Single Exposure to Cocaine Impairs Reinforcement Learning by Potentiating the Activity of Neurons in the Direct Striatal Pathway in Mice

Abstract

Plasticity in the glutamatergic synapses on striatal medium spiny neurons (MSNs) is not only essential for behavioral adaptation but also extremely vulnerable to drugs of abuse. Modulation on these synapses by even a single exposure to an addictive drug may interfere with the plasticity required by behavioral learning and thus produce impairment. In the present work, we found that the negative reinforcement learning, escaping mild foot-shocks by correct nose-poking, was impaired by a single in vivo exposure to 20 mg/kg cocaine 24 h before the learning in mice. Either a single exposure to cocaine or reinforcement learning potentiates the glutamatergic synapses on MSNs expressing the striatal dopamine 1 (D1) receptor (D1-MSNs). However, 24 h after the cocaine exposure, the potentiation required for reinforcement learning was disrupted. Specific manipulation of the activity of striatal D1-MSNs in D1-cre mice demonstrated that activation of these MSNs impaired reinforcement learning in normal D1-cre mice, but inhibition of these neurons reversed the reinforcement learning impairment induced by cocaine. The results suggest that cocaine potentiates the activity of direct pathway neurons in the dorsomedial striatum and this potentiation might disrupt the potentiation produced during and required for reinforcement learning.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12264-021-00687-8.

Introduction

It is well acknowledged that chronic cocaine abuse is a devastating neuropsychiatric disorder causing a wide range of emotional and cognitive deficits [1–4]. In contrast, the prolonged deleterious effects of occasional or recreational use of cocaine is still in debate [5–7]. Recent studies found that recreational users are less able to inhibit overt manual and covert attentional responses, and show more switching costs and perseverative errors [8, 9]. It has also been found that a single injection or exposure of mice to cocaine affects excitatory synapses, although this dosage of cocaine is not sufficient to make mice addicted [10–12]. Such deficits also impair a series of adaptive behaviors, including reinforcement learning [13, 14], which relies on the proper functioning of the basal ganglia-thalamo-cortical loop [15, 16]. In addition, synapses in this loop are the primary targets of many addictive drugs, including cocaine [17–19]. However, the neuropathological changes in the basal ganglia circuits induced by acute cocaine exposure and the relationship between these changes and the impairment of reinforcement learning are still unclear.

Basal ganglia circuits are essential for action selection and the encoding of action-outcome relations. Lesions or disorders affecting different nodes of these circuits lead to a variety of deficits in action initiation and goal-directed movement [17–21]. These same circuits are able to undergo plastic changes, ranging from plasticity in synaptic strength and excitability to the remodeling of microcircuits, in response to environmental challenges [22, 23]. Therefore, the basal ganglia circuits are not only important for the initiation and performance of actions, but also necessary for reinforcement learning and behavior remodeling. It has been reported that the dorsomedial striatum (DMS) is heavily involved in reinforcement learning, while the dorsolateral striatum is closely associated with the execution of habituated behaviors [20, 24]. Two major pathways constitute the cortico-striatal-basal ganglia loops: the direct pathway, a monosynaptic gabaergic projection from medium-sized spiny neurons expressing dopamine 1 (D1) receptors (D1-MSNs) to the substantia nigra pars reticulata, and the indirect pathway, a polysynaptic projection from MSNs expressing dopamine 2 (D2) receptors (D2-MSNs) to the substantia nigra pars reticulata through the external globus pallidus and subthalamic nucleus [20]. Striatal D1-MSNs and D2-MSNs are differentially modulated by dopamine and work in an antagonistic manner to facilitate or suppress movement, respectively [20]. A large body of experimentation has demonstrated that long-term plastic changes of the excitatory synapses on striatal D1-MSNs facilitate movement, positive reinforcement, and reward, while long-term plastic changes of the excitatory synapses on striatal D2-MSNs mediate inhibition of motion, negative reinforcement, and punishment [13, 14, 16], showing the fundamental roles of striatal plasticity in behavioral adaptations, including reinforcement learning [16, 25].

Single administration of cocaine has been frequently used in animal experiments to study the acute effects of the drug on the plastic changes in neurotransmission in different brain regions [12, 26, 27]. It has been demonstrated that cocaine binds with the dopamine transporter and inhibits the re-uptake of dopamine into dopaminergic terminals, therefore increasing the extracellular dopamine concentration, which in turn results in the potentiation of dopaminergic neurotransmission and an exaggerated effect of dopamine on postsynaptic neurons. The persistent neuroadaptive effects left by such an acute action mediate the emotional and behavioral changes of single cocaine exposure [28, 29]. Previous studies have found that acute exposure to cocaine exerts significant effects on the glutamatergic transmission onto D1-MSNs and/or the D2-MSNs in the nucleus accumbens and the dopaminergic neurons in the ventral tegmental area [12, 30, 31].

A few studies also indicate that both the effects of acute cocaine and the completion of reinforcement learning rely on plastic changes in the glutamatergic transmission on striatal MSNs [21, 24, 32]. Therefore, the plastic changes induced by acute exposure to cocaine may disturb the formation of the neural plasticity required for reinforcement learning, causing impairments in the learning. Negative reinforcement learning is essential to animals for behavioral adaptation, it has been widely used as a standardized experimental behavioral paradigm, and substance dependence or even single injection of cocaine affects negative reinforcement learning (NRL) [33–35], but the mechanisms are still unclear. The present work aimed to study the effects of a single in vivo exposure to cocaine on NRL in mice. To reach this goal, the changes in the glutamatergic transmission on DMS D1-MSNs and D2-MSNs induced by cocaine, reinforcement learning, and reinforcement learning at 24 h after exposure to cocaine were recorded. Since both the exposure to cocaine and the experience of reinforcement learning facilitated striatal glutamatergic transmission on DMS D1-MSNs, the sufficiency and necessity of the facilitation induced by cocaine in the learning impairment was also investigated by manipulation of the activity of DMS D1-MSNs.

Materials and Methods

Animals

Male C57BL/6J mice between 7 to 8 weeks old (from the Model Animal Research Center of Nanjing University, China) were used to record the effects of cocaine on reinforcement learning and excitatory neurotransmission on striatal MSNs. In the chemogenetic activation or inhibition of D1-MSNs, heterozygotic male progeny of dopamine 1 receptor (D1R)-Cre (Drd1-Cre, 262, Gensat) mice (gifts from the laboratory of Professor. Fuqiang Xu) were used and produced by mating transgenic male mice with C57BL/6J females. Mice were housed in a pathogen-free facility maintained at a constant temperature and on a 12-h light/dark cycle (light on from 08:00 to 20:00). Water and food were available ad libitum. Intraperitoneal injection of 20 mg/kg cocaine (Qinghai Pharmaceutical Factory, China) or 0.9% saline was given to the mice 24 h before experiments. All procedures were approved by the Medicine Animal Care and Use Committee of Shaanxi Normal University and conformed to the Guide for the National Institutional Animal Care.

Behavioral Tests

Open Field Test (OFT)

Each mouse was gently placed in a corner of an illuminated (10 lux) square box (50 × 50 × 35 cm³), facing the opaque walls. Its movements were automatically recorded for 10 min with a video camera above the box and analyzed with EthoVision software (Version 1.9, Noldus Information Technology, USA). The locomotor activity was evaluated as the total distance travelled, and anxiety-like behavior was evaluated by measuring the time spent and total number of entries in the center area.

Negative Reinforcement Learning to Escape Foot-Shocks (NRL)

NRL was performed as described previously [36]. All learning experiments were conducted in operant chambers (30 × 24 × 30 cm³; MED Associates, USA), which were positioned in sound-attenuating boxes, equipped with a ventilation fan, a house light, an observation window, and two nose-pokers located 2 cm above the metal grid floor. Briefly, a day before the reinforcement learning experiment, each mouse was placed in an operant chamber and allowed to explore the chamber freely for 100 min for adaption to the environment, during which no shock was delivered. On the day of reinforcement training, foot-shocks were delivered through the metal grid floor. A very mild intensity (0.15 mA) was used. One of the two nose-pokers was designated randomly as “active” and was illuminated by a light-emitting diode (LED, 20 lux) during the shock period and its activation triggered shock termination. Shock and the light began simultaneously and terminated whenever the learning mouse poked the active nose-poker, signaled by a 1.5-s tone (2.9 kHz, 65 dB) and turning off the LED; this was followed by a pseudorandom timeout period ranging from 30 s to 60 s. The learning procedure consisted of 50 trials, amounting to a total session duration of approximately 100 min. We defined one successful session as consisting of eight consecutive correct responses, similar to previous studies [37]. The trials continued until each mouse completed 50 trials. The total number of right or wrong (poking the inactive nose-poker) responses, and the escape latency were recorded.

Electrophysiological Recording

Brain Slice Preparation

Acute brain slices for analyzing synaptic function were prepared as previously described [38, 39]. Briefly, each mouse was anesthetized with isoflurane and decapitated. The brain was then rapidly removed and glued to a cutting stage immersed in artificial cerebrospinal fluid (ACSF) containing (in mmol/L): 125 NaCl, 2.5 KCl, 25 glucose, 25 NaHCO₃, 1.25 NaH₂PO₄, 2 CaCl₂, and 1 MgCl₂, gassed with 5% CO₂/95% O₂. Sagittal slices (300 μm) through the striatum were cut with a microslicer (VT 1200S, Leica, Germany) at an advance speed of 0.06 mm/s. The slices were then transferred to a holding chamber filled with oxygenated ACSF at 34 °C and allowed to recover for at least 0.5 h before use.

Electrophysiology

Individual slices were transferred to a recording chamber and continuously perfused with ACSF for the duration of the experiment. The DMS area in each slice was visually identified according to The Mouse Brain in Stereotaxic Coordinates (second edition). DMS MSNs were visualized under an upright microscope (DM LFSA, Leica, Germany). The recording pipettes had 3–8 MΩ resistance when filled with the RNase-free intracellular solution (in mmol/L): 140 CsCH₃SO₃, 10 HEPES, 2 QX-314, 2 MgCl₂, 0.2 EGTA, 4 MgATP, 0.3 Na₂GTP, 10 Na₂-phosphocreatine (pH 7.2–7.4 with CsOH). All experiments were carried out in the presence of 100 µmol/L picrotoxin (PTX, Sigma, USA). Series and input resistances were determined with each afferent stimulus and were monitored for stability throughout each experiment. Recordings were obtained using a Multiclamp 700B amplifier and a Digidata 1550 (Molecular Devices, USA). Data were filtered at 2 kHz and digitized at 10 kHz. After recording, D1-MSNs and D2-MSNs were identified by single-cell PCR as described in the methods for single-cell PCR. A patch pipette filled with RNase-free intracellular solution positioned close to the tissue in the recording chamber was used as the negative control.

For miniature excitatory postsynaptic current (mEPSC) recording, the cell membrane potential was clamped at −80 mV in the presence of 1 μmol/L TTX (Hebei Fishery Science and Technology Development Co., China).

For AMPAR/NMDAR (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor/N-methyl-D-aspartate receptor) ratio calculation, a stainless-steel bipolar microelectrode was located on the white matter between the cortex and the DMS close to the recording electrode and stimulated at a baseline frequency of 0.1 Hz. The AMPAR- and NMDAR-mediated current ratio was recorded in the presence of 100 μmol/L PTX at a holding membrane potential of −80 mV and +40 mV, respectively. The AMPAR/NMDAR ratio was calculated as the ratio of the average peak EPSC amplitude at −80 mV to the average EPSC amplitude recorded at +40 mV (averaged at 50 ms after afferent stimulation). All EPSCs used for analysis were averaged from 10 consecutive traces.

Long-term potentiation (LTP) was induced in the DMS in the slices by using the following high-frequency stimulation (HFS) protocol: 4 trains of 100 Hz paired with postsynaptic depolarization at 0 mV. Data were excluded when the series resistance changed by >20%. In this experiment, MgCl₂ was omitted from the ACSF.

In the pharmacological modulation study, current-clamp recording was used to measure evoked action potentials in Clozapine N-oxide (CNO; BrainVTA, China) activation or inhibition experiments. After applying currents in 25-pA steps, ranging from –100 pA to 300 pA and 1000 ms in duration, neurons were allowed to recover for 5 min before the slices were perfused with ACSF containing 10 µmol/L CNO. The same current-clamp procedure was performed 15 min after CNO perfusion. All recording sessions were recorded and analyzed using Clampfit 10.5 software (Molecular Devices, USA).

Single-Cell PCR

Single-cell PCR was performed as previously reported with minor modification [40, 41]. Briefly, after recording, the solution containing an ejected cell was transferred into a PCR tube containing 3 μL of RNase-free water and 0.5 μL of 40 U/µL RNasin (Promega, USA).

Single-strand cDNA was synthesized in PCR tubes containing 2 μL mixed dNTPs (2.5 mmol/L each), 0.5 μL oligo(dT) primer (50 μmol/L), and 0.5 μL random primer (100 μmol/L) (all from Takara, Japan). The mixture was heated to 65 °C for 5 min and then cooled on ice for 1 min, then 2.5 μL 5× RT Buffer and 0.75 μL Maxima Reverse Transcriptase (200 U/µL; Thermo Scientific, USA) were added and held at 25 °C for 10 min; 50 °C for 30 min; and 85 °C for 5 min; then kept at 4 °C.

A multiplex single-cell nested-PCR was carried out for detection of dopamine receptors (Drd1, Drd2) and glutamic acid decarboxylase 67 (GAD67). Primers and amplicons are listed in Table 1. The first-round PCR was started after adding 2× PCR Master Mix, ddH₂O, and primer 1 (4 μmol/L each) to the RT product (final volume 20 μL). Forty cycles were performed (denaturation at 94 °C, 3 min; annealing at 59 °C, 1 min; extension at 72 °C, 1 min; final elongation at 72 °C, 10 min). An aliquot (2 μL) of the first-round PCR product was used as a template for the second-round PCR (35 cycles; annealing at 58 °C, 30 s; extension at 72 °C, 30 s). The PCR reaction mix profiles were the same as in the first-round except that primer 1 was substituted by primer 2. The second-round PCR products were identified by 2% agarose gel electrophoresis. All PCR reagents were from Takara (Japan).

Table 1.

Oligonucleotide primers used for single-cell PCR.

Gene	GeneBank accession no.	Primer name	Primer sequence	Product length (bp)
Gad67	XM_011239023	Gad67-F1 Gad67-R1	TGTTCCTTTCCTGGTGAGTGC GGTAGGAAGCATGCATCTGGT	296
		Gad67-F2 Gad67-R2	CTTGGCTGTAGCTGACATCTG TGCATCAGTCCCTCCTCTCTA	207
Drd1	NM_010076	Drd1-F1 Drd1-R1	TCCGATAGTTGGGCTCATCG CTGTTGCAATACCCCCACCC	372
		Drd1-F2 Drd1-R2	ATAGTTGGGCTCATCGCTGG ACCGGGAAGGGGTTCTTCTA	222
Drd2	NM_010077	Drd2-F1 Drd2-R1	AACACACGCTACAGCTCCAA TCATGTCCTCAGGGTGGGTA	325
		Drd2-F2 Drd2-R2	CCCACTGCTCTTTGGACTCA GCTTGCGGAGAACGATGTAG	152

F, forward; R, reverse; 1, primers for first-round PCR; 2, primers for second-round PCR.

Stereotaxic Surgery and Viral Injections

Surgery was performed under anesthesia using isoflurane (4% for induction and 1% for maintenance). Each animal was then mounted in a stereotaxic frame with non-puncturing ear bars (RWD Life Science Inc., China). The viruses AAV-hSyn-DIO-hM3Dq(Gq)-mCherry (5.63 × 10¹² GC/mL, BrainVTA) or AAV-hSyn-DIO-hM4Di(Gi)-mCherry (4.72 × 10¹² GC/mL, BrainVTA) or AAV-hSyn-DIO-mCherry (5.96 × 10¹² GC/ml, BrainVTA) were bilaterally injected into the DMS through borosilicate glass pipettes connected to a 10-µL microsyringe (Gaoge, China) at the coordinates AP: 0.6 mm; ML: ± 1.5 mm; DV: 2.7 and 2.9 mm; AP: 1.0 mm; ML: ± 1.2 mm; DV: 2.6 and 2.8 mm. A total volume of 90 nL was injected at each desired depth at 30 nL per min, and the needle was held at the site for an additional 10 min. After surgery, mice were allowed to recover for 3 weeks before behavioral tests.

Statistical Analysis

All data were transferred to GraphPad Prism for analysis and graphing. Behavioral and electrophysiological data are presented as the mean ± standard error of the mean (SEM). Data were analyzed using a two-tailed unpaired t test, paired t test, or two-way or three-way repeated-measures ANOVA followed by Tukey’s multiple comparisons test. A P < 0.05 was considered significant and all data used a confidence level of 95%.

Results

Single Exposure to Cocaine Impairs Reinforcement Learning in Mice

Two groups of mice were used in this behavioral experiment: one was subjected to the behavioral tests 24 h after a single exposure to 20 mg/kg cocaine (Cocaine+NRL) and the other was subjected to the tests 24 h after saline injection (Saline+NRL). Since repeated cocaine administration induces neural adaptation in the dopamine system and causes a progressive increase in locomotor activity [42, 43], before subjecting the animals to the reinforcement learning, the effects of cocaine on the locomotor activity was first evaluated in the OFT (Fig. 1A). The total distance moved in the Cocaine+NRL group did not differ from that in the Saline+NRL group (Fig. 1C, t₁₈ = 1.636, P = 0.1192), suggesting that the cocaine exposure did not produce behavioral sensitization 24 h after administration. Then, 30 min after the OFT, the two groups were subjected to reinforcement training (Fig. 1A, B). The escape latency for the 50 training trials in the Cocaine+NRL group was significantly longer than that in the Saline+NRL group, resulting in a significant difference between the learning curves of the two groups (Fig. 1D, interaction F_{49, 900} = 2.638, P <0.0001; Trials F_{49, 900} = 7.390, P <0.0001; Treatment F_{1, 900} = 445.7, P <0.0001). Consistently, both the mean escape latency (Fig. 1E, t₁₈ = 8.109, P <0.0001) and the total number of mistakes (Fig. 1F, t₁₈ = 2.989, P = 0.0079) in the Cocaine+NRL group were larger than those in the Saline+NRL group. With the success of learning defined as the completion of eight consecutive correct responses [37], the time to reach the first successful learning session by the Cocaine+NRL group was also longer than that by the Saline+NRL group (Fig. 1G, t₁₈ = 6.437, P <0.0001). To exclude other possible cocaine-induced deficits in reinforcement learning, the effects on anxiety-like behaviors and pain sensitivity were evaluated. To assess anxiety-like behaviors, the time spent in the center zone and total number of entries in the OFT were analyzed first, and there were no significant differences in the Cocaine group before and 24 h after the single injection (Fig. S1A, B, Time in the center zone (Fig. S1A): interaction F_{1, 36} = 0.0283, P = 0.8675; Treatment F_{1, 36} = 0.3395, P = 0.5637; Pre & Post F_{1, 36} = 0.1271, P = 0.7235; Total number of entries (Fig. S1B): interaction F_{1, 36} = 0.2639, P = 0.6106; Treatment F_{1, 36} = 0.9214, P = 0.3435; Pre & Post F_{1, 36} = 0.1757, P = 0.6776). In addition, in the elevated plus maze test, there were no significant differences between two groups in the time spent in the open arms (Fig. S1C, interaction F_{1, 36} = 0.7687, P = 0.3864; Treatment F_{1, 36} = 0.1199, P = 0.7312; Pre & Post F_{1, 36} = 0.1125, P = 0.7392) and the percentage of entries before and 24 h after the single injection (Fig. S1D, interaction F_{1, 36} = 3.124, P = 0.0856; Treatment F_{1, 36} = 0.0143, P = 0.9054; Pre & Post F_{1, 36} = 0.2800, P = 0.6000). To test if the cocaine-treated animals were less sensitive to the electric shock, which might lead to the prolonged escape latency, we compared the sensitivity to thermal and mechanical stimuli before and 24 h after the single injection of cocaine; the results showed that the responses to thermal and mechanical stimuli did not differ between the groups (Figs. S2A–B, Thermal withdrawal latency (Fig. S2A): interaction F_{1, 36} = 0.2108, P = 0.6489; Treatment F_{1, 36} = 0.0030, P = 0.9566; Pre & Post F_{1, 36} = 1.244, P = 0.2721; Mechanical withdrawal threshold (Fig. S2B): interaction F_{1, 36} = 0.5991, P = 0.4440; Treatment F_{1, 36} = 2.317, P = 0.1367; Pre & Post F_{1, 36} = 3.226, P = 0.0809). Based on the above results, effects of anxiety-like behaviors and pain sensitivity on the Cocaine group could be excluded. During the habituation period, there was no significant difference between the total number of spontaneous nose-pokes on the active and inactive nose-pokers in the Saline+NRL and Cocaine+NRL groups, showing no preference for either of the nose-pokers (Fig. 1H, interaction F_{1, 36} = 0.0002, P >0.9999; Nosepoke F_{1, 36} = 0.1086, P = 0.7436; Treatment F_{1, 36} = 0.8473, P = 0.3634). In contrast, the total numbers on the active nose-pokers in the two groups were significantly larger than those on the inactive nose-pokers, showing the preference for the active nose-poker induced by the mild aversive foot-shocks. In addition, the total number of active nose-pokes in the Cocaine+NRL group was significantly lower than that in the Saline+NRL group, indicating that the Cocaine+NRL group had fewer correct responses than the Saline+NRL group (Fig. 1I, interaction F_{1, 36} = 11.42, P = 0.0018; Nose-poke F_{1, 36} = 111.0, P <0.0001; Treatment F_{1, 36} = 18.99, P = 0.0001; followed by Tukey’s post hoc test: Saline+NRL vs Cocaine+NRL, active nose-pokes P <0.0001). The above results clearly showed that the reinforcement learning was markedly impaired by a single dose of cocaine administered 24 h before the learning. The possible mechanism underlying the impairment was then studied by using electrophysiological recordings in brain slices 30 min after the learning (Fig. 1A).

Fig. 1.

Single cocaine injection-induced impairment in reinforcement learning in mice. A Experimental schedule for cocaine administration and behavioral tests. B The operant chamber. C Open field test (OFT): there is no difference in total distance between the Saline+NRL and Cocaine+NRL groups. D During the 50 trials of reinforcement learning, the escape latency curve is significantly increased by a single dose of cocaine administered 24 h before the learning. E–G Mean escape latency (E), total numbers of errors (F), and total time to criterion (G) are significantly greater in the Cocaine+NRL group than in the Saline+NRL group. H There is no preference between the active and the inactive nose-pokers in the Saline+NRL and Cocaine+NRL groups during the habituation period. I The number of active nose-pokes are significantly larger than that of the inactive nose-poker in both groups, while the number of active nose-pokes is larger in the Saline+NRL group than in the Cocaine+NRL group during the learning period (Saline+NRL, n = 10 mice; Cocaine+NRL, n = 10 mice). Data represent the mean ± SEM (*P <0.05, **P <0.01, ***P < 0.001; two-tailed unpaired t test in C, E, F, G and two-way RM ANOVA in D, H and I).

Single Exposure to Cocaine Enhances Striatal Glutamatergic Transmission and Prevents the Plastic Changes Produced by Reinforcement Learning

The basal glutamatergic transmission on DMS D1-MSNs and D2-MSNs in the DMS slices was evaluated by whole-cell voltage clamp recording of mEPSCs in the following groups: Saline, Cocaine, Saline+NRL, and Cocaine+NRL. In addition to the Saline+NRL and Cocaine+NRL groups used in the behavioral tests described above, slices were also prepared from the Saline and the 20 mg/kg Cocaine groups, prepared in the same way as the NRL groups before slicing but without NRL. D1-MSNs and D2-MNSs were identified post hoc by single-cell PCR using cytoplasm collected immediately after recording (Fig. 2A, D, and G).

Fig. 2.

Single exposure to cocaine inhibits the potentiation of glutamatergic transmission on D1-MSNs produced during reinforcement learning. A, D Representative recordings of mEPSCs from D1-MSNs (A) and D2-MSNs (D) in the groups with saline, cocaine, reinforcement learning after saline (Saline+NRL), and reinforcement learning after cocaine (Cocaine+NRL). B, E Summary data for mEPSC frequency with cumulative probability plots of inter-event intervals in D1-MSNs (B) and D2-MSNs (E). C, F Summary of mEPSC amplitude with cumulative probability plots in D1-MSNs (C) and D2-MSNs (F). G Representative image of the agarose gel electrophoresis of single-cell PCR with nested primers applied to single GABAergic MSNs with Dopamine type 1 receptor (Drd1) and Dopamine type 2 receptor (Drd2) (NC: Negative control). Double-expressing cells in single-cell PCR were not counted. D1-MSNs: Saline, n = 20 cells from 10 mice; Cocaine, n = 23 cells from 10 mice; Saline+NRL, n = 26 cells from 10 mice; Cocaine+NRL, n = 17 cells from 10 mice. D2-MSNs: Saline, n = 19 cells from 11 mice; Cocaine, n = 10 cells from 6 mice; Saline+NRL, n = 22 cells from 10 mice; Cocaine+NRL, n = 21 cells from 10 mice. *P <0.05, **P <0.01, ***P <0.001, two-way RM ANOVA.

The frequency of mEPSCs in the Saline+NRL group was significantly greater than that in the Saline and the Cocaine groups (Fig. 2B, interaction F_{1, 82} = 7.347, P = 0.0082; NRL F_{1, 82} = 14.48, P = 0.0003; Treatment F_{1, 82} = 9.263, P = 0.0031 in D1-MSNs: followed by Tukey’s post hoc test, P = 0.00004 vs Saline; P = 0.000008 vs Cocaine), showing that the experience of reinforcement learning enhanced the excitatory transmission on D1-MSNs by increasing the presynaptic release of glutamate. There was no significant difference between the frequency of mEPSCs in the Cocaine group and that in the Saline group (P = 0.9952), suggesting that the presynaptic release of glutamate was not enhanced by exposure to cocaine. Surprisingly, the increase in the frequency of mEPSCs in D1-MSNs in the Saline+NRL group was absent from the Cocaine+NRL group (P = 0.0007), suggesting that exposure to cocaine prevented the increase of the presynaptic release of glutamate produced during reinforcement learning.

The average mEPSC amplitude in the Saline+NRL group was significantly larger than that in the Saline and Cocaine+NRL groups (Fig. 2C, interaction F_{1, 82} = 27.87, P <0.0001, NRL F_{1, 82} = 1.627, P = 0.2057, Treatment F_{1, 82} = 1.629, P = 0.2054; followed by Tukey’s post hoc test, vs Saline, P <0.0001; vs Cocaine+NRL, P = 0.0315), showing that reinforcement learning enhanced the excitatory transmission on D1-MSNs by increasing the amplitude of the postsynaptic current. The average mEPSC amplitude in the Cocaine group was also larger than that in the Saline (P <0.0001) and Cocaine+NRL groups (p = 0.0377) (Fig. 2C), indicating that exposure to cocaine also enhanced the excitatory transmission on D1-MSNs by increasing the postsynaptic current. There was no significant difference between the average mEPSC amplitude in the Cocaine+NRL group and that in the Saline group (Fig. 2C, P = 0.3358), showing that the enhancement of postsynaptic current induced by reinforcement learning was also prevented by administration of cocaine 24 h before learning.

The above results showed that a single exposure to cocaine enhanced the postsynaptic current of glutamatergic synapses on DMS D1-MSNs and this enhancement may prevent the formation of the increase in both the presynaptic glutamate release and the postsynaptic current that were produced during reinforcement learning (Fig. 1).

On the other hand, no significant differences were found among the values of frequency and average amplitude of mEPSCs in D2-MSNs among the Saline, Saline+NRL, Cocaine, and the Cocaine+NRL groups, which suggests that the excitatory transmission on D2-MSNs might not be involved in the impairment of reinforcement learning induced by exposure to cocaine (Fig. 2E, F; Frequency: interaction F_{1, 68} = 3.249, P = 0.0759; NRL F_{1, 68} = 0.0091, P = 0.9245; Treatment F_{1, 68} = 0.0893, P = 0.7660; Amplitude: interaction F_{1, 68} = 5.487, P = 0.0221; NRL F_{1, 68} = 0.2401, P = 0.6257; Treatment F_{1, 68} = 0.5894, P = 0.4453).

To further confirm the above changes in the average amplitude of mEPSCs, the ratio of the amplitudes of evoked AMPAR- and NMDAR-EPSCs (A/N ratio) in D1-MSNs and D2-MSNs was recorded in DMS slices (Fig. 3A, C). In D1-MSNs, the A/N ratio in the Saline+NRL group was significantly larger than those in the Saline and the Cocaine+NRL groups (Fig. 3B; interaction F_{1, 70} = 32.90, P <0.0001; NRL F_{1, 70} = 0.2215, P = 0.6393; Treatment F_{1, 70} =2.177, P = 0.1446, followed by Tukey’s post hoc test: vs Saline, P = 0.0039; vs Cocaine+NRL, P <0.0001), further suggesting that reinforcement learning enhanced the excitatory transmission on D1-MSNs by increasing the AMPAR-mediated postsynaptic current. The A/N ratio in D1-MSNs in the Cocaine group was also significantly larger than those in the Saline and the Cocaine+NRL groups (Fig. 3B; vs Saline, P = 0.0345; vs Cocaine+NRL, P <0.0001), confirming that exposure to cocaine also enhanced the excitatory transmission on D1-MSNs by increasing the AMPAR-mediated postsynaptic current. There was no significant difference between the A/N ratio in D1-MSNs in the Cocaine+NRL group and that in the Saline group (Fig. 3B; P = 0.5528).

Fig. 3.

Single exposure to cocaine enhances the AMPAR- and NMDAR-EPSCs ratio, resulting in disruption of HFS-induced LTP in D1-MSNs, then inhibits the potentiation of the ratio by reinforcement learning. A, C Representative traces of NMDAR EPSCs at +40 mV (upper traces) and AMPAR EPSCs at –80 mV (lower traces) in D1-MSNs (A) and D2-MSNs (C) from the Saline, Cocaine, Saline+NRL, and Cocaine+NRL groups. B, D Statistics of the ratio of AMPAR to NMDAR EPSCs in D1-MSNs (B) and D2-MSNs, (D) corresponding to A and C, respectively (D1-MSNs: Saline, n = 13 cells from 7 mice; Cocaine, n = 18 cells from 7 mice; Saline+NRL, n = 21 cells from 9 mice; Cocaine+NRL, n = 22 cells from 9 mice. D2-MSNs: Saline, n = 22 cells from 8 mice; Cocaine, n = 14 cells from 7 mice; Saline+NRL, n = 21 cells from 10 mice; Cocaine+NRL, n = 15 cells from 9 mice). E HFS induces LTP in DMS D1-MSNs in single-dose saline- or cocaine-treated mice. Left: schematic of a striatal slice with the stimulating electrode (Stim) and recording electrode (Record) in the dorsomedial subregion (DMS); right: dark traces represent the baseline EPSC average from 0 to 10 min (labeled “1”), and light traces represent the average EPSC from the last 10 min after LTP induction (labeled “2”). F Summary of the magnitude of HFS-LTP induction in the Saline and Cocaine groups (comparison between baseline and the last 10 min of recording). (D1-MSNs, Saline, n = 11 cells from 5 mice; Cocaine, n = 10 cells from 5 mice; *P <0.05, **P <0.01, ***P <0.001, two-way RM ANOVA for B, D, E, paired t test for F).

Consistent with the findings for the average amplitude of mEPSCs, in D2-MSNs, there were also no significant differences in the A/N ratio of the postsynaptic current among the Saline, Cocaine, Saline+NRL, and Cocaine+NRL groups (Fig. 3D; interaction F_{1, 68} = 3.701, P = 0.0586; NRL F_{1, 68} = 4.135, P = 0.0459; Treatment F_{1, 68} = 0.6395, P = 0.4267), suggesting that the excitatory transmission on D2-MSNs might not play a key role in the impairment of reinforcement learning after exposure to cocaine.

Alterations of synaptic AMPAR and NMDAR subunits on the postsynaptic membrane contribute to the expression of long-term changes in synaptic strength, as in LTP and LTD [28]. To further explore whether the history of synaptic activation caused by single-dose cocaine interferes with the induction of potentiation by NRL, we examined HFS LTP induction in D1-MSNs in vitro after single-dose cocaine exposure. We found that the HFS protocol (4 trains of 100 Hz given at 10-s intervals paired with depolarization of the neuron to 0 mV) failed to elicit LTP in D1-MSNs approximately 24 h after cocaine exposure; instead, this protocol tended to produce LTD (Fig. 3E–F, eEPSC amplitude in the last 10 min relative to baseline, Saline: 155.9 ± 4.20%, t₁₀ = 4.337, P = 0.0015; Cocaine: 75.25 ± 2. 32%, t₉ = 3.725, P = 0.0047; Saline vs Cocaine 40–50 min t₁₉ = 5.375, P <0.0001). The results showed that exposure to cocaine enhanced the excitatory transmission in DMS D1-MSNs and thereby prevented further synaptic potentiation in the same population. In addition, this potentiated state elicited by cocaine may leave an enduring trace that affects the subsequent induction of plasticity, such as the response to the HFS protocol or activity-dependent paradigms (NRL as in Fig. 1). If cocaine-evoked potentiation is causally involved in the impairment of NRL behavior, then potentiating these synapses may artificially simulate the role of cocaine in the learning behavior deficits, while de-potentiating these synapses may reverse the behavioral change.

Activation of D1-MSNs is Sufficient for the Impairment of Reinforcement Learning

The results presented above strongly suggest that a single exposure to cocaine enhances the excitatory transmission on DMS D1-MSNs and impairs reinforcement learning in mice. As both LTP and LTD are depolarization-dependent, which might arise from alterations in the activity of postsynaptic neurons, both synaptic transmission and the postsynaptic neuronal activity jointly contribute to the activity level changes of the striatal direct pathway. To comprehensively assess the role of activity in neurons of the direct striatal pathway (D1-MSNs) in the impairment of NRL by cocaine, we used the designer receptors exclusively activated by designer drugs (DREADD) approach (Fig. 4A). We first tested whether the activation of D1-MSNs, mimicking the effects of cocaine, produces an impairment of reinforcement learning similar to that induced by cocaine. AAV viral vectors expressing the Gq-coupled human M3 muscarinic receptor (hM3Dq) or AAV-hSyn-DIO-mCherry were injected bilaterally into the DMS in D1-Cre mice [44, 45], targeting the D1-MSNs (Fig. 4B). Three weeks after virus injection, the spiking response to current stimulation in the hM3Dq-expressing D1-MSNs was significantly increased in brain slices by bath application of CNO (10 µmol/L), while that in the control mCherry-expressing D1-MSNs was unaffected (Fig. 4C). The rheobase of spike generation in the hM3Dq-expressing D1-MSNs was reduced by CNO (Fig. 4D; t₉ = 7.584, P <0.0001), while that in the mCherry-expressing D1-MSNs was unaffected (t₈ = 0.000, P >0.9999). The excitability of the hM3Dq-expressing D1-MSNs was significantly enhanced by CNO, showing that the application of CNO activated the hM3Dq-expressing D1-MSNs, compared to the pre-CNO baseline (Fig. 4E; F_{1, 216} = 110.4, P <0.0001).

Fig. 4.

Activating D1-MSNs with DREADD-hM3Dq impairs reinforcement learning in D1-Cre mice. A Experimental paradigm. B Injection site of AAV-DIO-hM3Dq-mCherry virus in the DMS. C Current-voltage relationship of representative D1-MSNs recorded before and after 10 µmol/L CNO perfusion. D The minimal injected current to induce action potentials (Aps) is decreased by CNO. E Number of Aps induced at different current steps (mCherry, n = 9 cells from 3 mice; hM3Dq, n = 10 cells from 3 mice). F Open field test: there is no difference in total distance traveled among the groups. G Escape latency is significantly increased in mice expressing hM3Dq after injection of CNO. H–J Mean escape latency (H), error numbers (I), and total time to the first successful session (J) in hM3Dq-expressing mice with CNO are significantly larger than those in the other three groups. K There is no preference between the active and inactive nose-poker in the mCherry and hM3Dq groups without saline or CNO during the habituation period of 100 min. L The number of active nose-pokes is significantly greater than that of inactive nose-pokes in the mCherry-expressing group and hM3Dq-expressing mice treated with saline, but the number of active nose-pokes in the hM3Dq-expressing group treated with CNO did not differ from that of inactive nose-pokes (Saline-mCherry, CNO-mCherry, Saline-hM3Dq n = 7 mice; CNO-hM3Dq n = 9 mice). Data represent the mean ± SEM (*P <0.05; **P <0.01; ***P <0.001, paired t test for D, two-way RM ANOVA for E, F, H, I, and J, and three-way RM ANOVA for G, K, and L.

Then, in vivo injection of CNO into mice expressing hM3Dq or mCherry was used to assess the effects of D1-MSN activation on reinforcement learning. In the OFT carried out 30 min after CNO injection (0.5 mg/kg) [46], there was no significant difference in the total distance traveled by mice expressing hM3Dq or mCherry with or without CNO (Fig. 4F; interaction F_{1, 26} = 0.6300, P = 0.4345; Treatment F_{1, 26} = 0.6295, P = 0.4347; Groups F_{1, 26} = 0.6109, P = 0.4415). Interestingly, during the 50 training trials, the escape latency in the hM3Dq-expressing mice with CNO was significantly longer than in those expressing Cherry or hM3Dq with saline, resulting in marked differences among the four learning curves (Fig. 4G; Trials × Treatment × Groups F_{49, 1274} = 0.7306, P = 0.9175; Treatment × Groups F_{1, 26} = 31.07, P <0.0001; Treatment F_{1, 26} = 30.68, P <0.0001; Groups F_{1, 26} = 35.30, P <0.0001). In addition, in the hM3Dq-expressing mice, the mean escape latency (Fig. 4H; interaction F_{1, 26} = 30.87, P <0.0001; Treatment F_{1, 26} = 34.92, P <0.0001; Groups F_{1, 26} = 30.66, P <0.0001), the total error numbers (Fig. 4I; interaction F_{1, 26} = 4.872, P = 0.0363; Treatment F_{1, 26} = 10.00, P = 0.0040; Groups F_{1, 26} = 9.259, P = 0.0053), and the time to the first successful session (Fig; 4J, interaction F_{1, 26} = 11.68, P = 0.0021; Treatment F_{1, 26} = 19.63, P = 0.0002; Groups F_{1, 26} = 14.15, P = 0.0009) were all significantly larger than those in those expressing Cherry or hM3Dq with saline. During the habituation period, no mice showed a preference for either the active or the inactive nose-poker (Fig. 4K; Nosepoke × Treatment × Groups F_{1, 26} = 0.0786, P = 0.7815; Nosepoke × Treatment F_{1, 26} = 0.0210, P = 0.8859; Nosepoke × Groups F_{1, 26} = 0.7586, P = 0.3917). However, during reinforcement learning, the number of active nose-pokes was still significantly higher than that of inactive nose-pokes in the mice expressing mCherry or hM3Dq with saline (Fig. 4L, Nosepoke × Treatment × Groups F_{1, 26} = 12.53, P = 0.0015; Nosepoke × Treatment F_{1, 26} = 70.76, P <0.0001; Nosepoke × Groups F_{1, 26} = 18.26, P = 0.0002; followed by Tukey’s post-hoc test, mCherry with Saline P <0.0001, with CNO P <0.0001; hM3Dq with Saline P <0.0001), but exhibited no difference in the hM3Dq-expressing mice treated with CNO (P >0.9999). These results indicate that selective activation of D1-MSNs in the DMS impairs reinforcement learning, similar to the effects of a single exposure to cocaine (Fig. 1).

Inhibiting D1-MSNs Reverses the Impairment of Reinforcement Learning Induced by Cocaine

In order to verify the necessity of the enhancement of D1-MSN activation in the impairment of reinforcement learning induced by exposure to cocaine, AAV viral vectors expressing the Gi/o-coupled human M4 muscarinic receptor (hM4Di) or AAV-hSyn-DIO-mCherry was bilaterally injected into the DMS [44, 45] in D1-Cre mice (Fig. 5A, B). Three weeks after the virus injection, the spiking response to current injection in D1-MSNs expressing hM4Di in brain slices was inhibited by bath application of CNO (10 µmol/L), while that in mCherry-expressing D1-MSNs was unaffected (Fig. 5C). The rheobase of spike generation in the hM4Di-expressing D1-MSNs was significantly increased by CNO (Fig. 5D, t₉ = 4.707, P = 0.0011), while that in the mCherry-expressing D1-MSNs was unaffected (t₇ = 1.000, P = 0.3506). The excitability of D1-MSNs expressing hM4Di was significantly reduced by CNO (Fig. 5E, F1, 216 = 10.00, P = 0.0018), showing that application of CNO inhibits the activity of hM4Di-expressing neurons.

Fig. 5.

Inhibiting the activity of D1-MSNs with the DREADD hM4Di reverses the impairment in reinforcement learning induced by exposure to cocaine. A The experimental paradigm. B Injection site of AAV-DIO-hM4Di-mCherry virus in the DMS. C Current-voltage relationship of representative D1-MSNs recorded before, and after 10 µmol/L CNO perfusion. D The minimal injected current to induce action potentials (Aps) is increased by CNO. E Number of induced Aps at different current steps (mCherry, n = 8 cells from 3 mice; hM4Di, n = 10 cells from 3 mice). F Open field test: the total distance traveled significantly decreases after CNO perfusion in the hM4Di group vs other groups. G Escape latency is reversed by CNO in hM4Di-expressing mice, but not in other groups. H–J Mean escape latency (H), error numbers (I), and total time to the first successful session (J) in the CNO+hM4Di group are significant smaller than those in the other three groups. K There is no preference between the active and inactive nose-pokers in the mCherry and hM4Di groups without saline or CNO during the habituation period of 100 min. L The number of active nose-pokes is significantly greater than that of inactive nose-pokes in the CNO+hM4Di group (Saline-mCherry, CNO-mCherry, Saline-hM4Di n = 7 mice; CNO-hM4Di n = 9 mice). Data represent the mean ± SEM (*P <0.05, **P <0.01, ***P <0.001, paired t test for D, two-way RM ANOVA for E, F, H, I, and J, and three-way RM ANOVA for G, K, and L).

We next investigated the effects of D1-MSN suppression on cocaine-induced reinforcement learning impairment through in vivo injection of CNO into mice expressing hM4Di or mCherry. Twenty-four hours after cocaine exposure and 30 mins after CNO injection, the total distance moved by the mice in the CNO+hM4Di group in the OFT was significantly shorter than that by the mice in the other three groups (Fig. 5F, interaction F_{1, 26} = 14.78, P = 0.0007; Treatment F_{1, 26} = 12.23, P = 0.0017; Groups F_{1, 26} = 11.53, P = 0.0022). This suggests that inhibition of D1-MSNs by CNO at 0.5 mg/kg significantly reduces the locomotor activity of hM4Di-expressing mice. In reinforcement training following the OFT, as shown in Fig 5G, the escape latency in all 50 training trials in the hM4Di-expressing mice treated with CNO was significantly shorter than that of the other three groups. This resulted in marked differences among the four learning curves (Trials × Treatment × Groups F_{49, 1274} = 0.7471, P = 0.9013; Treatment × Groups F_{1, 26} = 11.59, P = 0.0022; Treatment F_{1, 26} = 5.304, P = 0.0295; Groups F_{1, 26} = 4.918, P = 0.0355). Furthermore, in the hM4Di-expressing mice with CNO, the mean escape latency (Fig. 5H; interaction F_{1, 26} = 7.374, P = 0.0116; Treatment F_{1, 26} = 6.380, P = 0.0180; Groups F_{1, 26} = 2.610, P = 0.1183), the total error numbers (Fig. 5I; interaction F_{1, 26} = 4.571, P = 0.0421; Treatment F_{1, 26} = 2.972, P = 0.0966; Groups F_{1, 26} = 9.770, P = 0.0043), and the time to the first successful session (Fig. 5J; interaction F_{1, 26} = 6.486, P = 0.0171; Treatment F_{1, 26} = 4.844, P = 0.0368; Groups F_{1, 26} = 3.292, P = 0.0812) were all significantly lower than those in the mice expressing mCherry with saline or CNO, and mice expressing hM4Di with saline. During the habituation period, no mice showed a preference for either the active or the inactive nose-poker (Fig. 5K; Nosepoke × Treatment × Groups F_{1, 26} = 1.710, P = 0.2025; Nosepoke × Treatment F_{1, 26} = 0.0244, P = 0.8770; Nosepoke × Groups F_{1, 26} = 1.648, P = 0.2105). However, during reinforcement training, the number of active nose-pokes was significantly higher than that of inactive nose-pokes in the hM4Di-expressing mice treated with CNO. To evaluate whether the effect of CNO on reinforcement learning in hM4Di-expressing mice was related to the reduction of locomotor activity, the total numbers of the nose-pokes were checked, and there were no significant differences in all four groups after CNO injection (Fig 5L). These results indicated that the effect on reinforcement learning by DREADD inhibition was not associated with decreased locomotor activity in this study (Fig. 5L; Nosepoke × Treatment × Groups F_{1, 26} = 0.2791, P = 0.6018; Nosepoke × Treatment F_{1, 26} = 3.566, P = 0.0702; Nosepoke × Groups F_{1, 26} = 0.7986, P = 0.3797; followed by post hoc Tukey’s test, mCherry with Saline P = 0.8712, with CNO P = 0.1765; hM4Di with Saline P = 0.6847; hM4Di with CNO P = 0.0022). The above results showed that in mice whose DMS D1-MSNs expressed mCherry or hM4Di without CNO injection, the impairment of reinforcement learning by cocaine was preserved, while in mice whose DMS D1-MSNs expressed hM4Di and were inhibited by CNO, the impairment was reversed (Fig. 1). These results indicate that enhancement of DMS D1-MSN activity is required for the impairment in reinforcement learning induced by a single exposure to cocaine.

Discussion

Reward-based or aversion-induced reinforcement of certain behaviors is essential for survival in a changing environment. Such behavioral adaptation relies on feedback modulation by the outcomes of immediately previous actions. The basal ganglia circuit plays a key role in linking the outcomes produced by previous actions to the selection and adjustment of future actions, thus enabling the organism to approach profitable and avoid harmful environmental events [20, 47, 48]. Neurotransmission in the striatum, the input nuclei of the basal ganglia circuit, is sensitive to the feedback information conveyed by midbrain dopamine signals. This sensitivity guarantees the flexibility of behavioral remodeling [21]. Unfortunately, the same sensitivity to modulation at this node also endows the basal ganglia circuit with vulnerability to addictive drugs. For example, even a single exposure to cocaine produces prolonged impairments in adaptive behaviors by inducing plastic changes in brain regions mediating reinforcement learning [1, 6, 9, 10, 33, 34]. In present study, using a classic instrumental learning paradigm in mice, we provide further novel evidence showing that a single exposure to cocaine remarkably impairs reinforcement learning even 24 h after administration of the drug (Fig. 1). Previous studies have reported that peak cocaine values occur approximately 5–15 min after drug administration, and cocaine half-life ranges from 16 to 72 min depending on species, drug dose, and experimental conditions [49]. Although cocaine is metabolized rapidly, its metabolites including benzoylecgonine and ecgonine methyl ester are still detectable 24 h after the first cocaine exposure in humans [50]. These results indicate that a single exposure to cocaine, either occasionally taken or used for recreational purposes, causes significant deficits in behavioral learning even after the drug is metabolized. The persistent traces left by the drug exist in the brain and can still strongly interfere in the normal functioning of neural circuits mediating adaptive behavioral learning.

Previous studies have reported that potentiation in excitatory synapses on striatal D1-MSNs mediate the positive, reward-induced, reinforcement learning [13, 20]. In the present study, we further found that potentiation of both the frequency and amplitude of mEPSCs in glutamatergic synapses on DMS D1-MSNs, but not on D2-MSNs, was produced during NRL. The postsynaptic A/N ratio in D1-MSNs was also enhanced. The results showed that potentiation of excitatory transmission in direct striatal pathway neurons is essential for the completion of both positive and NRL (Figs. 2, 3). We also found a significant enhancement in the amplitude of mEPSCs in D1-MSNs, but not in D2-MSNs, was produced by a single exposure to cocaine. This result is similar to previous reports on ventral tegmental area dopamine neurons and locus coeruleus norepinephrine neurons, in which a single exposure to cocaine was found to enhance the AMPAR-mediated postsynaptic current, but did not affect the presynaptic glutamate release probability [12, 28]. Most importantly, previous studies indicated that acute cocaine exposure is sufficient to alter the AMPAR subunit composition in D1-MSNs in ventral striatum that may contribute to the synaptic potentiation [51, 52].

Interestingly, when mice were subjected to reinforcement learning 24 h after exposure to cocaine, the learning was significantly impaired. In electrophysiological recordings immediately after the reinforcement learning, enhancement in the frequency and amplitude of mEPSCs in DMS D1-MSNs was found in mice pre-exposed to saline but was absent in those pre-exposed to cocaine. These results suggest that enhancement of the AMPAR-mediated postsynaptic current induced by cocaine interferes with or inhibits the presynaptic and postsynaptic potentiation produced during reinforcement learning, and this disruption might be an important reason for the impairment of reinforcement learning induced by exposure to cocaine. The mechanisms underlying this inhibition remain to be elucidated.

Ample evidence has shown that the efficiency of synaptic transmission in the central nervous system is dynamically modulated in many physiological and pathophysiological processes [13, 14, 24, 53, 54]. Previous studies have demonstrated that de-potentiation from potentiated synaptic strength is involved in a series of normal and abnormal states. For example, when the glutamatergic transmission in the hippocampus is enhanced to a potentiated level, experience or a stimulation protocol normally resulting in LTP can only induce further de-potentiation, reversing the established potentiation to the previous base-line transmission level. Such de-potentiation mechanisms have been strongly implicated in erasing hippocampal-dependent fear memories [55, 56]. Consistent with the mechanism responsible for this deficiency of LTP reported by previous studies, our results confirmed that a single cocaine injection was sufficient to drive the inhibition of subsequent LTP in D1-MSNs (Fig. 3E, F). We hypothesize that single-dose cocaine, as a priming signal, contributes to the neuronal activity, persistently altering the response to a subsequent plasticity-inducing event, such as NRL. This synaptic plasticity might be the reason why mice pre-treated with cocaine perform poorly during NRL. To test this hypothesis, we used DREADD systems that could selectively activate or depress D1-MSNs in the DMS. First, the impairment of reinforcement learning by selective activation of D1-MSNs expressing hM3Dq (Fig. 4) might result from a mechanism similar to the reported de-potentiation. After enhancement of the D1-MSN activity, the experience of reinforcement learning normally resulting in potentiation in the glutamatergic synaptic transmission on those D1-MSNs might be reversed to de-potentiate those synapses and thus interfere with the dynamic modulation of synaptic plasticity during the learning process, leading to an impairment of learning. It is well established that cocaine strengthens the action of dopamine on D1-MSNs through occupying dopamine transporters and increasing the extracellular concentration of dopamine. Activation of dopamine D1 receptors then potentiates glutamatergic transmission on the D1-MSNs by increasing the calcium concentration in the postsynaptic cytoplasm and up-regulating the AMPAR-mediated postsynaptic current [25, 30, 57]. In such a potentiated condition induced by cocaine, the experience of reinforcement learning producing potentiation of glutamatergic transmission on D1-MSNs might be reversed to induce de-potentiation of the established potentiation state. Although this assumption is supported by the present results showing that the inhibition of hM4Di-expressing D1-MSNs successfully reversed the impairment of reinforcement learning by cocaine (Fig. 5), it still needs to be verified with cellular and molecular evidence in future. Previous studies found that the administration of CNO at 0.7 mg/kg is a commonly used dose in mice, which significantly reduces their locomotor activity [46]. In our results, administration of CNO at 0.5 mg/kg also caused inhibition of locomotor activity in hM4Di-expressing mice, but not in hM3Dq-expressing mice. On the other hand, we found the dose of 0.5 mg/kg CNO combined with D1R significantly improved the behavior of mice in NRL. And the total number of the nose-pokes were not affected in the CNO-treated hM4Di-expressing mice (Fig. 5L), indicating that the nose-poke-related behaviors were not reduced. These results indicated that the inhibition of locomotor activity in the CNO-treated hM4Di-expressing mice was not the reason for the improvement in NRL. However, further studies are needed to answer this question.

In a number of previous studies, an artificial bacterial chromosome (BAC) was introduced into mice to label D1-MSNs or D2-MSNs with GFP [58, 59]. Some of those studies also suggested that there could be behavioral abnormalities in the BAC transgenic mice [60, 61]. In order to study the neural plasticity in D1-MSNs and D2-MSNs using mice with an identical wild-type genotype, the present work collected the cytoplasm content immediately after the patch-clamp recording and assessed the transcription levels of dopamine D1 and D2 receptors (Fig. 2). This strategy successfully identified the two types of neurons, with ratios of the two subpopulations (data not shown) very similar to previous reports [17, 62]. In the present study, intraperitoneal delivery of cocaine at 20 mg/kg was chosen through a set of preliminary experiments to assess the effects on reinforcement learning. This dose is also consistent with those used in a number of animal studies [63, 64]. The effects of exposure to cocaine on reinforcement learning in the present study is also comparable with a series of previous human studies, which reported the impairment in various instrumental and go-directed behavior patterns [34, 65]. However, the present study provided electrophysiological evidence indicating that dynamic modulation of the excitatory transmission on direct striatal pathway neurons may underpin the effects of cocaine. By specifically manipulating the activity of D1-MSNs, we also demonstrated that activation of those neurons is sufficient to impair reinforcement learning, while their suppression reverses the effects of cocaine, showing that the enhancement of direct striatal pathway activity is required for the effects of cocaine.

Supplementary Information

Below is the link to the electronic supplementary material.

Conflict of interest

The authors declare no conflicts of interest.