Abstract
l-DOPA-induced dyskinesia (LID) is a debilitating motor side effect arising from chronic dopamine (DA) replacement therapy with l-DOPA for the treatment of Parkinson’s disease. LID is associated with supersensitivity of striatal dopaminergic signaling and fluctuations in synaptic DA following each l-DOPA dose, shrinking the therapeutic window. The heterogeneous composition of the striatum, including subpopulations of medium spiny output neurons (MSNs), interneurons, and supporting cells, complicates the identification of cell(s) underlying LID. We used single-nucleus RNA sequencing (snRNA-seq) to establish a comprehensive striatal transcriptional profile during LID development. Male hemiparkinsonian mice were treated with vehicle or l-DOPA for 1, 5, or 10 d, and striatal nuclei were processed for snRNA-seq. Analyses indicated a limited population of DA D1 receptor–expressing MSNs (D1-MSNs) formed three subclusters in response to l-DOPA treatment and expressed cellular markers of activation. These activated D1-MSNs display similar transcriptional changes previously associated with LID; however, their prevalence and transcriptional behavior were differentially influenced by l-DOPA experience. Differentially expressed genes indicated acute upregulation of plasticity-related transcription factors and mitogen-activated protein kinase signaling, while repeated l-DOPA-induced synaptic remodeling, learning and memory, and transforming growth factor-β (TGF-β) signaling genes. Notably, repeated l-DOPA sensitized Inhba, an activin subunit of the TGF-β superfamily, in activated D1-MSNs, and its pharmacological inhibition impaired LID development, suggesting that activin signaling may play an essential role in LID. These data suggest distinct subsets of D1-MSNs become differentially l-DOPA-responsive due to aberrant induction of molecular mechanisms necessary for neuronal entrainment, similar to processes underlying hippocampal learning and memory.
Significance Statement
These data establish a comprehensive transcriptional profile of the striatum across the development of l-DOPA-induced dyskinesia at the level of individual cells in a mouse model of parkinsonism, indicating that unique subclusters of striatal neurons differentially respond to experience with l-DOPA. These neurons have a profile enriched for markers of synaptic plasticity, neuronal entrainment underlying learning and memory, and activin signaling. Negative modulation of activin receptors dampened l-DOPA-induced dyskinesia development, suggesting that activin directly modulates aberrant behavioral sensitization to chronic l-DOPA.
Introduction
Nigrostriatal dopamine (DA) neuron loss is the cardinal hallmark of Parkinson’s disease (PD) and is commonly treated with levodopa (l-DOPA) therapy. While initially beneficial, l-DOPA treatment frequently results in the progressive development of abnormal involuntary movements, known as l-DOPA-induced dyskinesia (LID). At least 40% of patients treated with l-DOPA will experience LID after 5 years of therapy (Ahlskog and Muenter, 2001). LID is believed to stem from an imbalance in the output of striatal medium spiny neurons (MSNs) from the kinetogenic direct (striatonigral) pathway at the expense of the inhibitory indirect (striatopallidal) pathway (Cenci et al., 2018; Castela et al., 2023). Supraphysiological peaks in l-DOPA-derived DA cause the preferential activation of DA D1 receptor–expressing MSNs (D1-MSNs) within the direct pathway while causing the deactivation of DA D2 receptor–expressing MSNs (D2-MSNs) to drive LID expression. Biochemical and synaptic sensitization are most prominently found in D1-MSNs following l-DOPA exposure, while D2-MSNs show minimal changes in function during LID development (Engeln et al., 2016; Fieblinger et al., 2018; Scarduzio et al., 2022). Manipulation of either pathway can affect LID expression; however, only stimulation of D1-MSNs can independently induce dyskinesia in the absence of l-DOPA (Alcacer et al., 2017; Castela et al., 2023).
In addition to the functional differences of MSNs based on the DA receptor expression, the striatum is somatotopically organized with centrolateral striatal regions associated with sensorimotor inputs, while medial regions play a more important role in associative- and goal-oriented behaviors (Burton et al., 2015). The striatum also contains compartments with unique inputs and outputs that are known as the patch (or “striosome”) and matrix (Brimblecombe and Cragg, 2017). Due to limitations in isolating specific neuronal populations, the differential roles that these subareas may play and their respective contributions to the precise microcircuitry underlying LID remain unclear.
In response to chronic l-DOPA, D1-MSNs undergo physiological and biochemical changes. Following the induction of LID in animal models, D1-MSNs exhibit long-term potentiation that has been shown to be uniquely resistant to synaptic depotentiation (Iravani et al., 2012; Thiele et al., 2014). Underlying these synaptic changes in D1-MSNs is a persistent biochemical sensitization in mitogen-activated protein kinase (MAPK) signaling following DA D1 receptor stimulation, causing increased extracellular signal-regulated kinase (ERK)/cAMP response element–binding protein (CREB) phosphorylation following l-DOPA administration (Pavon et al., 2006; Westin et al., 2007; Santini et al., 2009, 2012; Cerovic et al., 2015; Mariani et al., 2019). Moreover, the altered dopaminergic signaling leads to long-term enhancements in the evoked transcription of several immediate early genes (IEGs), including Zif268, Arc, Fos, and Fosb, exclusively in D1-MSNs following l-DOPA treatment (Westin et al., 2007; Carta et al., 2008; Bastide et al., 2014; Fieblinger et al., 2018; Girasole et al., 2018). This biochemical and transcriptional sensitization is crucial for the manifestation of LID, and inhibition of this sensitization can block LID expression in animal models (Darmopil et al., 2009; Santini et al., 2009; Figge and Standaert, 2017). Although enhanced signaling in D1-MSNs has been implicated in LID, a clear understanding of the transcriptional response to l-DOPA across LID development is lacking. In addition, the response of the MSN subpopulations, interneurons, and non-neuron populations remains unknown. Thus, a comprehensive understanding of the cellular landscape and transcriptional network affected by l-DOPA exposure is necessary to fully understand the underlying pathophysiological mechanisms.
Although bulk RNA sequencing (seq) techniques have identified numerous changes in gene expression following LID development (Sodersten et al., 2014; Smith et al., 2016; Dyavar et al., 2020), due to the cellular complexity of the striatum (Gokce et al., 2016; Martin et al., 2019; Stanley et al., 2020), the particular cells specifically involved remain unknown. Recent advancements in single nuclei (sn) RNA-seq technologies have facilitated the exploration of complex tissues with significant cellular heterogeneity, enabling the generation of cell-type–specific transcriptional profiles during disease modeling and drug exposure. Using snRNA-seq, we studied the kinetics of the striatal transcriptional response during LID development in a 6-hydroxydopamine (6-OHDA) mouse model of PD. Our unbiased approach provides insights into cellular subtypes and gene modules involved in LID entrainment and stable expression, thereby highlighting a key role for specific D1-MSN subpopulations.
Materials and Methods
Mice
Male C57Bl/6J mice (8 weeks of age, N = 101 mice) purchased from Jackson Labs were utilized. Mice were group-housed with a 12 h light/dark cycle. Food and water were provided ad libitum. All studies were approved by and complied with the University of Alabama at Birmingham Institutional Animal Care and Use Committee's guidelines and protocols. The sample sizes for the experiments were based on pilot data and prior publications. All individuals responsible for behavioral measurements and data analysis were blind to the experimental group.
Experimental design
Experiment 1: Cell-type–specific transcriptional profile of LID development
6-Hydroxydopamine lesion surgeries
A time course of the experimental design for Experiment 1 is included in Figure 1A. To induce hemiparkinsonism, mice (N = 51) were subjected to a unilateral infusion of 6-OHDA into the medial forebrain bundle to deplete striatal DA by utilizing standard stereotaxic survival surgery techniques, as previously described (Thiele et al., 2012). Mice were treated with desipramine HCl (25 mg/kg, ip) 15 min prior to surgery to protect noradrenergic neurons. 6-OHDA HBr (0.2 µl in 0.2% ascorbic acid and 0.9% saline; 3 µg/µl) was injected with a Neuros 33 gauge 10 µl syringe (Hamilton Apparatus) at the following coordinates: AP, −1.2; L, −1.1; and V, −5.0 mm (Franklin and Paxinos, 1997). Mice received buprenorphine analgesic and lactated Ringer's solution (sc) to prevent dehydration and were allowed to recover on a heating pad for up to 8 h postsurgery. Any mice that expressed distress or dehydration during the 7 d recovery period were treated with additional fluids and/or additional heating pad time. Some mice did not recover from surgery despite these interventions (∼10–20% across all experiments).
Clustering analysis reveals all major striatal cell subtypes. A, Time course of experiment and graphical description of experimental groups for snRNA-seq experiments. B, ALO AIMs and (C) contralateral rotations (±SEM) for the first 60 min of the final behavioral session for each group of animals before tissue harvesting. D, Percentage of TH+ cells in nigral sections on intact and DA lesioned side (±SEM). E, Integrated UMAP of striatal cellular subclustering across all experimental groups (>97,000 nuclei). F, Dot plot examining % of nuclei expressing cluster-defining genes (size of the dots) and their average expression (color of the dots) within each cluster. See Extended Data Tables 1-1 and 1-2 for more details.
Table 1-1
Marker genes utilized for identification of cellular subtypes. Download Table 1-1, XLSX file.
Table 1-2
Cluster defining genes in order of population dominance. Download Table 1-2, XLSX file.
Vehicle-induced rotations
Vehicle-induced rotations (Boix et al., 2015) were utilized to separate animals into equal groups without exposing them to a dopaminergic drug that could induce a priming effect. Two weeks post-surgery, mice were individually placed into clear acrylic cylinders. After 30 min of habituation, mice received an injection of saline (0.2 ml, sc) and were immediately returned to the cylinder. For 5 min postinjection, ipsilateral and contralateral rotations were counted, with a tendency toward ipsilateral rotations considered indicative of hemiparkinsonism. Mice were ranked based on the number of ipsilateral–contralateral rotations and separated into equal groups as follows: vehicle-treated or l-DOPA-treated for 1 (Acute), 5 (Subchronic), or 10 d (Chronic).
Abnormal involuntary movements scale
Three weeks postsurgery, mice (n = 6/group) commenced 10 d of treatment with either vehicle (0.9% sodium chloride plus 0.1% ascorbic acid, Sigma-Aldrich) or l-DOPA methyl ester hydrochloride (4 mg/kg plus 15 mg/kg benserazide hydrochloride, Sigma), as indicated in Figure 1A. LID severity was assessed with the AIM rating scale (Lundblad et al., 2002, 2005) for 3 h on Days 1 and 5 and 1 h on Day 10 prior to harvesting of striata. In brief, mice were individually placed into the same clear acrylic cylinders utilized for vehicle-induced rotations. After 30 min of habituation, mice received an injection of vehicle or l-DOPA and were returned to the cylinder. Axial, limb, and orolingual (ALO) Abnormal involuntary movements scale (AIMs) were separately quantified after observation for 1 min every 20 min on a scale of 0–4 (0 = absent; 1 = present for <30 s; 2 = present for >30 s, but <1 min; 3 = present for 1 min, but interruptible by a tap on the cylinder; 4 = present for 1 min, but uninterruptable by a tap on the cylinder). Contralateral rotations were also counted as a traditional measure of differential DA sensitivity (Norman et al., 1990). Mice that displayed a total AIM score <10 were removed prior to snRNA-seq processing to avoid the introduction of variability.
Nuclei preparation and snRNA-seq
One hour after the last vehicle or l-DOPA injection, mice (n = 2–4 mice/sample, two samples/group, four groups) were anesthetized with isoflurane and decapitated. All animals included in pooled samples representing l-DOPA-treated groups (Acute, Subchronic, and Chronic) were required to display a summed ALO AIM score of >10 from the 0 to 60 min time points prior to tissue harvesting for inclusion. The hindbrain was removed and placed into cold 4% paraformaldehyde in 0.1 M phosphate-buffered saline for fixation and consequent quantification of DA neuron depletion in the substantia nigra. Left (lesioned) and right (intact) striata were removed by microdissection and pooled with 1–3 other mice that were in the same treatment duration group. Nuclei isolation was completed, as previously published (Savell et al., 2020). The tissue was chopped orthogonally and lysed for 15 min in 15 ml of lysis buffer containing 10 mM Tris-HCl, 10 mM NaCl, 3 mM MgCl2, and 0.1% Ipegal in nuclease-free water. The reaction was quenched with 5 ml of Hibernate-A (Thermo Fisher Scientific) containing B27, GlutaMAX (Life Technologies), and 0.2 U/µl of RNAse inhibitor (Lucigen). The tissue was then triturated 15 times through a fire-polished Pasteur pipette and passed through a 40 µM filter to remove cell debris and large chunks. The filtered solution was centrifuged at 500 rcf for 10 min; the supernatant was removed; and pelleted nuclei were washed with 10 ml 1× phosphate-buffered saline with 1.0% bovine serum albumin and 0.2 U/µl RNase inhibitor. This was centrifuged at 500 rcf for 5 min, and the nuclei pellet was resuspended in 300 µl of the wash buffer. Nuclei were then tagged with propidium iodide and purified using flow-assisted cell sorting (BD FACS Aria II, 70 µM nozzle, BD Biosciences) and brought to a concentration of 1,000 nuclei/µl. An average total of 10,000 nuclei/sample was loaded into each well of the Chromium Single Cell B Chip (10X Genomics). Single nuclei libraries were constructed according to the instructions provided in the Chromium Single Cell 3′ Library Construction Kit (10X Genomics). Nuclei were sequenced on Illumina NovaSeq with a minimum depth of 15,000 reads/nuclei. Sequencing files were processed and mapped to mm10, and count matrices were extracted using the Cell Ranger Single Cell Software (v3.1.0).
snRNA-seq analyses
All analyses were conducted in R using Seurat v3.1, as previously published (Schonhoff et al., 2023). Briefly, all data underwent quality control, filtering any genes expressed in <3 cells and excluding cells with <300 unique transcripts or >3% mitochondrial genes. The datasets were then integrated based on the 2,000 most variable genes. The data were normalized, and the top 20 principal components were used for UMAP dimensional reduction. Clustering was performed following the identification of nearest neighbors at a resolution of 0.7. Clusters predominantly composed of either cells with low RNA content or doublet cells were removed, and the datasets were reclustered following dimensional reduction. Marker genes for each cluster were determined using the FindAllMarkers function of Seurat with a minimum log2 fold change threshold of ±0.25 with the Wilcoxon ranked-sum test. For the direct comparisons between clusters, we used FindMarkers with similar thresholds. Filtered and merged datasets were imported into Monocle 3 (Trapnell et al., 2014). Module analyses were conducted using standard Monocle commands. GO analyses were completed using Metascape (Zhou et al., 2019) and/or PantherDB (Thomas et al., 2022). The dataset was compared with two additional datasets to (1) indicate the D1-MSN cell-subtype specificity of prior D1-MSN RNA-seq data in a model of LID (Heiman et al., 2014) and (2) compare the transcriptional profile of D1-MSNs differentially activated by l-DOPA treatment to hippocampal engram cells in a model of learning and memory (Marco et al., 2020). Data from the Marco et al. (2020) dataset were reanalyzed to generate DEGs from nRNA-seq data from initial exposure compared with reactivation as described above. Transcription factor enrichment was determined using a LISA Cistrome database TR ChIP-seq combined model (Qin et al., 2020) on the top 200 DEGs (based on LogFC) obtained from the pseudobulk-seq analysis of D1-MSN subpopulations.
Experiment 2: Independent verification of top gene upregulated in chronic versus acute l-DOPA-treated D1-MSNs
Fluorescence in situ hybridization
We utilized RNAScope Multiplex Fluorescent Assay (ACD Biosciences) in a separate group of 6-OHDA-lesioned mice to verify the enhanced expression of Inhba in D1-MSNs that also expressed Fos in response to chronic compared with acute l-DOPA treatment. Mice (n = 8–9/group) were treated with vehicle or l-DOPA (4 mg/kg plus benserazide, 15 mg/kg, sc) for 1 or 10 d and killed 1 h after injection. AIMs were assessed on Days 1 and 5 for 2–3 h and Day 10 for 1 h. Brains were removed and flash-frozen on dry ice. Sections (20 µM) were obtained using a cryostat, collected on slides, and refrozen immediately. Tissue was fixed with 4% paraformaldehyde, followed by ethanol dehydration and treatment with hydrogen peroxide and protease III (ACD). Tissue was then incubated according to the prescribed protocol (ACD) with probes targeted for Inhba, Fos, and Drd1. Slides were coverslipped with ProLong Gold Antifade Mounting Media with DAPI (Thermo Fisher Scientific), and images were captured at 20× objective with a Nikon confocal microscope (n = four animals/treatment, three sections/animal). All settings were held constant across all groups for a given experiment. Images were analyzed using QuPath (Courtney et al., 2021). Cell detection was performed using a DAPI signal following optimized parameters of QuPath defaults. Next, we performed cell classification based on fluorescent intensity thresholds of each of the selected channels: Inhba (fluorescein), Fos (Cy3), and Drd1 (Cy5). We applied the “object classification function” and created a single measurement classifier for each channel. We labeled cells based on the presence of fluorescence, considering the intensity threshold, for each of the applied channels (Inhba, Fos, and Drd1). After this last process, the data were exported containing the number of detections in annotation measurements.
Experiment 3: Validation of the role of the activin pathway in LID
Pharmacological approach
To verify the functional role of activin in LID indicated by our snRNA-seq analysis, we designed an LID development experiment in the presence and absence of the activin receptor ALK4/TGFβ1R inhibitor, SB431542 (Fig. 5E). Briefly, a separate group of mice (n = 6–8/group), which underwent unilateral 6-OHDA MFB lesions, was assessed for vehicle-induced rotations 3 weeks later as in Experiment 1 and separated into two groups: vehicle + l-DOPA and SB431542 + l-DOPA. Mice received either vehicle (20% DMSO in 0.9% NaCl) or SB431542 (4.2 mg/kg, ip) 15 min prior to injection with l-DOPA (2 mg/kg + benserazide, 15 mg/kg, sc). A lower dose of l-DOPA was utilized compared with that used for the snRNA-seq experiments to allow for the detection of either an improvement or worsening of LID due to ALK4/TGFβ1R inhibition. On Days 1, 3, 5, and 10 of daily treatment, AIMs and rotations were assessed as in Experiment 1. On Day 11, all animals were killed 1 h after their respective treatments according to group assignment by isoflurane overdose and transcardial perfusion with 0.1 M PBS, followed by 4% paraformaldehyde. Brains were postfixed overnight in 4% paraformaldehyde and then cryoprotected with 30% sucrose in 0.1 M PBS.
Immunohistochemistry
Cryoprotected brains were sliced on a freezing sliding microtome (40 µm) and stored in 50% glycerol/0.1 M PBS. Every sixth slice containing striatum was washed in TBS before blocking for 2 h with 5% normal donkey serum. To indicate on-target efficacy of the ALK4/TGFβR1 inhibitor, primary antibodies for phospho-SMAD2 (Cell Signaling Technology) and NeuN (to indicate neuronal involvement; Invitrogen-Thermo Fisher Scientific) were exposed overnight at 4°C to free-floating sections (three sections/animal) at concentrations (1:500–1:1,000) and incubated with corresponding fluorescence-conjugated secondary antibodies (1:1,000; Invitrogen-Thermo Fisher Scientific). Sections were also treated with primary antibodies for Iba1 to indicate microglial involvement (Invitrogen-Thermo Fisher Scientific). Sections were coverslipped using ProLong Gold Antifade and stored at 4°C. Images were captured using a Nikon Ti2-C2 confocal microscope at 20× objective. Images were analyzed using QuPath, as described for fluorescence in situ hybridization (FISH). Nigral sections were utilized as in Experiment 1 to verify unilateral DA cell loss.
Statistical analyses
Behavior obtained in Experiments 1 and 2 was analyzed using one-way ANOVA of averaged ALO AIM scores for the first 60 min of behavioral ratings immediately before tissue was harvested for snRNA-seq or FISH. TH+ immunofluorescence was analyzed by one-way ANOVA. No post hoc analyses were warranted in either measure. Behavior in Experiment 3 was analyzed using two-way ANOVA of the summed averages of ALO AIM scores for the entire 3 h rating period with the day of exposure and drug treatment as the independent variables, followed by Bonferroni’s post hoc analyses. Immunofluorescence and FISH data were analyzed using two-way ANOVAs of the percentage of total cells expressing specific mRNA/protein averaged across each slice per subject and subsequently across conditions, followed by Bonferroni’s post hoc analyses.
Results
l-DOPA-induced stable expression of dyskinesia in 6-OHDA-lesioned mice
l-DOPA, but not vehicle, treatment in 6-OHDA-lesioned mice induced dyskinetic movements and contralateral rotations (Fig. 1B,C; FALO AIMs(3,20) = 234.8, p = 0.0005; FRotations(3,20) = 14.12, p < 0.0001). ALO AIMs were prominent even in the Acute-1 d l-DOPA-treated group compared with vehicle (p < 0.0001); however, they did not significantly increase over time, despite the low dose of l-DOPA (4 mg/kg; Acute-1 d vs Subchronic-5 d, p = 0.6568; Acute-1 d vs Chronic-10 d, p = 0.1243; Subchronic-5 d vs Chronic-10 d, p = 0.2086). Likewise, contralateral rotations were also prominent from the first l-DOPA treatment compared with vehicle (p = 0.0004) and did not increase significantly over time (Acute-1 d vs Subchronic-5 d, p = 0.8693; Acute-1 d vs Chronic-10 d, p = 0.7522; Subchronic-5 d vs Chronic-10 d, p = 0.2917). Immunofluorescence for TH+ nigral cells revealed a similar extent of loss across all groups when comparing the lesioned side to the unlesioned side (Fig. 1D; FTH Loss(3,20) = 1.013, p = 0.4076).
Cluster analyses of transcription reliably identified cell-type heterogeneity of the dorsal striatum
After quality control and normalization, we analyzed 97,848 cells (Intact-Naive: 13,549; DA Lesioned-Naive: 14,362; DA Lesioned-Acute 1 d l-DOPA: 26,604; DA Lesioned-Subchronic 5 d l-DOPA: 28,221; DA Lesioned-Chronic 10 d l-DOPA: 15,112). Unbiased cluster analysis revealed 27 unique clusters (Fig. 1E). These clusters were characterized based on their unique marker genes and comprised all major cell types, including MSNs, interneurons, progenitor cells, and glia (Fig. 1F, Extended Data Tables 1-1, 1-2). Maturity status within each cell population was indicated by the expression of Tnr, an extracellular matrix glycoprotein associated with inhibition of morphological change and highly expressed after birth (Jakovcevski et al., 2013). Pdyn+/Drd1+ and Penk+/Drd2+ MSNs (putatively D1- and D2-MSNs, respectively) were found to segregate independently and included multiple unique clusters. The expression of Sema5b and Sgk1 has been previously associated with the patch and matrix subareas of the striatum, respectively (Martin et al., 2019; Stanley et al., 2020), and readily defined transcriptionally unique subclusters of D1- and D2-MSNs in our data set. Likewise, among matrix-localized MSNs, Cnr1 has been associated with centrolateral topographical localization, while Crym is more prevalent among matrix MSNs with medial localization (Martin et al., 2019; Stanley et al., 2020). We were also able to detect disparate subclusters of MSNs with these genes enriched. Other subclusters of cells sharing MSN marker genes were further identified based on their commonalities with so-called “eccentric” MSNs (Ecc; Martin et al., 2019; Stanley et al., 2020), although little is known of their function. Furthermore, some clusters of Drd1+ and Drd2+ MSNs were transcriptionally distinct and subclustered independently from other MSNs but did not meet the characteristics of Ecc MSNs. These neurons were enriched in activity-dependent gene expression programs elicited in striatal neuron cultures following KCl depolarization for 1 h (Phillips et al., 2023) and were, therefore, labeled as Activated (Act).
l-DOPA exposure, but not DA lesion, differentially affects the cellular proportions of striatal cells
Comparison of intact and DA-lesioned vehicle-treated striata revealed minimal changes in major cell-type population ratios (i.e., MSNs, interneurons, glia, oligodendrocytes, and progenitors; Fig. 2A,B; Extended Data Table 2-1). In l-DOPA-treated striata, the proportion of nuclei isolated from oligodendrocytes and glia nearly doubled after 1 d (Acute) and 5 d (Subchronic) l-DOPA exposure, though cellular proportions approached l-DOPA-naive levels after 10 d (Chronic) of l-DOPA exposure. The oligodendrocyte effect was largely driven by oligodendrocyte progenitors and immature oligodendrocyte subtypes; radial glia, astrocytes, and microglia populations were all altered l-DOPA treatment duration (Fig. 2B, Extended Data Table 2-1). These data are likely indicative of the supportive network that is necessary for synaptic and cellular remodeling of neurons upon l-DOPA treatment that leads to LID and will require future investigation. In addition, glial activation has been indicated previously in similar LID models and suggests an immune and/or inflammatory reaction in response to l-DOPA treatment (Kuter et al., 2020; Ferrari et al., 2021; Pinna et al., 2021; Morissette et al., 2022; Elabi et al., 2023; Nascimento et al., 2023). Although the overall populations of MSNs were not affected by l-DOPA treatment, the proportion expressing marker genes consistent with activation was increased in terms of D1-MSNs and reduced in terms of D2-MSNs compared with l-DOPA-naive mice (Fig. 2B, Extended Data Table 2-1).
l-DOPA treatment alters major and minor subpopulation proportions and transcription within striatal cells. A, Representative UMAPs of the striatal cellular distribution within each experimental group. B, Heat map depicting the population proportions of major and minor cellular subpopulations within each l-DOPA exposure group depicted as a percentage of the population sizes of the DA-lesioned, vehicle-treated group. See Extended Data Table 2-1 for more details. C, Effect of l-DOPA exposure for 1 (Acute), 5 (Subchronic), or 10 d (Chronic) on the number of discrete differentially expressed genes (DEGs; log2 fold change > 0.10) depicting either decreased or increased transcription in each striatal cell subpopulation. See Extended Data Table 2-2 for more details.
Table 2-1
Effect of 6-OHDA lesion and exposure to L-DOPA treatment on discrete populations of striatal cell subtypes in order of population prevalence. Download Table 2-1, XLSX file.
Table 2-2
Differentially expressed genes (adjusted p value <0.05, absolute log fold2 change >0.10) by group versus vehicle-treated, dopamine-lesioned striata. DEGs with increased expression are in red, decreased expression in blue. Download Table 2-2, XLSX file.
l-DOPA exposure, but not DA lesion alone, differentially affects the transcription of striatal cells
Transcriptionally, DA lesion alone resulted in <150 differentially expressed genes (DEGs) in any individual cluster (Extended Data Table 2-2). The populations most affected by DA lesion were neural progenitor cells and subtypes of oligodendrocytes and glia, with very few changes observed in neurons. In contrast, acute l-DOPA treatment induced a vast upregulation of genes across most cellular subtypes, with the majority exhibiting >150 DEGs when compared with l-DOPA-naive, DA-lesioned mice (Fig. 2C). Consistent with previous findings, interneurons and D2-MSNs were minimally affected by l-DOPA treatment, whereas D1-MSNs, glia, and oligodendrocytes were predominantly affected. These effects were similarly seen following 5 d (Subchronic) of exposure to l-DOPA. By Day 10 (Chronic) of l-DOPA treatment, while there were fewer DEGs observed overall (although still more than l-DOPA-naive, DA lesioned striata), there were roughly equivalent numbers of DEGs up- and downregulated among individual subpopulations. Among D1-MSNs, the preponderance of upregulated DEGs was highest among Act D1-MSNs. Among non-Act matrix D1-MSNs, the medially localized (Crym+) D1-MSNs expressed fewer DEGs overall than their centrolaterally localized (Cnr+) counterparts.
To grasp the holistic effects of l-DOPA treatment on striatal transcription, a pseudobulk-seq analysis was completed on all striatal cell subtypes (Extended Data Table 2-3). In general, the transcriptional differences were more pronounced acutely, with the number of unique transcripts the highest at this time point. As observed previously, IEGs were upregulated by acute l-DOPA treatment (Fos, Fosb, Junb, etc.). Other genes commonly associated with LID were also detected, including Pdyn, Arc, and Cytb (Sodersten et al., 2014; Smith et al., 2016; Dyavar et al., 2020). Glial- and oligodendrocyte-associated genes, such as Cx3cr1, Ptn, and Mbp, were further upregulated by acute l-DOPA treatment consistent with their increased cellular proportions following the l-DOPA exposure. Pseudobulk analysis of repeated (Subchronic-5 d and Chronic-10 d) l-DOPA-treated striata indicated a substantial reduction in the number of DEGs compared with acutely l-DOPA-treated (1 d) striata, although Fos and Pdyn transcription remained elevated.
Table 2-3
Download Table 2-3, XLSX file.
Distinct subpopulations of D1-MSNs are altered by l-DOPA treatment and undergo subtype- and exposure-dependent transcriptional changes indicative of cellular activation and remodeling
Although overall proportions of D1- and D2-MSNs were not altered by DA lesion or l-DOPA treatment, individual subclusters underwent extreme rearrangement in response to l-DOPA treatment that was dependent upon the length of exposure to l-DOPA (Fig. 3A–C; Extended Data Table 2-1). l-DOPA exposure in lesioned striata, but not DA lesion alone, was associated with an increased proportion of Act D1-MSNs and a decreased proportion of Act D2-MSNs (Fig. 3D; Extended Data Table 2-1). In response to l-DOPA treatment, D1-MSNs expressing markers from both patch and matrix displayed activation-dependent gene expression that included enhanced IEG expression causing them to subcluster separately. These different subtypes of D1-MSNs were differentially affected in their population proportion by the number of l-DOPA exposures (Fig. 3A–C). The proportion of Act Patch D1-MSNs increased due to l-DOPA experience and stayed elevated through Day 10 (Chronic) compared with lesioned, untreated striata. Nonactivated Patch D1-MSNs were nearly nonexistent on Days 1 (Acute) and 5 (Subchronic) of exposure to l-DOPA, indicating that a large proportion of this subpopulation was identified with the IEG+ cluster (Fig. 3A,B). Matrix-localized Act D1-MSNs transcriptionally diverged into two unique subclusters (Act Matrix D1-MSNs-1 and Act Matrix D1-MSNs-2). These subclusters exhibited many of the same CDGs (158 genes) but were transcriptionally distinct enough to cluster separately from one another. In general, Act Matrix D1-MSNs-1 shared more CDGs with CeLa (83 genes) and Me (23 genes) D1-MSNs compared with Act Matrix D1-MSNs-2 (71 and 12 genes, respectively), suggesting that these populations are indicative of a spectrum of activation. Indeed, both subpopulations of Matrix Act D1-MSNs acutely increased due to l-DOPA experience, although Population 2 remained elevated and Population 1 ebbed to untreated levels after 10 d (Chronic) of l-DOPA. While the matrix D1-MSN subtypes that Act Matrix D1-MSNs likely arise from can only be speculated upon, based on the ratios of populations, it appeared that acute exposure to l-DOPA reduced cellular populations of Me D1-MSNs and CeLa D1-MSNs. Upon further experience with l-DOPA (Subchronic-5 d and Chronic-10 d), the proportion of Act Matrix D1-MSNs slightly reduced with concomitant increases in non-Act Me D1-MSNs and CeLa D1-MSNs, with a greater proportion of Me D1-MSNs returning to untreated levels by Day 10 (Chronic). As such, it seems likely that the Matrix D1-MSNs that were in an activated state were arising predominantly from the centrolateral striatum. In contrast, Act D2-MSNs were proportionally reduced by l-DOPA treatment, regardless of the duration of exposure (Fig. 3C).
MSN subpopulations are specifically altered by l-DOPA exposure that leads to LID. A, Representative UMAPs of the striatal MSN populations within each experimental group. B, C, Population ratios of D1-MSN (B) and D2-MSN (C) subpopulations compared with intact, l-DOPA-naive striata. D, UMAPs of expression of the immediate early genes Arc, Fos, and Fosb in each experimental group. The order of UMAPs from left to right follows that depicted in A. E, Gene module analysis of integrated MSNs utilizing MonocleTM. See Extended Data Table 3-1 for more details. F, Expression profiles for genes within Modules 3, 4, 6, and 9 that were enriched in Act MSNs. Selected module genes are listed next to UMAPs for each module in italics. Selected significant terms from gene ontology analysis for Modules 3, 4, 6, and 9 are denoted beneath each UMAP. See Extended Data Table 3-2 for more details. G, Expression profile of DEGs observed in RNA-seq derived from a Drd1a-BAC-TRAP rat model of LID (Heiman et al., 2014) mapped onto MSNs from the current snRNA-seq dataset. Left side, Violin plot depicting the ratio of coexpressed DEGs within integrated MSN subpopulations. Right side, Mapped enrichment of the MSNs from the current dataset compared with that published by Heiman et al. (2014). H, Expression profile of DEGs induced in hippocampal cells by retrieval of a previously entrained stimulus–response association observed in nRNA-seq derived from an Arc-BAC-TRAP mouse model of associative learning (Marco et al., 2020) mapped onto MSNs from the current snRNA-seq dataset. Left side, Violin plot depicting the ratio of coexpressed DEGs within integrated MSN subpopulations. Right side, Mapped enrichment of the MSNs from the current dataset compared with that published by Marco et al. (2020).
Table 3-1
Module analysis of MSNs by MonocleTM. Download Table 3-1, XLSX file.
Table 3-2
Gene Ontology and transcription factor enrichment analyses of modules 3, 4, 6, and 9 from Extended Data Table 3-1. GO was conducted using Metascape. Transcription factor analyses conducted via LISA CistromeDB ChIP-seq database on top 200 genes (based on LogFC value). Download Table 3-2, XLSX file.
In order to better understand the transcriptional profile of each subcluster of MSNs, a module analysis was completed on the integrated data (Fig. 3E; Extended Data Table 3-1). Act D1-MSNs were hierarchically distinct from all other MSN subclusters, as determined by this unbiased analysis. Four modules (Modules 3, 4, 6, and 9) were also hierarchically distinct from other gene modules and differentially enriched in Act D1-MSNs (Fig. 3F). Gene ontology analyses were conducted on the genes included in each module (Fig. 3F; Extended Data Table 3-2). Act Matrix D1-MSNs-1/2 had the highest enrichment of Module 4, which included several IEGs (Fos, Junb, Fosb, etc.) and was enriched in gene ontology (GO) terms associated with “learning” and “positive regulation of transcription.” In comparison with Act Matrix D1-MSNs-1, Act Matrix D1-MSNs-2 expressed more genes associated with Modules 3 and 6. These modules included genes associated with “long-term memory” and “neuron projection development.” These modules were most enriched in Act Patch D1-MSNs.
To verify our results indicating the greater proportion of l-DOPA-induced transcriptional change was within D1-MSNs assigned to the activated subpopulations, we determined the enrichment of DEGs detected by Heiman et al. (2014) to our snRNA-seq data (Fig. 3G). Heiman et al. (2014) utilized a Drd1-TRAP methodology in 6-OHDA-lesioned, hemiparkinsonian rats chronically treated with l-DOPA. We mapped the expression of DEGs observed within their low-dose (6 mg/kg) l-DOPA-treated group onto our integrated MSN UMAP. DEG coexpression was enriched in Act D1-MSNs, particularly Act Patch and Act Matrix-2 subclusters.
Transcriptional effects of DA lesion and l-DOPA exposure on striatal D1-MSNs
In order to observe the overall DA lesion and l-DOPA exposure–induced changes in D1-MSNs and D2-MSNs, a pseudobulk analysis was completed using MonocleTM. Changes in gene transcription in D2-MSNs were minimal, while those in D1-MSNs were robust (Extended Data Table 4-1). Likewise, DA lesion alone did not positively affect the transcription of many genes in MSNs (Extended Data Table 4-1), but in D1-MSNs led to a reduced expression of active transcription of genes involved in cellular activation and synaptic transmission, including glutamatergic signal transduction and G-protein coupled signaling (Extended Data Table 4-2). In addition, genes associated with synaptic remodeling (Reln, Ntrk, Homer1) were also reduced by DA lesion, indicative of a loss of synaptic plasticity (Picconi et al., 2003; Belujon et al., 2010; Thiele et al., 2014).
The comparison of acute (1 d) and repeated (5 and 10 d) l-DOPA exposure to vehicle treatment in D1-MSNs revealed an evolution of transcriptional activity across the development of LID (Fig. 4A–C). Early in l-DOPA exposure, D1-MSNs exhibited a coordinated, bidirectional gene response indicative of enhanced DNA transcription activation and reduced protein degradation (Fig. 4A2). The expression of Drd1 was reduced by acute l-DOPA, which was perhaps a compensatory response to the influx of DAergic stimulation. The upregulation of genes in D1-MSNs by acute (1 d) l-DOPA was much stronger in comparison, with >120 genes with over two times greater expression in comparison with those isolated from vehicle-treated striata (Fig. 4A1). This massive upregulation of genes was coupled with enhanced expression of multiple transcription factors involved in the complex formation of the pioneer transcription factor AP-1 (Jun, Fos, Fosb, etc.) and transcriptional activation (Crem, Stat6, Ell2, Smarca5, Tet3, Ebf1, Per1, etc.; Fig. 4A3). The PD risk gene, Lrrk2, was upregulated and associated strongly with other upregulated genes associated with protein modification. TF enrichment through LISA Cistrome analysis indicated that Creb1 was highly associated with acute (1 d) l-DOPA exposure, as well as Fos, Fosb, and Mef2a, among others (Fig. 4B3).
Transcriptional effects of DA lesion and LID development and stable expression revealed by the pseudobulk-seq analysis of D1-MSNs. Naïve, lesioned (green) compared with (A) acute (red; 1 d) l-DOPA-treated, lesioned or (B) repeated (blue; 5 and 10 d) l-DOPA-treated, lesioned striatal D1-MSNs. C, Acute compared with repeated l-DOPA-treated, lesioned striatal D1-MSNs. (1) Volcano plots depicting DEGs (log2 fold change > 0.10) with selected genes highlighted. (2) Selected significantly enriched terms chosen from GO analyses of DEGs enriched in l-DOPA-treated groups. (3) LISA-generated enriched TFs associated with DEGs. See Extended Data Tables 4-1–4-4 for more details.
Table 4-1
Pseudobulk-seq of D1-MSNs by group. Download Table 4-1, XLSX file.
Table 4-2
Gene Ontology analyses of D1-MSN pseudobulk-seq from Extended Data Table 4-1. GO was conducted using Metascape. Download Table 4-2, XLSX file.
Table 4-3
Transcription factor enrichment of DEGs from Extended Data Table 4-1 conducted using LISA CistromeDB ChIP-seq database on top 200 genes (based on LogFC value). Download Table 4-3, XLSX file.
Table 4-4
Differentially expressed genes by subcluster in DA-lesioned, Acute (1 d L-DOPA) versus DA-lesioned, Chronic (10 d L-DOPA) striata. DEGs with increased expression are in red, decreased expression in blue. Download Table 4-4, XLSX file.
Repeated (Subchronic-5 d and Chronic-10 d) l-DOPA exposure was positively associated with many of the same DEGs, as revealed by acute (1 d) l-DOPA, although the strength of the response (LogFC) was attenuated in most cases (Fig. 4B1, Extended Data Table 4-1). DEGs related to physical changes in neural structure and modulation of synaptic transmission were enriched after repeated l-DOPA in D1-MSNs, while those associated with regulation of membrane potential were negatively associated compared with untreated mice (Fig. 4B2, Extended Data Table 4-2). TF enrichment of Fos, Fosb, and Mef2a/c was indicated in D1-MSNs from mice undergoing repeated l-DOPA treatment (Fig. 4B3).
To better understand the processes that differentiate entrainment (first exposure, Acute-1 d) and reconsolidation (repeated exposure, Subchronic-5 d and Chronic-10 d), the D1-MSN response to acute (1 d) l-DOPA was compared with that observed in Subchronic-5 d and Chronic-10 d groups (Fig. 4C1, Extended Data Table 4-1). Unsurprisingly, IEGs were significantly enriched in D1-MSNs from striata of the acute group compared with the repeated exposure groups. Other genes more strongly associated with entrainment were involved in protein dephosphorylation and folding, histone demethylase activity, and glutamate receptor activity (Fig. 4C2, Extended Data Table 4-2). In contrast, repeated exposure resulted in the enrichment of genes linked to mRNA splicing and modulation of synaptic transmission. The identification of putative TFs responsible for the differences among acute and repeated exposure groups in D1-MSNs suggested a role for Creb1, Srf, Myc, and Foxo1 in the regulation of the genes expressed following acute l-DOPA, whereas repeated exposure-induced genes that were dependent on Mef2a/c, Fos, Junb, and Fosb for their expression (Fig. 4C3, Extended Data Table 4-3).
Unique role for activins within striatal D1-MSNs indicated in the stable expression of LID
Pseudobulk analyses of D1-MSNs indicated that the gene most differentially affected by repeated exposure to l-DOPA in comparison with a single exposure was Inhba, which encodes for the protein inhibin subunit βA, a member of the transforming growth factor-β (TGFβ) superfamily (Fig. 4C1, Extended Data Table 4-1). UMAP mapping (Fig. 5A) and cluster-specific DEG analyses (Extended Data Tables 2-2, 4-4) revealed that Inhba was robustly induced in patch and matrix Act D1-MSN subclusters as animals received repeated exposure to l-DOPA. To verify this result, a separate group of hemiparkinsonian mice treated acutely or chronically with l-DOPA or vehicle was utilized for localization and quantification. Analysis of striatal tissue using FISH indicated an effect of the l-DOPA experience on gene colocalization (FGroup(2,8) = 32.87, p = 0.0001; FTranscript(1.266, 10.13) = 18.28, p = 0.0010; FInteraction(4,16) = 28.15, p < 0.0001). Both acute and chronic l-DOPA led to upregulation of coexpression of Fos transcript in Drd1+ cells compared with untreated mice (p = 0.0257 and p = 0.0248, respectively), albeit levels being the highest in acutely treated mice (Fig. 5C,D). This result was also indicated in our snRNA-seq dataset (Fig. 5B). A significant induction of Inhba transcription in Drd1+ cells was observed in animals with chronic l-DOPA exposure, in comparison with acutely treated or untreated mice (Fig. 5C,D; p = 0.0111 and p = 0.0065, respectively). Furthermore, chronic l-DOPA induced a significant increase in Inhba transcript in those Drd1+ cells that were also Fos+ (likely Act D1-MSNs; p = 0.0325 and p = 0.0253 in comparison with Untreated-0 d and Acute-1 d treated mice, respectively), corroborating results obtained using snRNA-seq.
Inhba/Activin/ALK4 signaling is enhanced by repeated l-DOPA treatment in Fos+ D1-MSNs and modulates LID development. A, UMAPs of the Inhba and Tgfb1 mRNA expressions in each experimental group. B, Confocal dorsal striatal images (20×) depicting enhancement of the Inhba transcript in Fos+/Drd1+ cells in animals treated for 10 d l-DOPA compared with those treated for 0 or 1 d with l-DOPA. C, QuPath-assisted quantification of confocal images depicting the Inhba, Fos, and Drd1 transcript coexpression in hemiparkinsonian mice treated for 0, 1, or 10 d l-DOPA. D, Time course of the experiment and graphical description of experimental groups utilized to establish the functional role of ALK4/TGFb1R signaling in LID development. E, Effects of ALK4 inhibition with SB431542 on summed ALO AIMs (top) and contralateral rotations (bottom) observed on Days 1, 3, 5, and 10 of cotreatment with l-DOPA. F, Confocal images (20×) and insets with arrows depicting pSMAD4 localization in NeuN+ and Iba1+ cells in the dorsal striatum of hemiparkinsonian mice treated with SB421542 (0 or 4.2 mg/kg) and l-DOPA for 10 d and QuPath-assisted quantification of confocal images depicting the pSMAD4, NeuN, and Iba1 coexpression in mice treated with SB421542 (0 or 4.2 mg/kg) and l-DOPA for 10 d.
ALK4R/TGFβR1 inhibition reduces development of LID
Inhba encodes for a preproprotein that is cleaved to form βA subunits of activin and inhibin protein complexes (Bloise et al., 2019). Dimerization of the βA subunit to another βA subunit results in the formation of the Activin A protein complex, while dimerization with an α subunit results in the formation of an inhibin B complex. The transcription of other subunits of inhibin was not significantly affected by l-DOPA in D1-MSNs; hence, we suspected that the formation of Activin A (βA/βA) protein complexes would most likely occur. Activin A has been shown to modulate cellular differentiation, immune function, and neuroplasticity (Ageta et al., 2010; Gancarz et al., 2015; Bloise et al., 2019). To explore the possibility of a functional role of the activin/TGFβ pathway in LID, as indicated by our snRNA-seq analysis and corroborated by FISH, we designed a LID development experiment in the presence and absence of the ALK4R/TGFβR1 antagonist, SB431542 (Fig. 5E; Inman et al., 2002). At doses previously utilized to influence hippocampal synaptic plasticity and did not interfere with the overall motor activity (Caraci et al., 2015), SB431542 dampened development of LID over the course of 10 d of treatment with l-DOPA compared with the vehicle (Fig. 5F; FDrug(1,19) = 13.43, p = 0.0016, FDay(3,57) = 18.75, p < 0.0001; FInteraction(3,57) = 1.038, p = 0.3826). The effect of drug reached statistical significance by Day 5 and continued to Day 10 (both p < 0.01), suggesting that reducing activin/TGFβ signaling can inhibit the stable expression of LID. Activin/TGFβ receptors typically signal via phosphorylation and nuclear translocalization of the receptor-regulated Smads, Smad2 and Smad3 (Massague, 1998). Although the precise Smad protein involved in LID remains unknown, Smad2 has been indicated in adult hippocampal plasticity and striatal development of MSNs (Maira et al., 2010; Gradari et al., 2021). As such, we used pSMAD2 nuclear localization to indicate the pharmacological efficacy of SB431542 to inhibit the activin/TGFβ pathway. In the context of LID, SB431542 reduced pSMAD2 nuclear localization in both neurons (Veh + LD: 425 ± 107.86 out of 1,249.7 ± 139.50 NeuN+ cells, SB + LD: 220.12 ± 79.91 out of 1,619.21 ± 200.41 NeuN+ nuclei; t10 = 2.527, p = 0.0300) and microglia (Veh + LD: 18.53 ± 5.52 out of 41.53 ± 7.94 Iba1+ cells, SB + LD: 16.57 ± 4.28 out of 125.19 ± 14.42 Iba1+ cells; t10 = 3.265, p = 0.0085; Fig. 5G), which have the highest expression of TGFβR1 in the brain (Extended Data Table 1-2).
Discussion
This comprehensive profile of the striatal transcriptional landscape upon progressive exposure to l-DOPA in DA-depleted animals identifies the subpopulations of D1-MSNs uniquely involved in LID development and maintenance. We corroborate prior investigations into striatal cellular heterogeneity (Gokce et al., 2016; Martin et al., 2019; Stanley et al., 2020), observing over 25 subpopulations, comprising both classes of MSNs, interneurons, progenitor cells, and glia (Fig. 1E,F). While DA cell loss induced minor transcriptional changes in our neurotoxin model, l-DOPA had wide-ranging effects on each cell population (Fig. 2). While there were limited effects on D2-MSNs following l-DOPA treatment, D1-MSNs were predominantly affected, forming new states of patch and matrix-residing D1-MSNs that expressed transcriptional profiles consistent with cellular activation (Fig. 3). The number of cells activated and the transcriptional behavior of both patch and matrix D1-MSNs were directly related to the l-DOPA experience (Fig. 4). In D1-MSNs, acute exposure to l-DOPA induced DEGs associated with neuronal plasticity and regulators of MAPK signaling, while the transcriptional behavior following chronic l-DOPA was related to synaptic stabilization and TGFβ signaling. Interestingly, we found a high degree of similarity between the transcriptional response of D1-MSNs to l-DOPA and that previously seen in hippocampal engram cells during the consolidation and retrieval of an associative memory (Figs. 3H, 4D; Marco et al., 2020). We then illustrated the translational applicability of these results, verifying that a subset of D1-MSNs shows enhancement of activin (Inhba) following chronic l-DOPA, and that pharmacological inhibition of activin/TGFβ signaling decreased dyskinesia severity (Fig. 5). These data illustrate the power of this comprehensive sequencing dataset for both hypothesis generation and validation in LID, and highlight a previously unappreciated role for activin/TGFβ in LID development.
The strength of this dataset lies in the comprehensive sequencing of the large number of cells sampled (>90,000), enabling the identification of glial and neuronal subpopulations previously challenging to distinguish (Fig. 1). Consistent with previous results (Heiman et al., 2014), DA loss had minimal effects on gene transcription, while l-DOPA exposure led to the emergence of three activation-associated D1-MSN subclusters, referred to here as Act Matrix-1, Act Matrix-2, and Act Patch (Fig. 3A,B). Following the l-DOPA exposure, there was also a minor decrease in the number of activated D2-MSNs, consistent with the inhibitory signaling downstream of DA D2 receptors (Fig. 3C). While acute l-DOPA led to the subclustering of D1-MSNs into all three activation subtypes, following repeated exposures, the number of cells in the Act Matrix-1 cluster abated, and Act Matrix-2 and Act Patch D1-MSNs remained relatively stable across 1, 5, and 10 d of l-DOPA. These data are suggestive of pruning of weakly stimulated neurons recruited in the initial circuit and a shift in the transcriptional behavior of the entrained neurons as the circuit matures: Act Matrix D1-MSNs-1 likely represents a population undergoing the first wave of activation-induced transcription, whereas Act Matrix D1-MSNs-2 may represent an accelerated transcriptional response, as the second wave of transcription is induced by first wave factors (Alberini, 2009). Transcriptionally, patch and matrix MSNs exhibited similar gene expression associated with activation, altered cell signaling, and long-term synaptic potentiation upon l-DOPA treatment. However, our results suggest that chronic l-DOPA treatment had an outsized effect on the activation state of patch D1-MSNs compared with matrix D1-MSNs, consistent with research suggesting that patch-localized MSNs play a more significant role in LID expression (Cenci et al., 1999; Saka et al., 1999; Henry et al., 2003; Crittenden et al., 2009; Mahmoudi et al., 2009).
Acute l-DOPA exposure induces a widespread cellular activation that, following repeated exposure, becomes focused on a select neuronal population (Engeln et al., 2016; Fieblinger et al., 2018; Girasole et al., 2018), processes akin to the early circuit refinement underlying learning and memory formation (Liu et al., 2012; Ramirez et al., 2013). Hippocampal-dependent memory relies upon unique ensembles of neurons known as “engrams,” which are recurrently activated during the initial formation and subsequent retrieval of a memory. The activation of engram cells is sufficient to drive conditioned behavior, even without a triggering stimulus (Josselyn and Tonegawa, 2020). Recent research has revealed that some D1-MSNs are recurrently activated during the initial and subsequent l-DOPA exposures. Inhibition and/or excitation of these cells can bidirectionally modulate the behavioral manifestation of LID consistent with putative “engram” cells (Engeln et al., 2016; Fieblinger et al., 2018; Girasole et al., 2018). Our data suggest that the two persistently activated D1-MSN subpopulations (Act Patch and Act Matrix-2) likely represent these “engram”-like cells, storing environmental exposures in a shared motor circuit dependent upon mechanisms akin to other models of synaptic plasticity, including learning and memory, drug addiction, and physiologic motor learning paradigms (Josselyn and Tonegawa, 2020). In support, we compared our data set to hippocampal nRNA-seq data from a mouse model, in which Arc+ cells are persistently tagged during entrainment of a hippocampal-dependent task, putative “engram” cells (Marco et al., 2020), and found that their transcriptional profile was enriched among the Act Patch and Act Matrix-2 D1-MSNs (Fig. 3H). These data are indicative of an engrammatic population encoding and retrieving the dyskinesia-inducing motor response to l-DOPA and strongly suggests the mechanisms of synaptic plasticity are shared mechanisms between LID and associative learning.
While many DEGs identified following acute l-DOPA exposure were also found after prolonged exposure (Fig. 4A,B), our analysis revealed several important transcripts uniquely expressed following chronic treatment (Fig. 4C). Initially, l-DOPA exposure led to the induction of several genes found to define the Act Matrix D1-MSNs-1 subpopulation, including multiple TFs necessary for synaptic plasticity, such as AP-1 and Egr family members, along with regulators of MAPK signaling (Fig. 3E). Conversely, repeated l-DOPA exposure predominantly induced two unique subpopulations, Act Matrix D1-MSNs-2 and Act Patch D1-MSNs (Fig. 3B), that were related to synaptic remodeling and transmission, such as Ntrk2, Elmo1, and Soat1 (Fig. 3E; Musumeci et al., 2009; Kim et al., 2011; Figge et al., 2016). Underlying these temporal differences in transcriptional behavior were distinct TF networks, with acute l-DOPA administration causing Creb1, Srf, Myc, and Foxo1 to induce the expression of IEGs, such as Fos, Jun, and Egr1. However, after repeated l-DOPA treatment, the transcriptional network became increasingly dependent upon the sustained activity of inducible factors: Mef2a/c, Fos, Junb, and Fosb. This transition away from calcium-dependent factors strongly resembles the patterns found in other models of long-term behavioral memory, where many of these same inducible factors play essential roles in establishing the necessary transcriptional network for multiple models of hippocampal and striatal plasticity (Cenci, 2002; Alberini and Kandel, 2014). In summary, these data indicate that early exposures to l-DOPA induce transcriptional activity to initiate synaptic remodeling that is dependent upon canonical neuronal signaling; however, as behavioral sensitization increases in concert with synaptic strength, the transcriptional network becomes increasingly dependent upon the expression of these previously temporarily inducible factors in a sustained transcriptional wave.
One of the DEGs specifically associated with chronic l-DOPA in D1-MSNs was Inhba (Figs. 4B,C, 5A), which encodes inhibin subunit βA, a member of the TGFβ superfamily that acts as a ligand for ALK receptors. In D1-MSNs, the Inhba expression was over 4× higher in mice repeatedly treated with l-DOPA compared with those with acute exposure, suggesting that it may be involved in behavioral sensitization to l-DOPA. Blocking ALK/TGFβ receptor signaling using SB431542 attenuated LID development, further indicating an essential role for activin in LID development (Fig. 5F,G). Although phosphorylation of Smad2 was utilized as a tool to indicate pharmacological efficacy, ALK/TGFβ signaling mechanisms are complex (Massague, 1998), and these findings do not preclude a role for Smad3 and/or other transduction pathways in the modulation of LID development, and further investigation of these mechanisms is necessary. While neuronal autocrine signaling is likely, the Inhba expression was also elevated in several glial populations, including oligodendrocytes and microglia (Extended Data Table 2-2). Activin has been implicated in microglial activation and oligodendrocyte maturation (Dillenburg et al., 2018; Morianos et al., 2019; Wang et al., 2022); however, little is known about the importance of glial or neuronal activin/TGF-β signaling in striatal plasticity, and future work identifying the transduction mechanisms and the role each cell type plays in this signaling pathway will be essential for therapeutic development.
Collectively, these data provide a comprehensive view of the transcriptional changes that underlie the aberrant striatal plasticity induced by l-DOPA that causes LID while also providing several promising avenues for future research. Due to the extensive transcriptional changes that correlate to LID development, in this study, we have focused on the disparate activation states of D1-MSNs induced by l-DOPA treatment, showing numerous parallels between the mechanisms necessary for associative learning and those underlying LID. These data strongly suggest that maladaptive synaptic plasticity directly contributes to progressive l-DOPA sensitivity and that the neurons activated during initial l-DOPA exposure are molecularly “primed” such that subsequent l-DOPA treatments unleash an imprinted motor program stored in these putative “engram cells,” causing dyskinetic behaviors. The targeting of these cells may hold promise for inhibiting LID development or extinguishing a previously formed cellular “memory” for l-DOPA, perhaps by targeting activin-mediated mechanisms. These data represent the first step toward establishing and delineating the unique roles of individual subsets of striatal cells in LID and start to define the differential gene modules necessary for the development of the aberrant circuit caused by dysregulated dopaminergic signaling.
Data Availability
The snRNA-seq data are available publicly from the NCBI's Gene Expression Omnibus (GEO) and are accessible at the following link https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE250364 and through the GEO accession number GSE250364, available by December 31, 2023.
Footnotes
We thank the University of Alabama at Birmingham (UAB) Flow Cytometry and Single-Cell Core Facility, particularly Dr. Shanrun Liu and Dr. Vidya Sagar Hanumanthu, for their assistance in protocol development, and Dr. Jeremy Day (UAB Comprehensive Neuroscience Center) and Dr. Ashley Harms (UAB Center for Neurodegeneration and Experimental Therapeutics), for their guidance in method development and interpretation. This work was supported by grants from the American Parkinson Disease Association and the Parkinson Association of Alabama to K.L.E.J., as well as the APDA Advanced Center for Parkinson Research at UAB.
D.G.S. has served as a consultant for or received honoraria from AbbVie, Curium Pharma, ApPello, Theravance, Sanofi-Aventis, Alnylam Pharmaceuticals, Coave Therapeutics, BlueRock Therapeutics, Biohaven, Eli Lilly, and F. Hoffmann-La Roche. All remaining authors declare no competing interests.
- Correspondence should be addressed to Karen L. Eskow Jaunarajs at karenej{at}uab.edu.
