|Home | About | Journals | Submit | Contact Us | Français|
Bursts of spikes triggered by sensory stimuli in midbrain dopamine neurons evoke phasic release of dopamine in target brain areas, driving reward-based reinforcement learning and goal-directed behavior. NMDA-type glutamate receptors (NMDARs) play a critical role in the generation of these bursts. Here we report LTP of NMDAR-mediated excitatory transmission onto dopamine neurons in the substantia nigra. Induction of LTP requires burst-evoked Ca2+ signals amplified by preceding metabotropic neurotransmitter inputs in addition to the activation of NMDARs themselves. PKA activity gates LTP induction by regulating the magnitude of Ca2+ signal amplification. This novel form of plasticity is associative, input specific, reversible, and depends on the relative timing of synaptic input and postsynaptic bursting in a manner analogous to the timing rule for cue-reward learning paradigms in behaving animals. NMDAR plasticity may thus represent a potential neural substrate for conditioned dopamine neuron burst responses to environmental stimuli acquired during reward-based learning.
The appropriate association between environmental cues and motivational valence is crucial for the brain to accurately guide behavior. Dopamine (DA) neurons, located in the substantia nigra pars compacta (SNc) and ventral tegmental area (VTA), are thought to assign positive values to objects and experiences in order to effectively influence decision-making strategies (Montague et al., 2004). In vivo experiments in non-human primates and rodents coupled with human functional imaging and computational modeling studies have suggested that this occurs through changes in DA neuron firing rate, which encode reward prediction errors (D’Ardenne et al., 2008; Montague et al., 1996; Pan et al., 2005; Schultz, 1998). As such, DA neurons transition from tonic single-spike firing (1–5 Hz) to burst firing (2–10 spikes at 10–50 Hz) in response to the unexpected presentation of primary rewards. Intriguingly, the burst response shifts in time to reward-predicting cues after conditioning with repeated cue-reward pairing. However, the locus of neural plasticity responsible for this conditioned DA neuron response remains elusive.
Glutamatergic inputs activating NMDA receptors (NMDARs) have been shown to drive the transition from slow, tonic firing to burst firing in DA neurons (Chergui et al., 1994; Morikawa et al., 2003; Overton and Clark, 1997; Tong et al., 1996; Zweifel et al., 2009), although AMPA receptors (AMPARs) may also play a role (Blythe et al., 2007). Therefore, potentiation of NMDAR-dependent excitation of DA neurons may contribute to the development of the conditioned burst response. Despite numerous studies describing the plasticity of AMPARs in DA neurons (Jones and Bonci, 2005; Kauer and Malenka, 2007), synaptic activity-dependent plasticity of NMDAR-mediated transmission has yet to be demonstrated [but see Borgland et al. (2006), Schilstrom et al. (2006), and Ungless et al. (2003) for enhancement of NMDAR function caused by metabotropic receptor agonists].
Ca2+ signaling, triggered by either postsynaptic action potentials (APs) or local synaptic events, is implicated in the plasticity of synapses throughout the CNS (Linden, 1999; Sjostrom and Nelson, 2002). We have previously shown that AP-evoked Ca2+ signals can be amplified by the activation of metabotropic glutamate receptors (mGluRs) and other neurotransmitter receptors coupled to phosphoinositide (PI) hydrolysis in DA neurons (Cui et al., 2007). This amplification results from an elevation in cytosolic inositol trisphosphate (IP3) levels, leading to enhanced Ca2+-induced Ca2+ release (CICR) through IP3 receptors (IP3Rs) located on intracellular Ca2+ stores. IP3, generated by activation of PI-coupled neurotransmitter receptors, and Ca2+, provided by AP-induced influx, thus synergistically coactivate IP3Rs (Taylor and Laude, 2002). In this study, we asked if this synergistic Ca2+ signaling could drive plasticity of NMDAR-mediated transmission onto DA neurons. We found that repeated pairing of sustained synaptic stimulation with burst firing results in long-term potentiation (LTP) of NMDAR EPSCs. The induction of LTP requires PI-coupled receptor-mediated facilitation of burst-induced Ca2+ signals and NMDAR activation. LTP induction is also gated by protein kinase A (PKA), which regulates IP3R sensitivity. We further show that NMDAR LTP is input specific, requires appropriately timed presynaptic and postsynaptic activity, and can be reversed by repetitive presynaptic stimulation without postsynaptic firing.
Whole-cell voltage-clamp recordings were made from DA neurons in the SNc (~90%) and VTA (~10%) using rat midbrain slices. Previous studies examining the conditioning of DA neuron responses in behaving animals, in which larger number of neurons were sampled from the SNc than the VTA, have reported similar response profiles in these two areas (Mirenowicz and Schultz, 1996; Pan et al., 2005; Schultz, 1998). A bipolar stimulating electrode was placed 50–150 μm rostral to the recorded neuron. Pharmacologically isolated NMDAR EPSCs were recorded at −62 mV in low Mg2+ (0.1 mM) to remove blockade of NMDARs. After 10 min of baseline EPSC recording, we delivered an LTP induction protocol consisting of a train of presynaptic stimulation (70 stimuli at 50 Hz) paired with a burst of 5 postsynaptic unclamped APs at 20 Hz, which mimics burst firing observed in behaving rats (Hyland et al., 2002). The onset of the burst was delayed by 1 s from that of the synaptic stimulation train. We found that repetitive synaptic stimulation-burst pairing (10 times every 20 s) resulted in LTP of NMDAR EPSCs in some but not all neurons tested (Figure 1A). The pattern of synaptic stimulation used in the induction protocol can augment AP-induced Ca2+ signals via activation of PI-coupled receptors, mainly mGluR1 (Cui et al., 2007). To address the role of AP-evoked Ca2+ signals in LTP induction, we measured small-conductance Ca2+-sensitive K+ (SK) currents (IK(Ca)) activated by unclamped APs (see Experimental Procedures). Immediately before induction, we tested each neuron for facilitation of IK(Ca) following synaptic stimulation by evoking a single AP at 60 ms after the offset of a 1-s stimulation train (example traces shown in inset of Figure 1B). The magnitude of NMDAR LTP, determined 30–40 min after the induction, was positively correlated with that of IK(Ca) facilitation (n = 31, r2 = 0.80) (Figure 1B). However, NMDAR LTP was not correlated with the size of IK(Ca) itself or with baseline EPSC amplitude (Figure S1). On average, NMDAR EPSCs were potentiated by 43% ± 6% in 21 neurons that exhibited IK(Ca) facilitation >15%, whereas no LTP was observed when IK(Ca) facilitation was <15% (1% ± 2% change, n = 10) (Figure 1C). The paired-pulse ratio (PPR, 50-ms interstimulus interval, expressed as EPSC2/EPSC1) and the coefficient of variation (CV, expressed as 1/CV2) of EPSCs were not significantly changed in 21 neurons that exhibited LTP (Figure 1D), suggesting a postsynaptic locus of LTP expression (Malinow and Tsien, 1990; Zalutsky and Nicoll, 1990). Repeated delivery of postsynaptic burst firing alone failed to induce LTP of NMDAR EPSCs (0% ± 3% change, n = 6), while synaptic stimulation alone produced a small but significant LTD (−8% ± 4% change, n = 6) (Figure 1E). Furthermore, LTP was not observed when the burst was replaced with a single AP during pairing (−1% ± 10% change, n = 5). Together, these results suggest that synaptic facilitation of burst-induced Ca2+ signaling is involved in the induction of NMDAR LTP. Due to the correlation between IK(Ca) facilitation and LTP in our initial finding, subsequent experiments were conducted in neurons that exhibited >15% IK(Ca) facilitation unless otherwise stated (see Table S1).
To confirm that this form of plasticity could be induced in physiological Mg2+, we recorded NMDAR EPSCs in 1.2 mM Mg2+ at slightly depolarized holding potentials (−47 to −62 mV) using Cs+-based internal solution to enhance the resolution of small NMDAR EPSCs (23 ± 3 pA, n = 8). Measurable IK(Ca) was not observed in these experiments, most likely due to the low permeability of SK channels to Cs+ (Shin et al., 2005). Pairing presynaptic stimulation with postsynaptic bursting produced LTP >10% in 6 out of 8 neurons tested in physiological Mg2+ (23%± 4% change, n = 6) (Figure S2).
We also examined the effect of the burst pairing protocol on AMPAR-mediated transmission. Here, AMPAR EPSCs were recorded at −62 or −77 mV in 1.2 mM Mg2+ with NMDARs intact, while synaptic stimulation-burst pairing was delivered at −62 mV. This resulted in LTD of EPSCs (−29% ± 3% change, n =5) (Figure S3). The magnitude of LTD showed no correlation with IK(Ca) facilitation by synaptic stimulation (r2 = 0.0003) (Figure S3C). Furthermore, there was no difference (p > 0.5) in the amount of LTD expressed when postsynaptic burst firing was omitted and neurons received the synaptic stimulation train alone (−26% ± 3% change, n = 3) (Figures S3B and S3C). It should be noted that the intracellular machinery responsible for the induction of AMPAR LTP may be “washed-out” during whole-cell recordings in DA neurons (Bonci and Malenka, 1999).
Activation of PI-coupled receptors facilitates AP-evoked Ca2+ signals in DA neurons via an increase in IP3 levels, which enhances IP3R-dependent CICR from intracellular stores (Cui et al., 2007). We thus examined the role of this Ca2+ signaling cascade in NMDAR LTP. Treatment of slices with cyclopiazonic acid (CPA, 10 μM), which depletes intracellular Ca2+ stores (Seidler et al., 1989), eliminated the facilitation of IK(Ca) by synaptic stimulation (0% ± 2%, n = 6) as well as the induction of NMDAR LTP (2% ± 4% change, n = 6) (Figure 2). Pharmacological blockade of mGluR1 together with muscarinic acetylcholine and α1-adrenergic receptors, other major PI-coupled neurotransmitter receptors expressed in DA neurons (Fiorillo and Williams, 2000; Paladini and Williams, 2004), also abolished IK(Ca) facilitation (2% ± 3%, n = 6) and NMDAR LTP (−6% ± 4% change, n = 6) (Figures 2B and 2C). We further confirmed that intracellular BAPTA (100 μM) blocked IK(Ca) facilitation (4% ± 1%, n = 6) and NMDAR LTP (−1% ± 5% change, n = 6). Together with the data presented in Figures 1B and 1E, these results demonstrate that Ca2+ store-dependent enhancement of burst-induced Ca2+ signals is critical for LTP induction.
IP3 sensitivity of IP3Rs can be increased by PKA-mediated phosphorylation (Tang et al., 2003; Wagner et al., 2008). To examine the role of PKA, we loaded recorded neurons with the specific PKA inhibitor PKI (100–200 μM) through the whole-cell pipette. We first tested the effect of PKI on IP3 sensitivity of IP3Rs by performing flash photolysis of caged IP3 using different UV pulse intensities (expressed in μF; see Experimental Procedures) to vary the concentration of IP3 released and measured the resulting SK-mediated outward current (IIP3) (Figure 3A). Intracellular PKI significantly increased the UV pulse intensity producing half maximal IIP3 amplitude (138 ± 12 μF in control, n = 5 vs. 220 ± 37 μF in PKI, n = 7, p < 0.05) (Figure 3B), suggesting that IP3 sensitivity is enhanced by tonic PKA activity. Although PKA is not known to modulate SK channel function, recent evidence indicates that PKA phosphorylation can regulate surface expression of SK2 channels (Lin et al., 2008; Ren et al., 2006). However, PKI failed to alter the maximal IIP3 amplitude (data not shown). This may be due to the predominant expression of SK3 channels in DA neurons (Wolfart et al., 2001).
PKI also significantly reduced the magnitude of IK(Ca) facilitation caused by bath perfusion of the mGluR agonist DHPG (1 μM) (92% ± 21% in control, n = 13 vs. 27% ± 10% in PKI, n = 10, p < 0.01) (Figures 3C and 3D). In 5 PKI-loaded neurons that exhibited <20% IK(Ca) facilitation (9% ± 3%) in response to 1 μM DHPG, higher concentrations of DHPG (5–10 μM), which should further elevate cytosolic IP3 levels, produced significantly larger IK(Ca) facilitation (221% ± 51%, p < 0.05) (Figure 3E), consistent with the idea that PKI reduced the IP3 sensitivity of IP3Rs.
We next tested the effect of PKI on NMDAR LTP. Intracellular PKI suppressed IK(Ca) facilitation by synaptic stimulation (9% ± 3%, n = 7) as well as the induction of LTP (−3% ± 6% change, n = 7) (Figures 3F and 3G). In contrast, intracellular dialysis with the PKC inhibitor chelerythrine (10 μM), which has been shown to block NMDAR LTP in the hippocampus (Kwon and Castillo, 2008), had no significant effect on IK(Ca) facilitation or NMDAR LTP (Figures 3D, 3F, and 3G). Together, these data demonstrate that PKA activity regulates the induction of NMDAR LTP by augmenting PI-coupled receptor-mediated facilitation of Ca2+ signals.
DA neuron bursts are thought to provide a plasticity signal in projection areas via phasic DA release, thus driving reward-based learning (Schultz, 1998). DA neuron bursts may also trigger Ca2+-dependent dendritic release of DA in the SNc (Beckstead et al., 2004; Chen and Rice, 2001). Furthermore, activation of DA D1/5 receptors can produce potentiation of NMDAR EPSCs (Schilstrom et al., 2006), raising the possibility that DA may play a role in LTP induction. However, significant NMDAR LTP was observed (38% ± 9% change, n = 5) even when the DA D1/5 receptor antagonist SCH 23390 (1 μM) was present during induction (Figure S4). The DA D2 receptor antagonist eticlopride (100–200 nM) was always present in the extracellular solution in this study to block D2 receptor-mediated IPSCs (Beckstead et al., 2004). Thus, burst-induced DA release is not involved in the induction of NMDAR LTP in DA neurons.
Recent studies on the plasticity of NMDARs at hippocampal mossy fiber synapses indicate that activation of NMDARs, in addition to activation of mGluRs, is required during LTP induction (Kwon and Castillo, 2008; Rebola et al., 2008). In order to test this possibility in DA neurons, we acutely blocked NMDARs with the NMDAR antagonist DL-AP5 (50–100 μM) during induction (Figure 4). Perfusion of DL-AP5 after 10 min of baseline recording rapidly and completely blocked NMDAR EPSCs (from 44 ± 11 pA to 2 ± 1 pA, n = 4), and the burst pairing protocol was delivered thereafter. DL-AP5 was washed out immediately after induction. Despite robust facilitation in all neurons tested (56% ± 13%, n = 4), none exhibited LTP of NMDAR EPSCs (μ1% ± 2% change) (Figure 4B). We confirmed that the washout of DL-AP5 (100 μM) was complete in ~30 min when no burst pairing protocol was delivered (n = 3). Therefore, the induction of NMDAR LTP requires the activation of NMDARs themselves.
We next examined if LTP induction is dependent on the relative timing between synaptic stimulation and burst firing. In our routine induction protocol, there is a 1-s delay between the onset of the 1.4-s synaptic stimulation train and that of the burst. When this delay was omitted, i.e., when the onset of the burst was shifted forward to coincide with that of synaptic stimulation, no LTP was induced (−3% ± 10% change, n = 4) (Figure 5A). Similarly, no significant LTP was observed when the burst was elicited with a delay of 200 ms after the onset of synaptic stimulation (3% ± 5% change, n = 3). However, sizable LTP was induced when the burst occurred with a 500-ms delay during the pairing protocol (20% ± 6% change, n = 5), although reduced in magnitude compared to the LTP induced with a 1-s burst delay. In line with these observations, we found that the magnitude of IK(Ca) facilitation gradually increased during 1-s synaptic stimulation in these neurons tested for the burst-timing dependence of LTP induction (Figure 5B), most likely reflecting gradual increases in cytosolic IP3 levels. In a separate series of experiments, we performed fluorescence imaging of burst-induced Ca2+ signals using fluo-5F (50 μM) as a Ca2+ indicator and examined the burst-timing dependence of facilitation produced by a 1.4-s synaptic stimulation train. The magnitude of facilitation of burst-evoked fluorescence change also gradually increased as the delay between the onset of synaptic stimulation and that of the burst was prolonged up to 1 s in 6 neurons tested (Figures 5C and 5D). Synaptic stimulation increased burst-evoked fluorescence change by 35% ± 5% (n = 6) at 1-s delay. This increase was abolished by CPCCOEt (75 μM, n = 2), consistent with the role of mGluR1 in synaptic facilitation (Cui et al., 2007), but was unaffected by DL-AP5 (50–100 μM, n = 4) (Figure S5). No AP5-sensitive fluorescence change was observed with synaptic stimulation alone, indicating that NMDAR-mediated Ca2+ influx was not detected with our imaging system.
Next, we delayed the burst until after the offset of the synaptic stimulation train. This resulted in a significant decrease in LTP with an interval of 60 ms (13% ± 14% change, n = 4) and near complete lack of LTP with a 120-ms interval (6% ± 5% change, n = 5) (Figures 5A and 5E). In the 5 neurons in which LTP induction was attempted with a 120-ms interval, facilitation of IK(Ca) at 120 ms after the offset of synaptic stimulation was indistinguishable from that at 60 ms, the interval routinely used to assess IK(Ca) facilitation (Figure 5B). Furthermore, in Ca2+ imaging experiments, facilitation of burst-evoked fluorescence change was not significantly reduced when the burst was elicited at 120 ms after the offset of the 1.4-s synaptic stimulation train (Figures 5C and 5D), indicating that the decrease in LTP is not due to a reduction in synaptic facilitation of Ca2+ signaling. Indeed, IP3-mediated enhancement of Ca2+ signals has been shown to last for hundreds of milliseconds because of prolonged lifetime of IP3 binding to IP3Rs (Sarkisov and Wang, 2008). In contrast, NMDAR EPSCs evoked by synaptic stimulation decayed by 80% ± 4% at 60 ms after the offset of stimulation in the 4 neurons tested at 60-ms interval for LTP induction, while the decay of NMDAR EPSCs was almost complete (96% ± 1%) at 120 ms in the 5 neurons tested for 120-ms interval. This implies that the burst may need to occur while NMDARs are activated during induction. Therefore, the burst-timing dependence of LTP induction described here is consistent with the requirement of both PI-coupled receptor-mediated facilitation of burst-induced Ca2+ signals (Figure 2) and activation of NMDARs (Figure 4).
Finally, we evoked burst firing before the onset of synaptic stimulation during induction. Interestingly, sizable NMDAR LTD was observed when the onset of the burst was placed 250 ms before that of the synaptic stimulation train (−22% ± 7% change, n = 4) (Figure 5A). There was no significant change in either PPR or 1/CV2 (0.84 ± 0.07 vs. 0.86 ± 0.06 and 41 ± 12 vs. 39 ± 10, respectively; p > 0.5 for both parameters), suggesting a postsynaptic locus of LTD expression as for LTP. When the interval between burst onset and synaptic stimulation was increased to 500 ms, where burst-induced Ca2+ rise had minimal overlap, if any, with synaptic stimulation, the magnitude of LTD was reduced to a level comparable to that induced by presynaptic stimulation alone (500 ms before onset: −10% ± 4% change, n = 4 vs. synaptic stimulation alone: −8% ± 4% change, n = 6, p > 0.5). Together, these results demonstrate that the relative timing between presynaptic stimulation and postsynaptic burst firing determines the direction and the magnitude of NMDAR plasticity.
The involvement of NMDAR activation in the induction of NMDAR LTP raises the possibility that NMDARs may be potentiated specifically at those synapses stimulated during induction. To test this possibility, we placed two stimulating electrodes >100 μm apart from each other. After confirming the independence of the two pathways (see Experimental Procedures), we monitored NMDAR EPSCs in each pathway for 10 min. Once a stable baseline was established, one pathway received sustained synaptic stimulation paired with burst firing while the other pathway was held silent (Figure 6). This produced LTP selectively in the paired pathway (paired pathway: 65% ± 16% change vs. unpaired pathway: 2% ± 2% change, n = 4, p < 0.05), demonstrating that NMDAR LTP can be input specific.
It has been shown that bath application of orexin A or DA D1/5 receptor agonists produces long-lasting increases in NMDAR EPSCs via changes in the composition of NR2 subunits of NMDARs in DA neurons (Borgland et al., 2006; Schilstrom et al., 2006). An activity-dependent switch in NR2 subunit composition has also been reported at neonatal hippocampal synapses (Bellone and Nicoll, 2007). We therefore tested if burst-dependent LTP of NMDARs in DA neurons is also associated with a change in the subunit composition by comparing the effects of NMDAR subunit specific antagonists on control NMDAR EPSCs versus potentiated EPSCs after successful LTP induction. We used three different NR2 subtype-specific antagonists: Ro 25–6981 (1 μM) and ifenprofil (3 μM), NR2B-containing receptor antagonists, and Zn2+ (100 nM), an NR2A-containing receptor antagonist (Fischer et al., 1997; Paoletti et al., 1997; Williams, 1993). None of these antagonists showed differential effects on control versus potentiated NMDAR EPSCs (Figure 7), suggesting that the burst pairing protocol induces NMDAR LTP without a change in the subunit composition of NMDARs.
Synaptic plasticity induced by correlated presynaptic and postsynaptic activity can be reversed by presynaptic stimulation in the absence of postsynaptic activation (Bellone and Nicoll, 2007; Massey and Bashir, 2007). To examine if NMDAR LTP can be reversed (depotentiated) in DA neurons, we repeatedly delivered synaptic stimulation alone (10 times every 20 s) 30 min after inducing LTP of NMDAR EPSCs (30% ± 6% change, n = 4) (Figures 8A and 8B). This depotentiation protocol rapidly depressed previously potentiated NMDAR EPSCs back towards baseline levels in all 4 neurons tested (baseline: 59 ± 9 pA, LTP: 75 ± 10 pA, post-depotentiation: 57 ± 9 pA). Depotentiation was not associated with a change in either PPR or 1/CV2 (0.99 ± 0.10 vs. 1.00 ± 0.10 and 19 ± 4 vs. 17 ± 4, respectively; p > 0.5 for both parameters). It should be noted that the same procedure, i.e., delivery of synaptic stimulation alone, also induced a small but rapid LTD of control EPSCs that had not undergone LTP induction (−8% ± 4% change, n = 6) (Figure 1E).
We next inserted a single AP into the depotentiation protocol at 1 s after the onset of synaptic stimulation, i.e., at the same timing as the burst in the burst pairing protocol (Figures 8C and 8D). Surprisingly, paring synaptic stimulation with a single AP completely prevented depotentiation in 4 out of 4 neurons (baseline: 53 ± 3 pA, LTP: 71 ± 3 pA, post-single AP pairing: 71 ± 3 pA). The same protocol also produced no change in control NMDAR EPSCs (1% ± 8% change, n = 5) (Figure 1E). Thus, synaptic stimulation-single AP pairing had no effect on NMDAR EPSCs regardless of whether they had been previously potentiated or not. Together, these results demonstrate that NMDAR LTP can be reversed by repetitive synaptic stimulation in the absence of postsynaptic firing activity.
Here we have demonstrated that repetitive pairing of sustained synaptic stimulation with burst firing induces LTP of NMDAR EPSCs in midbrain DA neurons. The induction of NMDAR LTP requires 1) synaptic facilitation of burst-evoked Ca2+ signals via mGluRs and other PI-coupled receptors generating IP3, and 2) activation of NMDARs. The burst-timing dependence of LTP induction is consistent with these two requirements in that 1) the burst needs to occur with a certain delay (~0.5–1 s) after the onset of synaptic stimulation, reflecting the time required for synaptic stimulation to cause a rise in IP3 levels, and 2) the burst also has to take place prior to or immediately (within tens of milliseconds) after the termination of synaptic stimulation so that NMDARs are activated at the time of the burst. Intriguingly, LTD of NMDAR EPSCs is induced when the burst precedes synaptic stimulation during the induction protocol, although the exact cellular mechanisms underlying bidirectionality of NMDAR plasticity remain to be determined (Harney et al., 2006). The activity-dependent plasticity of NMDARs in this study represents a novel mechanism for long-term regulation of DA neuron output that may also integrate with other forms of synaptic plasticity dependent on NMDAR activation (Engblom et al., 2008; Nugent et al., 2007; Zweifel et al., 2008).
It is well established that Ca2+ signals triggered by postsynaptic APs play a critical role in the induction of synaptic plasticity (Linden, 1999; Sjostrom and Nelson, 2002). APs can propagate and trigger Ca2+ influx in DA neuron dendrites with high efficiency (Hausser et al., 1995; Wilson and Callaway, 2000). Interestingly, the induction of NMDAR LTP in this study requires burst firing, as pairing synaptic stimulation with a single AP was ineffective at driving plasticity. In addition, burst-induced Ca2+ signals need to be amplified by preceding activation of PI-coupled receptors, which recruits CICR via IP3Rs on intracellular stores, to effectively induce LTP. Why are Ca2+ transients resulting from burst-induced Ca2+ influx insufficient to drive plasticity by themselves? Perhaps the mechanism is similar to that described for localized Ca2+ signaling and LTD of AMPAR EPSCs at parallel fiber synapses on cerebellar Purkinje neurons (Sarkisov and Wang, 2008; Wang et al., 2000). Here, climbing fiber activation and subsequent dendritic Ca2+ spike generation do not evoke large enough Ca2+ transients in dendritic spines to reach the threshold for plasticity induction unless CICR is triggered by parallel fiber inputs activating mGluRs and producing local IP3 increases in spines. The main difference between NMDAR LTP in DA neurons and AMPAR LTD in Purkinje neurons is the involvement of NMDAR activation in the induction. At parallel fiber-Purkinje neuron synapses, which lack NMDARs, chemical compartmentalization offered by dendritic spines restricts IP3 and Ca2+ signaling to individual spines, thereby mediating synapse specificity of plasticity (Nimchinsky et al., 2002; Wang et al., 2000). However, such compartmentalization of IP3-dependent Ca2+ signaling may not be easily attained at glutamatergic synapses on DA neurons, which are mostly formed on dendritic shafts [(Carr and Sesack, 2000; Charara et al., 1996), but also see (Sarti et al., 2007)]. Indeed, synaptic activation of mGluRs augments burst-induced Ca2+ transients throughout individual dendrites in DA neurons (Cui et al., 2007). Therefore, the localized signal underlying the input specificity of NMDAR LTP is presumably provided by NMDARs causing Ca2+ influx only at activated synapses, which would be below the spatial resolution of the confocal imaging system used in the present study. In support of this idea, synaptic activation of ionotropic glutamate receptors (i.e, Ca2+-permeable AMPARs) has been shown to produce highly localized (~1 μm) Ca2+ transients in aspiny dendrites mediating input-specific Ca2+ signaling and plasticity (Goldberg et al., 2003; Soler-Llavina and Sabatini, 2006). The requirement for coactivation of NMDARs and mGluRs, together with the dependence on intracellular Ca2+ stores, is in line with recent studies demonstrating input-specific LTP of NMDAR EPSCs at hippocampal mossy fiber synapses (Kwon and Castillo, 2008; Rebola et al., 2008). It should also be noted that Ca2+ transients resulting from NMDAR-induced Ca2+ influx can be amplified via an mGluR- and IP3-dependent CICR mechanism at Schaffer collateral synapses on hippocampal CA1 pyramidal neurons (Dudman et al., 2007).
A number of studies have reported LTP of NMDAR-mediated transmission in the hippocampus (Bashir et al., 1991; Bellone and Nicoll, 2007; Harney et al., 2008; Kwon and Castillo, 2008; Rebola et al., 2008), yet none of these studies have addressed the role of postsynaptic APs in LTP induction. A delayed NMDAR LTP has been observed in cortical pyramidal neurons, which is induced by simultaneous presynaptic and postsynaptic burst firing and is dependent on preceding AMPAR LTP (Watt et al., 2004). In the present study, NMDAR LTP required a delay between the onset of presynaptic stimulation and postsynaptic burst firing and was independent of AMPARs. Thus, NMDAR LTP in DA neurons represents a form of Hebbian plasticity of NMDAR-mediated transmission that has not been previously described.
Ample evidence indicates the important role of PKA in regulating different aspects of synaptic plasticity (Nguyen and Woo, 2003). In particular, PKA has been shown to gate the induction of AMPAR LTP by modulating CaMKII and SK2 channels in the hippocampus and amygdala (Blitzer et al., 1998; Faber et al., 2008). Our data show that PKA gates the induction of NMDAR LTP in DA neurons through enhancement of IP3R function. LTP induction may also be affected by PKA regulation of NMDAR-mediated Ca2+ influx (Skeberdis et al., 2006).
Burst-dependent potentiation of NMDARs appears to be expressed postsynaptically by a mechanism distinct from that previously described for metabotropic receptor-induced potentiation of NMDAR EPSCs in DA neurons (Borgland et al., 2006; Schilstrom et al., 2006; Ungless et al., 2003). For example, activation of orexin-1 receptors induces PKC-dependent translocation of NR2A-containing NMDARs to the synapse (Borgland et al., 2006). PKC-mediated recruitment of NMDARs has also been implicated in NMDAR LTP at hippocampal mossy fiber synapses (Kwon and Castillo, 2008). However, PKC blockade failed to affect NMDAR LTP in our study. Furthermore, the effects of NR2A- and NR2B-specific antagonists on NMDAR EPSCs were not altered after LTP expression. Although we cannot rule out potential changes in NR2C/2D subunits (Harney et al., 2008), these subunits make small contributions to NMDAR EPSCs in DA neurons (Borgland et al., 2006). Therefore, enhanced function of individual NMDAR channels and/or increased synaptic expression of existing NMDARs with no change in the subunit composition likely mediate the expression of LTP (Chen and Roche, 2007).
In behaving animals, DA neurons “learn” to respond to inherently neutral environmental cues with synchronized bursts of activity after repeated cue-reward pairing (Pan et al., 2005; Schultz, 1998). Several modeling studies have addressed the neurobiological substrates underlying the conditioning of DA neuron responses (Brown et al., 1999; Contreras-Vidal and Schultz, 1999; Houk et al., 1995). One of these models postulates that plasticity of synapses onto DA neurons is involved in this learning process (Contreras-Vidal and Schultz, 1999). It has also been shown in awake rats that excitatory responses of pedunculopontine tegmental nucleus neurons to auditory cues, which play an important role in driving DA neuron burst responses to those cues, remain unaltered during cue-reward learning (Pan and Hyland, 2005). Since the pedunculopontine tegmental nucleus gives rise to direct glutamatergic (and cholinergic) inputs to DA neurons (Charara et al., 1996), this raises the possibility that plasticity of glutamatergic synapses onto DA neurons may play a role in the development of conditioned burst responses. Therefore, in light of the prominent role of NMDARs in the generation of DA neuron bursts (Chergui et al., 1994; Morikawa et al., 2003; Overton and Clark, 1997; Tong et al., 1996; Zweifel et al., 2009), the activity-dependent plasticity of NMDARs described in this study may contribute to the acquisition of cue responses. It should be noted that the synaptic stimulation-burst pairing protocol emulates the neural activity evoked during the cue-reward pairing paradigm. Here, sustained synaptic stimulation mimics the working memory-type persistent input activated by the presentation of the cue (Brown et al., 1999; Funahashi et al., 1989), whereas the postsynaptic burst corresponds to that triggered by the reward during conditioning. In this model, potentiated NMDARs at those synapses activated by the cue, accompanied by certain termination mechanism(s) (e.g., SK channel activation), mediate the transient burst response to the cue after conditioning.
Of particular interest is the burst-timing dependence of the induction of NMDAR plasticity, which appears analogous to the timing rule governing cue-reward learning in behaving animals. In the standard and most effective training paradigm, termed delay conditioning, there is a delay of hundreds of milliseconds to several seconds between the onset of the cue and that of the reward, with the two stimuli overlapping in time (Fiorillo et al., 2003; Schwartz et al., 2002). For NMDAR LTP in DA neurons, the requirement of the delay (~0.5–1 s) and the overlap between synaptic stimulation and burst firing during induction most likely reflects the involvement of PI-coupled receptors and NMDARs, respectively. Furthermore, induction of LTD when the burst precedes synaptic stimulation during the burst pairing protocol is congruent with the ineffectiveness of backward conditioning in which the reward is presented prior to the cue (Schwartz et al., 2002). The timing rule described here is distinct from that for the spike timing-dependent plasticity reported in a variety of neurons (Dan and Poo, 2004; Sjostrom and Nelson, 2002), including DA neurons (Liu et al., 2005; Luu and Malenka, 2008), in which the plasticity is sensitive to the timing of presynaptic and postsynaptic spikes on a timescale of tens of milliseconds, much shorter than the timescales encountered during behavioral conditioning (Drew and Abbott, 2006).
It is of note that the same induction protocol that caused NMDAR LTP resulted in LTD of AMPAR EPSCs in this study. Since AMPAR LTD did not require postsynaptic bursting, it presumably corresponds to the mGluR-dependent but postsynaptic activity-independent AMPAR LTD mediated by a shift in the AMPAR subunit composition in DA neurons (Mameli et al., 2007). This LTD has been shown to reverse the persistent and global potentiation of AMPARs produced by cocaine administration paired with environmental cues, and thus may act to reset AMPAR-mediated transmission to enable AMPAR plasticity required for future learning (Bellone and Luscher, 2006; Chen et al., 2008; Stuber et al., 2008). Therefore, simultaneous NMDAR LTP and AMPAR LTD may work in concert to promote the learning of new environmental cues in animals previously conditioned with powerful reinforcers such as addictive drugs. However, it is important to point out that the exact, and perhaps differential, roles of NMDAR plasticity vs. AMPAR plasticity in vivo remain to be determined.
The neural mechanisms underlying behavioral learning are thought to involve both reversible and irreversible components (Medina et al., 2002; Pan et al., 2008). Our results show that NMDAR LTP can be reversed, or depotentiated, by repeated delivery of synaptic stimulation alone, which is reminiscent of the extinction of learned responses when the conditioning cue is repeatedly presented without the expected reward. It is remarkable that the expression of LTP is maintained when synaptic stimulation is repeatedly paired with a single AP, suggesting that single AP-evoked Ca2+ transients, facilitated by IP3-dependent CICR, can serve to prevent depotentiation. Therefore, a pause in tonic single-spike activity of DA neurons, as observed at the time of the expected reward when the learned cue is presented alone, may be necessary to induce extinction of phasic burst responses to the cue (Pan et al., 2008; Tobler et al., 2003).
Horizontal midbrain slices were prepared from male Sprague-Dawley rats (4–7 weeks old). Recordings were made at 34–35°C in a chamber perfused at ~2.5 ml/min with recording solution containing (in mM): 126 NaCl, 2.5 KCl, 1.2 NaH2PO4, 1.2 or 0.1 MgCl2, 2.4 CaCl2, 11 glucose, 21.4 NaHCO3, saturated with 95% O2/5% CO2 (pH 7.4, ~295 mOsm/kg). The pipette solution contained (in mM): 115 K-gluconate or K-methylsulfate, 20 KCl, 1.5 MgCl2, 10 HEPES, 0.025 EGTA, 2 Mg-ATP, 0.2 Na2-GTP, and 10 Na2-phosphocreatine (pH 7.25, 285 mOsm/kg).
Cells were visualized using an upright microscope with IR-DIC optics (Olympus). Whole-cell voltage-clamp recordings were made from electrophysiologically identified DA neurons at a holding potential of −62 mV, corrected for a liquid junction potential of −7 mV. Pipette resistance was 2.0–2.5 MΩ. Pipette capacitance was neutralized but series resistance was left uncompensated. Input resistance (typically ~250 MΩ) and holding current (typically 0 to −100 pA) were monitored continuously; experiments were discarded if they changed by more than 25% or 60 pA, respectively, or if series resistance increased above 16 MΩ. A Multiclamp 700B amplifier (Molecular Devices) was used to record the data, which were filtered at 2–10 kHz, digitized at 4–20 kHz, and collected using AxoGraph X (AxoGraph Scientific).
Synaptic stimuli were applied at 0.05 Hz using bipolar tungsten electrodes (100–120 μm tip separation) and pharmacologically isolated NMDAR EPSCs were monitored.
Immediately before LTP induction, the effect of sustained synaptic stimulation on IK(Ca), evoked by a single unclamped AP, was evaluated in each neuron. A 2-ms depolarizing pulse from −62 mV to −7 mV was used to elicit an unclamped AP. The integral of the outward tail current, i.e., IK(Ca), was calculated between 20 ms and 400–600 ms after the depolarizing pulse. We have shown previously that IK(Ca) thus measured is completely eliminated by TTX and also by apamin, a selective blocker of Ca2+-sensitive SK channels, and hence can be used as a readout of AP-induced Ca2+ transients (Cui et al., 2007). The magnitude of IK(Ca) facilitation by synaptic stimulation was calculated by comparing Ik(Ca) evoked 60 ms after a 1-s train of 50-Hz synaptic stimulation, after subtracting the trace elicited by synaptic stimulation alone, with Ik(Ca) evoked in isolation. A single AP, instead of a burst of APs, was used in evaluating synaptic facilitation of IK(Ca) in order to avoid potential influence on LTP induction.
The LTP induction protocol consisted of sustained synaptic stimulation (70 stimuli at 50 Hz) paired with a postsynaptic burst of 5 APs at 20 Hz, where the burst was delayed by 1 s from the onset of the synaptic stimulation. Synaptic stimulation was extended 200 ms beyond the end of the burst, i.e., until burst-evoked Ik(Ca) mostly decayed, to ensure that synapses were activated while cytosolic Ca2+ concentration was elevated. Synaptic stimulation-burst pairing was repeated 10 times every 20 s. The same stimulation intensity used for monitoring NMDAR EPSCs was used for synaptic stimulation during induction. The magnitude of LTP was calculated by comparing averaged EPSC amplitudes from 10-min windows (30 traces) immediately before and 30–40 min after LTP induction. These windows were also used to assess PPR and 1/CV2.
To test the independence of inputs in the two-pathway experiments (Figure 6), we used cross paired-pulse analysis. We first determined the PPR for each input. We then substituted the opposing input for the second pulse and confirmed the absence of interaction between the two inputs.
Caged IP3 (100 μM) was loaded into the cytosol through the whole-cell pipette. A 1-ms UV pulse was applied using a xenon arc lamp (Cairn Research) to rapidly release IP3 and the resulting SK-mediated outward current (IIP3) was measured. The amount of photolysis is known to be proportional to the UV pulse intensity, which is proportional to the capacitance of the capacitor feeding current to the flash lamp. This capacitance was varied (50–4050 μF) to adjust the UV pulse intensity.
Fluorescence imaging of intracellular Ca2+ was performed using fluo-5F (50 μM) loaded into the cytosol via the whole-cell pipette. Images were captured at 15 Hz with a spinning disk confocal imaging system (Olympus). Ca2+ signals from selected ROIs were expressed as ΔF/F = (F − Fbaseline)/(Fbaseline − Fbackground).
Additional methodological details are described in Supplemental Experimental Procedures.
We thank Drs. Rishikesh Narayanan and Michael Roberts for comments on the manuscript. We also thank Dr. Michael Mauk for suggesting the single-spike depotentiation experiment. This work was supported by National Institutes of Health Grant DA015687. M.T.H. was supported by a National Science Foundation Graduate Research Fellowship.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.