The Role of BTBD9 in Striatum and Restless Legs Syndrome

Abstract Restless legs syndrome (RLS) is a sensory-motor neurological disorder characterized by uncomfortable sensations in the extremities, generally at night, which is often relieved by movements. Genome-wide association studies (GWAS) have identified mutations in BTBD9 conferring a higher risk of RLS. Knockout of the BTBD9 homolog in mice (Btbd9) and fly results in motor restlessness and sleep disruption. Clinical studies have found RLS patients have structural and functional abnormalities in the striatum; however, whether and how striatal pathology contributes to the pathogenesis of RLS is not known. Here, we used fMRI to map regions of altered synaptic activity in basal ganglia of systematic Btbd9 knock-out (KO) mice. We further dissected striatal circuits using patch-clamp electrophysiological recordings in brain slices. Two different mouse models were generated to test the effect of specific knockout of Btbd9 in either striatal medium spiny neurons (MSNs) or cholinergic interneurons (ChIs) using the electrophysiological recording, motor and sensory behavioral tests. We found that Btbd9 KO mice showed enhanced neural activity in the striatum, increased postsynaptic currents in the MSNs, and decreased excitability of the striatal ChIs. Knocking out Btbd9 specifically in the striatal MSNs, but not the ChIs, led to rest-phase specific motor restlessness, sleep disturbance, and increased thermal sensation in mice, which are consistent with results obtained from the Btbd9 KO mice. Our data establish the role of Btbd9 in regulating the activity of striatal neurons. Increased activity of the striatal MSNs, possibly through modulation by the striatal ChIs, contributes to the pathogenesis of RLS.


Introduction
Restless legs syndrome (RLS) is a sensorimotor neurologic disease affecting up to 10% of the general population (Garcia . Characteristic symptoms of RLS include an urge for patients to move their legs often accompanied by, or felt to be caused by, uncomfortable sensations in the legs (Lanza and Ferri, 2019). The symptoms of RLS generally occur or worsen at rest or inactivity in the evening, which can be at least partially relieved by movements (Trenkwalder et al., 2018). Previous studies have emphasized the major role of iron in the disease (Connor et al., 2017). In addition, one of the primary medications for the disease is D 2 /D 3 dopamine (DA) agonists (Garcia- Borreguero and Cano-Pumarega, 2017), whereas refractory RLS can be treated with opioids (Silber et al., 2018).
To date, no neurodegeneration has been found in RLS patients. However, emerging studies suggest that changes in the striatum may underlie the pathogenesis of RLS Lanza et al., 2017;Rizzo et al., 2017). The striatum, which is comprised of the caudate and putamen, serves as the first recipient for most of the excitatory input from the cortex and thalamus to the basal ganglia (Haber, 2016). Approximately 95% of striatal neurons are GABAergic medium spiny neurons (MSNs), which are traditionally subdivided into two subtypes by DA receptor expression (Purves et al., 2001b). Generally, D 1 DA receptor (D 1 R)-expressing MSNs function in the direct pathway and are thought to facilitate wanted movements and pronociceptive effects (Soares-Cunha et al., 2016). In contrast, MSNs in the indirect pathway mainly express D 2 DA receptors (D 2 Rs). These MSNs are likely to be involved in the suppression of unwanted movement and the generation of antinociceptive effects (Soares-Cunha et al., 2016). Additionally, 1-2% of striatal neurons are cholinergic interneurons (ChIs; Zucca et al., 2018). MSNs regulate the activity of ChIs through GABA and endogenous opioids (Lim et al., 2014), while ChIs influence the activity of MSNs through M1, M4, and possibly nicotinic acetylcholine (ACh) receptors on MSNs (Bordia et al., 2016). Previous clinical studies have shown decreased D 2 R expression but increased phosphorylated TH in the putamen ). Brain imaging studies show decreased striatal DA transporter (DAT; Earley et al., 2011) and D 2 R binding potential (Michaud et al., 2002;Earley et al., 2013). Despite these observed alterations of the striatal dopaminergic system associated with RLS, the exact functions of the striatum, especially MSNs, in the disease development are largely unknown.
Genome-wide association studies (GWAS) have implicated up to 19 risk loci, including BTBD9, as genetic risk factors of RLS (Schormair et al., 2017;Jiménez-Jiménez et al., 2018). BTBD9 codes for a protein belonging to the BTB (POZ) protein family, which modulates transcription, cytoskeletal arrangement, ion conductance and protein ubiquitination (Stogios and Privé, 2004;Stogios et al., 2005). An alteration in hippocampal synaptic plasticity and neurotransmission has been found in Btbd9 knockout (KO) mice (DeAndrade et al., 2012b). Loss of the BTBD9 homolog in Drosophila melanogaster results in increased motor activity, decreased DA levels, and disrupted sleep patterns (Freeman et al., 2012). Similarly, the systematic Btbd9 KO mice showed motor restlessness, thermal hypersensitivity, and a disruption in sleep structure (DeAndrade et al., 2012a). Therefore, the systematic Btbd9 KO can be considered as a valuable disease mouse model to study the pathophysiology of RLS (De-Andrade et al., 2012a;Allen et al., 2017).
To elucidate the impact of striatal neurons on the generation of RLS-like phenotypes, we performed in vivo manganese-enhanced MRI (MEMRI) with the systematic Btbd9 KO mice and mapped the functional neural activity in the basal ganglia circuits. We next conducted electrophysiological recordings to observe both intrinsic and spontaneous firing activity of MSNs and ChIs in the striatum. Moreover, we selectively deleted Btbd9 in either striatal MSNs or ChIs and conducted behavioral studies.

Mice
The generation of the systematic Btbd9 KO mice The homozygous Btbd9 KO male mice used in MRI imaging were generated as described previously (DeAndrade et al., 2012a). The systematic Btbd9 KO mice used in the electrophysiological recording were generated from a line of Btbd9 loxP mice imported from the European Mouse Mutant Archive (EMMA; ID: 05554). In this line, the fourth exon of the Btbd9 gene was flanked by loxP sites (floxed). We first removed neomycin selection cassette by crossing with FLP mice (The Jackson Laboratory stock 126 no. 003946) to obtain Btbd9 loxP mice, which was then crossed with a general cre deletor to obtain Btbd9 KO allele. Heterozygous Btbd9 KO mice were interbred to produce experimental homozygous Btbd9 KO mice and the wild-type (WT) littermate controls.
The ChI-specific Btbd9 KO mice (Btbd9 ChKO) were generated by breeding Btbd9 loxP mice with Chat-cre mice (The Jackson Laboratory; stock 006410), in which the neo cassette had been removed by crossing with FLP mice. Breeding was conducted as outlined for Btbd9 sKO above. PCR was used for genotyping the Chat-cre (forward: ATC TCC GGT ATT GAA ACT CCA GCG C; reverse: CAC TCA TGG AAA ATA GCG ATC). To confirm the specific deletion of Btbd9 in the striatum, we dissected out brain regions following the protocol (Spijker, 2011) and PCR was conducted with primers specific for recombined locus (forward: AAG GCG CAT AAC GAT ACC ACG AT; reverse: TGG TGA TTC AAA TCT CCT TCC AAC ACA; Fig. 4C).
[AQ3] Experimental mice were housed in standard mouse cages at 21°C under normal 12/12 h light/ dark cycle (12 LD) condition. The protocol for the study received prior approval by the Institutional Animal Care and Use Committee, and all studies were conducted in accordance with the United States Public Health Service's Policy on Humane Care and Use of Laboratory Animals.
Quantitative RT-PCR (qRT-PCR) qRT-PCR was performed as described before (Yokoi et al., 2011) to determine whether exon 4 was deleted in mice after cre-mediated recombination. In brief, three Btbd9 sKO and three control adult male mice were sacrificed, and several brain regions (striatum, cerebral cortex and cerebellum) were harvested and flash frozen in liquid nitrogen. RNA was extracted using an RNAeasy Mini kit (QIAGEN) according to the manufacturer's instructions. Next, cDNA was made using SuperScript III reverse transcriptase (Invitrogen). PCR primers specific to Btbd9 exons 4 and 5 (forward: GAC TCT TGT CTC CGG ATG CT; reverse: TCA CAA CCT GAG CCC CAT AC); ␤-actin (forward: CAC CCG CGA GCA CAG CTT CTT TG; reverse: AAT ACA GCC CGG GGA GCA TCG TC). The expression of Btbd9 mRNA was measured and normalized by Bio-Rad CFX manager 3.1.

MEMRI
MnCl 2 (manganese chloride) pretreatment MEMRI was performed as described previously Zubcevic et al., 2018). Before the treatment, the animals used in the experiment were handled every two weeks and acclimatized to the investigator. Manganese (II) chloride tetrahydrate (Sigma-Aldrich Chemical Co.) was dissolved in distilled deionized water and sterilely filtered before administered intraperitoneally at a dose of 70 mg/kg/ml. After injections, mice were returned to their home cage and imaged after 20 -24 h as previously reported (Perez et al., 2013).

MRI
Images were collected by a 4.7 Tesla Magnex Scientific scanner under the control of Agilent Technologies VnmrJ 3.1 console software. A 38-mm quadrature transmit/receive radio frequency coil tuned to 200 MHz was used (Insight NeuroImaging Systems, LLC). Mice were anesthetized with 2.0% (0.1 l/min) delivered in 100% oxygen for 30 -60 s. Then the level of isoflurane was maintained between 1.0% and 1.25% throughout the entire setup and imaging session, during which the respiratory rates were monitored continuously and sustained between 20 and 30 beats per minute by adjusting isoflurane levels between the range. Placed prone on custom-size plastic bed with a respiratory pad placed underneath the abdomen, body temperatures of the mice were maintained using a warm air recirculation system (SA Instruments, Inc.). The head and incisors of mice were secured on the front end of the plastic bed to minimize motion. The front half of the bed was aligned and clamped inside the quad RF coil and placed inside the isocenter of the scanner. Images were acquired at 4.7 Tesla using a T1-weighted spin echo pulse sequence with the following parameters: repetition time ϭ 300 ms, echo time ϭ 12 ms, the field of view ϭ 19.2 ϫ 19.2, slice thickness ϭ 0.8 mm, 12 slices. Total scan time per mouse was 30 min.

Slice preparation
Experiments were conducted as described previously (Pappas et al., 2015;Augustin et al., 2018). Recordings of MSNs were conducted with three Btbd9 KO and five WT male littermates at an average age of four months. Recordings of ChIs were performed with four Btbd9 KO and five WT male littermates with an average age of four months, or with three Btbd9 ChKO and three control males with an average age of five months. Investigators who conducted the experiments were blind to the genotypes. Animals were sacrificed, and the brains were rapidly removed; 300 m-thick coronal brain slices containing the dorsal striatum were cut in ice-cold, oxygenated cutting saline: 180 mM sucrose, 2.5 mM KCl, 1.25 mM NaH 2 PO 4 , 25 mM NaHCO 3 , 10 mM D-glucose, 1 mM CaCl 2 , 10 mM MgCl 2 , and 10 mM glucose with a Vibratome (Leica VT 1000s). Slices were recovered in a holding chamber for 30 min at 35°C with artificial CSF (ACSF). Final concentrations of ACSF: 126 mM NaCl, 2.5 mM KCl, 1.25 mM NaH 2 PO 4 , 25 mM NaHCO 3 , 2 mM MgCl 2 , 2 mM CaCl 2 , and 10 mM glucose, pH 7.3 with KOH, osmolality 290 -300 mOsm. The slices were then incubated at room temperature.

Cell identification
The slices were placed in a recording chamber and continuously perfused with ACSF that was bubbled via 5% CO 2 and 95% O 2 at a rate of 1.5 ml/min while being visualized with an upright microscope (Zeiss) using a 40ϫ water-immersion objective with infrared optics. MSNs were identified by the somatic size and basic membrane properties including input resistance, membrane capacitance, and time constant. ChIs were recognized based on morphology and size, as they are irregularly polygonal with large cell soma (Ͼ20 m).

Cell-attached and whole-cell recordings
For MSNs, all experiments were recorded at 32°C by a dual automatic temperature controller (TC-344B). Cellattached recording patch pipette (6 -10 M⍀) contained following solutions: 125 mM K-gluconate, 8 mM NaCl, 10 mM HEPES, 2 mM MgATP, 0.3 mM NaGTP, and 0.2 mM EGTA (pH 7.25-7.3, osmolality 290 -300 mOsm) and was used for voltage and current clamp recordings. Access resistances were Ͻ30 M⍀. Spontaneous postsynaptic currents were recorded in ACSF. To minimize the contribution of GABA A receptors, we held cells at Ϫ70 mV with an application of 50 M picrotoxin solution, which can abolish the activation of GABA A receptors. Next, at holding potential -65 mV, injection of depolarizing 50-pA current pulse of 300-ms duration evoked spike firing when the membrane potential reaches the firing threshold under current clamp configuration in a brain slice. This process was repeated at 10 increasingly depolarized potentials with incremental current steps (50 pA).
For ChIs, electrodes for cell-attached recordings were filled with a K-gluconate-based solution containing the following concentrations: 112.5 mM K-gluconate, 4 mM NaCl, 17.5 mM KCl, 0.5 CaCl 2 , 5 mM MgATP, 1 mM NaGTP, 5 mM EGTA, and 10 mM HEPES; with pH 7.2 (270 -280 mOsm) and resistance of 5-10 M⍀. Positive pressure was applied to the patch electrode as it approached the ChIs. Suction was applied to the electrode to create a seal (Ͼ5 G⍀) between the recording pipette and cell membrane. Action potential current was recorded in a voltage-clamp mode that maintained an average of 0-pA holding current. After breaking through the cell membrane, cellular properties (capacitance, input resistance, and time constant) were recorded at a membrane potential of Ϫ70 mV. Electrode access resistance was maintained throughout at Ͻ30 M⍀. Resting membrane potential was recorded in current clamp mode. Action potential for current step recording was triggered using depolarizing currents steps of 300 ms.
Data acquirement and detection were the same as previously described (DeAndrade et al., 2012b). Recordings were made from targeted cells in the striatum using infrared differential interference contrast microscopy and an Axopatch 1D amplifier (Molecular Devices). Data were acquired using pClamp 10 software. Signals were filtered at 5 kHz, digitized at 10 kHz with a DigiData 1440 (Molecular Devices). Events were detected using the Mini Analysis Program (Synaptosoft) with parameters optimized for each cell and then visually confirmed before analysis. The peak amplitude, 10 -90% rise time and the decay time constant were measured based on the average of all events aligned by the rising phase.

Immunofluorescence staining
Rgs9-cre mice were bred with GFP mice imported from The Jackson Laboratory (stock 007906) to obtain Rgs9cre and GFP double heterozygous mice to map Rgs9-crepositive neurons. As described previously (Dang et al., 2006), the mice were anesthetized and perfused with ice-cold 0.1 M PBS (pH 7.4) followed by 4% paraformaldehyde in 0.1 M PB (pH 7.4). The brains were soaked in 4% paraformaldehyde-PB at 4°C overnight and then incubated in 30% sucrose in 0.1 M PBS at 4°C until the brains sank to the bottom. The brains were frozen with dry-ice powder and cut coronally into 40-m sections with a Histoslide 2000 sliding microtome (Reichert-Jung). Sections were sequentially rinsed 5 min each in 0.5% Triton X-100, 0.02 M PBS; 0.1% Triton X-100, 0.02 M PBS; 10 mM glycine in 0.1 M PBS for three times; 0.5-ml 2% gelatin in 0.1 M PBS; 10 mM glycine in 0.1 M PBS; and 0.1% BSA in 0.1 M PBS. Then tissues were incubated with the primary antibody, 1:50 goat anti-choline acetyltransferase (AB144P; Millipore), dissolved in 100 l 1% BSA, 0.1 M PBS at 4°C overnight. The next day, tissues were washed for six times with 0.1% BSA, 0.1 M PBS, followed by incubation with secondary antibody, 1:200 Cy3conjugated AffiniPure donkey anti-goat IgG (705-265-003; Jackson ImmunoResearch). After washing, the sections were mounted on glass slides with VECTASHIELD Antifade Mounting Medium (H-1400) and cover-slipped.

Behavioral studies Thirty-minute open field
Eight Btbd9 sKO male mice and seven male littermates with an average age of 15 months, or seven Btbd9 ChKO (three males, four female) and nine controls (five males, four female) with an average age of seven months, were used in the 30-min open field analysis as previously described (DeAndrade et al., 2012a). Briefly, each mouse was placed in the center of a VersaMax Legacy open field apparatus connected to a computerized Digiscan System (Accuscan Instruments, Inc.) and continuously monitored for 30 min. The apparatus contains infrared sensors along the walls that detect any breaks in the beams. Bright illumination (ϳ1 k lux at the center by a 60-W white bulb) was focused on the center of each field.

Wheel running
Eight Btbd9 sKO mice (seven males, one female) and 13 control mice (eleven males, two females) with an average age of four months, or six Btbd9 ChKO male mice and seven male littermates with an average age of two months, were maintained on a 12 LD cycle for 7 d. Wheelrunning activity (DeAndrade et al., 2012a) was recorded as the number of wheel revolutions occurring during 5 min bins and analyzed using Lafayette Instrument Activity Wheel Monitor software. The activity from the last 4 d was included in the data analysis, grouped by light phase and dark phase.

Continuous open field
Seven male Btbd9 sKO mice and five male littermates with an average age of five months or three Btbd9 ChKO male mice and four male littermates with an average age of two months were used in the long-term open field analysis modified from 30-min open field test (Meneely et al., 2018). Each mouse was placed in the center of a VersaMax Legacy open field apparatus with enough corncob bedding, food, and water. Breaks in the beams were decoded by VERSDATA version 2.70-127E (AccuScan Instruments Inc.) into behavioral patterns. Batch 1 data for Btbd9 ChKO mice were collected every 1 h. Other data were recorded every 15 min throughout the experiment. Data from the last 4 d were separated into light and dark phases, and the total distances during each phase were combined and coded as day 4 -7, and night 4 -7, respectively. The analysis was conducted based on all four periods in each phase. Separately, the total distances for each 15 min of the last 4 d were recoded for sleep analysis. If the total distance traveled in 15 min was 0, the mouse was considered as sleeping, and the data were coded as 0; otherwise, the mouse was considered as awake, and the data were coded as 1.

Tail flick test
Nine male Btbd9 sKO mice and nine male littermates with an average age of eight months were tested for the perception of warm stimuli. Each mouse was placed in an acrylic restrainer with the distal end of its tail protruding on a metal surface maintained at 55°C. The timer was turned on once the tail touched the surface and immediately stopped when the mouse flicked its tail away from the heat. The latency to respond was limited to 90 s to prevent injury to the mouse.

Experimental design and statistical analysis
Images were processed and analyzed as previously reported (Perez et al., 2013). Mn 2ϩ accumulation in active neurons produces signal intensity increases in T1 images. However, as this is a non-quantitative approach to measure activity and because there is scan-to-scan intensity variation independent of Mn 2ϩ , we normalized images based on their individual variance. Using this normalization approach, where surpassing a normalized threshold value of 1 indicates increased activity associated with Mn 2ϩ administration, we have observed significant differences between Mn 2ϩ administered and non-treated rodents. Image processing was conducted using ITK-SNAP (http://www.itksnap.org), and image math scripts were available on FSL (fslmaths; http://www.fmrib.ox.ac.uk/ fsl/). Scans were aligned with a segmented atlas of the adult mouse brain using an automated affine linear registration tool from FSL (Jenkinson et al., 2012). Each scan was converted to a z value map through a voxel-wise normalization procedure. The mean signal intensity across the entire extracted brain volume (x) was subtracted from each voxel (xi) and then divided by the variance (). A pre-set threshold of z Ն 1 was selected based on prior observation of individual datasets and a close inspection of their intensity distribution histograms. All voxels with z score values below this threshold were set to zero. Thus, the voxels exceeding the threshold value of z Ն 1 were considered in our statistical analysis as having higher signal intensities (quantified as the number of voxels above a z value of 1). Mean number of voxels for each region of interest (ROI) was compared using an unpaired two-tailed t test (homoscedastic variances, ␣ Յ 0.05).
Electrophysiological data were analyzed by logistic regression (not normally distributed) or mixed model ANOVA (SAS statistic package, normally distributed) with cell identification number nested within animal identifica-tion number. Open field data were analyzed by mixed model ANOVA and adjusted for multiple comparisons using the Benjamini-Hochberg-Yekutieli false discovery rate (FDR; p Ͻ 0.05). Data obtained from wheel running study was analyzed by logistic regression with a negative binomial distribution. Total distances of continuous open field were analyzed by logistic regression with a gamma distribution while sleep analysis was conducted with binomial logistic regression modeling the probability of waking. Tail flick data were processed by logistic regression with a gamma distribution. GEE model in the logistic regression normalized WT or control groups in terms of current steps, wheel running, continuous open field, and tail flick to 0 without the error bar. Age and gender were used as covariates in all analysis.
To generate the hourly wheel-running activity presented in Figure 5B, we summed the interval counts during each hour for each animal. Wheel-running activity during the last 96 h was analyzed. Hence each animal had 4 data points for each hour. The average interval counts within each hour were calculated for each genotype. The p values, calculated by the unpaired Student's test, were marked above hours in the figure. To generate the hourly probability of waking in Figure 5I, we determined the sleep status by the total distance traveled during 15 min in the long-term open field test as mentioned above. Therefore, there were 4 data points for each animal during each hour, which were coded from 1 and 4 as "sample." Last 4 d' open field activity was analyzed. Hence each animal had 4 d of data, which were coded from 1 to 4 as "period." The probability of waking was calculated from a genotype and hour two-way interaction with repeated measurement of the period, hour, and sample using SAS logistic regression with a binomial distribution. For each genotype, SAS normalized probability of waking during each hour to the probability of the last hour either during the day or night period. The probability of waking at 6 P.M. of the control mice was set as 1, and the probabilities of the waking of other hours of the control mice were calculated relative to that of 6 P.M. To calculate the relative difference between ChKO mice and controls during each hour, we sorted the data by the hour and analyzed the probability of waking for each genotype with repeated measurement of period and sample using SAS logistic regression with a binomial distribution. The p values were marked above hours in the figure. The probability of waking of the ChKO mice was derived from the relative difference between the ChKO mice and the control mice.

Increased striatal neural activity in fMRI study with the systematic Btbd9 KO mice
To study the role of the striatum in RLS pathogenesis, we first used MEMRI imaging to determine the striatal neural activity in the systematic Btbd9 KO mice. MEMRI has been extensively used to track Ca 2ϩ -dependent synaptic activity (Lu et al., 2007;Hsu et al., 2008;Chiu et al., 2015;Dudek et al., 2015;Perrine et al., 2015). As a calcium analog, Mn 2ϩ enters active synapses through voltage-gated calcium channels (Fukuda and Kawa, 1977;Narita et al., 1990) and is sequestered and transynaptically transported antero-and retrogradely across active neural circuits (Sloot and Gramsbergen, 1994;Pautler et al., 1998;Takeda et al., 1998a, b;Saleem et al., 2002;Murayama et al., 2006). The presence of the paramagnetic Mn 2ϩ ion in the brain increases longitudinal relaxation rates and enhances signal intensity in T1 weighted scans, and is used for functional mapping of synaptic activity (Duong et al., 2000). Here, after Mn 2ϩ was injected into the mice, the images of ROI were acquired (Fig. 1A). The result showed increased neuronal activity in the cerebral cortex of the systematic Btbd9 KO mice, indicating an increased cortical input to the striatum. This will be addressed in detail in a separate manuscript. In the striatum, which is the focus of the current study, there was a significant increase in the caudate/putamen (Fig. 1B, p ϭ 0.006, unpaired two-tailed t test), indicating increased neural activity specifically in the striatum. Increased entry of Mn 2ϩ likely through voltage-gated Ca 2ϩ channels also suggest an increase in Ca 2ϩ -dependent neural activity.

More excitable MSNs but decreased ChI activity in the systematic Btbd9 KO mice
As the first recipient for the excitatory input from almost all of the cortex, the striatum is mostly composed of MSNs (Ͼ95%; Cox and Witten, 2019). To determine the source of increased neural activity in striatum revealed by fMRI study, we did whole cell patch-clamp recording in brain slices. The result indicated that the resting membrane potential of the systematic Btbd9 KO cells was higher than the WTs (Fig. 2D, p ϭ 0.03, logistic regression with a gamma distribution). No change was found in membrane capacitance ( Fig. 2A, p ϭ 0.93, logistic regression with a gamma distribution), input resistance (Fig. 2B, p ϭ 0.58, logistic regression with a gamma distribution) and decay time constant (Fig. 2C, p ϭ 0.58, logistic regression with a gamma distribution). We then tested whether the intrinsic excitability of striatal MSNs was affected by the loss of Btbd9. Depolarizing current steps were injected to the MSNs of both Btbd9 KOs and WTs (Fig. 2E). There was no significant difference in the frequency-current relationship (Fig. 2F,G, p ϭ 0.51, logistic regression with a negative binomial distribution), am-plitude, rise, and decay time (data not shown). Our results indicated that there was no change in the intrinsic excitability of the KO MSNs.
MSNs are strongly driven by glutamatergic inputs (Purves et al., 2001a). To test whether loss of BTBD9 affects excitatory synaptic transmission in MSNs, we recorded sEPSCs (Fig. 2H). There were no alterations in the rise time (Fig. 2K, p ϭ 0.60, logistic regression with a gamma distribution), the decay time (Fig. 2L, p ϭ 0.33, logistic regression with a gamma distribution), and the frequency (Fig. 2I, p ϭ 0.11, logistic regression with a gamma distribution) of spontaneous postsynaptic currents. However, the systematic Btbd9 KO mice showed a significantly larger amplitude of sEPSC (Fig. 2J, p ϭ 0.004, logistic regression with a gamma distribution). The increased amplitude may be due to the increased presynaptic quantal size, increased postsynaptic functional AMPA receptor, or both. The result suggests that there are significantly enhanced excitatory inputs to the KO MSNs. In combination with the increased resting membrane potential found in the KO MSNs, the result indicates that BTBD9 deficiency may cause striatal MSNs to be more excitable.
Abnormality in ACh neurotransmission plays an important role in movement disorders like Parkinson's disease and dystonia (Bohnen and Albin, 2011;Dang et al., 2012;Lim et al., 2014;Eskow Jaunarajs et al., 2015). Furthermore, another RLS susceptibility gene, MEIS1, has been linked to the development of striatal ChIs (Spieler et al., 2014). Here, to better understand the striatal physiology in Btdb9 KO mice and RLS, we recorded both spontaneous activity (Fig. 3C) and intrinsic excitability of ChIs (Fig. 3A). We found that ChIs of the systematic Btbd9 KO mice had decreased intrinsic excitability (Fig. 3B, p ϭ 0.049, logistic regression with a negative binomial distribution) and spontaneous firing activity (Fig. 3D, p ϭ 0.04, ANOVA). It is known that MSNs inhibit the activity of ChIs through GABA and opioid receptors (Lim et al., 2014). Therefore, the alterations found in ChIs can either be a cellautonomous effect of BTBD9 deficiency or the response to increased activity of MSNs. This was further explored in Btbd9 ChKO mice.

Generation and molecular characterization of Btbd9 sKO mice
To further study the critical role of the striatum in RLS, we generated two conditional KO mouse models in which the Btbd9 gene was selectively knocked out either in the MSNs or ChIs (Fig. 4A). Specifically, we interbred Rgs9cre mice with Btbd9 loxP mice to obtain Btbd9 sKO mice. The transcription of Btbd9 gene was quantified by qRT-PCR. As expected, there was a significant reduction of Btbd9 mRNA in the striatum compared to control littermates (Fig. 4B, p ϭ 0.047, paired two-tailed t test), but not in the cerebral cortex or cerebellum (Fig. 4B, cerebral cortex, p ϭ 0.51; cerebellum, p ϭ 0.28; both paired two-tailed t test). To future confirm tissue specificity of the KO, we dissected out different brain regions from the Btbd9 sKO mice and their controls. DNAs were extracted from these different brain regions, and PCR reactions were conducted. Only the DNA extracted from the striatum of Btbd9 sKO mice showed the recombined band (Fig. 4C), indicating that the KO was restricted to the striatum. It is worth noting that the remaining Btbd9 expression in the striatum can be accounted for by the small subset of interneurons (Ͻ4%) that do not express cre, and non-neuronal cells such as glial cells. To determine whether ChIs express Rgs9-cre, we crossed Rgs9-cre mice with GFP indicator mice and generated Rgs9-cre-GFP mice. There were no overlaps between 100 randomly selected, ChAT-positive neurons with any of the GFPpositive neurons (Fig. 4D). The results suggest that Rgs9cre does not induce gene recombination in ChIs and Btbd9 gene is mostly knocked out in the MSNs of Btbd9 sKO mice.

RLS-like phenotypes in Btbd9 sKO but not in Btbd9 ChKO mice
A principal feature of RLS is a desire to move (Ferré et al., 2019). Previous mouse or fruit fly models of RLS have shown increased activity levels (DeAndrade et al., 2012a;Freeman et al., 2012). Therefore, we used an open field activity chamber to assess the total activity levels of the Btbd9 sKO and Btbd9 ChKO mice. In the short-term 30-min open field test, we observed that although there was no alteration found with Btbd9 ChKO mice [ Fig. 5F, total distance, p ϭ 0.69; clockwise (CW), p ϭ 0.29; counter-CW (CCW), p ϭ 0.70; all ANOVA], Btbd9 sKO mice exhibited significantly increased total distance traveled (Fig. 5A, total distance, adjusted p ϭ 0.01, ANOVA) and vertical activity (Table 1, adjusted p ϭ 0.049, ANOVA) compared with control mice. Vertical activity here represents rearing behavior (Tatem et al., 2014). Furthermore, Btbd9 sKO mice had a significant increase in CW circling, while there was no statistical difference in CCW circling compared with control mice (Fig. 5A, CW, adjusted p ϭ 0.02, ANOVA; CCW, adjusted p ϭ 0.095, ANOVA). Finally, there were no significant differences in stereotypical behavior or anxiety in the mice (Table 1). The increased activity level suggests that Btbd9 sKO mice are hyperactive. Furthermore, alterations in circling behaviors indicate imbalances in the striatal dopaminergic system (Forna-guera and Schwarting, 2002). In the long-term open field test, Btbd9 sKO mice showed no change (Fig. 5C, left panel, p ϭ 0.35, logistic regression with a gamma distribution) in the total distance traveled in the light phase, when mice are usually sleeping or resting. However, the sleep analysis indicates that the probability of waking of Btbd9 sKO mice significantly increased in the light phase (Fig. 5D, left panel, p ϭ 0.03, logistic regression with a binomial distribution), but did not change in the dark phase, when mice are usually active (Fig. 5D, right panel, p ϭ 0.63, logistic regression with a binomial distribution). The symptoms of RLS patients usually occur or become worse in the evening or at night (Garcia . With opposite day-night rhythms to human, Btbd9 sKO mice had the motor restlessness with a similar circadian predominance as patients. In contrast, although Btbd9 ChKO mice did not show any difference in total distance traveled compared with the controls (Fig. 5H, left panel, p ϭ 0.28; right panel, p ϭ 0.16; both logistic regression with a gamma distribution), they had a decreased probability of waking during the light phase (Fig.  5I, left panel, p Ͻ 0.0001, logistic regression with a binomial distribution), especially at 1, 3, and 5 P.M. (Fig. 5I, right panel), and increased probability of waking during the dark phase (Fig. 5I, middle panel, p ϭ 0.0029, logistic regression with a binomial distribution), suggesting that the animals sleep better than their controls. It should also be noticed that the increased probability of waking found in Btbd9 sKO mice mainly appeared during the second half of the rest phase (data not shown), which is consistence with clinical observations. Next, we conducted a wheel running study to measure the voluntary activity of these mice under the normal 12 LD condition. Btbd9 sKO mice showed a significantly elevated level of activity compared with controls during the light and rest phase (Fig. 5B, light phase, p ϭ 0.0004, logistic regression with a negative binomial distribution), but a similar level of activity as control mice during the dark and active phase (Fig. 5B, dark phase, p ϭ 0.32, logistic regression with a negative binomial distribution). On the other hand, Btbd9 ChKO mice showed a significantly decreased activity level during the light and rest phase (Fig. 5G, left panel, p Ͻ 0.0001, logistic regression with a negative binomial distribution) but increased activity level during the dark and active phase (Fig. 5G, right panel, p ϭ 0.0015, logistic regression with a negative binomial distribution). These data are consistent with the long-term open field test and suggest that only Btbd9 sKO mice have an increase in voluntary activity during their rest period. The increased voluntary activity mainly appeared during the second half of the rest phase and the first half of the active phase. Taken together, both total activity and voluntary activity were increased in the Btbd9 sKO mice in the rest phase, which resembles aspects of nocturnal RLS activity found in patients. The circadian component-involved behavior of Btbd9 ChKO mice is completely contradictory to Btbd9 sKO mice. Loss of BTBD9 protein only in MSNs, but not ChIs, can cause diurnal motor restlessness in mice.
Uncomfortable sensations in lower limbs are another common phenotype of RLS (Garcia . Therefore, we tested the Btbd9 sKO mice for abnormalities in the sensory system using the tail-flick test. The mutant mice had a higher level of sensitivity to the heat stimuli (Fig. 5E, p ϭ 0.005, logistic regression with a gamma distribution), indicating that mice lacking BTBD9 specifically in MSNs developed alterations in thermal sensation as the systematic Btbd9 KO.

Increased ChIs excitability in Btbd9 ChKO mice
The behavioral data showed that BTBD9 deficiency in MSNs, but not in ChIs, can lead to RLS-like phenotypes. Therefore, it is likely that the KO of Btbd9 in ChIs alone did not contribute directly to the behavioral and electrophysiological alterations observed in the systematic Btbd9 KO mice. To explore this, we recorded the ChIs in Btbd9 Open triangles indicate the FRT sites that were incorporated to remove the neo cassette. Btbd9 loxP mice were crossed with Rgs9-cre or ChAT-cre mice to obtain double heterozygotes. The double heterozygotes were crossed with Btbd9 loxP homozygotes to obtain Btbd9 sKO or Btbd9 ChKO mice. In conditional KO mice, exons 4 is deleted in specific types of neurons where cre is expressed. Recombination occurs in these cells, while other brain regions and the rest of the body still retain the intact exons. B, Btbd9 sKO mice showed a decreased level of Btbd9 mRNA in the striatum, but not in the cerebral cortex and cerebellum. Bars represent means plus SEs; ‫ء‬p Ͻ 0.05. C, Tissue-specific deletion of Btbd9 exon 4 in Btbd9 sKO mice was confirmed by PCR using DNA isolated from each brain region. The deletion was detected only in the striatum of Btbd9 sKO mouse as predicted (⌬). D, A representative immunohistochemical image of a coronal section of the striatum from an Rgs9-cre/GFP mouse. Scale bars represent 25 m. Enlarged images captured with a 40ϫ objective lens showed that the ChAT staining (red) did not overlap with GFP staining (green). The results suggest that Rgs9-cre does not have cre-mediated recombination in ChIs.
ChKO mice for both spontaneous activity (Fig. 6C) and responses to the injected currents (Fig. 6A). We did not find the decreased activity of ChIs in Btbd9 ChKO mice, as observed in the systematic Btbd9 KO (Fig. 3). Instead, spontaneous activity of ChIs in the mutant mice was significantly increased (Fig. 6D, p ϭ 0.02; 6E, p ϭ 0.74; all ANOVA), although there was no change found in intrinsic excitability (Fig. 6B, p ϭ 0.70, logistic regression with a negative binomial distribution). Hence, a lack of Btbd9 in ChIs alone led to elevated ChI activity.

Discussion
In this study, we determined how the loss of Btbd9 affects striatal physiology and focused on the role of MSNs in RLS pathogenesis. Using both brain imaging in vivo and electrophysiological recording in vitro, we found that the systematic Btbd9 KO mice had enhanced neural activity in the striatum, more excitable MSNs, and decreased activity in the ChIs. In addition, specific loss of Btbd9 in the MSNs was sufficient to cause RLS-like phenotypes. When Btbd9 was conditionally knocked out in the ChIs, the mice showed neither RLS-like behavioral phenotypes or maintained decreased activity in the ChIs as observed with the systematic Btbd9 KO mice. The results suggest that activity changes in the ChIs in the systematic Btbd9 KO mice are not cell-autonomous but result from alteration of striatal circuits, including changes  in the MSNs. It should be noted that the current study is not aimed at generating new animal models for RLS. Instead, our findings demonstrate that striatum, especially MSNs, is critically involved in the development of RLS-like phenotypes in mice. Our imaging study revealed increased neural activity in the systematic Btbd9 KO striatum. Close to 95% of cells in the striatum are MSNs. Therefore, the result suggests a possible overall increase in the Ca 2ϩ -dependent neural activity of MSNs. Furthermore, electrophysiological results showed an increased excitatory synaptic transmission onto striatal MSNs while their intrinsic neuronal properties were not altered. No change in the frequency, but the higher amplitude of sEPSC suggest that the MSNs of the KO mice seem to have increased excitatory synaptic inputs. Combined, our data suggest an enhancement of corticostriatal or thalamostriatal synaptic activity. Similarly, in iron deprived rat, which is thought to be an RLS rodent model, corticostriatal excitability is elevated (Yepes et al., 2017). Here, it is not known whether inputs from corticostriatal, thalamostriatal, or both contributed to the increased striatal sEPSC amplitude in the systematic Btbd9 KO mice. Differential alterations of thalamostriatal and corticostriatal synapses have been found in mouse models of Huntington's disease (Deng et al., 2014;Kolodziejczyk and Raymond, 2016;Parievsky et al., 2017), an MPTP-treated monkey model of parkinsonism (Raju et al., 2008) and a rat model with L-DOPA-induced dyskinesias (Zhang et al., 2013). Future studies will be focused on dissecting the differential effects of the systematic Btbd9 KO on corticostriatal and thalamostriatal synaptic transmission.
The electrophysiological recording revealed a decreased excitability in striatal ChIs in the systematic Btbd9 KO mice. It has been found that Gbx2, a gene which is essential for the proper development of striatal ChIs, is downed-regulated in heterozygous Meis1 KO embryos (E12.5; Spieler et al., 2014). MEIS1 is another gene implicated in RLS (Schormair et al., 2017). With conditional KO mice and electrophysiological recordings, we found that like Meis1 mice, Btbd9 mutation in the ChIs alone can cause functional abnormalities in striatal ChIs, yet it was insufficient to produce RLS-like behaviors. In the rest phase, Btbd9 ChKO mice had decreased locomotor activity with increased excitability of striatal ChIs, while the systematic Btbd9 KO mice showed increased locomotor activity (DeAndrade et al., 2012a) with decreased excitability in striatal ChIs. The results suggest a critical role of ChIs in movement control in the rest phase, and the decreased ChI activity found in the systematic Btbd9 KO mice is not a cell-autonomous effect. It is known that ACh can regulate striatal circuit through its receptors on MSNs, GABAergic interneurons, glutamatergic and DA terminals (Lim et al., 2014). It is possible that excess release of ACh in Btbd9 ChKO mice downregulated activities of MSNs directly through the nAChRs (Liu  (Howe and Surmeier, 1995) or indirectly by increasing GABAergic inhibition on MSNs (English et al., 2011). As the sole output of the striatum, MSNs may lead to decreased locomotion through their decreased activities. However, when Btbd9 was knocked out systematically, neuronal activities were changed in both MSNs and ChIs. Our results showed that MSNs became more excitable and may have increased activity. These changes in MSNs were likely to be dominant and overcame the influence of ChIs. MSNs, via their inhibition on the ChIs, in turn, led to decreased activity in ChIs as showed by the electrophysiological recording. Although it is not clear if the activity level of MSNs in the Btbd9 sKO mice is the same as what we found in the systematic Btbd9 KO, Btbd9 sKO mice did show an opposite output in behavioral tests as Btbd9 ChKOs, which support the overwhelmingly inhibitory effect of MSNs to ChIs as mentioned above.
MSNs-specific Btbd9 KO was found to be sufficient to induce rest-phase specific hyperactive movement, sleep disturbance, and increased thermal sensation in mice. The findings support the idea that striatum is critical for the pathogenesis of RLS motor phenotypes. Postmortem studies comparing RLS patients and control group show a decreased D 2 R expression in the putamen, but increased phosphorylated tyrosine hydroxylase (TH), a ratelimiting enzyme for DA synthesis, in both putamen and SN ). Brain imaging studies indicate decreased membrane-bound DAT level and D 2 R binding potential in the striatum (Rizzo et al., 2017). Additionally, changes in iron hemostasis have been found in SN and putamen of RLS patients Earley et al., 2014). Striata of iron-deprived RLS rodent models show a reduced density of DAT and DA receptors (Erikson et al., 2000(Erikson et al., , 2001 and enhanced release at corticostriatal terminals (Yepes et al., 2017). Taken together, these studies suggest functional alterations in the striatum are correlated with RLS (Rizzo et al., 2017). In addition, a recent GWAS study demonstrated that striatal MSNs are associated with insomnia (Jansen et al., 2019), which can be caused by RLS.
However, dysfunctional striatal circuit caused by the loss of BTBD9 might not be the only mechanism in the systematic Btbd9 KO mice. Btbd9 is also expressed in the spinal cord, although our genomic PCR failed to detect any deletion of the Btbd9 gene in the spinal cord of the Btbd9 sKO mice. Alterations in the striatum can lead to alterations in the spinal cord. The basal ganglia output modulates the spinal cord through feedback to the cortex. Additionally, recent evidence has suggested that a microcircuit exists between the corticostriatal tract and the corticospinal tract, starting in the striatum and ending in the spinal cord (Kiritani et al., 2012). Spinal neural circuits have been proposed to be central in RLS development (Clemens et al., 2006;Zwartbol et al., 2013;Koblinger et al., 2014;Kumru et al., 2015). It has been found that a lesion in a dorsoposterior hypothalamic dopaminergic A11 cell group, which is the sole source of spinal DA, leads to increased wakefulness across the rest phase (Ondo et al., 2000) and a long-lasting reduction in sensory thresholds (Clemens et al., 2006) in rats. D 3 receptors are mostly present in the dorsal spinal cord where it has been shown to modulate sensory pathways (Meneely et al., 2018). Mice knocked out of D 3 receptors (D 3 KO) show increased locomotion (Accili et al., 1996;Clemens et al., 2006) and decreased paw withdrawal latencies to thermal pain stimulation. Furthermore, D 3 KO mice have a reduction in frequency-dependent modulation of the longerlatency reflex (LLRs) in the spinal cord (Keeler et al., 2012). Pharmacological experiments indicate that D 3 KO mice exhibit a reversal of the modulatory actions of DA on spinal reflexes from depression to facilitation (Clemens and Hochman, 2004), and are responsive to D 1 and D 2 receptor agonists (Keeler et al., 2012). These findings emphasize the role of spinal DA in the etiology of RLS.
Our results provide a novel mechanism for the efficacy of dopaminergic agonists as treatments for RLS. The dopaminergic system plays a critical role in RLS pathogenesis. In addition to the evidence mentioned above, RLS patients have upregulated levels of L-DOPA metabolites and TH activity in CSF Allen et al., 2009). RNAi-mediated knock-down of BTBD9 homolog gene in a subset of dopaminergic neurons can reproduce sleep phenotype in fruit flies (Freeman et al., 2012). Ropinirole rescues the sensory deficit found in the systematic Btbd9 KO mice (DeAndrade et al., 2012a). It is possible that increased activity of MSNs leads to a decreased firing in the ChIs and a lower level of ACh. Nigrostriatal DA terminals express both ionotropic nicotinic ACh (nACh) and G q/11 -coupled muscarinic ACh (mACh) receptors (Cachope and Cheer, 2014). The activation of nACh receptors elicits DA release, while the downstream pathway of mACh receptor inhibits DA release. Overall, optogenetic stimulation of ChIs evokes DA release in slice preparation (Cachope et al., 2012;Cachope and Cheer, 2014). Therefore, a decreased level of ACh may cause a deficiency in DA release. It has also been found that activation of MSNs through AMPA receptor leads to the generation of a diffusible messenger, hydrogen peroxide (H 2 O 2 ), which can inhibit DA release via ATP-sensitive potassium channels (Avshalumov et al., 2008;Sulzer et al., 2017). Either way, increased MSNs activity is predicted to reduce striatal DA release in the systematic Btbd9 KO mouse model, which is consistent with the clinical finding that DA agonists can be used to treat RLS.
In summary, our results suggest that alteration in the striatal circuit, especially increased activity of the striatal MSNs, could potentially serve as a main pathogenetic mechanism of the motor and sensory dysfunction in RLS. It also supports an indirect role of the striatal ChIs in the disease development. Finally, these data present a plausible explanation for the therapeutic efficacy of the DA receptor agonists in RLS that includes the inhibition of the MSNs activity, the regulation of ChI excitability, and the resetting of the striatal circuit. Further investigations are needed to dissect the complex interactions among the striatal neurons and modulatory neurotransmitters, which will aid the development of highly selective anti-RLS drugs with fewer side effects.