Chronic nicotine impairs sparse motor learning via striatal fast‐spiking parvalbumin interneurons

Abstract Nicotine can diversely affect neural activity and motor learning in animals. However, the impact of chronic nicotine on striatal activity in vivo and motor learning at long‐term sparse timescale remains unknown. Here, we demonstrate that chronic nicotine persistently suppresses the activity of striatal fast‐spiking parvalbumin interneurons, which mediate nicotine‐induced deficit in sparse motor learning. Six weeks of longitudinal in vivo single‐unit recording revealed that mice show reduced activity of fast‐spiking interneurons in the dorsal striatum during chronic nicotine exposure and withdrawal. The reduced firing of fast‐spiking interneurons was accompanied by spike broadening, diminished striatal delta oscillation power, and reduced sample entropy in local field potential. In addition, chronic nicotine withdrawal impaired motor learning with a weekly sparse training regimen but did not affect general locomotion and anxiety‐like behavior. Lastly, the excitatory DREADD hM3Dq‐mediated activation of striatal fast‐spiking parvalbumin interneurons reversed the chronic nicotine withdrawal‐induced deficit in sparse motor learning. Taken together, we identified that chronic nicotine withdrawal impairs sparse motor learning via disruption of activity in striatal fast‐spiking parvalbumin interneurons. These findings suggest that sparse motor learning paradigm can reveal the subtle effect of nicotine withdrawal on motor function and that striatal fast‐spiking parvalbumin interneurons are a neural substrate of nicotine's effect on motor learning.


| INTRODUCTION
Tobacco addiction is the main cause of preventable death worldwide. 1 Understanding the behavioral consequences of nicotine and the neural mechanism thereof are critical to treatment development. A core characteristic of nicotine is its diverse and complex impact on behavior, [2][3][4] which suggests the necessity for fine examination of nicotine's effect on phenotype. Recently, nicotine has been shown to control a variety of motor functions, including motor activity, motor stereotypy, and motor learning. [4][5][6] Interestingly though, chronic nicotine did not affect motor learning with a daily training regimen, 7,8 suggesting that nicotine-induced motor learning deficit may be a subtle phenotype that should be revealed through a unique behavioral paradigm. Moreover, little is known about the neural mechanism underlying the potential effect of chronic nicotine on motor function.
The dorsal striatum is the major output of basal ganglia and is deeply involved in adaptive motor control. 9,10 Aberrant neural processing within the dorsal striatum can impair motor learning as observed in movement disorders. 11 Previous studies have found that chronic nicotine can disrupt neuronal activity within the dorsal striatum ex vivo, 12,13 suggesting that nicotine-dependent alteration of motor behavior could require striatal activity. However, the in vivo influence of chronic nicotine on the neural activity within the dorsal striatum and the potential role that striatal neurons play in nicotinedependent alteration of motor learning remains elusive.
In this study, we adopted 6 weeks of longitudinal single-unit recording and a sparse training regimen on rotarod to study the impact of chronic nicotine exposure and withdrawal on the striatal neural activity in mice. Specifically, we compared the in vivo activity of striatal medium spiny neurons (MSNs) and fast-spiking interneurons (FSIs) before, during, and after chronic nicotine treatment. MSNs comprise $95% of the striatal neuron population and are the output neurons of the striatum, 14 whereas FSIs are the main source of GABAergic feedforward inhibition in the striatum and marked by parvalbumin immunoreactivity. [15][16][17] Subsequently, we applied the chemogenetic approach to selectively modulate the activity of striatal fast-spiking parvalbumin interneurons to examine their role in nicotine-dependent alteration of motor learning. Here, we applied a sparse motor learning paradigm.
The sparse training regimen renders the behavioral task more demanding, which has been shown to be more sensitive to subtle deficit in behavior. [18][19][20] Furthermore, sparse training is a valid means of enhancing motor skill since the retention of motor learning can last for more than 7 days. 21,22 2 | MATERIALS AND METHODS

| Subjects
For in vivo single-unit recording, 3-month-old C57BL/6J male mice (KIST Research Animal Resource Center, Seoul, Republic of Korea) were single-housed and extensively handled for more than 2 weeks before tetrode microdrive implantation surgery. For behavioral experiments, 2-to 3-month-old C57BL/6J male mice (KIST Research Animal Resource Center) or B6.129P2-Pvalb tm1(cre)Arbr /J (PV-Cre) heterozygous male mice (Stock No. 017320; The Jackson Laboratory, ME, USA) were group-housed (two to four mice per cage) and handled for more than 1 week prior to the beginning of experiment. All mice were

| Nicotine
Mice were exposed to nicotine via subcutaneous implantation of osmotic mini-pump (Model 1004; Alzet, CA, USA). Briefly, mice were anesthetized with isoflurane (4% in pure oxygen for 3 min for induction, and 1.5% for maintenance). Nicotine solution was prepared immediately before osmotic pump implantation. (−)-Nicotine ditartrate (Tocris Bioscience, MO, USA) was dissolved in physiological saline, and pH was adjusted to 7.4. Concentration was adjusted to deliver free-base nicotine at 24 mg/kg/day for 2 weeks as previously described. 23,24 An incision was made in the back of the mice, the nicotine solution-filled osmotic pump was inserted subcutaneously, and the wound was closed with silk suture. After 2 weeks of implantation (chronic nicotine exposure), the osmotic pump was surgically removed under isoflurane anesthesia to induce spontaneous nicotine withdrawal. For behavior tests, the control group underwent sham surgery in which mice were implanted with a mini-pump without nicotine solution.
Single-unit recording and behavior tests were conducted over 6 weeks (Figures 2A, 4A, 5C, and S4). The schedule consisted of weekly recording or behavior test sessions. Data collected during the first 2 weeks before osmotic pump implantation were pooled and used as "Baseline" phase. Data collected during the next 2 weeks after osmotic pump implantation were used as "Nicotine exposure" (or "Exposure") phase. Data collected during the last 2 weeks after osmotic pump removal were used as "Nicotine withdrawal" (or "Withdrawal") phase.  26 with small modifications. The spike clusters were determined as a unit based on the waveform parameters, averaged spike waveform, autocorrelogram, and interspikeinterval histogram. The identified units were classified based on mean firing rate, spike half-width, and peak-to-valley as in previous studies. 17,27 Those units with mean firing rate < 6.5 Hz, half-width 0.1-0.16 ms, and peak-to-valley 0.25-0.37 ms were classified as putative MSNs, and those with half-width 0.075-0.13 ms and peak-to-valley 0.12-0.17 ms were classified as putative FSIs ( Figure 1B). Only those clusters with no interspike interval < 1 ms ( Figure 1C), isolation distance > 18 ( Figure 1D), L-ratio < 0.09 ( Figure 1E), 28 and firing rate > 0.2 Hz were included in the analysis.

| In vivo tetrode single-unit recording
The firing rates and averaged spike waveform parameters of MSNs and FSIs were compared among three phases of recording sessions (Baseline vs. Exposure vs. Withdrawal) ( Figure 2A). As depicted in Figure 2A, the units isolated from two recording sessions were pooled to represent each phase. Spike waveforms were drawn by averaging the spike traces of units identified as either MSNs or FSIs across the three phases of recording sessions. For spike waveform analysis, we analyzed halfwidth, peak-to-valley, spike amplitude, and amplitude ratio as in previous studies. 29,30 Briefly, half-width is the time duration between halfmaximal amplitude points, peak-to-valley is the time duration between peak and valley, spike amplitude is the voltage difference between peak and valley, and amplitude ratio is the absolute value of the peak amplitude divided by valley amplitude.
In addition, the burst activity of MSNs were analyzed in parallel ( Figure S1). Burst activity of MSNs is thought to be correlated with behavior-relevant information including movement initiation and regulation. 31,32 A single burst event of MSNs was defined based on a previous report, 33 with a small modification: A train of at least five spikes at an average firing rate of more than 20 Hz, and within which the interspike interval did not exceed 100 ms. The burst activity was analyzed through burst rate (the number of burst events per minute), intraburst firing rate (IBFR), and burst duration.

| Local field potential
Local field potential (LFP) recording was performed as described previously, 34 with small modifications. Briefly, the obtained data was bandpassed at 1-120 Hz and analyzed by Matlab (Mathworks; Natick, MA, USA). Thirty epochs (2 s each) of field potentials were randomly selected for analysis. For the power spectral density (PSD) analysis, the frequency ranges of delta (2-4 Hz), theta (4.5-8 Hz), alpha (8.5-12 Hz), beta , low gamma , and high gamma (60-85 Hz) were chosen. Sample entropy (SampEn) analysis was conducted following a previous report. 35 SampEn measures the predictability of a signal using the conditional probability that two template sequences with similar points remain similar at the following points. In the computation of SampEn, the parameter r was used as the arbitrary threshold of similarity between two template sequences.

| Nissl staining
After the single-unit recording was completed, mice were deeply anesthetized by intraperitoneal administration of Avertin (2,2,2-Tribromoethanol, 250 mg/20 ml/kg) (Sigma-Aldrich, MO, USA), and microlesions were made on the recording sites by passing an electrolytic current (10 μA, 30 s, cathodal) through one channel of each tetrode. Subsequently, the whole brain was quickly isolated and freeze stored at −80 C. The frozen brain was coronally microdissected on a cryostat (CM3050S, Leica Microsystems, Wetzlar, Germany) to a thickness of 30 μm at −18 C. The brain sections were mounted on Superfrost Plus slide and air-dried for $10 min at room temperature. Then, the whole slide was sequentially incubated in xylene for 20 min, 100% ethanol for 5 min, 95% ethanol for 5 min, 70% ethanol for 5 min, distilled water for 5 s, cresyl violet (1.0 mg/ml, syringe filtered) for 10 min, distilled water for 2 min, 70% ethanol for 5 min, 95% ethanol for 5 min, 100% ethanol for 5 min, and xylene for 20 min. Finally, the slide was mounted with Permount mounting medium mixed with xylene at 1:1 ratio. The sections were imaged under light microscopy ( Figure 1A).

| Behavior
Mice were acclimated to the behavior test room for at least 30 min before the beginning of each experiment. The behavioral apparatus was wiped with 70% ethanol and distilled water before and between experimental sessions or blocks. The whole behavioral sessions were video recorded. As depicted in Figures 4A and 5C, the data obtained from two behavioral sessions were pooled to represent each phase (Baseline, Exposure, and Withdrawal).

| Light-dark transition
The luminosity on the floor of the light chamber was adjusted to $350 lux. Mice were placed in the dark chamber; then, the door to the light chamber was opened, and mice were allowed to freely explore the chambers for 10 min. The time spent in the light chamber was measured.

| Rotarod
Based on previous studies showing that the motor learning retention can be maintained for at least a week, 21,22 we established a sparse motor learning protocol with small modifications. A session consisted of three blocks with a 1-h interblock interval. In the first block, mice were habituated on the rotarod with a fixed rate of 4 rpm for 5 min.
In the second and third blocks, each block consisted of two consecutive trials. In a trial, mice were subjected to 4 rpm fixed-speed training for 30 s and then were immediately subjected to accelerating rotarod (from 4 to 30 rpm, accelerating over 5 min). The better score out of two trials in the second block and the same in the third block were averaged to gain the latency to fall for the session.
To quantitatively assess the within-group, sparse trainingdependent improvement in motor skill, we analyzed motor learning index based on previous studies, 36

| Experimental design and statistical analyses
Before proceeding to statistical comparisons of single-unit spiking properties, Shapiro-Wilk normality test was conducted to analyze the skewness of distribution of each group (Table S1). For the within-group comparison of firing rates (Figure 2), spike waveform parameters (Table 1), and burst activity of MSNs ( Figure S1), Kruskal-Wallis nonparametric test followed by Dunn multiple comparison was applied because the sample distribution of spike properties was determined to be nonnormal (Table S1). For all empirical tests, p < 0.05 was considered statistically significant, and significance was denoted as # p < 0.05, ## p < 0.01, ### p < 0.001, * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001.    Table 1 Previous studies have well-established that the striatal FSIs are marked by parvalbumin immunoreactivity. 16  studies have demonstrated that nicotine does not affect motor learning at daily training schedule. 7,8 Therefore, we adopted sparse motor learning with a weekly training regimen ( Figure 4A) based on previous findings that sparse training can unveil subtle behavioral deficits [18][19][20] and motor learning retention can be maintained for over a week. 21,22 First, we found that chronic nicotine did not significantly affect Further analysis of the sparse motor skill learning through learning index revealed that motor performance gradually increased as sparse training regimen progressed (Control in Figure 4F; F (2,52) = 16.20; *** p = 0.0006, **** p < 0.0001), but the sparse learningdependent improvement in motor performance was hindered by chronic nicotine (Nicotine in Figure 4F).

Results
In association with the data from single-unit recording, we In accordance, the motor learning index gradually and substantially increased through sparse motor learning in PV-Cre mice (Control in Figure 5G; F(2,98) = 31.49; **** p < 0.0001), but chronic nicotine led to only marginal increase in the learning index (Nicotine in Figure 5G;  (Figure S4A,B). In addition, CNO alone did not affect sparse motor learning and did not alter Finally, the motor learning index gradually increased irrespective of CNO treatment (CNO in Figure S4D; F(2,86) = 26.44; ** p = 0.0088 for Baseline vs. Exposure, ** p = 0.0066 for Exposure vs. Withdrawal, **** p < 0.0001 for Baseline vs. Withdrawal) whereas the chronic nicotine-dependent reduction in the learning index was unaffected by CNO (Nicotine + CNO in Figure S4D).
In sum, we discovered that chronic nicotine withdrawal impaired sparse motor learning and that striatal fast-spiking parvalbumin interneurons mediate the nicotine-induced deficit in sparse motor learning ( Figure 6). These data collectively indicate that chronic nicotine causes There are a number of possibilities that could link together chronic nicotine, FSIs, and the deficit in the striatal network activity.
First, the nicotine-induced reduction of firing in striatal FSIs could directly shape the striatal network responses. As previously found, nondesensitizing nicotinic acetylcholine receptors can directly depolarize striatal FSIs, 49 and striatal FSIs can modulate both the striatal delta oscillation as well as the range of network responses. 43,44,47,48 Second, nicotine could work through other neurons to impact striatal FSIs and network activity. There are a variety of interneurons in the striatum, 50,51 each of which could be affected by nicotine 52,53 and can control striatal FSIs and striatal neural activity. 54 In addition, the excitatory cortical/thalamic input to the dorsal striatum could be altered in response to chronic nicotine, thereby modulating both FSIs and striatal network response. [55][56][57] Moreover, the nicotine-dependent alteration of dopaminergic action on striatal FSIs can also contribute to the overall process. 58,59 In the future, the existence and exact nature of causal relationship among chronic nicotine, striatal FSIs, and striatal network activity remain to be elucidated.
Here, we found that chronic nicotine reduces the activity of striatal fast-spiking parvalbumin interneurons, which was accompanied by increased spike half-width and peak-to-valley that gave rise to spike broadening. Previous studies have shown that spike broadening in the fast-spiking neurons is largely correlated with reduced potassium currents from calcium-activated BK channels and voltage-gated Kv3 channels, and that potassium channel blocker can reduce the firing rate of fast-spiking neurons. 60 Interestingly, spike broadening has also been associated with increased calcium influx and enhanced neurotransmission, 61,62 which might compensate for the potential F I G U R E 6 Schematic summary. Sparse motor learning requires the activity of striatal fast-spiking parvalbumin interneurons (FSI). In the process, FSI activity may be correlated with striatal delta oscillation power and entropy. Chronic nicotine impairs striatal FSI firing, which is accompanied by reduced delta power and entropy. Moreover, chronic nicotine causes sparse motor learning impairment in mice. Chemogenetic reversal of nicotineinduced deficit in striatal FSI firing rescues sparse motor learning, suggesting that the dorsal striatum is critical for motor learning and nicotine-induced motor impairment reduction in GABAergic neurotransmission from reduced activity of striatal FSIs and hence explain the unchanged firing of MSNs. In overall, these data imply that the potassium channels may underlie chronic nicotine-dependent spike broadening as well as reduced activity of the striatal FSIs. However, the impact of nicotine on BK channels or Kv3 channels in the striatal FSIs remains unknown and should be investigated in the future.
The chronic nicotine-mediated reduction of firing in striatal fastspiking parvalbumin interneurons contradicts the previous study demonstrating that acute activation of nicotinic receptors depolarizes the striatal FSIs ex vivo. 49

| The effect of chronic nicotine on behavior
Adopting 6 weeks of sparse motor learning paradigm, we found that chronic nicotine impairs motor learning in rotarod during withdrawal.
This finding demonstrates that the motor learning deficit should be accounted for in the research of nicotine addiction, particularly during nicotine withdrawal. This finding is also important in light of the previous studies showing that chronic nicotine does not impact daily motor learning. 7,8 Sparse training is more sensitive to subtle deficits in behavior, [18][19][20] which could explain the nicotine-induced impairment in sparse motor learning but not in daily motor learning. In addition, our data supports the previous studies showing that the timescale of learning is a critical factor in behavioral assays. [83][84][85][86] Moreover, the dissociation of findings between daily and sparse motor learning suggests that continuous and sparse motor learning may be mediated by distinct neural mechanisms, which would be an interesting topic to explore in the future.
The most robust difference in motor skill performance was observed only during withdrawal. However, it is difficult to pinpoint whether the deficit in motor performance during nicotine withdrawal reflects a nicotine withdrawal-specific motor deficit or an impairment in consolidation of learning during nicotine exposure. From careful examination of the data, a trend towards reduction in motor performance was observed during nicotine exposure, followed by a ceiling effect on performance during nicotine withdrawal. We suspected two possibilities: The motor deficit during nicotine withdrawal could be (1) an extension of subtle motor deficit from nicotine exposure or (2) a withdrawal-specific deficit that was newly arisen after nicotine exposure. Supporting the former idea, the subtle reduction in motor performance observed during nicotine exposure implies that the impaired motor learning may have already been in place during exposure and lasted throughout withdrawal. However, the slope of learning was positive from baseline to exposure, whereas the slope of learning reached towards 0 from exposure to withdrawal. This suggests that motor learning was partially intact during nicotine exposure but was evidently absent during nicotine withdrawal. Collectively, we support the latter view that an exposure-independent, withdrawal-specific motor deficit caused the chronic nicotine-dependent impairment in sparse motor learning.
In our study, anxiety-like behavior was unaffected by chronic nicotine treatment. This finding is consistent with previous reports showing that chronic administration of nicotine does not affect anxiety-like behavior in male mice. 87,88 However, it is also inconsistent with the previous reports showing that the ablation of FSIs leads to increased anxiety-like behavior. 89

| Limitation
Only the trait phenotype was examined in the single-unit recording experiment. The difficulty of single-unit recording during rotarod test precluded us from examining the state phenotype, that is, neu- Most importantly, the absence of relevant experimental groups in behavioral data limits us to make conservative interpretations.
First, we did not use female animals in our study, precluding us from generalizing the data to the overall population. Although nicotine discrimination and seeking behaviors does not seem to differ between males and females, 94,95 mice do show sex differences in motor behaviors including the latency to fall on rotarod. 96 Thus, the interaction between nicotine, motor skill, and sex may be an interesting topic to explore. Second, although we have demonstrated that CNO does not affect the chronic nicotine-dependent impairment in sparse motor learning, the interactive effect of nicotine with CNO has not been thoroughly evaluated due to the limited scope of our study. The back-conversion of CNO to clozapine could cause complex effects including its interaction with other pharmacological agents including nicotine. 40 CNO alone does not affect rotarod learning, 97,98 and the dose used in our study (0.3 mg/kg) has been reported not to significantly affect baseline behaviors in rodents. 40,99 However, the potential interaction between CNO and nicotine should still be carefully inspected.

| CONCLUSION
We employed a sparse motor learning paradigm to identify that chronic nicotine withdrawal leads to impairment in motor learning. We

DISCLOSURE/CONFLICT OF INTEREST
The authors declare no conflict of interest.