Modeling the physiological roles of the heart and kidney in heart failure with preserved ejection fraction during baroreflex activation therapy

Abstract

Heart failure (HF) is a leading cause of death and is increasing in prevalence. Unfortunately, therapies that have been efficacious in patients with HF with reduced ejection fraction (HFrEF) have not convincingly shown a reduction in cardiovascular mortality in patients with HF with preserved ejection fraction (HFpEF). It is thought that high sympathetic nerve activity (SNA) in the heart plays a role in HF progression. Clinical trials demonstrate that baroreflex activation therapy reduces left ventricular (LV) mass and blood pressure (BP) in patients with HFpEF and hypertension; however, the mechanisms are unclear. In the present study, we used HumMod, a large physiology model to simulate HFpEF and predict the time-dependent changes in systemic and cardiac hemodynamics, SNA, and cardiac stresses during baroreflex activation. The baseline HFpEF model was associated with elevations in systolic BP, diastolic dysfunction, and LV hypertrophy and stiffness similar to clinical HFpEF. Simulating 12 mo of baroreflex activation resulted in reduced systolic BP (−25 mmHg) and LV mass (−15%) similar to clinical evidence. Baroreflex activation also resulted in sustained decreases in cardiac and renal SNA (−22%) and improvement in LV β₁-adrenergic function. However, the baroreflex-induced reductions in BP and improvements in cardiac stresses, mass, and function were mostly attenuated when renal SNA was clamped at baseline levels. These simulations suggest that the suppression of renal SNA could be a primary determinant of the cardioprotective effects from baroreflex activation in HFpEF.

NEW & NOTEWORTHY Treatments that are efficacious in patients with HFrEF have not shown a significant impact on cardiovascular mortality in patients with HFpEF. We believe these simulations offer novel insight into the important roles of the cardiac and renal nerves in HFpEF and the potential mechanisms of how baroreflex activation alleviates HFpEF disease progression.

INTRODUCTION

Heart failure (HF) prevalence continues to rise and currently affects more than 6 million people in the United States with a 5-years mortality rate of 40% (1). HF with preserved ejection fraction (HFpEF) represents half of all HF cases with similar mortality as HF with reduced ejection fraction (HFrEF; 1). Patients with HFpEF are typically older, female, and overweight, and frequently have hypertension (70%) and renal dysfunction (50%; 2).

HF is associated with an increase in sympathetic nerve activity (SNA) and has been shown to be present especially in the heart, but also in the kidney, adrenal glands, and periphery (3). Patients with HF with higher SNA have worse prognoses (4). Unfortunately, neuromodulation and other therapies that have been efficacious in patients with HFrEF have not convincingly shown a reduction in cardiovascular mortality in patients with a higher ejection fraction (EF, >50%; 5). Nevertheless, manipulating the neural system in HFpEF has shown promise. For example, results from a small trial suggested that renal SNA may play an important role in HFpEF as shown by the improvements after renal denervation (6). β-Adrenergic blockers are prescribed to most patients with HFpEF and improve left ventricle (LV) hypertrophy and survival in animal models of HFpEF, but the mechanisms are unclear (7, 8). Clinical studies examining the impact of β-blockade in HFpEF (EF > 50%), however, have not shown significant benefits on cardiovascular mortality (9).

The baroreflex is a key mechanism for the acute regulation of SNA and blood pressure (BP), but its long-term importance is likely limited because of resetting. However, devices that electrically stimulate the carotid baroreflex have been shown to chronically reduce systemic BP and SNA in the periphery, heart, and kidney in animal models of hypertension (10). In patients with resistant hypertension, baroreflex activation therapy (BAT) significantly decreases chronic SNA and BP (10, 11). Recently, BAT has been approved by the Food and Drug Administration for treatment of HFrEF based on its ability to improve HF symptoms, increase ejection fraction, and reduce markers of congestion (12). Similarly, BAT chronically reduces BP and LV mass in patients with hypertension with symptomatic HFpEF (13, 14); however, the mechanisms of how this device lowers BP or improves diastolic function in patients with HFpEF are unknown.

Mathematical modeling provides the ability to analyze complicated interrelated effects across multiple physiological systems that are difficult or impossible to test experimentally. HumMod is a large physiological simulator that has been used to investigate mechanisms responsible for the BP lowering effects of BAT (15). The major goal of this study is to propose possible mechanistic insights into the relative roles of the cardiac and renal nerves during BAT in HFpEF. We hypothesized that alterations in cardiac SNA during BAT play a primary role in the chronic improvements in heart function and hypertrophy during BAT in HFpEF.

METHODS

All simulations were performed using HumMod, a model of human physiology composed of mathematical relationships derived from experimental and clinical data. This model accurately reproduces the acute and chronic physiological responses to baroreflex activation (15, 16) and the characteristics of many pathophysiological states (16–19). The entire model, its code, mathematical derivations, documentation, and instructions on how to run simulations are available for academic download as a single ZIP file at http://hummod.org/heartfailure.zip. In brief, organs and tissues that make up the model’s peripheral circulation include the kidneys, heart, skeletal muscle, gastrointestinal tract, liver, bone, brain, fat, skin, and lungs. Blood flow through these organs is determined by the BP gradient and vascular resistance. Vascular resistances are modulated by SNA, angiotensin II, and local tissue oxygen. Both kidneys in HumMod are separated into both vascular and tubular components. There is differential control of efferent neural outflow to the renal, hepatic, cardiac, adrenal, and splanchnic territories. The quantitative relationships between changes in renal SNA and sodium excretion were based on the experimental studies by Miki et al. (20, 21). Finally, the quantitative relationship between changes in atrial pressure and plasma atrial natriuretic peptide (ANP) concentration was based on data generated in chronically instrumented dogs during the control of atrial pressure (22). Physiological responses to ANP in the model were derived from studies conducted in both dogs and rats (23–26). Determinants of renin secretion included sodium chloride delivery to the macula densa, β-adrenergic stimulation, and circulating ANP.

Model Assumptions

Sympathetic activation during HF in the model is a function of chronic pulmonary venous pressure. Afferent receptors in the heart and pulmonary circulation are complex, and the signals that contribute to the sustained increase in SNA in HF are unknown. Nevertheless, the quantitative relationship between pulmonary vein pressure and chronic SNA outflow in the model is based on experimental evidence showing that acute increases in pulmonary vein pressure increases cardiac SNA (27, 28). Cardiac function is modeled after the Suga and Sagawa heart model (29). Diastole and systole are impacted by cardiac fibrosis, which is a determinant of LV stiffness and end-diastolic pressure (Fig. 1). Total ventricle mass was divided into right and left ventricles, which are then split into contractile protein mass and fibrotic tissue mass. Cardiac fibrosis is defined as the ratio of fibrotic mass to contractile protein mass. Hypertrophy of both of these tissue types was determined by diastolic stress, systolic stress, and cardiac SNA with growth and loss constants derived to match the changes in LV mass after BAT in patients with HFpEF (Fig. 2). LV stresses were calculated under the assumption of an ellipsoidal LV geometry based on Grossman et al. (30):

$$\sigma =\frac{\text{P}\times R_{\text{i}}}{2t(1+\frac{t}{2R_{\text{i}}})}$$

where P is either systolic or diastolic pressure, R_i is the inner radius of the ventricle, and t is the wall thickness at either systole or diastole. Cardiac SNA-induced contractile protein growth in the model was determined by the chronic level of β₁-adrenergic activation, whereas fibrotic mass growth was a function of chronic α-adrenergic activation. These assumptions were based on studies suggesting that SNA-induced cardiac hypertrophy (contractile protein) in HF is driven by β₁-signaling (31, 32), whereas α-adrenergic signaling has been shown to play a key role in cardiac fibrosis (8, 33).

Figure 1.

Cardiac model describing diastolic and systolic function and influence of fibrosis and stiffness on diastolic function. The fibrosis variable is derived from the ratio of left ventricular (LV) fibrosis to LV contractile mass (top, right). The break variable derived from a function of baseline LV stiffness and fibrosis (bottom, right). EDP, end-diastolic pressure; EDV, end-diastolic volume; ESP, end-systolic pressure; ESV, end-systolic volume; HFpEF, heart failure with preserved ejection fraction; LAP, left atrial pressure; MAP, mean arterial pressure; SBP, systolic blood pressure; SV, stroke volume.

Figure 2.

Model relationships and equations that determine the gain and loss of ventricular contractile protein mass and fibrotic mass and their impact on cardiac function. The fibrosis effect is also a multiplier on stiffness as shown in Fig. 1. Contractile, fibrotic, and cardiac functions show baseline control (blue) and heart failure with preserved ejection fraction (HFpEF) values (black). EDP, end-diastolic pressure; ESP, end-systolic pressure; LV, left ventricular.

The increased SNA in HFrEF is associated with myocardial β₁-adrenergic receptor downregulation (up to 50%; 34), but adrenergic receptor changes in HFpEF are relatively unknown. Downregulation occurs in parallel with significantly less stored norepinephrine in neuronal endings due to impaired norepinephrine reuptake (as fast as 5 days in animal experiments; 35, 36). Therefore, cardiac β₁-adrenergic signaling in the model was allowed to desensitize up to 50% with chronic (7 days) stimulation. α-Adrenergic receptors, which do not downregulate (37), did not desensitize.

Simulation Protocols

In each simulation, the model had continuous water, food, and electrolyte intake, where thirst was a function of osmolality. In addition, sodium intake was increased throughout the simulations (270 mmol/day). Baseline conditions were produced to mimic HFpEF with hypertension and renal disease. This was done by increasing baseline LV stiffness 2.7× normal levels based on Doppler echocardiography estimates in in vivo human studies and direct measurements in cardiac myocyte studies in humans with HFpEF (38, 39). In addition, the majority of patients with HFpEF have hypertension (>70%) and renal disease (>60%; 40). This was recapitulated by increasing salt intake 50% and reducing kidney mass to reach a glomerular filtration rate of 70 mL/min (within the ranges found in clinical HFpEF populations, Supplemental Table S1; all Supplemental material is available at https://doi.org/10.6084/m9.figshare.20464977). These conditions were changed in the model and then simulated over 12 mo to produce the HFpEF model.

Baroreflex activation was achieved by gradually increasing carotid baroreceptor input into the central nervous system to 40% above normal over the first 3 mo in all simulations. This level of baroreceptor afferent input was chosen to produce a ∼25-mmHg fall in systolic BP seen in patients with resistant hypertension with signs of HFpEF treated with BAT (14; Fig. 3) and a ∼30% decrease in muscle sympathetic nerve activity (MSNA) seen in patients with HFrEF after BAT (41). Twelve-month responses to BAT were determined under the following four conditions: 1) control: HFpEF alone; 2) BAT: HFpEF with BAT; 3) BAT-CSNA: HFpEF with BAT with cardiac SNA, vagus activity, and all cardiac adrenergic receptor activities clamped at baseline levels; and 4) BAT-RSNA: HFpEF with BAT with renal SNA and all renal adrenergic receptor activities clamped at baseline levels.

Figure 3.

Comparison of model response to patients with hypertension with evidence of heart failure with preserved ejection fraction treated with baroreflex activation therapy (13). Clinical data are presented as means ± SE except for ejection fraction (EF), which is expressed as medians and interquartile ranges. HR, heart rate; LV, left ventricular; SBP, systolic blood pressure.

Additional analyses are included in the supplemental material that examine the role of ANP during baroreflex activation as well as the impact of a low-salt diet (Supplemental Figs. S4 and S5). HumMod is a deterministic model and therefore generates a single trajectory for a given intervention. Statistical analyses were not an appropriate metric for comparing the different simulations.

RESULTS

Baseline HFpEF was characterized by normal EF, elevated BP, and moderate renal impairment, conditions similar to clinical HFpEF (Table 1). In addition, HFpEF was associated with elevated LV mass and thickness, elevated end-diastolic pressure, and increased SNA relative to normal (Table 1). Further comparison between clinical distributions and model outputs of cardiac chamber dimensions and function can be found in the supplement (Supplemental Table S1).

Table 1.

Cardiovascular, neurohormonal, renal, and fluid responses in heart failure and during 3 and 12 mo of baroreflex activation therapy

	Baseline			BAT
	Normal	HFpEF	Range (Refs)	3 mo	12 mo	Change	Range (Refs)
SBP, mmHg	121	188	180 ± 25 (13)	161	163	−25 mmHg	−26 ± 6 (13)
HR, beats/min	72	83	72 ± 10 (13)	78	78	−6 beats/min	−3 ± 2 (13)
PCWP, mmHg	10	14	12 ± 5 (13)	14	14	0 mmHg
ESV, mL	35	39	[50–59] (39, 42)	37	37	−2 mL
ESP, mmHg	121	188	[136–172] (38, 39)	160	163	−25 mmHg
EDV, mL	112	113	[110–151] (39, 42, 43)	117	118	6 mL
EDP, mmHg	5	13	[16–26] (38, 39, 43)	14	12	−2 mmHg
EF, %	69	65	[62–68] (13)	68	68	3%	[0%–4%] (13)
LV mass, g	174	306	302 ± 92 (13)	277	259	−15%	−17 ± 3% (13)
LV wall thickness, mm	10.5	14.6	[13–15] (13)	14.2	12.8	−13%	[−14% to −7%] (13)
LV stiffness (×Normal)	1	2.7	2.7 (38, 39)	2.8	2.3	−17%
MSNA (×Normal)	1	1.9	1.8 ± 0.1 (44)	1.5	1.5	−22%	−34 ± 7% (41)
NE, pmol/L	237	451	439 ± 225 (45)	329	318	−29%
GFR, mL/min	128	68	[64–83] (39, 43)	68	68	0%

Values are means ± SD or [interquartile ranges]. Normal represents normal model outputs. Heart failure with preserved ejection fraction (HFpEF) indicates baseline and is compared with clinical HFpEF. Change refers to the difference between control and BAT at 12 mo. Reference 41 indicates HF with reduced EF patients treated with BAT for 6 mo. EDP, end-diastolic pressure; EDV, end-diastolic volume; EF, ejection fraction; ESP, end-systolic pressure; ESV, end-systolic volume; GFR, glomerular filtration rate; HR, heart rate; LV, left ventricle; MSNA, muscle sympathetic nerve activity; NE, norepinephrine; PCWP, pulmonary capillary wedge pressure; SBP, systolic blood pressure.

BAT Simulation

After 12 mo of baroreflex activation, changes in systolic BP (−25 mmHg), heart rate (−6 beats/min), ejection fraction (+3%), and LV mass (−15%) were comparable with responses to baroreflex activation in patients with resistant hypertension and symptomatic HFpEF (13; Fig. 3). In addition, baroreflex activation decreased end-diastolic pressure (−2 mmHg) despite increases in plasma volume and end-diastolic volume (Fig. 4, Table 1). At 3 mo, diastolic wall stress increased (18%), but systolic wall stress decreased (−12%) compared with HFpEF at baseline (Fig. 5). After 12 mo of baroreflex activation and the resultant changes in LV wall thickness (−8%), LV wall stresses went back toward baseline levels (Table 1, Fig. 5). The relative levels of these variables are also shown for the normal state in Fig. 5 for comparison.

Figure 4.

Systemic and cardiac and pressure and volumes during baroreflex activation. Heart failure with preserved ejection fraction without treatment (control, black) is compared with baroreflex activation therapy (BAT, dashed blue), BAT with cardiac sympathetic activity clamped at baseline including influence from circulating catecholamines (BAT-CSNA, red), and BAT with renal sympathetic activity clamped at baseline including influence from circulating catecholamines (BAT-RSNA, dashed red). BAT indicates start of baroreflex activation therapy at time 0. ANP, atrial natriuretic peptide; LV, left ventricular.

Figure 5.

Cardiac pressures and left ventricular (LV) thickness and resulting wall stress during baroreflex activation. LV wall stress was calculated based on pressure, intraventricular diameter, and wall thickness assuming an ellipsoidal geometry. Heart failure with preserved ejection fraction without treatment (control, black) is compared with baroreflex activation therapy (BAT, dashed blue), BAT with cardiac sympathetic activity clamped at baseline (BAT-CSNA, red), and BAT with renal sympathetic activity clamped at baseline (BAT-RSNA, dashed red). Normal (no heart failure) model values are indicated on y-axis. EDP, end-diastolic pressure; ESP, end-systolic pressure.

Changes in SNA and adrenergic receptor downregulation are shown in Fig. 6. HFpEF was associated with elevated pulmonary venous pressures at baseline (13 mmHg) leading to SNA activation (Fig. 6). This activation from pulmonary congestion was slightly increased with baroreflex activation (5%); however, overall cardiac SNA was reduced (−22%) and was associated with reductions in LV α₁-adrenergic (−21%) and β₁-adrenergic (−13%) stimulation (Fig. 6). Figure 7 demonstrates the changes in renal parameters during baroreflex activation. Renal SNA was elevated in the HFpEF model at baseline but reduced with baroreflex activation. This was associated with afferent arteriolar vasodilation but no activation of the renin-angiotensin system (Fig. 7) due to the suppression of renal β₁-adrenergic receptor activity, despite the renin-stimulating effect of reduced delivery of renal tubular sodium to the macula densa (Supplemental Fig. S1). The changes in afferent arteriolar conductance during baroreflex activation were primarily due to the effects on tubuloglomerular feedback (Supplemental Fig. S2).

Figure 6.

Cardiac sympathetic nerve activity and adrenergic function during baroreflex activation. Increases in sympathetic nerve activity (SNA) from increased baseline pulmonary venous pressure is shown (top, left). Heart failure with preserved ejection fraction without treatment (control, black) is compared with baroreflex activation therapy (BAT, dashed blue), BAT with cardiac sympathetic activity clamped at baseline (BAT-CSNA, red), and BAT with renal sympathetic activity clamped at baseline (BAT-RSNA, dashed red). BAT indicates start of baroreflex activation therapy at time 0. LV, left ventricular.

Figure 7.

Changes in renal sympathetic nerve activity and renal vascular resistance during baroreflex activation. Heart failure with preserved ejection fraction without treatment (control, black) is compared with baroreflex activation therapy (BAT, dashed blue), BAT with cardiac sympathetic activity clamped at baseline (BAT-CSNA, red), and BAT with renal sympathetic activity clamped at baseline (BAT-RSNA, dashed red). Normal (no heart failure) model values are indicated on y-axis. SNA, sympathetic nerve activity.

Figure 8 summarizes the determinants of LV contractile mass growth. Relative to normal, HFpEF at baseline was associated with a 70% increase in LV contractile protein mass due to effects from cardiac SNA (1.2× normal), systolic wall stress (1.2× normal), and diastolic wall stress (1.3× normal) (Fig. 8). The determinants of LV fibrosis growth and its impact on LV stiffness are shown in Fig. 9. HFpEF at baseline was associated with 2× LV fibrotic mass, primarily due to the effects of cardiac SNA (1.5× normal), systolic wall stress (1.2× normal), and diastolic wall stress (1.3× normal; Fig. 9). Interestingly, after baroreflex activation, the absolute contractile and fibrotic masses decreased, but the ratio of LV fibrotic to contractile mass also decreased, resulting in decreased LV stiffness (Fig. 9).

Figure 8.

Determinants of left ventricular contractile mass. Contractile tissue mass was based on a growth constant and effect multipliers from β₁-adrenergic activity, systolic stress, and diastolic stress as shown in Fig. 2. Heart failure with preserved ejection fraction without treatment (control, black) is compared with baroreflex activation therapy (BAT, dashed blue), BAT with cardiac sympathetic activity clamped at baseline (BAT-CSNA, red), and BAT with renal sympathetic activity clamped at baseline (BAT-RSNA, dashed red). LV, left ventricular; SNA, sympathetic nerve activity.

Figure 9.

Determinants of left ventricular (LV) fibrotic mass and resultant effects on LV stiffness. Fibrotic tissue mass was based on a growth constant and effect multipliers from α₁-adrenergic activity, systolic stress, and diastolic stress as shown in Fig. 2. Heart failure with preserved ejection fraction without treatment (control, black) is compared with baroreflex activation therapy (BAT, dashed blue), BAT without contributions from cardiac sympathetic activity (BAT-CSNA, red), and BAT without contributions from renal sympathetic activity (BAT-RSNA, dashed red). Normal (no heart failure) model values are indicated on y-axis. SNA, sympathetic nerve activity.

BAT with Cardiac SNA Clamped at Baseline Levels

Baroreflex activation was also simulated for 12 mo but without any change in cardiac SNA or cardiac adrenergic activity (BAT-CSNA simulation), resulting in heart rate, cardiac SNA, and adrenergic receptor downregulation similar to baseline levels (Fig. 6). Interestingly, BP and LV mass still decreased to comparable levels as the normal BAT simulation. This was associated with a persistent suppression of renal SNA (Fig. 7) and occurred in tandem with a lower SNA activation from pulmonary congestion (Fig. 6) due to the faster heart rate and stronger LV contractility relative to BAT (Supplemental Fig. S3). However, without the decrease in cardiac SNA, LV fibrotic mass did not decrease to the level of the BAT simulation and resulted in relatively increased LV stiffness (Fig. 9).

BAT with Renal SNA Clamped at Baseline Levels

When renal SNA and all renal adrenergic receptor activity were fixed at baseline levels, the BP fall and regression of LV hypertrophy induced by baroreflex activation was largely abrogated (Fig. 4). This was despite the exaggerated plasma ANP levels from baroreflex activation. In the BAT-RSNA simulation, the SNA activation from pulmonary congestion was relatively greater as compared with the BAT simulation (Fig. 6). However, the improvements in LV β₁-adrenergic downregulation and decreases in overall LV adrenergic receptor simulation still occurred because of the suppression of cardiac SNA (Fig. 6). In this simulation, there was greater renal vascular resistance and angiotensin II relative to the BAT simulation (Fig. 7). This was also associated with increased systolic and diastolic pressures and greater LV wall stresses as compared with the BAT simulation (Fig. 5). Finally, without the suppression of renal SNA during baroreflex activation, both the decreases in LV contractile and fibrotic masses were blunted, resulting in greater overall LV mass, but favorable improvements in LV stiffness because of cardiac SNA suppression (Fig. 9).

DISCUSSION

Originating from the work of Guyton and Coleman (46), our physiological model has been continually developed and used for a variety of different research areas for the last 50 years (17–19, 46–50), including baroreflex activation. Now called HumMod, this model replicates many physiological and pathophysiological states and was not specifically constructed for HF. Regardless, the current model of HFpEF was associated with similar baseline cardiovascular pathophysiology and similar responses to baroreflex activation as compared with clinical evidence (Table 1). However, many physiological responses to baroreflex activation in patients with HFpEF are unknown, leaving many unanswered questions. The predictions from our model attempt to fill many of these gaps and provide novel insight into the treatment of diastolic dysfunction. These predictions suggest 1) in a physiological model of hypertension, high cardiac stiffness and pressures, and renal dysfunction, 12 mo of baroreflex activation decreases BP and LV mass, similar to clinical evidence (13, 14); 2) suppression of cardiac SNA may play an important role in reducing cardiac stiffness in HFpEF, but may not be obligatory for reductions in BP or LV mass; and 3) if renal SNA is not suppressed during BAT, the reduction in BP and LV hypertrophy is impaired. These simulations indicate that the suppression of renal SNA during baroreflex activation may be the primary mechanism for the improvements in symptoms, cardiac workload (hypertension or afterload), and LV hypertrophy in these patients.

The heart and kidney interact in a complex and tight relationship in HF. Most patients with HF have chronic kidney disease (CKD), one of the strongest independent risk factors and predictors of mortality in patients with either HFpEF or HFrEF (51, 52). Patients with HFpEF with concomitant renal dysfunction are associated with increased hypertrophy and higher indexes of LV diastolic pressures and cardiac fibrosis (53). In the current mathematical model, HFpEF was modeled with CKD stage 2 (60–90 mL/min) and a 50% above normal salt intake level. These perturbations in the model, along with the cardiac dysfunction, induced a pulmonary congestion that increased sympathetic outflow to the heart, kidney, adrenal glands, and peripheral vasculature, similar to patients with HF (3).

Although afferent signals from the heart are thought to be the sustained stimuli for overt sympathetic activation and disease progression in HFrEF (54), the role of direct cardiac sympathetic overactivation during HFpEF is unknown. Sympathetic outflow is increased in patients with HF associated with decreased myocardial norepinephrine stores (35), myocardial β₁-adrenergic downregulation (34), and cardiac desensitization. This occurs in parallel with less norepinephrine stored in neuronal endings because of impaired norepinephrine reuptake (35). Here we show that baseline HFpEF was associated with LV β₁-adrenergic receptors that were downregulated 10% below normal levels, which blunted the total contractility of the heart (Supplemental Fig. S3). Furthermore, LV α₁-adrenergic receptors, which did not reset, played an important role in the model for the sustained decrease in LV fibrotic mass during baroreflex activation. However, the BP and LV mass were still reduced even with cardiac SNA remaining at high levels (BAT-CSNA simulation), further suggesting multifactorial mechanisms for the efficacy of baroreflex activation in HFpEF. To confirm the present model predictions, experiments are needed to quantitate the time-dependent rate and magnitude of adrenergic receptor desensitization during HFpEF, as well as cardiovascular and renal responses to BAT during HFpEF.

Renal sympathetic nerve activity may play an especially important role in HFpEF. For example, renal denervation has been shown to significantly improve the elevated BP, BNP levels, and LV wall thickness in patients with HFpEF with hypertension (6). Indeed, in the current study, the decreases in systolic BP and LV mass were mostly attenuated during baroreflex activation when renal SNA was fixed (BAT-RSNA) despite a greater increase in ANP (Fig. 4). In fact, if ANP was also clamped at baseline levels with renal SNA, there was no change in long-term BP or LV mass during baroreflex activation (Supplemental Fig. S4). These data indicate a crucial importance of renal SNA in HFpEF and warrants future investigation into the types of patients with HF with elevated renal SNA (younger, obese, etc.) and on treatments that modulate the renal nerves (10). To be sure, the clear importance of renal SNA suppression in baroreflex activation shown in the current simulations does not rule out the importance of cardiac SNA in these patients and the resultant potential to improve cardiac fibrosis, adrenergic receptor downregulation, and contractility (Fig. 10).

Figure 10.

Working hypothesis on progression of heart failure with preserved ejection fraction (HFpEF) and the proposed role of baroreflex activation therapy (BAT). Impaired filling due to increased relative wall thickness and increased left ventricular stiffness leads to congestion, which increases sympathetic nerve activity (SNA) through unknown mechanisms. BAT suppresses both cardiac and renal SNA as well as potentiates atrial natriuretic peptide (ANP), all of which improve cardiac function.

Salt intake was increased 50% above normal in the current HFpEF model. Lowering salt in patients with HF is controversial, and many studies have reported conflicting results (55). Studies examining low-salt diets specifically in patients with HFpEF are limited. For example, aggressive salt restriction (35 mmol/day) produced no benefit for 1 wk (56). Conversely, in 14 patients with HFpEF with CKD, low-salt diet (65 mmol/day) for 3 wk improved diastolic function and LV stiffness, independent of any changes in LV mass (57). Our model predicted a fall in BP and LV mass during 12 mo of low-salt diet in HFpEF, despite activation of the renin-angiotensin system (Supplemental Fig. S5). Baroreflex activation during the low-salt diet potentiated these responses. To the best of our knowledge, the impact of chronic low-salt intake (months) on LV mass has not been reported in patients with HFpEF with hypertension, providing another yet another important insight from the model.

The role of angiotensin in HFpEF, independent from its effect on BP, remains unknown. A small percentage of patients with HFpEF have high plasma renin activity (PRA; 10%), and PRA in these patients is highly variable, partly due to adjunctive medications that block the renin-angiotensin system (58). In the present simulations, there was little or no increase in renin secretion during baroreflex activation in the control BAT simulation despite a 25 mmHg fall in SBP (Supplemental Fig. S1). However, experimental studies in dogs show that reductions in renal perfusion pressure ≥ 15 mmHg lead to substantial increases in renin secretion (59, 60). Conversely, increases in PRA do not occur when BP decreases ≥20 mmHg during chronic baroreflex activation (61) or during ANP infusion (59). These data suggest that the mechanisms proposed in the manuscripts (renal and or cardiac SNA suppression and increases in ANP) play important roles in preventing activation of the renin-angiotensin system during the −25 mmHg change in BP. In conditions when there are neural-mediated increases in PRA at baseline, as in obesity-induced hypertension, baroreflex activation has been shown to significantly decrease PRA and reduce BP back to normal (62).

Limitations

The physiological model, HumMod, reproduces many of the physiological responses to BAT in HFpEF. Although these simulations and results are clinically relevant, the predictions presented in these simulations are to be considered hypotheses until confirmed with the experimental and clinical investigation. Patients with HFpEF are characterized by increased systemic inflammation (∼60%; 63) that could have implications in cardiac remodeling and fibrosis, which our current model did not take into consideration. However, clinical trials investigating the impact on several different anti-inflammatory therapies consistently failed to demonstrate any positive outcomes in patients with HFpEF and in some cases did worse than placebo (64). Breakdown of cardiac extracellular matrix by matrix metalloproteinases may play an important role in cardiac fibrosis regulation after myocardial infarction or HFrEF but were not included in the current model. Activation of LV α₁-adrenergic receptors was the only mechanism considered in the model for SNA potentiating fibrosis, though there is evidence that β₂-adrenergic receptors may play a role in HFpEF by proliferating cardiac fibroblasts (65) and by activating immune cells (66), which we did not include in the model. Therefore, modeling selective or nonselective adrenergic therapies with this current model should be done with caution. Finally, the model does not currently consider cardiomyocyte length or diameter, which may limit the interpretation of the relative predictions when modeling HFrEF.

Perspectives

Baroreflex activation has been shown to significantly improve symptoms in patients with HFpEF and normotension without affecting BP (67, 68) and is currently FDA-approved for advanced HFrEF (12). Interestingly, baroreflex sensitivity is impaired in HFrEF but maintained in patients with HFpEF (69), suggesting there may be greater sensitivity to this device and better efficacy in HFpEF. However, very few studies have investigated the impact of BAT in patients with HFpEF. Chronic BAT has been shown to significantly reduce BP in patients with hypertension and significantly more so in patients with hypertension with congestive HF (70). BP and LV mass have been shown to significantly decrease with BAT in patients with HFpEF and in patients with hypertension with symptoms of HFpEF (13, 14), but these trials have been small without blinding or control groups. The current model of HFpEF was associated with severe hypertension and LV hypertrophy with moderate diastolic dysfunction (end-diastolic pressure of 13 mmHg and pulmonary capillary wedge pressure of 14 mmHg). We believe the current simulations offer novel insight into the use of BAT in hypertensive stage-B HFpEF (structural heart disease but most likely without symptoms) and the potential mechanisms of how this device attenuates HFpEF progression. Unfortunately, there is a lack of female data for modeling purposes (71). Sex was not considered in the current model, but estrogen does play an important role in HF. Indeed, estrogen deficiency precipitates diastolic dysfunction and increases the incidence of HFpEF in postmenopausal women (72), whereas estrogen replacement/treatment has been shown to inhibit cardiac fibrosis in animals (73) and reduce the risk of HF in postmenopausal women (74). Interestingly, the clinical data used to validate the current HFpEF model’s responses to baroreflex activation included approximately half of females (n = 15/34), suggesting this device is effective in both males and females (13). Currently, there is a large trial ongoing (NCT02876042) using BAT for the treatment of HFpEF.

SUPPLEMENTAL DATA

Supplemental Table S1 and Supplemental Figs. S1–S5: https://doi.org/10.6084/m9.figshare.20464977.

GRANTS

This work was supported by National Institute on Minority Health and Health Disparities Grant R00 MD014738 and National Institute of General Medical Sciences Grant P20 GM104357.

DISCLOSURES

W.A.P. is the CSO for HC Simulation, LLC, which does not alter the authors’ adherence to all the American Physiological Society’s policies on sharing data and materials.

AUTHOR CONTRIBUTIONS

J.S.C. and W.A.P. conceived and designed research; J.S.C. performed experiments; J.S.C. analyzed data; J.S.C. interpreted results of experiments; J.S.C. prepared figures; J.S.C. drafted manuscript; J.S.C. and W.A.P. edited and revised manuscript; J.S.C. and W.A.P. approved final version of manuscript.