Modulation of striatal cAMP levels: A key pathway in the treatment of hyperkinetic movement disorders

Summary

Hyperkinetic movement disorders, like dystonia, chorea, myoclonus, dyskinesia, and tremor, can be extremely disabling, impairing quality of life. Our translational approach in humans and mice investigates the link between cyclic AMP (cAMP) signaling pathway alterations in striatal neurons and hyperkinetic movement disorders. ADCY5 encodes adenylyl cyclase 5, key enzyme for striatal cAMP synthesis. Pathogenic variants result in mixed hyperkinetic movement disorders (MxMD-ADCY5). We prove caffeine therapeutic effect in a prospective trial in two patients with MxMD-ADCY5 and generated an Adcy5^R419W mouse model harboring the most frequent human pathogenic variant to understand underlying mechanisms. In patients, movements’ severity is increased in absence of caffeine. In mice, caffeine improves motor symptoms through adenosine A_2A receptor blockade. Mutation increases cAMP signaling in striatal projection neurons, an effect selectively corrected by A_2A receptor antagonism in indirect pathway neurons. Fine modulation of neuronal cAMP levels represents a key target to treat hyperkinetic disorders, especially dystonia/dyskinesia.

Graphical abstract

Highlights

• •Mutations of Adcy5 encoding the enzyme for striatal cAMP synthesis cause movement disorders
• •Translational approach in humans and mice shows that mutated AC5 increases cAMP signaling
• •A2A receptor antagonism specifically corrects the phenotype in humans and mice
• •Fine modulation of striatal cAMP level: a key target pathway to treat hyperkinetic disorders

Pharmacology; Molecular biology; Neuroscience; Movement Disorders

Introduction

Movement disorders can be classified into two types of movements: hyperkinetic and hypokinetic movements. Hyperkinetic movement disorders are characterized by the presence of abnormal involuntary movements, comprising most notably of dystonia, chorea, myoclonus, dyskinesia, and tremor. Their presence has significant implications for diagnosis and treatment, and they can be extremely disabling. Possible causes are numerous, including autoimmune disorders, infections of the central nervous system, metabolic disturbances, genetic diseases, drug-related causes, and functional disorders.^1,2 Dysregulation of the cAMP pathway and associated striatal dysfunction have been implicated in numerous hyperkinetic disorders, including many genetic causes such as Huntington disease and drug-induced causes such as levodopa-induced dyskinesia in Parkinson disease.^1,2,3 Thus, studying a disorder with a primary disturbance of the striatal enzyme producing cAMP would give a unique opportunity to unravel the common mechanisms underlying the different types of hyperkinetic movement disorders.

ADCY5-mixed movement disorder (MxMD-ADCY5) is a rare genetic disease that begins in childhood. MxMD-ADCY5 is characterized by a combination of hyperkinetic movements that can include chorea, myoclonus, dystonia, tremor, and paroxysmal dyskinesia.^4,5 One particular feature of the disease is the occurrence of paroxysmal dyskinesia during night-time,^6,7 which worsens at sleep/wake interphases.⁸ Although controversial, an associated psychiatric phenotype has been suggested, especially anxiety and/or depression.^4,9 The disease is caused by pathogenic variants in the ADCY5 gene, encoding the adenylyl cyclase 5 protein (AC5). MxMD-ADCY5 is autosomal dominant in most patients, but autosomal recessive cases have been occasionally reported.¹⁰ Biallelic pathogenic variant carriers tend to have more severe manifestations than monoallelic carriers.^11,12

The striatum plays a critical role in the control of movement, through balanced and coordinated activity of two populations of striatal projection neurons (SPNs), direct striatonigral (dSPNs) neurons and indirect striatopallidal neurons (iSPNs), which facilitate and inhibit movement execution, respectively.^13,14,15 In rodents, AC5 is enriched in both SPN populations, which represents 95% of striatal neurons.¹⁶ AC5 is the main enzyme producing cAMP in striatal neurons. AC5 can be either activated or inhibited by various G-protein-coupled receptors (GPCRs) upon binding of neurotransmitters. In dSPNs, AC5 is mostly activated downstream of dopamine D1 receptors (D1Rs) by the alpha subunit of the stimulatory olfactory type G protein (Gαolf) and inhibited downstream of adenosine A₁ receptors (A₁Rs) by the alpha subunit of the inhibitory Gi/Go protein (Gαi/o).^2,17,18,19 In iSPNs, AC5 is mostly activated downstream of adenosine A_2A receptors (A_2ARs) by Gαolf and inhibited downstream of dopamine D2 receptors (D2Rs) by Gαi/o. In neurons, downstream of AC5 activation, cAMP will bind to protein kinase A (PKA), a heterotetrametric serine-threonine kinase, which will become fully active to phosphorylate its targets,²⁰ among which are the dopamine- and cAMP-regulated neuronal phosphoprotein (DARPP-32) and the GluA1 subunit of glutamate AMPA receptor. Phosphorylation of ribosomal protein S6, a component of the small 40S ribosomal subunit implicated in mRNA decoding, also occurs via the activation of PKA and phosphorylation of DARPP-32.

Although there is no curative treatment for MxMD-ADCY5, oral administration of caffeine, a non-selective A_2AR antagonist, has been reported to improve patients with MxMD-ADCY5 through retrospective or anecdotal observations.^21,22,23 In a retrospective study on 30 patients, caffeine was well tolerated, and more than 80% of patients reported sizable improvements after caffeine administration.²² Anecdotal reports suggested that the A_2AR antagonists istradefylline and theophylline could also have positive effects on disease motor manifestations.^24,25,26 These observations suggest a potential effect of caffeine through its action on A_2ARs. In vitro studies found that five disease-causing pathogenic variants, including the p.R418W variant, were associated with an increased AC5 activity.^27,28 A_2AR antagonists might improve the phenotype through a reduction of AC5 activity. However, this hypothesis has not been proven yet in patients nor tested in animal models.

In the present study, we used a translational approach to characterize the pathophysiology and treatment of ADCY5-related disorder, for which little is known to date. To prove the putative effect of caffeine on MxMD-ADCY5, we conducted a first-in-class, proof-of-concept, prospective double-blind randomized crossover trial in two patients. To understand the underlying mechanisms of the therapeutic effects, we then generated a novel mouse model of MxMD-ADCY5, carrying the Adcy5^R419W mutation, reproducing the most frequent human pathogenic variant, p.R418W. We characterized this mouse model with a special focus on motor phenotype and the activity of the mutant AC5. We then characterized the impact of the modulation of the striatal cAMP signaling pathway, including the impact of A_2AR modulation in SPNs.

Results

Caffeine improves symptoms in MxMD-ADCY5 patients

The effect of caffeine was prospectively assessed in two patients in a double-blind randomized monocentric crossover pilot clinical trial, comparing the effects of caffeinated versus decaffeinated coffee on movement disorders with multimodal assessments (video, clinical scales, and motion sensors) (Figures 1A and 1B). At baseline, patients with MxMD-ADCY5 showed high clinical scores on the Unified Dyskinesia Rating Scale (UDysRS), cumulative objective UDysRS, Clinical Global Impression of Severity (CGI-S), and Patient Global Impression of Severity (PGI-S), reflecting intense abnormal hyperkinetic movements (Table S1; Video S1). Volumes of motion capture sensors displacements in four tasks (sitting with elevated arms sideward [1], standing still eyes opened [2], sitting in rest [3], turn head to right then to left while sitting [4]) strongly correlated with the clinical cumulative objective UDysRS and Abnormal Involuntary Movement Scale (AIMS) scores (Table S2).

Figure 1.

Caffeine reduces movements of patients with MxMD-ADCY5

(A and B) Excerpt from 3D view of motion sensor recording (A) and video recording (B) of Patient 01-01 during the task 1 “sitting with elevated arms sideward.”

(C) z axis displacement of body parts following either decaffeinated coffee or caffeinated coffee intake (same patient and task).

(D–G) Volume of sensors displacement of the full body for the two patients, following either decaffeinated coffee or caffeinated coffee intake, during task 1 (D), task 2 “Standing still eyes opened” (E), task 3 “Sitting in rest” (F), and task 4 “Turn head to right then to left while sitting” (G). Statistics: (C–G) Data were centered and normalized. (D–G) Data were expressed as ratios relative to the corresponding baseline (Visit V0) for each task. Results are presented as mean ± SEM. Linear mixed model: ∗∗p < 0.01, ∗∗∗p < 0.001 for placebo vs. caffeine treatment (see Table S5 for detailed statistics).

Gray curves and symbols are for the placebo (decaffeinated coffee) condition and orange curves and symbols for the active treatment (caffeinated coffee) condition.

Video S1. Video of clinical trial Patient 01-01 during task 1 “sitting with elevated arms sideward”, related to Figures 1 and S1

Video recording (left panel) and 3D view of Motion Sensor recording (right panel) of Patient 01-01 during the Task 1 “sitting with elevated arms sideward” on two different days following either placebo (decaffeinated coffee) condition (part 1 of the video) or the active treatment (caffeinated coffee) condition (part 2 of the video).

Treatment by caffeinated coffee significantly decreased the clinical scores (cumulative objective UDysRS and PGI-S), with a trend to also decrease AIMS, CGI-S, and CGI and PGI of Change (CGI-C and PGI-C) scores as compared to decaffeinated coffee (Table S1). Caffeine significantly reduced the volumes of sensors displacements at the full body scale by 43%–91%, depending on the task assessed (Figures 1C–1G; Video S1). Motion capture also refined caffeine efficacy analysis by an objective quantification of displacement of individual body segments, with a decrease of 34%–96%, depending on the body part assessed (Figures S1). Both volumes of motion capture sensors displacements and clinical scales scores were inversely correlated with caffeine plasma concentration levels, as assessed by pharmacokinetic parameters, but reaching significance only in task 4, due to the small sample size (Table S2). Based on the minimal correlation observed, we could estimate that treating 12 patients could allow to reach significance for most of the tasks. Effects were observed in caffeinated coffee but not decaffeinated coffee. These results, in controlled trial conditions, objectively show for the first time the efficacy of caffeine in MxMD-ADCY5 patients that had been previously reported in uncontrolled observations.^22,23 These effects strongly suggest a role of caffeine in improving the patients’ symptomatology, although we cannot rule out the potential contribution of other chemical differences between the two types of brews. Other results support the role of caffeine in patients’ improvement. This compound is a non-selective but strong A_2AR antagonist, and positive preliminary results have been obtained by treating patients with caffeine tablet, not only coffee beverage.²² Other reports indicate effects of other antagonists of these A_2ARs receptors, such as istradefylline and theophylline.^24,25,26

Adcy5 mutation in mice alters motor behavior and learning

We first investigated motor function in the p.R419W mice, using the accelerating rotarod, a widely used test of motor abilities and learning.^29,30 Adcy5^+/+ (wild-type) mice showed a normal learning curve, their latency to fall gradually increasing between day 1 and day 7 (Figure 2A). By contrast, both Adcy5^R419W/+ and Adcy5^R419W/R419W (mutant) mice performed very poorly over 7 days. Their latency to fall did not change (Figure 2A). On day 8, mutant mice performed very poorly in the fixed speed rotarod as compared to Adcy5^+/+ mice, even at the lowest speed (5 rpm), despite previous training (Figure 2B). Altogether, these results suggest that both Adcy5^R419W/+ and Adcy5^R419W/R419W mice have a combination of motor disorders and motor learning deficit.

Figure 2.

Motor and behavioral activities are impaired in Adcy5 mutant mice

(A) Accelerating rotarod performances are impaired in Adcy5 mutant mice. Performances of daily trials over 7 days on the accelerating rotarod only improved in Adcy5^+/+ mice. Average performances on the first and last days of accelerating rotarod show that latency to fall only increases in Adcy5^+/+ mice. (B) Performances on the rotarod at different fixed speeds. On day 8, performances of Adcy5^+/+ mice were higher than mutant mice at all speeds. (C) Activity- and ingestion-like behaviors recorded over 72 h in HomeCageScan. Shaded gray areas represent the 12 h dark period (night). Adcy5^R419W/+ and Adcy5^R419W/R419W mice showed decreased activity-like behaviors during night phases (active phases for mice) but spent more time on ingestion-like behaviors than Adcy5^+/+ mice. (D) Grip strength of forelimbs may be decreased in mutant mice. Statistics: data are presented as mean ± SEM. Three-way (A), two-way (A–C), and one-way ANOVAs (D) with repeated measures and Tukey’s post hoc test. $ represents genotype, § day or speed effects, and ∗∗∗ represents night effect. ∗∗∗p < 0.0001 for latencies of Adcy5^+/+ mice at D1 vs. D7 and ∗p < 0.05 forelimb grip strength between Adcy5^R419W/+ and Adcy5^+/+ mice (see Table S5 for detailed statistics). Adcy5^R419W/R419W are in light blue (n = 8–11), Adcy5^R419W/+ in orange (n = 8–11), and Adcy5^+/+ littermates in black (n = 5–11) (see Table S5 for detailed statistics).

Spontaneous behaviors of Adcy5 mutant mice are altered during active phases of the circadian cycle

We then assessed the spontaneous behaviors in an ecological situation, using home cage recording. The behaviors were recorded for three consecutive days and classified into four categories: activity-, ingestion-, exploration-, and resting-like behaviors as previously described³¹ (Figures 2C and S2A). Activity-like behaviors, in particular the Hanging behavior, of both Adcy5^R419W/+ and Adcy5^R419W/R419W mice decreased during night phases compared to Adcy5^+/+ (Figures 2C and S2A). Forelimb grip strength was slightly decreased in mutant mice compared to Adcy5^+/+ (Figures 2D and S2C). Mutant mice spent more time on ingestion-like behaviors than Adcy5^+/+ mice, but their food and water intakes were unchanged (Figures 2C and S2B). No significant genotype-related difference was observed in exploration- or resting-like behaviors over the full circadian period (Figure S2A). Because “hanging” and “eating” both require adequate motor skills for animals to stand up or grab, the results of home cage recordings provide further evidence of motor deficits in Adcy5-mutant mice.

An anxiety-like phenotype is also associated to Adcy5 mutation

The strong defects seen in the motor performance of mutant mice could be due to several contributing parameters, and clinical observations suggest that patients with MxMD-ADCY5 might have a psychiatric phenotype in addition to motor manifestations.^4,9 Therefore, we investigated anxiety-like behaviors in mutant mice with the open field and elevated plus maze.³² In the open field, mutant mice displayed thigmotaxis, spending less time in the center as compared to Adcy5^+/+ mice (Figures S3A–S3C). In the elevated plus maze, Adcy5^R419W/R419W mice entries into the center were decreased compared to Adcy5^+/+ mice (Figures S3D–S3F). We then investigated depressive-like behaviors with the tail suspension and forced swim tests,^33,34 which showed no increased immobility of mutant mice compared to Adcy5^+/+ (Figures S3G and S3H). Hence, the mutant mice may have an anxiety-like phenotype, avoiding central areas in the open field and elevated plus maze, that could participate in the motor defects observed in the previous experiments. In contrast, we found no depression-like phenotype in mutant mice.

Caffeine and the selective A_2AR antagonist istradefylline increase locomotor activity in Adcy5 mutant mice through A_2AR

So far, the most promising treatments of MxMD-ADCY5 are A_2AR antagonists, particularly caffeine,^21,22,23 and possibly istradefylline²⁴ and theophylline.^25,26 We thus decided to investigate the effects of caffeine and istradefylline on the locomotor activity of mutant mice (Figures 3A and 3B). In the absence of drug administration, no difference in locomotor activity was observed among genotypes. Caffeine (15 mg/kg) induced a prolonged increase in locomotor activity of Adcy5^R419W/R419W or Adcy5^R419W/+ mice compared to Adcy5^+/+ (Figure 3A). Istradefylline (2.5 mg/kg) made an even stronger effect in Adcy5^R419W/R419W mice (Figure 3B). These results suggested that caffeine effects were linked to its action on A_2AR.

Figure 3.

Caffeine increases locomotor activity and reverses motor impairment of Adcy5 mutant mice, through inhibition of A_2AR

(A–D) Spontaneous horizontal locomotor activity under the administration of vehicle or different treatments: caffeine 15 mg/kg (A), A_2AR antagonist istradefylline 2.5 mg/kg (B), caffeine 15 mg/kg then A_2AR agonist CGS21680 10 mg/kg sequential administration (C), or caffeine 15 mg/kg + CGS21680 10 mg/kg simultaneous administration (D). Caffeine and the A_2AR antagonist istradefylline increase locomotor activity in mutant mice. The suppression of caffeine-induced hyperlocomotion by the A₂AR agonist CGS21680 shows that the stimulatory effect of caffeine results from A_2AR inhibition.

(E and F) Timeline (E) and experimental setting of notched beam test (F). Open field was performed between the 3^rd (t₉₀) and 4^th (t₁₈₀) notch beam trials (see Figures S3A and S3B).

(G and H) Success rate (G) and time to cross performances (H) on the notched beam under vehicle or treatment conditions. Motor deficits of mutant mice in the notched beam test are reversed by caffeine or istradefylline, and this improvement is reversed by A_2AR stimulation. Statistics: data are presented as mean ± SEM except for (G) as percentage. Three-way (A–D) and two-way ANOVA (H) with repeated measures, followed by Tukey’s post hoc test. (G) Generalized linear mixed model. (A–D) ∗∗p < 0.01, ∗∗∗p < 0.001 for vehicle vs. drug treatment in D1GFP-Adcy5^R419W/R419W or Adcy5^R419W/R419W mice at each time point, #p < 0.05, ##p < 0.01, and ###p < 0.001 for vehicle vs. drug treatment in D1GFP-Adcy5^R419W/+ or Adcy5^R419W/+ mice at each time point.

(H) $ and § represent genotype and time effects, respectively. For clarity, only genotype post hoc comparisons are indicated when both genotype and time effects are found. ∗∗p < 0.01 and ∗∗∗p < 0.001 for Adcy5^+/+ vs. Adcy5^R419W/R419W comparison. ##p < 0.01 for Adcy5^+/+ vs. Adcy5^R419W/+ comparison, †††p < 0.001 for t-30 (before treatment) vs. other time points comparison (see Table S5 for detailed statistics). D1GFP-Adcy5^R419W/R419W and Adcy5^R419W/R419W (n = 5–9) groups are in light blue, D1GFP-Adcy5^R419W/+ and Adcy5^R419W/+ (n = 5–8) in orange, and D1GFP-Adcy5^+/+ and Adcy5^+/+ (n = 5–9) in black. The D1GFP-Adcy5 double mutant mouse strain was only used in (B) for subsequent immunohistochemistry experiment shown in Figures 6, S5, and S6. Filled symbols for vehicle groups and open for pharmacological treatment. Black arrows indicate timing of treatment administration. Shaded light blue and light gray areas represent the time periods after first or second drug administration if applicable, respectively.

To confirm that caffeine effects on locomotor activity were related to A_2AR blockade, mice were injected, either sequentially or simultaneously, with caffeine (15 mg/kg) and the A_2AR agonist, CGS21680 (10 mg/kg). The locomotor activity, increased by caffeine in Adcy5^R419W/+ and Adcy5^R419W/R419W mice, started to decrease 5 min after CGS21680 sequential administration and returned to baseline levels within 20 min (Figure 3C). Simultaneous administration of CGS21680 and caffeine prevented caffeine-induced increase in locomotor activity in both homozygous and heterozygous mutant mice (Figure 3D).

Altogether, these experiments in mice show that caffeine increases locomotor activity in Adcy5 mutant mice through its antagonistic action on A_2AR, which is mimicked by a selective A_2AR antagonist and prevented by a selective A_2AR agonist.

Caffeine and istradefylline reverse deficits in motor coordination and balance in Adcy5 mutant mice

Based on the effects of A_2AR antagonists on mutant mice locomotor activity, we then tested the effects of caffeine and istradefylline on their motor skills. Because following drug injection mutant mice were too agitated to undergo the rotarod test, we assessed A_2AR antagonists’ effects on the notched beam test, also evaluating motor skills (Figures 3E and 3F). Vehicle, caffeine, istradefylline, or a mix of caffeine and CGS21680 were administrated at t₀ (Figure 3E). Vehicle-injected Adcy5^R419W/R419W mice showed a lower success rate and an increased time to cross the beam, compared to Adcy5^+/+, which were not improving over test repetition (Figures 3G and 3H). When administrated with either caffeine or istradefylline, all genotypes increased their success rates, with shorter time to cross (Figures 3G and 3H). Importantly, as early as 20 min after injection with these A_2AR antagonists, mutant mice reached the performance level of Adcy5^+/+ mice. We tested the effect of CGS21680 administered immediately after the third crossing (t₉₀) to mice that had been injected with caffeine at t₀ (Figure 3E). The beneficial effects of caffeine were counteracted by CGS21680 (Figures 3G and 3H). When both caffeine and CGS21680 were simultaneously injected, the initial effect of caffeine was not observed and tended to resume after 90 min, reflecting half-life differences between caffeine and CGS21680 (Figures 3G and 3H).

During these evaluations, we also performed an open-field test, which showed, after the co-administration of CGS21680 with caffeine, either simultaneously or sequentially, that all genotypes showed a dramatic reduction of total traveled distance during the test duration (Figures S4A and S4B).

These experiments indicate that motor deficits of mutant mice in the notched beam test can be reversed by caffeine or istradefylline and that the improvement induced by caffeine is reversed by A_2AR stimulation.

Striatal cAMP signaling is altered in Adcy5 mutant mice

After demonstrating motor deficits in mutant mice, we investigated AC5 activity and the striatal molecular alterations associated with the Adcy5 p.R419W mutation (Figures 4A, 4B, 5A–5K, S5A, S5B, and S6A–S6H).

Figure 4.

Mutant AC5 has a reduced expression level but an increased specific activity in response to Golf-mediated stimulation

(A) AC5 protein levels, quantified by immunoblot in striatal homogenates, are decreased in Adcy5^R419W/R419W. GAPDH (glyceraldehyde-3-phosphate dehydrogenase) is used as housekeeping protein.

(B) cAMP production in the presence of various concentrations of D1R (SKF81297) or A_2AR (CGS21680) agonists or AC5 activator Forskolin in fresh striatal homogenates containing neuronal membranes. cAMP production was quantified by an HTRF cAMP assay, normalized to total protein levels, expressed as percentage of unstimulated basal cAMP production in Adcy5^+/+ striata, and normalized to relative AC5 protein levels in Adcy5^+/+ and Adcy5^R419W/R419W to estimate AC5 activity (%). cAMP production levels suggest that mutant AC5 activity was significantly increased in response to SKF81297 or CGS21680 (D1R or A_2AR agonists) as compared to wild-type, whereas response to forskolin (AC5 activator) remained similar for mutant and wild-type AC5s. Statistics: data are presented as mean ± SEM. Student test (A) and three-parameter logistic regression (B). ∗∗∗p < 0.001 (see Table S5 for detailed statistics). Fresh striatal homogenates from Adcy5^R419W/R419W are in light blue (n = 3–5) and from Adcy5^+/+ littermates in black (n = 2–4).

Figure 5.

The p.R419W mutation changes gene expression in the striatum and affects cAMP signaling actors both upstream and downstream of AC5

(A and B) Differential gene expression (FDR< 0.05, log2 fold-change >0.5), by RNA-seq analysis of striatal transcriptome of Adcy5^R419W/R419W vs. Adcy5^+/+ mice. Heatmaps of all differentially expressed genes (genes upregulated in red, downregulated in blue) (A). Expression of genes coding for proteins involved in AC5 or SPN signaling pathways (tpm: transcripts per million) (B). (Full listings available in Table S3).

(C–J) Protein levels, quantified by immunoblot in the striata of Adcy5^R419W/R419W, Adcy5^R419W/+, and Adcy5^+/+ mice. AC5 (C) and Gαolf (D) protein levels in the striatum and their correlation (E). A_2AR (F), D2R (G), D1R (H), DARPP-32 (I), and DAT (J) protein levels. AC5, Gαolf, A_2AR, and D2R decreased protein levels indicate down-regulation upstream of cAMP production by AC5 in the striatum of mutant mice. (K) Ratio of phosphorylated (pGluA1: Phospho-Serine-845-GluA1) over total GluA1 protein levels. The increased phosphorylation levels of GluA1 reflect up-regulation of effectors downstream of AC5 activation in mutant mice. Statistics: data are presented as mean ± SEM. (A and B) Comparison between genotypes carried out with DESeq2 (only genes with at least two normalized counts-per-million in two samples were included); adjusted p value: FDR<0.05 and log2FC cutoff>0.5; n = 3 animals per group. (C, D, F, G, I, and J) One-way ANOVA, followed by Tukey’s post hoc test when significant. (E) Linear regression (R² = 0.803, p < 0.001).

(H and K) Kruskal-Wallis test followed by Dunn’s post hoc test when significant. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001 for difference between genotypes (see Table S5 for detailed statistics). Adcy5^R419W/R419W are in light blue (n = 8–14), Adcy5^R419W/+ in orange (n = 10), and Adcy5^+/+ littermates in black (n = 9–12).

First, we measured the expression of AC5 in striatal extracts. We found decreased levels of AC5 in striatal extracts from Adcy5^R419W/R419W mice as compared to Adcy5^+/+ (Figures 4A and S5A). We then measured cAMP production in striatal extracts in response to D1R (SKF81297) or A_2AR (CGS21680) agonists or to the AC5 activator Forskolin (Figure S5B). To estimate AC5 activity in the striatum, we normalized cAMP production to total protein and AC5 expression levels (Figure 4B). Normalized cAMP production was significantly increased in response to D1R (SKF81297) or A_2AR (CGS21680) agonists in striatal homogenates from mutant mice as compared to wild type, whereas it remained similar after stimulation by the direct adenylyl cyclase activator, forskolin (Figure 4B). These results indicated an increased activity of the mutated AC5 in response to Gα stimulation (Figure 4B). This is in agreement with previous reports showing that in transfected cells, some ADCY5 mutations (including p.R418W) increase AC5 activity,^27,28 presumably through interference with the Gβγ binding site.³⁵ These findings demonstrate that p.R419W variant increases AC5 activity, whereas its expression levels were decreased in the striatum (Figures 4 and S5).

We then evaluated the consequences of the mutation on global gene expression using whole-transcriptome RNA sequencing (Figures 5A and 5B, S6G, and S6H; Tables S3). The expression of several hundred genes was modified between Adcy5^R419W/R419W and Adcy5^+/+ striatum (FDR<0.05, logFC>0.5, 283 upregulated, 312 downregulated) (Figures 5A and 5B; Table S3). Gene ontology analysis identified potassium transport and receptor regulation among the top regulated pathways and functions (Figures S6G and S6H; Table S4). We then focused on genes coding for AC5 or proteins regulating its activity or effects (Figure 5B; Table S3). Adcy5 expression was not significantly decreased, and the genes coding for the associated G protein subunits enriched in SPNs (Gnal, Gnab2, and Gnag7) were unchanged. We also examined the main proteins known to be regulating AC5 activity in striatal neurons. The protein levels of Gαolf, the Gα subunit associated to AC5 in SPNs, were decreased, proportionally to AC5 levels, in mutant mice as compared to Adcy5^+/+ mice (Figures 5C–5E). Therefore, AC5 and Gαolf (coded by Gnal) protein decreases (Figures 5C and 5D) resulted from post-transcriptional regulations. In contrast, the expression of Adora2a and Drd2, coding for upstream receptors A_2AR and D2R, respectively, was decreased (Figure 5B; Table S3). The expression of Rgs, regulators of G-protein signaling, was also modified, Rgs4 being decreased and Rgs10 and 14 being increased (Figure 5B; Table S3). The protein levels of A_2AR and D2R receptors were decreased, with the same trend for D1R (Figures 5F–5H). DARPP-32, a marker of SPNs, was also decreased (Figure 5I). Contrasting effects were observed in the mRNA regulation of other components of SPNs signaling pathways, including strong upregulation of Pdyn and Cartpt and coding for dynorphin- and cocaine-amphetamine-regulated transcript, respectively, two peptides known for their functional role in the striatum (Figures 5B; Table S3). An overall alteration of SPNs was ruled out by quantifying densities of striatal neurons and DARPP-32-positive neurons i.e., SPNs, which did not decrease significantly in mutant Adcy5^R419W/R419W and Adcy5^R419W/+ mice from Adcy5^+/+ mice at various ages (Figures S6A–S6F). The dopamine active transporter (DAT) protein levels were not changed, indicative of the absence of alteration of dopamine innervation (Figure 5J).

These transcriptome and protein changes showed that p.R419W mutation had a profound influence on striatal neurons with some changes at the transcriptional and/or post-transcriptional levels, possibly counterbalancing the increased activity of AC5.

Mutant-AC5-induced activation of cAMP signaling is selectively corrected in iSPNs by blockade of A_2AR

To evaluate the overall functional consequences on cAMP signaling in vivo, we investigated the phosphorylation pathways downstream of AC5. We first measured the phosphorylation of Ser845 in the GluA1 subunit of glutamate AMPA receptor, a well-characterized PKA substrate.³⁶ The phosphorylation level of this protein was increased in Adcy5^R419W/R419W with intermediate levels in Adcy5^R419W/+ striatal extract, as compared to wild type (Figure 5K). This result showed that despite a reduction in AC5 expression levels and alteration in multiple components of the pathway, the p.R419W mutation induced an activation of cAMP signaling in striatal neurons.

To disentangle the effects of the p.R419W mutation on striatal cAMP signaling in the two populations of SPNs, Adcy5 mice were crossed with Drd1a-EGFP mice to obtain D1GFP-Adcy5 double mutant mice. The phosphorylation of rpS6 at Ser235/236, a marker of striatal cAMP pathway activation,^37,38 was examined in the dorsolateral striatum (DLS) and dorsomedial striatum (DMS) with similar results in both territories of the striatum (Figures 6A–6E; S7A–S7E). In homozygous Adcy5^R419W/R419W mice, the proportion of pSer235/236-rpS6 positive SPNs was increased in both D1-GFP positive and D1-GFP negative striatal neurons, showing that signaling is activated in both iSPN and dSPN populations of projection neurons in the whole dorsal striatum (Figures 6D, 6E, S7D, and S7E). The acute injection of istradefylline reversed this increase specifically in D1-GFP-negative striatal neurons, probably the iSPNs (Figures 6B–6E and S7B–S7E).

Figure 6.

Increased cAMP signaling in mutant mice is reversed by A_2AR inhibition in the dorsolateral striatum

(A) Immunostaining of GFP in a coronal section from a D1GFP-Adcy5^+/+ mouse injected with vehicle. White square indicates the dorsolateral striatum (DLS) region of interest.

(B and C) Confocal coronal sections in the DLS showing double immunostaining for phosphorylated ribosomal protein S6 (p-rpS6 = Phospho-Serine-235/236-rpS6, gray), a marker of striatal cAMP pathway activation, and GFP (magenta) and DAPI staining (blue) from D1GFP-Adcy5^+/+, D1GFP-Adcy5^R419W/+, and D1GFP-Adcy5^R419W/R419W mice injected with either vehicle (B) or the selective A_2AR antagonist istradefylline 2.5 kg/mg (C). The specific expression of GFP under the D1R promoter was used as a marker of direct striatal projection neurons (SPNs).

(D and E) Quantification of the percentage of p-rpS6-positive SPNs among direct (GFP-negative) (D) and indirect (GFP-positive) (E) SPNs in the DLS. After an acute injection of istradefylline, the abnormal proportion of p-rpS6-positive cells among the SPNs populations in D1GFP-Adcy5^R419W/R419W mice only decreased in the indirect SPN but not in the direct SPN. It also decreased in D1GFP-Adcy5^R419W/+. Scale bars, (A) 200 μm; (B and C) 50 μm; inset in (B and C) 10 μm. Statistics: data are presented as mean ± SEM. Two-way ANOVA, followed by Tukey’s post hoc test. $ represents genotype effect, § represents treatment effect, ∗p < 0.05, ∗∗∗p < 0.001 (see Table S5 for detailed statistics). D1GFP-Adcy5^R419W/R419W (n = 3–4) in light blue, D1GFP-Adcy5^R419W/+ (n = 4–5) in orange, and D1GFP-Adcy5^+/+ (n = 5–6) wild type in black. Behavioral data for the same groups are reported in Figure 3B. Filled symbols for vehicle and open symbols for pharmacological treatment.

These results provide evidence that the behavioral alterations observed in Adcy5 mutant mice, improved by caffeine or istradefylline, result from increased cAMP signaling in iSPNs, which can be corrected by the blockade of A_2AR, supporting the role of this receptor as a therapeutic target in patients.

Discussion

Through a translational approach in humans and mice, our results help to understand the link between alterations of cAMP signaling pathways in striatal neurons and hyperkinetic movement disorders. In a randomized double-blind cross-over trial, we proved that the motor phenotype of two patients with MxMD-ADCY5 was significantly improved by caffeine and that the magnitude of abnormal movements was inversely correlated to caffeine blood levels. We then investigated the underlying mechanisms to explain this therapeutic effect. We generated a relevant mouse model of MxMD-ADCY5 with motor disorders, as observed in the human disease. In this model, an increase of striatal AC5 activity was associated with the Adcy5 p.R419W(mice)/ADCY5 p.R418W(human) disease-causing mutation. The motor phenotype of mutant mice was partially reversed with caffeine. We demonstrated that this positive effect occurred through the blockade of A_2AR, thus implicating the role of iSPNs in the altered motor function and its correction. Our translational experiments thus provide new insights into the pathogenesis and treatment of hyperkinetic movement disorders.

In patients, previous observations suggested an effect of caffeine and other A_2AR antagonists (istradefylline and theophylline) on MxMD-ADCY5.^{21,22,24,25,39} These findings raised questions about the exact intrinsic amplitude of this effect and the possible contribution of a placebo effect. Here, we provide a demonstration of the strong effect of oral administration of caffeine, a non-selective A_2AR antagonist. Of note, occasional reports suggested that a small proportion of patients with MxMD-ADCY5 do not respond or even worsen with caffeine, possibly due to the effect of specific pathogenic variants on AC5 activity.²² However, this first-in-class proof of concept trial was an important step to further investigate the mechanisms underpinning the effect of antagonizing A_2AR.

We generated a MxMD-ADCY5 mouse model with a human mutation altering striatal AC5 function, providing a reliable pathogenic and mechanistic validity. Mutant mice showed various motor deficits. In the brain, AC5 is largely expressed in the striatum, which has an important role in motor control and the early and late phases of motor learning.⁴⁰ Hence, variants in ADCY5 could induce both motor control and/or motor learning alterations. Mutant mice displayed abnormal performance in several tests of motor function, which did not normalize following training, indicating deficient performance and learning. The possible anxiety-like phenotype of these mice may also have partly influenced their poor performances on motor tests. Although we did not identify spontaneous hyperkinetic movements, the presence of a motor phenotype in different behavioral tests and its correlation to underlying molecular changes indicate a reliable and strong face validity of this model.⁴¹ Mouse models have been largely used to study various motor disorders, unravel their complex biochemical phenotypes, and evaluate the effect of therapeutic interventions.^42,43,44 We observed a dose-related pathogenic effect of the variant, which offers opportunities, such as testing the effect of an experimental drug on a large range of phenotype severity.

Mutant mice displayed evidence supporting the fact that striatal mutant AC5 is a hyperactivated enzyme compared to its wild-type form, in spite of many potential homeostatic adaptations. cAMP production normalized to AC5 levels was increased in response to Gα stimulation. It resulted, in vivo, in an increased phosphorylation of some target proteins involved in cAMP signaling, despite the decreased protein levels of several effectors such as AC5, the Gαolf-associated stimulatory G protein and A_2AR. In addition, among the hundreds of mRNAs modified in the striatum of mutant mice, several suggested a downregulation of cAMP pathways. These protein and transcriptomic effects, which may correspond to a homeostatic response were not sufficient to prevent the in vivo signaling alterations and behavioral impairments. The hyperactivable mutant AC5 enzyme could lead to differences in local signaling microdomains, dependent on the dendritic geometry, as previously shown by Li et al.,⁴⁵ that would be responsible for local peaks of cAMP levels, critical for the phenotypic readout. Our findings in an in vivo mouse model are consistent with previous in vitro studies in transfected cells, suggesting that an increased activity in response to Gαolf-/Gαs-mediated stimulation is associated with five disease-causing pathogenic variants including the p.R418W variant.^27,28 In addition, Gαi/o-mediated inhibition of mutant AC5 was altered in these heterologous systems,^27,28 as well as Gβγ activation.³⁵ The binding of Gαolf, Gβγ, or Gαi changes the conformation of AC5, thereby triggering its activation, conditional activation, or inhibition, respectively. The R418 residue, which we focused our current study on, is located in the coiled-coil domain of AC5, a domain critical for the positioning of the catalytic domains of ACs,⁴⁶ and in the vicinity of the AC5-Gβγ interface.³⁵ The p.R418W variant is associated with conformational changes of AC5 that appear to mimic the Gβγ-mediated release of its auto-inhibitory state.³⁵ Whether this mutation indirectly influences affinity for Gαolf/s and Gαi or their regulatory efficacy remains to be determined. These have important implications for experimental therapeutics, suggesting that treatment should target G-protein-mediated modulation of AC5 activity.

Caffeine decreased movement amplitude and duration in patients with MxMD-ADCY5, increased locomotor activity, and rescued the motor defect observed during notched beam experiments in mutant mice. In patients, caffeine blood levels were inversely correlated to the amount of abnormal movements. In mice, we showed that the effects of caffeine occurred through A_2AR blockade. In the mouse model, istradefylline, a selective A_2AR antagonist, had a similar effect as caffeine, whereas CGS21680, a selective A_2AR agonist, reversed or abolished caffeine effects when co-administered sequentially or simultaneously to caffeine, respectively. Our results show that although cAMP signaling was altered in the two populations of SPNs, its selective reversal by istradefylline in iSPNs was sufficient to improve the behavioral phenotype. Thus, correction of the basal cAMP signaling could restore an adequate balance between the direct/indirect striatal pathways and, in turn, improve motor control (Figure 7). Altogether, these results support the use of caffeine or other more specific and selective A_2AR antagonists to treat patients with MxMD-ADCY5.^{21,22,23,24,25,39,47}

Figure 7.

Overview of the molecular effects of MxMD-ADCY5 and proposed model for consequences on motor symptoms with and without A_2AR antagonist treatment

In the control situation, AC5 (faded dark blue) converts ATP to cAMP upon an equilibrated regulation by Gαolf and Gαi/o. cAMP triggers the activation of PKA, which phosphorylates downstream targets. The phosphorylation level of these downstream effectors controls an appropriate balance between the direct (light blue background) and indirect (light pink background) pathways, leading to normal motor performances. In MxMD-ADCY5, the mutated AC5 (bright blue) is associated with (i) an increased activity after Gαolf-mediated stimulation (thick black arrows) presumably through interfering with the Gβγ binding site; (ii) a negative feedback, which downregulates the protein levels of A_2AR, Gαolf, and AC5 itself; (iii) an increased phosphorylation of PKA targets; (iv) abnormal movements and/or poor motor performance due to the disruption of the balance between the direct and indirect pathways. Treatment with A_2AR antagonists (caffeine or istradefylline) rescues the molecular and motor phenotypes in MxMD-ADCY5. The excessive activity of the mutated AC5 after Gαolf-mediated stimulation is circumscribed in iSPNs (thinner black arrows in light pink background). cAMP production and phosphorylation level of PKA targets are decreased as well in iSPNs. Restoration of balance between the direct and indirect pathways improves the motor phenotype.

Abbreviations: AC5, adenylyl cyclase 5; D1R, dopamine receptor D1; D2R, dopamine receptor D2; A₁R, adenosine A₁ receptor; A_2AR, adenosine A_2A receptor; Gαi/o, G protein subunit alpha i/o; Gαolf, Golf protein subunit alpha; PKA, cAMP-dependent protein kinase A.

Conclusions

We found a behavioral and molecular phenotype in MxMD-ADCY5, partly reversed by a pharmacological approach in both rodents and patients. Our results provide an experimental support to the therapeutic use of A_2AR antagonists in patients with MxMD-ADCY5. The biological convergence of multiple dystonia genes implicated in common molecular pathways in SPNs suggests that intracellular signaling of these neurons, especially the cAMP signaling pathway, is critical for the pathogenesis of hyperkinetic movements.^2,48,49 Accordingly, dysregulation of the cAMP pathway and associated striatal dysfunction have been implicated in numerous genetic hyperkinetic disorders and levodopa-induced dyskinesia in Parkinson disease.^2,3 Thus, demonstrating the potent effect of pharmacological modulation of the cAMP pathway in MxMD-ADCY5 has important implications for future therapeutic research, in the broader field of hyperkinetic movement disorders, far beyond this particular disease. Fine modulation of striatal cAMP levels represents a key target in the treatment of hyperkinetic movement disorders, especially dystonia and dyskinesia.

Limitations of the study

Despite its translational extensive aspect and proof of concepts, our study shows some limitations that will require further studies. Though our clinical trial proves for the first time that an important clinical effect of A_2AR modulation is possible in patients with MxMD-ADCY5, it does not answer the question of the proportion of patients in which this treatment could be effective nor the mutations for which this treatment will have a significant impact. Our rodent model was prioritized on the main mutation in patients but did not allow to explore the effect of other mutations present in patients. Thus, generalizing these results to all patients with MxMD-ADCY5 could be refined by further functional studies on other mutations. Although we prove the effect of A_2AR antagonism, other antagonists could be further tested in patients to find the most appropriate, in terms of benefit-risk ratio, and the most efficient treatment for patients with MxMD-ADCY5 and other similar hyperkinetic movement disorders.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Louise-Laure Mariani (louise-laure.mariani@icm-institute.org).

Materials availability

This study did not generate new unique reagents except the mouse line. The mouse line generated in this study has been deposited on the European Mouse Mutant Archive (EMMA; https://www.infrafrontier.eu/emma/) mouse repository by the Mouse Clinical Institute/Institut Clinique de la Souris (MCI/ICS, Illkirch) part of the French National Infrastructure PHENOMIN. The mouse line generated in this study will be distributed via the INFRAFRONTIER consortium (the European Research Infrastructure for phenotyping and archiving of model mammalian genomes; https://www.infrafrontier.eu/).

Data and code availability

Any additional information required to reanalyze the data reported in this paper is available from lead contact upon reasonable request.

• •The RNAseq data are available on the Genomic database GEO (Gene Expression Omnibus) with the GEO database access number GSE298638. Original western blot images have been deposited at Mendeley Data: https://data.mendeley.com/datasets/mrh3r5kh5m/1 (https://doi.org/10.17632/mrh3r5kh5m.1) and are publicly available as of the date of publication. Microscopy data reported in this paper will be shared by the lead contact upon request. Local law prohibits depositing raw datasets derived from human samples outside of the country of origin.
• •Code: this paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

Acknowledgments

The authors are very grateful to the patients and their families for their participation in the research and to the patient association ADCY5.org. Sponsor of the clinical trial was « Assistance Publique - Hôpitaux de Paris ». The present work was funded by the French Foundation for rare diseases (Fondation maladies rares; grant MOM-RD_20170405), the Paris Brain Institute (grants Neurocatalyst 2020 and Big Brain Theory 2021), the patient association ADCY5.org, Amadys, the Dystonia Medical Research Foundation DMRF grant 2020, Merz Pharma and Aguettant, Enjoy Sharing, and Agence Nationale de la Recherche (Project ANR-16-CE37-003).

The authors thank all members of the lab for their support. Generation of the ADCY5 mouse model was carried out by the Mouse Clinical Institute/Institut Clinique de la Souris (MCI/ICS, Illkirch), part of the French National Infrastructure PHENOMIN. Part of this work was carried out by the Paris Brain Institute Data Analysis Core (DAC) platform (https://dac.institutducerveau-icm.org, RRID: SCR_026138). This work benefited from equipment and services from the iGenSeq core facility (Genotyping and Sequencing), the ICM.QUANT core facility (quantitative Cellular and Molecular Imaging platform), and the CENIR PANAM core facility (Physiology and ANAlysis Of Movement) at PBI (Paris Brain Institute). Animal work was conducted at the PBI Pheno-ICMice Core Facility. The Core is supported by “Investissements d’avenir” (ANR-10- IAIHU-06 and ANR-11-INBS-0011-NeurATRIS) and the “Fondation pour la Recherche Médicale”. We gratefully acknowledge Nadege Sarrazin for assistance with animal behavior facility use.

Author contributions

Conceptualization, R.Y., E.R., and L.-L.M. Animal behavior, biochemical and imaging analysis, and investigation, R.Y., E.R., M. Doulazmi, C.D., S.L., M. Delmas, B.K., J.M., Y.N., A.P., M.V., J.-A.G., D.H., and L.-L.M. Clinical trial evaluations, motion capture, and pharmacokinetics investigations and analysis, M. Doulazmi, C.T., A.M., M.R., N.Z., and L.-L.M. Writing, R.Y., E.R., M. Doulazmi, J.-A.G., C.D., and L.-L.M. All authors contributed to manuscript editing and support the conclusions.

Declaration of interests

R.Y. received a PhD grant from the ADCY5.org patient association and the Fondation pour la Recherche Médicale during the current work, has been employed by LEO Pharma, and is currently employed by Quinten Health outside of the submitted work.

E.R. declares no competing interests related to the present work. A patent application related to the present work (Publication No. US20220096486, Application No. 17/441,405 with a claimed priority of EP3714888) was previously filed by AP-HP. However, the application is no longer active due to anteriority issues. He received honorarium for speech from Orkyn, Aguettant, Elivie, Merz-Pharma, Janssen, Teva, and Everpharma and for participating in advisory boards from Merz-Pharma, Elivie, Teva, and BIAL. He received research support from Merz Pharma, Orkyn, Elivie, EVER Pharma, Amadys, and Aguettant.

C.T. received a PhD grant from the Fondation pour la Recherche Médicale outside of the submitted work.

A.M.: a patent application related to the present work (Publication No. US20220096486, Application No. 17/441,405 with a claimed priority of EP3714888) was previously filed by AP-HP. However, the application is no longer active due to anteriority issues. She received honoraria for speeches from Merz-Pharma and travel funding from Elivie and Merz-Pharma.

L.-L.M.: a patent application related to the present work (Publication No. US20220096486, Application No. 17/441,405 with a claimed priority of EP3714888) was previously filed by AP-HP. However, the application is no longer active due to anteriority issues.

Disclosures of financial interests unrelated to the present work: received research support grants from the CNES, INSERM, JNLF, The L’Oreal Foundation, the French Parkinson Association, Fondation of France, Paris Brain Institute, and Axa foundation; received honorarium for scientific advice or lectures for Digital Medical Hub, Elivie, Sanofi-Genzyme, Accure therapeutics, Owkin, Emeis, and Paris Brain Institute; received travel funding from the Movement Disorders Society, ANAINF, Merck, Merz Pharma, Medtronic, Teva, and AbbVie, and is a co-inventor of a patent outside of the submitted work.

Declaration of generative AI and AI-assisted technologies in the writing process

No generative AI or AI-assisted technology was used for this study.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti-AC5 1:500 (WB)	Sigma-Aldrich	Cat# MABS2049
Rabbit monoclonal anti-pSer235/236-rpS6 1:1000 (IF)	Cell Signaling Technology	Cat# 4858
Chicken polyclonal anti-GFP 1:1000 (IF)	Thermo Fisher Scientific	Cat# A10262
Rabbit monoclonal anti-DARPP32 1:1000 (IF, WB)	Cell Signaling Technology	Cat# 2306
Rabbit polyclonal anti-D2R 1:500 (WB)	Frontier Institute	Cat# AF960
Goat polyclonal anti-A2AR 1:500 (WB)	Frontier Institute	Cat# AF700
Mouse monoclonal anti-D1R 1:500 (WB)	Millipore	Cat# MAB5290
Rabbit polyclonal anti-Gαolf 1:1000 (WB)	Corvol et al.18 Hervé et al.19	N/A
Mouse monoclonal anti-GluA1 1:1000 (WB)	Santa Cruz Biotechnology	Cat# Sc-13152
Rabbit polyclonal anti-pSer845-GluA1 1:1000 (WB)	ProSciψ	Cat# XPS-1013
Rabbit polyclonal anti-DAT 1:1000 (WB)	Covalab	Cat# pab0375
Mouse monoclonal anti-β-Actin 1:20000 (WB)	Abcam	Cat# ab6276
Mouse monoclonal anti-α-Tubulin 1:20000 (WB)	Sigma-Aldrich	Cat# T9026
Rabbit monoclonal anti-GAPDH	Cell Signaling Technology	Cat# 2118
Goat Alexa Fluor 488-coupled anti-chicken 1:500 (IF)	Invitrogen	Cat# A11039
Goat Cy3-coupled anti-rabbit 1:500 (IF)	Jackson ImmunoResearch Laboratories	Cat# 111-165-144
Anti-rabbit IgG DyLightTM 800 or 680 conjugated antibodies 1:5000 (WB)	Rockland Immunochemicals, Pottstown, PA, USA	800: Cat# 611-145-002 680: Cat# 611-744-002
Anti-mouse IgG DyLightTM 800 or 680 conjugated antibodies 1:5000 (WB)	Rockland Immunochemicals, Pottstown, PA, USA	800: Cat# 610-745-002 680: Cat# 610-744-002
Anti-chicken IgG DyLightTM 800 or 680 conjugated antibodies 1:5000 (WB)	Rockland Immunochemicals, Pottstown, PA, USA	800: Cat# 603-145-002 680: Cat# 603-144-002
Anti-goat IgG DyLightTM 800 or 680 conjugated antibodies 1:5000 (WB)	Rockland Immunochemicals, Pottstown, PA, USA	800: Cat# 605-745-002 680: Cat# 605-744-002
Chemicals, peptides, and recombinant proteins
Forskolin	Tocris	1099/10
CGS21680	Tocris	1063/10
SKF81297	Tocris	1447/10
Caffeine	Tocris	2793/100
Istradefylline	Tocris	5147/10
FASTDIGEST HINDIII	Thermo Scientific	FD0504
Critical commercial assays
HTRF cAMP Gs dynamic kit	Cisbio/Revvity	62AM4PEB
Qiagen RNeasy Plus Micro Kit	Qiagen	74034
Bicinchoninic acid (BCA) assay	Pierce™ BCA Protein Assay Kit	N/A
Deposited data
RNAseq	Genomic database GEO (Gene Expression Omnibus)	GEO access number GSE298638
Western blot	Mendeley Data at https://data.mendeley.com/datasets/mrh3r5kh5m/1	doi: https://doi.org/10.17632/mrh3r5kh5m.1
Experimental models: Organisms/strains
Mouse: Adcy5R419W/+, Adcy5R419W/R419W; CRISPR/Cas9-mediated insertion of R419W mutation in Adcy5 in C57BL/6N background	MCI/ICS, Illkirch. Phenomin, This paper	Kus6795-12-PM
Mouse: Drd1a-EGFP mice carrying drd1a-EGFP bacterial artificial chromosome (BAC) transgene in C57BL/6N background	GENSAT (Gene Expression Nervous System Atlas), Rockefeller University.	Gong et al.50
Oligonucleotides
Primers for Adcy5 genotyping: F: CAGGCCCGCTGTGTTCCTGAGTGTC R: GGACTGGGCAGAGGGATGAGCATGG	This paper	N/A
Software and algorithms
RNAseq	FastQC	Andrews.51 Babraham Bioinformatics
RNAseq	Star v2.5.3a	Dobin et al.52
RNAseq	rsem 1.2.28	Li and Dewey53
RNAseq	edgeR (R 4.2.1)	R Foundation for Statistical Computing, Vienna, Austria
RNAseq	DESeq2 (R 4.2.1)	R Foundation for Statistical Computing, Vienna, Austria Love et al.54
RNAseq	clusterProfiler R package	R Foundation for Statistical Computing, Vienna, Austria Yu et al.55
GraphPad Prism v 6-10		Dotmatics
HomeCageScan	HomeCageScan 3.0	Cleversys Inc
Grip strength	Grip Strength Meter	Bioseb, BIO-GS3
Elevated plus maze	TopScan	Cleversys Inc
Open-Field	TopScan	Cleversys Inc
Tail suspension	TST-5 software	Bioseb
Spontaneous locomotion activity	Cyclotron	Imetronic
Image analysis	Fiji	N/A
Image acquisition	SP8 confocal microscope	Leica
WB image acquisition	Odyssey Clx	Licor
WB quantification	Image Studio Lit	Licor
cAMP Assay signal detection	HTRF program	Mithras LB 940
Data processing, summarization, visualization, and analysis	Python version 3.12.0 R software version 4.3.2 Graph pad prism 10	R Foundation for Statistical Computing, Vienna, Austria
Motion capture data	Nexus version 2.16	Vicon Motion Capture System
Pharmacokinetic parameters	validated Phoenix® WinNonlin® software version 5	N/A

Experimental model and study participant details

Clinical trial of caffeine effects in ADCY5 patients

Two male patients with MxMD-ADCY5 (variants c.1252C>T^22,56 and c.2088+1G>A^22,23 respectively) participated in the SENSeo-ADCY5 study (NCT05136495).⁵⁷ This prospective double-blind randomized monocentric crossover pilot clinical trial was conducted between July and October 2022, at the Neurosciences Clinical Investigation Center in the Pitié-Salpêtrière University Hospital (Paris, France) to compare caffeinated versus decaffeinated coffee effects on movement disorders with multimodal assessments (video, clinical scales, motion sensors). Both patients signed informed consent to participate in the study. This trial was conducted in accordance with the provisions of the Declaration of Helsinki and Good Clinical Practice guidelines. The trial protocol and all appending documents were approved by the French ethics committee CPP IDF VII.

Inclusion criteria were: ADCY5 pathogenic variant carriers; age above 15 years old and 3 months; informed consent from the patient or/and a legal representative when appropriate; affiliated with a social security system or beneficiary of such a regime, daily caffeine consumer. Exclusion criteria were: Hypersensitivity to caffeine or to xanthine derivatives; Heart condition contraindicating coffee intake; Liver failure; Impaired comprehension interfering with an informed consent; Positive pregnancy test for women of childbearing potential; Patient treated either by Enoxacin, Ciprofloxacin, Norfloxacin, Cimetidine or Phenytoin at inclusion.

The study consisted of 3 visits: V0, V1 and V2 on three consecutive days (48 hours between V0 and V2). Patients were asked not to ingest any coffee or caffeine-containing products for at least 24 hours before the inclusion visit (V0). They did not have any breakfast on the morning of each visit, and fasted during the visits’ assessments so that the absorption and other pharmacokinetic parameters of caffeine and its metabolites could be as standardized as possible. At each visit with coffee ingestion (V0 and V1), the UDysRS, PGI and CGI scales were assessed at baseline then, the Abnormal Involuntary Movement Scale AIMS, cumulative objective UDysRS and the Global impression scales (CGI-S, CGI-C, PGI-S and PGI-C) were assessed concomitantly to the duration of motion capture using the VICON™ system that started 2h after coffee ingestion. The dose of caffeine intake for each patient was roughly adapted to correspond to their usual daily morning coffee intake (Table S1). The beverages ingested at visits V0 and V1 were sampled to confirm their exact caffeine concentration and total caffeine intake at each visit. Blood samplings used for pharmacokinetics analysis were performed at V0 and V1 visits before coffee intake, and at 30 min, 1h±10 min, 2h±10 min, 4h±10 min, 8h±10 min.

Neither the patient nor the clinicians knew if a caffeinated coffee or a decaffeinated coffee was being served. A balanced randomization list was created by a person independent from the study according to the constraints defined by the statistician of the study. The randomization stratified patients into two arms following a 1:1 ratio. Patients randomized in one group would start with coffee with caffeine on V0 then switched to decaffeinated coffee on V1 and patients of the other group would do the opposite.

The main objective was to assess the use of motion capture to objectively quantify movements disorders observed in patients with ADCY5-mutations. The main secondary objective was to assess the effects of caffeinated coffee on abnormal movements. The other secondary objectives were to determine the individual pharmacokinetics of caffeine and its metabolites in each patient during the recording of abnormal movements by motion capture; To evaluate the correlation between displacements of motion sensors and blood levels of caffeine and its metabolites; To assess the global impression of patients and clinicians on the effect of caffeinated coffee on abnormal movements.

To assess the clinical effect of caffeine, changes in clinical scales scores (Cumulative objective UDysRS and AIMS, PGI and CGI) and of volumes of motion capture sensors displacements after caffeinated coffee vs decaffeinated coffee were assessed. The cumulative objective UDysRS (range=0 to 128 (worse)) was derived from the UDysRS as the sum of all objective impairment UDysRS subscores of each body parts and the disability subscores, each subscore ranging from 0 (no dyskinesia) to 4 (incapacitating dyskinesia). The historical UDysRS subscores were evaluated at baseline during V0, in the specific circumstances of this disease, where the OFF movements are when the patient has no medication and his state could be worsened, and the ON movements are in the case the patient is under medication and his state could be improved, which is the opposite to what happens in levodopa-induced dyskinesia in Parkinson Disease. The AIMS score (range=0 to 40 (worse)) was the sum of scores of items 1 to 10, each ranging from 0 (none) to 4 (severe). The PGI and CGI scales are Likert scales ranging from 1 (very much improved or not sick at all) to 7 (very much worsened or among the most ill patients). Each score or subscore item was independently evaluated and assessed by two blinded movement disorders specialists (LLM, AM). When scores were discordant, cases were reviewed by the two specialists until a consensus was reached.

Animals

Housing

Animals, 2-6 per cage (except for the period of isolation for the ones participating to Home cage monitoring), were maintained in a 12:12 light-dark cycle, in stable conditions of temperature (21 ± 1°C), with access to food and water ad libitum. All the described experiments were conducted with both male and female adult littermates aged 2-6 months-old, except when mentioned otherwise. Except for the home cage monitoring, all the behavior experiments were performed between 8 am and 6 pm. All animal procedures were in accordance with the guidelines for French Animal Welfare and were reviewed and approved by the French Agriculture and Forestry Ministry (APAFIS #27244-2020091611215078, APAFIS #29507-2021012615272491, APAFIS #37798-2022072210337537, APAFIS #37798-2022072210337537).

Method details

Motion capture in the patient study

The recording of videos during clinical scales and questionnaires (AIMS and cumulative UDysRS) was done during motion capture with the Vicon™ system using Vicon™ motion sensors to obtain an objective measure of the evolution of movements. Vicon™ Motions Systems Limited is a set of class I Medical devices with a CE-mark.

Multiple sensors (from 4 to 14 mm in diameter), positioned on the face, chest, upper and lower limbs of the subject, were tracked by infrared cameras to generate kinematics parameters. A reconstruction of the position of the different sensors in time and space allows a 3D analysis of movements. Volume of displacement of each body part (left upper limb, right upper limb, left lower limb, right lower limb, trunk, full body) during four tasks (Standing still eyes opened, Sitting in rest, Sitting with elevated arms sideward, Turn head to right then to left while sitting) were quantified and correlated to clinical scores available (MDS-UDysRS, and AIMS) and their variations. Changes in volume of displacement after caffeinated coffee vs decaffeinated coffee were compared. The correlations between volumes of displacement and PK parameters for caffeine and its metabolites in relation to caffeinated versus decaffeinated coffee were assessed.

Motion data processing

Pre-processing motion data

To ensure signal quality and consistency before analysis, the motion data underwent a series of preprocessing steps, including noise reduction using a Butterworth filter and data normalization.

Butterworth filtering

Raw motion data often contains high-frequency noise caused by sensor inaccuracies and environmental factors. To minimize this noise while preserving essential motion characteristics, a second-order low-pass Butterworth filter with a cutoff frequency of 10 Hz was applied. This filtering process effectively smooths the signal, reducing unwanted fluctuations without introducing significant phase distortion.

Normalization

To standardize the data across different subjects and trials, Z-score normalization was applied. This method transforms the data by subtracting the mean and dividing by the standard deviation, ensuring comparability and mitigating variability due to individual differences.

By applying these preprocessing techniques, the motion data were refined to be smooth, noise-free, and standardized, enabling robust and reliable analysis.

Pharmacokinetic parameters in the patient study

The analysis of caffeine and its metabolites (Theobromine, Paraxanthine, Theophylline) blood levels were conducted for each sampling time using a tandem mass spectrometry method. The assay method was validated according to FDA criteria.⁵⁸ Individual pharmacokinetic parameters for each patient were determined using the WinNonlin software version 5, employing the trapezoidal method to calculate the following parameters: Concentration at baseline (C0) in μg/mL; maximum concentration (Cmax) in μg/mL; minimum concentration (Cmin) in μg/mL; area under the curve over 8 hours (AUC) in μg.h/mL; Time to peak concentration (Tmax) in hours; Elimination half-life (t_1/2) in hours; Clearance (CL) in mL/h; Steady-state volume of distribution (Vss) in mL.

Generation of the Adcy5^R419W mouse line

The Adcy5^R419W mouse line was generated on a C57BL/6N background using CRISPR/Cas9-mediated genome editing at the Institut Clinique de la Souris – PHENOMIN (http://www.phenomin.fr). Mutations were introduced into exon 2 of the Adcy5 gene (ENSMUSG00000022840), located on chromosome 16.

Four mutations were simultaneously introduced into wild-type C57BL/6N embryos (Figure S8A). A CRISPR ribonucleoprotein complex—comprising a crRNA (5′-GCCGAGCCTGGATACACTCC-3′; MIT score: 88, as predicted by the CRISPOR website), tracrRNA, and wild-type Cas9—was co-electroporated with a 128-nucleotide single-stranded donor DNA into fertilized C57BL/6N oocytes. Resulting founders were selected based on genotyping and confirmed by Sanger sequencing. Among these, the c.1255C>T (p.R419W) substitution recapitulates the human p.R418W variant, accounting for the one-residue offset between mouse and human proteins. To prevent Cas9 recutting after homology-directed repair, two silent mutations—c.1236C>A and c.1239G>T—were incorporated to disrupt the PAM sequence. Additionally, the c.1251G>A change introduced a HindIII restriction site to simplify genotyping.

Heterozygous Adcy5 mutant mice from the selected founder line were backcrossed to the C57BL/6N background for at least 10 generations. Genotype distribution among Adcy5 littermates followed Mendelian ratios, and no lethality was observed in homozygous or heterozygous mutants during embryogenesis, at birth, or in the pre-weaning period.

Drd1a-EGFP mice were obtained from the GENSAT (Gene Expression Nervous System Atlas) program at Rockefeller University.⁵⁰ D1GFP-Adcy5 mutant mice were generated by crossing Adcy5^R419W/+ mice with heterozygous Drd1a-EGFP mice. All the D1GFP-Adcy5 mice used in this study were heterozygous for the Drd1a-EGFP allele.

Genotyping of the Adcy5^R419W mouse line

PCR products were obtained from genomic DNA with the F1 (CAGGCCCGCTGTGTTCCTGAGTGTC) and R1 (GGACTGGGCAGAGGGATGAGCATGG) primers with QIAGEN Multiplex PCR Kit following the program: 95°C 15 min (1 cycle); 95°C 30s, 62°C 30s, 72°C 45s (35 cycles); 72°C 10 min (1 cycle). These products were then digested with HindIII (Thermo Scientific FD0504) for 15 min at 37°C. The product from the mutant allele is cut in 2 fragments of similar weights (159 bp and 160 bp) while the product from the WT allele is uncut (319 bp) (Figure S8B).

Drugs administration in animal studies

Selective A_2AR antagonist Istradefylline (Istradefylline; No.5147/10), non-selective A_2AR antagonist caffeine (No.2793/100), selective A_2AR agonist CGS21680 (No.1063/10), selective D1R agonist SKF81297 (No.1447/10) and AC activator (including AC5) Forskolin (No.1099/10) were purchased from Tocris. Istradefylline, caffeine, and CGS21680, used in investigating locomotion activity, motor coordination, and anxiety-like behaviors, were dissolved in the vehicle solution containing 5% dimethylsulfoxide (DMSO), 5% Tween 20, 15% polyethylene glycol in H₂O. All drugs were administrated to animals via intraperitoneal injections (i.p.). For cAMP Assay, drugs were dissolved in homogenization buffer as described in the corresponding section.

Behavioral studies

Rotarod

For the accelerating rotarod test, mice were placed on a rotarod (Bioseb) accelerating from 4 to 40 rpm over a 5-min period; 7 trials were performed daily for 7 consecutive days with a 5-min intertrial interval. On day 8, for the fixed speed rotarod test, mice were placed on the rotarod at fixed speeds of 5, 10, 15, and then 20 rpm consecutively; only 1 trial was performed at each speed. In both tests, each trial lasted 5 min or ended earlier if the mouse fell off the rotarod. The rotarod test procedure is adapted from.⁵⁹

Home cage monitoring

Mice were isolated in individual cages 2 weeks before the experiment. Then, each home cage was put into a cabinet with identical light/dark cycle, temperature, feeding and drinking conditions as in the breeding zone, acclimated for 24-h before continuous monitoring of behavioral activities through infrared light panels and infrared cameras for 3 consecutive days. Recording and behavior analysis was performed with HomeCageScan 3.0 (Cleversys Inc), and was manually verified. Analysis of the phenotype span and temporal distribution of four categories of behaviors were performed as previously described.³¹

Grip strength test

One week after the home cage monitoring experiment, mice were used for the grip strength test. A Grip Strength Meter (Bioseb, BIO-GS3) was used to measure forelimb and fore- and hindlimb grip strength via a T-bar for anterior paws and a grid for both anterior and posterior paws, respectively. After 3 days of acclimatization to the apparatus, the forelimb strength and the fore- and hindlimb strength were each measured three times for each mouse. The two closest values among the 3 measurements were selected and averaged as the final force (g).

Elevated plus maze

The elevated plus maze apparatus consisted of two open (30 x 5 cm²) and two closed arms (30 x 5 x 15 cm³), made of light grey plastic walls and floor, orthogonal to each other in a cross-shaped maze, extending from a central platform (5 x 5 cm²). The maze is elevated 40 cm from the ground. Mice started the test on the central platform, facing an open arm opposite to the experimenter. The test lasted for 5 min and was recorded by a camera positioned above the experimental set up. The number of entries, time spent, traveled distance, and central and peripheral velocities were automatically analyzed with the TopScan software (Cleversys Inc). Trajectories were automatically generated. After 3 days of rest, mice participated then successively in the tail suspension and forced swim tests with another 3-4 days intra-test interval.

Tail suspension test

Mice tails were taped with a paper hook at the end of the tail and suspended, dangling, by hooking it on the top of an individual open cabinet (55 x 15 x 11.5 cm³) for 6 min. The duration of immobility during the test were recorded automatically by the TST-5 software (Bioseb).

Forced swim test

Mice were placed for 6 min in a semi-transparent cylinder (20 x 20 x 35 cm³) containing 4 L of water at room temperature (25°C ± 0.5). Once the test was over, mice were taken out of the water, dried with clean towels, and put back into their home cage. The performance in water was filmed by a camera and the duration of immobility (defined as floating with the absence of any obvious movements except for those necessary to keep the nose above water, as described previously³⁴ was manually analyzed.

Spontaneous locomotion activity

Each chamber of the cyclotron apparatus (Imetronic, France) was made of a transparent plastic circular corridor, with four infrared beams each placed every 90° from each other. Locomotor activity was quantified as the number of ¼ turns per 5 min. Each time animals crossed two consecutive infrared beams counted as a ¼ turn. For habituation, the mice were handled after a 30-min free exploration in the cyclotron chamber at day-2. They were touched at the injection point on the abdomen after a 30-min free exploration in the cyclotron chamber at day-1, and injected with vehicle solution at day0 (after a 30-min free exploration). They were then placed back into the apparatus for either 60 min or 90 min of free exploration depending on the respective experiment total duration. After this 3-day habituation phase, mice were then subject to either a Istradefylline administration, or a caffeine and CGS21680 administration either co-administered or sequentially administered depending on the experiment. The D1GFP-Adcy5 double mutant mouse strain was used for the experiment with Istradefylline administration: on day 1 of the experiment, after a 30-min free-exploration-period, mice were injected either with Istradefylline 2.5 mg/kg i.p. or vehicle and placed back into the chamber for 60 min of free exploration. On day 8, after a 30-min free exploration-period, mice were injected either with Istradefylline 2.5 mg/kg i.p. or vehicle, placed back into the chamber for 20 min of free exploration, and immediately sacrificed for histochemistry studies (see below). Their performances under caffeine treatment were assessed in comparison to a vehicle-injected-group on the same day. The Adcy5 mutant mouse strain was used for the experiment with caffeine and CGS21680 administration: on day 1 of the experiment, after a 30-min free exploration-period, mice were injected with caffeine 15 mg/kg i.p. and placed back into the chamber for 90-min free exploration. On day 8, after a 30-min free exploration, mice were injected with caffeine 15 mg/kg i.p. and placed back into the chamber for free exploration. Forty-five min later, mice received a sequential injection, this time of CGS21680 15 mg/kg i.p., and were placed back into the chamber again for a 45-min free exploration. On day 15, after a 30-min free exploration, mice were injected with a mix of caffeine 15 mg/kg and CGS21680 10 mg/kg i.p., and placed back into the chamber for 90 min of free exploration. Their performances under each of these treatments were assessed in comparison to their performances on the vehicle-injection day (Habituation Day 3).

Notched beam test

The notched beam test was a modified version of the beam walking test, to detect fine deficits in motor coordination, balance, and equilibrium.^60,61 Neither food nor a safe box were placed at the end of the bar to focus on motor evaluation and prevent biases linked to the striatal rewarding system. A white plastic elevated notched beam (1-m long, 16-mm wide, with a flat upper surface engraved with 26 regular, square, 16-mm deep, 16-mm long incisions spaced every 16 mm) was placed 38 cm above the ground. The notched beam was cleaned with 70% ethanol between each subject evaluation. Mice were always placed on the same end of the beam, and their crossing to the opposite end of the beam was recorded with a camera over a 150-s trial duration. The mutant mice displayed frequent freezing during this task, so the percentage of mice of each genotype able to reach the opposite end of the beam (success rate), and total time to cross (maximal value of 150 s in case of failure) were analyzed. For each condition, mice were tested for notched beam crossing at time points t_-30, t₂₀, t₉₀, and t₁₈₀ min. For each experiment with a drug injection, mice were placed on the notched beam, 30 min before and 20, 90, and 180 min after the acute administration of either the vehicle, caffeine 15 mg/kg, Istradefylline 2.5 mg/kg, or caffeine 15 mg/kg co-administered with CGS21680 10 mg/kg. Istradefylline 2.5 mg/kg acute administration was done one week after the initial acute administration of caffeine 15 mg/kg. The sequential administration of caffeine 15 mg/kg and CGS21680 10 mg/kg was performed one week after the combined injection of caffeine 15mg/kg and CGS21680 10 mg/kg.

Open-field

Two types of open-field apparatus were used in this study following the same experimental procedures. White opaque cylinders with a white base (40 cm diameter x 60 cm height) were used in the circular open-field test, whereas grey plastic squared boxes (50 x 50 x 40 cm³) with a white base were used in the open-field test performed during the notched beam test, between the 3^rd and 4^th trials (t₉₀ and t₁₈₀ min). Mice were initially placed in the center of the apparatus, allowed to freely explore the environment for 5 min. Their behaviors were recorded by a camera positioned overhead. Time, distance, and velocity in the center and periphery were automatically analyzed with the TopScan software (Cleversys Inc). Trajectories were automatically generated.

Biochemical studies

RNA extraction

Fresh dorsal striatal tissues were extracted after mice decapitation. The brain was carefully removed from the skull and placed on a brain matrix. Two blades were put on the 4^th and 9^th crest of the brain matrix to obtain a 2.5 mm thick coronal section containing the whole striatum. The dorsal striatum tissues (both left and right sides) were punched out with a pre-chilled metal stick. The tissues (one mouse per sample) were homogenized and lysed in pre-chilled TRIzol (Life Technologies) with loose and tight glass-glass 2 mL Dounce homogenizers. The total RNA was extracted using Qiagen RNeasy Plus Micro Kit (Cat No./ID: 74034). RNA quality was checked using Agilent TapeStation.

RNA sequencing and analysis

RNA sequencing (RNA-Seq) and analysis were performed with the help of Genotyping and sequencing core (iGenSeq) and the Data Analysis Core (DAC) facilities of the Paris Brain Institute. Reverse-transcribed mRNA (500 ng) was used for mRNA library preparation, realized following manufacturer’s recommendations (NEBnext Ultra 2 mRNA Kit from New England Biolabs). Final samples pooled library prep were sequenced on Nextseq 500 ILLUMINA with MidOutPut cartridge (2 x 130 millions of 75 bases reads) with 2 runs (4plex and 4plex), corresponding to 2 x 30 millions of reads per sample after demultiplexing.

Quality of raw data was evaluated with FastQC.⁵¹ Poor quality sequences and adapters were trimmed or removed with Fastp tool, with default parameters, to retain only good quality paired reads. Star v2.5.3a⁵² was used to align reads on reference genome GRCm38 - mm10 (mus musculus version 10) reference genome using default parameters except for the maximum number of multiple alignments allowed for a read which was set to 1. Quantification of gene and isoform abundances were done with RESEM 1.2.28⁵³ on RefSeq catalogue, prior to normalization with DESeq2⁵⁴ bioconductor package. RNAseq data was mapped to 13,297 identified genes. Downstream analyses and plots generation were performed with the DAC graphical interface, Quby. Comparison between genotypes carried out with DESeq2 on genes with at least 2 normalized counts-per-million in 2 samples. Multiple hypothesis adjusted p-values were calculated with the Benjamini-Hochberg procedure to control false discovery rate (FDR). Then, enrichment analysis was conducted with clusterProfiler R package (v4.2.2)⁵⁵ with over-representation analysis and Gene Set Enrichment Analysis, on gene ontology database http://geneontology.org.

Western Blot

Mice were decapitated and their heads were instantly frozen for 12s in liquid nitrogen and stored at -80°C before use. Fur and skull were eliminated on dry ice and the frozen brains were cut into 210 μm sections using a cryostat (Leica). Dorsal striatal tissue samples were extracted using a pre-chilled metal hollow pipe with an inner diameter of 3 mm. Samples were stored at -80°C before use. The frozen tissues were sonicated in 1% SDS, 1 mM sodium orthovanadate solution (40 Hz, 10 pulses) to prevent undesirable dephosphorylation, and boiled at 95°C for 10 min. Then, 20 μg of total proteins were separated on 4-15% Tris-HCl precast gel (Bio-rad) and transferred onto nitrocellulose membranes (Bio-rad). Membranes were blocked in a 3% bovine serum albumin (BSA) solution for 1h at room temperature, then incubated with primary antibodies (Key Resources Table) in a 3% BSA solution at 4°C overnight. Blots were incubated with secondary anti-rabbit, anti-mouse, anti-chicken or anti-goat IgG DyLight™ 800 or 680 conjugated antibodies (1:5000; Rockland Immunochemicals, Pottstown, PA, USA). The signal from the secondary antibody was acquired using Odyssey Clx (Licor) and quantified using Image Studio Lite (Licor). The signal was normalized to a housekeeping protein used as loading control (β-actin, tubulin or GAPDH depending on the molecular weight of the protein of interest studied) and normalized to the mean value of an internal control sample in each blot.

cAMP assay

After decapitation, mice brains were extracted and placed on ice. The right striatum was dissected and homogenized in 2 mL of homogenization buffer (2 mM Tris Maleate pH 7.2, 2 mM EGTA pH 7.2, 1% sucrose). Ten μL of each striatal sample were added to 10 μL of stimulant (either the AC5 activator Forskolin, A_2AR agonist CGS21680 or D1R agonist SKF81297, diluted in homogenization buffer), and 10 μL of incubation medium (25 mM Tris Maleate pH 7.2, 0.5 mM ATP, 1 mM MgSO₄, 10 mM creatine phosphate, 0.3 mg/mL creatine kinase, 0.4 u/mL adenosine deaminase, 0.1 mM papaverine, and 1 μM GTP) and incubated 10 min at 30°C. For each homogenate, the reaction was conducted in the presence or not of various concentrations of SKF81297 or CGS21680 or Forskolin. The reaction was stopped by 20 μL of cold Lysis & Detection Buffer (Cisbio/Revvity, HTRF cAMP Gs dynamic kit) and stored at -20°C. cAMP concentration was determined by cAMP – Gs Dynamic kit according to the manufacturer’s protocol (Cisbio Bioassays). Protein concentration in tissue homogenates was measured using a bicinchoninic acid (BCA) assay (Pierce™ BCA Protein Assay Kit).

cAMP production (pmol/min/mg of protein) was calculated from the cAMP concentration adjusted by the total protein concentration (measured using the BCA assay) and reaction time. This cAMP production was normalized to unstimulated basal cAMP production in Adcy5^+/+ striata (%), and then to relative AC5 protein levels in Adcy5^+/+ and Adcy5^R419W/R419W striata (quantified by immunoblot) to estimate AC5 specific activity (%).

Histochemistry

Brain sections

Mice were anesthetized using pentobarbital i.p. and perfused intracardially by 4% paraformaldehyde (PFA). Brains were collected and post-fixed in 4% PFA at 4°C overnight, and then cut into coronal serial sections (30 μM thick) with a Vibratome (Leica). Brain sections were conserved in cryoprotective solution (20% glycerol, 30% ethylene glycol in PBS) at -20°C before use.

Immunofluorescence

Free-floating mouse brain sections were washed in Tris-Buffered Saline (TBS)/NaF solution before a 5-min peroxide treatment (10% methanol, 3% H₂O₂ in TBS/NaF solution). Sections were washed in TBS/NaF before a 15-min permeabilization treatment (0.2% Triton-X100 in TBS/NaF solution). After blocking for 30 min at room temperature (3% bovine serum albumin in TBS/NaF), sections were incubated with the primary antibodies described above (key resources table) at 4°C overnight. They were then incubated for 45 min at room temperature with goat Cy3-coupled anti-rabbit (1:500; Jackson ImmunoResearch Laboratories) and goat Alexa Fluor 488-coupled anti-chicken (1:500; Invitrogen) secondary antibodies and mounted in Vectashield (H-1000-10, Vector laboratories).

Image acquisition and quantification

378 × 378 μm² and 290 μm x 290 μm² images were acquired with a SP8 confocal microscope (Leica) for visualizing immunofluorescence. For each section, one image was taken in the dorsal lateral and in the dorsal medial striatum regions on both the right and left sides of the brain. One mosaic image of the left striatum was taken for each genotype and condition. The neuronal quantification was performed by counting DAPI stained nuclei of SPNs. DARPP-32, GFP and Phospho-Serine-235/236-rpS6 positive cells were manually counted by an observer unaware of the drug administration group and genotype of mice. Density of neurons (in n/μm²) was obtained by dividing the number of neurons counted by the surface of the field on each image. The specific expression of GFP under the D1R promoter was used as a marker of dSPNs in D1GFP-Adcy5 mice.^62,63 In the striatal areas assessed in D1GFP-Adcy5 mice, GFP-positive neurons were reported as dSPNs and GFP-negative neurons were reported as iSPNs.

Quantification and statistical analysis

Statistics

For clinical analysis, the effect of treatment (caffeine or placebo) on outcomes of patients (continuous or binary) was analyzed using generalized linear mixed models (under Gaussian or binomial distributions). Treatment and sequence were considered fixed effects, while the subject was treated as a random effect in the model. Spearman correlation coefficients were used to examine associations between variables. For mice analysis, continuous variables were analyzed using Student’s t-test for comparing two groups, one-way ANOVA or linear mixed models followed by Tukey’s post hoc test for multiple comparisons. For dose-response curves analysis, the three-parameter logistic regression model was performed to estimate Max, ED50, and the slope. A p-value of less than 0.05 was considered statistically significant. Results are presented as the mean (standard error, SEM), unless otherwise stated. The specific statistical tests used are described in each figure legend and detailed in Table S5.

Additional resources

SENSEO-ADYC5 study in patients Clinicaltrials.gov ID is NCT05136495

https://clinicaltrials.gov/study/NCT05136495.

Published: December 18, 2025

Supplemental information

Table S3. Excel file containing data too large to fit in a PDF, related to Figures 5 and S6

Expression of genes in Adcy5^R419W/R419W mutant mice compared to Adcy5^+/+ mice

Differential gene expression (FDR<0.05, log2 fold-change >0.5), by RNA-seq analysis in the striatum of Adcy5^R419W/R419W vs. Adcy5^+/+ mice. RNA-seq data were mapped to 13,297 identified genes and comparison between genotypes carried out with DESeq2 (only genes with at least 2 normalized counts-per-million in two samples were included)

(A) Listing of all 13,297 identified genes. With an FDR ≥0.05, 595 differentially expressed genes were identified. Genes (B) downregulated (n = 312) and (C) upregulated (n = 283) in Adcy5^R419W/R419W mice. Statistics: comparison between genotypes carried out with DESeq2 (only genes with at least two normalized counts-per-million in two samples were included); adjust p value: FDR<0.05 and log2FC cutoff>0.5; n = 3 animals per group.

Table S4. Excel file containing data too large to fit in a PDF, related to Figures 5 and S6

Gene set enrichment analysis in Adcy5^R419W/R419Wmutant mice presented in Functional Gene Ontologies analysis of (A) GOBP, (B) GOCC and (C) GOMF components.

Abbreviations: GOBP: Gene Ontology “Biological Process”; GOCC: Gene Ontology “Cellular Component”; GOMF: Gene Ontology “Molecular Function.”

Table S5. Detailed statistics of experiments described in all figures, related to Figures 1–7, S1–S8 and Tables S1 and S2

Excel file containing data too large to fit in a PDF.