Myocardial Phosphodiesterases and their Role in cGMP Regulation

Abstract

Cyclic nucleotide phosphodiesterases comprise an 11-member superfamily yielding near 100 isoform variants that hydrolyze cAMP or cGMP to their respective 5’-monophosphate form. Each plays a role in compartmentalized cyclic nucleotide signaling, with varying selectivity for each substrate, and conveying cell and intracellular specific localized control. This review focuses on the five PDEs expressed in the cardiac myocyte capable of hydrolyzing cGMP and that have been shown to play a role in cardiac physiological and pathological processes. PDE1, PDE2, and PDE3 catabolize cAMP as well, whereas PDE5 and PDE9 are cGMP selective. PDE3 and PDE5 are already in clinical use, the former for heart failure, and PDE1, PDE9, and PDE5 are all being actively studied for this indication in patients. Research in just the past few years has revealed many novel cardiac influences of each isoform, expanding the therapeutic potential from their selective pharmacological blockade or in some instances, activation. PDE1C inhibition was found to confer cell survival protection and enhance cardiac contractility, while PDE2 inhibition or activation induces beneficial effects in hypertrophied or failing hearts, respectively. PDE3 inhibition is already clinically employed to treat acute decompensated heart failure, though toxicity has precluded its long-term use. However, newer approaches including isoform specific allosteric modulation may change this. Lastly, inhibition of PDE5A and PDE9A counter pathological remodeling of the heart and are both being pursued in clinical trials. Here we discuss recent research advances in each of these PDEs, their impact on the myocardium, and cardiac therapeutic potential.

Introduction

Phosphodiesterases (PDEs) are a superfamily of enzymes that hydrolyze the cyclic nucleotides adenosine 3’,5’-cyclic monophosphate (cAMP) and/or cyclic guanosine 3’,5’-cyclic monophosphate (cGMP). Both cyclic nucleotides are generated within intracellular nanodomains by corresponding cyclases, and are in turn catabolized by members of the PDE superfamily. Both synthesis and catabolism of cAMP and cGMP are altered by physiological and pathological stress, and this plays essential homeostatic roles as well as contributes to heart disease. Therapeutic benefits from direct stimulation of either cardiac cyclic nucleotide, as by beta-adrenergic agonism, organo-nitrates or nitic oxide donors, soluble guanylyl cyclase activators, or natriuretic peptides, are clinically used to trigger associated signaling. One disadvantage of these approaches is their diffuse impact on many cells, such that cardiomyocyte regulation often takes a back seat to changes in blood pressure, heart rate, and other changes.

The alternative to enhancing cyclic nucleotide synthesis is to selectively block their hydrolysis by inhibiting the relevant PDEs. Despite there being only two primary cyclic nucleotides, there are >100 different PDE members/isoforms to modulate them. These differ primarily in their N-terminus regulatory domain¹ to control localization and regulation, with the catalytic domain conferring substrate specificity ^{2, 3}. PDEs are very amenable to family-member selective potent small molecule inhibition, and many such inhibitors are being used or studied as therapeutics (Supplemental Table 1). This selectivity has its limitations, most notably that isoform and splice variants in a given species are equally susceptible as they share common catalytic domains. Another is that the relevant cyclic nucleotide must first be synthesized in order for a particular PDE inhibitor to have an impact. This is not required when this synthesis is itself being stimulated.

There are seven PDEs so far reported to be expressed in myocardium. PDE1, 2, and 3 are dual substrate esterases, PDE5 and PDE9 are selective for cGMP, and PDE4 and PDE8 are selective for cAMP. Preclinical studies support a role for each of these species in the heart, while existing clinical data pertain to PDE3 and PDE5. Each are expressed in myocytes, and many are also expressed in fibroblasts, vascular smooth muscle, and in some cases endothelial cells (see Supplemental Table 2). Importantly, many of these PDEs can contribute to cyclic nucleotide dysregulation in diseased heart, and so have become therapeutic targets. In this review, we focus on PDEs capable of hydrolyzing cGMP, PDEs 1-3, 5, and 9, highlighting recent research revealing novel roles to normal physiology and contributions to heart disease.

Cardiac role of cyclic nucleotides and their associated protein kinases

Cyclic AMP and cGMP control a broad range of myocardial properties including heart rate, cell growth and survival, interstitial fibrosis, vascular tone, endothelial permeability and proliferation, and muscle contractility and lusitropy. Cyclic AMP is generated by adenylate cyclase (AC type 5 and type 6 in the heart) and activates one of three cognate proteins: protein kinase A (PKA) expressed as one of two isoforms PKA-I and PKA-II, or the exchange protein directly activated by cAMP (Epac). PKA-I is primarily engaged in adrenergic stimulated phosphorylation of proteins that control excitation-contraction coupling and sarcomere function. These include troponin I ⁴, titin ⁵, myosin binding protein C ^{6, 7}, phospholamban ⁸, the ryanodine receptor (RyR2) ⁹, and the L-type calcium channel ¹⁰. Epac is a guanine nucleotide exchange factor (GEF) protein that activates calcium-calmodulin activated kinase II (CaMKII) to influence calcium cycling and gene transcription ¹¹.

Cyclic GMP is generated by either one of two guanylyl cyclases; GC-1, which is stimulated by nitric oxide, or GC-A which resides in the intracellular domain of the natriuretic peptide receptor (NPR1). Cyclic GMP in turn activates cGK1α (also PKG-1α) by binding to N-terminus regulatory domains. This kinase phosphorylates many similar myocyte calcium regulatory and sarcomere proteins and specific residues as protein kinase A (e.g. phospholamban, TnI, titin, myosin binding protein C). However, contractility is not generally altered as calcium transients are minimally changed. cGK1α also opposes neurohormonal stimulation by phosphorylating and/or binding to regulator of G-protein signaling 2 and 4 (RGS2, RGS4) to counter Gq- and Gi-protein receptor coupled agonism ^{12, 13}. cGK1α also phosphorylates transient potential receptor canonical type 6 (TRPC6) to block calcineurin/NFAT signaling^14-16, RhoA to suppress Rho-kinase signaling ¹⁷, and tuberin (TSC2) to block pathological growth and enhance autophagy in myocytes with stress-activation of the mechanistic target of rapamycin complex-1 (mTORC1) ¹⁸. The latter appears particularly important, as genetic mutation of a single phosphorylation site (S1364 human) on TSC2 to prevent its phosphorylation (serine to alanine substitution) was sufficient to eliminate the antihypertrophic efficacy of cGK1α activation against pressure-overload stress in vitro and in vivo. Lastly, activated cGK1α also enhances protein quality control by increasing proteasome activity to enhance clearance of damaged ubiquitinated proteins¹⁹.

PDE1: From cGMP to cAMP modulation in the heart

PDE1 is Ca²⁺/calmodulin (CaM)-activated²⁰ and a dual cAMP/cGMP esterase expressed as one of three different isoforms. Of these, PDE1A and 1C are expressed in the heart. PDE1A has a 25-fold lower Km for cGMP than cAMP ²¹ so is more cGMP-selective, whereas PDE1C has an equal affinity for both cyclic nucleotides ²². PDE1A represents the dominant cardiac isoform in mice and rats, whereas PDE1C predominates in larger mammals such as humans, dogs, and rabbits²³. The N-terminus contains two Ca²⁺/CaM binding sites and a phosphorylation site at Ser120, the latter modified by PKA in PDE1A and CaMKII in PDE1B to suppress sensitivity to Ca²⁺/CaM and thus PDE activation. PDE1A and PDE1C expression are upregulated in hypertrophied rodent hearts and myocardium from human heart failure^{21, 23-25}. In mice, β-AR stimulated hypertrophy is suppressed by PDE1 inhibition by a mechanism most compatible with cGK1 activation²⁶. PDE1A is also upregulated in rodent myofibroblasts after myocardial infarction, and PDE1 inhibition blocks expression of pro-fibrotic genes²⁷. The anti-fibrotic response involves both cGMP and cAMP signaling. This is summarized to the center-right in Figure 1.

Figure 1:

Cardiac myocyte regulation by PDE1A and PDE1C. PDE1A hydrolyzes both cAMP and cGMP, with greater affinity for cGMP derived from either GC-A (natriuretic peptide coupled) GC-1 (nitric oxide coupled) cyclases. Cyclic GMP activates cGK1α which phosphorylates to inhibit TRPC3/6, and suppresses G_q-coupled protein receptors by stimulating RGS2/4 proteins. Both results in attenuated hypertrophic signaling. PDE1A inhibition also enhances a cAMP that activates Epac to suppress profibrotic transcriptional regulation. PDE1C has equal affinitity for cAMP and cGMP. Its calcium-dependent activation is coupled to calcium influx through TRPC3, and colocalizes with adenosine Type 2 receptors (A₂R). PDE1C hydrolyzes the cAMP coupled to A₂R stimulation. The consequent PKA stimulation suppresses cell death pathways, as well as stimulates a positive inotropic response. The latter is independent of cAMP coupled to β-adrenergic stimulation. Abbreviations: NOS- nitric oxide synthase, NFAT – nuclear factor of activated T-cells, AngII – angiotensin II, Phe – phenylephrine, ET-1 endothelin-1, Epac: exchange protein directly activated by cAMP; CamKII – calcium calmodulin-dependent kinase II, GPCR – G protein coupled receptor, NP – natriuretic peptide, NPR – NP receptor, Cm – calmodulin, Cn – calcineurin.

The selective role of PDE1C, the isoform most prominent in larger mammalian hearts, was first suggested in genetic PDE1C-knockout mice. Despite having very low levels of basal expression of the 1C isoform, myocytes lacking the gene were protected against cytotoxic stress-induced apoptosis, and in vivo hearts subjected to pressure-overload displayed less pathological hypertrophy, fibrosis, apoptosis, and dysfunction ²⁴. Interestingly, this was not coupled to cGMP signaling enhancement, but rather to cAMP-PKA activation.

Two recent studies have further identified a critical role of PDE1C in controlling a cAMP-modulated signaling cascade²⁸. Continuing their work, the Yan laboratory uncovered a key mechanism underlying PDE1C-cAMP dependent protection against myocyte stress induced apoptosis. A key feature was identifying the source of the regulated cAMP to be coupled to type A2 adenosine receptors (A2R), and for PDE1C activation to occur by Ca²⁺ delivered via transient receptor potential canonical cation channel type 3 (Trpc3)²⁹. The authors identified a novel protein complex linking PDE1C, TRPC3, and A2R, providing a regulatory pathway to modulate adenosine-mediated anti-apoptotic cytoprotection via PDE1C inhibition.

In a separate study reported in the same journal issue, the Kass laboratory explored the cardiovascular impact of inhibiting PDE1 (all isoforms) in larger mammals²³. In both dogs and rabbits, they showed PDE1C>>PDE1A expression as observed in humans but opposite the PDE1A>>PDE1C expression in mice and rats. Using the selective PDE1 inhibitor ITI-214, they found positive inotropic, lusitropic, chronotropic, and vasodilator effects from PDE1 blockade in normal conscious dogs, with all but heart rate change being preserved in the same animals when their hearts were induced into failure following 3-4 weeks of tachycardia pacing. Similar profiles were observed in intact rabbits. PDE1 inhibition did not augment dobutamine stimulation, nor were its effects prevented by marked beta-receptor blockade (with the exception of heart rate that was suppressed by this blockade). Rather, cardiac and systemic arterial dilator effects from ITI-214 were blocked by A_2B receptor inhibition. In myocytes, PDE3 but not PDE1 inhibition augmented resting and isoproterenol stimulated cell contraction and calcium.

These data have shifted the focus of PDE1 regulation in the intact heart from one emphasizing cGMP to cAMP signaling, with the latter being coupled to adenosine and not β-AR receptors (Left side, Figure 1). The protective effects of A₂ receptor stimulation against ischemic injury are well known³⁰, and the new data suggest such protection maybe leveraged by PDE1 inhibition. Furthermore, the positive inotropy and lusitropy may differ from that coupled to β-AR-cAMP signaling, with the potential for a different safety/toxicity profile than observed with PDE3 inhibitors. A Phase Ia-IIb clinical trial testing safety and hemodynamic responses to single dose of ITI-214 (placebo+ dose escalation, Clinicaltrials.gov: NCT03387215) in patients with dilated heart failure initiated in mid 2018 and continues.

PDE2: Signaling ambiguities and alternative roles

PDE2 is a dual substrate phosphodiesterase that hydrolyzes cAMP and cGMP at similar maximal rates. A single gene Pde2a gives rise to three isoforms (PDE2A1, 2A2, and 2A3) which are differentially localized in cytosol, membrane, and mitochondrial sub-domains. A primary characteristic of PDE2A is its allosteric activation by binding of cGMP to GAF domains residing in the N-terminus that augments catalytic activity of cAMP by 10-30 fold. The reverse, cAMP mediated activation for cGMP hydrolysis, does not occur given the low affinity of the GAF domain for cAMP. This positions the PDE for particular relevance to cGMP/cAMP crosstalk.

In keeping with its regulatory capacity, studies have found the consequences of PDE2 regulation to vary depending on the precise conditions (Figure 2). In quiescent adult cardiomyocytes, PDE2 inhibition at the plasma membrane augments cGMP generated by guanylyl cyclase GC-A coupled to natriuretic peptide receptors (NPR1, NPR2), or by GC-1 which is activated by nitric-oxide. This was first revealed in myocytes expressing a cGMP-sensitive cation channel (Cnga2)³¹. The role of PDE2 in compartmentalizing cGMP generated by ANP stimulation of NPR1 was also revealed using nanodomain ligand/receptor activation of the receptor in adult cardiomyocytes. The NPR1 receptor (activated by A-type and B-type NP) was found in T-tubule membranes, and cGMP detected by a FRET sensor shows it to be normally constrained to the membrane near where ANP stimulation had been introduced. However, following PDE2 inhibition, ANP-stimulated cGMP diffused more generally into the cytosol³². This local regulation was further confirmed using a cGMP-FRET-sensor targeted to the caveolin-enriched nanodomains where NPR1 resided. One consequence of such compartmentation was that while ANP stimulation did not normally alter phosphorylation of the calcium cycling regulating protein phospholamban, it did if PDE2 was concomitantly inhibited³². Coupling of PDE2 with NP but not NO-dependent signaling has also been demonstrated in pulmonary vascular smooth muscle³³.

Figure 2:

Regulation of cardiac myocyte function by PDE2 and PDE3. PDE2 hydrolyzes cGMP and cAMP, and its activity is further stimulated by cGMP binding to regulatory domains. PDE2 inhibition can activate PKA-type II resulting in NFAT phosphorylation to prevent its nuclear translocation and hypertrophic signaling. It also stimulates PKA-type I that stimulates cardiac contraction and confers mitochondrial protection. PDE2 cGMP hydrolysis can provide antihypertrophic signaling effects. PDE3A can also hydrolyze cGMP and cAMP, the latter coupled to beta-adrenergic receptors. PKA- type I is activated phosphorylating L-type calcium channel (LTCC), phospholamban (PLN), and contractile proteins including troponin I, Tn; myosin binding protein C, MyBPC, and titin. This promotes contraction and relaxation. A deletion (del) in the PDE3A1 promoter prevents ATF3 (activating transcription factor 3) binding, disinhibiting effects of cAMP/CREB binding to enhance synthesis of PDE3A1. PDE3B hydrolyzes cAMP that has an impact on protection against ischemia/reperfusion injury linked PKA-I mitochondrial modifications. AC – adenylate cyclase, CREB – cyclic AMP responsive element binding protein.

The role of PDE2 in hydrolyzing cGMP generated by nitric-oxide stimulation of GC-1 has eluded detection by similar fluorescent sensor methods, most likely primarily due to insufficient sensitivity to the low amplitude cGMP signal. However, functional data support such interaction. For example, NO-stimulated cGMP, coupled to activation of the β₃-adrenergic receptor, was shown to activate PDE2 (cGMP binding to regulatory GAF domains) in neonatal cardiomyocytes. This enhanced its hydrolysis of cAMP and thereby countered catecholamine stimulated contraction³⁴. PDE2 inhibition also enhances cAMP-induced dilation induced by prostaglandin I2, supporting a role for cAMP hydrolysis in vascular tissue ³³. Yet another study reported antihypertrophic effects from PDE2 inhibition in isolated neonatal cardiomyocytes exposed to angiotensin II, but this was observed when cGMP was co-stimulated using an NO-donor but not by a NPR1 agonist (ANP or BNP)³⁵. Similarly, upregulation of PDE2 (genetic overexpression) in fibroblasts did not alter cGMP stimulated by NP agonism, but had a slight significant effect upon NO stimulation³⁶. These latter studies are hard to reconcile with earlier data coupling PDE2 to NPR1 signaling, but they may reveal a critical role of experimental conditions and cAMP/cGMP balance.

PDE2 also hydrolyzes cAMP, as demonstrated by its control of β-adrenergic stimulation³⁴ (Figure 2). In addition, PDE2 controls cAMP-dependent modulation of myocyte hypertrophy, fibrosis, and mitochondrial function. The primary effector of β₁-adrenergic stimulation is cAMP-stimulated protein kinase A-I (PKA-I) that when sustained induces myocyte hypertrophy. By contrast, the PKA-II isoform confers anti-hypertrophic effects due to phosphorylation of nuclear factor of activated T-cells (NFAT) blocking its nuclear translocation and thereby myocyte hypertrophy³⁷. PDE2 inhibition has been shown to enhance this cAMP-PKA-II pathway³⁷. PDE2 overexpression in fibroblasts reduces cAMP levels with or without their co-stimulation, and results in an activated myofibroblast phenotype³⁶. Interestingly, this transition is offset by co-stimulation of cGMP via NO or NP routes, indicating that PDE2 upregulation in this setting is preferentially via hydrolyzing cAMP. Lastly, PDE2A2 inhibition impacts mitochondrial elongation, increasing the transmembrane potential and thereby resistance to pro-apoptotic stimuli in MEFs and NRVMs. These changes are mediated by cAMP-PKA-dependent phosphorylation of dynamin-related protein 1³⁸.

The various features of PDE2 regulation identified in cellular systems have been examined in vivo in small rodents. PDE2 expression increases in diseased myocardium in various experimental animal models and in human heart failure^{35, 39}. While there is general agreement that this plays a pathophysiological role in heart disease, there is controversy as to whether therapeutic benefit will derive from PDE2 inhibition or from enhancing its activity further⁴⁰. The answer may ultimately depend on the nature of the cardiac disease. In models of pressure-load induced right or left ventricular hypertrophy, PDE2 inhibition reduces hypertrophy, fibrosis, and can improve cardiac function ^{33, 35, 37}. Even here, the explanation varies, with some studies coupling this effect to enhancement of the cAMP-PKA-II pathway³⁷, while others suggest this results from reduced cGMP hydrolysis and thus anti-hypertrophic signaling afforded by protein kinase G activation^{33, 35}.

As in the cell studies, the source of cGMP regulated by PDE2 inhibition in vivo also varies among studies. In models of pulmonary hypertension and RV hypertrophy, gene deletion of the NPR1 eliminated efficacy of oral PDE2 inhibition to counter RV disease, whereas the latter remained effective despite NOS inhibition³³. By contrast, in mice subjected to descending aortic constriction, LV hypertrophy was reduced by PDE2 inhibition independent of GC-A (GC coupled to the NPR1) expression, but was lost in mice lacking NO-stimulated GC-1α³⁵. In contrast to pressure-stress hypertrophy models, in heart disease induced by catecholamine hyper-stimulation, or by myocardial infarction, PDE2 appears to play a protective role principally by suppressing adverse β-adrenergic stimulation^{41, 42}. These studies propose upregulation of PDE2 as a protective component of β-AR desensitization limiting damage due to catecholamine hyper-stimulation, spawning efforts to further activate the PDE⁴¹. In mice with myocyte-targeted PDE2A overexpression, resting heart rate was reduced, catecholamine stimulated arrhythmia abetted, and cardiac function following myocardial infarction was improved over littermate controls⁴². The converse, e.g. reversal of downregulated β-AR signaling by inhibiting PDE2 has not been found in acute studies⁴³, though it might occur with more chronic in vivo inhibition.

Viewed together, recent studies regarding the cardiac role of PDE2 and its modulation to treat heart disease reveal complexities regarding cyclic nucleotide substrate selectivity, the source of the substrate, and even whether blocking or activating the PDE is beneficial. This suggests the role of PDE2 in vivo depends very much on the ambient conditions, which may alter the balance of cGMP versus cAMP and the compartment being regulated. While these factors are more easily constrained in animal models, even these studies show considerable discrepancies. While intriguing, the basic work does raise concerns for translation into human diseases, where the underlying pathophysiology is generally complex, so predicting the response to PDE2 modulation may be difficult.

PDE3 – Isoform-specific signaling for therapeutic targeting

PDE3 is the other dual-substrate phosphodiesterase proposed to regulate cGMP in the heart. While its affinity for cAMP and cGMP is similar, it has a higher turnover rate for cAMP, providing its primary regulatory footprint in larger mammalian hearts including human. By competitive binding at the catalytic site, cAMP hydrolysis is inhibited by cGMP, yielding a mechanism for cGMP-dependent contractility augmentation at low levels of cGMP⁴⁴. The PDE3 enzymes are transcribed from two genes, PDE3A and PDE3B. PDE3A exists as three isoforms that vary by alternative transcription and translation start sites, while PDE3B exists as only one isoform^{45, 46}. All of the isoforms differ in their N-terminus where there are hydrophobic loops (PDE3A1 and PDE3B) to mediate lipid membrane insertion, and phosphorylation sites that promote protein-protein interactions. The latter is important for determining the intracellular localization of the isoforms. In particular, PDE3A has been found to complex with PI3Kγ and SERCA2A, and is mainly localized in the cardiomyocyte at the sarcoplasmic reticulum. By contrast, PDE3B localizes to myocyte T-tubule membranes (Figure 2).

Given its role in hydrolyzing cAMP, initial interest in PDE3 cardiac regulation focused on the potential for non-isoform selective inhibitors to enhance cardiac contractility and lusitropy while simultaneously producing venous and arterial vasodilation. These inodilator attributes were particularly attractive as therapy for heart failure with ventricular dilation and depressed systolic function. Acute hemodynamic studies in the 1980’s confirmed a favorable hemodynamic profile⁴⁷, and the PDE3 inhibitor milrinone remains used in patients with acute decompensated heart failure. However, early optimism was countered by a major study of chronic PDE3 inhibition employing the identical compound that reported increased mortality, with concerns raised regarding lethal arrhythmia linked to excess calcium and energy demands⁴⁸.

The concept was resurrected over a decade later in the context of a transformation in HF treatment that now routinely included use of a β-AR receptor antagonist such as metroprolol and carvedilol. The expectation was that in such patients, the potentially toxic effects of cAMP enhancement via PDE3 inhibition would be mitigated by concomitant β-blockade. A subsequent ~2000 patient multicenter international trial of enoximone, another PDE3 inhibitor, reported no adverse impact on survival or other clinical outcomes, however, there was also no significant benefit⁴⁹. This lack of benefit may be due in part to a 29-nt INS/DEL polymorphism in the Pde3a promoter region⁵⁰. The transcription factor ATF3 normally binds to the promoter insertion site to repress cAMP response element activity, such that PDE3A1 transcription is not upregulated in a feedback manner by cAMP. In patients with the polymorphism that deletes this binding site, the cAMP response element maintains activation enhancing PDE3A1 mRNA. Since inhibiting PDE3 increases cAMP, those with the DEL polymorphism would offset this by enhancing PDE3A expression. This may provide a genetic basis for why some HF patients develop tolerance to PDE3 inhibitors over time, though direct proof of this remains lacking (lower left, Figure 2).

While short-term use of oral PDE3 inhibition in severe HF patients is used as a bridge to transplantation ⁵¹, the therapy is now also being explored in patients with HF and a preserved ejection fraction. This is related to the capacity of PDE3 inhibition to increase cardiac output reserve at lower ventricular filling pressures during exercise⁵². A new extended release form of milrinone (CRD-102) has been developed for advanced heart failure with some early favorable clinical results⁵³ similar to those reported with enoximone⁵¹. Whether this formulation will ultimately alter the safety-efficacy profile of milrinone leading to a renaissance as a HF therapeutic remains to be determined.

An alternative approach to small-molecule PDE3 inhibition is to impede cAMP hydrolysis by competing at the catalytic site with cGMP. This happens when low levels of cGMP compete with cAMP catalytic binding and thereby enhance contractility^{54, 55}. In the presence of adrenergic stimulation or co-existent PDE3 inhibition, PDE3 primarily impacts cAMP, and similar cGMP changes now antagonize cAMP-stimulated contraction⁵⁶ primarily by activating PDE2 to reduce cAMP (as discussed)⁵⁷. This exemplifies crosstalk regulation between PDE2 and PDE3 to control the fate of cGMP/cAMP modulation of contraction. Recent vascular smooth muscle findings suggest NO-GC-1 signaling also enhances PDE3A expression, as genetic deletion of GC-1 in these cells reduces expression and activity of the PDE in aortas⁵⁸. This may help maintain cAMP in the absence of homeostatic regulation of the phosphodiesterase by GC-1 generated cGMP. Whether this applies to the myocardium remains unknown. Such crosstalk regulation requires co-signaling in the appropriate nanodomain. This likely explains the lack of impact from NPR1 stimulation by ANP or BNP to inhibit or enhance β-AR coupled signaling⁵⁹, whereas NPR2 stimulation by CNP enhances cAMP signaling via PDE3 inhibition (much like that via NO), and improve contractility in normal and failing hearts⁶⁰.

Another third approach to improve the safety and efficacy from PDE3 inhibition is to develop isoform selective inhibitors. Polidovitch et al⁶¹ studied mice with either global gene deletion of PDE3A or PDE3B that were then subjected to pressure overload induced by trans-aortic constriction (TAC) and also given the PDE3 inhibitor, milrinone. WT-TAC mice treated with milrinone had less maladaptive ventricular remodeling, and PDE3A^−/− mice demonstrated similar protection with no added benefit from milrinone. By contrast, PDE3B^−/− mice responded similarly to TAC as WT, highlighting the selective role of PDE3A in this disease model. By contrast, PDE3B gene deletion is protective against myocardial ischemia/reperfusion injury⁶², with a mechanism linked to cAMP/PKA-induced mitochondrial protection⁴⁴.

As the catalytic site is identical between PDE3 isoforms, existing small molecule inhibitors are equipotent against them⁶³. However, differences in the intracellular localization and protein partnering provides the potential for an alternative approach based on disrupting local cAMP controlled by individual isoforms⁴⁴. PDE3A is a component of a multiprotein complex in the sarcoplasmic reticulum along with AKAP18, phospholamban, and SERCA2⁶⁴ (Figure 2). When PKA is activated by cAMP, PKA phosphorylates phospholamban and promotes its dissociation from the complex, leading to an increase in SERCA2 activity. PDE3 inhibition is capable of potentiating that phosphorylation and stimulation. PDE3A1 itself is phosphorylated by PKA, and that phosphorylation promotes its interaction with the SERCA2 complex⁶⁴. If this interaction is disrupted, then selective cAMP elevation near phospholamban and SERCA2 could potentiate inotropic effects perhaps while avoiding toxic effects from general PDE3 inhibition. This remains to be tested. Disruption of PDE3B from its complex with EPAC1 and p84-regulated PI3Kγ has been studied in vascular endothelial cells⁶⁵. Here, introduction of a small peptide to displace EPAC1 from PDE3B augmented PI3Kγ signaling by increasing cAMP binding to EPAC1. Whether such small peptide disrupters can evolve into a viable cardiac therapeutic remains to be seen, but early proof-of-concept studies hold promise.

PDE5 – Myocardial regulation and compartmentation

PDE5A is a cGMP-specific PDE expressed as one of three isoforms PDE5A1, A2, and A3, all three present in human and mouse and varying in their N-terminus⁶⁶. In vitro biochemical function is similar among the isoforms, though there are organ-specific differences in expression, and evidence that this may also influence subcellular localization⁶⁶. PDE5A cGMP-catalytic activity is stimulated by cGMP binding to GAF regulatory domains and by phosphorylation by PKG at serine S102 in human (S92 in mouse)⁶⁷. In cardiac myocytes, localization of PDE5A appears in a striated banding pattern that co-localizes with the z-disk protein α-actinin. This localization normally favors hydrolysis of cGMP generated by the nitric oxide-GC-1α pathway as compared to natriuretic peptide (GC-A or GC-B) pathway. However, studies in both adult canine and murine cardiomyocytes found this localization changes to a more diffuse cytosolic one in models of heart failure and hypertrophy, or if NOS3 is pharmacologically or genetically suppressed^{68, 69}. Interestingly, its normal z-disk localization is restored even in hearts with chronic NOS inhibition by directly co-activating GC1α to generate cGMP⁷⁰. The precise mechanism for altered intracellular PDE5A localization with heart disease remains unknown.

Inhibition of PDE5A is widely used to induce vascular smooth muscle relaxation, most notably in the corpus cavernosum and pulmonary vasculature. However, studies over the past two decades have documented substantial impact for countering cardiac structural and functional pathology in response to a broad range of myocardial disease (Figure 3). Indeed, studies mostly performed with the PDE5 inhibitor sildenafil have revealed its importance to myocardial cGMP-PKG activation and consequent modulation of pathological stress-mediating proteins^{12, 14, 69, 71-74}. Among these proteins are regulators of G-protein signaling (RGS2 and RGS4) that counter Gq-receptor coupled stimulation by enhancing GTP removal from the activated Gα-subunit¹². Another is transient receptor potential canonical channel type 6 (TRPC6), a non-selective non-voltage gated cation channel which principally conducts calcium and activates pro-fibrotic and pro-hypertrophic signaling coupled to calcineurin/NFAT. PKG phosphorylation inhibits the channel¹⁴. TRPC6 is also linked to abnormal mechano-stress responses in models of Duchenne muscular dystrophy, where it transduces PKG amelioration of this pathophysiology⁷⁵. Sildenafil also blocks the accumulation of misfolded proteins, such as a mutated ab crystalin by enhancing their clearance by the ubiquitin-proteasome system¹⁹. PDE5A inhibitors reduce infarct size and cardiomyocyte apoptosis in ischemia/reperfusion models, effects that require targeting of mitochondrial function and structure^76-78, ameliorate toxicity associated with doxorubicin^{79, 80}, and reduce cardiometabolic disease in the Zucker diabetic fatty rat⁸¹. The precise mechanisms, e.g. which exact proteins and residues are targeted, have not been identified for these effects, but PKG regulation of metabolism and mitochondrial function is now being actively pursued.

Figure 3:

Cardiac myocyte regulation by PDE5A and PDE9A. Both enzymes are highly selective for cGMP; PDE5A hydrolyzes principally GC-1α (NO-coupled) and PDE9A GC-A (NP coupled) derived cGMP. PDE5A inhibition leads to cGK1α that in turn phosphorylates sarcomeric proteins to enhance myocardial compliance and counter adrenergic stimulated contractility, proteasome activity, reduce hypertrophic-stimulated microRNA (miRNA) changes, RGS2/4 and TRPC6 to suppress Gq-protein coupled receptors and calcineurin (Cn) signaling, and tuberous sclerosis complex 2 (TSC2) to negatively control mechanistic target of rapamycin complex 1 (mTORC1) and its downstream effects on autophagy and hypertrophic growth stimulation. PDE9A shares many of these targets, notably those converging on hypertrophic stimuli, though there are differences (e.g. no impact on miRNAs, less contractility modulation). cGK1α also confers mitochondrial protection via K_ATP channels. Oxidation of cGK1α by reactive oxygen species (ROS) results in a loss of function of the pathway, by oxidative changes in NOS, GC-1α, and cGK1α. This results in depressed anti- hypertrophic signaling. Other abbreviations are as in other panels.

In 2019, PDE5A inhibition was shown for the first time to suppress the mechanistic target of rapamycin complex 1 by cGK1α-dependent phosphorylation of tuberous sclerosis complex protein 2 (TSC2)¹⁸. This occurred at one (or both) of two adjacent serines (S1364, S1365 in human) to activate TSC2 and in turn suppress mTORC1 by a Rheb-dependent pathway. This study is the first to date in which a specific amino acid residue target of cGK1α was mutated in vivo using a gene knock-in approach, and the impact on PDE5 inhibition tested. The result is striking in that homozygote KI mice with a TSC2 serine-alanine substitution (S1365 mouse, corresponds to S1364 in human) exposed to pressure-overload displayed marked hypertrophy and dysfunction that was unresponsive to PDE5 inhibition. Heterozygotes expressing the same mutation exhibited similarly worsened responses to pressure-overload, however, the availability of one normal allele for phosphorylation was sufficient to rescue the therapeutic efficacy of PDE5A inhibition by sildenafil. Similar results were obtained in isolated cardiac myocytes. In addition to expanding the regulatory control of pathological muscle growth and associated dysfunction by PDE5-regulated cGK1α signaling, this study also revealed its impact on macro-autophagy coupled to mTORC1 regulation. Sildenafil stimulated autophagy in a cGK1α-dependent manner and this was coupled to TSC2 phosphorylation at S1364.

While the majority of evidence supporting ameliorative effects from chronic PDE5A inhibition have been derived from small rodents – mostly mice, large animal model data also exist. For example, in tachypacing induced HF in sheep, tadalafil was found to improve contractile function, beta-adrenergic reserve, myocyte peak calcium transients, and restructure normal T-tubular architecture⁸².

Despite the multitude of studies revealing molecular, cellular, and intact organ impact from PDE5A regulation (both overexpression models and inhibition) in the myocardium, controversy remains regarding whether this PDE is in fact expressed in myocytes, and if so at levels sufficient to explain the observed changes. The major line of evidence used to claim a minimal role is based on lack of gel electrophoresis protein detection in mouse, canine, and human myocytes and myocardium^{83, 84}. However, this may reflect the assay specifics, as others have reported expression in each of these species, and in particular in humans with cardiac hypertrophy⁸⁵, and dilated failure including in single ventricle^{86, 87}. The second line of evidence fueling controversy is that PDE5A inhibition is impactful in some models of heart disease but not others^{83, 88}. It is possible that some studies utilized inadequate dosing, as mice have nearly 100x greater catabolism than humans, and much higher doses than those used clinically are required to achieve therapeutic yet selective free plasma concentrations⁷². However, model specifics are also likely key. PDE5A regulation of cGMP by the NO-stimulation pathway depends upon the extent the pathway is activated, and loss of this cGMP pool negates amelioration from blocking PDE5A⁶⁹. This was shown in female mice exposed to TAC that lost the anti-hypertrophic effects of sildenafil if their ovaries were removed, but had it restored if the mice also received exogenous estrogen replacement⁸⁹. PDE5A regulation of heart disease also requires that sufficient stress is generated so relevant PKG-targeted proteins are engaged. In the case of mild pressure-load induced myocardial disease, PDE5A inhibition has minimal impact, yet it yields ameliorative effects with more severe disease progression as PKG-modifiable pathological signaling is activated⁹⁰. In the often cited neutral clinical human trial of sildenafil in patients with heart failure and a preserved ejection fraction (HFpEF), only a minority of subjects had ventricular hypertrophy, and overall blood pressures were only slightly above normal limits⁹¹. Unlike HF with a reduced EF, there is no clear data supporting PDE5A upregulation in this form of disease. Thus, lack of efficacy from PDE5A inhibition in HFpEF may be due to a lack of adequate upregulation in this syndrome. Lastly, oxidative stress common to various forms of heart disease may influence the therapeutic efficacy from PDE5A inhibition. This works in both directions. Oxidative changes in NO synthase and cGC1 both result in reduced cGMP generation⁹², and oxidation of cGK1α at C42 residues to form a disulfide in the N terminus between its homodimers depresses its amelioration of cardiac disease^{79, 93}. This appears due to changes in intracellular localization of oxidized cGK1α versus a C42S form that cannot be oxidized and favors a plasma membrane distribution in adult cardiomyocytes. Together, these redox changes compromise the impact of PDE5A inhibition in the diseased heart. On an opposing side, mouse hearts and myocytes expressing C42S mutant cGK1α fail to show enhanced cGK1α activation and corresponding beneficial effects from PDE5A inhibition, whereas stimulating cGK1α directly is beneficial independent of its oxidation status⁹⁴. So in this situation, oxidative stress plays a positive role to sildenafil efficacy. The mechanism again appears related to altered intracellular localization of PKG1a as oxidation is needed for optimal engagement with cGMP regulated by PDE5A.

Clinical translation of PDE5A inhibitors for cardiac disease remains a work in progress. A meta-analysis of reported controlled clinical trials (555 patients in 13 studies) found significant improvement on clinical outcome, exercise capacity, and pulmonary hemodynamics in HF patients with a reduced ejection fraction (HFrEF)⁹⁵. A caveat is that these were mostly single center trials. HFpEF patients did not benefit but this was mostly based on the sole multicenter trial⁹¹.

PDE9A - Differential Effector of NO versus NP-generated cGMP

PDE9A is a cGMP-specific PDE encoded by a single gene that is then alternatively spliced to produce multiple isoforms⁹⁶. It is expressed throughout the body, including lung, kidney, heart, and skeletal muscle, but most prominently in cerebellar Purkinje neurons and at lower levels in cortex, hippocampus, and striatum. Brain expression is thought to play a role in synaptic plasticity⁹⁶, and studies to date have focused on this role and potential therapeutic use for disorders with cognitive disease such as Alzheimer’s and schizophrenia. There is isoform specific expression in different tissues, with brain expressing PDE9A6/13 and three higher molecular weight isoforms – PDE9x-100, −120, and −175. These are not found in the heart, which instead expresses PDE9A2, and 9A9⁹⁶. Isoform-specific functional differences have yet to be identified, and inhibitors of the enzyme do not differentiate between them.

The role of PDE9A in the heart was first reported in 2015⁶⁹. Expression at both mRNA and protein level was demonstrated in human tissue, where it was also found to be increased in failing and hypertrophied hearts. It is distributed differently from PDE5A in the cardiac myocyte, exhibiting a longitudinal staining pattern that colocalizes with SERCA2A but not with α-actinin (the opposite holding for PDE5A). Rather than modifying cGMP synthesized by NO-stimulated GC-1, PDE9A regulates cGMP coupled to natriuretic peptide signaling in myocytes (Figure 3). Mice lacking Pde9a are protected against sustained pressure-overload, exhibiting reduced hypertrophy, fibrosis, and enhanced cardiac function, and similar protection was observed in WT mice treated with a selective oral PDE9A inhibitor (PF-00047943). The latter occurs whether NOS is concomitantly inhibited by L-NAME or not, revealing independence from NO-cGMP dependent signaling. This contrasts to effects from sildenafil, a PDE5A inhibitor, that is only effective in countering pressure-overload if NOS signaling remains intact.

Comparisons between cGMP and consequent cGK-1 regulation by inhibition of either PDE5A or PDE9A in the heart have yielded intriguing similarities and differences. Both have a strong impact on suppressing hypertrophic and fibrosis-related gene expression, notably signaling coupled to TRPC6 and downstream calcineurin activation⁶⁹. However, the phosphokinome modified by either inhibition strategy also displays many unique targets, consistent with differences in cellular localization and substrate preference between the two PDEs. Among the most striking disparities is how PDE5A or PDE9A inhibitor treatment impacts microRNA expression altered by pressure-overload⁹⁷. The loading stress results in the upregulation of many pro-hypertrophic/pro-fibrotic miRNAs (e.g. 34c, 21a, 199a, 208b) and anti-hypertrophic miRNAs (e.g. MiR 1a, 30b, 133a). Of 111 miRs significantly impacted by pressure-overload, PDE5A inhibition reduced most of them. By contrast, PDE9A inhibition had negligible impact on the miRNA profile. This was true despite both treatments conferring near identical phenotypic improvement in cardiac function, histology, and morphology. RNAseq analysis revealed ~70-90% of the genes with altered expression from each respective PDE inhibitor were different, though they converged on similar signaling pathways. Disparities in miR expression occurred in the cytosol, in that neither Pri-miRNA or Pre-miRNA showed differences between PDE-inhibitor treatments.

In addition to a myocardial role, recent studies suggest PDE9A may contribute to vascular smooth muscle (VSM) cGMP regulation⁹⁸. Similar to brain, VSM cells were found to express PDE9A6/A13 but at a much lower level, as well as what may be PDE9x-100. PDE9A inhibition augmented cGMP coupled to NO stimulation but not ANP or CNP stimulation in rat arterial smooth muscle cells⁹⁸, opposite to the substrate selectivity found in cardiomyocytes and the intact heart⁶⁹. The reasons for this different are unknown, but might relate to expression of different splice variants and thus intracellular localization which has not been examined in VSM to date. The vascular study did not confirm changes in cells lacking the PDE9A gene. It is worth noting that PDE9A inhibition in vivo (both mice and humans) has not been found to impact systemic blood pressure. Ongoing studies with a PDE9A flox’d mouse will be helpful to sort out tissue specific roles.

Cardiovascular and renal effects of acute PDE9A inhibition were recently reported in sheep with normal or failing hearts⁹⁹. A different PDE9A inhibitor compound (PF- 04749982) was administered intravenously, and this dose dependently increased plasma cGMP/natriuretic peptide ratio, urinary cGMP, and in heart failure conditions, urine volume, sodium excretion, and creatinine clearance. An observed increase in cardiac output and arterial vasodilation in animals with failing hearts administered the highest dose may reflect PDE9A-mediated changes, though off target effects on PDE1 are also possible. No prior studies in humans have reported blood pressure changes. These data in a large mammal support interactions between PDE9A and NP-derived cGMP, and suggest HF benefits may derive from cardiac and renal improvement.

Clinical exploration of the cardiovascular role of PDE9A is in early stages. Several small molecule inhibitors have been generated, and at least three have been studied in humans for treatment of Alzheimer’s disease¹⁰⁰, schizophrenia, or prostatic hypertrophy. None of these indications proved sufficiently encouraging to continue development. However, Cardurion Pharmaceuticals initiated safety and dose-finding human studies in heart failure patients in late 2018, using the PDE9A inhibitor CRD-733, and these trials are ongoing. Another potential avenue is to synergize the regulation of NP-dependent cGMP levels resulting from neutral endopeptidase (neprilysin) inhibition with suppression of hydrolysis of the synthesized cGMP by a PDE9A inhibitor. Preclinical studies are underway to assess this interaction.

Summary

In just the past few years, studies have revealed striking new roles of various PDEs on cardiovascular physiology and disease. With PDE1, the disparity in isoform expression between rodent and larger mammal may explain the cAMP-related changes observed in the latter but not former species. After near abandonment for several decades, new therapeutic opportunities for PDE1 inhibition to treat myocardial disease and heart failure have led to clinical studies. PDE3A inhibition by non-conventional means, including allosteric protein disruptors or altered pharmacokinetics may reignite interest. The role of PDE5A and PDE9A inhibition continues to be elucidated, but their impact on molecular signaling in the heart, and differential control over cGMP synthesized by NO versus NP pathways, support ongoing efforts to test their therapeutic utility for heart disease.