Motor learning and metaplasticity in striatal neurons: relevance for Parkinson’s disease

Abstract

Nigro-striatal dopamine transmission is central to a wide range of neuronal functions, including skill learning, which is disrupted in several pathologies such as Parkinson’s disease. The synaptic plasticity mechanisms, by which initial motor learning is stored for long time periods in striatal neurons, to then be gradually optimized upon subsequent training, remain unexplored. Addressing this issue is crucial to identify the synaptic and molecular mechanisms involved in striatal-dependent learning impairment in Parkinson’s disease. In this study, we took advantage of interindividual differences between outbred rodents in reaching plateau performance in the rotarod incremental motor learning protocol, to study striatal synaptic plasticity ex vivo. We then assessed how this process is modulated by dopamine receptors and the dopamine active transporter, and whether it is impaired by overexpression of human α-synuclein in the mesencephalon; the latter is a progressive animal model of Parkinson’s disease. We found that the initial acquisition of motor learning induced a dopamine active transporter and D1 receptors mediated long-term potentiation, under a protocol of long-term depression in striatal medium spiny neurons. This effect disappeared in animals reaching performance plateau. Overexpression of human α-synuclein reduced striatal dopamine active transporter levels, impaired motor learning, and prevented the learning-induced long-term potentiation, before the appearance of dopamine neuronal loss. Our findings provide evidence of a reorganization of cellular plasticity within the dorsolateral striatum that is mediated by dopamine receptors and dopamine active transporter during the acquisition of a skill. This newly identified mechanism of cellular memory is a form of metaplasticity that is disrupted in the early stage of synucleinopathies, such as Parkinson’s disease, and that might be relevant for other striatal pathologies, such as drug abuse.

striatal plasticity, motor learning, dopamine active transporter, alpha-synucleinopathy, long-term depression and potentiation

Introduction

Learning of new motor skills is an incremental process characterized by an initial phase during which performance is still not optimal and sensitive to interference. Additional training progressively leads to optimization of motor performance until a plateau is reached; at this stage the motor skill can be automatically performed (Kargo and Nitz, 2004; Graybiel and Grafton, 2015).

Skill learning is mediated by dopamine action in the dorsal striatum. However, the synaptic plasticity mechanisms by which initial motor learning is stored for long-lasting periods in striatal neurons to be, upon subsequent training, gradually optimized, remain unexplored. William James suggested that plasticity underlying habit formation requires ‘the possession of a structure weak enough to yield to an influence, but strong enough not to yield all at once’ (James, 1890), hypothesizing the importance of a ‘plasticity window’ during initial learning needed to allow the acquisition and concatenation of new sequences of action until performance becomes fully automatic.

Striatal dopamine action on D1 and D2 receptors is crucial for striatal synaptic plasticity. Initial and optimal skill learning have been suggested to require D1-mediated long-term potentiation (LTP) in the dorsomedial striatum and D2-mediated long-term depression (LTD) in the dorsolateral striatum, respectively (Shen et al., 2008, 2014). This dichotomy is supported by data showing that in the dorsolateral striatum tonic dopamine release, as well as higher D2 expression, predominates as compared to other striatal compartments (Szele et al., 1991; Zhang et al., 2009; Siciliano et al., 2014). However, dopamine D1 receptors are also expressed in the dorsolateral striatum where they mediate LTD and LTP in medium spiny neurons (Calabresi et al., 2007). Furthermore, previous studies suggested that the dorsolateral striatum could be concurrently activated during all stages of incremental motor learning (Thorn et al., 2010). The switch between D1 toward D2 receptors pathway activation would be the consequence of phasic and tonic dopamine release associated with early and late learning, respectively (Shen et al., 2008).

Herein we addressed the synaptic and molecular mechanisms underlying the transition from initial toward optimal learning within the dorsolateral striatum, which remain almost completely unexplored.

We used a protocol of motor learning-induced changes in synaptic plasticity, referred to as metaplasticity, and discovered that early motor learning in the rotarod task induced LTP under a synaptic plasticity protocol that in untrained animals induced LTD in dorsolateral striatal medium spiny neurons. This form of learning-induced metaplasticity, here referred to as rotarod-induced LTP, as well as motor learning, was regulated by dopamine active transporter (DAT) and by D1 receptor pathway activation.

Motor learning is impaired in human pathologies characterized by striatal dopaminergic dysfunction, such as Parkinson’s disease. Therefore, we tested the translational relevance on this form of metaplasticity in a progressive model of Parkinson’s disease, made by selectively overexpressing human wild-type α-Syn (hu-α-Syn) in the midbrain of normal mice. In this model, the accumulation of α-Syn leads to early changes in DAT functions, as well as striatal dopamine synthesis and release (Adamczyk et al., 2006; Lundblad et al., 2012). This early stage progresses toward loss of dopaminergic substantia nigra pars compacta (SNpc) neurons, accompanied by bradykinesia and other motor dysfunctions, recapitulating all stages of the human pathology.

Here, we report, for the first time, that midbrain hu-α-Syn overexpression reduced striatal DAT, impaired the acquisition of performance plateau in the rotarod task, and prevented the rotarod-induced LTP before leading to neurodegeneration.

Material and methods

Subjects

Animals were group-housed (Charles River), with ad libitum access to food and water during a 12 h light/dark cycle. Testing was performed during the light phase. All procedures related to animal care and treatments conformed to the guidelines and policies of the European Communities Council and were approved by the Italian Ministry of Health. Animals were randomly assigned to each experimental group. Each experiment included the relative control group. Unless specified, experiments were performed in 9–16-week-old CD1 outbred mice. The number of animals for each experimental group is reported in the respective figure legends. Each experiment was repeated at least twice with at least two subjects in the experimental group, until the final number of subjects was reached. The experimenter was blind to the treatment.

Experimental study design

Experiment 1

CD1 naïve outbred male mice were trained on the incremental rotarod test. The apparatus (Ugo Basile) consisted of a rod suspended horizontally at a height of 14 cm from the floor. The rod (5 cm in diameter) was accelerated from 4 rpm to 40 rpm in 300 s and the latency to fall from the rod was measured with a cut-off time of 600 s. After 3 days of training, we applied a posteriori criterion to distinct animals in Early or in Plateau performers. We included in the Plateau group only the animals whose latency to fall from the rod was 600 s for at least two of four trials by the third training day; animals that did not reach this criterion were defined as Early. An Untrained group, handled for the same number of days and training time was used as the control group; 5–15 days after training, animals were sacrificed and in vitro electrophysiology was performed on both groups, applying a high frequency stimulation (HFS) protocol to the medium spiny neurons recorded in the dorsolateral striatum that in naïve animals led to LTD (here referred as ‘LTD protocol’). The animals were trained in two groups of three to five; 5 days after training we sacrificed one animal per day for electrophysiological recordings, for 10 working days (which corresponded to 2 weeks). The groups were randomized and a post hoc analysis was performed to evaluate the effect of time on learning-induced plastic change. The distribution of LTP and LTD obtained when applying the LTD protocol was not dependent on the day of sacrifice (data not shown).

Experiment 2

To verify whether early mice could reach performance plateau with additional training we used two training protocols; the first consisted of an identical training protocol as used in Experiment 1, performed for 10 days instead of three. The second, here defined as the ‘Overtrained’ group, consisted of giving each training trial an additional trial of a maximum of 300 s at the maximum running speed (40 rpm).

Experiment 3

To verify whether the rotarod-induced LTP in synaptic plasticity was peculiar of CD1 mice, we replicated this finding in outbred 2-month-old male Wistar rats whose striatal plasticity has been extensively characterized (Lovinger et al., 1993; Manahan-Vaughan and Kulla, 2003; Calabresi et al., 2007; Twarkowski and Manahan-Vaughan, 2016), using exactly the same procedure as described for Experiment 1.

Experiment 4

To evaluate dopamine receptors and presynaptic dopaminergic machinery, tyrosine hydroxylase (TH) and DAT, five groups of naïve mice were sacrificed immediately after the last trial of the rotarod training, the striatum dissected and subjected to fractionation methods for biochemical analysis on homogenate and synaptoneurosomes. The experimental groups were: Untrained, Early and Plateau as described in Experiment 1. To dissociate the effects of the number of training days and/or the total amount of time on the rotarod from that of the learning stage, we included two additional groups: Overtrained, as described in Experiment 2, and the Early 1 day group, which was trained as described in Experiment 1, but only for 1 day instead of 3 days. To assess whether the observed biochemical changes in striatal protein expression immediately after training persisted after 5 days (the same time interval used for LTD), three further groups of animals (Untrained, Early and Plateau) were sacrificed 5 days after rotarod training, and the striatum processed for western blotting.

Experiment 5

To verify whether 1 day of training was sufficient to induce changes in striatal plasticity, and whether this was dependent on DAT activation, we subjected two groups of animals to the rotarod training as described in Experiment 1 for 1 day (Early 1 day group) 15 min after injecting them with vehicle or with the DAT antagonist GBR-12909 [2.5 mg/kg, intraperitoneal (i.p.)]. An Untrained group receiving GBR-12909 without rotarod training was used as control (data not shown). Animals were subjected to the ex vivo LTD protocol as described for Experiment 1, 5–15 days after training.

Experiment 6

To select a dose of GBR-12909 that did not affect spontaneous locomotor activity, animals were injected with saline, followed by 20 min habituation to the open-field testing cage; immediately after they received systemic injections of 2.5 mg/kg of GBR-12909; motor activity was then monitored for a further 40 min (Supplementary Fig. 2C) according to a within-group treatment experimental design. ANY-maze video tracking software (Stoelting) was used to collect the behavioural data. Total distance travelled (m) and total immobility time (s) were used as parameters of basal motor activity.

Experiment 7

We have previously shown that dopaminergic drugs affect striatal synaptic plasticity; blockade of D1 receptors blocks LTD and LTP induction, while blockade of D2 receptors blocks LTD and potentiates LTP induction in medium spiny neurons of untrained animals (Calabresi et al., 1992; Centonze et al., 2001). Therefore, to avoid any confounding effect of these drugs on basal synaptic plasticity, we tested the effects GBR-12909 (10 µM), SCH-23390 (10 µM) and sulpiride (10 µM) on rotarod-induced LTP maintenance. Naïve animals were subjected to the rotarod training as described in Experiment 1; the Early group underwent an LTD protocol in striatal medium spiny neurons 5–15 days later, which consistently led to LTP in this group of animals. Twelve minutes later the tetanus drugs were applied on striatal slices.

Experiment 8

To test the effects of these drugs in vivo on rotarod learning, each day, 15 min pre-training, we injected sulpiride (10 mg/kg, i.p.), SCH-23390 (0.01 mg/kg i.p.) and GBR-12909 (2.5 mg/kg, i.p., injected only the first training day) or vehicle in different groups of animals and subjected them to the rotarod training for 3 days, as described in Experiment 1. To test whether the inhibition of DAT could favour the activation of the D1 or D2 pathway, we added a fourth training day in which GBR-12909 was co-administered with SCH-23390 and sulpiride (GBR-12909 was given 20 min before training, followed, 5 min later, by the dopamine antagonist).

Experiment 9

We reproduced a progressive animal model of Parkinson’s disease by performing bilateral injections of recombinant adeno-associate viral vector (rAAV)-hu-α-Syn and rAAV-GFP in the SNpc/VTA (ventral tegmental area). To separate different stages of the pathology, animals were assessed at 28 weeks (long-term experiment) and 4 weeks (short-term experiment) after the injection. The experimental design is shown in Fig. 3A. These animals were subjected to a battery of behavioural tasks to evaluate working memory (6-different objects task and 6-identical objects task), anxiety (elevated plus-maze), muscle strength (wire hanging test), locomotor activity (open-field), sensorimotor gating (pre-pulse inhibition), and motor learning (rotarod test) (data not shown). Seven to 10 days after completion of the behavioural sessions, brains were processed for post-mortem analysis. For reasons of brevity, here we report only the results of the open-field and rotarod tests. Hu-α-Syn correct expression was assessed in all animals. Representative groups of these animals (as indicated in the figure legends) were used for biochemical and immunohistochemical experiments.

Experiment 10

To test whether skill learning-induced striatal metaplasticity was impaired in the early stage of Parkinson’s disease, we used animals injected on one site with rAAV-hu-α-Syn and on the other site with rAAV-GFP (defined as unilaterally injected animals). Five weeks after surgery (short-term experiment) they were subjected to the rotarod test as described in Experiment 1, and then to ex vivo electrophysiology for evaluating LTD. This allowed us the great advantage of having a comparable level of rotarod-induced plasticity in medium spiny neurons ipsilateral or contralateral to the rAAV-hu-α-Syn injection.

Experiment 11

Bilaterally rAAV-hu-α-Syn- and rAAV-GFP-injected mice were used to evaluate changes in basal synaptic plasticity, LTD and LTP, 5–8 weeks after surgery (short-term experiment).

Experiment 12

To assess changes in mRNA expression in the substantia nigra and on dopamine tissue and its metabolite levels in the striatum, the mesencephalon and the striatum were dissected 5 weeks after unilateral viral injection (short-term experiment) and processed for real-time quantitative RT-PCR and high-performance liquid chromatography (HPLC) as previously described (De Leonibus et al., 2007, 2009; Volpicelli et al., 2012).

General methods

Recombinant AAV viral vector

The rAAV2/6-hu-α-Syn and rAAV2/6-GFP vectors were kindly provided by Professor Anders Björklund and were the same used in studies previously described (Decressac et al., 2012; Lundblad et al., 2012). The expression of the transgene is driven by the synapsin-1 promoter and enhanced using a woodchuck hepatitis virus post-transcriptional regulatory element (WPRE). Vector production was performed as described previously (Decressac et al., 2012; Lundblad et al., 2012). Injected vector titre was 7.7 × 10¹⁴ genome copies/ml (gc/ml) for both rAAV-hu-α-syn and rAAV-GFP vectors.

Surgery

All surgical procedures were performed as previously described (De Leonibus et al., 2007). Vector solution was bilaterally injected within the SNpc-VTA complex using a 0.2 mm-gauge stainless steel injector connected to a 5 μl Hamilton syringe. Unilateral injection of the rAAV-hu-α-Syn vector was always accompanied by rAAV-GFP vector injection in the contralateral side. In all experimental groups, the rAAV was injected in a volume of 1 μl/side at a rate of 0.2 μl/min. The stereotaxic coordinates used (flat skull position) were: AP = −3.1 mm; ML = ±0.8 mm, DV = −5.2 mm and AP = −3.1 mm; ML = ±1.3 mm, DV = −4.8 mm relative to the bregma, according to the atlas of Paxinos and Franklin (2001). Only animals with correct injection placements, verified by analysing immunofluorescence staining of consecutive coronal brain sections, were included in the statistical analysis.

Drugs

GBR-12909 [dissolved in sterile isotonic (0.9%) solution] and SCH-23390 (dissolved in distilled water) were purchased from Sigma-Aldrich; sulpiride (dissolved in 100 mM of DMSO) was purchased from Tocris. All drugs were, unless specified, injected intraperitoneally (i.p.).

Immunohistochemistry

Mice were perfused with 4% paraformaldehyde (PFA). Free-floating coronal striatal sections 50-μm thick were incubated with 1.6% H₂O₂ in 100% ethanol, and blocked with phosphate-buffered saline (PBS) containing 0.3% Triton^TM X-100, 1% bovine serum albumin (BSA) and 5% normal goat serum (NGS). Rabbit primary antibodies directed to TH (1:1000, AB152; Millipore) were used for single-labelling experiments and the sections were then incubated with an anti-rabbit IgG peroxidase-labelled (1:200). Brain sections were finally exposed to avidin-biotin complex (Avidin/Biotin Blocking Kit SP-2001, Vector Laboratories).

Double immunolabelling

Primary antibodies of DAT (1:500, MAB369; Millipore) or hu-α-Syn (1:1000, sc-12767, Santa Cruz Biotechnology) in combination with hu-α-Syn phosphorylated at Ser129 (phospho S129-hu-α-Syn, 1:1000, ab51253, Abcam) or TH (1:1000, AB152, Millipore) were used. Brain slices were incubated with the proper secondary antibody (or a mixture of them): goat anti-rabbit Alexa Fluor® 647, goat anti-mouse Alexa Fluor® 488, goat anti-rat fluorescein (1:300, Millipore). DAPI as a nuclear counterstain was used. Images of the striatum (×40 magnification) and VTA-substantia nigra (SN) complex (×40 magnification) were taken using a fully motorized microscope Leica DM6000B with Leica digital camera DFC 480 RGB and DFC 350FX B/W and Leica application Suite X (LAS X) software. Striatal TH- and DAT-immunoreactive fibres were quantified by densitometry in five brain slices from 1.10 to 0.14 anterior to bregma (250 µm apart) per animal, throughout the rostrocaudal extent of the dorsal portion of the striatum. Specifically, fluorescence intensity within three rectangular shaped regions of interest in medial, central and lateral dorsal striatum was measured and corrected for background by subtracting the adjacent corpus callosum using ImageJ (NIH, 1.47). In addition, for each animal, four brain slices from 2.7 to 3.64 posterior to bregma, at a distance of 250 µm, representative of the VTA-SN complex were processed for TH immunoreactivity and incorporated in the counting procedure. The VTA and SNpc were delineated and were analysed by counting the number of TH-positive cells included in the demarcated areas. For brain area identification and definition of subregions, the mouse brain atlas Paxinos and Franklin (2001) was used.

Synaptoneurosome preparation and western blotting

The synaptoneurosome fraction was prepared using a modified version of the Borreca et al. (2016) protocol. Striatal samples were homogenized in the homogenization buffer (0.32 M sucrose, 1 mM EDTA, 1 mg/ml BSA, 5 mM HEPES pH 7.4, proteinase inhibitor cocktail) (Roche). A fraction (∼20%) of the homogenate represented the homogenate fraction. Other fractions were centrifuged and pellets resuspended in Krebs-Ringer buffer (140 mM NaCl, 5 mM KCl, 5 mM glucose, 1 mM EDTA, 10 mM HEPES pH 7.4) and Percoll (45% v/v) and centrifuged at 14 000rpm for 2 min. The enriched synaptoneurosomes were recovered and resuspended in RIPA buffer (50 mM Tris-HCl pH 7.4, 1% Triton^TM X-100, 0.5% Na-deoxycholate, 0.1% SDS, 150 mM NaCl, 2 mM EDTA and protease inhibitor cocktail). Proteins (10 µg) were then analysed by western blots using standard procedures. Membranes were incubated overnight at 4°C with primary antibodies: anti-TH, anti-DAT, anti-D1 receptor (1:500, sc-14001, Santa Cruz), anti-D2 receptor (1:500, ab5084P, Millipore) and anti-β-actin (1:5000, MAB1501, Millipore). Immunoreactivity was detected by chemiluminescence and bands quantified by densitometry using ImageJ software. Western blotting was performed on both striatal synaptoneurosomes and homogenate fractions; the results relative to the homogenate fraction are not shown when no differences were observed.

Electrophysiology

Intracellular recordings with sharp electrodes and whole-cell patch-clamp recordings were performed from medium spiny neurons of corticostriatal slices of rats and mice as previously described (Bagetta et al., 2011). In brief, the stimulating electrode was located in the striatum to activate corticostriatal fibres. The recording electrodes were placed within the dorsolateral striatum. All the whole-cell patch-clamp experiments were conducted in the continuous presence of 50 µM picrotoxin. Input resistances and injected currents were monitored throughout the experiments. Variations of these parameters >20% lead to the rejection of the experiment. As conditioning HFS protocol to induce, in striatal medium spiny neurons, LTD and LTP we used three trains (3 s duration, 100 Hz, at 20 s intervals). During tetanic stimulation, the intensity was increased to suprathreshold levels. For the LTP protocol, at the beginning of intracellular recordings, magnesium ions were omitted from the medium to increase the N-methyl-d-aspartate (NMDA)-mediated component of excitatory postsynaptic potential (EPSP). Quantitative data on EPSP modifications induced by high protocol are expressed as a percentage of control, the latter representing the mean of responses recorded during a stable period (15–20 min) before the tetanus.

Statistical analysis

The results are expressed as mean ± standard error of the mean (SEM). Statistical analyses were performed using Statview 5.0 (SAS Institute Inc., North Carolina, USA), SigmaPlot 11.0 (Systat Software, Chicago, IL, USA), InStat 3.0 (GraphPad, San Diego, CA, USA) and Statistica 10.0 (Statsoft, Tulsa, OK, USA). The statistical significance was assessed using two-tailed paired or unpaired t-test, ANOVA or repeated measured ANOVA, followed by the Duncan post hoc comparison test when appropriate. No exclusion criteria were applied, but some animals were excluded from statistical analysis in case of procedural problems, such as lost videos, etc. For all statistical analysis in this study, data distributions were assumed to be normal, but this was not formally tested. No statistical methods were used to predetermine sample sizes, but our sample sizes were similar to those generally employed in the field. The number of animals used for each experiment, the degrees of freedom and the number of recorded cells used for statistical analysis are reported in the figure legends. P < 0.05 was considered significant.

Results

Early incremental motor learning induces metaplasticity in dorsolateral striatum neurons

Outbred mice when subjected to a rotarod incremental motor learning protocol show interindividual differences in performance: ∼47% of the animals rapidly reached the performance plateau (‘Plateau’) and ∼53% did not reach it by the third training day (‘Early’) (Fig. 1A). The Early group, although slower, learned the task, as evidenced by a significant increase in the latency to fall off the rod across days (Fig. 1B) (Experiment 1); it slowly continued to improve towards performance plateau across additional training days (Supplementary Fig. 1A) (Experiment 2). The Early and the Plateau groups differed in the amount of training necessary to reach maximal performance, as also demonstrated by the experiment showing that, with an overtraining protocol, 100% of CD1 mice reached performance plateau in only 2 days (Supplementary Fig. 1B) (Experiment 2).

Figure 1

Early incremental motor learning induces LTP using a LTD protocol in vitro in dorsolateral striatal neurons. (A and B) Performance in the rotarod test in CD1 naïve mice belonging to the Plateau or to the Early groups, across trials of each day (A) or averaging trials values (B) [repeated measures ANOVA, Group, F(1,17) = 38.83, P < 0.0001; Day, F(2,34) = 71.77, P < 0.0001; Group × Day, F(2,34) = 16.57, P < 0.0001; Trial, F(3,51) = 20.27, P < 0.0001; Early, n = 10; Plateau, n = 9; Duncan post hoc test]. (C) LTD protocol applied to neurons from CD1 naïve untrained mice results in LTD [repeated measures ANOVA, Time, F(15,105) = 5.93, P < 0.0001; Duncan post hoc test]. At the top, representative electrophysiological traces are shown. (D) LTD protocol applied to neurons from CD1 naïve mice belonging to the Plateau groups (neurons, n = 7 from n = 5 mice) and to the Early group (neurons, n = 8 from n = 5 mice) results in LTD and LTP, respectively [repeated measures ANOVA, Group, F(1,13) = 6.31, P = 0.026; Time, F(15,195) = 2.44, P = 0.0027; Group × Time, F(15,195) = 5.22, P < 0.0001; Duncan post hoc test]. At the top, representative electrophysiological traces are shown. (E and F) Performance in the rotarod test in Wistar rats belonging to the Plateau or to the Early groups, across trials of each day (B) or averaging trials values (F) [repeated measures ANOVA, Group, F(1,14) = 21.07, P = 0.0004; Day, F(2,28) = 74.87, P < 0.0001; Group × Day, F(2,28) = 18.43, P < 0.0001; Trial, F(3,42) = 13.02, P < 0.0001; Early, n = 13; Plateau, n = 3; Duncan post hoc test]. (G) LTD protocol applied to neurons from Wistar Untrained rats results in LTD [repeated measures ANOVA, Time, F(15,75) = 12.97, P < 0.0001; Duncan post hoc test]. At the top, representative electrophysiological traces are shown. (H) LTD protocol applied to neurons from Wistar rats belonging to the Plateau (neurons, n = 13 from n = 5 rats) and to the to Early (neurons, n = 8 from n = 5 rats) groups results in LTD and LTP, respectively [repeated measures ANOVA, Group, F(1,19) = 13.15, P = 0.0018; Time, F(15,285) = 0.31, P = 0.99; Group × Time, F(15,285) = 3.74, P < 0.0001; Duncan post hoc test]. Representative electrophysiological traces are shown at the top of the figure. Data represent mean ± SEM; *P < 0.05 and ***P < 0.001 versus Early; ^#P < 0.05, ^##P < 0.01 and ^###P < 0.001 versus Day 1, T1 within treatment and day, pre-HFS. EPSP = excitatory postsynaptic potential; MSN = medium spiny neurons.

We took advantage of these interindividual differences to dissociate the effects of early versus plateau performance stages on dorsolateral striatum neuronal plasticity. To address the rotarod-induced changes in striatal synaptic plasticity, 5–15 days after the rotarod test, we applied an HFS protocol to the medium spiny neurons of the dorsolateral striatum that led to LTD in Untrained mice (Fig. 1C) and Plateau animals (Fig. 1D, black circles), as expected. In contrast, the same HFS protocol induced LTP when learning was incomplete (Early group) (Fig. 1D, white circles), suggesting that early learning induced a change in striatal plasticity from LTD to LTP (Experiment 1). We will refer to this change as rotarod-induced LTP.

We confirmed this new identified mechanism of metaplasticity in outbred Wistar rats, which is the most used animal model to study striatal plasticity (Experiment 3). The two outbred animal models, different from inbred C57BL/6 mice (data not shown), showed similar behavioural results when challenged by the same learning protocol (Fig. 1E and F). As well as in CD1 mice, stimulation protocol that induced LTD in Untrained (Fig. 1G) and Plateau groups of rats (Fig. 1H, filled squares), induced an LTP instead of LTD in the Early group (Fig. 1H, open squares). The only major difference between these two species was the increased percentage of cells that under the motor learning/LTD protocols showed no change (<10% mean variation from baseline defined as ‘blocked-LTD’) (Supplementary Fig. 1C, CI and D), suggesting a block of LTD.

Metaplasticity in striatal neurons is mediated by DAT regulated dopamine action and D1 receptors

Striatal dopamine is crucial for striatal-dependent learning and neuronal plasticity, and for regulating the activation of D1 and D2 receptors (Calabresi et al., 2007). Therefore, we tested the effects of rotarod training on the key components of the presynaptic dopamine machinery, DAT and TH (Experiment 4). DAT was increased in the Early learning group; a similar increase was observed in animals that were trained on the rotarod for only 1 day (1 day Early and Early bars, Fig. 2A); this DAT increase was almost significant even 5 days after the rotarod training (Supplementary Fig. 2A). In contrast, DAT significantly decreased in animals reaching performance plateau, by standard training or overtraining protocol (Plateau and Overtrained bars, Fig. 2A). These findings further confirmed that the level of DAT, in the different training conditions, was regulated by the learning stage, rather than by general habituation to the rotarod or the number of training days (total time rotarod running: 1 day Early = 1049 ± 220 s; 3 days Early = 3016 ± 311 s; 3 days Plateau = 5290 ± 506 s; 2 days Overtrained = 3976 ± 184 s). TH expression was increased in both Early and Plateau groups (Fig. 2A), but it did not persist 5 days after training (Supplementary Fig. 2A); if any, a tendency to reduced TH expression in both groups was observed 5 days after training.

Figure 2

Early learning-induced metaplasticity in the dorsolateral striatum is mediated by DAT-regulated dopamine action and D1 receptors. (A) Densitometric quantification of protein in the striatal synaptoneurosome fraction, immediately after rotarod training, shows, compared to the Untrained mice, a significant increase in DAT in mice belonging to the Early groups—after 1 [unpaired t-test, t(1,12) = 6.79, P = 0.023, Untrained (U), n = 7; 1 day Early, n = 7] or 3 days [unpaired t-test, t(1,11) = 8.14, P = 0.016, Untrained, n = 7; Early (E), n = 6] of rotarod training. Compared to DAT levels after 1 day of rotarod test, there is a significant decrease in DAT in mice belonging to the Plateau (Pla) groups—after 3 days [unpaired t-test, t(1,9) = 6.30, P = 0.033, Plateau, n = 4; 1 day Early, n = 7] or 2 days of overtraining [unpaired t-test, t(1,11) = 5.18, P = 0.044, Overtrained (Over), n = 6; 1 day Early, n = 7] on the rotarod. In contrast, compared to the Untrained mice, there is an increase in TH protein level after 1 [unpaired t-test, t(1,12) = 9.70, P = 0.0089, Untrained, n = 7; 1 day Early, n = 7] and 3 days [unpaired t-test, t(1,11) = 6.41, P = 0.028, Untrained, n = 7; Early, n = 6] of rotarod training in mice that did not reach the performance plateau, that persists in mice that reached performance plateau [unpaired t-test, t(1,9) = 9.051, P = 0.015, Untrained, n = 7; Plateau, n = 4]. The gel was cut in the middle to eliminate redundant lanes, as indicated by the dashed lines. (B) One day of rotarod training in naïve mice is sufficient to induce, 5–15 days later, LTP in response to the LTD protocol applied to medium spiny neurons (MSNs, neurons, n = 5 from n = 4 mice); 15 min pre-rotarod training (Day 1) administration of GBR-12909 (2.5 mg/kg, i.p.) (GBR, neurons, n = 4 from n = 4 mice) prevents the LTP using a LTD protocol in vitro [repeated measures ANOVA, Group, F(1,7) = 5.80, P = 0.047; Time, F(15,105) = 1.98, P = 0.023; Group × Time, F(15,105) = 2.08, P = 0.016; Duncan post hoc test]. Representative electrophysiological traces are shown at the top of the figure. (C) The application of GBR-12909 (10 µM) to neurons (n = 5 from n = 4 mice) significantly decreases LTP induced by a LTD protocol in the Early group [repeated measures ANOVA, Time, F(15,60) = 5.49, P < 0.0001; Duncan post hoc test]. Representative electrophysiological traces are shown at the top of the figure. (D) Densitometric quantification of protein in the striatal synaptoneurosome fraction, immediately after the rotarod training, shows a significant increase in D1 receptors compared to the Untrained in the Early [unpaired t-test, t(1,8) = 13.30, P = 0.0065, Untrained, n = 6; Early, n = 4] and Plateau [unpaired t-test, t(1,8) = 6.58, P = 0.033, Untrained, n = 6; Plateau, n = 4] groups, but not D2 receptors [unpaired t-test, t(1,8) = 0.23, P = 0.65, Untrained, n = 6; Early, n = 4; unpaired t-test, t(1,8) = 1.63, P = 0.24, Untrained, n = 6; Plateau, n = 4]. The gel was cut in the middle in order to eliminate redundant lanes, as indicated by the dashed lines. (E) The application of SCH-23390 (10 µM) but not sulpiride (10 µM) to neurons (SCH-23390, n = 4 from n = 4 rats; sulpiride, n = 4 from n = 4 rats) significantly decreases the rotarod-induced LTP in the Early group [SCH-23390, repeated measures ANOVA, Time, F(15,45) = 4.15, P = 0.0001; Duncan post hoc test; sulpiride, repeated measures ANOVA, Time, F(15,45) = 10.93, P < 0.0001; Duncan post hoc test]. (F) Percentage of Plateau and Early mice in the rotarod test after in vivo pre-training injection of sulpiride, SCH-23390 or GBR-12909, compare to vehicle-treated mice. (G) Latency to fall from the rod in Plateau mice after in vivo pre-training injection of SCH-23390 or GBR-12909 compared to vehicle-treated mice [repeated measures ANOVA, Group, F(2,13) = 1.72, P = 0.22; Day, F(2,26) = 69.57, P < 0.0001; Group × Day, F(4,26) = 0.90, P = 0.48; Vehicle, n = 10; SCH, n = 3; GBR, n = 3; Duncan post hoc test]. (H) Latency to fall from the rod in Early mice not reaching the performance plateau after in vivo pre-training injection of sulpiride, SCH-23390 or GBR-12909 compared to vehicle-treated mice [repeated measures ANOVA, Group, F(3,24) = 1.27, P = 0.31; Day, F(2,48) = 20.01, P < 0.0001; Group × Day, F(6,48) = 3.00, P = 0.014; Vehicle, n = 9; SCH, n = 6; GBR, n = 6; sulpiride, n = 7; Duncan post hoc test]. Administration of GBR on the fourth day of training in sulpiride and SCH-23390 Early groups impaired and improved performance, respectively in these groups as compared to vehicle [repeated measures ANOVA, Group, F(1,11) = 0.74, P = 0.76; Day, F(1,11) = 0.64, P = 0.44; Group × Day, F(1,11) = 11.37, P = 0.0062; Duncan post hoc test]. Data represent mean ± SEM; *P < 0.05 versus 1 day Early; ^#P < 0.05, ^##P < 0.01 and ^###P < 0.001 versus Untrained, Day 1, pre-HFS; ^§P < 0.05 versus Day 3. MSN = medium spiny neurons.

These findings suggest that inhibition of DAT, during early motor learning, might have impaired the early learning-induced LTP in dorsolateral striatum. To test this hypothesis (Experiment 5), we first verified that 1 day, as well as 3 days, of rotarod learning was sufficient to trigger LTP under the LTD protocol (Fig. 2B). We then showed that the inhibition of DAT on the first day of training with systemic injection of the DAT selective inhibitor, GBR-12909, impaired 1 day rotarod-induced LTP (Fig. 2B). Systemic administration of the same dose of the drug did not affect LTD in Untrained animals (data not shown). The same dose of GBR-12909 was void of effects either on medium spiny neurons basal synaptic transmission (data not shown) or basal locomotor activity (Supplementary Fig. 2C) (Experiment 6). GBR-12909 (10 µM) applied in striatal slices downscaled the rotarod-induced LTP, suggesting a role for DAT in both the induction and maintenance of the rotarod-induced LTP (Fig. 2C).

To address whether the observed presynaptic alterations were paralleled by changes at postsynaptic level, we measured D1 and D2 expression in striatal synaptoneurosome of Early and Plateau groups (Experiment 4). In line with previous findings using a similar approach, D1 receptor expression increased in both Early and Plateau groups (Yin et al., 2009; Sommer et al., 2014), as compared to the Untrained group (Fig. 2D); the observed increase in D1 persisted 5 days after training in the Early group, but not in the Plateau group (Supplementary Fig. 2B). Consistently, the D1 receptors inhibitor, SCH-23390 (10 µM) (Fig. 2E), but not the D2 receptors inhibitor sulpiride (10 µM) (Fig. 2E), fully downscaled rotarod-induced LTP (Experiment 7).

Altogether, these data show a role of DAT in both the induction and maintenance of rotarod-induced LTP, likely regulating dopamine action on D1 receptors. Indeed, pre-training injection of these drugs (Experiment 8), at doses that did not affect basal locomotion (Adriani et al., 2000), showed that sulpiride completely prevented the acquisition of performance plateau (Fig. 2F), but did not affect performance in the Early group (Fig. 2H and Supplementary Fig. 2D, purple circles). In contrast, SCH-23390 did not prevent the acquisition of performance plateau (Fig. 2F and G, yellow circles) in a small percentage of animals, but impaired early learning in the rest of them (Early group) as compared to vehicle-injected animals [Fig. 2H and Supplementary Fig. 2D(I), yellow circles]. Interestingly, the effects of GBR-12909 paralleled those of the D1 receptor antagonist (Fig. 2F–H). Although it was injected only during the first day of training, which was sufficient to prevent the rotarod-induced shift to LTP and to impair learning on the same day [Day 1 × Treatment: F(3,39) = 4.1; P = 0.01, Supplementary Fig. 2D(II)], the GBR-12909 group was impaired as compared to the control group during the following days, when the animals were OFF-drug [Fig. 2H and Supplementary Fig. 2D(II)].

These data suggest that by inhibiting dopamine reuptake we might have mimicked D1 receptor blockade. Inhibition of DAT through GBR-12909 in the dorsolateral striatum has been demonstrated to increase tonic dopamine (Zhang et al., 2009). Interestingly, administration of GBR-12909 to the fourth training day to the SCH-23390 and the sulpiride treated groups improved and impaired performance, respectively (Day 4 in Fig. 2H). This finding suggests that inhibition of DAT, through GBR-12909 application, shifted dopamine action from D1 towards D2 receptor pathway activation.

Altogether, these findings show that early stages of motor learning induce an LTP under an LTD protocol through DAT-regulated dopamine action in the dorsolateral striatum neurons; behavioural experiments suggest that DAT may regulate the different stages of motor learning by ‘gating’ dopamine action on D1 and D2 pathways.

Midbrain hu-α-Syn overexpression impairs metaplasticity in striatal neurons and performance plateau acquisition

To model the time course of the progression from early toward advanced stages of Parkinson’s disease patients (Lundblad et al., 2012), we used SNpc-VTA bilateral injection of rAAV-hu-α-Syn. Consistently with previous critical findings, long-term overexpression of rAAV-hu-α-Syn (long-term experiment, timeline protocol in Fig. 3A) led to dopamine neuronal loss and bradykinesia [Supplementary Fig. 3A-F(I)], as compared to rAAV-GFP (Experiment 9). In contrast, short-term overexpression of rAAV-hu-α-Syn (short-term experiment, timeline in Fig. 3A) selectively impaired motor learning as compared to rAAV-GFP mice (Fig. 3B and C), in the absence of bradykinesia in the open field (data not shown).

Figure 3

Short-term overexpression of hu-α-Syn in midbrain impairs incremental motor learning-induced striatal plasticity. (A) Schematic representation of the experimental design to assess the neurobehavioural consequences of short- and long-term overexpression of hu-α-Syn in midbrain. Two separate groups of animals were bilaterally injected in the SNpc/VTA complex with rAAV vectors encoding either hu-α-Syn or GFP and sacrificed at two different time points. All animals were subjected to a number of behavioural tests. Seven to 10 days after completion of the behavioural sessions, brains were processed for post-mortem analysis. (B and C) rAAV-hu-α-Syn- (n = 20) and rAAV-GFP-injected mice (n = 16) are behaviourally tested. Overexpression of hu-α-Syn induces a motor learning defect in the rotarod test. Latency to fall do not increase across trials (T1-4) and days in the rAAV-hu-α-Syn group, except on Day 1 [B, repeated measures ANOVA, Treatment, F(1,34) = 13.93, P = 0.0007; Day, F(2,68) = 16.20, P < 0.0001; Treatment × Day, F(2,68) = 7.97, P = 0.0008; Trial, F(3,102) = 11.97, P < 0.0001; Day × Trial, F(6,204) = 3.88, P = 0.0011; Treatment × Trial, F(3,102) = 1.37, P = 0.25; Treatment × Trial × Day, F(6,204) = 1.13, P = 0.35; Duncan post hoc test]. Averaging trials values, a significant difference emerges during Day 3 with Duncan post hoc test (C). (D) LTD protocol is applied to neurons from unilaterally injected mice—on one side rAAV-GFP (neurons, n = 4 from n = 5 mice) and on the other side rAAV-hu-α-Syn (neurons, n = 9 from n = 5 mice), trained for 3 days on the rotorad test and recorded after 5–15 days: rotarod training produces LTP in the rAAV-GFP site, and in the rAAV-hu-α-Syn site LTD with an initial increase in excitatory postsynaptic potential (EPSP) only in the first two time points [repeated measures ANOVA, Site of injection, F(1,11) = 5.03, P = 0.046; Time, F(15,165) = 3.02, P = 0.00026; Site of injection × Time, F(15,165) = 2.82, P = 0.0006; Duncan post hoc test]. Representative electrophysiological traces are shown at the top of the figure. (E) The application of a high frequency stimulation (HFS) protocol to the medium spiny neurons (MSNs) of dorsolateral striatum, recorded in both rAAV-GFP (neurons, n = 10 from n = 5 mice) and rAAV-hu-α-Syn mice (neurons, n = 8 from n = 5 mice) not subjected to the rotarod test, produces a normal LTD [repeated measures ANOVA, Treatment, F(1,16) = 0.37, P = 0.55; Time, F(15,240) = 8.08, P < 0.0001; Treatment × Time, F(15,240) = 0.39, P = 0.98; Duncan post hoc test]. Representative electrophysiological traces are shown at the top of the figure. (F) Application of HFS protocol in magnesium free condition leads to a normal LTP in rAAV-hu-α-Syn mice (neurons, n = 5 from n = 5 mice), similar to that recorded in rAAV-GFP mice (neurons, n = 5 from n = 6 mice) [repeated measures ANOVA, Treatment, F(1,8) = 0.16, P = 0.70; Time, F(15,120) = 5.84, P < 0.0001; Treatment × Time, F(15,120) = 0.85, P = 0.62; Duncan post hoc test]. Representative electrophysiological traces are shown on the right of the graphs. Data represent mean ± SEM. *P < 0.05 versus Uni rAAV-GFP; ^#P < 0.05, ^##P < 0.01 and ^###P < 0.001 versus T1 within treatment and day, pre-HFS. ST = short-term; LT = long-term.

To assess whether this rotarod impairment prevented the rotarod-induced LTP, we used animals injected on one side with rAAV-GFP and on the other side with rAAV-hu-α-Syn (Experiment 10), as they were indeed impaired in motor learning to the same extent as bilaterally-injected animals (Supplementary Fig. 4A). After the rotarod test, in the medium spiny neurons recorded from the contralateral side (injected with rAAV-GFP), the rotarod-induced LTP was observed (Fig. 3D). This is in line with a comparable level of performance on the first training day, which led to a comparable level of rotarod-induced LTP in these two groups [Supplementary Fig. 4B and B(I)]. Nevertheless, the unilaterally-injected group did not improve performance across days, due to rAAV-hu-α-Syn expression (Supplementary Fig. 4B). The rotarod-induced LTP in the neurons recorded from rAAV-hu-α-Syn-injected mice showed a short-term potentiation that dropped soon after and did not reach a normal LTD (10 min post-tetanus, Fig. 3D). Impaired rotarod-induced LTP in the rAAV-hu-α-Syn site was not due to a general defect in the capability to potentiate and decrease synaptic transmission, as Untrained rAAV-hu-α-Syn mice presented normal LTD, and normal LTP when HFS was applied in magnesium-free conditions (Fig. 3E and F) (Experiment 11).

Hu-α-Syn-mediated impairment in metaplasticity is associated to reduced DAT expression before dopamine neuronal loss

The deficit in motor learning-induced metaplasticity observed in rAAV-hu-α-Syn mice (Fig. 3C) recapitulates the DAT and D1 inhibitors-induced LTP block (Fig. 2C and E). We therefore evaluated the impact of α-Syn on components of the synthesis and release of dopamine pathway (Experiment 12). rAAV-mediated hu-α-Syn midbrain expression (short-term experiment) transduced TH+ neurons, and consistently with its presynaptic function abundantly reached the striatum (Fig. 4A). The pathological α-Syn phosphorylated form, Ser(129)-phosphorylated, was detected within the cell bodies of hu-α-Syn overexpressing cells, exclusively in the injection site, already at this early time point (Fig. 4B). Nevertheless, counting of TH+ neurons in the SNpc and VTA did not reveal any difference between groups in the short-term experiment (Fig. 4C and D), suggesting that this was a ‘pre-neurodegeneration’ stage. Co-localization markers showed that hu-α-Syn was present in the cell body of TH+ midbrain neurons (Fig. 4E). This was associated with reduced midbrain mRNAs of both TH and DAT, but not of the vesicular monoamine transporter-2 (VMAT2) (Fig. 4F) (Experiment 12). At the level of the striatum, immunohistochemical analysis showed reduced expression of TH protein in all dorsal striatum subregions (Supplementary Fig. 4C-CII, D and DI), which was confirmed by western blot analysis showing a significant reduction of TH protein in the homogenate but not in the synaptoneurosome fraction (Fig. 4G). This suggested that it might be a compensatory increase in TH transport to the synapse, or reduced degradation. This mechanism might be sufficient to maintain striatal tissue levels unchanged. Indeed, dopamine level in striatal tissue, which was quantified by HPLC, was not significantly affected by rAAV-hu-α-Syn (Fig. 4H) (Experiment 12). In contrast, DAT expression was dramatically reduced in both the homogenate and the synaptoneurosome fractions (Fig. 4I). A tendency toward D1, but not D2, receptors increase was observed in the rAAV-hu-α-Syn group (Fig. 4J).

Figure 4

Short-term overexpression of hu-α-Syn in midbrain leads to decreased transcription and reduced striatal protein levels of TH and DAT, associated with altered response to DAT inhibitors. (A) Immunofluorescence for hu-α-Syn and TH displays high levels of hu-α-Syn at the injection site (SNpc and VTA) and within the striatum. (B) Positive staining of the phosphorylated form (phospho-Ser129-hu-α-Syn) is found exclusively in the cell bodies of the SNpc/VTA complex. (C and D) Counting of TH+ neurons within both the SNpc and VTA does not reveal a reduction of dopamine neurons in rAAV-hu-α-Syn injected mice [unpaired t-test, SNpc, t(1,10) = 0.63, P = 0.45, rAAV-GFP, n = 4; rAAV-hu-α-Syn, n = 8; VTA, t(1,8) = 0.45, P = 0.52, rAAV-GFP, n = 4; rAAV-hu-α-Syn, n = 6]. (E) A higher magnification (×100) reveals the presence of hu-α-Syn in TH+ neurons in the SNpc/VTA. (F) Quantitative RT-PCR performed on mRNA extracted from the SNpc/VTA complex shows reduced expression of TH [paired t-test, 55 ± 13%; t(4) = 3.55, P = 0.024] and DAT (Slc6a3) [paired t-test, 38 ± 15%; t(4) = 4.01, P = 0.015], but not VMAT2 (Slc18a2) [paired t-test, 94 ± 31%; t(4) = 0.20, P = 0.85], in the rAAV-hu-α-Syn as compared to the contralateral rAAV-GFP injected site (n = 5; unilaterally injected). (G) The unpaired t-test on relative band quantifications of striatal homogenate and synaptoneurosome fractions show a significant reduction of TH in the homogenate [t(1,6) = 11.64, P = 0.014; rAAV-GFP, n = 4, pulled from eight mice; rAAV-hu-α-Syn, n = 4, pulled from eight mice] but not in the synaptoneurosome striatal fraction [t(1,6) = 2.61, P = 0.16; rAAV-GFP, n = 4, pulled from eight mice; rAAV-hu-α-Syn, n = 4, pulled from eight mice]. (H) Levels of dopamine and its metabolites, 3,4-dihydroxyphenylacetic acid (DOPAC) and homovanillic acid (HVA), assayed by HPLC, in the striatum ipsilateral to rAAV-hu-α-Syn nigra overexpression do not differ from that ipsilateral to rAAV-GFP nigra overexpression [unpaired t-test, dopamine (DA), t(1,8) = 1.58, P = 0.24; DOPAC, t(1,8) = 0.90, P = 0.37, HVA, t(1,8) = 0.027, P = 0.87; n = 5; unilaterally injected]. (I) The unpaired t-test on densitometric quantification of protein in striatal homogenate and synaptoneurosome fractions show a significant DAT reduction in both fractions [homogenate, t(1,6) = 29.74, P = 0.0016; synaptoneurosome, t(1,6) = 14.17, P = 0.0027; rAAV-GFP, n = 4, pulled from eight mice; rAAV-hu-α-Syn, n = 4, pulled from eight mice]. (J) The unpaired t-test on densitometric quantification of protein in striatal synaptoneurosome fraction an almost significant effect for D1, but not for D2 receptors [D1R, t(1,6) = 5.81, P = 0.052; D2R, t(1,6) = 0.15, P = 0.71; rAAV-GFP, n = 4, pulled from eight mice; rAAV-hu-α-Syn, n = 4, pulled from eight mice]. Data represent mean ± SEM. *P < 0.05 and **P < 0.01 versus rAAV-GFP.

Altogether, these data suggest that the identified pre-synaptic changes in dopamine signalling machinery, in the very early stage after midbrain rAAV-hu-α-Syn overexpression, are not due to dopamine neuronal loss, but likely to early altered expression and function of TH and DAT protein, as previously reported (Eriksen et al., 2010; Decressac et al., 2012; Lundblad et al., 2012; Oaks et al., 2013). These presynaptic changes might be sufficient to alter striatal postsynaptic receptors changes, as observed for D1 receptors expression.

Discussion

In this study, we describe a novel synaptic mechanism by which the transition from initial toward plateau performance is represented at cellular level in the dorsolateral striatum. This mechanism may have broad relevance for several striatal-related diseases.

Previous findings suggested that practice-induced behavioural optimization is mediated by a transition from dorsomedial striatum toward dorsolateral striatum neuronal activation (Yin et al., 2009; Shiflett et al., 2010; Thorn et al., 2010; De Leonibus et al., 2011); nevertheless, lesions of the dorsolateral striatum impairs both stages of motor learning (Yin et al., 2009). How these different stages regulate synaptic plasticity within the dorsolateral striatum has not been clarified yet.

In this study, by using individual differences in learning speed in outbred mice and rats, we discovered that early motor learning acquisition makes medium spiny neurons of dorsolateral striatum susceptible to LTP, rather than to LTD. This form of metaplasticity is no longer elicited when performance plateau is reached. Therefore, we provide evidence of a cellular mechanism in support of in vivo findings showing that the dorsolateral striatum is recruited during initial learning (Yin et al., 2009), and that decreased activity in these neurons reflects a neural mechanism underlying improvement in movement efficiency with overtraining (Tang et al., 2009; Shan et al., 2014, 2015). Previous studies that have used a learning induced-striatal metaplasticity approach in C57BL/6 mice were not able to identify the early learning-induced shift in dorsolateral striatum plasticity. This might be due to the use of inbred mice, which do not overlap the span in behavioural performance observed in outbred strains, or to the use of different neuronal plasticity protocol (Yin et al., 2009; Shan et al., 2014, 2015).

Previous findings suggested that the change from early toward habit learning is regulated by D1 pathway in the dorsomedial striatum and D2 pathway in the dorsolateral striatum, respectively (Yin et al., 2009; Shan et al., 2014, 2015). By showing that early motor learning-induced LTP in the dorsolateral striatum is mediated by D1, but not D2, receptors activation, we herewith provide evidence that early motor learning activates D1 pathway also within the dorsolateral striatum.

How D1 and D2 receptors are differently activated during different steps of skill learning is still unknown. Early and extensive rotarod training might regulate the activation of D1 and D2 receptor pathways, likely by changing their sensitivity to dopamine release (Yin et al., 2009; Shan et al., 2014, 2015; Sommer et al., 2014). Accordingly, we have found here that early learning was associated to long-lasting increase in D1 striatal expression. However, additional pre-synaptic mechanisms have been suggested to regulate the selective activation of these pathways. Although dopamine binds to both D1 and D2 receptors, the relative activation of either subtype depends on the synaptic dopamine concentration and the respective affinities of the receptors for the neurotransmitter (Richfield et al., 1989; Thivierge et al., 2007; Nieto-Mendoza and Hernandez-Echeagaray, 2015). D2 receptors, which have higher affinity than D1 receptors for dopamine, mediate tonic dopaminergic signalling. D1 receptors are activated at high dopamine concentrations during phasic increases in extracellular dopamine (Dreyer et al., 2010). Phasic and tonic dopamine release is regulated in the striatum by different mechanisms, among which the dopamine transporter DAT is the best one characterized. DAT is highly expressed in the dorsolateral striatum and the inhibition of DAT with GBR-12909 has been demonstrated to increase tonic dopamine in the dorsolateral striatum, which led to preferential activation of D2 receptors (Zhang et al., 2009). Accordingly, we reported that early versus optimal motor learning leads to high versus low striatal DAT levels, paralleled by increased TH, which is consistent with the need of a highly tuned regulation of dopamine release during the very early stage of learning. Therefore, concomitant increase in DAT and D1 receptors expression, during initial learning, might allow to maintain dopamine in a phasic state, and to avoid a premature switch from D1 to D2 receptor pathway activation. By showing here that early learning through DAT and D1 makes dorsolateral striatum synapses more prone to show LTP under an LTD protocol, we provide the first cellular mechanism through which sensorimotor learning controls the transition from early versus optimal learning within the dorsolateral striatum. Indeed, premature DAT lowering under normal D1 activation, although it allows learning across days, does not enable the performance plateau in most of the animals to be reached, as shown in control animals injected with GBR-12909 and in overexpressing human α-Syn mice. In contrast, when learning is complete a decrease in DAT might favour the activation of tonic dopamine action on D2 receptors, which is necessary for induction and maintenance of LTD (Calabresi et al., 2007).

It must be noted, however, that a small percentage of animals can reach plateau performance under GBR-12909 and under SCH-23390 treatment. This finding suggests that there might be a parallel signalling pathway to reach performance plateau, which does not require the early training-induced metaplasticity, but requires D2 receptors activation. The latter might rely, in the early stages, on the activation of other receptors classes such as cannabinoid or glutamate receptors (Centonze et al., 2003; Calabresi et al., 2007; Kreitzer and Malenka, 2007; Shan et al., 2015); future studies are needed to address the contribution of these pathways in this form of metaplasticity.

Altogether, our findings add an important piece of information to previous evidence showing that automatization of behaviour involves a change from D1 pathway in the dorsomedial striatum to the D2 pathway in the dorsolateral striatum. We speculate that DAT-regulated activation of the D1 pathway in the dorsolateral striatum, during the early stage of incremental motor learning, cooperates with D1 pathway activation in the dorsomedial area to prevent premature shifting to habit learning. Thus, early learning-induced LTP in the dorsolateral striatum might be a mechanism to prevent a premature activation of the indirect pathway; at the same time it might also be a mechanism of ‘synaptic tagging’ (Wang et al., 2010) for dopamine released during consolidation or subsequent training, until a shift to tonic dopamine release will allow the activation of D2 receptors leading to the dominance of the indirect pathway over the direct one. Further longitudinal in vivo studies will allow us to correlate the observed plasticity and behavioural changes.

These data provide a novel insight into the synaptic mechanisms through which the dorsolateral striatum regulates its role in different stages of habit formation and may be relevant for all those pathologies involving habit learning dysfunctions, such as neurodegenerative disorders affecting basal ganglia.

In animal models of neurodegeneration, deficits in synaptic plasticity in the early stages of the disease can be uncovered by ‘challenging’ the neuron with learning protocols (D'Amelio et al., 2011). This approach allowed us to find that overexpression of α-Syn impairs motor learning by altering DAT expression, before leading to dopamine neuronal loss and bradykinesia.

The neuronal and behavioural deficits observed in rAAV-hu-α-Syn mice strongly parallel those induced by pre-training pharmacological inhibition of DAT, although in the former TH protein level is also reduced. In vitro, direct interaction between α-Syn and DAT proteins leads to decreased dopamine uptake (Wersinger and Sidhu, 2003, 2005; Adamczyk et al., 2006; Qian et al., 2008). Moreover, using the same experimental approach (rAAV-α-Syn rats), previous studies identified a reduction in the dopamine reuptake rate and increased its striatal tonic release before any major loss of dopaminergic neurons. Therefore, our findings are in line with previous evidence showing dysfunctions of dopamine terminals preceding the dopamine neuronal loss in this mouse model (Lam et al., 2011; Lundblad et al., 2012; Volakakis et al., 2015).

In line with previous findings (Tozzi et al., 2016), we did not find any synaptic plasticity impairments in basal conditions at this early stage in spiny neurons. Conversely, permanently reduced DAT protein expression and increased D1 expression in hu-α-Syn overexpressing mice recapitulates the pharmacological inhibition of rotarod-induced LTP of the DAT-inhibitor GBR-12909, as well as its in vivo-induced learning deficit in control mice. The identification of a cellular mechanism of motor memory, which is impaired by α-Syn overexpression, has a particular relevance for human studies showing that healthy SNCA (α-Syn) gene duplication human carriers display impaired instrumental learning compared with non-carriers (Keri et al., 2010). The findings on motor learning impairment in this study and the accumulating evidence that α-Syn pathology contributes to early learning impairment in Parkinson’s disease (Skogseth et al., 2015), greatly increase the relevance of this model. Reduced DAT activity in patients is considered to be a direct index of dopaminergic neurodegeneration. Our preclinical findings, in line with accumulating experimental evidence (Kirik et al., 2002; Lundblad et al., 2012; Oliveras-Salva et al., 2013; Song et al., 2015), suggest the intriguing hypothesis that it might instead be an early sign of synucleinopathy not necessarily correlated with dopamine neuronal loss. Our findings are in line with clinical studies suggesting that patients with Parkinson’s disease show deficits during the extended practice or automatization phases (Doyon et al., 1998, 2009; Agostino et al., 2004; Wu and Hallett, 2005). This impairment might be explained assuming that the role of the corticostriatal system increases as a function of learning (Doyon et al., 2009). Accordingly, continued practice seems to cause a shift in activation from the associative to the sensorimotor regions of the striatum and these regions are particularly affected in the early stages of Parkinson’s disease (Lehericy et al., 2005).

Our findings, together with the previous literature, support a reorganization of cellular plasticity within the dorsolateral striatum that is induced by the acquisition of a skill via dopamine receptor pathway-specific change. Our experimental findings show that these changes are controlled by DAT, which is impaired by mesencephalic overexpression of human α-Syn.

Although learning-induced striatal metaplasticity in Parkinson’s disease has been previously hypothesized, we report here for the first time the molecular features of this form of synaptic plasticity demonstrating its impairment in a very early stage of the disease preceding neurodegeneration.

Acknowledgements

We thank Prof. Anders Bjorklund for inspiring this study and for kindly providing the vector, and Jenny G. Johansson for packaging the vector. We thank Vincenza Bagetta for her contribute in the original design of the project. We thank Dr Valentina Ferretti, Prof. E. Illingworth and Dr G. Diez-Roux for critically revising the manuscript, and Prof. E. Illingworth and Miss A. Aliperti for English revision of the text. We also thank De Leonibus’ lab for critical revision of the manuscript. We thank Martina Colucci and Eleonora Sacco for helping with histological experiments, and IGB’ Animal House and Microscopy Facilities.

Funding

This work was supported by grants from Italian Ministry of Education University and Research (MIUR), Progetto di Ricerca di Interesse Nazionale (PRIN) 2011 (prot. 2010AHHP5H) (to P.C.), PRIN 2015 (prot. 2015FNWP34) (to E.D.L. and P.C.) and Fondazione con il Sud 2011- PDR-13 (to E.D.L. and P.C.).

Supplementary material

Supplementary material is available at Brain online.

Abbreviations

Abbreviations
 
• DAT

  dopamine active transporter

 
• GFP

  green fluorescence protein

 
• HFS

  high frequency stimulation

 
• LTD

  long-term depression

 
• LTP

  long-term potentiation

 
• hu-α-Syn

  human wild-type alpha-synuclein

 
• SNpc

  substantia nigra pars compacta

 
• VTA

  ventral tegmental area