Therapeutic targeting of 3’,5’-cyclic nucleotide phosphodiesterases: Inhibition and beyond

Abstract

Phosphodiesterases (PDEs), enzymes that degrade 3’,5’-cyclic nucleotides, are being pursued as therapeutic targets for several diseases, including those affecting the nervous system, cardiovascular system, fertility, immunity, cancer, and metabolism. Clinical development programmes have focused exclusively on catalytic inhibition, which continues to be a strong focus of ongoing drug discovery efforts. However, emerging evidence supports novel strategies to therapeutically target PDE function, including enhancing catalytic activity, normalizing altered compartmentalization, modulating post-translational modifications, as well as the potential use of PDEs as disease biomarkers. Importantly, a more refined appreciation of the intramolecular mechanisms regulating PDE function and trafficking is emerging, making these pioneering drug discovery efforts tractable.

Introduction

Conventional 3’,5’-cyclic nucleotide phosphodiesterases (PDEs) are members of a highly conserved superfamily of enzymes that degrade the canonical cyclic nucleotides 3’,5’-cyclic adenosine monophosphate (cAMP) and 3’,5’-cyclic guanosine monophosphate (cGMP)¹, as well as the non-canonical cyclic nucleotides 3’,5’-cCMP, 3’,5’-cUMP, 3’,5’-cIMP and c-di-GMP^2–4 (Figure 1). As extensively reviewed elsewhere¹, there are 11 families of PDEs that are grouped based on the homology of their C-terminal catalytic domain, and each PDE family has multiple isoforms that differ in terms of the length and complexity of their N-terminal regulatory domains (Figure 2). PDEs do not simply control the total cellular content of cyclic nucleotides, but rather create individual pockets or nanodomains of cyclic nucleotide signaling. It is this subcellular compartmentalization of cyclic nucleotide signaling that enables a single cell to respond discretely to multiple extracellular and intracellular signals. Thus, PDEs regulate a myriad of physiological processes, and their dysfunction has been associated with a number of pathophysiological states including those affecting the nervous system, cardiovascular system, fertility, immunity, cancer, and metabolism (Box 1). Because the location of a PDE is just as important to its overall function as is its catalytic activity, how the location of a given PDE isoform can change based on tissue type, age, or disease status—possibly due to factors such as activation of receptors, alterations in calcium signaling, or elevations in cyclic nucleotides—is of paramount importance when considering the therapeutic potential of a given PDE isoform (e.g.,^5–12; for full review, see Table S1).

Figure 1. 11 families of PDEs degrade cyclic nucleotides.

Both of the canonical cyclic nucleotide signaling pathways (cAMP and cGMP) are composed of numerous molecules responsible for the synthesis, execution, and breakdown of their signals. cAMP is synthesized by transmembrane adenylyl cyclases (ACs) that are activated by Gαs and inhibited by Gαi³¹⁵ as well as soluble ACs that are activated by bicarbonate and calcium³¹⁶. cGMP is synthesized by particulate guanylyl cyclases (pGCs) that are activated by natriuretic peptides and soluble guanylyl cyclases (sGCs) that are activated by nitric oxide (NO)³¹⁷. Both cAMP and cGMP activate cyclic nucleotide gated channels and allosterically modulate activity of select PDEs¹⁹⁶. In contrast, only cGMP stimulates protein kinase G (PKG); whereas, cAMP activates protein kinase A (PKA), exchange protein activated by cAMP (Epac) and popeye domain-containing proteins (POPDC)³¹⁸. Signaling through either the cAMP or cGMP pathways ultimately leads to phosphorylation of a myriad of downstream targets, including the transcription factor cAMP response element binding protein (CREB). In addition to cAMP and cGMP, several PDEs also hydrolyze the non-canonical cyclic nucleotides (not included) cUMP (PDE3A, PDE3B, PDE9A), cCMP (PDE7A), and c-di-GMPa (bacterial PDEs), albeit with much lower affinity^{2, 3}.

Figure 2. The 21 phosphodiesterase (PDE) genes are grouped into families (name and substrate specificity listed to right of each illustration) based on the homology of their C-terminal catalytic domain (represented as a semi-ellipse).

Due to alternate promoters and splicing events, each PDE family has multiple isoforms that differ in terms of the length and complexity of their N-terminal regulatory domains (depicted with different shapes), which are thought to regulate subcellular trafficking, substrate affinity, and catalytic activity. The relative size and domain distances were drawn based on estimations from the Pfam/uniprot database, with the exception of the REC domain of PDE8 (estimated from³¹⁹) and the second CaM domain of PDE1 (estimated from³⁰²). Illustrations represent the longest isoform for gene A of each PDE family. CaM, calmodulin-binding domain; GAF, cGMP-binding PDEs Anabaena adenylyl cyclases and E. coli FhlA; TM, transmembrane domain of PDE3; UCR, upstream conserved region; REC, signal receiver domain; PAS, Per-Arnt-Sim domain.

Box 1. Physiological roles of PDEs.

Because PDEs are ubiquitously distributed and are the only enzymes to degrade 3’,5’-cyclic nucleotides, this superfamily of enzymes plays a role in numerous biological processes in health and disease. That said, the biological roles played by a given PDE isoform are distinct due to its unique expression pattern at the level of tissue/organ, cell-type and subcellular compartment (Table S1).

Most PDEs families are expressed in the nervous system where they regulate neurodevelopment and apoptosis, neuronal excitability, synaptic transmission and neuroplasticity^{70, 95}. Every brain region expresses more than one PDE, but no two PDEs exhibit the exact same regional distribution⁷⁰. For instance, PDE11A is the only PDE with brain expression restricted to the hippocampus. While PDE11A regulates how well an individual responds to the mood stabilizer lithium, it does not regulate basal anxiety- and depression-related behaviors¹⁷⁴. In contrast, PDE4D is expressed in most brain regions and does appear to regulate basal depression-related behaviors (e.g.,²⁹³). PDE1B, PDE2A, PDE7B, PDE8B and PDE10A are enriched in the striatum relative to other brain regions⁷⁰, and each has been implicated in regulating basic motor function; whereas, PDE5A and PDE9A that are enriched in the cerebellum have not (for review, see^{70, 294}). In the retina, PDE6 function is central in mediating activation of the light response in rod and cones photoreceptors, and PDE6 mutations cause photoreceptor degeneration in retinitis pigmentosa^{224, 295}.

In the cardiovascular system, PDE2, PDE3 and PDE4 isoforms control different subcellular pools of cyclic nucleotides to regulate important cardiac functions from myocardial contraction/relaxation to chronic cell growth and survival, and disruption of this PDE signaling has been associated with disease (for review, see¹). For example, heart failure has been associated with reduced levels of PDE3A and PDE4D, which results in myocyte apoptosis and cardiac arrhythmias, respectively^{222, 236}.

Many cancers have been associated with reduced levels of cAMP and/or cGMP secondary to an elevation in PDE activity. For example, chronic lymphocytic leukemia cells exhibit increased PDE7B expression; leukemia, colon cancer, and glioma cells overexpress one or more isoforms of PDE4; and colon cancer cells and adenocarcinmoas exhibit elevated PDE5 activity^{16, 296, 297}.

Inflammation of numerous tissue types can be enhanced by a drop in cAMP levels that is caused by an increase in cAMP-PDE activity. For example, activity of PDE4—the predominant cAMP-hydrolyzing enzyme in the immune system—is elevated in the context of various inflammatory diseases, including psoriasis, COPD and asthma^{298, 299}.

PDEs are also implicated in reproductive health. Several isoforms are present in granulosa cells as well as in oocytes in preovulatory follicles of mammalian ovary regulating the meiotic cell cycle³⁰⁰. Furthermore, many PDEs are expressed in cells of the spermatogenic pathway where they may regulate sperm motility^{301, 302}, and PDE5 is expressed in the contractile tissues of the male excurrent tract and accessory where its increased activity contributes to erectile dysfunction (e.g.,²³).

Importantly, no two PDE isoforms share the exact same combination of substrate specificity, tissue expression profile and subcellular localization (Table S1). This is quite important because there are a number of diseases where compartment-specific defects in cyclic nucleotide signaling have been identified. For example, the function of soluble guanylyl cyclase, but not particulate guanylyl cyclase, is significantly impaired in brains of Alzheimer’s patients and in vitro models of Alzheimer’s disease pathology^13–15, which would be expected to decrease cytosolic pools of cGMP. In contrast, in colon cancer, membrane-bound guanylyl cyclase appears to be dysregulated/suppressed¹⁶ and membrane-enriched PDE10A appears to be overexpressed¹⁷, both of which would decrease membrane-proximal pools of cGMP. That said, cytosolic pools of cGMP may also be decreased in colorectal cancer cells, as PDE5A is also overexpressed in these cells¹⁷, and—at least in heart and brain—PDE5 regulates pools of cGMP that are downstream of soluble guanylyl cyclases^{18, 19}.

With regard to compartment-specific changes in cAMP signaling, studies examining brain tissue from patients with bipolar disorder show no change in membrane but increased signaling in the cytosol, which may be normalized by the classic mood stabilizer lithium (e.g.,^{20, 21}). Other disease states where compartment-specific defects in cyclic nucleotide signaling have been implicated include—but are not limited to—erectile dysfunction²², hypertension²³, cardiac hypertrophy^{18, 24}, acrodysostosis^{25, 26}, and Huntington’s disease^{27, 28}. The unique substrate/localization profile offered by each PDE isoform offers multiple degrees of freedom in the context of therapeutic targeting. As such, isoform-specific targeting could enable selective restoration of cyclic nucleotide signaling in affected compartments (i.e., provide efficacy) without disturbing cyclic nucleotide signaling elsewhere (i.e., avoid side effects).

As reviewed in detail below, there has been and continues to be strong interest in developing PDE-targeted therapeutics for a number of diseases. Unfortunately, the majority of PDE-targeted therapeutics on the market are simply competitive blockers of substrate binding at the catalytic site that lack the ability to selectively target a specific isozyme within a single PDE family or sub-family. That said, novel therapeutic strategies are currently being explored to increase the selectivity and specificity with which PDEs are targeted (e.g., by targeting protein-protein interactions). Further, PDE activators are now being considered as agents for treating select diseases, as are the exploitation of PDEs as biomarkers for diagnosis and/or patient selection (Box 2). Here, we review the clinical successes and failures of PDE inhibitors to date and describe a number of ways in which the field is moving beyond pharmacological inhibition of PDEs for therapeutic gain. This review largely focuses on PDE function as it relates to the canonical cyclic nucleotide substrates, cAMP and cGMP, as little is known about the pathophysiology of non-canonical cyclic nucleotides and what is known has been recently reviewed elsewhere⁴.

Box 2. PDEs as disease biomarkers.

PDEs are also being explored as both diagnostic and patient-selection biomarkers. This super family of enzymes, like other genes, can be genotyped from blood samples to assess risk for specific diseases (e.g., high suicide risk,³⁰³; PDE8A, ). PDE mRNA and protein expression can be measured ex vivo in an isoform-specific manner from biopsied samples (e.g, excised tumors). Further, imaging compounds can be engineered with relative ease to selectively target a given PDE family/sub-family in vivo. Thus, PDEs can be explored as biomarkers in diseases where tissue is readily biopsied (e.g,. cancers) and diseases where tissue is not (e.g., diseases of the brain).

Ex vivo biomarkers.

Measurement of PDE4D7 is seen as a valuable biomarker for both pre-surgical and post-surgical risk stratification to optimize treatment decisions^{303, 304} (Table 4). Studies in patient samples showed that PDE4D7 expression is initially upregulated with the development of primary tumours but then is downregulated when the disease progresses to an androgen-independent state (e.g., in castration-resistant tumours)^{305, 306}, consistent with in vitro reports using prostate cancer cells¹⁹⁰. The analysis of PDE4D7 expression in biopsy/surgery samples has been applied to develop InformMDx™ (licensed by MDxHealth from Philips), a tissue-based prognostic prostate cancer biomarker test to stratify patients by risk of disease progression and secondary tumors and, thus, inform post-biopsy/post-surgery treatment decisions (https://mdxhealth.com/press-release/mdxhealth-launch-agreement-philips-prognostic-prostate-cancer-biomarker; accessed 12/03/18).

PDE3A may also represent a useful cancer biomarker as it is greatly expressed in certain cancer cell types such as squamous carcinoma cell lines or gastrointestinal stromal tumour (GIST) cells^{307, 308}. Furthermore, cancer cell lines with the highest PDE3A expression proved the most susceptible to the chemotherapeutic effects of PDE3i’s³⁰⁹. Thus, PDE3A expression could qualify as a biomarker for patient selection which improves patient care by reducing exposure to ineffective drugs and accelerates clinical development of novel therapeutic agents by testing them in targeted populations.

In vivo biomarkers.

Altered cyclic nucleotide signaling has been implicated in a variety of age-related diseases of the brain (c.f.,²³⁵). PDE10A is widely reported as downregulated in both the striatum and cortex of patients with Huntington’s disease, with the extent of PDE10A loss corresponding to the genetic burden associated with the disease^{28, 67, 310, 311}. A loss of PDE10A expression has also been observed in the basal ganglia of patients with Parkinson’s disease³¹². Importantly, highly-specific PDE10A positron emission tomography (PET) tracers shows that PDE10A expression in HD patients continues to decline over years²⁸. Thus, PET imaging of PDE10A could be a useful biomarker for assessing the initial diagnosis and subsequent progression of these neurodegenerative diseases³¹³. PET ligands also exist for PDEs 2, 4, 5, and 7³¹⁴.

Therapeutic PDE inhibition

Several PDE family-selective inhibitors have successfully reached the market, targeting diseases such as psoriasis, COPD and erectile dysfunction (Table 1). Conversely, many potent and selective PDE inhibitors have also failed when tested in clinical trials (Table 2). Below, we will summarize the main achievements and pitfalls in the development of marketed PDE inhibitors to consider factors that currently limit the effectiveness of such therapeutic agents. Given the clinical successes of some PDEi’s as discussed, traditional PDEi’s are still very much being pursued as potential therapeutics, particularly in the context of the central nervous system (CNS), cardiovascular system, reproductive system, cancer and metabolic disorders (Table 3).

Table 1.

Marketed PDE inhibitors

Compound (popular trade names)	Indication	Approval date (USA, Europe and Asia markets)	Side effects
Non-selective
Theophylline [inhibits PDE3, 4, 7, adenosine 2 receptors] (Theolair, Slo-Bid, Theo 24)	Asthma and Bronchoconstriction	FDA (1937), Europe (e.g. Spain, 1922), Asia (e.g. India, 1969)	Nausea, vomiting, diarrhea, headache, irritability, flushing and palpitations.
Aminophylline [inhibits PDE3, 4, 7, adenosine 2 receptor] (Phyllocontin)	Asthma and Bronchoconstriction	FDA (1940), Europe (e.g. Hungary, 1935), Asia (e.g. India, 1950)	Stomach pain, diarrhea, headache, irritability, restlessness and insomnia
Oxtriphylline [inhibits PDE3, 4, 7, adenosine 2 receptors] (Choledyl)	Asthma and Bronchoconstriction	FDA (1981), Europe (only in Greece, 2003)	Stomach pain, nausea, vomiting, diarrhea, headache, irritability, restlessness, insomnia flushing, and increased urination
Dyphylline [inhibits PDE3, 4, 7, adenosine 2 receptors] (Dilor, Lufyllin)	Asthma and Bronchoconstriction	FDA (1951), some countries in Europe (e.g. Spain 1968) and Asia (e.g. Japan, 1952)	Stomach pain, nausea, vomiting, diarrhea, headache, irritability, restlessness, insomnia flushing, and increased urination
Pentoxifylline [inhibits PDE4, 5, adenosine 2 receptors] (Trental, Pentoxil)	Intermittent Claudication	FDA (1984), some countries in Europe (e.g. Spain, 1978) and Asia (e.g. India, 1975)	Belching, bloating, upset stomach, nausea, vomiting, indigestion, dizziness, and flushing, headache
Ibudilast [highest affinity for PDE10A, 4, 11, 3] (Ketas, Pinatos, Eyevinal)	Asthma and dizziness related to cerebral infarction	Asia (Japan, 1989; South Korea, 1998; China, 2003)	Nausea, diarrhea and abdominal pain, depression, rash and fatigue
	Allergic conjunctivitis	Japan (2000)
Tofisopam [highest affinity for PDE4, PDE10, 3, 2] (Emandaxin, Grandaxin)	Anxiety	Some countries in Europe (e.g. Hungary, 1974) and Asia (e.g. Japan, 1985)	Nausea, stomach discomfort, dry mouth, skin rash, insomnia, vomiting and drowsiness
Dipyridamole [highest affinity for PDE8, 1, 3, 2 adenosine deaminase and ENT1] (Persantine)	Post-operative thromboembolism	FDA (1961), some countries in Europe (e.g. Spain 1986) and Asia (e.g. India, 1964)	Headache, dizziness, nausea, diarrhea, muscle pain and vomiting
PDE1
Vinpocetine (Cavinton)	Cerebral vascular disorders and memory impairment	Some countries in Europe (e.g. Spain, 1997) and Asia (e.g. India 2002). USA as an over-the-counter dietary supplement	Flushing, rashes, and minor gastrointestinal disturbances
PDE3
Cilostazol (Pletal, Ekistol)	Intermittent Claudication	FDA (1999), some countries in Europe (e.g. UK, 2000) and Asia (e.g. South Korea, 1990)	Headache, palpitations, diarrhea, dizziness, nasal irritation and pharyngitis
Milrinone (Primacor, Corotrope)	Congestive Heart Failure	FDA (1987), EMA (2016), Asia (e.g. Japan, 1996)	Ventricular/supraventricular arrhythmias, hypotension and headache
Amrinone  (Inamrinone, Inocor)	Congestive heart failure	FDA (1984), some countries in Asia (e.g. India, 1988)	Thrombocytopenia, nausea, diarrhea, hepatotoxicity, arrhythmias and fever
Enoximone (Perfan)	Congestive heart failure	Some countries in Europe (e.g. France, 1987)	Headache, diarrhoea, insomnia, hypotension, vomiting, nausea, tachycardia and arrhythmias
Olprinone (Coretec)	Heart failure	Japan (1996)	Cardiac dysrhythmias and thrombocytopenia
Pimobendan  (Acardi)	Heart failure	Japan (1994)	Headache, palpitation, nausea and ventricular arrhythmias
Anagrelide [also inhibits phospholipase A2] (Agrylin, Xagrid)	Thrombocythemia	FDA (1997), EMA (2004), some countries in Asia (e.g. South Korea, 2004)	Headache, diarrhea, unusual weakness/fatigue, hair loss, nausea and dizziness
PDE4
Roflumilast (Daliresp, Daxas)	Chronic Obstructive Pulmonary Disease (COPD)	FDA (2011), EMA (2010), some countries in Asia (e.g. India, 2014)	Diarrhea, weight loss, nausea, headache, insomnia, decreased appetite
Apremilast (Otezla)	Psoriasis and psoriatic disorders	FDA (2014), EMA (2014), some countries in Asia (e.g. Japan, 2016)	Diarrhea and vomiting, weight loss, mood changes
Crisaborole (Eucrisa)	Moderate Atopic dermatitis (patients >2 years old)	FDA (2016)	Hypersensitivity reactions of the skin
Drotaverine [also inhibits L-type voltage-operated calcium channel] (No-Spa, Doverin)	Functional bowel disorders and alleviating pain caused by smooth muscle spasm	Some countries in Europe (e.g. Hungary, 1963) and Asia (e.g. China, 1999)	Fainting, nausea, vomiting and dry mouth
PDE5
Sildenafil (Viagra, Revatio)	Erectile Dysfunction	FDA (1998), EMA (1998), Asia (e.g. Japan, 1999)	Headache, flushing, dyspepsia, nasal congestion, and impaired vision, including photophobia and blurred vision
	Pulmonary arterial hypertension (PAH)	FDA (2014), EMA (2005), Asia (e.g. Japan, 2008)
Vardenafil (Levitra, Staxyn, Vivanza)	Erectile Dysfunction	FDA (2003), EMA (2003), some countries in Asia (e.g. Japan, 2004)	Headache, flushing, and dyspepsia
Tadalafil (Cialis, Adcirca)	Erectile Dysfunction, benign prostatic hyperplasia	FDA (2003), EMA (2002), some countries in Asia (e.g. India, 2003)	Headache, dyspepsia, back pain and myalgia
	PAH	FDA (2009), EMA (2008), some countries in Asia (e.g. India, 2009)
Avanafil (Stendra, Spedra)	Erectile Dysfunction	FDA (2012), EMA (2013), some countries in Asia (e.g. South Korea, 2011)	Headache, flushing, and nasopharyngitis
Udenafil (Zydena)	Erectile dysfunction and hypertension	Some countries in Asia (e.g. South Korea, 2005)	Headache, dizziness, reddening, nasal congestion, dyspepsia and impaired vision
Mirodenafil (Mvix)	Erectile Dysfunction	South Korea (2007)	Flushing, headache, nasal congestion, eye redness, nausea and dizziness
PDE10A
Papaverine (Pavabid, Pavagen)	Visceral spasm and vasospasm and erectile dysfunction	FDA, Europe (e.g. Hungary, 1933), Asia (e.g. Japan, 1953)	Ventricular tachycardia, diarrhea, somnolence, vertigo, flushing and headache

Source: Drugs.com, drugcentral.org, drugbank.ca, kegg.jp, and approval dates by FDA (Food and Drug Administration), EMA (European Medicines Agency), ANSM (French Agency for the Safety of Health Products), MHRA (Medicines and Healthcare products Regulatory Agency), AEMPS (Spanish Agency of Medicines and Medical Products), EOF (Greek National Organization for Medicines), OGYÉI (National Institute of Pharmacy and Nutrition), PMDA (Pharmaceuticals and Medical Devices Agency), CDSCO (Central Drugs Standard Control Organisation), MFDS (Ministry of Food and Drug Safety) and CFDA (China Food and Drug Administration).

Table 2.

Clinical trials involving PDE inhibitors that failed to reach the market for the described indication

Compound / Manufacturer	Indication	Phase / Status	Clinical trial ID	Cause of failure
PDE2
PF-05180999 / Pfizer	Healthy volunteers (schizophrenia)	Phase I / Completed 2011
	Healthy volunteers	Phase I / Completed 2012
	Migraine	Phase I / Terminated early 2014		Safety concerns
	Migraine	Phase I / Withdrawn prior to enrollment 2014
PDE2/5
Exisulind (Aptosyn) / Cell Pathways	Breast cancer neoplasms and metastases	Phase I/II / Completed 2003
	Breast cancer	Phase II / Completed 2008
	Non-small cell lung cancer	Phase I/II/III / Completed 2003–2007	, ,
	Small cell lung cancer	Phase II / Completed 2008	326	Lack of efficacy
	Prostate cancer and prostatic neoplasms	Phase II / Completed 2006–2011	, , , , 327, 328	Safety and efficacy deficiencies
	Melanoma	Phase II / Completed 2011	329	Lack of efficacy
PDE3
Cilastozol (Pletal) / Otsuka Pharmaceuticals	Type 2 diabetes polyneuropathy	Phase IV / Completed 2009	48	Lack of efficacy
PDE4
ASP9831 / Astellas Pharma	Non-alcoholic steatohepatitis (NASH)	Phase II / Completed 2010	50	Lack of efficacy
PDE5
Sildenafil (Viagra or Revatio) or Tadalafil (Cialis or Adcirca) / Pfizer and Eli Lilly, respectively	Duchenne or Becker Muscular Dystrophy	Early Phase I/Phase I / Completed 2013	, 57
		Phase II/III / Terminated early 2014–2017	, 330–332	Lack of efficacy
		Phase IV / Completed 2012	58
	Impaired glucose tolerance	Phase IV / Terminated early 2016	333	Safety concerns
Vardenafil (Levitra) / Bayer/GSK	Type 2 diabetes	Phase II / Completed 2014	149	Lack of efficacy
PDE9
BI 409306 / Boehringer Ingelheim	AD	Phase II / Completed 2017	, 63, 334	Lack of efficacy
PF-04447943 / Pfizer	AD	Phase II / Completed 2010	59	Lack of efficacy
PDE10
PF-02545920 (a.k.a. MP-10) / Pfizer	Schizophrenia	Phase I / Completed 2007
	Schizophrenia	Phase II / Terminated early 2008		Safety concerns
	Healthy volunteers (glucose metabolism)	Phase I / Completed 2011
	Schizophrenia	Phase II / Completed 2011	66, 335	Lack of efficacy
	Schizophrenia	Phase I / Terminated early 2012		Results from other clinical study
	Schizophrenia	Phase I / Completed 2013
	Schizophrenia	Phase II / Terminated early 2014	336	Lack of efficacy
	Healthy male volunteers (PET imaging)	Phase I / Completed 2014	337
	Huntington’s Disease	Phase II / Completed 2015–2016	,	Lack of efficacy
	Huntington’s Disease	Phase II / Terminated early 2017		Results from
OMS643762 / Omeros	Schizophrenia	Phase II / Completed 2014
	Huntington’s Disease	Phase II / Terminated early 2016		Results from
PBF-999 / Palobiofarma	Huntington’s Disease	Phase I / Completed 2015
	Huntington’s Disease	Phase I / Terminated early 2016		Change in therapeutic indication (cancer)

Reported on Clinicaltrials.gov (accessed 05/28/19). Information is included for all clinical trials involving molecules whose pursuit was terminated after April 2009.

Table 3.

Selected clinical trials involving PDE inhibitors pursued for new indications

Compound / Manufacturer	Indication	Phase / Status	Clinical trial ID (Refs)
PDE3, 4, 10, 11
Ibudilast (AV-411, MN-166) / MediciNova	Opioid withdrawal	Phase II / Completed 2012–2017	,
	Methamphetamine-dependence	Phase I / Completed 2013	77, 78
	Alcohol use disorder	Phase I / Completed 2015	80
	Alcohol use disorder	Phase II / Recruiting
	Opioid abuse	Phase II / Completed 2017	79, 338
	Amyotrophic lateral sclerosis (ALS) (Biomarker study)	Phase II / Active, not recruiting
PDE1
ITI-214 / Intracellular Therapies	Schizophrenia	Phase I / Terminated early 2014
	PD	Phase I/II / Completed 2018
	Healthy volunteers (CNS engagement)	Phase I / Recruiting
	Systolic heart failure	Phase I/II / Recruiting
Vinpocetine / Rxmidas Pharmaceuticals/ Nootrobox	Ischemic stroke	Phase II/III / Completed 2013–2015	, 339
	Cognition enhancement	Not Applicable / Completed 2017
PDE2
TAK-915 / Takeda	Healthy volunteers (PET imaging, schizophrenia)	Phase I / Completed 2016
	Healthy volunteers	Phase I / Completed 2016
PDE3
Cilostazol (Pletal) / Otsuka Pharmaceuticals	Type 2 diabetic atherosclerosis	Phase IV / Completed 2010–2012	, 340
	Chronic tinnitus	Not applicable / Completed 2013	94
	Alzheimer’s Disease	Phase IV / Completed 2013	341
	Atherosclerotic events in type 2 diabetes	Phase IV / Unknown
	Mild Cognitive Impairment	Not applicable / Completed 2015
	Ischemic events in type 2 diabetic artery obstruction	Phase IV / Recruiting
	Antiplatelet aggregation in type 2 diabetes	Phase IV / Active, not recruiting
	Antiplatelet aggregation in type 2 diabetes	Phase IV / Recruiting
	Antiplatelet aggregation in type 2 diabetes	Phase IV / Unknown
PDE4
Apremilast (Otezla) / Celgene Corp.	Vitiligo	Phase II / Active, not recruiting	,
	Lichen Planus of Vulva	Phase II / Not yet recruiting	,
BPN14770 / Tetra Discovery	Alzheimer’s Disease	Phase I / Completed 2016–2017	,
	Alzheimer’s Disease	Phase II / Now recruiting
	Fragile X Syndrome (FXS)	Phase II / Recruiting
Crisaborole (Eucrisa) / Pfizer	Morphea	Phase II / Recruiting
GSK356278/ GlaxoSmithKline	Huntington’s Disease	Phase I / Completed 2012	,
Roflumilast (Daxas or Daliresp) / Astrazeneca	Polycystic Ovary Syndrome	Phase IV / Completed 2014	, 108, 109
	Cognitive deficits in schizophrenia	Phase II / Completed 2015	112
	Cognition (Dementia)	Phase II / Completed 2013–2015	, 2013–001223–39 (EudraCT)111, 113, 342
	Insulin and Blood Sugar Levels in Prediabetic Overweight and Obese Individuals	Phase I/II / Completed 2017	343
HT-0712 / Dart Neuroscience	Age-associated memory impairment (AAMI)	Phase II / Completed 2015
N/A	McCune-Albright Syndrome (PET imaging)	Phase I/II / Recruiting
TAK-648 / Takeda	Type 2 diabetes	Phase I / Completed 2015	, ,
Zembrin / ND	Aged Individuals	Phase I / Completed 2012	39
PDE5
ND	Contrast Media-induced Nephropathy (CMN)	Not Applicable / Unknown
ND	Diabetic nephropathy	Not Applicable / Unknown
Sildenafil (Viagra or Revatio) / Pfizer	Diabetic cardiomyopathy (type 2 diabetes)	Phase IV / Completed 2009	145–148
	Metabolic syndrome (Skeletal muscle insulin signaling)	Phase IV / Completed 2016
	Urolithiasis/urinary stones	Phase IV / Active, not recruiting
	Migraine aura	Early Phase I / Recruiting
	Solid tumors	Phase I / Active, not recruiting	151
	mTBI or concussion	Phase I / Recruiting	,
Sildenafil cream (SST-6007) / Strategic science and technologies/Dare	Sexual arousal disorder	Phase II / Completed 2017
Tadalafil (Cialis or Adcirca) / Eli Lilly	Type 2 Diabetes (Postprandial Hyperglycemia)	Phase I / Terminated early 2011
	Head and neck squamous cell carcinoma	Phase II / Completed 2012–2016	153
	Head and neck squamous cell carcinoma	Phase II / Active, not recruiting
	Obesity	Phase IV / Completed 2015	144
	Insulin secretion/ sensitivity in obesity	Phase IV / Completed 2015	344
	Aortic stenosis (AS) left ventricular remodeling/hypertrophy	Phase IV / Terminated early 2017
	Multiple myeloma (MM)	Phase II / Terminated early 2017
	Insulin Resistance in Type 2 Diabetes	Phase II / Recruiting
	Diabetic cardiomyopathy (DC)	Phase IV / Recruiting
	Endocrine cardiomyopathy in Cushing Syndrome (CS)	Phase II / Recruiting	345
	Access sheath deployment (Nephrolithiasis/kidney stones)	Phase IV / Enrolling 2019
	Lower urinary tract symptoms (prostatic hyperplasia)	Phase IV / Recruiting
	Obesity-related cardiometabolic dysfunction	Phase II / Recruiting
	Anti-tumor Mucin 1 vaccine efficacy in head and neck squamous cell carcinoma (HNSCC)	Phase I/II / Recruiting
	Small Vessel Disease	Phase II / Active, not recruiting	346
PDE9
PF-04447943 / Pfizer	Stable sickle cell disease	Phase I / Completed 2016	347
BI 409306 / Boehringer Ingelheim	Healthy volunteers	Phase I / Completed 2011–2018	, , , , , , , , 61, 348, 349
	Alzheimer’s disease, schizophrenia	Phase I / Completed 2017
	Schizophrenia	Phase I / Completed 2013–2016	, 60, 62
	Schizophrenia or attenuated psychosis syndrome	Phase II / Recruiting	,
	Drug-drug interactions	Phase I / Completed 2016–2017	, , , ,
PDE10
[18F]MNI-659	Huntington’s Disease (PET imaging)	Early Phase I / Completed 2016/2017	, , 2012–003808–13 (EudraCT)350
PBF-999 / Palobiofarma	Cancer	Phase I / Recruiting
EVP-6308 (now FRM-6308) / En Vivo Pharmaceuticals (now Forum Pharmaceuticals)	Schizophrenia	Phase I / Completed 2014	,
RO5545965 / Hoffmann-La Roche	Schizophrenia	Phase I / Completed 2013–2017	, , , ,
TAK-063 / Takeda	Schizophrenia	Phase I / Completed 2014	, , 163
	Schizophrenia	Phase II / Completed 2016	351

Reported on Clinicaltrials.gov (accessed 05/28/19) with an end date after April 2009

PK/PD, Pharmacokinetics and Pharmacodynamics; ND, not described

Marketed PDE inhibitors

The non-selective PDE1 inhibitor (PDE1i) vinpocetine is not FDA approved but is available in over-the-counter supplements (e.g., Cavinton or Intelectol, Richter Gedeon; Cognitex, Life Extension) claiming to improve memory and recovery from stroke, likely due to increasing vasodilation²⁹. As extensively reviewed elsewhere²⁹, a number of clinical trials have examined the cognition-enhancing effects of vinpocetine—either alone or in combination with another compound (e.g., caffeine or Ginko Biloba)—and have generally found improvement in healthy volunteers, individuals with cerebral hypofusion, and possibly aged individuals, but no improvement in AD patients. Reports of side effects associated with vinpocetine have generally been minimal (Table 1,³⁰).

Several PDE3i’s are currently marketed, with Cilostazol and Milrinone perhaps the most widely known. Cilostazol received FDA approval in 1999 for intermittent claudication, but off-label uses include secondary prevention of cerebrovascular accident, percutaneous coronary intervention and coronary stent stenosis (c.f.,³¹). Cilostazol improves function across a number of domains, but it is associated with serious side effects (Table 1) and so is contraindicated for patients with severe heart failure, hepatic impairment, or renal impairment³². Milrinone increases contractility of the heart and is FDA approved for short-term management of severe congestive heart failure. It is particularly used in the context of end-stage heart failure for patients who prove resistant to optimal therapy and for those awaiting cardiac transplant³³. That said, the clinical utility of milrinone has been limited by significant side effects (Table 1) and the fact that it is cleared through the kidneys (i.e., generally not used in patients with renal failure)³³.

Three so-called “second-generation” PDE4i’s are currently FDA approved, with a 4^th compound marketed as an over-the-counter supplement. Roflumilast is an add-on therapy for chronic obstructive pulmonary disorder (COPD; Table 1). Although it causes gastrointestinal and weight loss side effects, making it a third line treatment for COPD, it improves sugar metabolism in obese patients and may decrease cardiovascular events in patients with COPD³⁴. Apremilast is used in the treatment of moderate to severe plaque psoriasis and psoriatic arthritis^35–37 and is also being tested in a Phase IV trial for active ankylosing spondylitis (see below). The most common side effects for both of these orally-administered PDE4i’s are the same that plagued first-generation PDE4i’s (i.e, gastrointestinal disturbances; Table 1), albeit with much improved therapeutic windows³⁵. Crisaborole is a topically-applied ointment for treatment of moderate atopic dermatitis in patients >2 years old. Given the topical nature of the drug, gastrointestinal side effects are avoided and, instead, hypersensitivity reactions are the major possible side effect. Clearly, there is an anti-inflammatory theme shared amongst these FDA-approved PDE4i’s. Zembrin in a non-selective PDE4 inhibitor (also acts as a 5-HT uptake inhibitor) that is a component of a number of herbal supplements claiming calming or mood-stabilizing properties (e.g., Calm, Doctor’s Best; Mood, Procera; Nutri-calm, Nature’s Sunshine)³⁸. fMRI imaging of the amygdala in humans supports an anxiolytic-like effect of Zembrin³⁸. Further, a Phase I trial found Zembrin was well tolerated and improved cognitive flexibility, executive function, mood and sleep³⁹. As noted below, a number of PDE4i’s are currently being pursued to improve cognitive functioning (see below).

There are 4 PDE5i’s currently FDA approved and marketed in the U.S., with 2 additional PDE5i’s marketed outside the U.S.. All 6 PDE5i’s were originally marketed for erectile dysfunction (Table 1), with the most recent approval for avanafil based on its much more rapid onset of action. Sildenafil later received a secondary approval for pulmonary hypertension (contraindicated for pediatric patients, veno-occlusive disease, or sickle cell disease), as did tadalafil. These PDE5i’s are generally considered safe and well tolerated, with no increase in cardiac mortality or myocardial infarction⁴⁰. They share largely similar side effect profiles (Table 1) with headache, flushing, dyspepsia, and vision disturbances being the most common adverse events⁴⁰. Interestingly, udenafil (Zydena, Mezzion Pharma)—one of the PDE5i’s used to treat erectile dysfunction in Korea, Russia and Philippines^{41, 42}--has also been reported to improve cognitive function in patients with erectile dysfunction^{43, 44}.

The success of a number of marketed PDEi’s validates PDEs as appropriate therapeutic targets in many pathological conditions. However, the presence of unwanted side effects resulting from the inability to target individual isoforms is the major limiting factor to success. It is notable that of the 11 PDE families, only agents that attenuate the activity of PDEs 1, 3, 4 and 5 have made it to market, despite significant efforts targeting the inhibition of other PDE families (see next).

Failed PDEi clinical trials

Despite the successes noted above, a number of PDEi’s that entered the clinic failed to make it to market. The selective PDE2i PF-05180999 was originally considered a candidate for cognitive impairments associated with schizophrenia based on its preclinical profile⁴⁵; however, it was brought into the clinic for migraine. Despite the completion of earlier Phase 1 safety and tolerability studies, additional trials were terminated early due to safety concerns (Table 2). Exisulind inhibits both PDE2A and PDE5A (which are overexpressed in a number of precancerous and cancerous cell types) and triggers apoptosis in precancerous/cancerous cells with minimal effects on healthy cells (c.f.,^{46, 47}). Despite promising findings in multiple clinical trials, exisulind failed to secure FDA approval due to deficiencies in safety and efficacy (Table 2, c.f.,^{46, 47}).

As noted above, Cilastozol has gained FDA approval for intermittent claudication; however, clinical trials for other indications, such as type 2 diabetes mellitus peripheral neuropathy, have failed⁴⁸. That said, cilostazol significantly improved walking speed in these patients, suggesting improved peripheral blood flow as would be expected based on its current approved indication⁴⁸.

The PDE4i cilomilast (Ariflo, GlaxoSmithKline) gained FDA approval in 2003 as a second-line treatment for COPD in patients who are poorly responsive to salbutamol⁴⁹. However, cilomilast never made it to market due to the severely dose-limiting nature of gastrointestinal side effects (e.g., nausea and vomiting, diarrhea, and abdominal pain³⁵). The fact that cilomilast elicited more pronounced side effects relative to the other systemically-delivered PDE4i’s described above may be related to preferential inhibition of the PDE4D family relative to the other PDE4 subtypes⁴⁹. A novel PDE4i ASP9831 was tested in Phase I and II trials for non-alcoholic steatohepatitis based on preclinical findings, but failed to improve biochemical biomarkers of the disease⁵⁰. As target engagement in the organ of interest was not confirmed⁵⁰, the reasons underlying the lack of efficacy remain unclear.

A number of clinical trials have attempted to extend therapeutic indications for PDE5i’s, but have failed. As reviewed extensively elsewhere²⁹, a number of trials have tested the effects of sildenafil or vardenafil on various measures of cognition in healthy volunteers or patients with schizophrenia and have largely found no effects^51–55. That said, one report from an Iranian clinical trial did report an improvement in negative symptoms in patients with chronic schizophrenia when sildenafil was administered in addition to risperidone⁵⁶. Several studies were initiated to study sildenafil and/or tadalafil in patients with Duchenne or Becker Muscular dystrophy, with the hopes that the vasodilatory properties of the drugs would improve muscular ischemia; however, clinical trial outcomes have been mixed (Table 2;^{57, 58}).

Two PDE9i’s have been tested in the clinic for cognition-enhancing effects. Although PF-04447943 was found to be safe and well-tolerated, it failed to improve either cognition or dementia-related behavioral disturbances in a Phase II clinical trial⁵⁹. Similarly, BI 409306 was reported as safe and well tolerated in healthy subjects as well as patients with AD or schizophrenia; however, no positive effects on cognition were observed in either patient population (https://www.boehringer-ingelheim.com/PDE9-Inhibition-in-AD, accessed 05/28/19;^60–63). BI 409306 is still being tested in the clinic for prevention of schizophrenia relapse and prevention of first psychotic episode (Table 3). The failure of PDE9i’s to improve functioning in AD may be related to the fact that brain PDE9A is enriched in the nucleus and membrane⁵ and, thus, is not in a position to directly regulate the cytosolic pools of cGMP that appear to be dysregulated in Alzheimer’s disease^13–15.

A number of clinical trials have tested the PDE10i PF-02545920 in schizophrenia and Huntington’s disease (Table 2–3). Despite widely replicated efficacy in a number of preclinical assays (e.g.,^{64, 65}), PF-02545920 failed to improve symptoms in patients with either exacerbated, stable, or sub-optimally treated schizophrenia (Table 2;⁶⁶). Further, in at least 1 trial, there were reports of motoric side effects such as dystonia⁶⁶. Similarly, despite decreased striatal expression of PDE10A being found in patients with Huntington’s disease (Box 2) and promising efficacy of PDE10i’s in preclinical models of the disease^{28, 67, 68}, PF-02545920 failed to improve symptoms in patients with Huntington’s (https://clinicaltrials.gov/ct2/show/results/NCT02197130?sect=X70156#outcome1, accessed 05/28/19), and so efforts for this disease indication were terminated (https://clinicaltrials.gov/ct2/show/NCT02342548, accessed 05/28/19). Several other PDE10i’s have also been pursued in the clinic for schizophrenia and/or HD, with some efforts subsequently suspended or terminated (Table 2) and others ongoing (Table 3—see more below).

The unsuccessful translation of these PDEi’s from promising preclinical data to human testing suggest that therapeutic approaches targeting PDEs need to extend beyond occlusion of the enzyme’s catalytic site. Of particular note are the numerous failures seen in nervous system disorders, even when target engagement was verified. Expression of PDEs in the brain is particularly complex, with PDE isoforms differentially expressed across circuits, cell-types, and subcellular domains^{5, 30, 69, 70}. Thus, the challenge in evaluating the clinical potential for the next generation of PDE-modulating drugs is to gain novel insights about disease-related changes in PDE structure, function and regulation to understand how PDEs should be targeted in a compartment-specific manner for therapeutic gain.

PDE inhibitors in development.

Ongoing efforts in the development of novel PDEis include creation of new chemical entities as well as the repurposing of existing entities. Advances in our understanding of structural differences that exist between PDEs, coupled with extensive medicinal chemistry efforts to optimize structure-activity relationships, have yielded recent vast improvements in terms of selectivity and potency (e.g., see work related to PDE10is and PDE4is^71–73). These efforts have also yielded PDEis with novel modes of action in some cases (i.e., acting as a negative allosteric modulator instead of direct catalytic inhibitor⁷²). In addition, there are still significant efforts to repurpose older PDEi’s. Drug repurposing efforts can be driven by computational or experimental approaches; however, most drug repurposing efforts have been driven either by a better understanding of pharmacology or by a retrospective analysis of clinical effects that were observed during trials or marketed use of a drug for its original indication (c.f.,⁷⁴). Indeed, the PDE5i sildenafil was originally brought into clinical trials for angina but—following observations made by clinicians in that trial—was later repurposed for erectile dysfunction. Drug repurposing has several advantages including reduced risk and substantially reduced timelines and cost due to the fact that the drug would already have passed preclinical and Phase I safety testing and possibly even formulation development⁷⁴. That said, there are a number of barriers to recouping expenses incurred by drug repurposing trials, particularly when they are carried forward by an entity other than the patent holder or following the expiration of the original patent (see⁷⁴ for further discussion).

Non-selective inhibitor

The non-selective PDE3-4-10-11 inhibitor Ibudilast, which also inhibits glial cell activation, is approved for use in Japan as a bronchodialator and has long been of interest as a therapeutic approach for neuropathic pain and substance abuse/withdrawal^{75, 76}. Recent clinical trials have tested Ibudilast in the context of amyotrophic lateral sclerosis, pain, as well as opiate, methamphetamine, and alcohol abuse (Table 3) and positive effects have been reported for all trials completed to date^77–80.

PDE1 inhibitors

The broad PDE1 inhibitor ITI-214, which shows picomolar IC50s for PDE1A, PDE1B and PDE1C in enzymatic assays and >1000-fold selectivity versus its nearest neighbor PDE4^{81, 82}, is being explored for CNS and non-CNS indications. ITI-214 demonstrates cognition-enhancing effects in rodent models of long-term memory and working memory deficits^81–83, mimicking effects of a dopamine receptor 1 (D1) agonist⁸³ and occurring at doses that leave efficacy of the antipsychotic risperidone intact⁸¹. Although the target mediating the cognition-enhancing effects of ITI-214 remains undetermined, PDE1B may be the most likely candidate given its expression in D1-expressing neurons⁸³, along with the fact that a PDE1B-selective inhibitor showed similar cognition-enhancing effects⁸⁴. ITI-214 was moved into the clinic, with potential applications for cognitive deficits associated with schizophrenia, AD, and Parkinson’s Disease⁸², with safety and tolerability established in healthy volunteers and patients with schizophrenia (Table 3). ITI-214 is also being explored in the clinic for heart failure () given its ability to improve cardiac function in dog and rabbit models of heart failure⁸⁵ as a consequence of its inhibition of PDE1C⁸⁵.

PDE2 inhibitors

A number of highly selective PDE2i’s have demonstrated cognition-enhancing, anxiolytic and anti-depressant like-effects in animal models (c.f.,⁸⁶). TAK-915 entered Phase I trials to correlate plasma exposures with central target engagement, with the purpose of informing dose selection for future trials targeting cognitive impairment in schizophrenia^87–89 (Table 3). Looking beyond the brain, PDE2i’s may hold relevance for cardiovascular function since elevated PDE2A expression has been found in failing human hearts as well as a large number of animal models of heart disease (c.f.,⁹⁰). Further, PDE2i’s may hold promise as an antifungal treatment for moderate to severe candidiasis infections, given that genetic deletion of Pde2a reduces virulence and biofilm integrity of the fungal pathogen (c.f.,⁹¹).

PDE3 inhibitors

Despite its existing FDA approval, the efficacy and safety of cilostazol is still very much a topic of investigation, with 27 active clinical trials registered on clinicaltrials.gov (accessed 04/30/2019) and 54 more drawn to a close within just the past 10 years. Numerous recent Phase IV studies appear focused on broadening the therapeutic indications of cilostazol to include vasculature-related insults and nephropathies associated with Type 2 diabetes (Table 3), and recent reports suggest largely positive effects^{31, 48, 92, 93}. This PDE3i also elicited some improvement in chronic tinnitus⁹⁴. Several prospective and retrospective studies have examined cilostazol as a primary or adjunctive treatment for cognitive deficits associated with AD and schizophrenia; the majority of which demonstrated positive effects on cognition (see²⁹ for review). The mechanism by which cilostazol elicits improved cognition has yet to be determined empirically. Given there is very little expression of PDE3A or PDE3B in the brain^{70, 95}, it may be likely that cognition-enhancing effects of cilostazol are driven by increased cerebral blood flow that comes with chronic—but not acute—dosing as opposed to inhibition of PDE3 isoforms directly in the brain (e.g.,⁹⁶).

Novel therapeutic applications of cilostazol are also being explored in preclinical studies. For example, oral cilostazol (30 mg/kg) improved retinal stress, ischemia, and ganglion cell death in a rat model of diabetic retinopathy⁹⁷. In addition, PDE3A knockout (KO) mice are infertile⁹⁸ and chronic administration of cilostazol blocks pregnancy in naturally-cycling swine⁹⁹, suggesting potential utility of PDE3i’s for birth control or regulating in vivo oocyte maturation in the context of assisted reproduction. Indeed, administration of cilostazol to superovulated mice improved in vitro fertilization rates of subsequently harvested oocytes, possibly by virtue of synchronizing the oocyte maturation¹⁰⁰.

Due to the promise of PDE3 as a therapeutic target, coupled with concerns over side effect associated with cilostazol, alternative PDE3 inhibitors are currently being developed^101–103.

PDE4 inhibitors

The PDE4 family is arguably the most studied of all the PDE families. A number of clinical trials have tested the effect of apremilast for indications beyond psoriasis and arthritis. Two Phase II studies are testing the efficacy of apremilast in combination with phototherapy to produce repigmentation in patients with Vitiligo (Table 3). Interestingly, a recent case report showed apremilast dramatically improved repigmentation in a woman with treatment-resistant Vitiligo¹⁰⁴. Multiple case reports have also described an ability of apremilast to improve symptoms in patients with treatment-resistant erosive oral lichen planus^105–107, perhaps motivating the recently registered Phase II study that will test the ability of apremilast to improve genital erosive lichen planus (Table 3).

Additional indications are also being explored for roflumilast. Phase IV studies showed roflumilast reduced fat mass and, thus, body weight in obese women with polycystic ovary syndrome (PCOS); however, these reductions were smaller than those elicited by liraglutide (Table 3;^{108, 109}). The PDE4 inhibitor TAK-648 is being tested in the clinic in patients with Type 2 diabetes, based on preclinical data¹¹⁰. Roflumilast has also been tested for its ability to improve cognition and information processing in healthy humans, with promising results observed at a dose previously indicated as being devoid of side effects¹¹¹. Patients with stabilized schizophrenia receiving adjuvant roflumilast in a small Phase II trial showed no improvement in working memory but did show some improvement in verbal learning and memory¹¹². Given these positive findings, roflumilast was tested in elderly subjects who demonstrated no change in spatial memory but improved verbal word memory with roflumilast treatment¹¹³. Numerous preclinical studies have supported the therapeutic potential of PDE4i’s in the context of schizophrenia and cognition^114–117.

Cognition-enhancing effects have also been reported for the PDE4i HT-0712, which improved long-term memory for word-lists without serious adverse events in elderly subjects experiencing cognitive decline (http://www.dartneuroscience.com/press_release/july_22_2008.pdf). The cognition-enhancing effect of HT-0712 in humans is consistent with previous reports in mice^{118, 119}. The PDE4D negative allosteric modulator BPN14770 is also being pursued for improving cognitive impairment and has been tested for safety and/or efficacy in healthy elderly subjects, healthy volunteers with scopolamine-induced cognitive impairment, and adult males with Fragile X Syndrome (Table 3). In a press release, BPN14770 was described as having good safety and oral bioavailability and an ability to improve working memory in healthy elderly adults (http://tetradiscovery.com/wp-content/uploads/2016/11/FINAL-Tetra-Phase-1-121616-FINAL.pdf; accessed 05/28/19). These effects in humans are consistent with preclinical studies showing BPN14770 improved a number of behaviors in a mouse model of Fragile-X Syndrome and antagonized the amnestic effects of scopolamine in mice^{120, 121}. Based on preclinical studies showing anxiolytic and cognition-enhancing effects¹²², the PDE4i GSK356278 entered Phase I safety trials for Huntington’s disease but adverse events limited the highest dose to that achieving only ~50% occupancy in brain (Table 3;¹²²). Other nervous system disorders for which preclinical evidence suggests a therapeutic potential of PDE4i’s include ischemic stroke^{118, 123–126}, traumatic brain injury¹²⁷, axon regeneration¹²⁸, and substance abuse disorders (both causes and consequences,^129–132).

McCune-Albright Syndrome is a disease affecting endocrine tissues, skin and bones and is caused by a mutation that results in constitutive activation of the G-protein alpha subunit Gαs (Gαs*). Preclinical studies show that while Gαs* triggers increased cAMP levels in some tissues, it actually results in decreased cAMP levels in other tissues due to a PKA-dependent upregulation of PDE activity, particularly that of PDE1 and PDE4^{115, 133, 134}. Consistent with this upregulation of PDE4 activity, the PDE4i rolipram was able to reverse deficits in Gαs mouse models^{115, 116}. A clinical trial measuring PDE4 expression in the brain and peripheral organs of patients with McCune-Albright Syndrome is currently underway (Table 3).

More recent work is examining PDE4 in the context of cancer. Both preclinical and clinical data suggest roflumilast may exhibit anti-tumor activity for B-cell lymphomas¹³⁵. The PDE4i rolipram, in combination with cAMP-elevating agents, has been shown to suppress triple negative breast cancer both in vitro and in vivo in mice¹³⁶. Apremilast similarly induced tumor regression in mouse models of colorectal cancer¹³⁷. Perhaps even more interesting, specific inhibition of PDE4D, either with genetic tools or the PDE4Di Gebr-7b, resensitized chemotherapy-resistant ER-positive breast cancer cells¹³⁸. These early studies provide promise for the chemotherapeutic potential of PDE4is.

PDE5 inhibitors

A number of recent clinical trials have explored additional disease indications that might benefit from the vasodilatory properties of PDE5i’s. A cream version of sildenafil has recently been tested in a Phase II study examining female sexual arousal disorder (Table 3) as well as a study in which improved blood flow in patients with secondary Raynaud phenomenon was observed¹³⁹. International consortiums are investigating the effects of sildenafil in intrauterine growth restriction, due to anticipated improvements in uteroplacental perfusion^{140, 141}. Initial results suggest sildenafil improves fetal growth and maternal blood pressure across species, including human, sheep, rabbit, and rodents^{142, 143}. Several studies have also explored the effects of sildenafil, tadalafil, or vardenafil in the context of metabolic disorders such as Type 2 diabetes and obesity, assessing glucose tolerance and insulin signaling as well as effects on elevated body weight, nephropathy, and cardiomyopathy (Table 3). Tadalafil improved insulin secretion, endothelial function, and abdominal lean mass content in non-obese men¹⁴⁴, and chronic sildenafil improved glycometabolic control, ameliorated visceral adiposity, and prevented remodeling in diabetic cardiomyopathy^145–148. That said, vardenafil failed to reduce cardiovascular risk in men with type 2 diabetes¹⁴⁹. Interestingly, the positive effects of sildenafil on adiposity and diabetic cardiomyopathy are suggested to be independent of sildenafil’s vasodilatory properties, rather being mediated by epigenetic signaling and/or a reduction of inflammatory chemokines^145–148. It is important to note, however, that one of the studies examining the effect of sildenafil on glucose homeostasis was terminated early due to safety concerns (Table 2). With regard to other indications related to nephropathy and cardiomyopathy, sildenafil has also been tested against media-induced nephropathy, and tadalafil is being explored in the context of kidney stones and endocrine cardiomyopathy (Table 3). A meta-analysis of older clinical studies suggest PDE5i’s could be an effective medical expulsive therapy for distal ureteral calculi, albeit not significantly improved relative to tamsulosin¹⁵⁰. With regard to the brain, two early-stage clinical trials are testing the ability of sildenafil to reverse concussion-related reductions in cerebrovascular reactivity (Table 3).

As described for PDE4i’s, a number of trials are exploring the therapeutic potential of PDE5i’s as chemopreventives for solid tumors, multiple myeloma, and head and neck squamous cell carcinoma. Early reports suggest combining sildenafil with the chemotherapeutic regorafenib is safe in patients with solid tumors¹⁵¹. Further, a number of in vitro and animal models of colorectal cancer suggest that PDE5i’s, either alone or as part of a multi-chemotherapeutic regimen, demonstrate an ability to prevent tumor growth (e.g.,^{17, 151, 152}). Similarly, reports suggest tadalafil promotes tumor immunity in patients with head and neck squamous cell carcinoma (Table 3,^{153, 154}). However, particularly with regard to colorectal cancer, PDE5is do not produce complete anti-tumorigenic effects¹⁶. This lack of complete efficacy may be related to the fact that membrane GCs are inhibited in colorectal cancer¹⁶, and PDE5 may be primarily regulating cytosolic rather than membrane GCs^{18, 19}. Alternatively, the overexpression of both PDE5 and PDE10A in colorectal cancer cells—the latter a membrane-enriched PDE^{69, 155}—may be involved¹⁷. Indeed, PDE10i’s also inhibit growth of colorectal cancer cells^{156, 157}; however, when both PDE5 and PDE10 are inhibited, anti-tumor efficacy is improved in preclinical models¹⁷. Although enthusiasm for PDE5i’s as chemopreventives is growing¹⁵⁸, it should be noted that PDE5i’s similarly prevented prostate carcinogenesis in preclinical models but did not appear to reduce risk or recurrence in clinical studies¹⁵⁹. Perhaps even more concerning, PDE5A appears to suppress melanoma cell invasion in mice¹⁶⁰ yet a recent systematic review and meta-analysis showed that PDE5i’s actually increase risk for melanoma and basal cell carcinoma in humans¹⁶¹.

PDE9 inhibitors

Although PDE9i’s thus far have failed in the clinic for brain diseases, they may hold therapeutic potential for cardiovascular diseases. In a mouse model of sickle cell disease, the PDE9i BAY73–6691 exerted immediate benefits on acute vaso-occlusive events¹⁶², and a Phase 1 clinical trial looking at safety, tolerability and PK/PD of the PDE9i PF-04447943 in patients with sickle cell anemia has recently been completed (Table 3). PDE9i’s may also hold therapeutic potential for cardiovascular indications as PDE9A expression is upregulated by cardiac hypertrophy and cardiac failure. Indeed, the PDE9i PF-04449613 reverses heart disease in animal models by controlling pools of cGMP downstream of pGCs¹⁸.

PDE10i inhibitors

Despite the PDE10i clinical failures described above, TAK-063 was tested in healthy controls and patients with schizophrenia. In healthy controls, TAK-063 was reported to be safe and well tolerated¹⁶³, altering the effects of ketamine on brain activity in healthy controls, particularly in the striatum, substantia nigra, and ventrolateral prefrontal cortex (https://clinicaltrials.gov/ct2/show/results/NCT01892189?sect=X70156#outcome1, accessed 05/28/19). In patients with schizophrenia, although TAK-063 failed to demonstrate a significant effect on the total PANSS score, there was a trend that mirrored effects sizes normally seen with risperidone¹⁶⁴. Furthermore, TAK-063 did significantly improve a number of secondary endpoints relative to placebo¹⁶⁴. It is not entirely clear why TAK-063 was able to succeed where PF-2545920 failed. While one study suggested that TAK-063 activates the striatal direct and indirect pathways in a balanced manner and PF-254920 activates the direct pathway more so than TAK-063¹⁶⁵, other studies have reported that PF-254920 activates these pathways equally^{64, 166}. It is notable that TAK-063—but not PF-254920—increased sensorimotor gating in rodents as measured by prepulse inhibition of acoustic startle (PPI)¹⁶⁵, suggesting PPI may more accurately predict antipsychotic-like effects of novel compounds. Preclinical studies are also exploring the therapeutic potential of PDE10i’s in the context of L-DOPA-induced dyskinesias¹⁶⁷ and alcohol abuse disorders¹³⁰.

Inhibition of PDE7, PDE8 or PDE11

Studies describing the physiological function of the PDE7, PDE8, and PDE11 families are now emerging; however, inhibitors have not yet reached the clinic. Like many of the PDE families discussed above, early research suggests that PDE7i’s and PDE8i’s may have positive effects in diseases where cognition, neuroprotection, neuroinflammation, and/or motor function are impaired (e.g., multiple sclerosis and/or Parkinson’s disease;^168–172). Similarly, PDE11i’s may hold potential for treating age-related cognitive decline^{70, 173} or as an adjunctive treatment to improve lithium responsiveness in patients with bipolar disorder^{174, 175}. PDE7i’s may also hold promise in treating leukemia^{176, 177}, and PDE8i’s may have potential for treating disorders associated with reduced androgen production in males as PDE8i’s, particularly when applied in combination with PDE4i’s, stimulate Leydig cell steroidogenesis^{178, 179}.

Therapeutic strategies beyond inhibition

PDE Activation

There are several disease states where PDE activation may be warranted. Tissue-, brain region-, and subcellular domain-specific decreases in PDE expression/activity and/or increases in cyclic nucleotide signaling have been implicated in select disease states, including some age-related deficits^{180, 181}, Huntington’s disease¹⁸², social isolation¹⁸³, migrane^184–188, retinitis pigmentosa¹⁸⁹, infertility⁹⁸, prostate cancer¹⁹⁰, melanoma and basal cell carcinoma¹⁶¹, cardiac hypertrophy^{24, 191}, acrodysostosis²⁵, and polycystic kidney disease¹⁹². PDE activators would be expected to have a greater impact in cells with higher cyclic nucleotide levels (either basal or stimulated) as opposed to cells with low cyclic nucleotide levels, although this remains to be empiracly established. Indeed, Mironid have developed PDE4 longform-specific activators (mechanism as yet unknown; Table 4) for the treatment of polycystic kidney disease where increased adenylate cyclase activity caused by overexpression of vasopressin V2R receptors results in elevated cAMP levels that drive cyst growth and disease progression¹⁹³. There are several natural mechanisms by which PDE activity can be activated (Figure 3), and it is our contention that these avenues could be manipulated phamacologically to trigger PDE activation.

Table 4.

Selected patents involving PDEs published in the last 5 years.

Patent # (Priority date d/m/y)	Subject	Assignee	Author
PDE3
US2019046528 (08/08/2017)	Method of preventing hair loss or promoting hair growth by using PDE3 inhibitor	Seoul Nat Univ Hospital	Kwon O, Choi HI, Jo SJ, Kim KH
WO2017186103 (26/04/2016)	Applications of PDE3A in determination of tumor treatment effect of Anagrelide	Shanghai Inst Materia Medica Cas	Yu Q, Liu J
PDE4
CN108904493A (12/08/2018)	PDE4 inhibitor and purpose for preparing novel anti-inflammatory drugs	Hu Y	Hu Y
WO2018167142 (16/03/2017)	Treatment of idiopathic pulmonary fibrosis [with a PDE4 inhibitor]	Takeda GMBH	Hanauer G, Nikam S, Hazama M
WO2017133713 (05/02/2016)	Application of PDE4 inhibitor ZL-N-91 in preparation of medications for lung cancer proliferation and metastasis	Guangzhou Sinogen Biomedical Tech Ltd	Zhao AZ, Gong S, Lin Y, Li F, Li X
WO2018060704 (28/09/2016)	Compounds and their use as PDE4 activators for the treatment of disorders requiring a reduction of cAMP	Mironid Ltd	Adam JM, Adams DR
WO2018037109 (26/08/2016)	Treatment of nonalcoholic fatty liver disease with PDE4 inhibitors	Takeda GMBH	Hanauer G, Nagabukuro H, Amano Y
CN107412214A (31/07/2017)	Application of PDE4 Inhibitor (FCPR16) for the treatment of PD	Guangzhou Lanssonpharm Jianzhi Tech Co Ltd	Xu L
US2017051291 (28/12/2005)	RNAi-mediated inhibition of PDE4 for treatment of cAMP-related ocular disorders	Arrowhead Pharmaceuticals Inc	Yanni JM, Chatterton JE, Gamache DA, Miller ST
WO2017017165 (29/07/2015)	PDE4 inhibitor for the treatment of diabetic nephropathy	Takeda GMBH	Hanauer G, Vollert S, Hazama M, Matsuo T
WO2016075543 (13/11/2014)	Treatment of multiple sclerosis with the combination of laquinimod and a PDE4 inhibitor	Teva Pharma, Piryatinsky V, Kaye J	Piryatinsky V, Kaye J
WO2015022417 (16/08/2013)	Treatment of cognitive impairment with the combination of a PDE4 inhibitor and an acetylcholinesterase inhibitor	Univ Maastricht	Yamada T, Prickaerts J, Van Duinen M, Sambeth A, Blokland A
PDE5
US2018221373 (16/09/2015)	Method of treating insomnia with PDE5 inhibitors	Rosenberg LI	Rosenberg LI
CA2975049 (10/08/2016)	PDE inhibitors (sildenafil) to repair brain and/or retinal injury in human newborns	Wintermark P	Wintermark P
CN107163052A (18/04/2017)	Immunodetection method for various PDE5 inhibitor drugs	Univ South China Agricult	Shen Y, Hua Y, Xu Z, Yang J, Wang H, Sun Y, Lei H
WO2014088326 (04/12/2012)	Composition comprising PDE5 inhibitor for inhibiting apoptosis of nerve cells	Aribio Inc, Sk Chemicals Co Ltd	Kim MH, Choung JJ, Ku SK
PDE6
JP2019047763A (12/09/2017)	Rhodopsin PDE as an optogenetic tool for light control of intracellular cyclic nucleotides	Nagoya Institute Of Technology	Kandori H, Tsunoda SP, Yoshida K
CN107287239 (11/04/2016)	Gene therapy vector and medicine (adenovirus encoding PDE6B) for retinal pigment degeneration	Shenyang Fuming Biological Tech Co Ltd	Pang J
PDE7
WO2018055140 (23/09/2016)	T cells with increased immunosuppression resistance [expressing PDE4C or 7A for the treatment of cancer]	Adaptimmune Ltd	Laugel B, Skibbe K
PDE9
WO2017070293 (20/10/2015)	PDE9 inhibitor and levodopa therapy for treating PD or Parkinsonism	Ironwood Pharmaceuticals Inc	Leventhal L, Townsend TM
PDE10
WO2019067955 (29/07/2017)	Compositions and methods for regulating let-7 microRNA targets, such as PDE10A, for treatment of cancers	Univ California	Roos M, Lowry W

Source: Espacenet and Google patents (accessed 05/28/19)

Figure 3. Mechanisms that activate phosphodiesterase (PDE) catalytic activity.

A) Calcium-calmodulin (CaM) binding to the CaM domains of PDE1 relieves N-terminal auto-inhibition of the catalytic site, thereby promoting enzymatic activity³²⁰. B) Cyclic nucleotides binding to GAF domains of dimeric PDEs (shown here: cGMP binding the GAF-B domain of PDE2) are thought to promote catalytic activity by inducing an outward rotation of the catalytic domains and, thus, enabling access to substrates³²¹. C) Phosphorylation by PKA or PKG activates several PDEs¹⁹⁶. In the case of PDE4D, phosphorylation of the UCR1 domain by PKA causes UCR1 to bind its own UCR2 domain instead of the catalytic site of the other monomer, thereby locking the enzyme in an active state. D) PDE activity can also be modulated by protein-protein binding interactions. One such well-characterized example involves membrane-bound PDE6, where the rhodopsin-activated G-protein α-subunit transducin displaces the inhibitory PDE6γ C-termini from the catalytic sites on PDE6αβ, thus, promoting cGMP hydrolysis³²².

Targeting GAF domains

One route to PDE activation is by way of tandem GAF (cGMP-specific and stimulated PDE, Anabaena adenylyl cyclases, and E. Coli FhlA) domains¹⁹⁴ (Figure 3B). Although GAF domains have been identified in over 7400 proteins, in mammals they are only are found in the PDE families 2, 5, 6, 10 and 11^{195, 196}. For a vast majority of non-PDE GAF domains the activating ligand is unknown, however for PDEs it is known that cyclic nucleotides bind to these pockets (Figure 3). PDE2 and PDE5 are activated when cGMP binds the GAF domain^197–199, and PDE10 is activated by cAMP binding the GAF domain²⁰⁰. In the context of activation, binding of cyclic nucleotides to GAF domains is thought to cause structural changes that relieve autoinhibition of the PDEs (Figure 3). In contrast, cGMP binding the GAF-A domain of PDE6 enhances protein-protein interactions that inhibit PDE6 catalytic activity²⁰¹. This suggests that blocking cGMP binding of the PDE6 GAF domain may provide a means of promoting PDE6 activity. It also suggests it may be possible to both activate and inhibit GAF-containing PDEs with small molecules at a site distinct from the catalytic domain. Indeed, PDE11A is activated when a cGMP analog—but not cGMP itself—binds the GAF domain²⁰⁰. Further, even though cGMP binding of the GAF domain activates PDE5¹⁹⁷, a number of other types of molecules that bind the GAF domain inhibit PDE5 in its activated—but not basal—state²⁰². This is consistent with the fact that GAF domains are known to bind a diverse array of small molecules that are unrelated to cyclic nucleotides¹⁹⁴. The fact that GAF domains are only found in PDEs in mammals¹⁹⁶ makes GAF domains of high interest in the context of drug targeting²⁰³. Importantly, mammalian GAF domains are sufficiently structurally divergent from one another (e.g., low degree of homology between PDE families and the tandem GAF domains are preceded by variable N-terminal stretches) as to allow selective pharmacological targeting of individual PDE families¹⁹⁷. Together, this suggests the GAF domains may provide an inroad for targeting reagents that selectively activate a given PDE isoform while avoiding off-target activity.

Preventing trans-capping

PKA or PKG phosphorylation of PDE3^{204, 205}, PDE4^{204, 206}, PDE5²⁰⁷, and PDE8²⁰⁸ is also known to activate catalytic activity in a negative feedback loop. In the case of PDE4, for example, catalytic activity is inhibited when the UCR2 regulatory domain “trans-caps” the catalytic site; thus, occluding cAMP from reaching the enzymatic core of PDE4^{72, 209} (Figure 3C). PKA phosphorylation of the UCR1 regulatory domain blocks the ability of UCR2 to trans-cap the catalytic site, which locks PDE4 in the active state⁷². Notably, select PDE4Di’s allosterically inhibit catalytic activity by promoting “trans-capping”¹²⁷; whereas, phosphatidic acid activates PDE activity by inhibiting trans-capping in a similar but mutually exclusive manner to PKA^210–212. Furthermore, the dominant negative peptide “UCR1C”, which corresponds to UCR1 sequence, also activates PDE4 activity by inhibiting trans-capping²¹³. These results provide proof of principle that activation of PDE4 may be achieved by either small molecules or biologics that prevent UCR2 from adopting a trans-capping conformation.

Manipulating protein-protein binding

PDEs may also be activated by manipulating protein-protein binding interactions. PDE6 is unique in the fact that the heterodimeric holo-enzyme includes two inhibitory subunits that span the catalytic pockets of the dimer, thus occluding cGMP from the catalytic site²¹⁴ (Figure 3D). Binding of the GTP-bound α-subunit of the heterotrimeric G-protein transducin relieves PDE6 inhibition by binding to the C-terminal region of PDE6 and its inhibitory subunits²¹⁵. The full crystal structure of PDE6 is not yet available²¹⁶; however, recent cryo-EM work²¹⁷ has confirmed the predicted structural organisation of the holo-enzyme, albeit without sufficient detail to inform pharmacological targeting. The success in upregulating PDE6 activity via gene transfer to combat retinitis pigmentosa¹⁸⁹ (see following section) suggests that PDE6 activation could be a viable therapeutic strategy for the treatment of vision loss. As discussed in greater detail below, it may also be possible to increase PDE activity by preventing the binding of PDEs to binding partners that sequester or suppress activity.

Gene therapy

Viral transfer of PDE genes, agents that silence PDE gene expression (e.g. antisense, silencing or microRNAs)^{160, 218–222}, or gene editing (e.g., Crispr/Cas9)²²³ might also prove a useful means of therapeutically targeting individual PDE isoforms (Figure 4). The best characterized PDE gene therapy approach to date targets PDE6 activity in the retina. A loss of transducin-mediated activation of PDE6 results in elevated cGMP levels, which causes the loss of primary rods and, ultimately, vision²²⁴. Expression of recombinant PDE6α in the retina via an adeno-associated virus (AAV-PDE6α) preserved retinal structure, photo-transduction, and vision in retinal degeneration (rd) mice, as did AAV-PDE6β^{225, 226}. AAV-PDE6α similarly rescued retinal deficits in a mouse model that mimics human retinitis pigmentosa mutations²²⁷. Experiments injecting AAV-PDE6γ into the retina have also proven successful in mice²²⁸. In dogs, delivery of recombinant PDE6α using a tyrosine capsid-mutant AAV8 was able to stabilize cGMP levels and improve survival of photoreceptor rods and cones in PDE6α-mutant dogs; however, several adverse effects related to the AAV injection were identified¹⁸⁹. The recent development of synthetic AAV vectors that target the retina in non-human primates may provide the answer to these problems in the future²²⁹. Notably, two clinical trials are underway testing the safety and efficacy of PDE6 gene therapy in retinitis pigmentosa (PhI , PhII ; clinicaltrials.gov accessed 5/29/19).

Figure 4. Methods for targeting phosphodiesterase signaling with increasing specificity.

Given the vast diversity of PDE isoforms, each with unique tissue expression profiles, subcellular compartmentalization, and protein-protein interactions, it is becoming clear that selective targeting of PDE function will be required to achieve efficacy while diminishing undesirable side effects. Small molecule inhibitors (e.g., cilomilast) are readily developed with family-specific selectivity (e.g., targeting PDE4 over PDE3); however, isoform specificity remains a challenge (e.g., cilomilast inhibits PDE4D with only 7-fold selectivity versus PDE4B)³²³. Conversely, gene therapy (i.e., expressing a recombinant construct to knock down or restore expression of a given PDE isoform) and dominant negative approaches (i.e., expressing a catalytically inactive PDE4D5 that displaces the endogenous isoform from its interacting partners) can target isoform subtypes exclusively (e.g., targeting PDE4D5 but not PDE4D3 nor PDE4B). That said, dominant negative approaches would influence signalling within all microdomains regulated by that isoform. The greatest specificity can be achieved with peptide/small molecule binding disruptors or mutagenesis approaches (not shown) that are designed to prevent a specific PDE isoform from binding a specific partner, thus, altering signaling only within one specific complex. As shown here, a disruptor peptide that specifically prevents the interaction between PDE4D5 and β-arrestin would lead to the recruitment of EPAC1 to β2 adrenergic receptors (βAR), but would leave PDE4D5 regulation of heat shock protein 20 (HSP20) and RACK1 complexes intact^{236, 324, 325}.

A rapidly evolving approach within the gene therapy field is optogenetic medicine, which combines viral delivery of recombinant, light-activated proteins with biomedical devices that emit light of the specific intensity and wavelength needed to activate those proteins²³⁰. With the field of personalized bioelectronic implants quickly evolving, optogenetic-based biomedical approaches are being pursued for neurological diseases, cancer, cardiovascular disease and metabolic disorders²³⁰. Given that optogenetic-based approaches have now entered clinical trials²³⁰, it is worth noting here that light-activated PDEs have been identified in lower organisms^231–233 and engineered in the lab²³⁴. Both are being explored as biological tools in higher organisms. Activating or inhibiting a given PDE by a spatially- and temporally-restricted light emission, as opposed to a systemically administered pharmacological agent, may prove an ideal approach for treating diseases where cyclic nucleotide signaling is down regulated in one tissue yet upregulated in another (e.g., aging; c.f.,²³⁵). It might also provide a means of avoiding side-effects associated with targeting PDE activity in a specific tissue (e.g., nausea/emesis associated with inhibiting PDE4 in the area postrema).

Targeting location

As production of cAMP is utilized by a variety of different Gs-coupled receptors to transduce signals, compartmentalization of signaling intermediates is crucial to define physiological outcomes specific to each receptor²³⁶. This compartmentalization of cyclic nucleotide signaling is achieved by virtue of PDEs being tethered to a precise cellular location via binding partners (Table S1). Thus, promoting or disrupting isoform-specific protein-protein interactions may prove a viable approach to therapeutically target PDEs in an isoform-specific manner, a level of specificity that has not been achieved with pharmacological inhibitors to date (Figure 4).

Dominant negative PDEs

Proof of principle for such an approach first emerged with studies using dominant negative (DN) PDEs, catalytically inactive mutants that would displace their endogenous PDE. Using specific DN-PDE4 isoforms, in vitro studies have successfully altered perinuclear cAMP signaling²³⁷, β-arrestin-dependent desensitization of the beta2-adrenergic receptor^{238, 239}, growth control of prostate cancer cells¹⁹⁰, prostanoid receptor-mediated cAMP signaling²⁴⁰, glucagon-like peptide-1 release²⁴¹, and cAMP gradients at the centrosome²⁴². DN-PDE4 tools have also yielded beneficial effects in vivo. For example, viral delivery of a DN-PDE4A5 to the mouse hippocampus was able to rescue localized cAMP signaling deficits and hippocampus-dependent memory impairments that were caused by sleep deprivation^243–245. In contrast, overexpression of a DN-PDE4B1 in the forebrain of mice did not affect hippocampus-dependent memory, although it did enhance hippocampal long-term potentiation in male mice²⁴⁶. This finding underscores the importance of understanding the role of specific PDE isoforms, because a homozygous mutation in PDE4B (Y358C) that greatly reduces activity of all PDE4B isoforms by virtue of attenuating interactions with the scaffold protein Disrupted In Schizophrenia 1 (DISC1) improves both long-term potentiation and memory as well as other mood-related behaviors²⁴⁷. It is interesting to note that nature has developed its own dominant-negative approach with PDE4A7, a PDE isoform that is targeted to specific subcellular compartments but is catalytically dead²⁴⁸.

One point to consider in adopting a DN approach is the fact that a single PDE isoform can contribute to more than one function in a cell via its participation in multiple distinct signaling complexes, which involve mutually exclusive protein-protein interactions²⁴⁹ (Table S1; Figure 4). For instance, PDE4D5 is involved in a number of processes common in almost all cells, such as cell growth, cell orientation, desensitization of Gs-coupled receptors, and inactivation of the phosphorylation of the ubiquitous chaperone HSP20²⁵⁰. The ability for PDE4D5 to have all these functions in a cell is a result of it being localized in different compartments by different anchors (e.g., RACK1 at leading edge of cells, beta-arrestin at transmembrane receptors, and HSP20 in the cytosol; Figure 4;²⁵⁰). It is clear that this is also the case for a number of other isoforms based on proven protein-protein binding interactions (e.g., in heart tissue/cells, PDE4D3 can bind to either the ryanodine receptor, HSP20, or an AKAP9/Potassium channel complex—see Table S1) or based on inference from the fact that the exact same isoform can be found localized to multiple subcellular domains (e.g., ~50% of PDE11A4 in neurons localizes to the cytosol while 25% is localized to the nuclear fraction and another 25% to the membrane compartment⁶⁹). Thus, a non-selective DN approach has the potential to influence multiple domains within the cell. To achieve a compartment-specific manipulation of a given PDE isoform, one could mutate the binding site(s) that mediates a particular protein-protein interaction. Mutating isoform-specific binding sites has also proven a useful approach, revealing that an integrin α5-PDE4A5 complex regulates endothelial inflammation²⁵¹, a PDE3A1-SERCA complex regulates myocardium contractility²⁵², and a DISC1-mediated sequestering of PDE4B regulates hippocampal function²⁴⁷. An alternative approach is to develop a peptide or small molecule that specifically competes for a given protein-protein binding site²⁵³. This approach would displace only a specific “pool” of a given PDE isoform, while leaving the vast remainder unfettered in their respective signaling complexes (Figure 4). Efforts have begun to identify small molecules capable of promoting or interfering with protein-protein interactions, but have not yet been published so it is too early to speculate on the required design characteristics at this stage.

Disrupting protein-protein interactions with peptides

A recent review suggests cell-permeable, peptide disruptors effectively manipulate specific PDE isoforms in a compartment-specific manner²⁵³ and evidence continues to build. More recently, a PDE4D-FAK disrupting peptide prevented direction sensing and invasion of melanoma cells^{254, 255} and a PDE8A-Raf1 disruptor retarded cancer cell growth promoted by a Ras mutation²⁰⁸. Interestingly, the same PDE8A-Raf1 peptide has also been used to target T cell adhesion and migration and was more potent than a PDE8-specific inhibitor in reducing inflammatory signaling²⁵⁴. The effectiveness of PDE displacement has also been demonstrated in vivo, where intraperitoneal injections of a PDE4-HSP20 disruptor significantly attenuated hypertrophy-induced cardiac remodeling²⁵⁶.

Disrupting PDE homodimerization

Disrupting PDE homodimerization (that is, a PDE monomer binding to itself) may also prove an effective way to target PDE function in a domain-specific manner. Disrupting PDE11A4 homodimerization using a peptide recognizing its GAF-B domain was shown to selectively remove PDE11A4 from membrane-bound complexes but not the cytosol, which may hold utility for improving responsivity to the mood stabilizer lithium¹⁷⁴ or age-related cognitive decline^{70, 173}. Conversely, a peptide or mutation that could stabilize PDE11A4 homodimerization might prove useful in treating the deleterious psychological effects of social isolation¹⁸³. Targeting homodimerization of PDE2, PDE4, or PDE5 may also relocate the enzymes by virtue of changing susceptibility to regulatory post-translational modifications^{196, 257}. Indeed, nature appears to have taken advantage of dimerization as a mechanism to regulate PDE trafficking. For example, when PDE10A2 heterodimerizes with PDE10A19, PDE10A2 is prevented from trafficking to the membrane as it normally does under conditions of homodimerization²⁵⁸. Such complex-specific targeting of PDE function may be required to achieve efficacy in absence of unwanted side effects, particularly in cases where multiple subfamily isoforms orchestrate a variety of physiologic responses by virtue of different protein-protein interactions (e.g., targeting various PDE3 isoforms for cardiovascular disease²⁵⁹).

Targeting post-translational modifications

As post-translational modifications (PTM) directly regulate PDE activity and location, PTMs could be considered a point of therapeutic control (Figure 5).

Figure 5. Phosphodiesterase (PDE) regulation by post-translation modification (PTM).

Cyclic nucleotide dynamics can be modulated by the addition of different functional groups to PDEs. Phosphorylation is a very common mechanism to control PDE activity as depicted by the action of PKA on PDE4D3. Both enzymatic activity and binding affinity of PDE4D3 for mAKAP are increased by PKA phosphorylation, allowing a faster signal termination in myocytes²⁶³. Palmitoylation of PDE10A2 in its N-terminal region translocates the enzyme to the plasma membrane, although its phosphorylation by PKA can prevent the action of the palmitoyl acyltransferase (zDHHC)¹⁵⁵. Ubiquitination can influence PDE function by controling stability. For example, the E3 ubiquitin ligase Smurf2 targets PDE4B for degradation which leads to the attenuation of liver fibrosis²⁶⁷. S-nitrosylation can also tag PDEs for destruction. Thus, the covalent incorporation of nitric oxide (NO) to the GAF-A domain of a PKG-phosphorylated and active PDE5, directs the enzyme to the proteasome²⁷². Hydroxylation of proline residues has emerged as another PTM to stimulate turnover of PDEs. Prolyl hydroxylase domain protein 2 (PHD2) action on PDE4D increases its recognition by E3 ligase complexes in cardiomyocytes²⁷⁴. Finally, SUMOylation can intensify the activity of PDE4A and PDE4D. The SUMO transfer from the E2 conjugase UBC9–E3 enzyme PIASy complex to the PDEs, enhances their activation by PKA phosphorylation and represses their inhibition induced by ERK activity²⁶⁹.

Phosphorylation

PKA or PKG phosphorylation of PDE3^{204, 205}, PDE4^{204, 206}, PDE5²⁰⁷, and PDE8²⁰⁸ will stimulate catalytic activity. Phosphorylation can also influence PDE cellular location by virtue of preventing other PTMs that promote membrane association (e.g., palmitoylation) or changing protein-protein binding interactions. A great example to illustrate these principals is PDE10A2. PDE10A2 is palmitoylated in its N-terminal region, which directs membrane targeting and trafficking to dendrites¹⁵⁵. If, however, PDE10A2 undergoes isoform-specific phosphorylation by PKA on Thr16²⁶⁰, palmitoylation of PDE10A2 is attenuated and the specific membrane localization is lost¹⁵⁵. Interestingly, phosphorylation on the same site also interferes with the scaffolding of PDE10A2 by AKAP150²⁶¹. Hence, although PDE10A catalytic activity is not directly affected by this PTM, cyclic nucleotide levels should increase within this compartment due to the absence of PDE10A2. Preliminary evidence also suggests that PDE11A4 can similarly be shuttled between membrane and cytosolic compartments by virtue of phosphorylation of N-terminal serines, although this is likely by virtue of altering protein-protein interactions as opposed to a direct insertion into the membrane²⁶². PKA phosphorylation of PDE4D3 drives an association with the mAKAP signaling complex to evoke rapid signal termination in the muscle compartment²⁶³, which may have therapeutic implications given that polymorphisms in the PDE4D3-mAKAP binding site lead to a higher susceptibility to cardiovascular disease²⁶⁴. PDE4D enzymes also get phosphorylated by both casein kinase 1 (CK1) and glycogen synthase kinase 3β (GSK3β) in the catalytic region on a motif known as a “phosphodegron”²⁶⁵. This action increases the affinity of the PDE for a ubiquitin ligase complex (Cullin 1 containing SCF E3 ligase) which promotes proteosomal degradation of the enzyme²⁶⁵. Hence PDE phosphorylation not only affects activity, localization, and protein-protein interactions, it also regulates protein turnover.

Ubiquitination and sumoylation

It has been known for some time that increases in PKA activity promote the proteosomal degradation of short-lived proteins, an action that can be enhanced by PDE inhibitors (c.f.,²⁶⁶). However, we are just starting to understand that the stability of PDEs can themselves be influenced by the ubiquitin-proteosome system (Figure 5). Ubiquitin conjugation is known to target proteins for degradation by the proteasome and specificity is introduced at the level of E3 ligase-substrate interaction. We now know of multiple instances where PDEs interact definitively with one of the over 600 E3 ligase types to dramatically shorten PDE half-life and this could be a new point at which to direct innovative therapeutics. As mentioned above, PDE4Ds can be degraded by virtue of an SCF E3²⁶⁵, whereas PDE4B levels can be down-regulated by a different E3, Smurf2, to promote anti-fibrotic signaling in the liver²⁶⁷. PDE4D5 can also be targeted for ubiquitin modification by the RING type MDM2 E3 ligase; however, this beta-adrenergic driven ubiquitination of PDE4D does not signal a degradation of the enzyme. Instead, it shifts PDE4 from binding RACK1 to binding β-arrestin²⁶⁸. PDE4s are similarly regulated by the ubiquitin-like protein SUMO (small ubiquitin-like modifier)²⁶⁹. SUMO-conjugation tends to alter the location, activity or protein-protein interactions of a protein rather than tagging for destruction via the ubiquitin-proteosome system²⁷⁰. Unlike ubiquitination, sites of SUMOylation can be predicted in the amino acid sequence of putative substrates as conjugation usually occurs on a lysine residue within the consensus h-K-X-D/E (where h is a bulky hydrophobic and X is any residue)²⁷⁰. PDE4s from subfamilies A and D contain the consensus motif, whereas subfamilies B and C do not. This adds an extra layer of sub-family-specific regulation as SUMOylation serves both to protect against the inhibitory phosphorylation by ERK/MAP kinases²⁷¹ and further enhances activity of the PKA phosphorylated longform PDE4 by locking it in the “open” non-UCR inhibited conformation²⁶⁹.

S-nitrosylation and proline hydroxylation

Two additional PTMs that trigger PDEs for destruction include S-nitrosylation and proline hydroxylation (Figure 5). PDE5 can be S-nitrosylated by NO on Cys220²⁷², which targets the enzyme to the proteasome and reduces PDE activity. Under conditions of reduced NO bioavailability, as in heart disease, PDE5 is upregulated due to a loss of this S-nitrosylation-induced degradation²⁷². Proline hydroxylation has also been identified as a modification that can tag substrates for recognition by E3 ligase complexes²⁷³. In the heart, proline hydroxylases domain-containing proteins (PHDs) hydroxylate surface-associated prolines on PDE4D enzymes, triggering their degradation²⁷⁴. In this way, direct binding of PHDs to PDE4s increase cAMP without affecting adenylate cyclase activity.

Challenges

Although this is clearly an exciting time in the PDE field, there is much work that remains to be done. For therapeutics to be efficiently developed, we need to more thoroughly understand precisely where cyclic nucleotide signaling is disrupted in a given disease—in which tissue, cell types, and subcellular compartments. We then need to target a PDE in a defined locale, with the understanding that subcellular compartmentalization of a given PDE may vary depending on species, age, tissue type, or disease status (e.g.,^5–12; for full review, see Table S1). This consideration is equally important in the evaluation of potential efficacy and potential side effects. To maximize potential efficacy while minimizing potential side effects, one would target a PDE that is enriched, if not exclusively expressed, in the tissue of interest and that controls the same pool of cyclic nucleotides that is altered by the disease. At the same time, efforts to unravel the intramolecular signals responsible for trafficking each PDE also need to continue to inform more sophisticated therapeutic approaches that can preferentially target a given PDE in a given subcellular compartment. Along these same lines, we need to grow our understanding of how to stimulate PDE activity and how to target the PDE catalytic activity of dual-specificity PDEs in a functionally-selective manner (i.e., target only its cAMP- or cGMP-hydrolytic activity, see²⁰³ for further discussion). Perhaps by increasing the specificity of our approach, we can retain efficacy while mitigating the numerous side effects described above that have plagued PDE inhibitors to date.

An additional challenge is to gain a better understanding of which physiological/disease processes are governed by PDE regulation of canonical versus non-canonical cyclic nucleotides. Research into the role of non-canonical cyclic nucleotides is rapidly evolving as new techniques and reagents facilitate functional studies⁴. Cyclic cytidine monophosphate (cCMP) and cyclic uridine monophosphate (cUMP) are synthesized by soluble guanylate and soluble adenylate cyclases in mammalian systems, although an as-yet-to-be identified generator likely accounts for the majority of production given the dissociation between sGC/sAC and cCMP/cUMP expression patterns⁴. ExoY, a bacterial nucleotidyl cyclase, is known to generate cUMP in non-mammalian systems². Hydrolysis of cCMP is catalyzed by PDE7A whereas cUMP is broken down by PDE3A, PDE3B and PDE9A². Functionally both nucleotides have been shown to activate PKA/PKG²⁷⁵, cyclic nucleotide gated channels²⁷⁶, and cCMP is described as a partial agonist of EPAC²⁷⁷. In a disease context, the non-canonical cyclic nucleotides play roles in promoting virulence of pseudomonas aeruginosa infections²⁷⁸ and triggering apoptosis of cancer cells²⁷⁹; however much more research is needed to accurately characterize the pathophysiology involving these signaling molecules. Difficult issues facing the field will be defining specific non-canonical cyclic nucleotide signaling systems that are aberrantly regulated in disease, determining the mechanisms by which PDEs might preferentially degrade non-canonical versus canonical cyclic nucleotides, and visualising the compartmentalisation of non-canonical signalosomes in specific locations within cells and tissues.

Finally, development of PDE-targeted therapeutics faces the same challenges as does all drug discovery—namely the high rate of failure in clinical trials. A recent study suggests that from 2000–2015, only 13.8% of all compounds made it from Phase I to approval²⁸⁰. When considering success rates for individual indications, we may gain insight into the likelihood that a PDE-targeted therapeutic will achieve success within a given disease area. For example, oncology saw only 3.4% of compounds made it from Phase 1 to approval, perhaps not surprising given the resistance and heterogeneity that dogs this area. In contrast, metabolic/endocrinology, cardiovascular, CNS, autoimmune/inflammation, genitourinary, and ophthalmology saw a success rate of 19.6%, 25.5%, 15%, 15.1%, 21.6%, and 32.6%, respectively²⁸⁰. This certainly paints a grim picture for pursuing any type of therapeutic in the context of cancer; however, this failure rate also underscores the desperate need to develop novel therapeutic options. It is then important to note that success rates were doubled for cancer compounds when patient selection biomarkers were employed in the trials; the success rate for cardiovascular compounds similarly benefited²⁸⁰. Importantly, expression or activity of PDEs themselves may prove viable biomarkers in this context (Box 2).

Outlook

Although the challenge of targeting localized pools of PDEs for purposes of correcting pockets of aberrant cyclic nucleotide signaling has proven difficult in the past, there are indications that innovative approaches and technological advances are making headway (see Table 4 for recent patent activity that includes PDE activators, biomarkers, and viral approaches). For example, agents that show remarkable selectivity for sub-families of PDE4 are showing promising results in mouse models of learning and memory and translation to human disease would be a game-changing advance¹²¹. Additionally, novel delivery systems are being developed that can transport PDE inhibitors to precise tissues or cell types, thereby abrogating complications associated with systemic distribution (e.g.,^{281, 282}). In the future, novel delivery systems such as these could be employed to deliver agents that specifically disrupt the anchoring of single isoforms. Another potential approach is the intelligent design of a new generation of PDE inhibitors that preferentially accumulate in or segregate from certain tissues or organs. The mild side-effect profile of Apremilast is largely due to its inability to penetrate the brain and engineering of similarly restricted distribution profiles may unlock the latent abilities of other PDE inhibitors. Gene therapy that seeks to abrogate or enhance activity of single PDE isoforms in a cell type-specific manner may provide a way to combat disease or fight complications associated with ageing. There is also the possibility that an improved therapeutic window might be achieved by combining sub-optimal doses of PDEi’s with ineffective doses of downstream-target activators, which would result in an effective combined dose only in tissues and subcellular compartments where both molecules were present (e.g.,²⁸³). Indeed, PKA, PKG, and Epacs have been implicated as therapeutic targets in their own right for a number of indications for which PDEi’s are being pursued, including diseases of the cardiovascular, immune, metabolic and nervous systems as well as cancer (e.g.,^{115, 116, 284–289}). Further, several studies in a variety of tissues have attributed the beneficial effects of PDEi’s to the activation of PKA, PKG and/or Epac (e.g.,^{121, 290–292}). As we learn more about the functional role and molecular interactions of each PDE splice variant, and how the function and/or localization of an individual variant may be altered in a given disease, it will become clearer how we can successfully target PDEs in a specific fashion to achieve efficacy while avoiding side effects. Only with this detailed level of knowledge will we be able to realize the full potential of PDEs as therapeutic targets.