The Emerging Role of Phosphodiesterases in Movement Disorders

ABSTRACT Cyclic nucleotide phosphodiesterase (PDE) enzymes catalyze the hydrolysis and inactivation of the cyclic nucleotides cyclic adenosine monophosphate and cyclic guanosine monophosphate, which act as intracellular second messengers for many signal transduction pathways in the central nervous system. Several classes of PDE enzymes with specific tissue distributions and cyclic nucleotide selectivity are highly expressed in brain regions involved in cognitive and motor functions, which are known to be implicated in neurodegenerative diseases, such as Parkinson's disease and Huntington's disease. The indication that PDEs are intimately involved in the pathophysiology of different movement disorders further stems from recent discoveries that mutations in genes encoding different PDEs, including PDE2A, PDE8B, and PDE10A, are responsible for rare forms of monogenic parkinsonism and chorea. We here aim to provide a translational overview of the preclinical and clinical data on PDEs, the role of which is emerging in the field of movement disorders, offering a novel venue for a better understanding of their pathophysiology. Modulating cyclic nucleotide signaling, by either acting on their synthesis or on their degradation, represents a promising area for development of novel therapeutic approaches. The study of PDE mutations linked to monogenic movement disorders offers the opportunity of better understanding the role of PDEs in disease pathogenesis, a necessary step to successfully benefit the treatment of both hyperkinetic and hypokinetic movement disorders. © 2021 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society

A BS TRACT: Cyclic nucleotide phosphodiesterase (PDE) enzymes catalyze the hydrolysis and inactivation of the cyclic nucleotides cyclic adenosine monophosphate and cyclic guanosine monophosphate, which act as intracellular second messengers for many signal transduction pathways in the central nervous system. Several classes of PDE enzymes with specific tissue distributions and cyclic nucleotide selectivity are highly expressed in brain regions involved in cognitive and motor functions, which are known to be implicated in neurodegenerative diseases, such as Parkinson's disease and Huntington's disease. The indication that PDEs are intimately involved in the pathophysiology of different movement disorders further stems from recent discoveries that mutations in genes encoding different PDEs, including PDE2A, PDE8B, and PDE10A, are responsible for rare forms of monogenic parkinsonism and chorea. We here aim to provide a translational overview of the preclinical and clinical data on PDEs, the role of which is emerging in the field of movement disorders, offering a novel venue for a better understanding of their pathophysiology. Modulating cyclic nucleotide signaling, by either acting on their synthesis or on their degradation, represents a promising area for development of novel therapeutic approaches. The study of PDE mutations linked to monogenic movement disorders offers the opportunity of better understanding the role of PDEs in disease pathogenesis, a necessary step to successfully benefit the treatment of both hyperkinetic and hypokinetic movement disorders. © 2021 The Authors.

Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society
Key Words: cyclic nucleotides; Huntington disease; ADCY5; adenylyl cyclases; PDE The cyclic nucleotides cyclic adenosine monophosphate (cAMP) and cyclic guanosine monophosphate (cGMP) act as intracellular second messengers for many signal transduction pathways and modulate a number of processes in the central nervous system (CNS), including neurogenesis, apoptosis, plasticity, as well as sleep, sensorimotor gating, mood stability, memory, and other cognitive functions. 1,2 cAMP and cGMP levels are tightly regulated through a fine balance between their synthesis, which is mediated by the activity of adenylyl and guanylyl cyclases (AC and GC, respectively), and their degradation. 1,3 Cyclic nucleotide phosphodiesterase (PDE) enzymes are responsible for the breakdown of cAMP and cGMP by hydrolysis of phosphodiester bonds, thereby directly regulating the intracellular levels of these second messengers. Over the years, mounting evidence has suggested that alterations in PDE expression, and consequently of cyclic nucleotides levels and their downstream targets, might occur with aging and in different age-related diseases, including Alzheimer's disease, 1,4 and more recently their implication also has been suggested for several movement disorders, including, but not limited to, Huntington's disease (HD) 5 and Parkinson's disease (PD). 6 The notion that PDEs might be intimately implicated in the pathogenesis of different movement disorders has been further corroborated by the evidence that mutations in several PDE genes are responsible for rarer genetic conditions primarily manifesting with movement disorders. [7][8][9] In this review, we aim to provide a translational overview of the preclinical and clinical data on PDEs, the role of which is emerging in the field of movement disorders, offering a novel venue for a better understanding of their pathophysiology and ultimately for the development of alternative therapeutic approaches.

Brain Expression and Function of PDEs
Several neurotransmitters exert their actions via membrane-bound G-protein-coupled receptors, triggering a series of intracellular signaling cascades that are linked to the regulation of second messenger cytosolic concentrations, among which cyclic nucleotides play a pivotal role. 10 Increased cytosolic concentrations of cyclic nucleotides drive the activation of protein kinase A (PKA) and protein kinase G (PKG) and their downstream pathways. 11,12 Among these, the cAMP-regulated phosphoprotein molecular mass 32 kDa (DARPP-32) is thought to play a critical role in mediating many of the neuromodulatory effects of dopamine (DA) on local striatal GABAergic and glutamatergic transmission. Furthermore, it induces the transcriptional activation of cAMP response element binding protein (CREB), which has been reported to play a major role in cognitive and motor functions (Fig.1). [13][14][15] PKG is also a potent activator of striatal DARPP-32. Its phosphorylation of DARPP-32 is rapid, transient, and requires a coincidental increase in intracellular calcium levels induced after glutamatergic stimulation of N-methyl-D-aspartate receptor, AMPA, and metabotropic glutamate 5 receptors. Acting together, the PKA/PKG-mediated pathways finely regulate long-term changes in striatal synaptic efficacy. 16 Alterations of these cascades have been demonstrated in basal ganglia disorders and are therefore likely to be critically involved in the planning and execution of complex motor behaviors. As mentioned earlier, CREB is one of the PKA/PKG target substrates that increases the levels of different neurotrophic factors, including the brain-derived growth factor (BDNF), thus enhancing neuronal growth and development on one side, but also initiates gene transcription involved in long-term synaptic plasticity on the other. [14][15][16][17][18] In the striatum, cyclic nucleotides signaling is controlled by a variety of different neurotransmitters, including DA, glutamate, and adenosine. Several lines of evidence would suggest that activation of AC-cAMP-PKA signaling in striatal medium spiny neurons (MSNs) facilitates corticostriatal transmission, enhancing the excitatory effects of glutamatergic transmission. The excitatory responses induced by N-methyl-D-aspartate receptor are further potentiated by D1 and reduced by D2 receptor stimulation. 10 It is likely that stimulation of postsynaptic AC-cAMP-PKA signaling is primarily driven by D1-like receptors given that they are positively coupled to AC, whereas D2 receptor activation produces the opposite effect by inhibiting AC-cAMP-PKA signaling at both presynaptic and postsynaptic levels (Fig. 1). 19 In contrast, the intracellular concentration of cyclic nucleotides is further determined by the activity of PDEs, which mediate their degradation into their corresponding monophosphate nucleosides. In mammals, PDEs are encoded by 21 genes, which produce 11 distinct PDE families (PDE1-PDE11), which are classified based on their structural similarities and splice variants. 20 PDEs can be further stratified based on their substrate specificity: PDEs 4, 7, and 8 are cAMP specific; PDEs 5, 6, and 9 are cGMP specific; whereas PDEs 1, 2, 3, 10, and 11 can hydrolyze both cyclic nucleotides (Fig. 1). [1][2][3]20,21 Recent work has been performed to characterize PDE expression in different CNS areas. It has been shown that the PDEs most abundant in the basal ganglia are PDE1B and PDE10A, which are equally expressed in the caudate nucleus. 21 The level of PDE1B mRNA in the caudate nucleus is 10-to 100-fold higher than PDE1C and PDE1A, respectively. 21 Similarly, PDE10A mRNA concentrations in the caudate nucleus and nucleus accumbens are at least 10-fold higher than in any other CNS areas. 21 Recent work has suggested that PDE2A is as highly expressed as PDE10A in the striatum, 9 but it is also highly enriched in the hippocampus and other cortical regions. Two other PDE isoenzymes, PDE4B and PDE8B, have been implicated in the pathogenesis of movement disorders. 1,5,19 PDE4B mRNA is the most highly expressed isoenzyme of the PDE4 family in the brain, with similar levels expressed in the substantia nigra, putamen, thalamus, and spinal cord. 21 The highest PDE8B levels in brain are found in the caudate, cortex, and hippocampus. 21 Compared with PDE7A that is highly expressed in many peripheral organs, PDE7B is dominant in the CNS, with the highest expression levels in the caudate, nucleus accumbens, and cortex as compared with relatively low levels in the cerebellum. 21 Conversely, PDE3 mRNA levels are higher in the cerebellum than in all other brain areas, while PDE5, PDE6, PDE9, and PDE11 expressions in the CNS are low in comparison with both other PDE levels and their expression in other tissues. 21 It is likely that different PDE isoforms act collectively in striatal neurons to modulate cyclic nucleotides signaling within specific subcellular compartments and to further regulate fine tuning of the electrical activity patterns and synaptic plasticity changes of striatal projecting neurons (for a review, see Threlfell and West 10 ). Given that these abnormal firing patterns have been demonstrated in disorders of the basal ganglia, modulation of PDEs might reverse abnormal striatal neuronal excitability and synaptic plasticity. For instance, this has been demonstrated in a genetic alphasynuclein-related model of PD, in which the application of the PDE5 inhibitor zaprinast fully restored aberrant long-term depression to normal conditions via the stimulation of a cGMP-PKG-dependent intracellular signaling pathway. 22 In the following sections, we present the role of PDEs in specific movement disorders. To this aim, a literature search was conducted on September 2020 using the PubMed-NCBI database and the following search terms: (1) "phosphodiesterase*" OR "PDE" AND "Huntington" OR "Parkinson*" OR "movement disorder" OR "basal ganglia" OR "striatum"; and (2) "phosphodiesterase*" OR "PDE" AND "gene mutation." Only articles written in English were considered, and the bibliographies of the retrieved articles and of relevant review articles were also checked. The final list of the included articles was identified according to the aim of the current work to provide an integrated and synthesized overview of the current state of knowledge.

Huntington's Disease
Findings in both animal models and patients would suggest an impairment of cyclic nucleotide signaling in HD. 5 Reduced cAMP levels and CREB activity have been demonstrated in both the cortex and striatum of the HD mouse model, 23 which could be possibly owing to increased cyclic nucleotide degradation by PDEs. 24 The most implicated PDE in HD is PDE10A given that it is particularly enriched in the striatum, 21,25 but there is also suggestion that PDE4A, PDE1B, PDE5, PDE9A, and PDE3 might be playing a role in HD pathogenesis. [26][27][28] Tables 1 and 2 summarize the evidence from preclinical and clinical studies, respectively, specifically investigating PDE activity and cyclic nucleotides signaling in HD.
The striatal mRNA expression of PDE4A in R6/2 mice (ie, one of the first transgenic HD models created by inserting a copy of the five untranslated regions, exon 1 with about 150 CAG repeats, and part of intron 1 of the human HD gene into the mouse genome) has been found to be proportionately decreased as the animals age. 29 However, significant decrease in striatal CREB activity has also been demonstrated in R6/2 mice. 29 Therefore, it has been suggested that reduced expression of PDE4 could be compensatory and would depend on the effect of mutant huntingtin (mHTT) protein that sequesters DISC1, a protein that normally binds to and inhibits PDE4B, which in turn may lead to reduced cAMP levels and PKA and CREB activity. 24 Of note, the PDE4 inhibitor rolipram increases CREB phosphorylation exerting neuroprotective effects in both the quinolinic acid (QA) rat model 30 and genetic R6/2 mouse model of HD. 31 As mentioned earlier, PDE4 is mainly expressed in the cortex and would drive the psychiatric, but not motor, phenotype of HD, 24 although rolipram has been further found to rescue motor coordination and activity deficits of R6/2 mice. 32 Compensatory changes caused by direct effects of mHTT onto cycling nucleotide signaling have been further postulated for PDE1B and PDE10A. PDE1B mRNA levels have been reported to be reduced in R6/2, 33 R6/1, 34 and N171-82Q HD transgenic mouse models, 35 as well as cDNA obtained from mRNA of symptomatic patients with HD, 36 suggesting a mHTTdependent decrease in PDE1B expression. The PDE1 inhibitor, vinpocetine, has been reported to ameliorate both biochemical and behavioral abnormalities in toxic models of HD. 37 A number of studies have shown that PDE10A inhibition is able to rescue behavioral, neurodegenerative, and electrophysiological deficits in HD animal models, which correlates with an increase of CREB activity and BDNF levels in both the striatum and the cortex. [38][39][40][41][42] However, conflicting with these beneficial results, PDE10A mRNA expression has been shown to be consistently reduced in both cell 23 and animal 29 models of HD, as well as in brain tissue or cerebrospinal fluid of patients with HD. 23,[43][44][45] Similarly to PDE4A and PDE1B, it has been speculated that this reduction may be compensatory to a primary decrease in cAMP levels driven directly by mHTT. 41 An alternative possible answer was suggested by a study investigating PDE10A levels in a specific subpopulation of striatal neurons. PED10A levels were found to be increased in MSNs and other subsets of striatal interneurons (ie, parvalbuminergic, somatostatinergic, and calretininergic interneurons) in R6/2 transgenic HD mice compared with the wild-type mice, whereas striatal cholinergic interneurons PDE10A levels were lower. 46 Importantly, densitometric studies of the whole striatum showed lower PDE10A immunoreactivity in the R6/2 than in the wild-type mice. 46 This suggests that, although PDE10A levels are lower in the striatum in toto, it might be overexpressed in MSNs, where it inhibits cyclic nucleotides and downstream signaling. In contrast, low levels of PDE10A found in cholinergic interneurons might be possibly related to their selective resistance to HD neurodegeneration. 5 It should be noted, however, that the observed increase of PED10A levels in MSNs occurred at 4 and 9 weeks of age in R6/2 mice, whereas a significant decrease was observed at 13 weeks of age, 46 which might further suggest that PDE10A levels might initially increase in a compensatory fashion and subsequently decrease throughout disease progression.
In humans, a positron emission tomography (PET) study has shown that striatal binding of the PED10A ligand, [ 18 F]MNI-659, was significantly lower in both patients with HD and premanifesting HD gene carriers than in healthy control subjects, and strongly correlated with brain atrophy, disease severity, and the burden of pathology (ie, an estimation of the genetic burden adjusted for age). 47 Longitudinal PET studies have further shown that the mean annualized rate of decline in [ 18 F]MNI-659 caudate binding was about 16% compared with less than 1% in healthy volunteers, suggesting it might efficiently track HD progression and in a more sensitive way than with both functional and volumetric imaging methods that are currently used. 48,49 Another study using the PED10A ligand, [11] C-IMA107, showed bidirectional changes in PDE10A expression in premanifest HD gene carriers, with lower binding in the striatum and pallidum and increased binding in motor thalamic nuclei than in healthy volunteers. 50 The ratio between higher PDE10A expression in motor thalamic nuclei and lower PDE10A expression in striatopallidal projecting striatum was found to be the strongest predictor of symptomatic conversion. 50 Notably, these changes were observed early before symptomatic onset and when no brain atrophy was detectable, 50 indicating that reduction of PDE10A expression does not merely reflect loss of MSNs. Decreased PED10A expression could be further observed in extrastriatal areas, which could account for the development of cognitive and psychiatric symptoms in HD. 51 Taken together, this body of work suggests that PDE10A levels in patients with HD are prominently and progressively reduced through disease progression and might be primarily linked to the phenotypic manifestations, which is also supported, as discussed in more detail later, by the identification of loss-of-function PED10A mutations causing chorea in humans. 52 This would contradict the preclinical evidence suggesting that PED10A alteration might be merely compensatory to cAMP/cGMP signaling   changes driven by mHTT, as well as the arguably weak evidence of neuroprotective effects of PDE10A inhibitors. In fact, in humans, the Pfizer compound PF-02545920 (ie, PDE10A inhibitor) is the only one that has currently completed a 6-month multicenter phase 2 study (AMARYLLIS) investigating two doses (5 and 20 mg) versus placebo and failed to meet its primary end point (ie, the total motor score of the Unified Huntington's Disease Rating Scale) as publicly announced by Pfizer in a press release. 53 Finally, the PDE5 inhibitor sildenafil has been found to reduce striatal projection neurons loss, increase CREB activity and BDNF levels, as well as improve the phenotype of the 3-nitroproprionic neurotoxic rat model of HD, 54 which further points to a significant role of cGMP signaling in HD pathogenesis, as also suggested by a recent study that used the selective PDE9A inhibitor PF-04447943 in transgenic HD rats and demonstrated a normalization of cortically evoked firing of striatal MSNs. 27 Because PDE9A is a cGMPspecific enzyme, its inhibitors facilitate striatal cGMP signaling and, in turn, corticostriatal transmission. Similarly to earlier, however, it remains to be clarified whether the observed alterations are primarily linked to HD pathogenesis or reflect compensatory changes.

Parkinson's Disease
There are somewhat conflicting results as to whether PDEs are implicated in the pathogenesis of PD (Tables 3 and 4 summarize the results of the preclinical and clinical studies investigating PDEs and cyclic nucleotides involvement in PD, respectively). An in vitro study showed that decreased BDNF expression determines loss of dopaminergic neurons in the substantia nigra, 55 suggesting that cyclic nucleotide signaling might be altered in PD. 56 In fact, there is evidence of a reduction of cGMP levels in the striatum and globus pallidus, along with an increase of striatal cAMP levels in 6-hydroxydopamine (6-OHDA) models. 56 Such alteration might depend on an increase of PDE1B 57,58 and a decrease of PDE10A expression, 57,59 respectively. This type of bimodal regulation, which has also been demonstrated in animal models unrelated to PD, 60,61 would be because of the fact that PDE1B is highly expressed in the striatonigral pathway, whereas       57 6-OHDA rat model • Dyskinetic animals had lower levels of cAMP/cGMP during the increasing phase of dyskinesias than eukinetic animals, but their levels increased in the extinction phase of dyskinesias • Both dyskinesias and the abnormal lowering of striatal cAMP/cGMP could be prevented by amantadine • Total cGMP-specific PDE activity was increased in dyskinetic animals • After L-dopa, PDE1B and PDE10A expression did not increase and, accordingly, PDE1B and PDE10A inhibitors failed to prevent dyskinesias cAMP/cGMP signaling is involved in the development of LID, but the striatal PDE isoforms contributing to this abnormality are to be determined Morales-Garcia (2015) 81 6-OHDA rat model and primary rat mesencephalic cells • S14 or BRL50481 (two PDE7 inhibitors) increased the level of CREB • S14 compound manifested mitogenic properties and elicited differentiation of stem cells toward a dopaminergic phenotype • S14 induced a significant increase in the SNpc neurons in rats lesioned with 6-OHDA PDE10A would be more expressed in the striatopallidal pathway. 62 A reduction of PDE10A activity also has been demonstrated in vivo in patients with PD: decreased striatal and pallidal binding of 11 C-IMA107 (ie, a highly selective PDE10A radioligand) was found in both patients with early and patients with more

First Author (Year of Publication) Model Main Results Conclusions
• Cilostazol revealed anti-inflammatory activity and promoted autophagy Russo (2018) 75 LRRK2 G2019S knock-in mice and transfected microglial cells • LRRK2 kinase activity was found to be a negative regulator of PKA activation state in microglia by affecting PDE4 activity and modulating cAMP content  advanced PD, which correlated with the duration of the disease, as well as the severity of both motor symptoms and motor complications. 63,64 However, these results have not been confirmed 65 and quite counterintuitively, inhibition of PDE10A 66,67 exerts anti-inflammatory and neuroprotective effects in -MPTP models of PD. Notably, however, differently from 6-OHDA models, an increase in both cAMP and cGMP activity has been found in some, 68,69 but not all, 70 studies using 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) models of PD. In fact, one study showed that motor deficits, as well as reduced cAMP/cGMP levels, could be rescued by the PDE1 inhibitor vinpocetine. 70 Notably, some classic antiparkinsonian medications, such as selegiline and amantadine, play an inhibitory effect on PDE1A2, 71,72 which suggests a possible additional mechanism for their symptomatic effect in PD. Moreover, different preclinical studies have shown the potential therapeutic role of different PDE4 inhibitors in PD: thus, MPTP neurotoxicity can be partially reversed by restoring DA levels in mesencephalic neurons, by attenuating neuroinflammation, and by triggering autophagy. 73,74 Notably, leucine-rich repeat kinase 2 recently has been found to modulate PKA activity by affecting PDE4 activity in microglia, modulating cAMP degradation, content, and its dependent signaling. 75 Interestingly, PDE4 inhibitors have been shown to improve motor function in a mouse model of Batten disease, a lysosomal storage disorder, which features parkinsonism, 76 but failed to improve parkinsonism in humans in a small, double-blind, crossover trial. 77 Recent PET studies using [ 11 C]rolipram did demonstrate loss of PDE4 expression in several brain areas, including the striato-thalamo-cortical circuit in patients with PD, that, however, was not correlated with the presence and severity of motor symptoms, but with working memory deficits 78 and excessive daytime sleepiness. 79 The potential of PDE inhibitors in preventing dopaminergic neuronal loss has been further tested in LPSinduced dopaminergic toxicity in rats. Hence S-14 and BRL 50481, both of which are PDE7 inhibitors, showed similar neuroprotective effects, reducing microglial activation and neuronal loss, which were paralleled by an elevation in cAMP levels. 80,81 These results would be supported by a recent study using different in vitro and in vivo models of PD that showed a significant PDE7 upregulation taking place mainly in degenerating dopaminergic neurons and in microglia cells. 82 Recent evidence also implicates PDE3 in PD: in a rotenone rat model, cilostazol (ie, a PDE3 inhibitor) exerted a potential protective effect by enhancing Nurr1 (a transcription factor that regulates expression of the gene encoding tyrosine hydroxylase and development of midbrain DA-producing neurons, the mutations of which cause a rare form of adult-onset PD 83 ) and by modulating apoptosis and autophagy. 84 Preclinical studies have also implicated abnormal cAMP/cGMP signaling in the development of levodopa (L-dopa)-induced dyskinesia (LID) in PD, suggesting a role of PDEs in the development of such motor complications. Thus, dyskinetic animals show lower striatal cAMP/cGMP levels at 60 minutes after L-dopa administration compared with nondyskinetic animals, which upregulates during the extinction phase of dyskinesia (90-150 minutes after L-dopa administration). 57 This is in line with the evidence that striatopallidal cAMP was significantly reduced at peak of dyskinesias compared with nondyskinetic animals. 85 Importantly, administration of the PDE inhibitor zaprinast (nonselective inhibitor of PDEs 5, 6, 9, 10, and 11) effectively reduced LID severity, preventing cyclic nucleotides level decrement. 85 Moreover, UK-343664 (ie, a PDE5 inhibitor) and MR1916 (ie, a PDE10A inhibitor) have also been reported to attenuate LID in 6-OHDA rats 86,87 and parkinsonian monkeys, 88 suggesting that PDE could, in fact, be targeted for preventing or treating LID in PD.

Tardive Dyskinesias
In animals chronically exposed to either haloperidol or clozapine, increased PDE1B, PDE4, and PDE10A levels have been documented in both membrane and cytosolic fractions, 89 suggesting that there is an upregulation of PDE activity contributing to abnormal cyclic nucleotide signaling. In line with this, the PDE4 inhibitor rolipram has been reported to suppress involuntary orofacial movements in haloperidol-induced tardive dyskinesia (TD) in rats. 90 However, the mechanisms whereby the improvement occurs are not entirely understood because, as mentioned earlier, PDE4 is mainly expressed at the cortex, which is consistent with the antipsychotic effects of PDE4 inhibitors. PDE4B genetic variants were, in fact, not associated with TD incidence and severity in patients with schizophrenia chronically treated with antipsychotics. 91 More work is therefore needed to fully understand the role of PDEs in the genesis and as a potential therapeutic target in TD.

Movement Disorders Resulting From PDE Gene Mutations
In 2010, a heterozygous mutation of PDE8B was first described in a German family as causative of autosomal dominant striatal degeneration (ADSD), a rare genetic disorder that is characterized by slowly progressive dysarthria, brisk deep tendon reflexes, and mild parkinsonism without tremor and poor response to Ldopa treatment. 7 ADSD, with onset in the fourth to fifth decade of life, is further characterized by a distinctive magnetic resonance imaging (MRI) pattern featuring symmetrical hyperintensities of the caudate, putamen, and nucleus accumbens. 7 A few other patients with the same phenotype and imaging findings have been reported. [92][93][94] In all cases, heterozygous frameshift mutations in PDE8B resulted in a stop codon leading to a severely truncated and dysfunctional protein. 7,[92][93][94] Because PDE8B is highly expressed in the striatum and regulates cAMP levels, which is the second messenger of DA receptors, and these patients do not respond to L-dopa therapy, it was proposed that ADSD is caused by a defect in DA signaling downstream of the DA receptors.
Heterozygous PDE10A mutations have been recently implicated in a childhood-onset nonprogressive choreic syndrome with bilateral striatal hyperintensity on MRI. 8,95-97 All identified missense mutations (p.F300L, p.F334L/C) have been demonstrated to recurrently affect amino acid residues that are completely conserved down to invertebrate species and are located in the GAF-B domain, likely altering the morphology of the cAMP binding pocket. 8 Interestingly, in vitro studies showed that these mutations do not substantially impair basal PDE10A activity but severely affect the positive regulatory mechanism of cAMP binding to the GAF-B domain on PDE catalytic activity. 8 It was therefore proposed that dominant PDE10A mutations may have a strong impact on the regulation of MSNs activity, especially when MSNs are activated by high levels of cAMP. 8 However, recent evidence showed that heterozygous mutations in the GAF-B domain may also drive misprocessing and misfolding of the protein. 98 Notably, another report demonstrated that loss-offunction biallelic PDE10A mutations, located in the GAF-A domain and resulting in severe striatal PDE10A loss, are responsible for an infantile-onset hyperkinetic movement disorder, indicating that loss of PDE10A activity is sufficient to cause chorea in humans. 52 Interestingly, patients with biallelic PDE10A mutations did not show the striking abnormalities of the striata observed on imaging of patients with dominant variants in the GAF-B domain, indicating that dominant and recessive mutations have different pathogenic mechanisms.
This work also showed that the knock-in mouse model carrying the homologous variant identified in humans had decreased striatal PDE10A and an impaired capacity to degrade cAMP, further highlighting the critical role of PDE10A in motor control across species. 52 However, surprisingly, these mice showed a clear hypokinetic movement disorder, suggesting that the same reduction in PDE10A activity results in radically opposite effects on movement production in humans and mice.
Finally, Salpietro et al 9 have demonstrated PDE2A homozygous mutations in a single case with childhoodonset chorea with epilepsy and cognitive impairment. In vitro functional studies showed that both cAMP and cGMP hydrolysis were decreased approximately 4-fold and 6.5-fold, respectively, for the mutant compared with the wild-type PDE2A. 9 In addition, it was reported that neither cAMP nor cGMP could increase enzyme activity, indicating a severe disruption of catalytic activity of mutated PDE2A. 9 The high expression of PDE2A in extrastriatal brain regions 9,21 might account for the presence of additional neurological signs, beyond chorea. Recently, PDE2A homozygous mutations have been confirmed to cause a complex infantile syndrome consisting of paroxysmal dyskinesia with cognitive disability and electroencephalographic abnormalities or overt epilepsy. 99

Unsolved Questions and Conclusions
There is mounting evidence of abnormal cyclic nucleotide signaling in the development of movement disorders. Although we focused in this review on the role of PDEs, additional evidence supporting the critical role of cyclic nucleotide signaling in basal ganglia circuitry regulation and movement disorder pathogenesis stems from the discovery that mutations in GNAL and ADCY5 are also a cause of hyperkinetic movement disorders, featuring dystonia, chorea, and myoclonus in variable combinations. [100][101][102][103][104] Interestingly, patients with ADCY5 mutations also share microstructural basal ganglia pathology with cases with PDE10A mutations. 105 GNAL and ADCY5 code, respectively, for Gα olf and AC5, two of the main molecules involved in cAMP signaling downstream of activation of DA and adenosine receptors in MSNs of the direct and indirect pathways, respectively. 106 Pathogenic GNAL mutations are loss of function, hence resulting in reduced cAMP synthesis in MSNs, 101 while the majority of pathogenic ADCY5 mutations cause an increase of AC5 activity, and hence an increase of cAMP production. 104,107 Importantly, the fact that mutations with opposite effect on cAMP production in MSNs result in somewhat similar movement disorders clearly indicates the extreme sensitivity of this pathway to perturbations in either direction, highlighting some of the potential pitfalls and difficulties for developing novel therapeutic strategies focused on MSNs cAMP metabolism. Nonetheless, given the relative ease of modulating cyclic nucleotide signaling, by either acting on their synthesis (ie, on AC) 107 or on their degradation (ie, on PDEs), these could be promising therapeutic targets, despite the negative results obtained so far.
More research is, however, needed to understand whether some of the observed alterations reflect compensatory rather than primary changes. For instance, it would seem that PDE10A levels are actually reduced in humans with HD, and an approach opposite to what has been pursued in the AMARYLLIS study would consist in enhancing rather than inhibiting its activity. However, a full understanding of the delicate balance between PDE10A activity in the direct and indirect pathways, in both physiological and pathological conditions, as well as how to target PDEs in both a brain region-specific and a compartmentspecific manner, is required. Finally, there is a limited number of studies investigating PDE isoform expression and function in humans. The conflicting results arising from preclinical studies using different models might be further owing to the specificity of the models under investigation, which might not exactly recapitulate the pathophysiology occurring in humans. For instance, the observation that the same loss-of-function PDE10A mutation causes a hyperkinetic movement disorder in humans and a hypokinetic motor phenotype in mouse models clearly suggests that there may be species-specific differences in the way PDE10A activity modulates the output of striatal circuits.
As far as PD is concerned, there is increasing evidence implicating cyclic nucleotide signaling in its pathogenesis, but there still remain a number of unsolved questions. If, on the one hand, PDE inhibition has been shown in vitro to exert neuroprotective effects, on the other, it is surprising that many studies targeted PDE isoforms, like PDE5, that are not among the most expressed in the striatum. Moreover, more work is needed to understand which specific PDE isoforms are implicated in PD pathogenesis and how each of them interacts with the underlying neurodegenerative process. For instance, when the activity of direct pathway striatal MSNs is reduced by D1 antagonism, PDE10A inhibitors increase the indirect pathway drive, therefore acting as D2-like antagonists. 108 Conversely, they exert D1-like agonism, when the activity of the indirect pathway is reduced by the administration of the D2-antagonist haloperidol. 89 PDE10A inhibitors can hence activate both striatonigral and striatopallidal MSNs, depending on their underlying state, 109 despite having a stronger impact on the latter. 18 Conversely, PDE1B has a preferential modulatory effect on the striatonigral MSNs. 19 Given that striatonigral and striatopallidal MSNs exert opposing effects on motor control, it is likely that these two PDEs interact to produce the appropriate motor output, but how this occurs remains unknown.
Therefore, more work is warranted to obtain a better understanding of the link between PDE functioning and the pathogenetic mechanisms in humans of each single condition examined in this review.

Data Availability Statement
No data are available.