Heterozygous Drd1-Cre transgenic rats (Sprague-Dawley genetic background) from Biocytogen (Beijing Biocytogen Co., Ltd., Beijing, China) were used in this study (Yu et al., 2019). To ensure that the rats had a consistent genetic background, we crossed the Drd1-Cre rats with wild-type rats for at least six generations. The genotype of the transgenic Drd1-Cre rat offspring was identified by polymerase chain reaction test on genomic tail DNA. Male rats 8–10 weeks of age and weighing 230–250 g were enrolled in the study. They were group-housed on a 12-h light/dark cycle with ad libitum access to water and rodent chow. All experimental protocols complied with the ARRIVE guidelines and were strictly conducted according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH Publications No. 8,023, revised 1978) and were approved by the Ethics Committee for Animal Experimentation of Xi’an Jiaotong University. All efforts were made to minimize the number of animals used and to reduce their suffering.

Rats were treated as shown in Figures 1A,B. All surgeries were performed with a stereotaxic frame using aseptic techniques. Rats were anesthetized with 5% isoflurane and anesthesia was maintained with 1–2% isoflurane. Unilateral injection of 6-hydroxydopamine hydrochloride (6-OHDA; dissolved in ice-cold normal saline containing 0.02% ascorbic acid, 12 μg/4 μl) was administered to the rats in the left medial forebrain bundle (coordinates: Anterior-posterior, -4.20 mm; Lateral, -1.25 mm; Dorsal, -7.80 mm relative to the bregma and the dural surface; (Paxinos and Watson, 2006) as previously described (Lindgren et al., 2010; Wang et al., 2014). Effects of the dopaminergic lesion were verified using apomorphine-induced (0.05 mg/kg, s.c.) contralateral circling test 2 weeks after the surgery (Wang et al., 2014).

The vector virus was injected into the dorsal striatum (Anterior-posterior, + 0.6 mm; Lateral, -3.6 mm; Dorsal, -3.6 mm; (Paxinos and Watson, 2006) on the side ipsilateral to the 6-OHDA lesion at 1-2 weeks after the apomorphine test. AAV5-EF1a-DIO-EYFP-WPRE-pA, AAV5-EF1a-DIO-eNpHR3.0-EYFP-WPRE-pA, AAV5-EF1a-DIO-GCaMP6m-WPRE-hGH-pA, AAV5-hSyn-DIO-mcherry and AAV5-hSyn-DIO-hM4Di-mCherry were deposited and packaged into viral vectors (BrainVTA Co., Ltd.; Hubei, China). The injection volume was 300-nl. For optogenetics and fiber photometry tests, optical ferrule fibers (outer diameter: 220 μm, OD, numerical aperture [NA]: 0.37) were implanted 0.3-0.5mm above the injection coordinates immediately after virus infusion (Figures 2A, 3A). After the implantation, dental cement (Super-Bond C&B, Sun Medical, Shiga, Japan) was used to secure the fiber to the skull. At least 4 weeks after the surgery, recovery of the animals and viral expression were allowed before the behavioral assays (Wang et al., 2019). One week after virus injection, levodopa plus benserazide (L-3, 4-dihydroxyphenylalanine methyl ester hydrochloride, 6 mg/kg; benserazide hydrochloride, 12 mg/kg, s.c; Sigma-Aldrich, St. Louis, United States) were injected once daily for 21 days (Lindgren et al., 2010). After 3 weeks of chronic levodopa treatment, each rat was observed every 20 min for 180 min and scores were assigned using the standard procedures of Abnormal Involuntary Movement scale (AIMs) for four subtypes of dyskinesia including axial, limb, orolingual, and contralateral circular locomotion on a scale from 0 to 4 (Figure 3B). The maximum AIMs score per time-point was 16 (Winkler et al., 2002). Rats with apparent dyskinetic symptoms and high AIMs scores were classified into the LID group, and rats with no apparent abnormal involuntary movements were classified into the non-dyskinetic (non-LID) group (Lindgren et al., 2010).

For monosynaptic rabies virus tracing, AAV2-EF1a-DIO-oRVG-WPRE-hGH-pA and AAV2-EF1a-DIO-H2B-eGFP-T2A-TVA-WPRE-hGH-pA (BrainVTA) were mixed at a ratio of 1:3 and then injected in the striatum. About 3 weeks later, EnvA G-deleted Rabies-dsRed (BrainVTA) was injected in the same place (Figure 4A; Callaway and Luo, 2015).

A dual-color multichannel fiber photometry system (Inper Hangzhou Bioscience Inc., Hangzhou, China) was used to detect and record the fluorescence signals of GCaMP6m. The recording mode is shown in Figure 2A. An optical fiber was integrated with a fiber-optic rotary joint for fiber photometry (FRJ_1 × 1_PT, Doric Lenses, Quebec, Canada) to guide the light between the fiber photometry system and the implanted optical ferrule fiber. Lights of 470 and 410 nm wavelengths from a light-emitting diode (Thorlabs Inc., Newton, NJ, United States) were bandpass filtered and delivered to the brain as excitation sources for the Ca²⁺-dependent and Ca²⁺-independent control measurements, respectively. Light intensity at the top of the fiber was set to 20 μW to minimize bleaching. Time-division multiplexing was implemented to acquire Ca²⁺-dependent GCaMP signals and Ca²⁺-independent control signals simultaneously (Gunaydin et al., 2014; Kim et al., 2016). The emission light that traveled through the same optical fiber was bandpass filtered (passing band: 535 ± 25 nm) and captured by a scientific complementary metal-oxide semiconductor camera (Basler; Ahrensburg, Germany) with a sampling frequency of 40frames/s (Figure 2A). Streamline fiber photometry data and real-time rat behavioral videos from a high-speed digital video camera (Logitech, Shanghai, China) were synchronized with Inper Studio data acquisition (Inper Hangzhou Bioscience Inc., Hangzhou, China). The initiation time for each dyskinetic event was determined using video scoring. The raw data of fiber photometry and real-time behavioral videos were saved in CVS and MP4 files, respectively.

Analysis was performed using custom MATLAB (MathWorks, Portola Valley, CA, United States) scripts and Photometry Modular Analysis Tool (pMAT) (Bruno et al., 2021). We subtracted the scaled 410-nm reference trace from the 470-nm signal to obtain the motion-corrected 470-nm signal. We calculated the normalized change in motion-corrected 470-nm signal (dF/F) by subtracting the median signal from the signal at each time point and dividing that value by the median signal. To calculate peri-event time heatmaps and histograms in the pMAT suite, the event window was set around the initiation time for each dyskinetic event in LID rats (Li et al., 2016; Vesuna et al., 2020). The baseline window was defined as 0.5 s preceding each event window (Figures 2F,G). The corresponding power spectral density (PSD) of the GCaMP signal (dF/F, 300 s) was then estimated using Welch’s method [pwelch() in MATLAB], with a window size of 10 fs (sampling rate of the signal). The average band power within the dominant frequency band (0.2–7 Hz) was then computed by integrating the PSD estimate [bandpower() in MATLAB] (Vesuna et al., 2020).

A yellow laser (Aurora120–589, 589 nm, 50 mW; Newdoon Inc., Hangzhou, China) was used to deliver light for optogenetic inhibition. A stimulus generator (STG4002, Multi-Channel Systems, Reutlingen, Germany) was used to control the frequency and pulse width of the laser light. The light was delivered to the brain through an optical fiber (200 μm diameter, NA: 0.37) connected to the implanted ferrule fiber using a zirconium sleeve. The light power in the brain regions 0.5 mm below the fiber tip was calibrated as described previously (Wang et al., 2019). The calibrated light power density (0.5 mm below the fiber tip) used in light deactivation experiment was 5 mW/mm² (Gradinaru et al., 2008; Stefanik et al., 2013; Lee et al., 2017). At 50–70 min after the administration of levodopa, the LID rats were placed individually in the center of the box and their dyskinetic behavior was tracked for 9min in optogenetic tests with 3 min of light inhibition (589 nm, 10.5 s-ON/2 s-OFF pulse, 0.08 Hz) applied at 3 min after the start (Figure 3A). All behaviors were videotaped and analyzed offline.

Clozapine-N-oxide (CNO, BML-NS105-0005; Enzo Life Sciences Inc., Farmingdale, United States) was freshly dissolved in saline (0.9% NaCl; 1 mg/ml). It was injected intraperitoneally at 5 mg/kg for hM4Di silencing. Behavioral tests were usually performed 30-40 min after CNO injection and immediately after levodopa administration. Saline was injected as the vehicle control. Dyskinetic behavior was tracked for 180 min in the chemogenetic tests (Figure 5A).

The rats were administered an overdose of 20% urethane at 80-100 min after the last levodopa administration and transcardially perfused with 200 ml of 0.01 M phosphate-buffered saline (PBS), followed by 300-ml of 4% paraformaldehyde. Brains were removed and placed in 4% paraformaldehyde for 8 h followed by immersion in 30% sucrose solution until sinking was observed. Brains were sectioned in 30 μm in the coronal plane using a cryostat slicer. The sections mounted on glass slides were used to identify the anatomical placement of the optical fibers and virus expression. To determine the extent of dopaminergic neuron degeneration in following injection of 6-OHDA, tyrosine hydroxylase (TH) immunofluorescence histochemistry of the SNc and striatum was performed (Figures 1C,D). Only the rats with almost total loss (>97%) of TH immunoreactivity in the left SNc were used to analyze the data. In addition, immunofluorescence histochemistry was used to observe the localization of c-Fos or D₁ staining in the striatum. The sections were incubated for 24 h at 4°C in monoclonal mouse anti-c-Fos primary antibody-containing solution (1:1,200; ab208942, Abcam, Cambridge, United Kingdom), rabbit polyclonal antibody for dopamine D₁ receptor (1:500; ab40653, Abcam) or polyclonal rabbit anti-TH primary antibody-containing solution (1:1,200; ab112, Abcam). After several washes, the sections were incubated with donkey anti-mouse IgG H&L (Alexa Fluor 488, 1:500; ab 150073, Abcam) or donkey anti-rabbit IgG H&L (Alexa Fluor 594, 1:1,000; ab 150108, Abcam) antibodies for 2 h at room temperature. The sections were then washed in PBS, mounted, and coverslipped using Fluoromount-G gel (SouthernBiotech, Birmingham, United Kingdom). The digital images were captured from the sections using a fluorescent microscope (BX51, Olympus, Tokyo, Japan). Co-localization and quantification of the fluorescence images were performed using the cell counter plug-in in ImageJ.

Data were represented as mean ± standard error of the mean or median as indicated. One-way analysis of variance (ANOVA) with post-hoc Holm-Sidak method or Kruskal–Wallis One-way ANOVA on ranks with post-hoc Student–Newman–Keuls method was used to compare three or more groups with a single variable. Two-way repeated measures ANOVA with post-hoc Holm-Sidak method was used for data with more than one independent variable over more than two time points. Statistical significance was set at P < 0.05. Data were analyzed using GraphPad Prism 8.0 (GraphPad Software, San Diego, United States) and SigmaStat 3.5 (Systat Software, Inc., San Jose, United States).

Following unilateral 6-OHDA lesion of the MFB and chronic levodopa (6 mg/kg) treatment (Figure 1A), we evaluated the effect of striatal dopaminergic denervation on total AIMs scores and striatal c-Fos expression at 80–100 min after subcutaneous injection of levodopa. The administration of levodopa (Figure 1I), but not saline (Supplementary Figure 1), increased striatal c-Fos expression of LID rats. The c-Fos expression after levodopa injection was confined in unilateral striatal areas with complete dopaminergic denervation. However, the percentage of striatal area with complete dopaminergic lesion showed no significant correlation with the total AIMs scores and striatal c-Fos⁺ neurons counts per mm² (Figures 1E,F) after almost complete TH denervation on ipsilateral striatum (Figure 1D). In contrast, the striatal c-Fos⁺ neurons counts/mm² strongly correlated with the total AIMs scores (Figure 1G).

To investigate the involvement of striatal D₁ receptor-positive neurons in the regulation of LID, we monitored c-Fos expression in D₁⁺ neurons after levodopa administration. The D₁-Cre LID rats were used in the assay at 4 weeks after intra-striatal injection of AAV5-hSyn-DIO-mCherry. Immunostaining for c-Fos revealed that levodopa administration induced substantial c-Fos expression in the striatal neurons of dyskinetic rats and D₁-Cre-mCherry marked a sizeable fraction of c-Fos⁺ neurons activated by levodopa (Figure 1I). Notably, the majority (66.97%) of the c-Fos⁺ striatal neurons activated by levodopa were positive for D₁-Cre-mCherry and 30.58% of the D₁-Cre-mCherry ⁺ striatal neurons showed c-Fos expression (Figure 1H).

In addition, we used the Cre-dependent monosynaptic retrograde rabies system to screen for the brain regions that send inputs to striatal D₁-Cre⁺ neurons (Figure 4A). Projection neurons from the cortex and thalamus, and dopaminergic neurons from the SNc were identified as the monosynaptic upstream of D₁-Cre⁺ neurons in the striatum (Figure 4B). An anterograde tracing study showed that striatal D₁-Cre-eYFP⁺ neurons sent their fibers directly to the basal ganglia output nuclei, internal globus pallidus, and substantia nigra pars reticulata. In contrast, D₁-Cre-eYFP⁺ fibers were sparse in the external globus pallidus (Figure 4C). The results confirmed that striatal D₁-Cre⁺ neurons were involved in the direct pathway of the basal ganglia.

We investigated how striatal D₁⁺ neurons responded during the dyskinetic phase in LID rats. In vivo calcium imaging with fiber photometry was used to record striatal D₁⁺ neuronal population activity of D₁-Cre LID and non-LID rats following levodopa injection and D₁-Cre normal rats without 6-OHDA lesion and levodopa administration. We virally expressed the genetically encoded calcium indicator GCaMP_6m in the D₁⁺ neurons of the striatum ipsilateral to the 6-OHDA lesion. GCaMP_6m expression was largely limited to the dorsolateral striatum (Figure 2A). We recorded stable striatal D₁⁺ neuronal GCaMP fluorescence signal from normal controls, non-LID rats, and LID rats following levodopa administration (Figures 2B–D). The signal of striatal D₁⁺ neuronal population activity from all three groups was mainly distributed in the low frequency band (0.2–7 Hz; Figure 2F). In contrast, the magnitude of striatal D₁⁺ neuronal GCaMP fluorescence signal was different among the three groups (Figures 2B–D).

Quantitative analysis showed that the average power (0.2–7 Hz) of striatal D₁⁺ neuronal GCaMP fluorescence signal in LID rats was significantly higher than that in non-LID and normal controls (Figure 2G). These findings indicated that the striatal D₁⁺ neurons in LID rats were overactivated by levodopa administration. However, it was unclear whether the neurons would be activated during each dyskinetic phase. To address this issue, we examined the temporal response profiles during the dyskinetic phases in LID rats by plotting heatmaps and peri-event time histograms (PETHs) of D₁⁺ neuronal GCaMP fluorescence signals aligned with the initiation of each dyskinetic movement. Sorting the population activity of striatal D₁⁺ neurons by the initiation time of dyskinetic movements revealed that the striatal D₁⁺ neuronal population activity peaked at each dyskinetic movement in LID rats (Figures 2E,H,I and Supplementary Video 1). We observed no associated changes in Ca²⁺-independent control fluorescence signals during dyskinetic behavior, indicating that the contribution of movement-induced artifacts to the fluorescence signal was negligible (Figures 2H,I; Kim et al., 2016).

Overactivity of striatal D₁⁺ neuronal population was observed during dyskinetic movements in LID rats, but the functional role of striatal D₁⁺ neurons in levodopa-induced dyskinetic behavior has not been fully explored. To address this issue, we virally expressed eNpHR3.0, a fast light-activated electrogenic Cl^– pump, unilaterally in the striatal D₁⁺ neurons of LID rats (Figure 3A). To determine the effects of striatal D₁⁺ neuronal deactivation on dyskinetic movements in LID rats, we delivered yellow light when the animals showed stable and severe dyskinetic symptoms (Figures 3A,B). When compared with the eYFP controls, optogenetic inhibition of striatal D₁⁺ neurons reduced locomotive, axial, limb, and total AIMs scores of eNpHR3.0-expressing LID rats (Figures 3C–F and Supplementary Video 2). However, yellow light did not affect the orolingual AIMs scores of eNpHR3.0–expressing LID rats (Figure 3G). These results suggest that the overactivity of striatal D₁⁺ neurons is necessary for most of the dyskinetic symptoms in LID rats.

We tested whether long-term inhibition of striatal D₁⁺ neuronal population activity in LID rats could affect their dyskinetic behavior after levodopa injection. A G_i-coupled designer receptor exclusively activated by designer drugs (DREADDi) conjugated with mCherry was expressed unilaterally in striatal D₁⁺ neurons following intra-striatal microinfusion of AAV5-Syn-DIO-hM4Di-mCherry-WPRE-pA (Figures 5A,B). AAV5-Syn-DIO-mCherry-WPRE-pA was used as a control. Following selective binding to CNO, DREADDi could hyperpolarize hM4Di-mCherry-expressing neurons (Dong et al., 2010). We examined the effect of striatal D₁⁺ neuronal deactivation on dyskinetic behaviors of LID rats by alternating injections of saline and CNO (5 mg/kg of body weight). LID rats were observed for 180 min and AIMs scores were assigned after systemic injections of CNO/saline and levodopa (Figures 5A,C). A two-way repeated measures ANOVA revealed significant overall differences among the three groups [F_{(2, 219)} = 8.641, P = 0.002] and across time points [F_{(9, 219)} = 381.419, P < 0.001], as well as a significant interaction between these parameters [F_{(18, 219)} = 2.767, P < 0.001]. In post-hoc comparisons (Holm-Sidak method), AIMs scores in the hM4Di+CNO group were significantly lower than those in the hM4Di+Sal and mCherry+CNO groups (P = 0.002 and P = 0.003, respectively; Figure 5C). In addition, we found that CNO injection significantly reduced the dyskinetic behavior of hM4Di-expressing LID rats in the early phase. Administration of CNO significantly increased the latency period of dyskinesia expression in hM4Di+CNO LID rats after levodopa injection when compared with hM4Di+Sal and mCherry+CNO groups (Figures 5C,D). CNO-mediated deactivation significantly reduced the duration of dyskinesia and the total AIMs score in hM₄D_i-expressing LID rats, but did not affect the highest AIMs score (Figures 5E–G).