cGMP Signaling and Modulation in Heart Failure

Abstract

Cyclic GMP (cGMP) represents a classic intracellular second messenger molecule. Over the past two decades, important discoveries have identified that cGMP signaling becomes deranged in heart failure, and that cGMP and its main kinase effector, Protein Kinase G, generally oppose the biological abnormalities contributing to heart failure, in experimental studies. These findings have influenced the design of clinical trials of cGMP-augmenting drugs in heart failure patients. At present, the trial results of cGMP-augmenting therapies in heart failure remain mixed. As detailed in this review, strong evidence now exists that Protein Kinase G opposes pathologic cardiac remodeling through regulation of diverse biological processes and myocardial substrates. Potential reasons for the failures of cGMP-augmenting drugs in HF may be related to biological mechanisms opposing cGMP, or due to certain features of clinical trials, all of which are discussed.

I. Introduction

The syndrome of heart failure (HF) has reached epidemic proportions in industrialized nations¹. Current medical treatments exist for this condition, but the mortality for HF remains high. Modulation of cGMP has emerged as a potential therapeutic strategy for HF. Initial pharmacologic efforts to target cGMP attempted to take advantage of the vasodilating effects of cGMP. More recently, the focus has shifted to exploiting potential direct effects of cGMP signaling in the myocardium. This review will first outline briefly the current definition and context of HF. We will next discuss the separate components of the cGMP signaling pathway, focusing on both the clinical evidence that this pathway becomes dysregulated in HF, and on preclinical mechanistic studies establishing the roles of the pathway in opposing the HF process. Next, we will review the clinical trials to date supporting the use of cGMP pathway augmenting drugs in HF and will then then devote particular attention to recent unsuccessful or inconclusive clinical trials of these agents. Finally, we will detail the current research on the cGMP-dependent protein kinase in basic studies and in clinical studies. We will propose mechanistic and clinical explanations for the discrepancies to date between pre-clinical models and clinical trials of HF and suggest areas for future investigation.

Ia. Rationale for Understanding cGMP Signaling in Heart Failure.

As described above, HF remains highly prevalent² and represents a leading cause of hospitalization in Western nations³. While the incidence and mortality from a number of cardiovascular conditions have been decreasing, the mortality for HF remains high and the prevalence continues to rise. As a clinical syndrome, HF occurs in response to diverse pathologic stimuli³. Multiple definitions of HF have been proposed, but one unifying definition holds that HF represents a syndrome in which the heart fails to maintain a normal cardiac output under conditions of normal filling pressures⁴. Consequences of reduced cardiac output include fatigue, hypotension, and end organ damage resulting from decreased perfusion. Effects of increased filling pressures include pulmonary edema (from increased LV end diastolic pressure) and extremity edema (due to increased right ventricular filling pressures).

The biological process leading to HF has been termed pathologic cardiac remodeling. The current concept of remodeling holds that in response to pathologic stress (such as chronic hypertension, cardiotoxic agents, myocardial infarction, or other insults), the heart, in particular the LV, undergoes derangements in structure (chamber hypertrophy and dilation) as well as reduction in systolic and diastolic function. At the tissue and cellular level, cardiac remodeling is marked by various stages of interstitial and perivascular fibrosis, cellular apoptosis, reversion to maladaptive fetal gene expression, and intercellular calcium dyshomeostasis⁵. In general, the overall hypertrophic and remodeling process represents a maladaptive condition which promotes, rather than protects from HF. As such, preclinical models of cardiac remodeling serve as important tools for HF investigation⁶ whereas medical therapies for HF seek to attenuate or reverse this process.

Ib. Heart Failure Classification

For the purposes of this review we will highlight the two major classifications of HF, though we direct the reader to several in-depth reviews on the subject^{7, 8}. Briefly, chronic HF can be classified as either HF with reduced LV ejection fraction (HFrEF) or HF with preserved LV ejection fraction (HFpEF), meaning abnormal or normal systolic function, respectively. The cornerstone of medical therapy for HFrEF includes agents which oppose pro-remodeling neurohormones (angiotensin II, norepinephrine, and aldosterone). Additional therapies include amelioration of symptoms with diuretics, and prevention of arrhythmia death through implantable defibrillators.

HFpEF, which represents a spectrum of conditions, often occurs in the setting of hypertension, obesity, and advanced age⁸. In contrast to HFrEF, no medical therapies for HFpEF yet exist which can improve the mortality and natural history of this condition. Therefore, treatment of HFpEF currently involves lifestyle modification, symptom treatment as for HFrEF, and treatment of contributing conditions (hypertension, diabetes, etc.).

Finally, whereas HFrEF and HFpEF define chronic conditions, in both cases, acute worsening of symptoms and signs can occur and has been termed acute decompensated HF. We will focus this review on chronic HF therapies for cGMP modulation, as these appear most relevant to molecular signaling in the LV, but we refer the interested reader to other reviews on the topic⁹.

II. Maladaptive alterations in cGMP generating and opposing molecules in cardiac remodeling and in heart failure.

Figure 1 outlines the major pathways in the cardiovascular system which promote or oppose cGMP generation, and describes the dysregulation of these pathways in the setting of HF. The current paradigm that cGMP signaling opposes pathological remodeling represents the synthesis of a disparate set of findings ranging from mechanistic preclinical studies to correlative observations in humans. Mechanistic data supporting the anti-remodeling effects of this pathway have been the subject of several comprehensive reviews over the past several years^{9, 10,11, 12}. We will outline these alterations briefly below.

Figure 1. Schematic summary of cGMP-generating and cGMP-opposing factors, and pathophysiological derangements occurring to the system in heart failure.

Blue boxes denote cGMP generating species, and orange boxes denote cGMP opposing molecules. Red text indicates alterations to these components observed in humans with heart failure, or in animal models of left ventricular remodeling and failure. NOS, nitric oxide synthase; NO, nitric oxide; ROS, reactive oxygen species; PDE, phosphodiesterase. NEP, neprilysin (neutral endopeptidase); NP, natriuretic peptide; NPR, natriuretic peptide receptor.

IIa. Guanylate cyclase activation in HF.

Natriuretic peptide receptors (NPR) and soluble guanylate cyclase (sGC) represent the predominant cGMP generators in the cardiovascular system. Natriuretic peptide (NP) interaction with NPR-A or NPR-B activates the receptors’ intracellular GC activity. Preclinical studies^{9, 12} as well as small studies in humans¹³, support anti-remodeling roles of NPs and NPR-A and B. NPs act directly at the level of the cardiac myocyte and cardiac fibroblast to oppose pathologic growth and maladaptive gene expression¹². In HF, circulating NP concentrations actually increase¹⁴. However, the net potency of these peptides likely becomes reduced through several mechanisms. First, though total NP levels increase in HF, the ratio of mature (active) to immature (inactive) NPs decreases in HF, likely due to altered processing by the enzymes corin and furin¹⁵. Further lines of evidence support that while total NP concentrations increase in HF, the cGMP generating receptors GC-A and B become desensitized in target tissues^{16, 17}. Activity of a third NP clearance receptor, which does not possess guanylate cyclase activity but targets NPs for intracellular degradation, also increases in HF¹⁸.

In addition to the dysregulation of NPs described above, extracellular membrane-associated proteases, such as neprilysin, degrade NPs at the cell surface, thus reducing their NP activation even if circulating levels remain high. The endopeptidase neprilysin (NEP) becomes upregulated on the surface of the failing heart¹⁹, and soluble NEP concentrations predict worse outcomes in HFrEF patients²⁰, suggesting that increased NEP degradation of surface and circulating NPs may contribute to HF pathogenesis. Importantly, recent studies in animal models have investigated the mechanistic effects of neprilysin inhibition with sacubitril/valsartan in preclinical HF models. The additive effects of the NEP inhibitor sacubitril to the efficacy of valsartan (or the angiotensin converting enzyme inhibitor enalapril) have now been established in rodent models of: myocardial infarction^{21, 22,23}; LV pressure overload^24–26; LV volume overload²⁷; renal insufficiency²⁸; and isoproterenol infusion²⁸. These studies specifically identify that sacubitril improves systolic and diastolic function, fibrosis, LV pathological gene expression patterns, as well as LV hypertrophy. Further, NEP inhibition increases cGMP circulation in the blood, and reduces circulating aldosterone²³. Recently, Burke and colleagues also demonstrated that NEP inhibition in isolated fibroblasts increases phosphorylation of the PKG substrate VASP²⁵, though the requirement of cGMP and PKG for the efficacy of NEP inhibition remain untested.

As with NP-NPR signaling, experimental studies have also confirmed that nitric oxide synthase (NOS), NO, and sGC, when functioning normally, all oppose pathological remodeling at least in part through important roles in the cardiac myocyte⁹. A large body of work supports that the NO, the activator of soluble guanylate cyclase, also becomes dysregulated in HF and in conditions predisposing to HF. First, ROS directly reduce cardiovascular tissue release of NO, which can occur in the absence of effects on NOS expression and activity²⁹. Correspondingly, human studies have shown that conditions associated with reduced endothelial NO release, such as diabetes, aging, and obesity, predispose to HF^{8, 30, 31}. In fact, based on this human data a current proposed model postulates that reduced NO-induced cGMP generation drives the pathogenesis of HFpEF³⁰. Animal models have also identified that endothelial nitric oxide synthase (eNOS) becomes uncoupled in hypertension³² as well as in the failing heart³³, leading to production of pathologic reactive oxygen species (ROS) rather than NO. In addition to derangements in upstream NOS function, multiple HF risk factors and conditions of oxidative stress can directly modify soluble guanylate cyclase, rendering it unable to generate cGMP in response to NO. Oxidation of sGC also alters its intracellular localization in the CM³⁴. In vivo studies in knockin mice replicating the oxidatively modified, NO-insensitive sGC, have confirmed the biological relevance of this alteration in vivo³⁵.

IIb. cGMP catalysis with phosphodiesterases

The above studies have investigated or targeted components of upstream cardiovascular cGMP regulation. However, apart from alterations in GC augmentation, the failing heart also exhibits changes in cGMP catalysis through upregulation of cGMP-specific phosphodiesterases. Phosphodiesterases (PDEs) comprise an 11 member family and catabolize cAMP, cGMP, or both, depending on the specific PDE¹⁰. Two cGMP-specific PDEs, PDE5 and PDE9 have been studied recently. PDE5 increases expression in the failing human LV³⁶ based on samples of end stage failing myocardium. Further, aside from protein expression, total cGMP phosphodiesterase activity increases in the LV of mice subjected to experimental TAC³⁷. Pharmacologic and genetic manipulation studies in mice also have confirmed the effectiveness of PDE5 inhibition in opposing cardiac remodeling and the pro-remodeling role of the enzyme in the CM^36,38. More recently, the cGMP-selective PDE9 was observed to increase in the failing human LV both in HFrEF and in HFpEF, as well as in the LV of mice subjected to TAC³⁹. Genetic deletion³⁹ or pharmacologic inhibition^{39, 40} of PDE9 improves LV remodeling after thoracic or abdominal aortic constriction as well as in response to isoproterenol infusion.

The collected observations above describe a general model in which alterations in cGMP-generating and anti-cGMP-generating components lead to net reduction of myocardial cGMP and subsequent cardiac remodeling and failure. Though difficult to prove causality in humans, studies from myocardial tissue of HFpEF patients have demonstrated reduced cGMP concentration, compared with nonfailing tissue from patients with aortic stenosis or with HFrEF tissue⁴¹. Moreover, these differences correspond with reduced kinase activity of the cGMP-dependent protein kinase, as well as negatively correlating with isolated cardiac myocyte passive stiffness⁴¹. Reduced myocardial cGMP has also been observed in a rat model of HFpEF⁴². Numerous other animal studies also identify net increase in cGMP which accompanies PDE5 and PDE9 inhibition^{37, 39}. Thus, the animal studies of cGMP regulating molecules, as well as direct observations in humans with HF, support that rationale behind pharmacological strategies to augment intracellular cGMP in patients with HF.

III. Clinical Successes of cGMP Modulating Drugs in Heart Failure

IIIa. Nitric Oxide Donors in HFrEF

Small studies in the 1970s and 1980s established the favorable effects of nitrates on pulmonary hemodynamics and on LV filling pressures in HF patients^43–45. Further, nitrates, in combination with hydralazine, improved mortality in men with HFrEF, compared with placebo or with an alpha adrenergic antagonist vasodilator⁴⁶. Though afterload reduction with ACE inhibitors proved superior to hydralazine/nitrates⁴⁷, the combination of hydralazine and nitrates remains a standard therapy for HFrEF patients who cannot take ACE inhibitors. Further, in black patients, hydralazine/isosorbide dinitrate combination therapy provided additive mortality reduction to standard therapy⁴⁸. Despite the safety and tolerability of nitrates, no large studies have tested the isolated effects of nitrates alone in chronic HF (in the absence of hydralazine), although clinical trials have been proposed⁴⁹. In addition to their combined use in chronic HFrEF, oral nitrates as well as sodium nitroprusside still remain accepted therapies for acute HF based on their demonstrated vascular unloading effects⁹.

IIIb. Neprilysin Inhibition in HFrEF

The PARADIGM trial compared the effects of a combination of the NEP inhibitor sacubitril and the angiotensin receptor blocker valsartan (LCZ696) with the ACE inhibitor enalapril in patients with HFrEF⁵⁰. The trial terminated early due to improved survival in the sacubitril/valsartan group. Initial studies of sacubitril/valsartan in HFpEF have proven encouraging as well. The Prospective comparison of ARNI with ARB on Management Of heart failUre with preserved ejectioN fracTion (PARAMOUNT) trial of sacubitril/valsartan in HFpEF demonstrated reduced n-terminal proBNP levels in HFpEF patients treated with sacubitril/valsartan versus valsartan alone, as well as reductions in left atrial area, suggesting sacubitril-induced improvement in LV filling pressures⁵¹. An ongoing phase 3 study⁵² will test the effect of sacubitril/valsartan on clinical outcomes in HFpEF.

IV. Clinical Challenges and neutral results of cGMP modulators in heart failure

IVa. Natriuretic Peptide administration in HF.

While the above clinical trials support the use of NO donors and NEP inhibitors as cGMP modulators in HFrEF, the results of other clinical trials of cGMP modulators have been less encouraging, and in some cases negative. Several clinical trials have tested the efficacy of direct natriuretic peptide administration in HF. Chronic outpatient infusion of nesiritide did not improve outcomes in phase 2 studies⁵³. In small studies, subcutaneous administration of BNP improved LV geometry and symptoms in HFrEF¹³, and improved preexisting diastolic dysfunction in select populations⁵⁴. However, at this point large multicenter studies of subcutaneous BNP for the above conditions have not been performed.

IVb. Nitric Oxide Donors and Modulators

Based on the hypothesis that reduced myocardial NO availability underlies the pathophysiology of HFpEF³⁰, a clinical trial evaluated the effect of the NO donor isosorbide mononitrate in HFpEF⁵⁵. This study specifically examined patient activity as measured by wearable accelerometers. Unexpectedly, nitrate treatment failed to improve ambulatory activity levels, or exercise capacity in HFpEF, and reduced patient activity. Thus, despite the basic, preclinical, and human observational data supporting that HFpEF represents a state of pathologically reduced myocardial NO, the largest clinical trial to date of nitrates in HFpEF supports no benefit, and possibly even worse outcomes in patients receiving NO donors in this condition. We will review potential explanations for these discrepancies in Section VI below.

IVc. Soluble Guanylate Cyclase Stimulators and Activators

The phenomenon of sGC oxidation and the potential contribution of NO insensitivity to cardiovascular diseases has influenced the development of small molecule sGC stimulators which can stimulate sGC enzymatic production of cGMP independently of NO. sGC stimulators currently have approval for use in vascular conditions such as pulmonary hypertension⁵⁶. Recently, two separate phase 2 dose finding studies tested the oral sGC stimulator vericiguat in patients with worsening chronic HFrEF⁵⁷ or with HFpEF⁵⁸. In HFrEF, vericiguat did not improve the primary endpoint of change in NT-proBNP over 12 weeks of treatment, though secondary analysis did suggest improvements of NT-proBNP with specific doses, compared with placebo. The more recently completed study in HFpEF also observed no change in the primary endpoints (again NT-proBNP change over 12 weeks of treatment), though a dose that did not affect blood pressure did improve quality of life measures. These two studies represent the largest efforts to date to explore the effects of sGC stimulation in HF. The findings that vericiguat appeared well tolerated, and that select doses correlated with improved secondary endpoints, have enabled a phase 3 trials of vericiguat in HFrEF⁵⁹.

IVd. Phosphodiesterase Inhibition in HF

Clinical trials of PDE5 inhibition in HF have yet to achieve the expected outcome results predicted by the basic science studies. In a two center randomized study in HFrEF patients, sildenafil treatment for 1 year reduced LV mass and pulmonary arterial pressures, and also improved exercise time as well as echocardiographic estimates of LV diastolic function⁶⁰. Unfortunately, the Phosphodiesterase Type 5 Inhibition with Tadalafil Changes Outcomes in Heart Failure (PITCH-HF), a multicenter, randomized, double blinded, placebo-controlled trial of the PDE5 inhibitor tadalafil in HFrEF, terminated early due to administrative cancellation by the funding organization⁶¹. Thus, despite the encouraging mechanistic evidence, no current phase 3 clinical trial data yet exist to support the routine use of PDE5 inhibition in patients with HFrEF.

Smaller trials in HFpEF patients also suggested benefits of PDE5 inhibition on LV structure and function⁶². However, the multicenter phase 2 RELAX-HF trial failed to improve the primary outcome of peak oxygen consumption at 6 months in HFpEF patients⁶³. Current clinical practice therefore does not support general use of sildenafil or other PDE5 inhibitors for HFpEF.

V. Mechanistic evidence for cGMP and cGMP-dependent protein kinase in cardiac remodeling and heart failure.

Va. Alterations to cGMP-PKG signaling in HF.

The mixed results described above for cGMP augmenting drugs raise the question of whether the predominant cGMP effector, PKG, actually opposes cardiac remodeling, and what physiologic and molecular processes it governs within the heart and cardiac myocyte. Despite the now large body of basic, translational, and clinical work investigating GC pathway biology in the heart, the direct role in vivo of cGMP or of PKG in the pathophysiology of cardiac remodeling and HF has only more recently been investigated. In animal models of HF such as TAC, cGMP and PKG enzymatic activity each increase in the failing LV³⁷. By contrast, in tissue samples from humans with HFrEF, HFpEF, or LVH but no overt HF, PKG activity and cGMP appeared most reduced in HFpEF, compared with HFrEF⁴¹. Importantly, this study did not measure PKG or cGMP in true normal hearts. Thus, the relative increase or decrease of PKG activity in HF patients compared with truly normal LVs remains unclear.

The potentially increased PKG activity in the failing heart would appear to serve as a cardioprotective mechanism. However, recent studies have detected ROS-induced alterations in PKG structure and function and indicate that despite increased overall enzymatic activity, PKG localization and substrate exposure likely become deranged under certain pathologic conditions⁶⁴. The PKGIαisoform functions as a homodimer under normal conditions, through allosteric interactions mediated by the hydrophobic PKGIα leucine zipper domain⁶⁵. However, under oxidative stress, modification of the cysteine 42 residue by hydrogen peroxide induces covalent disulfide bonding and homodimerization of PKGIα⁶⁶, which alters tertiary structure compared with the non-oxidized dimer. Importantly, this oxidized homodimer increases in relative proportion in failing human and animal LV tissue⁶⁴. The oxidized PKGI homodimer retains increased kinase activity but cannot be activated by cGMP⁶⁷. Further, oxidized PKGIα localizes to different domains within the CM than does the non-oxidized form, and therefore phosphorylates a separate set of effectors than normal PKGIα⁶⁴. Other studies have observed that in response to TAC, PKGIα initially migrates to the CM cell membrane, but over the course of pathologic remodeling becomes more associated with the cytosol⁶⁸. Thus, available evidence to date supports that despite the increased net PKG activity observed in pressure overloaded TAC hearts, oxidation of PKGIα likely decreases activation of PKG by cGMP and disrupts normal PKGIα intracellular localization and substrate composition.

Vb. Mechanistic studies of PKGI in models of HF.

Mechanistic studies in cultured myocytes implicate PKGI as functioning within the CM to oppose cellular hypertrophy and remodeling. cGMP directly inhibits norepinephrine-induced CM hypertrophy in cultured cells⁶⁹, supporting a role for PKG as a cGMP effector, in mediating this process in the CM. The mouse models studying the effects of PKG mutation in HF are summarized in Table 1. Whole body deletion of PKGI results in early mortality from GI dysfunction, precluding investigation of the resultant cardiac phenotype⁷⁰. Due to the importance of the PKGIα LZ domain in mediating interaction with, and phosphorylation of, kinase substrates, the role of this domain in cardiac remodeling was studied in a mouse whole body knockin model of LZ domain disruption termed the PKGIα leucine zipper mutant (LZM) mouse⁷¹. The LZM mice developed rapid mortality within the first 48 hours after TAC, and displayed more severe systolic and diastolic dysfunction at 48 hours and 7 days after pressure overload, compared with WT TAC littermates⁷². These findings directly identified a critical role of PKGIα, and of its LZ domain, in mediating both early compensation to LV pressure overload and inhibition of LV remodeling. Other mouse models of PKG mutation have produced different results. For example, smooth muscle cell restoration of the PKGIβ isoform on a whole body PKGI knockout background did not result in increased LV remodeling either with TAC or with an angiotensin II infusion model^{73, 74}. Interestingly, though, these beta rescue mice experienced less reduction of myocardial fibrotic gene expression with sildenafil treatment⁷³, suggesting in these models that PKGI does mediate the anti-fibrotic effects of sildenafil.

Table 1.

Mouse models of mutation of PKGI in heart failure models.

Mouse Model	Specific Mutation	Model of HF	Mutant Phenotype versus Wild Type	Mutant Molecular Signaling in LV versus Wild Type
PKGI Knockout	Whole body deletion of gene encoding PKGI	Baseline	Early mortality with deranged intestinal motility Unable to study cardiac structure/function
PKGI Beta Rescue Mouse (BRM)	PKGI whole body deletion with smooth muscle-restricted PKGIβ expression	Isoproterenol (7 day) TAC (21 day)	No difference in LV mass	-comparable LV PDE-5 activity between genotypes
		Angiotensin II Infusion (7 day)	Reduced cardiac hypertrophy Increased Mortality	Loss of sildenafil effect on fibrotic gene expression: tgf1b; ctgf;colla1; fn1
PKGIα Leucine Zipper Mutant (LZM)	Whole body knockin for mutant PKGIα with retained kinase activity but disrupted leucine zipper substrate interaction domain	Baseline	Adult-onset hypertension LV hypertrophy Elevated LV End diastolic pressure Normal LV systolic function	Reduced LV AKT phosphorylation
		TAC (2 day; 7 day; 21 day)	Accelerated Mortality in LZM Day 2: More severe systolic and diastolic dysfunction Day 7: Increased LV Hypertrophy and systolic dysfunction	Reduced myocardial MKK4 and JNK activation at 2 days post-TAC
PKGIα Redox Dead	Knockin point mutation at cysteine 42 (C to S) abolishing H2O2-mediated PKGIα disulfide bonding and homodimerization	Baseline	Hypertension and reduced H2O2-induced vascular relaxation
		Baseline	Reduced LV diastolic function in Langendorf Preparation In vivo LV diastolic dysfunction	Reduced PLB Ser-16 phosphorylation
		TAC (21 day)	Preserved LV function after TAC Reduced CM hypertrophy Reduced LV fibrosis	Reduced LV calcineurin, and reduced phosphorylation of CaMKII, ERK, AKT
		Doxorubicin	Improved LV systolic function	Reduction in CM apoptosis as assayed by TUNEL Stain Increased inhibitory phosphorylation of RhoA
PKGI CM KO	aMHC-restricted Cre on PKG I fl/fl background, leading to deletion of both PKGIα and PKGIβ in CM	Isoprenalin (7 day) Ang II (14 day) TAC (21 day)	Isoprenalin: No differences vs WT Angiotensin II: Increased LV fibrosis, LV systolic dysfunction, LV diastolic dysfunction TAC: Increased LV systolic dysfunction	Isolated CMs: Reduced CNP-mediated PLB-Ser16 phosphorylation Preserved Iso-mediated Ser16 phosphorylation
		TAC (21 day)	Increased lung weight indicating heart failure Improved LV hypertrophy reduction and LV functional improvement with mineralocorticoid antagonist eplerenone Increased LV fibrosis
PKGIα CM KO	αMHC-restricted Cre on PKG Iα flox/flox background	Baseline	Basal cardiac hypertrophy Early mortality by 6 months age LV systolic dysfunction

Based on the in vivo observation that oxidative stress induces excess PKGIα disulfide bonding⁶⁶, mutant mouse models have demonstrated the relevance of this modification in experimental HF. In the PKGIα redox dead model, a cysteine to serine mutation at amino acid 42 prevents oxidation-induced PKGIα disulfide dimerization⁶⁷. The resultant PKGIα “redox dead” mouse develops reduced LV remodeling and dysfunction after pressure overload⁶⁴, as well as preserved LV function in response to doxorubicin cardiotoxicity⁷⁵. These findings support both a direct role of intact, reduced PKGIα in opposing adverse cardiac remodeling, and further provide in vivo evidence that PKGIα oxidation-induced dimerization in the LV represents a pathologic event.

Whereas the above data in whole body mouse mutant models support the role of WT PKGIα in counteracting pathologic LV hypertrophy and remodeling, others have also investigated a CM-specific role of PKGI in this process. CM-specific deletion of PKGI α and β led to increased LV dysfunction and chamber dilation in response to either TAC or Angiotensin II, but not isoproterenol, indicating a direct requirement of PKG within the CM for mediating LV compensation and for opposing pathological remodeling in vivo⁷⁶. These mice displayed no differences in LVH in response to pathologic stimuli. Unlike WT mice, PKGI CM-KO mice did display improvements in LV hypertrophy and remodeling with antagonism of the mineralocorticoid receptor, suggesting that some of the pathology with CM PKGI deletion arises due to disinhibition mineralocorticoid signaling in the CM⁷⁷.

In comparing the discrepant results of the in vivo PKG mutation studies, a pattern emerges in which disruption of PKGIα and Iβ in the whole body has neutral results, whereas whole-body disruption only of PKGIα produces strong effects (Table 1). In support of this, our own studies with a mouse model harboring CM-specific deletion of only the PKGIα isoform have detected basal, age-dependent LV dysfunction, hypertrophy, dilation, and premature mortality (Figure 2), indicating that PKGIα may function within the CM as a tonic inhibitor of pathological LV hypertrophy and remodeling in vivo. When contrasted with the lack of basal phenotype observed in the CM-specific PKGIα/β deletion model, these findings suggest possible separate or even opposing roles of PKGIα and β isoforms in the cardiac response to stress.

Figure 2. Cardiac myocyte specific excision of PKGIα produces cardiac hypertrophy, reduced LV systolic function, and early mortality.

(A) Agarose gel of PCR showing DNA excision of exon 1 of the PKG-I sequence (specific for the Iα subunit) from Cre+ CMs, but not ear tissue. Below is a schematic showing location of primers and predicted PCR fragment sizes for each genotype. Three primers were used, indicated by different colors. The top right gel indicates genotyping for presence of loxP sites, and the bottom right panel indicates genotyping for presence or absence of the α-MHC Cre gene. (B) Representative LV cross section of hearts of male Cre- and Cre+ littermates, as well as summary data of masses of whole heart normalized to tibia length. (C) Load independent contractile index LV end systolic pressure volume relation (ESPVR) obtained by pressure volume loop analysis in 3-month-old mice, as well as survival of male Cre- and Cre+ littermates to 7 months. fl/fl, flanked by LoxP; Cre+, αMHC-Cre bred to PKGIα fl/fl; het, heterozygous; WT; wild-type. This study was performed in accordance with the Tufts University Institutional Animal Care and Use Committee

Vc. PKG regulation of LV systolic and diastolic function in vivo.

LV pathologic remodeling involves gradual structural abnormalities ranging from the whole LV to the single CM. Thus, identifying the degree to which LV dysfunction represents a consequence of remodeling versus the primary initiator and driver of the remodeling process can be difficult. Several studies do, however, support a primary role of PKG in rapidly modulating LV and CM systolic function in addition to opposing chronic remodeling. A screen for myocardial PKGIα leucine zipper (LZ)- interacting proteins revealed the cardiac myosin binding protein C (cMyBP-C) as precipitating with the wild type PKGIα LZ domain, but not with a binding mutant of this domain⁷⁸. Expressed selectively in the CM, cMyBP-C intercalates between actin and myosin in the sarcomere, and fine tunes actin myosin contraction and relaxation⁷⁹. The M domain of cMyBP-C contains 3 highly conserved serines which when phosphorylated change M domain affinity for myosin⁸⁰. Reduced cMyBP-C phosphorylation has been observed in human and experimental HF⁸¹. Importantly, M-domain phosphorylation mediates the inotropic effect of pacing and of beta 1 adrenergic stimulation⁸². PKGIα phosphorylates the conserved M-domain serines in vitro, both at baseline and to a greater extent with cGMP stimulation⁷⁸. Augmentation of cGMP either in cultured CMs, or in vivo also increased the phosphorylation at these sites in the CM or LV tissue, respectively⁷⁸. These findings identify one of the few CM-specific PKGIα substrates reported to date, and support a direct mechanism through which PKGIα can modulate systolic function through phosphorylation of cMyBP-C. These observations also suggest that reduced cMyBP-C phosphorylation, as has been reported in human failing hearts⁸⁰, represents an important pathologic consequence of reduced cGMP/PKG signaling in human HF.

The whole body LZM mice and the CM-specific PKG deletion model both display primary functional abnormalities after TAC. Interestingly, deletion of the molecule Regulator of G Protein Signaling 2 (RGS2), an established cardioprotective PKGIα substrate⁸³, also induces rapid and severe mortality after TAC with evidence of early HF based on lung weight increases, again suggesting an important early contribution of PKG signaling to the initial LV compensation to pressure overload⁶⁸. Finally, cellular and in vivo findings demonstrate that PKGIα mediates rapid myocardial JNK activation after pressure overload⁷². Several lines of evidence support requirements of CM JNK^{84, 85} as well as upstream JNK regulating proteins, such as MEKK1⁸⁶, CDC42⁸⁷, MKK4⁸⁸, MKK7⁸⁹, and MLK3⁹⁰, for LV functional compensation to LV pressure overload. Thus, molecular and in vivo evidence supports a requirement for PKGIα and downstream PKGIα effectors for promoting rapid functional compensation to LV pathologic stress, through specific signaling in the LV and the CM.

Further lines of evidence implicate PKGI as directly modulating diastolic function, which is of particular relevance to HFpEF. PKG LZ mutant mice develop elevated end diastolic pressure with normal LV systolic function, suggesting an element of diastolic dysfunction⁹¹. The large sarcomere and structural protein titin, when phosphorylated by PKG, becomes more compliant, and this structural property directly lowers CM passive stiffness, a key component of diastolic function^{92, 93}. Passive LV stiffness, besides being regulated by CM intrinsic stiffness, can be modulated by extracellular matrix fibrosis and collagen content. In line with this, PKGI directly regulates fibrotic transcription factors such as Smads in fibroblasts to regulate the myofibroblast transition^{94, 95}.

Besides effects on LV passive stiffness, PKGIα appears to modulate energy-dependent, or active, early diastolic relaxation (also termed lusitropy), again through direct functions in the CM. For example, targeted disruption of the PKGIα LZ domain leads to rapid increase in the time constant of LV relaxation tau after TAC, indicating worsening LV diastolic function⁷². The PKGIα substrate cMyBP-C also mediates lusitropy through PKGIα-phosphorylated sites⁹⁶. PKGIα phosphorylates the sarcoplasmic reticulum protein phospholamban, leading to a net improvement in diastolic sarcoplasmic reticulum Ca2+ reuptake, a key component of diastolic function^{76, 97}.

Thus, besides the established regulation by PKG of chronic cardiac remodeling (discussed below), PKGIα also likely directly promotes early preservation of LV systolic and diastolic function after pressure overload. We speculate that, rather than reflecting only a consequence of chronic structural remodeling, the derangements in cGMP and PKG signaling observed in human HF actually contributes early after pathologic stress to decompensation, and that this loss of protective function serves as a critical driver of subsequent pathologic remodeling.

Vd. PKG modulation of other pathologic mechanisms in vivo.

PKG also opposes fundamental processes of CM hypertrophy and pathologic remodeling, through multiple mechanisms involving: inhibition of nuclear localization of nuclear factor of activated T cells (NFAT)⁹⁸, a central transcriptional regulator of pathologic fetal gene reexpression; RGS2-mediated inhibition of the CM Gq receptor⁶⁸; and direct inhibition of TRPC6 channels⁹⁹, which normally promote influx of a calcineurin-specific pool of Ca2+ sufficient to activate CN-mediated NFAT activity. Further, cGMP, presumably through PKG, regulates cysteine-rich LIM-only protein 4, which itself functions to attenuate pathological AngII/Gq signaling in the heart¹⁰⁰. More recently, Ranek et al identified that PKG inhibits stress-responsive signaling of the mTOR complex (mTORC) in the CM, and that this occurs through PKGI phosphorylation of and activation of the tuberous sclerosis complex subunit 2 (TSC2) protein¹⁰¹. These shared findings reveal mechanisms through which cGMP-PKG regulates nodal components of the both gene regulation and molecular signaling in pathologic CM and LV hypertrophy.

In addition to opposing CM hypertrophy through the overlapping processes described above, several lines of evidence support that PKGI signaling inhibits CM death^{75, 102, 103} and promotes CM survival signaling^{102, 104}. Beyond the attenuation of these canonical cardiac remodeling mechanisms, PKG has more recently been implicated in regulation of CM processes just beginning to become established as relevant to remodeling. These mechanisms include: promotion of protein turnover through activation of CM ubiquitin ligase¹⁰⁵; upregulation of CM autophagy through activation of TSC2¹⁰¹; promoting CM optimal metabolic substrate utilization¹⁰⁶; and PKG-mediated downregulation of pathological mineralocorticoid genomic signaling within the CM⁷⁷.

VI. Understanding discrepancies between preclinical successes and therapeutic challenges

As described above, current evidence from cell culture, pharmacological studies, whole body and cell restricted models now supports that: 1) In HF, PKG and upstream cGMP generating molecules become generally pathologically altered and mislocalized, leading to a net reduction of normal PKGI phosphorylation of anti-remodeling substrates; and 2) PKG and its upstream activators oppose pathological LV dysfunction and remodeling in response to various pathological stresses. These findings raise the critical question: why have pharmacological agents targeting PKG not universally improved outcomes in humans with HF?

VIa. Biological Mechanisms.

This raises the related question of why NO donors and NEP inhibition have proven efficacious in HF, whereas NP administration, PDE inhibition, and sGC stimulation remain of unproven benefit. One possibility could be that the successful agents do not actually require cGMP and PKG for their therapeutic effects. For example, NEP inhibition with sacubitril presumably improves outcomes through inhibition of proteolysis of ANP, BNP, and CNP, permitting net increase in LV and cardiovascular cGMP and cardioprotective signaling. However, NEP also degrades multiple other circulating peptides, such as adrenomedullin and bradykinin, both of which themselves oppose LV hypertrophy. Thus, it remains unknown at present whether NEP inhibition truly improves HF outcomes through NP-cGMP-PKG signaling.

Isosorbide dinitrate represents the only other known PKG modulator to date to improve HFrEF outcomes in randomized double-blind placebo-controlled trials. While encouraging from the clinical standpoint, it remains uncertain whether either nitrates or hydralazine alone might be sufficient to improve outcomes in HFrEF. Others have proposed potential complementary mechanisms through which hydralazine prevents NO oxidation^{49, 107}. Addressing the requirement of PKG for the therapeutic effects of nitrates or sacubitril does not simply represent an academic exercise. If in fact NEP inhibition or NO donation ameliorates LV remodeling or HF through PKG-independent effects, such knowledge could inform future therapeutic target research and provide potential explanations for the unsuccessful trials of other PKG activating compounds in human HF. Further, though multiple preclinical studies support an overall beneficial effect of NO-sGC signaling in the LV, the non-sGC mediated effects of NO may actually worsen HFpEF-associated phenotypes¹⁰⁸. These mechanisms could potentially contribute to the worsened outcomes in nitrate-treated patients in HFpEF trials.

The relatively neutral and disappointing results of recent HFpEF and HFrEF trials of cGMP modulating agents could be due to the diffuse derangements in cGMP generators and in PKG described above in the setting of HF (Figure 1). For example, even if NO, NPs, and GCs exert beneficial effects on LV structure and function through PKG activation, one could envision that the potentially overlapping dysregulation of cGMP augmenting and opposing factors (Figure 1) could reduce the efficacy of targeting a single component with a drug. Further, if the oxidation-induced homodimerization of PKGIα observed in the human failing LV prevents PKGIα activation by cGMP, then one could expect none of the cGMP augmenting drugs to have full efficacy if their effect truly requires cGMP induction of PKG kinase activity. Experimental support for this concept comes from the finding that inhibition of NO generation with the NOS inhibitor L-NAME disrupts the anti-remodeling effect of sildenafil in TAC³⁹. Based on the human observations of reduced NO production in conditions such as HFpEF, it therefore remains plausible that some cGMP modulating drugs have reduced efficacy in the setting of low NO states in humans. Interestingly, PDE9 inhibition appeared to improve LV remodeling even in the setting of L-NAME, suggesting that the PDE9 regulated pool of cGMP might circumvent this limitation³⁹. PDE9 inhibitors have reached clinical trials for other conditions, including dementia¹⁰⁹, and appear safe and well tolerated. At this time, however, no large-scale clinical trials of PDE9 inhibitors in HF have been performed.

Several additional possibilities could contribute to these explanations. First, the inherent biological differences between humans with chronic HF, compared with other species subjected to experimental models, may contribute to the discrepant outcomes. Further, the effects of genetic modulation in mice may not replicate the effects of pharmacologic manipulation of these signaling pathways. On a related note, many preclinical experiments used doses several orders of magnitude higher than those administered to humans³⁷. In the case of sildenafil, for example, the 100 mg/kg/day dose of this PDE5 inhibitor differs from the approximately 1–2 mg/kg/day dose in humans^{37, 63}. At the same time, however, the reported serum concentration of sildenafil achieved in preclinical studies with this dose mirrors that achieved with human administration, and likely relates to difference in drug metabolism between species³⁷. These and other features do underscore the inherent differences between small animal and human biology, and support using caution when extrapolating conclusions from preclinical disease models to outcomes of patient trials.

In summary, several biological mechanisms may explain the lack of efficacy of some cGMP modulating drugs in HF, including: 1) redundancy and overlap of cGMP degrading/opposing pathways; 2) downstream oxidative modifications of PKG signaling thus blunting upstream efficacy; and 3) reduced NO availability and reduced PKG augmentation in conditions such as advanced age or postmenopausal status¹¹⁰. Thus, rather than focusing on upstream cGMP activation, future research efforts might be better focused on modulation of downstream PKG substrates. Such a strategy could circumvent the pathologic effects of oxidized PKG, but would require a more complete understanding of the druggable PKG substrates which truly mediate its antiremodeling effects in the LV.

VIb. Clinical explanations: the timing hypothesis

In addition to mechanistic explanations, perhaps trial duration influenced the outcomes of some of the above studies. As displayed in Table 2, the unsuccessful or borderline trials of nesiritide (synthetic BNP), sildenafil, vericiguat, and isosorbide mononitrate each lasted for 6 or fewer months⁹, whereas the duration of the successful NEP inhibition and nitrate/hydralazine trials in HFrEF extended to more than 24 months. The shorter duration studies better enabled analysis of endpoints potentially influenced by the acute vasodilating effects of PKG stimulation, such as exercise duration. However, it remains possible that beneficial effects of PKG on sarcomeric protein phosphorylation, cellular metabolism, LV hypertrophy, and fibrosis, for example, might require a more chronic duration of therapy.

Table 2.

Summary of phase 2 and 3 clinical trials of cGMP modulating drugs in Heart Failure with Reduced Ejection Fraction (HFrEF) and Heart Failure with Preserved Ejection Fraction (HFpEF).

Drug	cGMP Pathway Effect	Primary Endpoint	Duration (months)	Improvement in Primary Endpoint?
HFrEF:
Nesiritide	Direct GC-A activator	Time to all-cause death or cardiovascular or renal hospitalization	3	No
Vericiguat	Direct soluble GC stimulator	Change from baseline to week 12 in log-transformed N-terminal pro-B-type NP	3	No
Sacubitril/ Valsartan	Sacubitril: Inhibitor of Neprilysin	Composite of cardiovascular death or hospitalization for heart failure	27 (median)	Yes
Isosorbide Dinitrate and Hydralazine	NO donor	Mortality	24	Yes
Isosorbide Dinitrate and Hydralazine	NO donor	Composite of death first HF hospitalization, and change in the quality of life at 6 months (self-described black patients)	18	Yes
HFpEF
Vericiguat	Direct soluble GC stimulator	1. Change from baseline in log-transformed N-terminal pro-B-type natriuretic peptide 2. Left atrial volume	3	No
Sildenafil	PDE5 inhibitor	Change in peak oxygen consumption after 24 weeks of drug	6 (3 month low dose, then 3 month high dose)	No
Isosorbide Mononitrate	NO donor	Daily activity level (measured by accelerometer)	1.5	No
Sacubitril/ Valsartan	Sacubitril: Inhibitor of Neprilysin	Change in NTproBNP from baseline to 12 weeks	9	Yes

If the requirement of PKG for acute LV compensation to pathologic stimuli observed in mice holds true for humans, this would suggest that augmenting cGMP and PKG activity prior to clinical decompensation or during pathophysiological stress would represent a more effective clinical strategy than starting a drug after the first clinical decompensation event. Current clinical practice guidelines do not recommend PKG activating drugs such as nitrates, sacubitril/valsartan, or NPs as first line agents in HF-promoting conditions such as hypertension, or in patients with features of LV remodeling but no overt decompensation or HF. But, since preexisting hypertension or LV remodeling do predict high risk for HF, it could be useful to test whether pharmacologic cGMP augmentation could serve as an effective primary prevention strategy for HF.

Finally, excess hypotension represents a complication of both the successful and unsuccessful trials of cGMP modulating drugs in HF⁹. In some cases, trials terminated early due to the excess hypotension associated with these agents¹¹¹. As described above, basic studies support that cGMP generating enzymes as well as PKG itself control beneficial molecular processes through direct effects in the CM. Additionally, PKG phosphorylates a number of CM-specific substrates, such as cardiac myosin binding protein and titin. Further, though sacubitril/valsartan did lower blood pressure in clinical HF trials, post-hoc analysis has shown that the outcome improvements afforded by this drug did not correlate with the degree of hypertension at baseline¹¹². These combined observations indicate that cGMP and PKG signaling oppose pathologic remodeling through mechanisms independent of blood pressure. Therefore, identifying myocyte-specific or at least non-vascular anti-remodeling mechanisms regulated by PKG may identify future therapeutic strategies for HF.

VII. Summary and Future Directions

The next decade will likely produce more exciting developments regarding the biology of cGMP signaling in the normal and failing heart. Several unanswered questions will hopefully be addressed during this time. First, it remains to be tested, to our knowledge, whether direct pharmacological activation of PKG can slow or reverse adverse cardiac remodeling. Answering this question could provide direct evidence about the role of PKG kinase activity in regulating cardiac remodeling and could identify a testable method of circumventing pathophysiological downregulation of cGMP production. Further, delineation of PKG substrates unique to different cardiovascular cell types (vascular smooth muscle cells, cardiac myocytes, cardiac fibroblasts, endothelial cells, and others) could suggest new therapeutic targets for various disease phenotypes. Finally, genetic or pharmacologic models of PKG disruption and deletion can be used to test the requirement of PKG for the therapeutic effects of new and established cGMP-modulating drugs in preclinical models of HF.

Additional clinical questions will be important to answer in the coming years. In the area of HFpEF, it could be informative to test whether longer-term augmentation of cGMP might have different clinical effects than the relatively short-term investigations of some of the negative studies to date. In other words, what therapeutic effect might PDE5 inhibitors, NO donors, or newer drugs have on longer time-points and on endpoints similar to those used in many HFrEF trials? In addition to this question, the potential role of circulating or urinary cGMP as either a marker of disease severity or of treatment response should be investigated. Finally, given the broad dysregulation of cGMP signaling described in Figure 1, it would prove highly informative to test whether simultaneously targeting both the NP and NO arms of the cGMP pathway could have synergistic effects in HF. Whether such an approach would allow lower doses and thus prevent hypotension, or conversely would actually produce excess limiting effects on blood pressure, remains unknown. Answering each of these translational questions will require basic science discovery, preclinical disease models, and clinical studies.

In conclusion, a large body of literature now supports that cGMP signaling becomes deranged in the failing LV, and that cGMP-PKG signaling generally opposes pathological remodeling and HF in experimental disease models, through regulation of diverse processes and substrates. The clinical applicability of cGMP modulating drugs in HF has yet to be fully realized. Future investigation of both PKGI downstream anti-remodeling substrates in the LV, and optimal timing and duration of cGMP modulating drugs in HF may suggest novel therapeutic strategies for patients suffering from HF.