Circulating Angiokines are Associated with Reverse Remodeling and Outcomes in Chronic Heart Failure

Abstract

Background:

We sought to determine whether circulating modifiers of endothelial function are associated with cardiac structure and clinical outcomes in patients with HFrEF.

Methods:

We measured 25 proteins related to endothelial function in 99 patients from the GUIDE-IT study. Protein levels were evaluated for association with echocardiographic parameters and the incidence of all-cause death and hospitalization for heart failure (HHF).

Results:

Higher concentrations of ANGPT2, VEGFR1 and HGF were significantly associated with worse function and larger ventricular volumes. Over time, decreases in ANGPT2 and, to a lesser extent, VEGFR1 and HGF, were associated with improvements in cardiac size and function. Individuals with higher concentrations of ANGPT2, VEGFR1, or HGF concentrations had an increased risk for a composite of death and HHF in the following year (HR 2.76 (95% CI 1.73 to 4.40) per 2-fold change in ANGPT2; HR 1.76 (95% CI 1.11 to 2.79) for VEGFR1; and HR 4.04 (95% CI 2.19 to 7.44) for HGF).

Conclusions:

Proteins related to endothelial function associate with cardiac size, cardiac function, and clinical outcomes in patients with HFrEF. These results support the concept that endothelial function may be an important contributor to the progression and recovery from HFrEF.

Graphical Abstract

Take Home Visual Graphic: Changes in Circulating Angiokines are Associated with Reverse Remodeling and Outcomes in Chronic Heart Failure

Lay Summary

Prior work has shown that failing hearts have a reduced density of blood vessels. We hypothesized that circulating proteins that regulate blood vessel growth and function might be altered in patients with heart failure. We measured protein levels in a group of patients from the GUIDE-IT study that had multiple assessments of heart size and function. We found that higher concentrations of ANGPT2, VEGFR1, and HGF in the blood are related to larger hearts with worse function. In hearts that recovered function over time, the concentrations of these three proteins decreased. In addition, patients with higher concentrations of ANGPT2, VEGFR1, and HGF were more likely to die or be hospitalized for heart failure. Our results suggest that changes in the blood vessels of the heart in response to the circulating factors ANGPT2, VEGFR1, and HGF could regulate the severity of heart failure.

Introduction

Heart failure with reduced ejection fraction (HFrEF) is associated with significant morbidity and mortality, but there is considerable variation in outcomes. Patients with significant reverse left ventricular (LV) remodeling, characterized by a reduction in LV volumes, have improved survival and fewer hospitalizations during follow-up.^1–5 Treatment with guideline-directed medical therapies (GDMT) can improve the probability of reverse remodeling in HFrEF, but this response to GDMT remains variable.^6,7 Furthermore, the mechanisms that mediate recovery of ventricular function are poorly defined. Better insight into the pathobiology of cardiac remodeling could allow for targeted interventions, could increase rates of reverse remodeling, and might improve outcomes among patients with HFrEF.

In normal hearts, an average cardiomyocyte is associated with approximately ~1-1.2 cardiac endothelial cells, such that nearly every cardiomyocyte is adjacent to a capillary.⁸ This close anatomic coupling suggests that myovascular cross-talk is a critical regulator of myocardial growth and adaptation to injury.^9–12 In preclinical models, decreased microvascular density, known as vascular rarefaction, has been implicated in the transition from compensatory LV hypertrophy to HFrEF.^13,14 Translational studies suggest a similar role in the human heart. Vascular rarefaction, with resultant decreases in myocardial blood flow, have been associated with advanced heart failure in patients with HFrEF^15–19 and HFpEF.²⁰ While preclinical studies to molecularly revascularize the heart have yet to be successfully translated into clinical success for HFrEF,²¹ several “pillars” of GDMT such as angiotensin-converting enzyme inhibitors, angiotensin receptor-neprolysin inhibitors, beta blockers, and sodium-glucose cotransporter-2 inhibitors have been associated with increases in myocardial capillary density.^22–26 The precise mechanisms that drive vascular rarefaction, endothelial dysfunction, or recovery of the microvasculature are not well delineated but single center studies have linked alterations of systemic angiogenic factors to both ischemic and non-ischemic HFrEF.^18,27,28 However, whether concentrations of angiogenic factors can be used to monitor treatment responses in HFrEF remains unknown.

We performed an exploratory analysis to test the hypothesis that concentrations of circulating modulators of endothelial growth and maintenance correlate with reverse remodeling and outcomes in patients with HFrEF. We utilized echocardiographic data and banked plasma samples from the Guiding Evidence Based Therapy Using Biomarker Intensified Treatment (GUIDE-IT) trial for this work.²⁹

Methods

Trial design and outcomes for the GUIDE-IT study have been previously reported.^29,30 Briefly, GUIDE-IT was a multicenter trial that randomized 894 patients with a left ventricular ejection fraction (LVEF) ≤ 40% to NT-proBNP-guided management or usual care. Participants were planned for 24 months of follow-up after enrollment; however, GUIDE-IT was stopped early for futility. 268 participants from GUIDE-IT also participated in a pre-specified echocardiographic substudy.³¹ These participants underwent echocardiograms within 30 days of study enrollment and again at 12 months. Of these patients, 99 participants had both serial blood samples and underwent concurrent echocardiograms at 0 and 12 months (Figure 1).

Figure 1: Consort Diagram.

Patients from GUIDE-IT included in this analysis were limited to those with both echo and plasma samples at 0 and 12 months from enrollment.

All participants included in this analysis provided written informed consent to participate in the GUIDE-IT trial and the accompanying echo and biomarker sub-studies. GUIDE-IT and its sub-studies were approved by institutional review boards at each site.

Echocardiographic Analysis

Protocolized echocardiograms from the echocardiographic sub-study of GUIDE-IT were analyzed in an independent core laboratory at the Duke Clinical Research Institute (Durham, NC), as previously reported. ³¹ All echocardiographic interpretations were performed by blinded readers in accordance with current guidelines.³²

Protein Profiling

All included participants provided EDTA-plasma samples at the time of their baseline and 12-month echocardiograms. Samples were stored at −80°C in a centralized biobank (Covance, Indianapolis, IN) until analysis. We employed a validated assay panel, termed the Angiome, to evaluate concentrations of key circulating protein biomarkers related to angiogenesis, inflammation, and extracellular matrix (ECM) remodeling. The Angiome panel has been validated and approved for use by the National Cancer Institute for use across a variety of cancer clinical trials.^33–35 The panel consists of enzyme linked immunoassays (ELISAs) using the Quanterix SPX platform (Quanterix; Billerica, MA), the Meso Scale Discovery platform (Meso Scale Diagnostics; Rockville, MD), and the Ella platform (ProteinSimple; San Jose, CA). All samples were assessed using a single batch printed multiplex plates to minimize assay variability. Included on each plate was a standard sample of pooled plasma from healthy volunteers to ensure consistency across plates.³⁴ Lab personnel were blinded to any demographic and clinical outcome data. An averaged value across two technical replicates was used for analysis. For this exploratory analysis, only proteins with a coefficient of variation greater than 0.75 across baseline samples were considered for additional analyses to minimize multiple testing. Proteins with incomplete data (>15% of samples below limit of detection) in the study cohort were also excluded. A full list of sample proteins is available in Supplemental Table 1.

Statistical Analysis

Continuous and categorical variables were compared between groups using the Wilcoxon rank sum test and the Fisher exact test, respectively. Biomarker concentrations are presented as median ± IQR unless otherwise noted; continuous variables are presented as mean ± SD; categorical variables are shown as counts and percentages. Prior to modeling, biomarker concentrations were log2 transformed. Mixed effect linear regression models were constructed to estimate protein effects on echocardiographic parameters, including protein concentrations as a fixed effect and subject number as a random effect to acknowledge repeated measurements from single participants over time.

For ANGPT2, VEGFR1, and HGF, we used linear regression to model changes in echocardiographic parameters by the change in biomarker concentrations over the same time frame. To determine effects on reverse remodeling, we examined patients with or without a relative decrease in LVESVi by ≥15% over 12 months, a threshold previously associated with better outcomes.³¹ The correlation between NT-proBNP concentrations and biomarker concentrations was evaluated by linear regression, with R values of <0.4 considered weak, ≥0.4 considered moderate, and those ≥0.6 considered strong. As participants had to survive to 12 months for participation in the echocardiographic sub-study, landmark analysis was set at 12 months, and clinical outcomes were evaluated from 12 through 24 months after study enrollment. For time-to-event analysis for the composite outcome of death and HHF and the outcome of death alone, Kaplan-Meier (KM) cumulative risk curves were used to estimate the event rate, and log-rank statistics were used for comparisons between angiokine tertiles. For time to event analyses using the HHF outcome, cumulative incidence curves were generated after adjusting for the competing risk of death and Gray’s method was used to compare angiokine tertiles.³⁶ Cox proportional hazard modeling was used to evaluate the hazard ratio (HR) for outcomes based on ANGPT2, VEGFR1 and HGF and NT-proBNP concentrations at 12 months. HRs are presented as change per 2-fold change in each biomarker. Cause specific hazard ratios are reported for all-cause death and HFH.

The threshold for statistical significance was 2-sided with a type I error rate of 0.05. Due to the exploratory nature of this work, no adjustment for multiple comparisons was performed. Analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, North Carolina) and R.³⁷

Results

Patient Population

99 participants from the GUIDE-IT Echo sub-study had available plasma for Angiome profiling. Cohort characteristics are presented in Table 1. The average age was 60.9 ± 11.2 years (mean ± SD); 29.3% (29/99) of patients were female; and 48.5% (48/99) of participants had an ischemic cardiomyopathy. Mean left ventricular ejection fraction (LVEF) at the time of enrollment was 28.2 ± 10.0% with improvement in mean LVEF over 12 months to 34.5 ± 11.9%. Median follow-up was 18.7 months (IQR 15.0-23.8 months). Overall, 7.1% (7/99) of all patients died, 10.1% (10/99) were hospitalized for HF , and 12.1% (12/99) experienced either death or a hospitalization for HF over the 12 month follow-up.

Table 1: Baseline Characteristics of the Study Participants Baseline Characteristics of Patient Cohort.

Continuous variables are presented as mean (SD) and categorical variables are presented as counts (%), with the exception of NT-proBNP concentrations, which are presented as median (IQR). Blue denotes p <0.05; purple <0.01, and green <0.001, p value is difference in means between those with and without remodeling. Body mass index (BMI); chronic obstructive pulmonary disease (COPD); left ventricular ejection fraction (LVEF); heart failure (HF); myocardial infarction (MI); month (mo); obstructive sleep apnea (OSA); peripheral arterial disease (PAD); ventricular tachycardia (VT).

Parameter	Overall (n=99)	<15% Reduction in LVESVi (n=48)	≥15% Reduction in LVESVi (n=51)	P Value
Age, years	60.9 (11.2)	60.1 (10.2)	61.6 (12.2)	0.36
Male	70 (70.7%)	33(68.8%)	37(72.5%)	0.85
Race
White	57(57.6%)	27(56.3%)	30(58.8%)	0.34
Black	38(38%)	17(35.4%)	21(41.2%)	0.31
Smoking History	38(38.4%)	16(33.3%)	22(43.1%)	0.43
BMI (kg/m2)	30.1(7.3)	29.9(7.04)	30.3(7.6)	0.94
Hypertension	74(74.7%)	39(81.3%)	35(68.6%)	0.23
Hyperlipidemia	58(58.6%)	35(72.9%)	23(45.1%)	<0.01
Type 2 Diabetes	44(44.4%)	27(56.3%)	17(33%)	0.04
Atrial Fibrillation	39(39.4%)	17(35.4%)	22(43.1%)	0.56
COPD	15(15.2%)	8(16.7%)	7(13.7%)	0.9
OSA	25(25.3%)	13(27.1%)	12(23.5%)	0.86
PAD	13(13.1%)	10(20.8%)	3(5.9%)	0.06
Chronic Kidney Disease	32(32.3%)	18(37.5%)	14(27.5%)	0.39
Prior MI	38(38.4%)	26(54.2%)	12(23.5%)	<0.01
Prior Stroke	8(8.1%)	2(4.2%)	6(11.8%)	0.31
Prior VT	18(18.2%)	10(20.8%)	8(15.7%)	0.69
HF duration, mo	64.4 (89.4)	90.6 (97.4)	42.1(76.3)	<0.01
Ischemic Cardiomyopathy	48(48.5%)	31(64.6%)	17(33.3%)	<0.01
Baseline LVEF,%	28.2(9.97)	28(9.6)	28.4(10.4)	0.96
12 mo LVEF, mean	34.5(11.9)	28.8(9.47)	39.8(11.6)	<0.001
Baseline NT-proBNP, (pg/mL)	2446(1394-4331)	2556(1496-5668)	2263(1321.5-3623)	0.22
12 mo NTproBNP, (pg/mL)	1305(507.8-2887)	2306(1133.8-5201.2)	666.3 (273.7-1438)	<0.001

Concentrations of Circulating Angiokines are Associated with Ventricular Size and Function

To determine whether systemic concentrations of endothelial modulators are related to ventricular structure, we assayed a panel of 25 proteins originally developed for monitoring responses to anti-angiogenic therapies in cancer patients.^33–35 We assayed plasma samples obtained at the baseline visit and at the 12-month follow-up visit of the GUIDE-IT study. Proteins with incomplete data (i.e., with >15% below limit of detection) in the study cohort and/or with little variance across baseline samples were filtered to minimize multiple testing, leaving 10 candidate proteins (Supplemental Table 1, Supplemental Figure 1). Using these proteins, we next evaluated for associations between factor concentrations and contemporaneous echocardiographic parameters. We noted multiple significant associations of protein concentrations with echocardiographic parameters (Figure 2, Supplemental Figure 2, and Supplemental Table 2).

Figure 2: Association of Circulating Modulators of Endothelial Function with Echocardiographic Markers of Cardiac Size and Function.

Point estimates and 95% confidence intervals depicting the effect of a 2-fold increase in each protein with echocardiographic parameters. A. Association of each protein with left ventricular ejection fraction (LVEF). B. Association of each protein with global longitudinal strain (GLS). C. Association of each protein with left ventricular end systolic volume index (LVESVi). D. Association of each protein with left ventricular end diastolic volume index (LVEDVi).

Among the 10 candidates analyzed, concentrations of three protein modulators of angiogenesis, or angiokines, were most consistently correlated with measures of cardiac size and function: ANGPT2, VEGFR1, and HGF (Figure 2, Supplemental Figure 2, and Supplemental Table 2). ANGPT2 and HGF had significant correlations with 9 of 10 echocardiographic measures tested, while VEGFR1 was significantly associated with 8 of 10 parameters. Associations were stronger for ANGPT2 than for VEGFR1 or HGF. In general, increases in circulating ANGPT2, VEGFR1, and HGF were associated with worse cardiac function (decreased LVEF and higher GLS) and larger cardiac size (increased LVESVi, LVEDVi, LASVi, and RA area). All three markers were also associated with higher right-sided pressures (RVSP) and worse right ventricular function (lower TAPSE). However, only ANGPT2 and HGF were significantly associated with higher left-sided filling pressures (E/e’). In addition to identifying proteins that were associated with more advanced HFrEF, we also noted a positive association of VEGFA concentrations with higher LVEF and decreased ventricular size (LVESVi).

Association of ANGPT2, VEGFR1, and HGF Concentrations with Reverse Remodeling

Based on the close correlation of ANGPT2, VEGFR1, and HGF concentrations with multiple morphologic and functional features, we focused our subsequent analyses on these three biomarkers. To determine whether biomarker concentrations related to ventricular remodeling, we modeled change in echocardiographic parameters with the change in ANGPT2, VEGFR1, and HGF concentrations overtime (Supplemental Figure 3). A 2-fold increase in ANGPT2 over 12 months was associated with a 2.31% decrease in LVEF (95% CI, −3.90 to −0.71%, p=0.005). Similarly, a 2-fold increase in VEGFR1 was associated with a 2.86% decrease in LVEF (95% CI −4.75 to −0.97, p=0.003), and a 2-fold increase in HGF was associated with a 2.01% decrease in LVEF (95% CI −3.84 to −0.18, p=0.03). We also observed a significant relationship between ANGPT2 and LVESVi, such that increasing ANGPT2 concentrations over time were associated with increased LVESVi and LVEDVi, with similar though less robust trends for VEGFR1 and HGF (Supplemental Figure 3). Taken together, these findings indicate that increases in ANGPT2, VEGFR1, and HGF concentrations are related to adverse changes in cardiac size and function over time within a given patient (Take Home Visual Graphic).

We next sought to determine how ANGPT2, VEGFR1, and HGF concentrations change in patients with significant reverse remodeling, defined as a 15% relative decrease in LVESVi based on previous work.³² Among the 99 patients included, 51 (51.5%) participants in our cohort had at least a 15% decrease in LVESVi (Table 1). Patients with reverse remodeling were less likely to have an ischemic cardiomyopathy (33.3% vs 64.9%, p= 0.004), less likely to have had a prior myocardial infarction (23.5% vs 41.7%, p=0.0034), and had a shorter duration of HF (42.1 ± 76.3 months vs 90.6 ± 97.4 months, p=0.003). Importantly, there was no significant difference between the remodeling and non-remodeling groups for baseline mean LVEF (28.4 ± 10.4% vs 28% ± 9.6%, p=0.964) or baseline median NT-proBNP concentrations (2263 pg/mL, IQR 1322-3623 pg/mL vs 2556 pg/mL, IQR 1596-5668 pg/mL, p=0.22). When we considered the change in ANGPT2, VEGFR1, and HGF, we identified significant differences in the change in protein levels depending on whether the definition for reverse remodeling was met (Figure 3). Specifically, we found that concentrations for all three proteins decreased in participants with a ≥15% reduction in LVESVi, and there was a significant difference in these biomarkers over time between patients with and without reverse remodeling (unadjusted analyses). Interestingly, VEGFR1 concentrations tended to be higher at baseline in participants who went on to reverse remodeling compared to those without (median 149 pg/ml, IQR 115-236 pg/ml vs 134, IQR 102-182 pg/ml, p=0.06).

Figure 3: Changes in ANGPT2, VEGFR1, and HGF Concentrations Based on the Presence of Reverse Remodeling Over the Study Period.

Mean biomarker concentrations across the study period are plotted based on the presence (orange line) or absence (blue line) of reverse modeling. Error bars indicate SEM. P values are the results of Wilcoxon rank-sum testing for a difference in change in biomarker concentrations between patients with and without reverse remodeling at time points 0 and 12 months. Panel A: ANGPT2 Panel B: VEGFR1 Panel C: HGF Left Ventricular end systolic volume index (LVESVi).

Relationship between ANGPT2, VEGFR1, and HGF with NT-proBNP Concentrations

The correlation of ANGPT2, VEGFR1, and HGF concentrations with cardiac structure and function is similar to associations observed with NT-proBNP.³⁰ Thus, we assessed the relationship between each biomarker and NT-proBNP concentrations (Supplemental Figure 4). We found that ANGPT2 concentrations were weakly, but significantly, correlated with NT-proBNP at baseline (R=0.32, p=0.001), and moderately correlated at 12 months and when assessed as change from baseline in both proteins (R=0.59 and 0.42, respectively, p<0.001 for both). VEGFR1 concentrations were weakly, but significantly, correlated with baseline, 12-month, and the change in NT-proBNP concentrations. HGF was only associated with NT-proBNP level at 12 months (R=0.40, p<0.001).

Higher ANGPT2, VEGFR1, and HGF Concentrations are Associated with Poor Outcomes

Morphologic and functional cardiac parameters are highly prognostic of outcomes in patients with HFrEF, and we sought to determine whether ANGPT2, VEGFR1, and HGF concentrations might be similarly prognostic. After grouping patients into tertiles for each biomarker at 12 months, we found that participants in the highest tertile of each biomarker had a significantly increased incidence of a composite endpoint of all-cause death and need for HHF (Figure 4). When we analyzed each marker for associations with each component of the composite endpoint, those with the highest tertile of ANGPT2 or HGF concentrations had a significantly increased incidence of both death and HHF (Supplemental Figure 5. By contrast, adverse outcomes for individuals with the highest tertile of VEGFR1 were primarily the result an increased incidence of HHF (Supplemental Figure 5). Overall, elevated ANGPT2, VEGFR1, and HGF concentrations are highly associated with adverse outcomes in this cohort of patients with HFrEF.

Figure 4: Association of ANGPT2, VEGFR1, and HGF Concentrations with a Composite of Death and Hospitalization for Heart Failure.

P value is the result of log-rank testing. Note that this was a landmark analysis, and all patients necessarily survived for the 12 months prior to these events; patients may have been hospitalized during this initial 12 month period).

When we assessed the cause-specific hazard ratios for all-cause death or HHF, and relative hazard ratio for a composite endpoint of all-cause death and HHF for ANGPT2, VEGFR1, HGF, and NT-proBNP, we found that 12-month concentrations for each of these biomarkers was highly associated with the composite outcome. When considering individual components of the composite, ANGPT2, HGF, and NT-proBNP concentrations were associated with both the hazard of death and HHF. However, VEGFR1 concentrations were primarily driven by an association with the incidence of HHF (Table 2).

Table 2. Hazard for Death and Hospitalization for Heart Failure by Biomarker Concentration.

Hazard ratios are shown with 95% confidence intervals and are per 2-fold change in each biomarker. Cause-specific HRs are provided for death and hospitalization for heart failure (HHF). Blue denotes p <0.05; purple <0.01, and green <0.001

	Composite		Death		HHF
	HR	p value	HR	p value	HR	p value
ANGPT2	2.76 (1.73 to 4.40)	<0.001	2.98 (1.57 to 5.67)	<0.001	3.02 (1.77 to 5.14)	<0.001
VEGFR1	1.76 (1.11 to 2.79)	0.016	1.62 (0.86 to 2.04)	0.14	1.98 (1.25 to 3.12)	0.003
HGF	4.04 (2.19 to 7.44)	<0.001	3.02 (1.58 to 5.80)	< 0.001	5.28 (2.60 to 10.7)	<0.001
NT-proBNP	2.28 (1.55 to 3.35)	<0.001	2.11 (1.29 to 3.44)	0.003	2.01 (1.35 to 3.00)	<0.001

Discussion

Much of the biology of reverse cardiac remodeling has focused on changes to the cardiomyocyte. For example, decreases in cardiomyocyte size, increased β-adrenergic receptor density, and improved calcium handling have all been reported with reverse remodeling.³⁸ Importantly, the fetal gene expression program that is induced in failing myocytes is attenuated, likely reflected clinically by decreased NT-proBNP concentrations during reverse remodeling.³¹ Other work has suggested that restoration of microvascular density occurs with reverse remodeling.¹¹ Thus, we hypothesized that systemic concentrations of angiogenic factors are associated with reverse remodeling.

In this study, we assessed concentrations of 25 modulators of endothelial homeostasis and function for their association with cardiac structure and function in a cohort of patients with HFrEF.³⁹ While we found several of these proteins to be associated with different echocardiographic parameters, ANGPT2, VEGFR1, and HGF were most consistently correlated with a broad array of echocardiographic parameters, such that higher concentrations of these biomarkers corresponded to larger cardiac volumes and worse function (Figure 2, Supplemental Table 2). Furthermore, we found that changes in ANGPT2 and, to a lesser extent, VEGFR1 and HGF, over time paralleled changes in markers of cardiac function (Supplemental Figure 3), and that changes in ANGPT2, VEGFR1 and HGF over time were associated with a clinically significant measure of reverse remodeling (Figure 3). Finally, we observed that ANGPT2, VEGFR1, and HGF concentrations are highly associated with a composite outcome of death and HHF (Figure 4). Taken together, these findings strongly support the role of ANGPT2, VEGFR1, and HGF as important mediators of progressive HF, with prognostic significance.

Our results are concordant with previous studies. For each factor, prior work has suggested higher concentrations are related to acute decompensation (ANGPT2²⁸ and HGF⁴⁰) or to worse clinical outcomes of patients with HFrEF (ANGPT2^41,42 HGF⁴³ and VEGFR1^44–47). However, to our knowledge, our work is the first to link these factors with reverse remodeling. ANGPT2, in particular, is an emerging biomarker for HFrEF. Peplinski et al recently reported that higher ANGPT2 concentrations are correlated with increases of pulmonary capillary wedge pressure and right atrial pressures in patients with left-sided heart failure,³⁹ and Haddad et al have additionally reported an association between ANGPT2 and presence of pulmonary hypertension,⁴⁸ findings that parallel our association of ANGPT2 with a higher RVSP and larger right atrial size. Additionally, previous reports have shown that patients with HFrEF have significantly higher concentrations of ANGPT2 than the general population,²⁷ even suggestive of a potential screening role.

Mechanistically, our work is suggestive of a blunted angiogenic program in the failing heart. Depending on its context, ANGPT2 can either stimulate angiogenesis, as in tumors, or inhibit endothelial cell proliferation.⁴⁹ In the heart, pre-clinical work indicates that ANGPT2 causally mediates microvascular regression after injury.⁵⁰ Mice subjected to experimental myocardial infarction demonstrate increased expression of Angpt2 in the border zone by cardiac endothelial cells and macrophages. Inhibition of ANGPT2 by either genetic deletion or treatment with an anti-ANGPT2 antibody prevents capillary loss and mitigates ventricular dysfunction after infarction. As a receptor, VEGFR1 can mediate vascular growth. However, the sFLT1 isoform of VEGFR1 that was measured here, is secreted into circulation where it limits angiogenesis by binding and trapping VEGF. Interestingly, VEGFA was the only protein for which higher concentrations were associated with better ventricular function and smaller chamber sizes (Figure 2, Supplemental Table 2). We noted that VEGFR1 concentrations trended to a higher concentration in individuals that would experience reverse remodeling. Hence, VEGFR1 may be a marker of at-risk, but recoverable myocardium. HGF, by contrast, is generally considered to be a mitogen for multiple cell types, including endothelial cells, in vitro and in pre-clinical models. ^51,52 HGF has been proposed previously as a potential therapeutic for ischemic heart failure.^51,53,54 However, our work and others linking higher HGF concentrations with more advanced heart failure raises questions about this approach.

We acknowledge several limitations to this exploratory work. First, our study was based on a subset of the GUIDE-IT study and is subject to biases of secondary analyses. Second, the absolute size of our cohort was relatively small, with relatively few events and an unusually high rate of reverse remodeling. Thus, our work may have been underpowered to identify significant relationships of additional proteins, with echocardiographic parameters or HFrEF outcomes. For example, whether VEGFR1 concentrations can be used to identify individuals likely to remodel should be explored further. Sample size and event rates also limited further analyses on the relationship between angiogenic proteins and GDMT use and the use of multivariable models to adjust for NT-proBNP and other clinical parameters know to be prognostic in HFrEF. Third, our studies focused on a handful of modulators of endothelial growth and maintenance with well-developed assays, but additional proteins are likely to be active in HFrEF and may have similarly predictive roles.

Taken together, our work supports a role for circulating modulators of endothelial homeostasis as mediators of HFrEF. Future work is needed to verify the ability of ANGPT2, VEGFR1, and HGF to prognosticate outcomes in patients with HFrEF in a larger dataset, and to understand if and how these protein concentrations change in response to GDMT for HF. Mechanistically, we speculate that circulating angiokines could link progressive HFrEF with vascular rarefaction. In addition to testing a broader panel of proteins on additional patient populations, fundamental work is needed to define the mechanisms that link endothelial function to progressive HFrEF. In particular, the effects of circulating angiokines ANGPT2 and VEGFR1 on the coronary microvasculature needs to be established. Nevertheless, our findings suggest that angiokines may be an important class of biomarkers in HFrEF, and potential mediators of progressive heart failure.

Conclusion

We found that circulating regulators of endothelial homeostasis and function have varying concentrations among patients with HFrEF, with dynamics that are closely associated with differences in cardiac structure and function. Of the factors assayed, ANGPT2, VEGFR1, and HGF concentrations were most robustly related to cardiac morphologic and functional parameters and were highly associated with the need for hospitalization for heart failure (HHF) and death. Our work suggests that dynamic changes in circulating modulators of endothelial function could play a role in the regulation of cardiac function in HFrEF and adds to the growing body of literature supporting circulating angiokines as prognostic biomarkers in HFrEF.

Clinical Perspective.

A decrease in cardiac capillary density has been associated with worsening heart failure. We identified associations between concentrations of proteins known to affect endothelial function with cardiac size and function in patients with HFrEF. Of measured factors, we found ANGPT2, VEGFR1, and HGF concentrations to be closely associated with ventricular remodeling and to be associated with future death and hospitalization for HFrEF. Our work suggests that endothelial remodeling is an important modifier of HFrEF and identifies new biomarkers with the potential to predict both clinical outcomes and reverse remodeling in HFrEF.

Abbreviations

ANGPT2

Angiopoietin 2

VEGFR1

Vascular endothelial growth factor 1

HGF

Hepatocyte growth factor

GLS

global longitudinal strain

HHF

hospitalization for heart failure

HFrEF

heart failure with reduced ejection fraction

LVEF

left ventricular ejection fraction

LVEDVi

left ventricular end diastolic index

LVESVi

left ventricular end systolic index

RVSP

right ventricular systolic pressure

TAPSE

tricuspid annular plane systolic excursion

Biography