Calcium-activated non-selective cation currents are involved in generation of tonic and bursting activity in dopamine neurons of the substantia nigra pars compacta


  • Ana Mrejeru,

    1. Committee on Neurobiology, University of Chicago, Chicago, IL 60637, USA
    Search for more papers by this author
  • Aguan Wei,

    1. Center for Integrative Brain Research, Seattle Children's Research Institute, Seattle, WA 98101, USA
    2. Department of Neurological Surgery, University of Washington, Seattle, WA 98101, USA
    Search for more papers by this author
  • Jan Marino Ramirez

    1. Committee on Neurobiology, University of Chicago, Chicago, IL 60637, USA
    2. Center for Integrative Brain Research, Seattle Children's Research Institute, Seattle, WA 98101, USA
    3. Department of Neurological Surgery, University of Washington, Seattle, WA 98101, USA
    Search for more papers by this author

Corresponding author J. M. Ramirez: Center for Integrative Brain Research, Seattle Children's Research Institute, 1900 Ninth Avenue, Seattle, WA 98101, USA.  Email:


Non-technical summary  The loss of dopamine-containing neurons within the substantia nigra has been implicated in Parkinson's disease. Thus, the question why these neurons are particularly vulnerable to excitotoxicity has received considerable attention. Under physiological conditions dopamine neurons can generate a burst of activity that seems to require NMDA receptors and that is activated by high frequency glutamatergic inputs. Here, we show in a brain slice preparation of mice that NMDA receptor activation further excites the neurons by recruiting a calcium-activated non-selective cation current (ICAN) and we hypothesize that ICAN is specifically mediated by a transient receptor potential (TRP) channel. We used RT-PCR methods to confirm expression of TRPM2 and TRPM4 mRNA in substantia nigra pars compacta. We propose that ICAN is selectively activated during burst firing to boost NMDA currents and allow plateau potentials. This boost mechanism may render DA cells vulnerable to excitotoxicity.


Abstract  Nigral dopamine neurons are transiently activated by high frequency glutamatergic inputs relaying reward-predicting sensory information. The tonic firing pattern of dopamine cells responds to these inputs with a transient burst of spikes that requires NMDA receptors. Here, we show that NMDA receptor activation further excites the cell by recruiting a calcium-activated non-selective cation current (ICAN) capable of generating a plateau potential. Burst firing in vitro is eliminated after blockade of ICAN with flufenamic acid, 9-phenanthrol, or intracellular BAPTA. ICAN is likely to be mediated by a transient receptor potential (TRP) channel, and RT-PCR was used to confirm expression of TRPM2 and TRPM4 mRNA in substantia nigra pars compacta. We propose that ICAN is selectively activated during burst firing to boost NMDA currents and allow plateau potentials. This boost mechanism may render DA cells vulnerable to excitotoxicity.


calcium-activated non-selective cation




flufenamic acid


interburst interval


interspike interval


substantia nigra pars compacta


voltage-gated calcium channel


Nigrostriatal dopamine (DA) release is vital for the initiation of movements, including feeding and drinking (Zhou & Palmiter, 1995). An animal's behaviour is altered by the temporal pattern of DA release, and it has been proposed that tonic dopamine levels provide a general motivating function while phasic DA permits the corticostriatal plasticity necessary for habit learning (Reynolds & Wickens, 2000; Wise, 2004; Niv et al. 2007). Thus, burst firing may encode a ‘reward’ signal during normal reinforcement learning and also pathological addictions (Mirenowicz & Schultz, 1996; Phillips et al. 2003).

However, in contrast to our understanding of the cellular mechanisms underlying tonic firing (Surmeier, 2007), relatively little is known about the mechanisms that give rise to bursting. Intracellular recordings from the substantia nigra pars compacta (SNc) have shown that dopamine cells employ a calcium-dependent bursting mechanism both in vitro and in vivo (Grace & Bunney, 1984b). Calcium influx through NMDA receptors and/or voltage-gated calcium channels causes a depolarization that is terminated by the subsequent activation of calcium-activated potassium channels (Overton & Clark, 1997). The substantial rise of intracellular calcium during bursting led us to hypothesize that a calcium-activated non-selective cation (CAN) current may also be activated. CAN channels are involved in the bursting behaviour of cells in neocortex, amygdala, subthalamic nucleus and pre-Bötzinger complex (Pena et al. 2004; Schiller, 2004; Zhu et al. 2004; Egorov et al. 2006; Pace et al. 2007). It is generally believed that CAN channels are activated downstream of glutamate receptors to drive plateau potentials and boost burst firing.

The molecular identity of the channel underlying ICAN has been proposed to be TRPM4 or TRPM5 (Launay et al. 2002; Ullrich et al. 2005). Pharmacological agents perturb individual subtypes of the TRP channel family, but with limited selectivity (Clapham et al. 2001). Flufenamic acid (FFA) is a broad-spectrum TRP blocker used to block ICAN in a variety of neuronal systems (Pena & Ordaz, 2008). Here we demonstrate that burst firing in DA cells is reduced by FFA and other ICAN antagonists. We provide the first evidence of TRPM2 and TRPM4 mRNA expression in SNc dopamine cells.

Given the potential role of elevations in cytosolic calcium concentration in creating metabolic stress and toxin susceptibility (Surmeier et al. 2010), the opening of calcium activated non-selective cation channels and subsequent depolarization could be a factor in loss of SNc DA neurons in diseases like Parkinson's disease.


All experiments conformed to the guidelines for the care and use of animals approved by the National Institutes of Health and the Animal Care and Use Committee at the University of Chicago.

Slice preparation

Juvenile (P8–18) CD-1 mice (Charles River Laboratories, Wilmington, MA, USA) were anaesthetized with ether or isoflurane and decapitated. Coronal midbrain slices (320 μm thick) were prepared using a vibrating-blade microtome (VT1000S; Leica, Nussloch, Germany). Slices were maintained at room temperature in the recording solution until use (1–2 h).

In some experiments, transgenic mice from the C57BL/6 strain expressing either TH-GFP (gift of D. S. Sulzer) or BAC-Drd2-EGFP (gift of D. J. Surmeier) were used to aid identification of dopamine cells. We did not observe any significant differences in spontaneous or burst firing between strains and pooled our data from all mouse strains.


Slices were mounted in a chamber on the stage of an upright microscope (Zeiss Axioskop, Oberkochen, Germany) and visualized by infrared differential interference contrast video microscopy through a 40× water immersion objective. An ORCA-ER digital camera (Hamamatsu, Shizuoka, Japan) and MetaMorph software (Molecular Devices, Sunnyvale, CA, USA) were used to capture images. The substantia nigra pars compacta recording location was lateral to the prominent fibres of the oculomotor nerve (Paxinos & Franklin, 2001).

Slices were superfused continuously at a rate of 6 ml min−1 with recording solution containing (in mm): 125 NaCl, 25 NaHCO3, 4 KCl, 1.2 NaH2PO4, 2.5 CaCl2, 1.2 MgSO4 and 17 glucose, pH 7.4 when gassed with 5% CO2–95% O2. The slice temperature was maintained at 32°C (±2°C) with an in-line heater and dual-channel temperature controller (Warner Instruments, Hamden, CT, USA). The volume of the chamber was 2 ml, requiring 2 min for complete wash-in of drugs. All chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA), with the exceptions of ZD7288 and SKF96365 from Tocris Cookson (Ellisville, MO, USA). Stock solutions of drugs in DMSO, ethanol or H2O were diluted 1000-fold in recording solution. Patch pipettes (2–6 MΩ) were pulled from filamented borosilicate glass (G150F-4; Warner Instruments) with a P-97 Flaming–Brown micropipette puller (Sutter Instruments, Novato, CA, USA). Internal solution contained (in mm): 138 potassium methane sulfonate, 10 Hepes, 10 EGTA, 4 Na2ATP, 0.3 Na-GTP, 2 MgCl2, 1 CaCl2 and in some cases 0.1% rhodamine-dextran (Molecular Probes, Eugene, OR, USA), pH 7.2 (300–315 mosmol l−1). The free calcium concentration with this internal solution was calculated to be 10 nm ( Whole-cell intracellular recordings were made from individual SNc neurons using an Axopatch 1D amplifier (Axon Instruments, Union City, CA). The current signal was digitized at 5 kHz and low-pass filtered before acquisition (Bessel characteristic of 1 kHz cutoff frequency) and transferred to a PC using a Digidata interface (Axon Instruments).

Data analysis

Putative dopamine neurons were identified based on spontaneous firing frequency below 5 Hz, strong Ih rectification >30 mV when measured from −120 mV, action potential durations greater than 2 ms with prominent AHP >10 mV, and sensitivity to the D2 receptor agonist quinpirole (2 μm).

Only cells with a stable baseline activity for 5 min were analysed for tonic firing. Membrane potential (Vm) was not corrected for the liquid junction potential of −8.4 mV. Spike times during 50–100 s of recording were obtained using Clampfit 8 (Axon Instruments) and Igor Pro (WaveMetrics, Lake Oswego, OR, USA) software with a customized script for spike detection. The coefficient of variation (CV) for interspike intervals (ISIs) was calculated as CV = SD of ISIs/mean ISI. Low CV values indicated a high regularity of spiking. Data are reported as means ± SEM. Statistical analysis compared firing before and after drugs by Student's paired t test with P < 0.05 considered significant (Prism, GraphPad Software, San Diego, CA, USA).

Bursts were considered trains of action potentials with two or more spikes and interspike intervals less than 250 ms, followed by a pause longer than 500 ms. An interburst interval (IBI):ISI ratio of at least 5:1 was necessary for clear separation of bursts. The number of spikes per burst, instantaneous spike frequency, spike adaptation and burst frequency were determined based on spike times during 100 s of recording. In addition to high frequency spiking dynamics, the underlying low frequency oscillations of membrane potential were analysed after spikes were low-pass filtered using a customized MatLab routine (The MathWorks, Natick, MA, USA). Each burst waveform was fit by a piecewise linear function calculated from the 10%–50%–90% points of maximum depolarization (Olsen & Calabrese, 1996), as outlined in Fig. 3. A minimum voltage trajectory depolarization of 15 mV was required for a burst.

Figure 3.

Dual analysis of burst spike times and underlying envelope of depolarization
A, burst envelope (outlined in green) was obtained by low-pass filtering raw traces to remove spikes. Amplitude was measured from the minimum Vm (blue squares) to the peak of depolarization. Envelope duration half-max was measured between points at 50% of peak amplitude (red triangles). Rise time and rise slope were measured from 10% to 90% of peak amplitude (yellow triangles). Burst area was calculated by integrating the area under the waveform between points at 10% of peak amplitude on the rising and falling phase. B, burst envelope was fit to a piecewise linear function, with points at 0%, 10%, 50%, 90% and 100% of peak amplitude. Four consecutive bursts (purple) from a cell were superimposed, and the average waveform is shown in red. C, bursting parameters in NMDA from n= 88 DA cells (juvenile mice). Spike times and burst envelope analyses were performed separately.

The underlying envelope amplitude, duration, area, frequency, duty cycle, cycle-by-cycle variation, rise and fall time, and slope were quantified. For each cell, bursts during drug application are reported as percentage change from baseline prior to drug. Statistical significance was determined by a one-sample t test using the mean percentage change compared to theoretical mean of 0%. P < 0.05 was considered significant (GraphPad Prism).


For identification of DA neurons, slices were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.4, overnight at 4°C and cryoprotected with 30% sucrose in PBS. Frozen sections (30 μm thick) were cut with a cryostat and dissolved in PBS. Sections were treated for 1 h with blocking solution containing 10% normal donkey serum (Vector Laboratories, Burlingame, CA, USA), 0.2% BSA and 0.03% Triton X-100 for permeabilization. Mouse anti-tyrosine hydroxylase (1:1000, BD, Franklin Lakes, NJ, USA) primary antibody was added to the blocking solution and incubated overnight at 4°C. Sections were washed with PBS and incubated with secondary antibody FITC anti-mouse (1:500, Vector) for 1 h at room temperature. Sections were washed with PBS and mounted with Vectashield mounting medium containing DAPI (Vector). Staining was analysed using a fluorescence light microscope with digital camera.


Substantia nigra pars compacta (SNc) tissue was microdissected from 300 μm coronal midbrain slices obtained from P15 mice. Tissue samples from five animals were pooled and processed for total RNA using RNAzol (Molecular Research Center, Cincinnati, OH, USA), and quantified by fluorometry (Qubit, Invitrogen). First-strand cDNA synthesis was generated with Moloney murine leukaemia virus reverse transcriptase (M-MuLV Reverse Transcriptase, New England Biolabs, Ipswich, MA, USA) from ∼0.4 μg of total SNc RNA, primed with 6 μm random oligo hexamers (Random Primer 6, New England Biolabs), following conditions recommended by the manufacturer. Assuming ∼10 pg of total RNA per neuron and a conservative 1:5 ratio of neuron to glia, this represents a sampling of pooled RNA equivalent to ∼8000 neurons. Approximately 0.1% of this SNc first strand cDNA reaction served as template for each PCR. Negative control PCR reactions included reactions assembled without reverse transcriptase or without the addition of PCR primers. PCR reactions utilized a ‘hot-start’ thermostable DNA polymerase (Phire, New England Biolabs), and followed standard conditions recommended by the manufacturer (0.5 μm of each primer, 0.2 μm dNTPs, and 1.5 mm MgCl2). PCR primers for each targeted gene were selected with the assistance of DNA sequence software (Lasergene, DNAStar, Madison, WI, USA) to generate short (300–600 bp) amplicons that would span at least one large intron (>1.0 Kb) to minimize the amplification of genomic DNA with the short extension times used. PCR reactions were initiated by a ‘hot-start’ step (98°C, 30 s), followed by 70 cycles of amplification (98°C, 5 s, denaturing; 63°C, 5 s, annealing; 72°C, 10 s, extension), and a terminal extension step (72°C, 1 min), in a PTC-100 thermocycler (MJ Research, Waltham, MA, USA). PCR products were electrophoresed on 1.5% agarose gels in 0.5× Tris/Borate/EDTA (TBE) buffer, and visualized with 0.01% Gel Green (Biotium, Hayward, CA, USA) on a Molecular Imager Gel Doc XR System (Bio-Rad). The primers used and expected sizes of PCR products were:

  • mouse TRPM2 (497 bp): (forward) 5′-TGGGGTGCGGCTCAAGGAGTT-3′,

  • (reverse) 5′-CAGAGCTGTCTGTATCTTCCTCCTCCTTG-3′; mouse TRPM4 (584 bp):

  • (forward) 5′-GCTGGCACTCACCTGCTTCCTG-3′, (reverse) 5′TGTAGCGCTGTGCCTTCCAGTAGAG-3′; mouse TRPM5 (392 bp): (forward)


  • 5′-TGCATTGCCTTGCACCACCTG-3′; mouse tyrosine hydroxylase (TH), (333 bp): (forward) 5′-GCACATTTGCCCAGTTCTCCCAG-3′, (reverse)

  • 5′-GCTGGATACGAGAGGCATAGTTCCTG-3′; mouse β-actin (ATCB) (557 bp): (forward) 5′-GGATGACGATATCGCTGCGCTGGTC-3′, (reverse)



Identification of dopamine neurons in vitro

We obtained whole-cell intracellular recordings from midbrain slices of juvenile mice (P8–18). Substantia nigra pars compacta (SNc) DA neurons were identified by a slow tonic firing rate less than 5 Hz, a rebound depolarization from −120 mV indicative of a hyperpolarization-activated inward current (Ih), and a silencing response to the D2 autoreceptor agonist quinpirole (2 μm) (Fig. 1). After recordings, cells were filled with a rhodamine-dextran dye and stained with a tyrosine hydroxylase primary antibody (1:1000) and FITC secondary antibody, as shown in Fig. 1D.

Figure 1.

Identification of putative DA neurons in vitro
A, spontaneous tonic firing occurred at a frequency of 2.6 ± 0.1 Hz (n= 104 cells). B, all DA cells exhibited a prominent Ih rectification measured at −120 mV, during injection of a current pulse of −100 to −200 pA. Mean sag amplitude was 40 ± 1 mV. C, application of the D2 agonist quinpirole (2 μm, in the bath) hyperpolarized the membrane potential and silenced all firing. D, after electrophysiological characterization, a portion of DA cells were loaded with rhodamine dextran (red) and processed for immunostaining. 40× merged image: the red cell was immunopositive for tyrosine hydroxylase (TH) labelled with FITC (green).

The spontaneous spiking rate of juvenile DA cells (2.6 ± 0.1 Hz, n= 104) ranged from 1 to 5 Hz (Fig. 1A). In order to confirm the presence of Ih, the rebound sag amplitude was measured at −120 mV by injecting a hyperpolarizing current step for 2 s (Fig. 1B). All DA cells included in this study (except one cell) had sag amplitudes greater than 30 mV (mean: 40 ± 1 mV) in current-clamp.

Burst firing induced with NMDA

Burst firing in vitro can be achieved by increasing the excitatory drive and/or reducing the inhibition of SNc neurons (Johnson et al. 1992; Mereu et al. 1997; Paladini et al. 1999; Komendantov et al. 2004; Blythe et al. 2007; Lobb et al. 2010). We applied the glutamatergic agonist NMDA in the presence of the GABAA receptor antagonist picrotoxin (50 μm). The majority of juvenile DA cells responded to NMDA with burst firing (77% of n= 138), while 23% of the DA cells continued to spike in a tonic mode. In half of the bursting DA cells, hyperpolarizing holding current (−30 to −150 pA) was necessary to sustain the bursting oscillations within the range of −33 to −100 mV, as described by Johnson et al. (1992). In the other half of the bursting cells, no holding current was necessary, as shown in Fig. 2A.

Figure 2.

NMDA caused a switch from tonic to burst firing
A, NMDA (15 μm, in the bath) caused the tonic firing and stable resting potential from time point i to depolarize and switch to high-frequency burst firing with regenerative oscillations of membrane potential (mean: 0.3 Hz ± 0.02, n= 88) by time point ii. Lower trace: DC injection was slowly ramped to −100 pA in order to hyperpolarize the cell and induce stronger oscillations starting at time point ii. B, ISI histograms of firing activity for a representative DA neuron. Tonic spiking had a unimodal distribution of ISIs at 250 ms (CV = 0.15). C, NMDA caused strong bursting with a bimodal ISI distribution, with peaks at 150 ms (ISI within a burst) and 2100 ms (ISI between bursts).

The voltage trajectory of each burst waveform was fit by a piecewise linear function calculated from the 10%–50%–90% points of maximum depolarization (Fig. 3A and B). The average amplitude of the NMDA-induced oscillation was 27 mV (envelope excluding spikes) and occurred in the range of −23 to −85 mV. The mean oscillation frequency was 0.29 ± 0.02 Hz, as summarized in Fig. 3C. The switch in firing mode from tonic to bursting was associated with a dramatic shift of the interspike interval (ISI) distributions. Spontaneous tonic firing was characterized by a unimodal ISI distribution, with relatively little dispersion (CV = 0.15) as shown in Fig. 2B. During bursting, on the other hand, the ISI distribution was bimodal with peaks at 150 ms and 2100 ms for the intraburst and interburst epochs, respectively, as shown in Fig. 2C. Burst firing ISIs were relatively irregular (CV = 1.8).

Burst structure

A spectrum of bursts was evoked by NMDA. Representative traces from three juvenile DA neurons are shown in Fig. 4. Although burst pattern (e.g. burst duration, number of spikes, spike frequency) varied between SNc DA cells, 97% of cells exhibited a constant burst pattern throughout the recording. ISIs within a burst (mean: 154 ± 8 ms) were at least 5-fold shorter than the interburst inhibitory period (ISI, mean: 2 ± 0.2 s).

Figure 4.

Juvenile DA cells exhibit heterogeneous bursting patterns in NMDA
A, a spectrum of burst patterns was observed during NMDA application (20 μm, in the bath) with a constant –DC current injection to maintain Vm from −60 to −80 mV. Bursting in 46% of DA cells exhibited a ramping depolarization with progressively shorter ISIs leading into a long duration burst, as shown in Cell 1. Shorter bursts, depicted in Cells 2 and 3, occurred in 51% of juveniles and all adults tested. B, scatter plot of individual DA cells bursting in NMDA. Burst duration had a strong positive correlation with the number of spikes per burst (correlation coefficient = 0.848, n= 79). C, the rise slope of bursts was negatively correlated with number of spikes per burst (correlation coefficient =−0.546, n= 79), indicating that bursts with steep slopes had fewer spikes. D, boxplot (min–25%–median–75%–max) of Ih sag amplitude measured at −120 mV. Ramping burst (sag mean: 42.6 ± 1.0 mV, n= 21) and non-ramping burst (sag mean: 38.9 ± 1.2 mV, n= 17) cell types were significantly different.

Roughly half of all bursting DA cells exhibited a ramping depolarization leading into a burst of up to 60 spikes (mean: 35 ± 3 spikes; range: 7 to 66), as shown in Fig. 4A (Cell 1). These cells had a slow rate of rise (mean: 9 ± 1 mV s−1, n= 38). Burst durations were long (mean: 5 ± 0.5 s), and the instantaneous firing rate increased as the burst progressed.

Half of all bursting DA cells had bursts with fewer spikes (mean: 13 ± 1 spikes; range: 2–30) and shorter durations (mean: 2 ± 0.2 s) (Fig. 4, Cells 2 and 3). For these cells, the rate of rise was rapid (mean: 18 ± 2 mV s−1, n= 43). The remaining 3% of all bursting DA neurons had an irregular burst pattern alternating between ramping and non-ramping bursts.

Bursting cells were designated as either ‘ramping’ or ‘non-ramping’ based on the presence of slow spikes leading into a burst, and these groups had significantly different burst durations (P < 0.001) and spikes per burst (P < 0.001). We also found a significant difference (P < 0.05) between the Ih sag amplitudes in these two groups (Fig. 4D), with ramping cells having a larger mean sag of 42.6 ± 1.0 mV (n= 21) than non-ramping cells with mean sag of 38.9 ± 1.2 mV (n= 17). However, both ramping and non-ramping burst types exhibited sag amplitudes greater than 35 mV, which largely correspond to calbindin negative dopamine neurons as described by Neuhoff et al. (2002).

The interplay between depolarizing (Ih, ICa, INMDA) and hyperpolarizing (SK) currents putatively determines burst duration and the number of spikes per burst (Komendantov et al. 2004; Ji & Shepard, 2006). Since these currents are differentially expressed in nigral DA cells (Wolfart et al. 2001; Neuhoff et al. 2002; Wolfart & Roeper, 2002), we attribute the different burst patterns to heterogeneous intrinsic properties.

Burst firing does not require HCN channel currents

To test whether the current attributable to hyperpolarization and cyclic adenosine monophosphate-activated (HCN) current that is essential for tonic firing in juvenile DA cells (Chan et al. 2007) is also important for burst firing, we applied the HCN channel blocker ZD7288 (50 μm, in the bath) to NMDA-induced bursting cells. Blocking HCN currents enhanced bursting (n= 5 cells) (Fig. 5), increasing their amplitude and duration (P < 0.05), while burst frequency was reduced (P < 0.05). ZD7288 also caused a 7 mV hyperpolarization of membrane potential. Thus, the HCN current modulates NMDA-induced bursting but is not essential for bursting.

Figure 5.

Effects of Ih blockade on bursting of juvenile DA neurons
A, ZD7288 (50 μm, in the bath) was applied to juvenile DA cells bursting in NMDA (25 μm). After several minutes, the bursting became even more pronounced in amplitude. B, the slow rising depolarization leading up to each burst was eliminated by ZD7288. C, ZD7288 eliminated the slow ramping spikes to increase IBIs by 3-fold. Burst duration increased by 69% and envelope amplitude increased by 34% in ZD7288. Overall burst frequency was reduced by 43%. The rise slope from 0 to 10% of peak amplitude was reduced by 30%.

Burst firing does not require voltage-gated calcium channels

Spontaneous tonic activity in adult SNc DA cells is influenced by calcium through Cav1.3 L-type channels (Nedergaard et al. 1993; Chan et al. 2007; Puopolo et al. 2007). The role of these channels in bursting is not clear. In vivo, voltage-gated calcium channels do not affect bursting in adult DA cells (Grace & Bunney, 1984b).

The potential role of Cav1.3 in burst firing was tested by bath-applying the selective Cav1.3 antagonist isradipine (30 μm), a dihydropyridine known to block all L-type calcium channels in DA cells (Chan et al. 2007). However, isradipine had no significant effects on burst spiking frequency or duration (P > 0.05) in juvenile DA cells (n= 5) (Fig. 6A). The non-specific calcium channel blocker Cd2+ (200 μm, in the bath) did not eliminate bursting in juvenile DA cells (n= 7) (Fig. 6B). In fact, oscillations increased in frequency (40% increase, P < 0.05) and had shorter burst envelope durations (43% reduction, P < 0.05). These effects are likely to be attributable to block of calcium channels controlling small conductance K+ channels (Amini et al. 1999; Wolfart & Roeper, 2002; Komendantov et al. 2004; Waroux et al. 2005).

Figure 6.

Blockade of voltage-gated calcium channels did not eliminate bursting
A, bursting in NMDA (15 μm) continued undisturbed after application of isradipine (20–30 μm, in the bath). Isradipine increased the duration of each burst envelope by 15%, and reduced burst frequency by 10%. Instantaneous spikes frequency and rise slope were unaffected. B, bursting in NMDA (15 μm) also persisted after application of Cd2+ (200 μm, in the bath). In juvenile DA cells, cadmium significantly increased the rise slope of a burst by 118%. Bursts had 43% shorter envelope durations, causing burst frequency to increase by 40%. C, in adult DA cells, cadmium did not block NMDA-induced burst firing. Cadmium reduced burst frequency by 46%. Envelope duration increased by 31%, and amplitude increased by 61%. All of these changes were significant.

The expression of L-type calcium channels in SNc DA cells increases with age (Chan et al. 2007). Therefore, we examined the effects of Cd2+ in older mice (P30–60) (n= 6). Cd2+ significantly slowed burst frequency and increased burst duration (P < 0.05) in these older neurons (Fig. 6C), suggesting an elevated role in burst generation. These results argue that voltage-gated calcium channels do not play a necessary role in burst generation.

Burst firing is terminated by ICAN blockade

DA cells in which burst firing was evoked with NMDA were tested for response to a broad-spectrum TRP channel blocker, flufenamic acid (FFA). Application of FFA (50–100 μm, in the bath) completely terminated bursting, causing some cells to revert to single spikes after 5–10 min, as shown in Fig. 7A. FFA caused a 3 mV depolarization of membrane potential (Vm), and –DC injection to maintain constant Vm did not alter the firing pattern. In these experiments, a low concentration of FFA (50 μm) was initially applied to minimize non-specific effects (Crowder et al. 2007). Of the 10 DA cells tested, four reverted to tonic firing in 50 μm FFA, and the remainder required 100 μm FFA to see this effect. A concentration-dependent effect of FFA on bursting has been reported for other cell types (Pace et al. 2007). With regard to heterogeneity of burst type, FFA blocked bursting of all varieties (ramping, n= 4; and non-ramping, n= 6). The ISI distribution during bursting in NMDA (bimodal peaks at ∼200 ms and ∼2 s) was disrupted by FFA, which skewed the distribution toward longer ISIs, as shown in Fig. 7B. FFA application reduced the portion of total spikes occurring in a burst from 99.5% to 47.9% (P < 0.01), as shown in Fig. 7C. Burst frequency (baseline 0.29 ± 0.05 Hz; FFA 0.06 ± 0.02 Hz) and instantaneous spike frequency (baseline 12.3 ± 2.1 Hz; FFA 4.6 ± 0.7 Hz) were decreased by 74% (P < 0.0001) and 58% (P= 0.0001), respectively (n= 10). The amplitude of depolarization also decreased by 70% (P < 0.0001). These effects were reversible after washout of FFA.

Figure 7.

Flufenamic acid caused a switch from burst to tonic firing
A, FFA (50–100 μm, in the bath) completely disrupted bursting in NMDA and caused the DA cell to resume tonic firing. Membrane potential became more depolarized by 3 mV, and –DC to maintain constant Vm did not alter firing pattern. Bursting was restored upon washout. B, top, DA cell bursting in NMDA has a bimodal ISI distribution with a ∼200 ms peak for ISIs and ∼2 s peak for IBIs. Bottom, FFA reduced the peak at 200 ms and scattered the ISI distribution across longer ISIs. Fewer spikes (events) occurred in FFA. C, top, envelope amplitude was reduced by 70% and burst frequency by 74%, with 5 of 10 cells exhibiting no bursts at all in FFA (50–100 μm, in the bath). Only 48% of spikes occurred during a burst, and instantaneous spike frequency within a burst decreased by 58%. Bottom, 9-phenanthrol (100 μm, in the bath) caused tonic firing in 5 of 6 cells. Burst frequency was reduced by 56%, and only 36% of spikes occurred within a burst. These changes were reversible upon washout. D, FFA (100 μm) slowed tonic firing frequency by 79% and caused silence. FFA also hyperpolarized the membrane potential by 6 mV. These changes were significant and reversible in 2 of 4 cells. Control recordings (whole-cell) showed a 38% increase of firing rate and 3 mV depolarization over the same duration of time as FFA (n= 4). E, 9-phenanthrol (100 μm) applied during cell-attached patch (CAP) recordings significantly decreased firing rate by 42%. Control recordings in CAP showed a 14% increase of firing frequency over the same duration of time.

In neurons of the entorhinal cortex and ventral medulla, the CAN channel is predicted to be a melastatin-like TRP channel, TRPM4 or TRPM5 (Egorov et al. 2002; Launay et al. 2002; Petersen, 2002; Crowder et al. 2007; Pace et al. 2007; Mironov, 2008). These TRP channels are directly activated by intracellular calcium and pass sodium ions to further depolarize the cell (Ullrich et al. 2005). To test the possibility that TRPM4 or TRPM5 is involved in DA cell bursting, we applied 9-phenanthrol, a drug that selectively blocks TRPM4 but not TRPM5 (Grand et al. 2008). 9-Phenanthrol (100 μm, in the bath) completely eliminated burst firing in 5 of 6 cells tested, and caused tonic firing in a manner similar to FFA (Fig. 7C). Burst frequency was reduced by 56% (P < 0.05) and only 36% of all spikes in the recording occurred in a burst (P < 0.05). These effects were reversed upon washout of 9-phenanthrol.

Tonic firing is reduced by TRP channel blockers

The effects of ICAN blockers on spontaneous tonic firing were also measured. In unperturbed control recordings, spontaneous tonic firing frequency (mean: 2.1 ± 0.4 Hz at mean Vm−49 mV, n= 4) during whole-cell recordings with 10 mm EGTA pipette solution became 38% faster (P < 0.05) 5 min after break-in (mean: 2.9 ± 0.5 Hz at mean Vm−46 mV, n= 4). Faster firing was associated with a 3 mV depolarization of membrane potential. CV of ISIs also increased in controls from 0.06 to 0.08 (P < 0.05), indicating less regular firing. In contrast, FFA (100 μm, in the bath) dramatically reduced tonic activity after 5–8 min (Fig. 7D). Tonic frequency decreased from a baseline of 3.1 ± 0.4 Hz at −42 mV (n= 4) to 0.7 ± 0.7 Hz at −48 mV (n= 4) after 5 min of FFA (79% reduction, P < 0.05). Membrane potential became hyperpolarized by 6 mV in FFA. These changes were reversible in 2 of 4 cells. A lower concentration of FFA (50 μm) caused no significant changes of tonic frequency (baseline 2.5 ± 0.4 Hz vs. 2.4 ± 0.4 Hz in FFA, n= 3) or in the CV of ISIs (P > 0.05). These data indicate a dose-dependent effect of FFA from 50 to 100 μm.

The whole-cell configuration may influence firing rate by altering the intracellular milieu. To circumvent this problem, we obtained cell-attached patch (CAP) recordings of tonic activity. After 10 min of CAP recording there was a non-significant increase in spike frequency (14%, n= 3, P > 0.05) (baseline 2.0 ± 0.3 Hz vs. 2.3 ± 0.3 Hz at 10 min) (Fig. 7E). In contrast, 9-phenanthrol (100 μm) significantly reduced firing frequency by 42% during cell-attached recordings (n= 3, P < 0.05; baseline 2.2 ± 0.2 Hz vs. 1.3 ± 0.2 Hz at 10 min).

TRPC blockers and ER store release did not affect burst firing

The effects of FFA (100 μm) on tonic and bursting DA cells were similar to FFA studies in substantia nigra pars reticulata (SNr) neurons. SNr GABA cells exhibit calcium-dependent plateau potentials that require ICAN (Lee & Tepper, 2007). SNr cells also express constitutively active TRPC3 channels that provide the depolarization necessary for rapid tonic firing (Zhou et al. 2008). Since several of the canonical TRP channels, TRPC1, TRPC5, and TRPC6, are expressed in nigral DA cells and boost EPSCs (Tozzi et al. 2003; Bengtson et al. 2004; Martorana et al. 2006; DeMarch et al. 2006; Giampa et al. 2007), we tested if TRPC underlies the ICAN-dependent bursting of DA cells.

Application of the TRPC blocker SKF96365 (50 μm, in the bath) had no effect on bursting (n= 3). None of the burst parameters measured (i.e. duration, amplitude, instantaneous frequency, rise slope) were significantly altered by SKF96365. The trivalent cation lanthanum (La3+) was also reported to block TRPC (Clapham et al. 2001). However, bursting in NMDA persisted in the presence of La3+ (100 μm, in the bath) (n= 3). The only significant change in La3+ was an 18% decrease in the number of spikes per burst (P < 0.01). Our TRPC blockade experiments indicate that although TRPC1 channels are involved in excitatory neurotransmission in DA cells, these channels are not specifically recruited during bursting.

In order to better identify the calcium sources involved in ICAN activation, we tested whether intracellular calcium stores contribute to burst firing. The SERCA pump blocker thapsigargin (5 μm, intracellular) was added to the internal recording solution in order to empty ER stores. Thapsigargin had no significant effects on burst firing after 35 min (P > 0.05, n= 3).

Rapid calcium chelation occludes burst firing in NMDA

The heavy reliance on calcium-activated potassium channels for repolarization suggests that spiking is quite sensitive to changes of intracellular calcium. We varied EGTA concentrations in the pipette solution (0.05, 5, 10 mm) and noted a clear effect on spike frequency and burst frequency. There were no significant differences between bursting in 5 mm (n= 18) and 10 mm EGTA (n= 63), and these cells were pooled in Fig. 3C. On the other hand, 0 and 0.05 mm EGTA (n= 3) caused short duration bursts (less than 1 s) to occur spontaneously at depolarized membrane potentials (−40 mV to −20 mV) (Fig. 8A). 0 and 0.05 mm EGTA increased burst frequency by 225% (P > 0.05) as compared to control cells in 10 mm EGTA, and decreased the burst duration by 73% (P < 0.01). Rise time decreased by 74% (P < 0.01) and rise slope increased by 122% (P > 0.05), while IBIs decreased by 45% (P > 0.05) as compared to cells in 10 mm EGTA. However, the burst envelope shape did not change dramatically from that in 10 mm EGTA, with acceleration of spikes sometimes leading to depolarization block (Fig. 8A).

Figure 8.

Low calcium buffering facilitates rapid bursting oscillations in NMDA
A, lowering the EGTA concentration in the pipette to 0.05 mm allowed faster burst oscillations to occur at a more depolarized membrane potential range of −40 to −20 mV. The lower trace (magnified time scale) shows that burst envelope shape in 0.05 mm EGTA is similar to the envelope shape in 10 mm EGTA, shown in panel B. C, burst envelope duration increased more than 3-fold as [EGTA] increased from 0 to 5–10 mm, and IBIs nearly doubled. As a result, bursts appeared more drawn-out. D, the burst frequency and rise slope showed an inverse correlation with [EGTA] (correlation coefficient: −0.820 and −0.737, respectively).

NMDA receptor activation presumably increases the intracellular calcium concentration, which in turn activates ICAN. To test this possibility in DA cells, we tested the effects of a fast calcium chelator on bursting. First, NMDA-induced bursting was measured in our standard pipette solution (10 mm EGTA) after 10–30 min compared to the start of recording (n= 10) (Fig. 9A). Bursts showed minor changes in envelope duration and amplitude (P > 0.05), but a significant 39% reduction of burst frequency (P < 0.05). The slower frequency may be attributed to a minor rundown of intracellular milieu during the whole-cell configuration. Recordings with 10 mm EGTA pipettes also had a 14% reduction of instantaneous spike frequency (P < 0.01) after 10–40 min due to the increased depolarization block in each burst.

Figure 9.

BAPTA-loaded pipettes caused a switch from bursting to tonic firing
A, EGTA (10 mm) caused minor disruptions of bursting after 10–50 min. Percentage change was calculated relative to the start of recording in NMDA. Burst duration and amplitude showed changes (2% increase and 2% decrease, respectively) that were not significant. Burst frequency decreased by 39% and instantaneous spike frequency by 14%. Both changes were significant. B, BAPTA (10 mm) injected through the recording pipette decreased the burst frequency by 66% and instantaneous spike frequency by 50%. Envelope amplitude was reduced by 38% and the rise slope decreased by 53%. All changes were significant. C, representative traces from two cells in which BAPTA disrupted bursting with NMDA (Cell 1: strong burst, and Cell 2: weak burst) and led to tonic firing. D, for Cell 1 in C, the ISI distribution at start of recording in NMDA had a signature peak at 161 ms. The firing mode was primarily tonic after 30 min in BAPTA, and the ISI distribution was shifted to larger ISIs (mean: 310 ± 104 ms).

In a second experiment, we replaced the slow buffer EGTA with an equal amount of the rapid chelator BAPTA and measured NMDA-induced bursting. BAPTA (10 mm) injected through the recording pipette perturbed NMDA-induced bursting within 15 to 30 min (n= 7) (Fig. 9B). Data are shown as the percentage change from baseline before significant amounts of chelator entered the cell (Fig. 9C). Burst frequency (baseline 0.63 ± 0.22 Hz; BAPTA 0.18 ± 0.07 Hz) was reduced by 66% (P < 0.01) with BAPTA and instantaneous spike frequency (baseline 11.0 ± 3.1 Hz; BAPTA 4.1 ± 0.9 Hz) was reduced by 50% (P < 0.01). Envelope amplitude (baseline 21.0 ± 1.7 mV; BAPTA 13.0 ± 1.7 mV) was decreased by 38% (P < 0.01) and fell below our 15 mV threshold for bursting. Rise slope decreased by 53% (P < 0.05) (baseline 28 ± 8 mV s−1; BAPTA 10 ± 2 mV s−1), which contributed to longer burst durations (62% increase, P < 0.05) (baseline 1.3 ± 0.3 s; BAPTA 1.9 ± 0.4 s).

The primary effect of BAPTA was to increase the ISIs (from baseline 148 ± 20 ms to BAPTA 873 ± 420 ms, P= 0.2) until tonic firing occurred, as shown in Fig. 9D. As the spikes become more spread out, we predict the cell is unable to maintain supralinear dopamine release (Chergui et al. 1994).

BAPTA reduced tonic firing frequency and regularity

The influence of BAPTA on spontaneous tonic activity (without NMDA) was tested because previous ICAN blockers FFA and 9-phenanthrol significantly slowed tonic firing frequency. BAPTA (10 mm) dialysed through the recording pipette for 30 min significantly decreased firing frequency (from baseline 2.5 ± 0.5 Hz at −60 mV to 1.6 ± 0.3 Hz at −58 mV, n= 5, P < 0.05). BAPTA also increased the CV of ISIs from 0.12 to 0.41 (P < 0.01) indicating less regularity. Control recordings (10 mm EGTA) showed a 51% increase of firing (from baseline 1.8 ± 0.4 Hz at −60 mV to 2.5 ± 0.5 Hz at −53 mV, n= 4, P > 0.05) after 30 min. CV changed from 0.08 to 0.22 (P > 0.05) in controls. Faster firing in controls (10 mm EGTA) may be attributed to a reduced SK current (Grace & Bunney, 1984a; Nedergaard, 2004).

TRPM2 and TRPM4 are expressed in SNc

We used the reverse transcriptase polymerase chain reaction (RT-PCR) to determine if calcium-activated TRPM channel subtypes are expressed in SNc neurons. SNc tissue was microdissected bilaterally from midbrain slices and pooled for RNA extraction. Total RNA was extracted from cells and found to contain abundant tyrosine hydroxylase mRNA (Fig. 10), confirming a dopamine neuron population. SNc tissue tested positive for TRPM2 and TRPM4, but not TRPM5. The presence of β-actin served as a positive control, and the omission of reverse transcriptase (Fig. 10, right lane for each reaction) or the omission of primers were negative controls with absent bands.

Figure 10.

TRPM2 and TRPM4 mRNA was expressed in SNc dopamine tissue
Microdissected SNc tissue showed positive mRNA expression for TRPM2 (497 bp), TRPM4 (584 bp), and TH (333 bp), but not for TRPM5 (392 bp). A positive control was β-actin (557 bp) mRNA expression. For negative controls, reverse transcriptase was omitted from each reaction (right lanes). Expected product sizes are in parentheses.


Here we demonstrate that burst generation in dopaminergic neurons depends on a calcium-dependent boosting mechanism. The calcium source for this mechanism derives from NMDA receptor activation. Burst generation is abolished by blockers of the CAN current as well as chelators of intracellular calcium, indicating that the NMDA mechanism alone is insufficient to generate bursting in dopamine neurons. The mechanisms that underlie bursting are strikingly different from those known to be critical for the generation of tonic activity. While the activation of voltage-dependent calcium channels and the HCN current is critical for generating slow tonic activity, these channels are not essential for activity patterns involving fast spiking (>10 Hz spiking) such as bursting.

Our study was conducted in acute slices of mouse brain. The NMDA-induced bursting in this in vitro model approximates some of the spiking features occurring in awake rats. Freely moving rats exploring novel environments or lever-pressing for a food reward typically exhibit bursts with 20–30 Hz spiking (Hyland et al. 2002). The mean instantaneous spike frequency of bursts in our slices was 10 Hz (range: 3–30 Hz), which was within an order of magnitude of in vivo frequencies. Furthermore, our slices bathed in NMDA had an average of 21 spikes per burst (range: 3–66 spikes), which compares to the 4–22 spikes per burst recorded in vivo during food-seeking (Hyland et al. 2002). Larger rewards or drugs of abuse may elicit even more than 22 spikes per burst in vivo. The duration (3.0 ± 0.3 s) and frequency (0.29 ± 0.02 Hz) of bursts that we recorded intracellularly from nigral DA cells closely match the timing of bursts recorded extracellularly in awake rats (Freeman et al. 1985; Hyland et al. 2002). This finding suggests that although synaptic inputs initiate a burst, the cell's intrinsic properties may set the burst duration (Kitai et al. 1999; Wolfart et al. 2001; Neuhoff et al. 2002; Wolfart & Roeper, 2002).

NMDA-induced bursting in our slices often exhibited spike frequency acceleration within a burst. Although this phenomenon was not reported in anaesthetized preparations (Grace & Bunney, 1984b), spike frequency acceleration was often seen in bursts from awake rats during reward-learning tasks (Hyland et al. 2002). At peak depolarization, some neurons exhibited sodium channel inactivation, which has also been seen in bursts evoked by electrical stimulation or iontophoretic application of glutamate (Blythe et al. 2007, 2009; Deister et al. 2009). Interestingly, subthalamic nucleus neurons exhibit an identical spike acceleration that has been attributed to progressive activation of ICAN (Beurrier et al. 1999). The waveforms we observed experimentally are consistent with computational models of bursting in which increased NMDA-induced current accelerated the fast component of the plateau (Komendantov & Canavier, 2002; Komendantov et al. 2004; Kuznetsov et al. 2006; Putzier et al. 2009; Lobb et al. 2010). This burst pattern has been observed in juvenile ventral tegmental area (VTA) cells during application of the cholinergic agonist carbachol (20 μm) (Zhang et al. 2005). The mGluR1 agonist (S)-3,5-dihydroxyphenylglycine (DHPG) can induce similar long-duration bursts in adult SNc DA cells (Prisco et al. 2002).

The instantaneous spike frequencies (10 ± 0.5 Hz) during bursting in our slices were slower than those reported in vivo, and the interspike intervals within a burst (153 ± 8 ms) were longer (Grace & Bunney, 1984b). Long ISIs may result, in part, from the clamping of intracellular calcium concentration (approaching 10 nm) by our intracellular recording solution, which contained 10 mm EGTA, 1 mm CaCl2 and 2 mm MgCl2. This recording solution is typically used in studies of burst firing (Pena et al. 2004; Zhang et al. 2005; Zhu et al. 2005). We showed that lowering EGTA to 0.05 mm caused an increase in burst frequency (Fig. 8). Less calcium buffering may facilitate bursting oscillations by allowing more rapid activation of calcium-activated potassium channels. Thus, the cell repolarizes more quickly to resume another burst cycle.

In vivo bursts occur transiently during baseline tonic activity. An IBI:ISI ratio of at least 3:1 has been used to define a burst both in vivo and in vitro (Grace & Bunney, 1984b; Johnson et al. 1992; Johnson & Wu, 2004). The larger this ratio becomes, the more discrete the burst separation, and the less tonic activity occurring between bursts. The IBI:ISI ratio in our slices (13:1) was well above the 3:1 threshold, indicating that discrete bursts occur more easily in the absence of synaptic inputs and neuromodulators. Glutamatergic inputs transiently activate DA cells in vivo, and we show that the continuous bath-application of glutamatergic agonist (NMDA) causes regenerative bursting oscillations with a constant cycle period. In a minority of DA cells (n= 3 of 105 bursters), the period exhibited more cycle-to-cycle variability.

A primary concern with the disruption of NMDA-induced bursting by FFA is that this drug may affect the NMDA receptor itself (Wang et al. 2006). The low concentrations of FFA (50–100 μm) used in our experiments have not been reported to alter NMDA currents or sodium channels (Lerma & Martin del Rio, 1992; Di Prisco et al. 2000; Wang et al. 2006; Yau et al. 2010).

Our data suggest that the HCN current plays an important role in modulating bursting. However, in contrast to the generation of tonic firing in juvenile DA neurons, Ih is not essential for burst generation. In fact, blockade of the HCN current with ZD7288 enhanced bursting. It is known that Ih actively shunts whole-cell currents during spontaneous activity (Magee, 1998; Ibanez-Sandoval et al. 2007). Therefore, we hypothesize that Ih normally limits NMDA-induced bursting in two distinct manners. The first effect of Ih-mediated shunt is to reduce burst amplitude, and therefore a decrease in calcium influx. Second, Ih alters the timing of bursts by slowing depolarization onset, and causing tonic spikes at the start of each burst.

We found some correlation between burst pattern (ramping vs. non-ramping) in NMDA and Ih sag. Previous studies have shown that Ih sag amplitude correlates well with calbindin expression in nigral dopamine cells, with over 80% of SNc being calbindin-negative (Neuhoff et al. 2002; Brown et al. 2009). However, the difference we noted in Ih sag amplitude between ramping and non-ramping cells (42.6 ± 1.0 mV vs. 38.9 ± 1.2 mV, respectively) was weak compared to the results obtained by Neuhoff et al. where the difference in the Ih sag amplitude is much more pronounced between the calbindin negative and positive cells (37.3 ± 0.7 mV vs. 25.3 ± 2.2 mV, respectively). Moreover, the proportion of cells expressing or not expressing calbindin in Neuhoff et al. (17 and 83%, respectively) does not fit with the proportion of ramping vs. non-ramping cells in this study (50% of ramping and 50% of non-ramping cells). Thus, we believe that a majority of bursting cells recorded here corresponded to calbindin negative cells, although direct confirmation by molecular methods will be necessary to further characterize calbindin expression with regard to bursting in the substantia nigra.

We noted an age-dependent difference in burst firing pattern, namely the slow ramping depolarization of a burst was observed only in juvenile DA cells (P8–18) (Figs 5 and 6). In contrast, adult DA cells had a steeper rise slope and shorter burst durations. This difference may reflect the developmental transition from Ih-dependent tonic firing to a calcium-dependent tonic mechanism in adults (Chan et al. 2007; Guzman et al. 2009).

Indeed, we showed that cadmium caused a significant reduction of burst frequency in adult mice. Our unpublished data also indicated burst parameter changes between juveniles (P8–18) and adults (P45–90). We found a gradual reduction of burst amplitude (by 55%), duration (by 53%), and spikes per burst (by 35%) as the mice matured, with a corresponding increase of burst frequency (by 86%) (the authors’ unpublished data). These data suggest that the cellular mechanisms underlying the burst firing in adult and juvenile might be different, with a greater reliance on calcium channels in adults.

In summary, we have altered several calcium buffering parameters to study bursting in SNc DA cells. First, we tested several EGTA concentrations (0.05, 5, 10 mm) and found an inverse correlation with burst frequency (correlation coefficient =−0.820) (Fig. 8D). The impact of EGTA on nearly all burst parameters suggests that one or more calcium-activated currents underlie bursting. Second, we altered the calcium buffering speed by comparing equimolar EGTA and BAPTA. The length constant for BAPTA (28 nm) is much shorter than for EGTA (419 nm), which represents the distance a calcium ion will diffuse before it is captured by the buffer (Naraghi & Neher, 1997). The ability of NMDA-induced bursting to proceed in 10 mm EGTA, but not 10 mm BAPTA, suggests that rapid local changes in calcium (<100 nm) are necessary for bursting (Borst & Sakmann, 1996). One intriguing mechanism for rapid calcium dynamics during burst firing is the dendritic colocalization of NMDARs and CAN channels.

Our BAPTA experiment showed a dramatic increase of ISIs in each burst with rapid calcium chelation. As a result, the physiological role of closely spaced spikes or ‘bursts’ to enhance phasic dopamine release into striatum is eliminated in the presence of BAPTA. The elimination of bursting with BAPTA in our study contradicts an earlier finding by Johnson et al. (1992), which reported that NMDA-induced bursting is not disrupted by intracellular injection of BAPTA (10 mm) through a patch pipette. The time course they tested was shorter than 30 min. One explanation for their finding is that BAPTA applied for short durations can reduce somatic calcium concentrations near the pipette, but does not sufficiently buffer dendritic calcium levels that may be spatially localized to microdomains such as spines.

NMDA receptor subunit composition changes dramatically during the first two postnatal weeks (Mereu et al. 1997; Brothwell et al. 2008). Our experiments were done in P8–18 mice, and no correlation between age and burst intensity (e.g. number of spikes, amplitude, or duration) was observed during this developmental time period. This suggests that the heterogeneity we observed in the nigral DA cell population is not attributable to NMDA receptor expression.

Here we provide the first evidence of TRPM channel mRNA expression in SNc neurons. The presence of TRPM2 is relevant to dopamine cell toxicity because this calcium-activated channel is a sensitive indicator of reactive oxygen species and mitochondrial death (Perraud et al. 2001; Hara et al. 2002; Kolisek et al. 2005). In addition, TRPM4 channels are also found in calcium-dependent bursting neurons of the pre-Bötzinger complex (Crowder et al. 2007; Guinamard et al. 2010). A conserved mechanism may exist by which glutamatergic activation of NMDA receptors triggers TRPM4 channel opening and plateau potentials. Similar to other systems such as the pre-Bötzinger complex, CAN currents may be heterogeneously expressed in nigral DA cells. Further studies using single-cell RT-PCR will be necessary to determine what percentage of nigral cells contain TRPM2 and TRPM4 and which ion channels or intracellular factors (like calbindin) are co-expressed along with TRPM.

TRPC channels do not appear to mediate NMDA-induced burst firing in nigral DA cells, as bursting was not perturbed by SKF96365 or lanthanum. The burst mechanism differs markedly from tonic firing of nigral DA cells, which is silenced by TRPC blockade (Kim et al. 2007). Our pharmacology data found bursting was sensitive to 9-phenanthrol, a selective blocker of TRPM4. We confirmed the expression of TRPM4 mRNA in SNc, and found no expression of the closely related channel TRPM5. Thus, TRPM4 or TRPM2 may be the CAN channels that boost bursting in NMDA.

If ICAN is activated primarily by intracellular calcium, then ICAN should trigger bursting with any manipulation that raises intracellular calcium. It follows that ICAN may be activated without NMDA receptors. Zhang et al. (2005) have recently shown that carbachol activates burst firing with a spike pattern that closely resembles our NMDA-induced bursts. Prisco et al. (2002) showed DHPG also induced bursting. We postulate that metabotropic receptors such as the mAChR and mGluR1 can be activated by muscarine or tACPD, respectively, to recruit ICAN (Yamashita & Isa, 2003, 2004). Further experiments will be necessary to investigate whether these bursts are FFA-sensitive and whether FFA has also direct effects on NMDA currents in these substantia nigra neurons. Furthermore, bursts can be induced in vivo with the neuropeptide cholecystokinin (CCK), which activates a metabotropic receptor (CCK1) (Freeman & Bunney, 1987; Wu & Wang, 1994a,b). This type of bursting may also result from internal calcium release, and the subsequent activation of ICAN.

The proposed burst mechanism for DA cells relies on NMDA receptors, or their effectors, for calcium influx (Zweifel et al. 2009). A large electrochemical gradient for calcium drives calcium ions through the NMDA receptor in SNc DA neurons (Mercuri et al. 1992). This mechanism of ICAN activation differs markedly from that of other cell types that rely instead on voltage-gated calcium channels (VGCCs) as their main calcium source. For instance, the bursting of DA cells was not blocked with cadmium, unlike the cadmium-sensitive bursters of the neocortex and pre-Bötzinger complex (Pena et al. 2004; Schiller, 2004). We hypothesize that the burst mechanism described in this study will exaggerate the known sensitivity of DA cells to excitotoxicity. This would be especially important for adult animals that express higher levels of Cav1.3. We extend the proposed model for tonic activity to include physiologically relevant firing patterns similar to those observed during reward learning (Mirenowicz & Schultz, 1996). Here, we suggest that ICAN is a novel player in the classic burst model (Overton & Clark, 1997; Kitai et al. 1999). The sodium-mediated CAN current that is selectively activated during burst firing may depolarize the cell and further promote calcium excitotoxicity in DA cells.


Author contributions

A.M. and J.M.R. designed the experiments and wrote the manuscript. A.M. performed all electrophysiological recordings, immunostaining, and data analysis. A.W. designed and executed the RT-PCR experiments, and provided editorial suggestions. All authors approved the final version of the manuscript for publication. Work was performed in laboratories at The University of Chicago, Chicago, IL, USA and Seattle Children's Research Institute and the University of Washington, Seattle, WA, USA.


We thank James Surmeier and several anonymous reviewers at The Journal of Physiology for thoughtful and thorough comments on the manuscript. We thank Andrew Hill for designing a customized Matlab script for burst analysis. Funding was provided by a Falk Foundation gift grant to J.M.R. and NINDS grant F31 NS062524 to A.M.