An Ultra-Short Dopamine Pathway Regulates Basal Ganglia Output

Substantia nigra pars reticulata (SNr) is a key basal ganglia output nucleus critical for movement control. Its GABA-containing projection neurons intermingle with nigral dopamine (DA) neuron dendrites. Here we show that SNr GABA neurons coexpress dopamine D1 and D5 receptor mRNAs and also mRNA for TRPC3 channels. Dopamine induced an inward current in these neurons and increased their firing frequency. These effects were mimicked by D1-like agonists, blocked by a D1-like antagonist. D1-like receptor blockade reduced SNr GABA neuron firing frequency and increased their firing irregularity. These D1-like effects were absent in D1 or D5 receptor knock-out mice and inhibited by intracellularly applied D1 or D5 receptor antibody. These D1-like effects were also inhibited when the tonically active TRPC3 channels were inhibited by intracellularly applied TRPC3 channel antibody. Furthermore, stimulation of DA neurons induced a direct inward current in SNr GABA neurons that was sensitive to D1-like blockade. Manipulation of DA neuron activity and DA release and inhibition of dopamine reuptake affected SNr GABA neuron activity in a D1-like receptor-dependent manner. Together, our findings indicate that dendritically released dopamine tonically excites SNr GABA neurons via D1–D5 receptor coactivation that enhances constitutively active TRPC3 channels, forming an ultra-short substantia nigra pars compacta → SNr dopamine pathway that regulates the firing intensity and pattern of these basal ganglia output neurons.


Introduction
The substantia nigra is an important component of the basal ganglia (DeLong and . It comprises two distinct parts: pars compacta (SNc) and pars reticulata (SNr). The major cell type in SNc is the dopamine (DA) neuron whose axon projects mainly to the striatum, forming the long-distance nigrostriatal DA pathway (Parent et al., 2000;Haber, 2003). SNr is populated primarily by a majority of GABA projection neurons and a minority of DA projection neurons (Fallon and Loughlin, 1995;Nelson et al., 1996;Parent et al., 2000). The GABA projection neurons fire spontaneous, mostly regular, high-frequency spikes (Hikosaka and Wurtz, 1983;Schultz, 1986).
SNr is a key output nucleus of the basal ganglia and inhibits its targets such as brainstem motor nuclei and thalamus (Hikosaka et al., 2000;McHaffie et al., 2005). It is critical to motor control, eye, and orofacial movements in particular. SNr neuronal activity is known to be influenced by the striatal and pallidal inputs that are themselves regulated by SNc-originated DA projection, thus forming the nigro-striato-nigral loop (Haber, 2003). Loss of SNc DA neurons in animal models of Parkinson's disease (PD) and human PD patients often alters the firing frequency and/or pattern of the SNr neurons (Wichmann et al., 1999;Obeso et al., 2000;Nevet et al., 2004).
SNc DA neurons extend their dendrites deep into SNr (see Fig.  1 A) (Björklund and Lindvall, 1975;González-Hernández and Rodríguez, 2000), in which tyrosine hydroxylase (TH)-positive DA dendrites and parvalbumin (PV)-positive GABA neurons intermingle closely. Such a proximity between DA dendrites and SNr GABA neurons may have important functional consequences because dendritically released DA may diffuse for a considerable distance as a result of the relatively low density of dopamine transporter (DAT) in SNr (Cragg et al., 2001).
These anatomical and physiological facts led to our hypothesis that dendritically released DA may directly activate DA receptors on SNr GABA output neurons, forming an ultra-short SNc 3 SNr DA pathway that provides a fast DA control over this key basal ganglia output nucleus. Because DA receptors do not di-rectly gate any ion channel, we further hypothesized that DA receptor activation may couple to and enhance TRPC3 channels, a type of cation channel that have been shown recently to be tonically active and regulate SNr GABA neuron activity (Zhou et al., 2008).

Materials and Methods
Mice. Wild-type 2-to 4-week-old male and female C57BL/6J mice were purchased from The Jackson Laboratory. D 1 knock-out (KO) mice and D 5 KO mice were generated, maintained, and genotyped according to established procedures (Xu et al., 1994;Hollon et al., 2002). All mice were housed in the animal facility on the campus of University of Tennessee Health Science Center (Memphis, TN).
Patch-clamp recording in midbrain slices. Coronal midbrain slices containing SNc and SNr were prepared according to well established procedures (Atherton and Bevan, 2005;Zhou et al., 2006Zhou et al., , 2008. The cutting solution contained the following (in mM): 220 sucrose, 2.5 KCl, 1.25 NaH 2 PO 4 , 25 NaHCO 3 , 0.5 CaCl 2 , 7 MgCl 2 , and 10 D-glucose. The slices were then transferred to a holding chamber containing the bathing solution (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH 2 PO 4 , 25 NaHCO 3 , 2.5 CaCl 2 , 1.3 MgCl 2 , and 10 D-glucose. This bathing solution was also the normal perfusing solution. Drugs were applied via the bathing solution. The perfusion rate was 2 ml/min controlled by a peristaltic pump. Patch electrodes had resistances of 2-3 M⍀ when filled with intracellular solution (in mM): 135 KCl, 0.5 EGTA, 10 HEPES, 2 Mg-ATP, and 0.2 Na 3 GTP. pH was adjusted to 7.3 with KOH. In a few cells, 125 mM K-gluconate replaced KCl on an equal molar basis. For perforated patchclamp recording, 50 g/ml gramicidin was added to the KCl-based intracellular solution that led to perforated whole-cell mode 10 -15 min after forming tight seal.
Recordings were made at 30°C under visual guidance of a video microscope (Olympus BX51W1) equipped with Nomarski optics and 60ϫ water-immersion lens. We targeted large cells (likely to be GABA projection neurons) for recording and selected cells with action potentials overshooting at least ϩ20 mV for analysis. Signals were recorded with Multiclamp 700B patch-clamp amplifiers, low-pass filtered at 5 kHz by the built-in Bessel filter in the Multiclamp 700B amplifiers, digitized at 5-20 kHz, and analyzed with pClamp 9 program (Molecular Devices). For display and for reducing file size for the computer to handle, data (except action potentials) were digitally refiltered at 1 or 2 kHz and resampled at lower sampling rates. Recordings with access resistance increase of Ն15% were rejected.
The summary data in Figures 2 B and 4 B were from two groups of selected neurons with stable baseline firing, good response to dopaminergic drugs, good recovery, and also matching temporal patterns in their response and recovery. A total of 21 neurons with significantly no-matching temporal patterns in their response and recovery were excluded to prevent the smearing-out of the response and recovery. The excluded cells had a response magnitude similar to that of included cells such that this exclusion did not affect our data or conclusion at all. Also, the recording of the initial several minutes was often not stable and thus was not included in analysis. The starting point of stable recording was defined as time ϭ 0 min.
Identification of substantia nigra pars reticulata GABA and DA neurons. Reliable and unambiguous identification of SNr GABA and DA neurons by their distinct electrophysiological characteristics has been described in detail in previous studies (Atherton and Bevan, 2005;Zhou et al., 2006;Lee and Tepper, 2007;Zhou et al., 2008).
Double immunostaining of GABA neurons and DA neurons. Under lethal anesthesia, mice were intracardially perfused with the ice-cold cutting solution described above and then 4% paraformaldehyde in 0.1 M phosphate buffer (PB). The removed brains were postfixed in the same fixative for 4 h at 4°C and then infused with 30% sucrose in 0.1 M PB, pH 7.4, at 4°C over two nights. Coronal midbrain sections (50 m in thickness) were cut on a Vibratome 1000 Plus. The sections then were processed for immunofluorescence detection of PV [a calcium-binding protein localized in GABA neurons but not in DA neurons in the SNr (González-Hernández and Rodríguez, 2000)] and TH (the rate limiting enzyme in DA biosynthesis). The free-floating sections were incubated with 2% fat-free milk, 1% bovine serum albumin (BSA), and 0.8% Triton X-100 in 0.01 M PBS for 1 h at room temperature to block nonspecific binding and permeate the cell membrane. After thorough rinsing, freefloating sections were incubated for 48 h at 4°C with the two primary antibodies (see below) and then rinsed in 0.01 M PBS three times for 10 min, followed by incubating with the two secondary antibodies (see below) for 2 h at room temperature in the dark. Both the primary and secondary antigen-antibody reactions occurred in 0.01 M PBS containing 3% normal donkey serum, 1% BSA, and 0.8% Triton X-100. The two primary antibodies were a polyclonal TH antibody raised in sheep (diluted at 1:1000; Millipore Corporation) and a monoclonal PV antibody raised in mouse (diluted at 1:2000; Sigma). The two secondary antibodies were as follows: (1) donkey anti-sheep IgG antibody, conjugated with red Alexa Fluor 594 (diluted at 1:200; Invitrogen), used for labeling TH, and (2) donkey anti-mouse IgG antibody, conjugated with green fluorescent Alexa Fluor 488 (diluted at 1:200; Invitrogen), used for labeling PV. Detection of intracellularly infused D 1 antibody was achieved by using an Alexa Fluor 594-tagged secondary antibody. An identical strategy was used in the study by Zhou et al. (2008).
Single-cell reverse transcription PCR. Our single-cell reverse transcription (scRT)-PCR procedures were described in detail by Zhou et al. (2008) and followed the principles and general methods of Surmeier et al. (1996). Patch pipettes were autoclaved to inactivate RNase. The intracellular solution (in mM: 135 KCl, 0.5 EGTA, 10 HEPES, and 1.5 MgCl 2 , pH adjusted to 7.3) was prepared with RNase-and DNase-free water. Brief (Յ5 min) recording was done to provide electrophysiological fingerprints for SNr GABA and DA neurons. Once a stable and reliable recording was obtained, gentle suction was applied to aspirate the cytoplasm.
The aspirated cell content was expelled into a 0.2 ml PCR tube, and potential genomic DNA was digested by DNase I (5 min at room temperature). Complete removal of genomic DNA was verified by RT-minus control in which no expected amplicon was detected when the reverse transcriptase was omitted, whereas all other reaction components were exactly the same. cDNA was synthesized using SuperScript III reverse transcriptase-based Cells-Direct cDNA Synthesis kit containing RNase inhibitor (Invitrogen). RT was performed at 50°C for 50 min, yielding 30 -35 l of cDNA solution. The reaction was terminated by heating to 85°C for 5 min and then chilling on ice. The 30 -35 l of cDNA solution was used either immediately for PCR amplification or stored at Ϫ80°C for later use. All steps ware performed according to the instructions of the manufacturer.
The cDNAs were amplified using a hot start, high specificity, and high yield Platinum PCR SuperMix (Invitrogen). To increase the detection probability of D 1 and D 5 mRNAs in a single neuron, the two-stage, nested PCR procedure was used, as detailed by Zhou et al. (2008).
To avoid potential interference among the primers, singleplex scRT-PCR was used in the initial screening ( Fig. 1 B). In later experiments, to confirm the coexpression of multiple mRNAs in a single cell, the limited amount (30 -35 l) of cDNA from a single neuron was divided, and separate PCR runs were performed to detect D 1 , D 5 , TRPC3, and glutamic acid decarboxylase 1 (GAD1) mRNAs in the same SNr GABA neurons (Fig. 1C).
PCR products from the second-stage PCR amplification were separated by 2% agarose gel electrophoresis and visualized by ethidium bromide under ultraviolet light. Then the bands on the gel were cut out, and the DNA fragments were extracted and purified with QiaQuick extraction kit (Qiagen) and sequenced. The obtained sequences were positively identified at http://blast.ncbi.nlm.nih.gov/Blast.cgi.
Primers were designed according to the sequences in GenBank (supplemental Table S1, available at www.jneurosci.org as supplemental material) and were accomplished using the web-based Primer3 software (Massachusetts Institute of Technology, Cambridge, MA). Whenever possible, intron-spanning primer pairs were used that help detect genomic DNA contamination. In cases in which negative scRT-PCR results were obtained, the effectiveness of the primer pairs was verified by whole-brain mRNAs that yielded positive products (supplemental Fig.  S1, available at www.jneurosci.org as supplemental material). All primers were synthesized by Integrated DNA Technologies.
To rule out contamination of extracellular debris that may contain mRNAs, the patch pipette was lowered into the tissue without actually aspirating cytoplasm, and then the pipette content was subjected to RT-PCR, yielding no product at all.
Statistics. Descriptive statistics such as mean and SE were used. Paired t test and Kolmogorov-Smirnov (K-S) test were also used to compare measurements in the same groups/cells when applicable. ANOVA was used for comparing the measurements obtained in two groups of mice. Linear regression analysis was used to analyze the linearity of the I-V curves and determine the reversal potential. p Ͻ 0.05 was considered to be significant.

SNr GABA neurons express D 1 and D 5 receptor mRNAs
The first question we addressed was: do SNr GABA neurons express any DA receptors? We chose scRT-PCR to answer this question because of its high sensitivity and specificity (Surmeier et al., 1996). SNr GABA neurons in brain slices were identified by their characteristic fast spiking properties (Atherton and Bevan, 2005;Zhou et al., 2008). Then the cell content was aspirated into the recording patch pipette and expelled into a PCR tube. Reverse transcription was performed, followed by two-stage nested PCR (Zhou et al., 2008). Singleplex PCR was used to avoid potential interference among primers. We also used detection of GAD1 (a key GABA synthesis enzyme) mRNA as an indicator of successful cytoplasmic aspiration and reverse transcription. If GAD1 mRNA was not detected, indicating failure in cytoplasm aspiration and/or reverse transcription, no additional detection of DA receptor mRNAs was pursued. Under these conditions, D 1 and D 5 mRNAs were detected in every GAD1 mRNA-positive SNr GABA neuron (n ϭ 16) ( Fig. 1 B), demonstrating that most, if not all, SNr GABA neurons express D 1 and D 5 mRNAs ( Fig. 1 B). In contrast, D 2 , D 3 , and D 4 receptor mRNAs were not detected (for positive control for primer effectiveness, supplemental Fig. S1, available at www. jneurosci.org as supplemental material). The mRNA for the key enzyme for DA synthesis, TH, was not detected in any of these electrophysiologically identified SNr GABA neurons (data not shown). Furthermore, after dividing the limited amount (30 -35 l) of cDNA from a single SNr GABA neuron, separate PCR runs detected D 1 , D 5 , TRPC3, and GAD1 mRNAs in the same single SNr GABA neurons (n ϭ 5) (Fig. 1C). These results indicate a coexpression of D 1 and D 5 receptors in SNr GABA neurons, pro-viding a molecular substrate for dendritically released DA to directly influence these basal ganglia output neurons and also for a potential interaction between DA signaling and TRPC3 channel activity.

Dopamine directly excites SNr GABA projection neuron firing
If functional D 1 and D 5 receptors are indeed expressed in SNr GABA neurons, then DA should induce D 1 -like excitation as seen in other brain areas (Aosaki et al., 1998;Centonze et al., 2003). To detect potential direct D 1 -like excitation in SNr GABA neurons, Figure 1. The SNc DA dendrites intermingle with SNr GABA neurons expressing D 1 and D 5 receptor and TRPC3 mRNAs. A, Confocal images of double-immunohistochemical staining for TH (red; a key enzyme for DA synthesis) and PV (green; expressed only in GABA neurons). Most TH-positive DA neurons are in SNc with their dendrites extending into SNr, in which most neurons are the PV-positive GABA neurons. Scale bar, 100 m. B, To ensure maximal detection, singleplex scRT-PCR was used to detect GAD1 mRNA in electrophysiologically identified individual SNr GABA neurons (B1). Positive GAD1 mRNA detection indicated successful cell content aspiration and RT, providing the necessary condition for subsequent detection of DA receptor mRNAs from the same cell sample (B2). All amplicons were of expected sizes. No TH mRNA amplicon of 197 bp was seen in GABA neurons, but it was detected in electrophysiologically identified DA neurons. D 5 /RT-, RT-minus negative genomic DNA control for the intronless D 5 gene. D 2 , D 3 , and D 4 receptor mRNAs were not detected (for positive control for their primers, see supplemental Fig. S1, available at www.jneurosci.org as supplemental material). C, After dividing the limited amount (30 -35 l) of cDNA from a single neuron, separate PCR runs detected D 1 , D 5 , TRPC3, and GAD1 mRNAs in the same single SNr GABA neuron, confirming the coexpression of D 1 , D 5 , GAD1, and TRPC3 mRNAs in these neurons.
we first blocked glutamate-and GABA-mediated fast synaptic transmission with 20 M D-AP-5, 10 M CNQX, and 100 M picrotoxin to remove complications of synaptic activity. Under this condition, bath application of 10 and 30 M DA reliably increased the firing rate of SNr GABA neurons by 16.2 Ϯ 1.9 and 31.0 Ϯ 2.1% from a baseline frequency of 10.7 Ϯ 0.9 Hz, respectively (measured over the 2-min peak enhancement period; p Ͻ 0.01; n ϭ 14) ( Fig. 2 A, B).
The question now was: how did DA induce this excitation? We found that DA did not affect action potential shape and, after adjusting membrane potential to the pre-DA level with hyperpolarizing current, it also did not alter the fast afterhyperpolarization (AHP) nor medium AHP. These results indicate that DA enhanced SNr GABA neuron firing without detectable effect on voltage-gated Na ϩ and K ϩ channels or Ca 2ϩ -activated K ϩ channels responsible for action potential generation and repolarization and the AHPs. Therefore, other mechanisms must be involved. We hypothesized that the primary mechanism of DA excitation could be a DA-induced direct depolarization in SNr GABA neurons.
To characterize the DA-induced depolarization, action potentials were blocked with 1 M tetrodotoxin (TTX) to obtain stable membrane potential in SNr GABA neurons (Fig. 2C). Under this condition, bath application of 10 and 30 M DA induced a clear depolarization of 1.8 Ϯ 0.1 and 2.7 Ϯ 0.2 mV from a baseline membrane potential of Ϫ49.3 Ϯ 1.3 mV, respectively (n ϭ 9) (Fig. 2C). These DA-induced direct depolarization amplitudes are similar to those in striatal cholinergic interneurons (Centonze et al., 2003) and medium spiny neurons (Yasumoto et al., 2002). Because SNr GABA neurons have a depolarized membrane potential close to the spike threshold, even a modest depolarization will affect neuronal firing and circuitry operation, conferring physiological significance onto this direct DA effect.
To further test our hypothesis, SNr GABA neurons were voltage clamped at Ϫ70 mV. At this holding potential, bath application of 10 and 30 M DA induced an inward current of 17.4 Ϯ 1.7 and 31.3 Ϯ 2.9 pA from a baseline membrane current of Ϫ160.1 Ϯ 16.5 pA, respectively (n ϭ 9) (Fig. 2 D) [much of the baseline holding current was attributable to the tonic TRPC3 current (Zhou et al., 2008)]. Hill equation fitting indicated an EC 50 of 13.3 Ϯ 0.7 M with a Hill slope of 1.2 Ϯ 0.1 (Fig. 2 E). Whole-cell conductance, monitored with Ϫ10 mV, 100 ms voltage pulses, was increased from 5.28 Ϯ 0.41 nS under control conditions to 6.55 Ϯ 0.66 nS during 30 M DA application (n ϭ 8; p Ͻ 0.05), suggesting an opening or enhancement of ion channels that mediate an inward current at Ϫ70 mV. All these DA-evoked effects were prevented by 5 M D 1 -like antagonist SKF83566 (8-bromo-2,3,4,5-tetrahydro-3-methyl-5-phenyl-1 H-3benzazepin-7-ol hydrobromide) (supplemental Fig. S2, available at www.jneurosci.org as supplemental material). A similar inward current was induced by 30 M DA in SNr GABA neurons voltage clamped at Ϫ70 mV under conventional whole-cell recording mode with K-gluconate-based intracellular solution  (Zhou et al., 2008). E, E1 shows the dose-response relationship of DA-induced inward current. The curve is from Hill equation fitting: EC 50 was 13.3 Ϯ 0.7 M and Hill slope of 1.2 Ϯ 0.1. n ϭ 5-9 for each data point. E2 and E3 are example inward currents induced by 5 and 100 M DA, respectively, in an SNr GABA neuron voltage clamped at Ϫ70 mV. The ramp was between Ϫ90 and Ϫ20 mV. The 30 M DA-induced net current (red trace) was obtained by subtracting the current generated under control conditions (black trace) from the current generated during DA application (blue trace). In other words, the red, net current trace ϭ the blue DA trace Ϫ the black control trace. Linear regression analysis indicated a reversal potential at approximately Ϫ37 mV. B, The same voltage ramp protocol revealed that 5 M SKF81297 (SKF) induced a similar linear current with a reversal potential also at approximately Ϫ37 mV. Identical subtraction procedure was performed. All traces were individual traces.
Next, a voltage-ramp protocol was used to determine the I-V relationship of the DA-induced current. The linear ramp was between Ϫ90 and Ϫ20 mV. After obtaining a stable baseline recording, the ramp protocol was activated and 30 M DA was bath applied. The net DA-induced current was obtained by subtracting the current generated under control condition from the current generated during DA application (Fig. 3A). Linear regression analysis indicated that the DA current I-V was linear between Ϫ90 and Ϫ20 mV with a reversal potential of Ϫ37.3 Ϯ 1.3 mV (n ϭ 5), suggesting a tonically active, mixed cation channel. It was not I h because I h has a voltagedependent activation with a threshold at Ϫ60 mV or more negative (Harris and Constanti, 1995). In SNr GABA neurons, I h is very small. Furthermore, the DA current I-V curve in the presence of ZD7288 (4-ethylphenylamino-1,2-dimethyl-6methylaminopyrimidinium chloride), a selective I h blocker (Harris and Constanti, 1995), was similar to that obtained without ZD7288. As will be discussed below (see Fig. 6), this current was likely mediated by TRPC3 channels that are tonically active cation channels in SNr GABA neurons (Zhou et al., 2008).

D 1 -like receptor agonists mimicked and an antagonist blocked D 1 -like effects in SNr GABA neurons
Based on the results described above, we reasoned that D 1 and/or D 5 receptors were mediating the direct excitation of DA in SNr GABA neurons. If this hypothesis was correct, then D 1 -like agonists should mimic the effects of DA. To test this reasoning, we investigated the effects of SKF81297 (6-chloro-2,3,4,5-tetrahydro-1-phenyl-1 H-3-benzazepine hydrobromide) and dihydrexidine, two selective, full agonists for D 1 -like receptors (Lovenberg et al., 1989;Aosaki et al., 1998). After a stable baseline recording, bath application of 5 M SKF81297 had a clear excitatory effect on SNr GABA neurons (Fig. 4), similar to that induced by 30 M DA (Fig. 2). The spontaneous spike firing rate was increased by 29.6 Ϯ 2.5% during SKF81297 application (n ϭ 9; p Ͻ 0.01) (Fig.  4 A, B). This effect was prevented by 5 M D 1 -like blocker SKF83566 (Fig. 4 B), clearly indicating an involvement of D 1like receptors.
To further characterize the depolarization that accompanied the SKF81297-induced increase in SNr GABA neuron firing rate, action potentials were blocked with 1 M TTX such that changes in membrane potential could be reliably detected. Under these conditions, bath application of 5 M SKF81297 induced a clear depolarization of 2.6 Ϯ 0.2 mV (n ϭ 9), and this effect was prevented by 5 M D 1 -like antagonist SKF83566 (Fig. 4C).
When the SNr GABA neurons were voltage clamped at Ϫ70 mV, 5 M SKF81297 induced an inward current of 30.8 Ϯ 3.8 pA (n ϭ 8) (Fig. 4 D), and whole-cell conductance was increased to 6.22 Ϯ 0.34 nS from 5.12 Ϯ 0.40 nS under control condition (n ϭ 6), indicating an opening of ion channels. These effects were prevented by 5 M SKF83566 (Fig. 4 D). Furthermore, in the presence of 1 M TTX, voltage ramp experiments showed that the SKF81297-induced current was linear between Ϫ90 and Ϫ20 mV with a reversal potential of Ϫ37.5 Ϯ 1.5 mV (n ϭ 4) (Fig.  3B). These characteristics were identical to those of the DAinduced current (Fig. 3A).
Similar current-clamp and voltage-clamp results were obtained with a second D 1 -like agonist, dihydrexidine, that is molecularly distinct from the phenylbenzazepines such as SKF81297 (Lovenberg et al., 1989).
Together, these data indicate that D 1 -like agonists mimic, and a D 1 -like antagonist blocks, DA-induced effects in SNr GABA neurons, supporting our idea that D 1 and/or D 5 receptor activation may directly depolarize these basal ganglia output neurons.
In contrast, in the presence of 5 M D 1 -like antagonist SKF83566, neither D 2 -like receptor agonist quinpirole (Lacey et al., 1987;Mercuri et al., 1997) (1-10 M) nor D 2 -like receptor antagonist sulpiride (1 M) affected the firing rate, membrane potential, or membrane current in SNr GABA neurons (n ϭ 2-3; data not shown). These results were consistent with our scRT-PCR data showing a lack of D 2 , D 3 , and D 4 mRNA in SNr GABA neurons. Quinpirole and sulpiride had typical effects in SNc and SNr DA neurons (n ϭ 3-6) as reported in the literature (Lacey et al., 1987). The SKF81297-induced firing enhancement was 29.6 Ϯ 2.5%, measured over the 2 min peak enhancement period. n ϭ 9. The D 1 -like excitatory effect was prevented by the D 1 -like antagonist SKF83566. Note that SKF83566 induced an inhibition indicated by the arrowhead, suggesting an endogenous tonic D 1 -like activity that will be discussed later. At the end of the recording, 30 M DA was applied to show that the neuron was still healthy. C, D, Example recordings showing that, after blocking spikes with 1 M TTX, SKF81297 induced a clear depolarization in current-clamp recording mode (C) and a clear inward current when voltage clamped at Ϫ70 mV (D). This depolarization or inward current was prevented by the D 1 -like antagonist SKF83566 (5 M). Also note that SKF83566 itself induced a hyperpolarization or outward current (arrowhead in C, D), suggesting an endogenous tonic D 1 -like activity (addressed later).
Absence of D 1 -like DA effects in SNr GABA neurons in D 1 or D 5 receptor KO mice D 1 and D 5 receptors are known to induce similar effects (Neve et al., 2004), and SNr GABA neurons express both D 1 and D 5 receptor mRNAs (Fig. 1 B, C). Thus, the D 1 -like effects described above may be mediated by D 1 and/or D 5 receptors. There is currently no ligand selective for D 1 or D 5 receptors. To overcome this difficulty, we used D 1 receptor KO mice and D 5 KO mice (Xu et al., 1994;Hollon et al., 2002).
In contrast to wild-type mice, in SNr GABA neurons from D 1 KO mice, 30 M DA had no significant effect on the spontaneous firing rate (9.9 Ϯ 0.8 Hz under control vs 9.7 Ϯ 0.8 Hz during DA; p Ͼ 0.05; n ϭ 5). In the presence of 1 M TTX, 30 M DA had no significant effect on the membrane potential (n ϭ 4) (Fig. 5A) or membrane current (n ϭ 5; data not shown) in SNr GABA neurons from D 1 KO mice. SNr GABA neurons (n ϭ 5) from wildtype littermates showed the same D 1 -like DA responses as in wild-type mice described previously. Similar to D 1 KO mice, in SNr GABA neurons from D 5 KO mice, 30 M DA did not affect the spontaneous firing rate (9.6 Ϯ 0.7 Hz under control vs 9.5 Ϯ 0.7 Hz during DA; p Ͼ 0.05; n ϭ 4), membrane potential (n ϭ 4) (Fig. 5B), or membrane current (n ϭ 5).
Also, in these same D 1 KO mice and D 5 KO mice, SNc DA neurons (n ϭ 4 -6) and SNr DA neurons (n ϭ 3-4) displayed the typical D 2 receptor-mediated autoinhibition when 30 M DA was bath applied, indicating that the lack of D 1 -like responses in SNr GABA neurons in these mutant mice was not attributable to unhealthy brain slices.
In aggregate, these results indicate an involvement of an apparent D 1 and D 5 receptor coactivation, i.e., both D 1 and D 5 receptors is required for the D 1 -like responses.

D 1 or D 5 receptor antibody inhibited the D 1 -like effects in SNr GABA neurons
To provide additional evidence from a different approach for the D 1 and D 5 receptor coactivation indicated by D 1 and D 5 receptor knock-out mice, we used D 1 and D 5 receptor antibodies against an intracellular domain of the receptor proteins and applied these antibodies intracellularly via the patch pipette. Intracellular infusion of proteins (enzymes and antibodies) is a well established technique (Tauc, 1980;Aragay et al., 1995). The antibody binds to and may thus interfere with the functioning of the corresponding receptor. This strategy has been used to identify receptor subtypes when subtypespecific ligands are not available (Amaral and Pozzo-Miller, 2007;Zhao et al., 2008;Zhou et al., 2008) (Dr. Jonathan Jaggar, personal communication). The antibodies, purchased from Santa Cruz Biotechnology, were diluted at 1:100 in intracellular pipette solution. We found that intracellular application of D 1 receptor antibody caused a clear hyperpolarization in current clamp (0.6 Ϯ 0.1 mV; n ϭ 5) (Fig. 5C). Effective infusion of the antibody into the recorded neuron was confirmed by immunohistochemical detection (Fig. 5C, inset). More importantly, in cells infused with D 1 receptor antibody, 30 M DA induced a depolarization of only 0.6 Ϯ 0.1 mV (n ϭ 5) (Fig. 5C), a 78% inhibition compared with that under control condition shown in Figure 2C. When voltage clamped at Ϫ70 mV, similar results that mirror the current-clamp data were obtained (data not shown). In contrast, a mixture of D 1 receptor antibody and its antigenic peptide (at a weight ratio of 1:1 or 1:2) had no effect by itself and did not inhibit the 30 M DA-induced depolarization (2.7 Ϯ 0.4 mV; n ϭ 3) (Fig. 5D). Thus, the effects of D 1 receptor antibody were specific.
Intracellularly applied D 5 antibody had effects similar to those of D 1 receptor antibody (n ϭ 5) (Fig. 5 E, F ). Furthermore, D 5 receptor antigen peptide did not neutralize the effects of the D 1 antibody (n ϭ 2), and D 1 antigen peptide did not neutralize the effects of the D 5 antibody (n ϭ 2), also indicating that the effects of D 1 and D 5 antibodies were specific. As shown in Figure 5, C and E, intracellularly applied D 1 or D 5 antibody, respectively, also induced a hyperpolarization, indicating a tonic D 1 and D 5 receptor-

KO mice (A) and D 5 KO mice (B). C, D,
In wild-type mice, intracellular infusion of D 1 receptor antibody caused a small but clear hyperpolarization, indicating a tonic D 1 -dependent depolarization. Once infused with D 1 antibody, the DA-induced depolarization was greatly inhibited (C). Inset, Conventional fluorescence pictures of immunostaining detection of D 1 antibody infused into a PV-positive SNr GABA neuron. In contrast, intracellular infusion of a mixture of D 1 antibody and its antigen peptide had no effect by itself and did not inhibit the DA response (D). E, F, Similarly, in wild-type mice, intracellular infusion of D 5 receptor antibody caused a hyperpolarization (0.6 Ϯ 0.1 mV; n ϭ 5), indicating a tonic D 5 -dependent depolarization. Once infused with D 5 antibody, 30 M DA only induced a depolarization of 0.6 Ϯ 0.1 (E). Intracellular infusion of a mixture of D 5 antibody and its antigen peptide had no effect by itself and did not inhibit the DA response with 30 M DA inducing a depolarization of 2.5 Ϯ 0.4 mV (F ). Comparable voltage-clamp data were also obtained. G, H, Intracellular infusion of D 1 receptor antibody into SNr GABA neurons in D 1 KO mice had no effect (G). Similarly, infusion of D 5 receptor antibody into SNr GABA neurons in D 5 KO mice also had no effect (H ). All traces were individual traces. dependent depolarization that was inhibited by either one of these two antibodies. Finally, none of these antibody-induced effects was seen in SNr GABA neurons in D 1 KO mice (n ϭ 3) and D 5 KO mice (n ϭ 3) (Fig. 5G,H ).
Together, our data clearly suggest that an apparent coactivation of D 1 and D 5 receptors was required to cause an inward current or depolarization in SNr GABA neurons. Inhibition of one of these two receptors led to the inhibition of the entire D 1 -like excitation.
Next we turned to this question: because D 1 and D 5 receptors do not directly gate any ion channel, what effector ion channel mediates the D 1 -like excitation? The experiments below were designed to identify the effector ion channel(s).

TRPC3 channels mediate the D 1 -like effects in SNr GABA neurons
Our data presented in Figure 3 indicate that the DA current I-V curve was linear between Ϫ90 and Ϫ20 mV with a reversal potential at approximately Ϫ37 mV, suggesting a constitutively active, mixed cation channel, similar to the I-V relationship of the TRPC3 channel-mediated current in SNr GABA neurons (Zhou et al., 2008). Our scRT-PCR data also showed that D 1 , D 5 , and TRPC3 mRNAs were coexpressed in the same SNr GABA neurons (Fig. 1C). Together, these data indicate that the apparent D 1 -D 5 receptor coactivation opened a type of depolarizing cation channels, consistent with the idea that DA may enhance the tonically active TRPC3 channels. To test this hypothesis, we investigated whether blocking TRPC3 channels also blocked the D 1 -like excitation (Fig. 6 A).
Because of the lack of selective agonists and antagonists, we applied a TRPC3 antibody intracellularly via the recording pi-pette as illustrated in Figure 6 A. The TRPC3 antibody (purchased from Alomone Labs) is known to interfere with or inhibit the functioning of TRPC3 channels Amaral and Pozzo-Miller, 2007;Zhou et al., 2008). Intracellular infusion of this antibody (diluted at 1:100 in the pipette solution) caused a hyperpolarization of 6.9 Ϯ 0.6 mV (n ϭ 6), probably by inhibiting the tonic TRPC3-mediated depolarization (Fig. 6 B, open arrow) (Amaral and Pozzo-Miller, 2007;Zhou et al., 2008). In these TRPC3 antibody-infused cells, the DA-induced depolarization was substantially inhibited with the residual amplitude being only 0.6 Ϯ 0.1 mV (n ϭ 6) (Fig. 6 B). The effects of TRPC3 antibody were neutralized by its antigenic peptide (antibody/ peptide ratio was 1:1 or 1:2 by weight) with an intact 30 DA M-induced depolarization (2.8 Ϯ 0.4 mV; n ϭ 3) (Fig. 6C), indicating that the inhibitory effects of TRPC3 antibody were specific. Together, these results strongly suggest that TRPC3 channels are the effector channel of D 1 -D 5 receptor coactivation in SNr GABA neurons such that an inhibition of TRPC3 channels leads to an inhibition of D 1 -like excitation.

D 1 -like receptor blockade decreased SNr GABA neuron firing frequency and increased their firing irregularity: evidence for tonic D 1 -D 5 receptor activity
Thus far, our data demonstrated unambiguously that exogenously applied DA-or selective D 1 -like agonists induced D 1 -like excitation in SNr GABA neurons. Now we turn to this critical question: can endogenous DA released from DA dendrites induce a similar D 1 -like excitation? Data addressing this question were already shown in Figures 4 and 5 and also in supplemental Figure  S2 (available at www.jneurosci.org as supplemental material) and will be discussed in combination with additional data.
We hypothesized that spontaneously released DA from DA neuron dendrites may induce a tonic activation of D 1 and D 5 receptors and exert a tonic influence on SNr GABA neurons. Consequently, blocking these D 1 and D 5 receptors may reveal this potential tonic excitation. To test this idea, we first blocked fast GABA-and glutamate-mediated synaptic transmission with picrotoxin, D-AP-5, and CNQX. After establishing stable baseline recording, 5 M SKF83566 (a D 1 -like blocker) was bath applied to probe the potential tonic D 1 /D 5 receptor activity. Indeed, SKF83566 significantly decreased SNr GABA neuron firing frequency by 8.6 Ϯ 1.7% (from 10.8 Ϯ 0.9 Hz under control conditions to 9.8 Ϯ 0.7 Hz during antagonist application; n ϭ 10; p Ͻ 0.01) (Fig. 8 A-D) (see also Fig. 4 B). These results suggest that there is indeed a tonic D 1 -like receptor activity that normally increases SNr GABA neuron firing rate.
After blocking action potentials with 1 M TTX and obtaining stable membrane potential, bath application of 5 M SKF83566 induced a clear hyperpolarization of 0.7 Ϯ 0.2 mV (n ϭ 9) in SNr GABA neurons (Fig. 4C). When the neurons were voltage clamped at Ϫ70 mV, bath application of 5 M SKF83566 induced a clear outward current (8.4 Ϯ 1.3 pA; n ϭ 8) (Fig. 4 D) (supplemental Fig. S2, available at www.jneurosci.org as supplemental material). Whole-cell conductance was also decreased from 5.08 Ϯ 0.37 nS under control condition to 4.72 Ϯ 0.35 nS during SKF83566 application (n ϭ 4). These effects were not seen in SNr GABA neurons in D 1 KO mice (n ϭ 2) and D 5 KO mice (n ϭ 2). Intracellular infusion of D 1 or D 5 receptor antibody also revealed a tonic D 1 and D 5 receptor-dependent excitation (Fig. 5C-H ). These data clearly indicate that, in SNr GABA neurons, inhibition of D 1 and D 5 receptors induced an outward current, most likely by downregulating the tonically active TRPC3 channels.
The SNr GABA projection neuron firing pattern is often altered in Parkinson's disease (Wichmann et al., 1999;Tang et al., 2005). Therefore, a question arises: can D 1 and D 5 receptor activation in SNr GABA projection neurons contribute to the regulation of the firing pattern in these neurons? If it does, then DA neuron degeneration will likely lead to degradation of this regulation, potentially contributing to the symptoms of Parkinson's disease. We noticed that, during SKF83566 exposure, the decrease in SNr GABA neuron firing frequency or the increase in interspike interval (ISI) was accompanied by an increase in the irregularity of ISI. To quantify this change, the coefficient of variation (CV) of ISI under control and during SKF83566 was calculated. CV was computed by dividing the SD of ISI by the mean ISI (Bennett and Wilson, 1999;Zhou et al., 2008). Indeed, bath application of 5 M SKF83566 increased the CV of ISI by 27.5 Ϯ 4.5% ( p Ͻ 0.05; n ϭ 8) (Fig. 8C,D). These results indicated that tonic activation of D 1 and D 5 receptors arising from endogenous DA release may help keep SNr GABA neurons sufficiently depolarized and spiking reliably at a regular interval.

Additional evidence for DA neurons directly influencing SNr GABA neurons
To provide more evidence for our hypothesis that SNc DA neurons can directly affect SNr GABA neurons, we did the following three additional sets of experiments.
In the first set of experiments, we stimulated SNc DA neurons and monitored responses in SNr GABA neurons. To evoke the maximal somatodendritic DA release and hence the maximal DA response, we adopted the published methods of burst-stimulating DA neurons (Chen and Rice, 2001;John et al., 2006). Each stimulation consisted of a burst of 10 50-Hz pulses delivered at every 5 min that prevents DA release depletion (Chen and Rice, 2001;John et al., 2006). Each pulse was 0.1 ms at 80 -120 A. The bipolar stimulating electrode was placed in the SNc. The recorded SNr GABA neurons were voltage clamped at Ϫ80 mV to increase driving force and prevent spontaneous spikes. The bathing solution was a blocking mixture (for details, see the legend of Fig. 9A). Under these conditions, burst stimulation of DA neurons induced a postsynaptic-like response in SNr GABA neurons. The peak amplitude was 7.1 Ϯ 0.8 pA (n ϭ 4) (Fig. 9A). The rise time was ϳ0.5 s and the decay time was ϳ8 s, although the small amplitude and consequently the interference from noise prevented precise measurement of these two parameters. This postsynaptic-like inward current was blocked by 2 M D 1 -like antagonist SKF83566 (Fig. 9A), indicating that somatodendritically released DA may activate D 1 -like receptors in SNr GABA neurons. In contrast, endogenous DA induced a postsynapticlike, G-protein-gated inwardly rectifying K ϩ channel-mediated, outward current in DA neurons by activating D 2 -like receptors in DA neurons (Beckstead et al., 2004).
In the second set of experiments, we examined the effects of D 2 -like agonist quinpirole that is known to inhibit DA neuron firing and presynaptic DA release via D 2 autoinhibition (Lacey et al., 1987;Cragg and Greenfield, 1997;Pothos et al., 1998). We found that quinpirole (1 M) inhibited SNr GABA neuron firing rate by 5.9 Ϯ 1.0% (n ϭ 11) (Fig. 9B). K-S tests on the distributions of firing rates in individual SNr GABA neurons indicate that the inhibition was significant ( p Ͻ 0.05). We also tested the effects of D 2 -like antagonist sulpiride that is known to increase DA neuron firing and presynaptic DA release (Pucak and Grace, 1994;Cragg and Greenfield, 1997). We found that sulpiride (1 M) increased SNr GABA neuron firing rate by 6.2 Ϯ 1.1% (n ϭ 10; p Ͻ 0.05) (Fig. 9B).
These results indicate that manipulation of DA neuron activity and DA release can affect SNr GABA neuron activity. The relatively weak effects are consistent with the established fact that DA neuron dendrites have Na ϩ -spike-independent, pace-making activity-driven, L-type Ca 2ϩ channel-mediated Ca 2ϩ oscillation (Chan et al., 2007) that may be essential for the dendritic DA release. Certainly, Ca 2ϩ release from internal stores may also contribute (Patel et al., 2009). Another factor is that the responses were recorded in 300-m-thick brain slices in which substantial numbers of DA neurons and GABA neurons and their dendrites were killed or injured during the tissue sectioning procedure, leading to a reduced magnitude of the endogenous DAmediated interactions. Thus, the endogenous DA-mediated interactions in intact animals are likely to be substantially larger.

Discussion
Based on a series of experiments using multiple different approaches, the major finding of this study is that nigral DA neurons form a novel ultra-short SNc 3 SNr DA pathway, in addition to the established long-distance nigro-striato-nigral loop (Fig. 9). In this ultra-short DA pathway, dendritically released DA exerts direct excitatory effects on SNr GABA projection neurons by an apparent coactivation of D 1 and D 5 receptors that subsequently enhances constitutively active TRPC3 channels in these neurons, leading to enhanced and more regular basal ganglia output.

Dendritically released dopamine induces tonic D 1 and D 5 receptor activity in SNr GABA neurons
SNr GABA projection neurons intermingle with DA dendrites that release DA (Björklund andLindvall, 1975, Geffen et al., 1976;Cheramy et al., 1981;Cragg et al., 2001;Falkenburger et al., 2001). Endogenously released DA has been suggested to activate D 1 -like receptors on GABA and glutamate afferent terminals in the nigral areas and, hence, facilitate GABA and glutamate release (Cameron and Williams, 1993;Ibañez-Sandoval et al., 2006). Cheramy et al. (1981) also indicated that DA may directly affect SNr GABA projection neurons. Iontophoretic studies on DA effects on SNr GABA neurons in intact animals have produced conflicting results. For example, early studies (Ruffieux and Schultz, 1980;Waszcak and Walters, 1983) found that DA exerted an excitatory effect on SNr neurons, whereas a recent study (Windels and Kiyatkin, 2006) reported that DA had no consistent effects. Note that the membrane potential during SKF83566 went below the initial baseline because SKF83566 blocked the tonic D 1 -like activity. Traces in A were digitized at 10 kHz in which the trace in D was digitized at 100 Hz. All traces are individual traces.
Another recent study (Kliem et al., 2007) indicated that local infusion of a D 1 -like agonist generally reduced SNr GABA neuron firing rate via presynaptic mechanisms. However, these studies used extracellular recordings and consequently could not conclusively determine whether the effects were a direct excitation in SNr GABA neurons or were local circuit-mediated effects. These studies might also have been complicated by the fact that DA may affect multiple, potentially opposing synaptic inputs (such as glutamate and GABA) that may mask each other's effects on SNr GABA neuron firing.
To ensure that the observed responses were direct effects in the recorded SNr GABA neurons, our present study was performed after blocking these fast glutamate and GABA synaptic inputs. Under this isolated condition, we found that DA or D 1 -like agonists (SKF81297 and dihydrexidine) induced similar D 1 -like excitatory effects (depolarization or inward current accompanied by increased whole-cell conductance) in SNr GABA neurons. All these effects were prevented by a D 1 -like receptor antagonist SKF83566. These effects were also absent in D 1 KO mice and D 5 KO mice and inhibited by intracellular application of D 1 or D 5 antibody. These results plus the data that SNr GABA neurons express both D 1 and D 5 mRNAs lead to the compelling conclusion that coactivation of D 1 and D 5 receptors induces a direct excitation in these basal ganglia output neurons.
In the present study, we also found that, after blocking fast GABA and glutamate receptors, D 1 -like antagonist SKF83566 induced a hyperpolarization in SNr GABA neurons and decreased their firing frequency accompanied by an increase in firing irregularity. These effects were absent in D 1 KO mice and also in D 5 KO mice, strongly suggesting that spontaneously released DA may induce a tonic coactivation of D 1 and D 5 receptors that influences SNr GABA neuron activity. Direct stimulation of DA neurons induced a direct inward current in SNr GABA neurons that was sensitive to D 1 -like blockade. Furthermore, manipulation of DA neuron activity and DA release with D 2 -like receptor ligands, and blockade of DA reuptake in particular, affected SNr GABA neuron firing. These data provide additional evidence that SNc DA neurons can directly influence SNr GABA neurons.
SNr GABA neurons may also feedback onto SNc DA neurons, as indicated by D 1 -like agonist SKF81297-induced increase in spontaneous IPSCs (sIPSCs) in DA neurons (supplemental Fig.  S3, available at www.jneurosci.org as supplemental material). This result is consistent with our data that SKF81297 increased SNr GABA neuron firing, although the sIPSCs in DA neurons might have originated in SNr GABA neurons and also other GABA afferents (Tepper et al., 1995;Mailly et al., 2003;Misgeld, 2004).

D 1 -D 5 receptor coactivation enhances TRPC3 channel activity in SNr GABA neurons
Our scRT-PCR consistently detected both D 1 and D 5 receptor mRNAs in SNr GABA neurons, although previous studies detected D 1 receptors only on striatonigral axon terminals but not in SNr GABA neurons (Mansour et al., 1992;Levey et al., 1993;Ariano, 1997). D 5 receptor expression in SNr was implied, although the cell type(s) expressing this receptor was not known (Choi et al., 1995;Khan et al., 2000). The use of the highly sensitive and specific scRT-PCR techniques in the present study may explain this difference in detection. D 1 and D 2 receptor coactivation and the underlying molecular mechanisms are relatively well established (Hopf et al., 2003;Rashid et al., 2007). Our functional demonstration of the requirement of both D 1 and D 5 receptors or coactivation of D 1 and D 5 receptors for the D 1 -like excitation is a novel finding. Despite a lack of direct molecular evidence, a recent review (Rashid et al., 2007) implied a D 1 -D 5 receptor coactivation that can stimulate G q/11 activation and lipid signaling mechanism. Our present study provides the first data supporting this implied novel signaling pathway associated with these DA receptors. The apparent D 1 and D 5 receptor coactivation we observed may arise from receptor dimerization and/or downstream signaling mechanisms. Future studies are needed to elucidate the molecular mechanisms underlying this apparent D 1 and D 5 receptor coactivation.
Because D 1 and D 5 receptors do not directly gate any ion channel, to what ion channel does this D 1 -D 5 coactivation couple? Our results show that the biophysical characteristics of the DA-induced inward current are almost identical to those of TRPC3 channel-mediated current: both are linear and voltage independent with reversal potential at approximately Ϫ37 mV (Fig. 3) (Zhou et al., 2008). Furthermore, intracellular application of TRPC3 channel antibody inhibited the D 1 -like effect in SNr GABA neurons (Fig. 6). These results strongly suggest that the D 1 -D 5 receptor coactivation may ultimately couple to and enhance the constitutively active TRPC3 channels (Fig. 6 A) that in turn regulate SNr GABA neurons. A DA regulation of TRPC3 channel activity is entirely possible because D 1 -D 5 coactivation may activate G q/11 and phospholipase C, thereby triggering lipids-mediated signaling mechanisms (Rashid et al., 2007) that are known to stimulate TRPC3 channels (Clapham, 2003;Hardie, 2007;Nilius et al., 2008).

Functional implications
Based primarily on the fact that DA dendrites may release DA in SNr, Cheramy et al. (1981) suggested that DA may directly affect SNr GABA projection neurons. Two pioneering studies indicated that iontophoretically applied DA exerts an excitatory effect on SNr neuron firing in intact animals (Ruffieux and Schultz, 1980;Waszcak and Walters, 1983). However, these studies used ex- Figure 10. Diagrammatic representation of the novel ultra-short SNc 3 SNr DA pathway identified and characterized by our present study. DA dendrites are known to release DA and induce D 2 -mediated autoinhibition in DA neurons. There is also the established long-distance nigro-striato-nigral loop. We have demonstrated that SNr GABA neurons express D 1 and D 5 receptors that ultimately couple to tonically active TRPC3 channels. Dendritically released DA also directly acts on SNr GABA output neurons, forming an ultra-short DA pathway that contributes to the regulation of the firing frequency and pattern of these basal ganglia output neurons. GPe, Globus pallidus external part; motor N, brainstem motor nuclei; BG, basal ganglia; STN, subthalamic nucleus; SC, superior colliculus. tracellular recordings, did not remove synaptic inputs, and consequently could not determine whether the effect was a direct excitation in SNr neurons or local circuit-mediated effect. Because of technical limitations at these early dates, these studies also could not determine the DA receptor subtype(s) responsible for the effect. Two recent studies revisited this topic but led to conflicting conclusions on DA regulation of SNr GABA neurons (Windels and Kiyatkin, 2006;Kliem et al., 2007). Our present results are consistent with those of the two early in vivo studies (Ruffieux and Schultz, 1980;Waszcak and Walters, 1983). More importantly, our data provide cellular and molecular mechanisms for these early in vivo observations.
The nigral DA neurons are commonly described to regulate basal ganglia output through the long distance nigro-striatonigral pathway (Parent et al., 2000;Haber, 2003). However, DA dendrites and SNr GABA neurons intermingle with each other. As illustrated in Figure 10, our study shows that there is also an ultra-short SNc 3 SNr DA pathway that directly regulates SNr GABA neuron output. Besides the well known D 2 receptormediated autoinhibition of nigral DA neurons, our data indicate that dendritically released DA induces a tonic D 1 and D 5 receptor coactivation-mediated hetero-excitation in SNr GABA projection neurons. This D 1 and D 5 receptor coactivation couples to and enhances tonically active TRPC3 channels, leading to the depolarized membrane potential, high firing rate, and regular firing pattern in SNr GABA projection neurons. Loss of DA neurons in Parkinson's disease may impair this novel ultra-short DA pathway, contributing to the complex neuropathophysiology and, hence, movement deficits. Indeed, a recent study found movement control deficits in TRPC3 channel knock-out mice (Hartmann et al., 2008). Although loss of TRPC3 channel in the cerebellum may be primarily responsible for the motor deficits, TRPC3 channel loss in SNr may also be involved.