Cortical Control of Striatal Dopamine Transmission via Striatal Cholinergic Interneurons

1Striatal dopamine (DA) transmission plays critical roles in action selection, reward-related learning, habit formation, and disorders such as Parkinson's disease and addiction. It is becoming increasingly appreciated that DA release is governed by striatal mechanisms in addition to those that govern action potential generation in substantia nigra (SN) or ventral tegmental area (VTA). The axonal propagation of action potentials could be moderated by thousands of branch points found on the axonal arbor of each DA neuron, and at release sites, the dynamic probability of DA release is influenced by presynaptic mechanisms including local inputs. Inputs from striatal cholinergic interneurons (ChIs) in particular are emerging as powerful players. ACh, by acting at nicotinic receptors (nAChRs) on DA axons, can filter how different firing frequencies are relayed into the release of DA and furthermore, synchronous activity in a population of ChIs can trigger striatal DA release directly, bypassing action potential generation in DA soma. We have identified further that glutamate inputs to ChIs from the predominantly parafascicular nucleus (Pf) of caudal intralaminar thalamus can gate striatal DA release through a mechanism that appears to require ionotropic glutamate receptors on striatal ChIs.

2It has long been thought that cortex might be able to regulate striatal DA transmission, with facilitatory and inhibitory effects observed, but the facilitatory role of ionotropic glutamate receptors in particular has had little explanation because these receptors have not been localized to DA axons. However, ChIs express a variety of glutamate receptors. A monosynaptic connection between cortical input and ChIs has been suggested from a wealth of electrophysiological experiments showing modifications to ChI activity after cortical input stimulation in vivo and ex vivo, but anatomical studies have suggested that these dendritic inputs are sparse. Although historically perceived to be a sparse input, more recent retrograde tracer experiments suggest that cortical innervation might account for a significant share of inputs to ChIs. We explored whether cortical inputs could modulate DA transmission.

3We show here that M1/M2 cortical inputs to striatum modulate DA release, requiring activation of ChIs via AMPA receptors, and downstream nAChRs. Furthermore, we show that Pf thalamic modulation of DA release via ChIs has greater dependence on NMDA receptors than cortical input. These findings indicate regulation of ChIs by thalamic versus cortical inputs through different integration strategies that might differently govern how ChIs modulate striatal function, including DA transmission.

4Heterozygous and homozygous adult (21–40 days) male and female CaMKIIa-Cre mice (B6.Cg-Tg(CaMK2a-Cre)T29-1Stl/J, stock 005359) were anaesthetized with isoflurane (2% w/o), and injected bilaterally with 400 nL of an adeno-associated virus backbone serotype 5 (AAV5, 8×10e12 particles/mL), carrying ChR2-eYFP (pAAV-EF1α-DI0-hChR2(H134R)EYFP-WPRE-pA) (Gene Therapy Center Virus Vector Core, University of North Carolina). Serotype AAV5 was used for good transduction efficiency and spread, and was used previously to successfully drive thalamostriatal inputs to ChIs and DA release. Injections for motor cortex (mCtx) were targeted to ML 1.7 mm, AP +1.3 mm, DV 1.0 mm from bregma. Injections targeted to the Pf nucleus of the caudal intralaminar thalamus were ML 0.7 mm, AP −2.3 mm, DV 3.5 mm from bregma. Coordinates were taken from the adult mouse brain atlas. Injection sites were confirmed with post hoc immunocytochemistry for eYFP expression and co-injected red polystyrene fluorescent spheres (Invitrogen). In all, 4–10 weeks were allowed for transgene expression.

5DA release was monitored using fast-scan cyclic voltammetry (FCV) in acute coronal slices as used previously. In brief, after brain removal following cervical dislocation and decapitation, 300 μm coronal slices were cut on a vibratome (Leica VT1200S) in ice-cold HEPES-buffered artificial cerebrospinal fluid (aCSF) saturated with 95% O2/5% CO2. Slices between +1.5 mm and +0.5 mm from bregma containing caudate-putamen and nucleus accumbens (i.e. striatum) were used for experimentation. In the recording chamber, slices were maintained at 32 °C in aCSF containing in mM: 130 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 2 MgCl2, 10 glucose. Striatal DA release following light activation of ChR2-expressing fibers was monitored with FCV at carbon-fiber microelectrodes (CFMs) using a Millar voltammeter (Julian Millar, Barts and the London School of Medicine and Dentistry). CFMs were fabricated in-house from epoxy-free carbon fiber 7–8 μm in diameter (Goodfellow Cambridge Ltd) enclosed in glass capillary and cut to final tip length of 50–100 μm. A triangular waveform was applied to carbon fiber scanning from −0.7 V to +1.3 V and back, against Ag/AgCl reference electrode at a rate of 800 V/s. Evoked DA transients were sampled at 8 Hz, and data were acquired at 50 kHz using AxoScope 10.2 (Molecular Devices) or Whole Cell Program (WCP, University of Strathclyde, Glasgow, UK). Recorded FCV signals were identified as DA by comparing oxidation (+0.6 V) and reduction (−0.2 V) potential peaks from experimental voltammograms with currents recorded during calibration with 2 μM DA dissolved in experimental media.

6Methods used for whole-cell patch clamp in acute coronal slices were as described previously. Briefly, mice were anaesthetized with pentobarbital and transcardially perfused with ice-cold high Mg2+ aCSF containing in mM: 85 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 0.5 CaCl2, 7 MgCl2, 10 glucose, 65 sucrose. Coronal 300 μm slices were cut on a vibratome (Leica VT1200S) at coordinates between +1.5 mm and +0.5 mm from bregma. Slices recovered at 32°C for 30–40 min after dissection and were subsequently kept at room temperature. Slices were maintained and recorded from in aCSF containing in mM: 130 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 2 MgCl2, 10 glucose. The aCSF was saturated with 95% O2/5% CO2; recordings were made at 32°C.

7Glass electrodes for whole-cell patch clamp (5–9 MΩ) were filled with an intracellular solution containing in mM: 120 K-gluconate, 10 KCl, 10 HEPES, 4 MgATP, 0.3 NaGTP, 10 Na-phosphocreatine and 0.5% neurobiotin. ChIs in the striatum were identified by their distinctive morphological features, that is, large somas (>20 µm) and their characteristic electrophysiological properties, that is prominent Ih, after-hyperpolarization (AHP) and broad action potential. A minimum negative holding current (<40 pA) was applied to prevent spike activity during recordings of laser-induced excitatory post-synaptic potentials (EPSPs). Recordings were acquired using a Multiclamp 700B at 10–20 kHz. All data were analyzed offline with Clampfit (pClamp10), and custom-written Matlab (R2013b) scripts. ChIs were confirmed by post hoc labelling for choline acetyltransferase (ChAT) and neurobiotin, as previously.

8ChR2-eYFP-expressing neuropil in striatum was activated using blue light pulses (473 nm, OptoLED, CAIRN or laser DL473, Rapp OptoElectronic) transmitted through the microscope objective via a fluorescence arm, pulse width 2 ms, generated out of phase with voltammetric scans. Full-field illumination with the LED system for FCV experiments covered an area 2.2 mm in diameter under ×10 immersion objective, and light power density was 25–30 mW/mm2. Spot laser illumination for patch-clamp experiments had a 30 µm diameter under ×40 immersion objective, and light power density between 150 µW/mm2 and 23 mW/mm2. Light power used reflected variation in ChR2-eYFP levels and the experimental measure, with lowest levels being used to probe EPSPs without evoking action potentials, and higher intensities were required to drive inputs to make cells spike. Stimulation frequencies used were 10 and 25 Hz (400 ms trains) in order to reflect the frequencies observed in vivo in input neurons during activated/arousal states for Pf and cortical inputs, respectively. Since there was no effect of pulse number or frequency on DA evoked or effects of pharmacological ligands, 10 and 25 Hz data were pooled unless otherwise specified. Protocols were repeated at intervals of 2.5 min. Recordings for Pf and mCtx groups were acquired from across the CPu in an unbiased manner in relation to site location, but there was a tendency for more successful recordings in more lateral sites for mCtx injections and in more medial sites for Pf. However, the distributions of striatal recording locations for experiments that identified differences in receptor pharmacology were comparable and spanned the dorsal striatum.

9To verify cortical injection sites, striatal coronal sections used for recordings were fixed in 4% PFA and 0.2% saturated picric acid (pH 7.2–7.4) for at least 24 h at 4°C. Sections were washed in PBS, and mounted on gelled slides using Vectashield (Vector Laboratories) for visualization. To verify thalamic injection sites, brain blocks containing thalamus were taken after preparing striatal slices used in recordings. Blocks were fixed in 4% PFA and 0.2% saturated picric acid (pH 7.2–7.4) for at least 72 h at 4°C. Blocks were washed in PBS, re-sectioned at 50–100 μm on a vibrating microtome (VT1000S, Leica Microsystems), washed in PBS (5 × 5 min) and mounted with Vectashield (Vector Laboratories). Injection sites (red spheres) and eYFP-tagged ChR2-transduced neuropil were visualized using a fluoescence microscope Olympus BX41 (Olympus medical) coupled to Q-Click cooled monochrome CCD camera and Q-capturePro 7.0 software. Red spheres confirmed targeted injections. ChR2-eYFP expression targeted to Pf thalamus was expressed most densely and predominantly in Pf with some limited expression in wider intralaminar thalamus. For simplicity, and as the center of our targeted ChR2 expression, we use the term Pf here. ChR2 expression in off-target regions could also occur including in non-injected thalamus and cortex from their reciprocal connectivity; however, we could not resolve any cell body labelling in off-target sites, and moreover, the physiological and pharmacological differences we reveal here confirm that distinct inputs to striatum are differentially activated when we target ChR2 injections to either thalamus or cortex.

10To verify that recorded neurons were ChIs, neurons filled with neurobiotin were co-labelled for ChAT. Acute striatal slices were fixed after recording, in 4% paraformaldehyde dissolved in PBS containing 0.2% picric acid. Slices were fixed overnight at 4°C and then stored in PBS. Free-floating sections were then washed in PBS 5 × 5 min and incubated in 0.5% Triton X-100 and 10% normal donkey serum. Slices were subsequently incubated with goat anti-ChAT 1:100 (Millipore) antibody dissolved in PBS containing 0.5% Triton X-100 and 3% normal donkey serum overnight. Sections were then washed with PBS 5 × 5 min and incubated for 2 h at room temperature with 1:1000 Alexa Fluor 568 donkey anti-goat (Invitrogen) antibody dissolved in PBS containing 0.5% Triton X-100 and 3% normal donkey. Alexa 488-conjugated streptavidin (Invitrogen) was included in the secondary antibody solution at a final concentration of 1:250 to identify the recorded neurons. Sections then washed with PBS and mounted on gelled slides with Vectashield mounting medium (Vector Labs) and imaged using an AxioSkop fluorescent microscope (Zeiss).

11Data are expressed as mean ± SEM. The n value is the number of experiments unless otherwise stated, each conducted in a minimum of three different animals. For DA release data, we included a minimum of four release events for each stimulus or condition at individual recording sites, unless otherwise stated. Drug data were normalized to control data, and frequency data were normalized to single pulses, before collating across experiments. DA peak concentrations were statistically compared using Wilcoxon or Kruskall–Wallis non-parametric tests. For Figure 1i, data for release variability indicate the minimum value in release seen as a % of the maximum release seen at each release site. For EPSP data, the maximum EPSP amplitude and the area under the curve were normalized to the control condition within each group. Data were compared with the control condition (hypothetical value 1.000, i.e. the normalized control group) using a one-sample t-test.

12D-APV, GYKI 52466 hydrochloride (GYKI), oxotremorine-M (Oxo-M), MCPG, bicuculline and saclofen were obtained from Tocris Bioscience or Ascent Scientific. Dihydro-β-erythroidine (DHβE) and other chemicals were obtained from Sigma Aldrich. Stock aliquots of drugs were prepared at 1000–10 000× final concentrations in de-ionized water or aqueous acid (GYKI) and stored at −20°C. None of the drugs applied altered electrode sensitivity to DA. Data from immediately prior to drug application were compared with data acquired when drug effects equilibrated, which was typically 10–20 min after application.

13CaMKIIa-Cre mice received injections of an AAV-packaged floxed construct for ChR2-eYFP into either Pf thalamic nucleus as previously (Fig. 1a) or mCtx (Fig. 1b). Subsequently, ChR2-eYFP expression was detected in the striatal neuropil (Fig. 1a,b). In acute coronal slices containing ChR2-eYFP expression, we recorded extracellular DA concentration ([DA]o) using FCV at CFMs at a range of sites from across the dorsal striatum (Fig. 1a,b) whilst stimulating striatum with blue light pulses. To firstly corroborate previous observations, we confirmed that activation of thalamostriatal ChR2-eYFP afferents in Pf-injected mice evoked the release of striatal DA (Fig. 1c–e). We then explored the effect of light-activation of ChR2-eYFP afferents in mCtx-injected mice. Light activation of corticostriatal afferents powerfully drove DA release (Fig. 1f–h). Light-evoked [DA]o for both inputs was TTX-sensitive (Fig. 1d,g), Ca2+-dependent and approximated the levels evoked by local electrical stimulation (data not illustrated). Evoked [DA]o did not vary with frequency (5–40 Hz for 5 pulses) for Pf or mCtx inputs (Fig. 1e,h), a feature seen following activation of ChR2-expressing ChIs. We noted however that in contrast to the activation of DA release by direct light activation of ChR2-expressing ChIs reported previously, levels of DA release evoked by thalamic or cortical inputs were not easily predicted by eYFP expression at recording sites: firstly, ChR2-eYFP expression was not always sufficient for DA release; and secondly, DA release seen after light-activation could be highly variable on a stimulus-by-stimulus basis at a given release site (Fig. 1i), with the minimum evoked [DA]o release seen at each site varying reversibly 50–60% of maximum level seen (averaged across sites), and to as little as 5%. These observations suggest an additional process, such as an intermediary circuit between glutamate inputs and DA axons, bearing some autonomous dynamic properties that lead to its varying recruitment.

14We identified the receptors involved for both thalamostriatal- and corticostriatal-evoked DA release. For both inputs, DA release required activation of nAChRs: DA release was reversibly abolished by an antagonist for β2-subunit-containing nAChRs, DHβE (1 µM) (Fig. 2a,c). These data indicate that ACh action at nAChRs is required for DA release driven by activation of either Pf or mCtx, implicating activation of ChIs as the common driving mechanism. DA release was also prevented by a broad-spectrum muscarinic agonist, oxotremorine-M (10 µM) (Fig. 2b,d), consistent with previous observations that M2 and M4-mAChRs can inhibit DA release by inhibiting ACh release from ChIs and ChI activity, but also consistent with presynaptic inhibition of glutamate input from corticostriatal terminals by M2/M4-mAChRs. Furthermore, application of an NMDA receptor antagonist (D-APV, 50 µM) followed by an AMPA-receptor antagonist (GYKI, 10 µM) showed that DA release driven by Pf involved both NMDA and AMPA receptors (10 or 25 Hz, Fig. 2e), whereas mCtx inputs predominantly relied on AMPA receptors (10 or 25 Hz, Fig. 2f). Note that DA release driven in CPu by light-activation of ChIs directly does not involve glutamate receptors; and therefore, the glutamate dependence of DA release driven by Pf or mCtx does not result from glutamate potentially co-released by ChIs. There was no detectable effect of antagonists for metabotropic glutamate receptors (mGluRs), or GABAA or GABAB receptors (Fig. 2g–l). Mean evoked [DA]o was similar following stimulation of Pf versus mCtx afferents (Fig. 2m).

15These data suggest a disynaptic circuit in which corticostriatal and thalamostriatal glutamate inputs activate AMPA receptors on ChIs, with or without significant involvement of NMDA receptors, which, in turn, activate nAChRs on DA axons to trigger DA release (Fig. 2n). The variability in glutamate-evoked [DA]o seen over time could be consistent with compounded thresholds of transmission from multiple neurons in a serial circuit (glutamate-ACh-DA) and also with the autonomous dynamic activity expected of the intermediary ChIs. Glutamate inputs might be expected to modulate only the timing of autonomous pacemaking by ChIs, and increase but not guarantee spike probability, with a potentially variable outcome on DA.

16We identified whether ChR2-expressing inputs from mCtx and Pf were activating ChIs with these protocols. Light-activation of either Pf or mCtx afferents (10 and 25 Hz) significantly increased spike probability and firing rate in individually patched ChIs (Fig. 3a–d) although spike probability per stimulus was <1 (unlike direct activation of ChIs; ). Spike latency differed between inputs: mean latency to spike (averaged following all individual pulses) was short after mCtx activation (typically <10 msec) but significantly longer for Pf afferents (spanning ~60 msec) (Fig. 3b–e). Correspondingly, the glutamate receptor types co-varied with input type and with the receptors that regulated DA release: EPSPs generated by light-activated Pf inputs were NMDA- and AMPA-receptor mediated, whereas those activated by mCtx inputs were predominantly AMPA-receptor-mediated (Fig. 3f,g). These data reveal that different subsets of ionotropic glutamate receptors are recruited by Pf thalamic versus M1/M2 cortical inputs to ChIs, consistent with the receptor dependence of striatal DA release. They suggest furthermore that the time window during which ChI firing could be modified, and even synchronized, by its inputs to modulate DA release or other aspects of striatal function differs between cortical and thalamic inputs due to different postsynaptic receptors.

17We show here that cortical glutamate input to the dorsal striatum can regulate striatal DA release via activation of ionotropic glutamate receptors on ChIs which serve as gatekeepers. The cortical input to the striatum has been viewed as providing a relatively sparse innervation of ChIs compared with thalamic Pf inputs, but we identify here that both types of input can effectively recruit ChIs, and drive DA transmission. Therefore, cortical and thalamic inputs might modify striatal function through mechanisms that extend to the regulation of striatal DA release.

18The finding that glutamate inputs to striatum can modulate DA transmission via modulation of ChI activity finally provides a mechanism for earlier reports of DA efflux following glutamate application despite an absence of ionotropic glutamate receptors on DA axons. By promoting activity in ChIs, glutamate inputs could drive local striatal DA release via activating nAChRs on DA axons, as we observed here. This multi-synaptic event (Glu-ACh-DA) was insensitive to firing frequency of the glutamate inputs, which is in keeping with activation of ChIs as the intermediary: DA release is frequency insensitive when driven by direct activation of ChIs. We note that extrinsic sources of ACh input to the striatum from brainstem have recently been identified but, in pilot experiments we performed in collaboration, these inputs do not drive or otherwise modulate striatal DA release (not illustrated).

19DA release driven here likely resulted from synchronization of activity in a small network of ChIs. ChIs are autonomous pacemakers but receive excitatory inputs from thalamus and cortex that can promote and even synchronize ChI firing. We saw that DA release at some sites varied over time; and therefore, there is apparently moment-by-moment variation in the size of the ChI population recruited. This variability is consistent with the expectation that in cells such as ChIs, where spiking is predominantly driven by a strong autonomous pacemaker, excitatory inputs would be expected to change spike probability by modifying the timing of spiking. We speculate that glutamate inputs might modulate DA signals in vivo variably according to the timing of their input relative to ChI activity phase, which will determine how large a population of ChIs can synchronously fire action potentials for driving DA release.

20We show that M1/M2 cortical and Pf intralaminar thalamic inputs use different receptors for recruiting activity in ChIs and, in turn, for modulating DA release. Cortical inputs recruit predominantly AMPA receptors, while thalamic strongly recruit both AMPA plus NMDA receptors. The difference in balance of receptor types is consistent with the receptors shown to dominate the regulation of striatal ACh release and ChI activity after electrical stimulation, and with the relative dominance of receptors activated by these inputs to other striatal neurons, the spiny projection neurons, in some studies (but see ), although an ultrastructural explanation for these observations has yet to be resolved. This finding also supports the hypothesis that corticostriatal and thalamostriatal projection systems code information in temporally distinct ways. We saw that cortical inputs show a short latency and time window of operation, consistent with AMPA receptors having fast gating and desensitization. This brief window for summation of cortical inputs by ChIs will limit temporal summation, which could promote broadcasting of highly coincident cortical input to striatum. By contrast, thalamic inputs have a longer latency and more protracted time window for input summation and action potential generation, lasting tens of milliseconds, consistent with the slow gating of NMDA receptors. This large time window may lead to relatively long timescales over which inputs can be integrated and might facilitate summation of inputs arriving within a relatively extended period. Information carried by individual Pf neurons, which may not arrive with perfect synchronization but in only relative temporal proximity due to characteristic low-frequency burst discharges of Pf neurons, could nonetheless succeed in modulating ChI activity. Therefore, the different receptors and consequent input integration strategies used by ChIs could influence different components of action selection on different timescales.

21In conclusion, cortical and thalamic inputs have effects on ChI excitability mediated by AMPA and NMDA receptors that can be sufficient to modulate striatal DA transmission via nAChRs. In turn, cortical and/or thalamic inputs might regulate behavior and plasticity partly via modulation of striatal DA, even in the absence of detectable changes in firing rate in DA neurons. There could be a greater range of functions for DA than predicted from the activities of DA neuron firing alone. In this regard, it is of note that a discrepancy has recently been identified between the DA signals detected in nucleus accumbens throughout an adaptive decision-making task and the underlying DA neuron firing rates expected from published literature. These observations are consistent with the hypothesis that there are local striatal drivers of DA release, such as ChIs, which we now show can be governed by corticostriatal and thalamostriatal inputs. Our findings might therefore have significant implications for the regulation of behavior and for dysfunction in disorders that extend beyond Parkinson's disease to those involving dysfunctional corticostriatal inputs including Huntington's disease, attention-deficit and hyperactivity disorder (ADHD), obsessive-compulsive disorder and addiction, whose management includes targeting DA.

22Medical Research Council UK (grant number MR/K013866/1); Parkinson's UK (grant numbers G1305, J-0901 Monument Trust Discovery Award); and a Clarendon Fund Studentship (to P.K.).