptides with high-frequency bursts of activity (Bartfai et al. 1988). The complex autonomous firing patterns in striatal LTS interneurons may be a specialization for the differential control of neurotransmitter release, in which the multiple firing patterns preferentially release various neurotransmitters that may produce completely different modulatory functions within the striatum.


This work was supported by National Institutes of Health (NIH) Grants NS-37760, NS-072197 (to C.J.W.), DA-015687 and AA-015521 (to H.M.). The multiphoton imaging facility at University of Texas at San Antonio was supported by NIH Grant RR-13646.


No conflicts of interest, financial or otherwise, are declared by the author(s).


J.A.B., M.A.S., H.M., and C.J.W. conception and design of research; J.A.B., M.A.S., and C.J.W. performed experiments; J.A.B., M.A.S., H.M., and C.J.W. analyzed data; J.A.B., H.M., and C.J.W. interpreted results of experiments; J.A.B. prepared figures; J.A.B. drafted manuscript; J.A.B., M.A.S., H.M., and C.J.W. edited and revised manuscript; J.A.B., M.A.S., H.M., and C.J.W. approved final version of manuscript.


View Abstract