Gating of dopamine transmission by calcium and axonal N‐, Q‐, T‐ and L‐type voltage‐gated calcium channels differs between striatal domains

The voltage‐gated Ca2+ channels (VGCCs) that catalyse striatal dopamine transmission are critical to dopamine function and might prime subpopulations of neurons for parkinsonian degeneration. However, the VGCCs that operate on mesostriatal axons are incompletely defined; previous studies encompassed channels on striatal cholinergic interneurons that strongly influence dopamine transmission. We define that multiple types of axonal VGCCs operate that extend beyond classic presynaptic N/P/Q channels to include T‐ and L‐types. We reveal differences in VGCC function between mouse axon types that in humans are vulnerable versus resistant to Parkinson's disease. We show for the first time that this is underpinned by different sensitivity of dopamine transmission to extracellular Ca2+ and by different spatiotemporal intracellular Ca2+ microdomains. These data define key principles of how Ca2+ and VGCCs govern dopamine transmission in the healthy brain and reveal differences between neuron types that might contribute to vulnerability in disease.


Introduction
Release of dopamine (DA) from mesostriatal DA neurons is critical to the selection and learning of our actions and motivations. Release of transmitters is catalysed by presynaptic VGCCs that provide a transient Ca 2+ microdomain but the VGCCs that govern DA transmission have not previously been resolved. Typically, N-type (Ca v 2.2) and P/Q-type (Ca v 2.1) VGCCs dominate in neurotransmission at CNS synapses (Rusakov, 2006), but it is increasingly evident that other VGCCs, including T-types (Ca v 3) and L-types (Ca v 1.2-4), may regulate neurotransmitter release from some neuron types (Pan et al. 2001;Tippens et al. 2008;Holmgaard et al. 2009;Zanazzi & Matthews, 2009;Tang et al. 2011).
Mesostriatal DA neurons possess a uniquely extensive arbour of DA release sites: axonal fields form the vast majority (ß99%) of the total membrane area of DA neurons (Matsuda et al. 2009;Henny et al. 2012) with individual rat DA neurons comprising approximately 0.5 m of axons (Matsuda et al. 2009) and half a million release sites (Arbuthnott & Wickens, 2007). These striking figures will be orders of magnitude higher in the human brain (Pissadaki & Bolam, 2013). The VGCCs that support axonal DA release will be critical to DA function, and might contribute to the Ca 2+ -dependent vulnerability of DA neurons to degeneration in Parkinson's disease (Mosharov et al. 2009;Collier et al. 2011). Subpopulations of DA neurons of the substantia nigra pars compacta (SNc) that innervate caudate putamen (CPu) are more vulnerable than those of the adjacent ventral tegmental area (VTA) innervating nucleus accumbens (NAc), and to date a somatodendritic L-type conductance (Ca v 1.3) in SNc but not VTA has been identified as a contributing factor (Chan et al. 2007;Guzman et al. 2009). The additional handling of Ca 2+ by axons might generate a substantial ionic load and metabolic or proteostatic burden Harris et al. 2012;Pissadaki & Bolam, 2013).
Previous studies in CPu identified that N-, P/Q-, R-and T-channels but not L-channels govern DA transmission (Phillips & Stamford, 2000;Chen et al. 2006;Sgobio et al. 2014) and suggested that VGCC subtypes might differ between striatal regions. It was also suggested that VGCC roles might differ during low tonic and high phasic firing frequencies to encode DA functions attributed to specific firing patterns (Phillips & Stamford, 2000). However, new insights suggest other factors could account for those findings. There are underlying short-term dynamics in DA release probability that govern frequency dependence that depend on Ca 2+ and vary between striatal regions (Cragg, 2003;Montague et al. 2004). Furthermore, an input from cholinergic interneurons (ChIs) to nicotinic acetylcholine receptors (nAChRs) on DA axons governs DA release probability, plasticity and frequency dependence (Zhou et al. 2001;Zhang & Sulzer, 2004;Cragg, 2006;Exley & Cragg, 2008;Threlfell et al. 2012) and these effects, alongside VGCCs on ChIs, will have confounded interpretation.
Here, we identify which VGCCs on mouse DA axons govern DA transmission in CPu versus NAc and the principles that underlie their dynamic participation. Furthermore, we reveal significant differences in the dynamic coupling of Ca 2+ to DA transmission.

Slice preparation
Male adult mice were C57Bl6/J wild-type (Charles River) or DA transporter (DAT)-Cre heterozygote mice used previously (Threlfell et al. 2012) of mean age 8 weeks, which did not differ for experiments between different regions. Mice were killed by cervical dislocation, the brains removed, and 300 μm coronal slices containing CPu and NAc prepared as described previously (Exley et al. 2012;Threlfell et al. 2012) in ice-cold Hepes-based buffer saturated with 95% O 2 /5% CO 2 , containing (in mM): 120 NaCl, 20 NaHCO 3 , 6.7 Hepes acid, 5 KCl, 3.3 Hepes salt, 2 CaCl 2 , 2 MgSO 4 , 1.2 KH 2 PO 4 and 10 glucose. Slices were incubated at room temperature for ࣙ1 h in Hepes-based buffer before transferral to the recording chamber. All procedures were carried out according to institutional guidelines and conformed to the UK Animals (Scientific Procedures) Act 1986.

Fast-scan cyclic voltammetry
DA release was monitored in acute slices using fast-scan cyclic voltammetry (FCV) as we have described previously (Exley et al. 2012;Threlfell et al. 2012). Slices were superfused in a recording chamber with bicarbonate-buffered artificial cerebrospinal fluid (aCSF) saturated with 95%O 2 /5%CO 2 at 31-33°C, containing (in mM): 124 NaCl, 26 NaHCO 3 , 3.8 KCl, 2.4 CaCl 2 , 1.3 MgSO 4 , 1.3 KH 2 PO 4 and 10 glucose. Evoked extracellular DA concentration ([DA] o ) was monitored using FCV at 7-10 μm diameter carbon fibre microelectrodes fabricated inhouse (tip length 50-100 μm) and a Millar voltammeter (Julian Millar, Barts and the London School of Medicine and Dentistry). In brief, a triangular voltage waveform (range −700 mV to +1300 mV vs. Ag/AgCl) was applied at 800 V s -1 at a scan frequency of 8 Hz. Electrodes were switched out of circuit between scans. Electrodes were calibrated using 1-2 μM DA in each experimental medium. Calibration solutions were prepared immediately before calibration from a 2.5 mM stock solution in 0.1 M HClO 4 stored at 4°C. Signals were attributable to DA by the potentials for peak oxidation and reduction currents (oxidation peak: +500-600 mV, reduction peak: ß−200 mV).

Electrical stimulation
Recordings were obtained from both dorsolateral CPu and NAc core. DA release was evoked by a local bipolar concentric Pt/Ir electrode (25 μm diameter; FHC Inc., Bowdoin, ME, USA) placed approximately 100 μm from the carbon fibre microelectrodes as previously (Threlfell et al. 2012;Exley et al. 2013). Stimulus pulses (200 μs duration) were given at 0.6 mA (perimaximal in control conditions). DA neurons in vivo exhibit a range of firing frequencies from ß1-40 Hz or higher. We applied either single pulses (1p) or five pulses (5p) at 5, 25, 40 and 100 Hz to span a full range of firing frequencies. Mean peak [DA] o evoked by 1p was equivalent to that of a 1 Hz train; 1p is used in frequency comparison to indicate maximum 1 Hz data. A frequency of 100 Hz may be supraphysiological but is useful as a tool for exposing changes in short-term plasticity (STP) that arise through changes in initial release probability . Electrical stimulations were repeated at 2.5 min intervals, which allow stable release to be sustained over several hours. Each stimulus type was repeated in triplicate in a random order.
All data were obtained in the presence of the nAChR antagonist, dihydro-β-erythroidine (DHβE, 1 μM) added to the recording aCSF, to inhibit nAChRs on DA axons and remove the confounding effects of VGCCs on cholinergic interneurons that regulate ACh release and ACh effects on DA Exley & Cragg, 2008). Experiments were conducted in the presence of 2.4 mM extracellular Ca 2+ unless otherwise stated. Muscarinic acetylcholine receptors do not regulate DA transmission during the stimulation protocols used here (Threlfell et al. 2010) and therefore it was not necessary to include a muscarinic acetylcholine receptor antagonist. Data were acquired without the use of DAT inhibitors, which are used in some studies to boost DA signals, as the DAT may modify the Ca 2+ dependence of DA release (Kile et al. 2010). Release was tetrodotoxin-sensitive as previously (Threlfell et al. 2010(Threlfell et al. , 2012, and was not modulated by glutamate or GABA antagonists as shown previously (Threlfell et al. 2010), and confirmed here (not illustrated).
We confirmed that the specificity of ChR2-eYFP expression in DAT-Cre animals was restricted to tyrosine hydroxylase-immunoreactive neurons, as shown previously (Threlfell et al. 2012). Midbrain slices were fixed in 4% paraformaldehyde before re-sectioning to 40 μm. Free-floating sections were washed in phosphate-buffered saline (PBS) and incubated in 0.5% Triton X-100 with 10% normal goat serum and 10% fetal bovine serum. Slices were incubated overnight at 4°C in primary antibody (1:2000 rabbit anti-tyrosine hydroxylase; Sigma-Aldrich, Dorset, UK) diluted in PBS with 0.5% Triton X-100, J Physiol 593.4 1% normal goat serum and 1% fetal bovine serum. Slices were washed in PBS then incubated in secondary antibody (1:1000 DyLight 594 goat antirabbit; Jackson ImmunoResearch, Suffolk, UK) diluted in PBS with 0.5% Triton X-100 and 1% normal goat serum and 1% fetal bovine serum for 2 h at room temperature. Slices were then washed in PBS then mounted on to gelled slides and cover-slipped with hard mount Vectashield (Vector Labs, Peterborough, UK) and imaged with an Olympus BX41 microscope with Olympus UC30 camera and filters for appropriate excitation and emission wavelengths (Olympus Medical, Southend-on-Sea, UK).

Incubation of EGTA-AM or BAPTA-AM
Coronal sections (300 μm) containing both CPu and NAc were bisected. Each hemisphere was incubated for 30 min in aCSF at room temperature (either 2.4 or 3.6 mM [Ca 2+ ]) containing 2-hydroypropyl-β-cyctodextrin (70 μM), probenecid (175 μM), pluronic acid (0.1%) and for EGTA-AM or BAPTA-AM conditions also contain EGTA-AM or BAPTA-AM (100 μM) (Ouanounou et al. 1999;Kukley et al. 2007). Following incubation, hemispheres were placed in the recording chamber with the recording electrode for 30 min before recordings were taken. Equivalent anatomical recording sites were chosen in each of the hemispheres, alternating between the control slice and EGTA-AM/BAPTA-AM-incubated slice: whether the control or EGTA/BAPTA slice was started with was alternated between animals. DA release data in EGTA-AM or BAPTA-AM were expressed as the percentage of paired control recording site.
For T-type inhibition, NNC 55-0396 was chosen in preference to mibefradil or Ni 2+ because of confounding effects: mibefradil obscures electrode sensitivity to DA (Chen et al. 2006) and Ni 2+ inhibits DA uptake (Brimblecombe & Cragg, 2014). NNC 55-0396 has fewer off-target effects than mibefradil (Li et al. 2005), and we show that its effects vary predictably with Ca 2+ concentration (see Fig. 7) corroborating its VGCC site of action. We also trialled the R-type blocker SNX 482 (100 nM), which was found to be electroactive, to reduce electrode sensitivity to DA and to lead to a paradoxical increase in evoked [DA] o in CPu and NAc (all stimulation frequencies, data not shown). As there are no alternative R-selective compounds commercially available to corroborate these paradoxical effects, it is difficult to interpret these data and they were excluded from further consideration here.

Data and statistical analysis
Data were acquired and analysed using Axoscope 10.2 (Molecular Devices, Berkshire, UK), Whole Cell Program (University of Strathclyde, Glasgow) and Excel macros written locally. Data are expressed as means ± SEM, and n = number of animals (for data in Figs 1-7) or number of sites per region (for data in Fig. 8). Data from each animal were obtained by averaging at least three recordings for each stimulus type and normalizing to mean control 1p conditions for each animal. Data presented are normalized for greater clarity and for more ready comparisons between regions. Note that the outcomes reported here in each region were the same using raw data.
Population means were compared using one-or two-way ANOVA with post-hoc Bonferroni's t test where appropriate using GraphPad Prism. Curve fits and EC 50 analysis were done using GraphPad Prism without constraints; EC 50 values were compared using an F test. Raw control data passed Shapiro-Wilk normality test.

N-type voltage-gated Ca 2+ channels have a major role in dopamine transmission that is frequency-dependent
We first explored the VGCC dependence of striatal DA release, evoked electrically by single pulses (1p) and short trains of five pulses (5p) at frequencies ranging from 5 to 100 Hz, in the presence of 2.4 mM extracellular Ca 2+ . The nAChR antagonist DHβE (1 μM) was present throughout to eliminate the confounding effects of the VGCCs on striatal ACh. We chose electrical stimulation (with nAChR block) in preference to optogenetic activation, as the Ca 2+ conductance associated with ChR2 might interfere with endogenous Ca 2+ -dependent processes. Peak-evoked [DA] o varied with stimulus frequency in CPu ( Fig. 1A and B; one-way ANOVA, F 4,26 = 62.7, P < 0.001) and NAc ( Fig. 1C and D; one-way ANOVA, F 4,14 = 36.3, P < 0.001) as shown previously Exley & Cragg, 2008;Exley et al. 2012). Single pulses evoked a mean peak [DA] o of 0.97 ± 0.10 μM in CPu and 0.67 ± 0.10 μM in NAc, in line with previous studies (Threlfell et al. 2010;Hartung et al. 2011).

P/Q-type voltage-gated Ca 2+ channels participate in dopamine transmission, with frequency dependence in caudate putamen
To block P-type channels we applied a low concentration of ω-Agatoxin IVA (15 nM), which had no effect on evoked [DA] o in either CPu or NAc (data not illustrated). A higher concentration of ω-Agatoxin IVA (200 nM) was subsequently used to block P + Q-type channels and significantly reduced peak evoked [DA] o in the CPu ( Fig.  2A and B; two-way ANOVA, effect of drug F 1,40 = 93.7, P < 0.001) and NAc ( Fig. 2C and D; two-way ANOVA, effect of drug F 1,40 = 24.1, P < 0.001). Evoked [DA] o was reduced by P+Q-block in CPu to ß30-50% of control, depending inversely on stimulation frequency ( Fig. 2A and B; two-way ANOVA, drug × frequency interaction F 4,40 = 3.8, P < 0.05) but in NAc was only reduced to ß70-80% of control and did not vary significantly with stimulation frequency (Fig. 2C and D; two-way ANOVA, drug × frequency interaction F 4,40 = 0.9, P > 0.05). The effect of P + Q channel blockade on [DA] o was significantly greater in CPu than NAc (Fig. 2E, two-way ANOVA, region F 1,40 = 29.3, P < 0.001).

T-type voltage-gated Ca 2+ channels control dopamine release in the caudate putamen
Application of the T-type VGCC blocker NNC 55-0396 (1 μM) (Tai et al. 2011)  and B; two-way ANOVA, drug × frequency interaction F 4,20 = 2.1, P > 0.05). These effects in the CPu were significantly different from those in the NAc ( Fig. 3E; two-way ANOVA, region F 1,20 = 37.9, P < 0.001). Higher drug concentrations (10 μM, NNC 55-0396) had no further effect in either region (not illustrated). The effect of T-block was unaffected by the presence of GABA and glutamate blockers (data not illustrated) confirming that VGCC effects were not via actions of GABA/Glu. Ni 2+ or mibefradil are used commonly to block T-type channels, but their use is precluded here: we have recently identified that Ni 2+ inhibits DA uptake (Brimblecombe & Cragg, 2014) while mibefradil renders electrodes insensitive to DA (Chen et al. 2006).

L-type voltage-gated Ca 2+ channels regulate dopamine release in the caudate putamen
In the CPu, blockade of L-type Ca 2+ channels with isradipine 5 μM (or nifedipine 10 μM, data not shown) modestly but significantly reduced electrically evoked [DA] o to ß75% of controls in the CPu for all stimulation frequencies ( Fig. 4A and B; two-way ANOVA, F 1,30 = 20.38, P < 0.001). These effects were reversible on drug washout (Fig. 4C). By contrast, in the NAc, isradipine did not modify evoked [DA] o under these conditions ( Fig. 4D and E; two-way ANOVA, F 1,20 = 1.18) (but see Fig. 7). These different outcomes of L-block in the CPu and NAc were statistically different ( Fig. 4F; two-way ANOVA, region F 1,20 = 206.7, P < 0.001). Higher isradipine concentrations (10 μM) did not unmask an effect in the NAc (not illustrated). Non-specific effects of isradipine were minimal in these experiments: The lack of effect of isradipine in the NAc suggest that any non-specific effects on other channels are negligible, and to confirm specificity, lower concentrations of isradipine (500 nM) in the CPu also significantly reduced 1p DA release, however with a protracted wash-on time (n = 3, t 4 = 3.9, P < 0.05, not illustrated). The effect of the L-block was unaffected by the presence of GABA and glutamate blockers (data not illustrated) confirming that VGCC effects were not via actions of GABA/Glu. To confirm that the L-type channels that regulate DA release in the CPu are located to DA axons rather than to other neuromodulatory inputs driven by the electrical field stimulus, we used an optogenetic approach to drive DA axons selectively. ChR2-eYFP was expressed by DA axons after injection of a Cre-inducible floxed construct into the SNc of DAT-Cre mice (Fig. 4G), as previously described (Threlfell et al. 2012). In drug-free control conditions, pulses of blue light (2 ms, one or five pulses at 5 or 25 Hz) evoked [DA] o of ß0.75-1.5 μM, which varied slightly with frequency ( Fig. 4H and I, effect of frequency, P < 0.01) as shown previously (Threlfell et al. 2012). Light-evoked [DA] o was not modulated by tonic ACh as the nAChR antagonist DHβE (1 μM) had no effect (data A CPu NAc (D) in control conditions (filled) or P/Q block (unfilled). Bonferroni post-tests vs. control: * P < 0.05, * * P < 0.01, * * * P < 0.001. E, mean peak [DA] o ± SEM after voltage-gated Ca 2+ channel block expressed as percentage of control at each frequency in CPu (circles) and NAc (squares) shows variable effect of block with frequency in CPu and a greater effect in CPu than NAc. Two-way ANOVA, CPu vs. NAc: * * * P < 0.001, n = 5 animals. ω-ATX IVA, ω-Agatoxin IVA; CPu, caudate putamen; DA, dopamine; NAc, nucleus accumbens. not illustrated). Isradipine (5 μM) significantly reduced light-evoked [DA] o to ß75% of control, invariantly with stimulus frequency (Fig. 4H and I; two-way ANOVA effect of drug, F 1,6 = 23.6, P < 0.01) as seen for electrical stimuli (Fig. 4I). These data corroborate a role in DA release for L-type VGCCs located on DA axons. Do voltage-gated Ca 2+ channel subtypes operate preferentially at specific frequencies?
Previous work suggested that the role of particular types of VGCCs in DA transmission might by differently recruited with frequency (Phillips & Stamford, 2000) and differently tune DA transmission during phasic or tonic modes of DA neuron activity. That study, however, did not account for the effects of the VGCC blockers on ACh release, which acts at nAChRs to modify DA release in a frequency-dependent manner (Zhou et al. 2001;Zhang & Sulzer, 2004;Threlfell et al. 2012). Our study excluded the effects of ACh by including nAChR antagonists, and none the less, detected variation in the effects of some VGCC blockers with stimulus frequency (e.g. N-and P/Q-blockers; Figs 1 and 2): The dynamic role of P/Q channels was significant in the CPu but not NAc, indicating it was not simply an intrinsic property of these channels, e.g. time constants for activation/inactivation. This dynamic property was detectable only for blockers that caused a large (>50%) suppression of DA release evoked by single pulses. At many synapse types, reductions in initial release probability is expected to relieve short-term depression of release and, at sufficiently short interpulse intervals, enhance short-term facilitation (STF) (Katz & Miledi, 1968;Schneggenburger & Neher, 2005). We hypothesized that the apparent frequency-specific functions of some VGCCs might then arise from the inverse relationship between initial DA release probability (P R ) and its dynamic STP during trains, that depends on Ca 2+ availability and interpulse intervals at DA synapses (Cragg, 2003; as at non-DA synapses. To explore whether frequency-specific effects of VGCC blockers were due to classical Ca 2+ -dependent changes to initial P R and STP, we tested the following three predictions that: (i) activity-specific outcomes of specific VGCC blockers should be mimicked by a sufficient reduction in extracellular Ca 2+ ; (ii) there should be an overall inverse relationship across the data sets as a whole between STP and initial P R ; and therefore (iii) a channel that does not ordinarily show activity-specific modulation of DA release should do so if P R is already sufficiently low to have modified STP. We explored each prediction in turn.
First, we compared the effect of N-channel blockade to that of a decrease in extracellular Ca 2+ . We lowered extracellular Ca 2+ from 2.4 mM to 1.0 mM to reduce peak [DA] o to levels comparable with that seen previously with the N-block (ß15-30% of control  N-block, did so in a manner that varied inversely with frequency in the CPu (Fig. 5A and B; two-way ANOVA, effect of treatment F 1,20 = 1.16, P > 0.05; effect of frequency F 4,20 = 10.2, P < 0.001) or NAc ( Fig. 5C and D; two-way ANOVA, effect of treatment F 1,20 = 3.75, P > 0.05; effect of frequency F 4,20 = 7.12, P < 0.01).
Secondly, we explored whether for the data set as a whole, there was a relationship between release evoked by 1p and its STP. We determined the ratio of [DA] o evoked by a burst (5p at 100 Hz) compared to a single pulse (the 5p:1p ratio) as a measure of STP, and plotted this ratio against [DA] o evoked by 1p, for all data sets combined that  from the relationships between Ca 2+ entry, DA P R and Ca 2+ -limited STF. Thirdly, we tested whether the L-type channel blocker isradipine, that modified DA release in the CPu in a similar manner across frequencies (no change in 5p:1p ratio; see Figs 4 and 5E), could be made to become activity-specific. We explored whether the previous reduction of DA P R with the N-blocker, to a point where any further reductions in P R should dramatically increase 5p:1p ratio, would result in isradipine increasing the 5p:1p ratio. Indeed, in the CPu, in the presence of the N-type blocker (ω-Conotoxin GIVA 100 nM), when 1p [DA] o was low and the 5p:1p ratio was >10 (Fig. 5E and F), the subsequent application of the L-blocker isradipine (5 μM) further decreased [DA] o evoked by 1p (by ß50%) but did not reduce [DA] o evoked by 5p/100 Hz, resulting in a significant increase in 5p:1p ratio to ß17 (Fig. 5F; linear regression R 2 = 0.79, P < 0.05). Therefore, we show here that the dynamic roles for subtypes of VGCCs in gating DA release are due to classical relationships between release probability and its STP.

Different voltage-gated Ca 2+ channel roles in the caudate putamen and nucleus accumbens are due to different Ca 2+ dependence
Across experiments, we noted that VGCC blockers caused more pronounced reductions in DA release in the CPu than the NAc. To date, there is no evidence that expression levels of VGCC subtypes differ correspondingly between DA neurons. To understand why the control of DA release was different, we explored whether DA release in the NAc was less sensitive than in the CPu to entry of extracellular Ca 2+ . Ca 2+ entry was required for DA release in both regions: brief application of zero extracellular Ca 2+ solution (with 100 μM EGTA) abolished DA release in the NAc or CPu to below detection levels, for trains and single pulses (Fig. 6A). However, by varying [Ca 2+ ] o we identified that 1p [DA] o varied with a steeper concentration-response in the CPu, with a lower EC 50 (CPu: 1.9 mM; NAc 2.9 mM) and larger Hill coefficient (CPu: 3.7 ± 0.7; NAc: 2.0 ± 0.9) than the NAc ( Fig.  6B; R 2 : CPu = 0.93, NAc = 0.87; comparison of fits: F 2,76 = 3.2, P < 0.05). These responses suggest there could be either greater co-operativity or overlap of Ca 2+ domains (Schneggenburger & Neher, 2005;Schneggenburger et al. 2012), or enhanced Ca 2+ entry or a reduced Ca 2+ buffering capacity in the CPu. They confirm also that DA release is more sensitive to reductions in [Ca 2+ ] o below 2.4 mM in the CPu than NAc (see Fig. 6B), consistent with the greater effects on DA release in CPu of limiting Ca 2+ entry with a wider range of VGCC blockers (Figs 1-4). We followed up two key implications of these observations.  (Fig. 7A and B) Fig. 7I; two-way ANOVA, effect of [Ca 2+ ] o F 1,40 = 1.44, P < 0.05). Therefore, the greater roles for T-and L-VGCCs in the CPu versus NAc is not due to an absence of these VGCCs at DA release sites in NAc terminals. These channels are present and can operate in the CPu and the NAc, if/when appropriate local Ca 2+ conditions are met. These data also corroborate the hypothesis that local handling of Ca 2+ varies between the CPu and NAc.

Endogenous buffering of how Ca 2+ couples to dopamine release differs in the caudate putamen and nucleus accumbens
The greater Ca 2+ sensitivity of DA transmission in the CPu compared to NAc, and greater roles for more VGCC types in the CPu, might arise from underlying mechanisms that could include enhanced co-operativity or 'domain overlap' of the intracellular Ca 2+ micro-/nanodomains of VGCCs, enhanced sensitivity of the exocytotic machinery and/or reduced intracellular buffering, which might govern spatial and temporal Ca 2+ summation (Blatow et al. 2003;Zhao et al. 2011;Eggermann et al. 2012;Hoppa et al. 2012;Pan & Ryan, 2012;Schneggenburger et al. 2012). The SNc and VTA DA neurons differ in expression of endogenous Ca 2+ buffers. Whereas the SNc DA neurons express higher levels of the buffer parvalbumin (Greene et al. 2005), a slow buffer (Schwaller et al. 2002), VTA DA neurons express higher levels of the buffer calbindin-D28k (Haber et al. 1995;Greene et al. 2005;Chung et al. 2005), a fast, high-affinity buffer (Nägerl et al. 2000;Schwaller et al. 2002;Eggermann et al. 2012). To explore if differences in Ca 2+ dependence of DA transmission in the CPu versus NAc were due to detectable differences in Ca 2+ handling we used the cell-permeable exogenous Ca 2+ chelators EGTA-AM (100 μM) and BAPTA-AM (100 μM), to identify the roles of slow and fast Ca 2+ transients respectively. These exogenous buffers have comparable affinity values but differ by a factor of ß40 in their buffering rates (reviewed by Eggermann et al. 2012). In 2.4 mM [Ca 2+ ] o , following brief incubation with EGTA-AM, evoked [DA] o was decreased to a similar extent in CPu and NAc ( Fig. 8A and B) indicating a role for slow/remote coupling between Ca 2+ source and sensor in both axon types. However, following brief incubation with BAPTA-AM, evoked [DA] o was decreased to a greater extent in the CPu than NAc, indicating greater coupling between fast/local Ca 2+ transients and DA release in the CPu (Fig. 8B; Bonferroni post-test BAPTA P < 0.05).
To explore whether the lesser effect of BAPTA-AM in the NAc results from buffering by fast endogenous buffers in the NAc, we increased [Ca 2+ ] o from 2.4 mM to 3.6 mM to saturate and overwhelm endogenous buffers, which should be evident through a preferential increase in effect of BAPTA-AM compared to EGTA-AM. First in the CPu,  8E; CPu post-test P > 0.05). By contrast, in the NAc at 3.6 mM [Ca 2+ ] o , there was a markedly enhanced effect of BAPTA-AM ( Fig. 8C and D), and a resulting significant increase in the ratio of [DA] o remaining in the presence of EGTA-AM vs. BAPTA-AM indicating an increase in the relative role for fast BAPTA-buffered transients, and suggesting an underlying role for endogenous fast buffers (Fig. 8F). This ratio at 3.6 mM [Ca 2+ ] o did not then differ between the NAc and CPu (Fig. 8E; post-test NAc, 2.4 vs. 3.6 mM, P < 0.05; post-test 3.6 mM NAc vs. CPu, P > 0.05). The different dominance of VGCC subtypes and apparent greater role for fast endogenous buffer in the NAc versus CPu is illustrated in the schematic in Fig. 8F.

Discussion
Here, we reveal the VGCCs on mesostriatal DA axons that regulate DA transmission and the principles that govern their dynamic roles, revealed when we exclude the obscuring effects of VGCCs on ChIs. We identify major underlying differences in the intracellular Ca 2+ dynamics that couple to DA transmission in the CPu versus NAc.

Multiple voltage-gated Ca 2+ channel subtypes regulate dopamine transmission in the striatum
It is generally assumed that presynaptic Ca 2+ influx required for synaptic neurotransmission occurs typically via the N-subtypes (Ca v 2.2) and P/Q-subtypes (Ca v 2.1) of the VGCCs, which, through synprint motifs, associate closely with SNARE proteins (Kim & Catterall, 1997;Vance et al. 1999;Zamponi, 2003). Properties defined at classical synapses cannot be assumed to apply universally across all synapse types, particularly for DA release sites, which can differ from classic models of point-to-point synapses in structural, junctional and functional aspects Arbuthnott & Wickens, 2007;Rice et al. 2011). None the less, we found major roles for N/P/Q-channels in the control of striatal DA release. The efficacy of high but not low concentrations of ω-Agatoxin IVA suggests that the P/Q component arises from Q-channels. We show also, however, that T-and L-type VGCCs participate in DA release, more readily in the CPu than NAc.
T-and L-type VGCCs are expressed by DA neurons (Takada et al. 2001;Wolfart & Roeper, 2002 et al. 2013) and our findings are consistent with accumulating evidence in other neuron types that L-and T-channels can regulate exocytosis. L-type VGCCs are the primary VGCC involved in neurotransmitter release from ribbon-type active zones of primary sensory neurons (Brandt et al. 2003;Zanazzi & Matthews, 2009) and they operate alongside N/P/Q-VGCCs within central areas that include the hippocampus (Murakami et al. 2002;Tippens et al. 2008) and supraoptic nucleus (Bhaukaurally et al. 2005). They regulate presynaptic plasticity in the amygdala and primary cortical neurons (Fourcaudot et al. 2009;Subramanian & Morozov, 2011). Their roles in action potential-dependent transmission are probably underestimated, however: neuronal L-type channels can be as fast and as efficient as the N-type channels at supporting calcium entry during action potential-like stimuli but they are inhibited only slowly by the dihydropyridine blockers usually applied to probe their functions (Helton et al. 2005). Previous studies of DA transmission or intracellular Ca 2+ that did not resolve the L-channel activity on DA axons (Phillips & Stamford, 2000;Chen et al. 2006;Sgobio et al. 2014) will have been confounded by the overriding effects of co-stimulating the ChI input to DA axons, and by the limited ability to resolve small changes in Ca 2+ that none the less translate to detectable changes in neurotransmission through a steep power relationship. T-type VGCCs are typically thought to activate in response to subthreshold membrane depolarizations and regulate membrane excitability, oscillatory activity and rebound firing. Mounting evidence suggests T-channels also participate in vesicular exocytosis from various neurons (Pan et al. 2001;Carbone et al. 2006;Tang et al. 2011;Weiss et al. 2012) and that like the classic 'high-voltage-activated' N/P/Q-channels, T-type channels also directly interact with synaptic proteins (syntaxin-1A and SNAP-25) (Weiss et al. 2012).

Dynamic changes in P R and short-term plasticity underlie frequency-specific roles of voltage-gated Ca 2+ channels
We showed that the VGCCs with greatest contributions to DA release had roles that depended inversely on frequency, suggesting that the VGCCs have a greater role in gating DA release by single or lower frequency activity than during phasic-like higher frequencies. We showed that this frequency-dependent role of a given VGCC could be explained, and predicted, by a classical underlying inverse relationship between initial DA P R and subsequent STF that depends on initial Ca 2+ entry. The greater the inhibition of release for a single pulse by a VGCC blocker, the greater the relief from short-term depression (STD) that limits re-release (Katz & Miledi, 1968;Cragg, 2003) and the greater the STF (5p:1p ratio >5). The effect of N-block on the DA varied more strongly with frequency than other VGCCs because the N-block caused the greatest suppression of initial DA P R and led to STF at sufficiently high frequencies. Blockers for channels that only slightly reduced DA P R (L/T-type channels) did not relieve the STD sufficiently to modify frequency dependence, but could do so if applied in combination, when the cumulative reduction in P R was sufficient to promote the STF. These data therefore indicate that VGCCs contribute to low-pass filtering of DA release and that their role is dynamic. Rather than having a fixed contribution to transmission, their roles are probably therefore to vary depending on the local driving forces that modulate P R , including recent history of synapse use. Different voltage-gated Ca 2+ channel function in the caudate putamen and nucleus accumbens is due to different Ca 2+ dependence and handling Differences in VGCC function between the NAc and CPu were observed that could not readily be attributed to variable VGCC expression or levels. Expression levels of the 10 potential α1 subunit genes that define VGCC subtype show little relevant difference between DA neurons (Grimm et al. 2004;Greene et al. 2005;Chung et al. 2005;Greene, 2006;Dryanovski et al. 2013). While one study reported that Ca v 3.1 (T-type) expression in SNc neurons is 1.2-fold that in VTA neurons (Greene et al. 2005), another indicated Ca v 3.3 (T-type) expression in VTA is twice that in the SNc (Grimm et al. 2004).
We showed that the different VGCC dependence of DA transmission in the CPu and NAc could be attributed to different underlying Ca 2+ sensitivity. The Ca 2+ concentration-response curve in CPu is leftward-shifted and steeper (higher co-operativity/Hill slope). Different Ca 2+ dependence of DA transmission in CPu and NAc prompted us to explore whether otherwise 'silent' Land T-VGCCs in NAc could be unmasked at higher [Ca 2+ ] o when the greater sensitivity to [Ca 2+ ] o makes small changes in the number of open Ca 2+ channels more effective yet in modulating transmitter release. These data reveal that L-and T-type VGCCs are not restricted to DA axons in the CPu but can be functional in NAc axons, under appropriate conditions. These conditions might be met physiologically in vivo, and dynamically. Such conditions might arise in vivo after previous synaptic activity or neuromodulatory inputs that promote Ca 2+ entry and shift DA release upwards on its Ca 2+ response curve, or through signalling pathways that modify phosphorylation states and conductance of the channels or their coupling to auxiliary subunits.
We explored whether differences between CPu and NAc were due to differences in how Ca 2+ dynamics couple to DA release, by exploring the effects of cell-permeable exogenous Ca 2+ chelators to chelate slow or fast Ca 2+ transients, using EGTA-AM or BAPTA-AM respectively. Our findings reveal critical differences in endogenous Ca 2+ buffering or signalling mechanisms. At 2.4 mM [Ca 2+ ] o , DA transmission in CPu and NAc was dependent on slow EGTA-sensitive Ca 2+ dynamics, suggesting that Ca 2+ source and Ca 2+ sensor can be remotely coupled within both axon types (Eggerman et al. 2012). Furthermore, DA transmission was coupled to an additional fast/local BAPTA-sensitive Ca 2+ source, more tightly in the CPu than in the NAc. However, when we provided more Ca 2+ , we enhanced the dependence of DA release on faster sources compared to slower Ca 2+ ones in the NAc but not the CPu. These data suggest that different Ca 2+ -dependent mechanisms operate in the NAc versus CPu, and that a fast mechanism normally limits the coupling of Ca 2+ entry to DA exocytosis in the NAc that can be exposed by overwhelming it with higher Ca 2+ . One explanation is that an additional fast endogenous buffer normally operates in NAc, and becomes overwhelmed by higher Ca 2+ . Calbindin-D28k, a fast, high-affinity and mobile buffer expressed by DA axons in the NAc but not CPu, is a possible candidate additional fast buffer. Calbindin is expressed in VTA DA neurons at ß2-3-fold higher levels than in the SNc (Haber et al. 1995;Greene et al. 2005;Chung et al. 2005). It can regulate Ca 2+ -dependent exocytosis from dissociated DA neurons (Pan & Ryan, 2012), and participate in spatial buffering in hippocampal neurons (Blatow et al. 2003;Müller et al. 2005), and because of its high Ca 2+ affinity, the Ca 2+ -binding sites of calbindin-D 28k are expected to become saturated after a sufficiently large Ca 2+ influx (Neher, 1998) in keeping with our experimental observations here. In summary, these data provide the first evidence that the spatiotemporal handling of Ca 2+ sources that couple to DA transmission differs significantly in DA axons in the CPu than in the NAc, and in particular, that buffering of fast Ca 2+ sources may be more limited in the CPu.
Ca 2+ dependence is not the only feature in which DA transmission from these two axonal populations differs. DA transmission in the CPu and NAc differs in release probability, STP and regulation by presynaptic neuromodulatory receptors (e.g. nAChRs, opioid receptors), synucleins and DA uptake (Cragg et al. 2000;Cragg, 2003;Britt & McGehee, 2008;Exley & Cragg, 2008;Anwar et al. 2011;Exley et al. 2012). We now show that this regional control of DA transmission extends to Ca 2+ coupling, which may even account for some of these properties.

Summary and further implications
In summary, we reveal key principles that underscore the regulation of striatal DA transmission by Ca 2+ and VGCCs. We show that multiple VGCC types in DA axons regulate DA release: N-and (P)Q-types dominate but L-and T-channels also participate. However, whether a given VGCC contributes to DA release is not fixed, but varies dynamically depending on neuron activity and factors that govern local Ca 2+ and DA P R . VGCC roles appear to vary with frequency because they contribute to low-pass filtering of DA release through a classic relationship between DA P R and its STP. A wider repertoire of VGCC types more readily regulate DA transmission in the CPu than in NAc due to an underlying greater Ca 2+ sensitivity in the CPu. Furthermore, we reveal the striking finding that these different axon types have different a spatiotemporal coupling of Ca 2+ dynamics to DA exocytosis, with endogenous fast Ca 2+ buffers apparently limiting DA transmission in the NAc more than in the CPu. These findings reveal that different Ca 2+ -handling mechanisms control DA function in limbic-versus motor-associated striatum.
As the axonal fields of DA neurons form the vast majority (ß99%) of the total membrane area of a DA neuron (Matsuda et al. 2009;Henny et al. 2012), their Ca 2+ handling might contribute to the differential susceptibility of different populations of DA neurons to degeneration in Parkinson's disease. L-channel antagonists are currently in clinical trial as neuroprotective therapies for Parkinson's disease (Simuni et al. 2010;Parkinson Study Group, 2013), and their actions on axonal L-channels might therefore contribute to neuroprotection but also modulate DA transmission.