Characterization of a robust mouse model of heart failure with preserved ejection fraction

Abstract

Heart failure (HF) is a leading cause of morbidity and mortality particularly in older adults and patients with multiple metabolic comorbidities. Heart failure with preserved ejection fraction (HFpEF) is a clinical syndrome with multisystem organ dysfunction in which patients develop symptoms of HF as a result of high left ventricular (LV) diastolic pressure in the context of normal or near normal LV ejection fraction (LVEF; ≥50%). Challenges to create and reproduce a robust rodent phenotype that recapitulates the multiple comorbidities that exist in this syndrome explain the presence of various animal models that fail to satisfy all the criteria of HFpEF. Using a continuous infusion of angiotensin II and phenylephrine (ANG II/PE), we demonstrate a strong HFpEF phenotype satisfying major clinically relevant manifestations and criteria of this pathology, including exercise intolerance, pulmonary edema, concentric myocardial hypertrophy, diastolic dysfunction, histological signs of microvascular impairment, and fibrosis. Conventional echocardiographic analysis of diastolic dysfunction identified early stages of HFpEF development and speckle tracking echocardiography analysis including the left atrium (LA) identified strain abnormalities indicative of contraction-relaxation cycle impairment. Diastolic dysfunction was validated by retrograde cardiac catheterization and analysis of LV end-diastolic pressure (LVEDP). Among mice that developed HFpEF, two major subgroups were identified with predominantly perivascular fibrosis and interstitial myocardial fibrosis. In addition to major phenotypic criteria of HFpEF that were evident at early stages of this model (3 and 10 days), accompanying RNAseq data demonstrate activation of pathways associated with myocardial metabolic changes, inflammation, activation of extracellular matrix (ECM) deposition, microvascular rarefaction, and pressure- and volume-related myocardial stress.

NEW & NOTEWORTHY Heart failure with preserved ejection fraction (HFpEF) is an emerging epidemic affecting up to half of patients with heart failure. Here we used a chronic angiotensin II/phenylephrine (ANG II/PE) infusion model and instituted an updated algorithm for HFpEF assessment. Given the simplicity in generating this model, it may become a useful tool for investigating pathogenic mechanisms, identification of diagnostic markers, and for drug discovery aimed at both prevention and treatment of HFpEF.

INTRODUCTION

Heart failure (HF) is a leading cause of morbidity and mortality particularly in older adults and patients with multiple metabolic comorbidities (1–4). HF can be distinguished clinically as HF with preserved ejection fraction (HFpEF) and HF with reduced ejection fraction (HFrEF), with both having distinct etiologies and pathophysiological mechanisms. Approximately half of all patients (with most of them over 65 yr of age and predominantly female) suffering from HF will have HFpEF (1, 5, 6). Although HFpEF is more common in females, males are more likely to die from HFpEF (5, 7–9). The pathophysiology of HFpEF is incompletely understood and may go unnoticed at initial stages of its clinical course leading to delayed diagnosis and therapeutic intervention (2, 10). Hemodynamic compromise in HFpEF often results in conditions that lead to fluid retention (pulmonary edema, systemic fluid accumulation) and symptomatology identical to HFrEF including exercise intolerance, dyspnea, and respiratory distress. Despite similar mortality and hospitalization rates among patients with HFpEF and HFrEF, these two entities have distinctive pathophysiological etiologies. Although HFrEF is largely characterized by primary deficits in cardiac contractile function, HFpEF is associated with dysfunctional myocardial relaxation and impaired ventricular filling because of chronic systemic changes that influence myocardial function (11, 12).

Pathophysiological Hallmarks of HFpEF and Challenges for Translational Research

Challenges in the recognition of HFpEF arise from its widely acknowledged origin as a multisystemic disorder. Previously thought to be a disease primarily affecting elderly woman with systolic hypertension and a small-volume hypertrophied heart, additional abnormalities including obesity, metabolic syndrome, and diabetes have shifted the “elderly” paradigm of HFpEF to a younger generation (13–17). To develop a mechanistic understanding of HFpEF pathogenesis, several hypothetical pathways have been proposed that focus on the roles of underlying hypertension, cardiac fibrosis, inflammatory responses, capillary insufficiency, metabolic derangements, and abnormal cGMP signaling (18). These abundant and heterogeneous pathophysiological hallmarks result in significant challenges in creating a single animal model of HFpEF that recapitulates a comprehensive human HFpEF clinical scenario.

Modeling and Identifying Major Features That Satisfy HFpEF Phenotypes

Despite numerous animal models showing cardiac hypertrophy and fibrosis, an “ideal” HFpEF mouse model should also demonstrate exercise intolerance, capillary rarefaction, diastolic dysfunction, and an alteration in lung physiology (12, 19, 20). Both large and small animal models of HFpEF have been developed (Table 1); however, small animal models have the advantage of lower cost, higher throughput enabling “omics” analyses, and the potential for genetic manipulation. Although the aldosterone-infused uninephrectomized mouse, the db/db mouse, and the ZSF1 rat fulfill many of the criteria for a HFpEF model, they require surgical manipulation or specific genetic backgrounds that limit their general use. The angiotensin II and phenylephrine (ANG II/PE) model (12, 20) has not been comprehensively evaluated as a model for HFpEF. We therefore performed a rigorous longitudinal characterization of the ANG II/PE model using comprehensive traditional phenotyping methods, as well as using additional methodologies such as advanced and novel echocardiographic analysis, multiorgan examination including histological analysis, and measurements of cardiovascular injury markers (Fig. 1A).

Table 1.

Animal model of HFpEF classified by animal type and subclassified by animal comorbidity

Model/Stimuli	Absolute Criteria for HFpEF (Conventional HFpEF Criteria)1	Additional and Novel HFpEF Criteria and Advantages of the Model (Nonconventional and Novel HFpEF Criteria)2	Mechanism and Pathophysiology of the Model (Major Pathophysiological Processes of HFpEF)3	Disadvantages of the Model4	Ref.
Comprehensive multi-hit murine HFpEF models					(12)
ANG II(1.5 µg/kg/day) + PE(50 µg/kg/day)	⌣	-Additional criteria present-Translatable to the bedside	-Multiple pathophysiological mechanisms-Specific mechanism for the model not discussed	-Attrition rate < 5%-Lacks metabolic component of HFpEF
HFD + l-NAME	⌣	-Additional criteria present	-Metabolic + hypertrophy + fibrosis-Specific mechanism discussed	-Attrition rate not discussed-l-NAME less applicable to human HFpEF pathophysiology	(21)
Disease-specific murine HFpEF models
Renal insufficiency/CKD
Subtotal nephrectomy	Models are not focused on HFpEF. Intended as a model for heart/renal failure and interorgan cross talk	Excellent review on multiple mouse and rat models that underwent nephrectomy. Advantage – combination of both renal and heart failure	-Hypertension and activation of RAAS	-Difficulty extrapolating to non-CKD-induced HFpEF ⌣	(22)
Subtotal nephrectomy	Majority of the major criteria of HFpEF are not discussed ⌣	Advantage – combination of both renal and heart failure	-Hypertension and activation of RAAS.	-Difficulty extrapolating to non-CKD-induced HFpEF ⌣	(23)
DOCA + angiotensin	Majority of the major criteria of HFpEF are not discussed	Histological myocardial assessment	-Hypertension + activation of RAAS.	-Attrition rate not discussed-Combination of DOCA + activation of RAAS is not easily translatable to bedside ⌣	(24)
Diabetes
db/db mouse	Majority of the major criteria of HFpEF are not discussed	Cardiac MRI. Systolic and diastolic hemodynamic parameters on ex vivo isolated heart	-Metabolic + fibrosis	-Attrition not discussed.-Hypertension did not present ⌣	(25)
Metabolic
Hyperlipidemia	Cardiac catheterization not performed	Histological assessment of cardiomyocyte size	-Metabolic + fibrosis-Lipotoxicity as a specific mechanism	-Attrition - 100% by 12 wk-Metabolic derangements as a hit factor ⌣	(26)
Hypertension
TAC model	Narrow margin on transitioning between HFpEF and HFrEF (HFpEF⇔HFrEF) ⌣	N/A	-Myocardial hypertrophy, fibrosis-AMPK pathway as a specific mechanism discussed	-Attrition - 30 % by 13th wk-Elevated BP due to nonphysiological processes ⌣	(27)
TAC model	All major criteria (physical intolerance, pulmonary edema etc.) of HFpEF not presented	Histological analysis of the myocardium	-Hypertrophy and fibrosis-PDE1C deficiency as a specific mechanism	-Attrition not discussed-Elevated BP due to nonphysiological mechanical processes ⌣	(28)
Comprehensive large-animal (swine) HFpEF models					(29)
Disease-specific HFpEF models
Hypertension
Phenylephrine infusion	All major criteria (physical intolerance, pulmonary edema etc.) of HFpEF not present	Cardiac MDCT imaging, histopathology, PV loops analysis	-Myocardial hypertrophy and fibrosis-Specific mechanism for the model is not discussed	-Elevated BP due to clinically irrelevant drug infusion ⌣	(30)
TAC model	All major criteria (physical intolerance, pulmonary edema etc.) of HFpEF not present	Cardiac cath and PV loops analysis, histopathology of the myocardium	-Myocardial fibrosis and hypertrophy-Specific mechanism for the model is not discussed	-Narrow margin on transitioning between HFpEF and HFrEF ⌣	(31, 32)
Multi-hit approach models
Göttingen Miniswine model + DOCA	⌣	Multisystemic involvement including kidneys, liver, lungs	-Multiple pathophysiological mechanisms	-Attrition rate > 30%-Longevity of follow up 20 wk ⌣	(33)
TAC model + metabolic stress	All major criteria (physical intolerance, pulmonary edema etc) of HFpEF not presented	N/A	-Pressure load and metabolic	-Attrition rate > 35% ⌣	(34)

ANG II, angiotensin II; BP, blood pressure; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; l-NAME, N^G-nitro-l-arginine methyl ester; PE, phenylephrine; PV, pulmonary veins; RAAS, renin-angiotensin-aldosterone system; RAS, renin-angiotensin system; TAC, transverse aortic constriction.

¹Echo criteria (E/E′, E/A, concentric hypertrophy, EF > 50%), increased left ventricular end diastolic pressure (LVEDP), exercise intolerance, pulmonary congestion; ²Multisystem organ involvement; nonconventional echocardiographic criteria [speckle tracking analysis, left atrium (LA) assessment]; cardiac catheterization with pressure-volume loop analysis; histological analysis; coronary flow reserve; ³Fibrosis, inflammation, myocardial hypertrophy, capillary rarefaction; ⁴Attrition rate; applicability to human pathophysiology; reference to HFpEF as only diastolic dysfunction vs. systemic disease.

Figure 1.

Mouse model and criteria for heart failure with preserved ejection fraction (HFpEF). A: major clinical, physiological, and histological features to fulfill a HFpEF model phenotype. B: schematic for angiotensin II and phenylephrine (ANG II/PE) treatment in which minipumps were inserted on day 0 and cardiac phenotypes were assessed at days 3, 10, and 28 after initiation of ANG II/PE infusion. RNA-seq was performed at days 3 and 10 after initiation of ANG II/PE infusion. C: physical endurance assessed by treadmill demonstrated decreased distance traveled in ANG II/PE-treated mice compared with control mice at 28 days. D and E: quantitative analysis of heart weights normalized to mouse body weight (D) and to tibial length (E). F: M-mode echocardiographic analysis (short axis) showing concentric cardiac hypertrophy in ANG II/PE-treated mice compared with control mice at 28 days. ***P < 0.001. G: quantitation of diastolic left ventricular (LV) wall thickness from echocardiographic data showing concentric LV hypertrophy. H: quantification of serum NH₂-terminal pro-brain natriuretic peptide (NT-proBNP) showing elevation in chronic ANG II/PE infusion. *P < 0.05, n = 5 for control and 14 for ANG II/PE-treated animals. Serum cytokine levels are in Supplemental Fig. S1.

Several experimental endeavors have focused on the development of myocardial fibrosis (one of the major features of HFpEF) and aspects of its treatment (35–37). However, other clinical criteria of HFpEF phenotypes were not investigated or satisfied. Many animal models lack some of the clinically significant common comorbidities that are associated with and could trigger human HFpEF. For example, the transverse aortic constriction (TAC) model robustly induces fibrosis but has a limitation in the ability to model the slow onset of HFpEF seen in humans and variability in the transition from HFpEF to HFrEF can lead to phenotype heterogeneity. On the other hand, administration of nonphysiological prohypertensive drugs [as in the N^G-nitro-l-arginine methyl ester (l-NAME) plus high-fat diet model] limits its relevance to clinically applicable scenarios. In the context of multiorgan failure in HFpEF, one could create a distinct model for each HFpEF phenotype depending on the predominant etiological stressor. In a recently described mouse model of HFpEF driven by hyperlipidemia (26), hypertension (a major stressor of HFpEF) was absent and multiorgan involvement was not observed. Other models that used ANG II with or without phenylephrine infusion lacked key echocardiographic criteria for HFpEF, as well as clinical correlation with physical endurance and multiorgan involvement (37).

In several studies, the ANG II/PE model was shown to induce robust cardiac fibrosis and an increased left ventricular (LV) weight ratio and lung wet weight; however, echocardiographic analysis to establish HFpEF was not reported (35, 38–42). In studies where only ANG II was infused into mice or rats, animals developed echocardiographic and histological (fibrosis) phenotypes consistent with HFpEF (43–45). However, these studies did not demonstrate decreased ejection fraction or rigorously address multiorgan involvement or exercise tolerance.

Experimental Model for the Present Studies

Our rationale for using the ANG II/PE-induced mouse model was to mimic increased cardiac workload and myocardial wall tension by increasing afterload (through selective α₁-adrenergic receptor agonism by phenylephrine and AT₁-receptor agonism via ANG II), and simultaneously creating ANG II-induced endothelium-dependent microvascular dysfunction (46), ANG II-aldosterone-dependent fluid retention and elevated cardiac preload, and an ANG II-regulated proinflammatory state, all of which are major pathophysiological features that are present in human HFpEF (47). In addition, hypertension-induced myocardial wall tension will lead to elevated metabolic demands and oxygen consumption resulting in chronic myocardial stress, and elevated afterload will drive left ventricular hypertrophy (LVH) and worsened LV myocardial compliance that will result in an altered ventricular pressure-volume relationship, culminating in diastolic dysfunction.

ANG II is an important stimulus of NADPH oxidase, an enzyme known for activation of production of mitochondrial reactive oxygen species (ROS) (48) and suppression of endothelial nitric oxide synthase (eNOS) resulting in lower levels of nitric oxide (NO) (49). Nitric oxide regulates vascular tone and blood flow by activating soluble guanylate cyclase (sGC) in vascular smooth muscle cells and can desensitize cardiac myofilaments leading to an increase in diastolic fiber length, improving Frank-Starling mechanisms of myocardial diastolic reserves (50, 51). Environments that exacerbate ROS production and proinflammatory endothelial cell phenotypes in the coronary microvasculature will lead to further reduction in NO, depression of cGMP pathways, impairment of active myocardial relaxation, and worsening of passive LV stiffness, which is thought to result from changes in cardiomyocyte diameter, density, and increased fibrosis (52, 53).

Noninvasive Methods to Assess Cardiac Diastolic Function

In addition to invasive diagnostic procedures (cardiac catheterization and myocardial biopsy), it is also essential to establish and validate noninvasive methods such as echocardiography to accurately evaluate diastolic dysfunction (a major clinical component of HFpEF). Speckle tracking echocardiography (STE) has emerged in recent years as a robust and clinically useful noninvasive image analysis technique to evaluate ventricular function in patients with HFpEF, and adaptation of this technique to quantify left atrial (LA) function in clinical investigations has been shown to add invaluable data to support the early diagnosis and treatment of HFpEF (54). To our knowledge, the application of STE to evaluate LA structure and function in mouse models has not been reported. Therefore, in our comprehensive characterization of the HFpEF phenotype, we extended the use of STE beyond quantitation of LV structure and function and demonstrate that similar to human studies, STE of the LA is a powerful tool to evaluate associated LA structural and functional abnormalities in mouse models.

METHODS

Mice

Mice were housed in a pathogen-free facility and handled in accordance with standard use protocols, animal welfare regulations, and the NIH’s Guide for the Care and Use of Laboratory Animals. Mice were housed with a 12-h:12-h light/dark cycle in a temperature (22 ± 1°C)- and humidity (55 ± 5%)-controlled room. Mice were allowed free access to water and a standard laboratory mouse diet (PicoLab Rodent Diet 20, Cat. No. 007688). All protocols were approved by the Washington University Animal Studies Committee. All mice were maintained on a C57BL/6J; 129X1 mixed, or F1 hybrid genetic background. Both male and female mice were used in these studies with approximately equal distribution.

Mouse Model of Heart Failure with Preserved Ejection Fraction

Mice (8- to 10-wk old) were implanted with an osmotic minipump (Alzet, 200 μL, Cat. No. 2004) to allow subcutaneous infusion of ANG II/PE for 28 days (Fig. 1B). Additional batches of mice were also implanted with smaller minipumps (Alzet 100 μL, Cat. No. 1002) to provide delivery of ANG II/PE for 3 or 10 days. Mice were randomly selected and assigned to different groups: control mice (either no pump implantation or implantation of a pump filled with saline), and mice implanted with an infusion pump with ANG II/PE. Osmotic minipumps were programmed to deliver ANG II (A9525, Sigma-Aldrich, 1.5 μg/g/day) and phenylephrine HCl (P6126, Sigma-Aldrich, 50 μg/g/day), consistent with previous studies (35, 39, 42). Loaded minipumps were primed in 0.9% NaCl for 24 h at 37°C before implantation.

Exercise Tolerance

Treadmill (Columbus Instruments Model Exer3/6 Treadmill) exercise was assessed in representative groups of animals at 28 days after initiation of the experiment following 3 days of acclimation to the treadmill. Exercise tests were carried out starting at a speed of 10 m/min for 3 min, with speed increasing to 13 m/min for 3 min, and subsequently increasing by 3 m/min every 3 min. Electrified bars delivered a mild shock when a mouse failed to keep pace and exhaustion was defined as the inability of the animal to return to running within 5 s after receiving the mild electric shock (set at 0.5 mA, 2 Hz). Mice were run until exhaustion. Total running time and distance were subsequently analyzed.

Echocardiography

Transthoracic echocardiography with conventional two-dimensional (2-D) and Doppler imaging and speckle tracking echo analysis was performed using a VisualSonics 3100 high-resolution in vivo imaging system at 0, 10, and 28 days, as previously described (55). Cardiac images were obtained by a handheld technique using tribromoethanol (100 mg/kg ip; Avertin) anesthetic in unrestrained mice. Note that this anesthetic dose is one-third of the dose used by Pachon et al. (56) and induces relatively light anesthesia with minimal effect on heart rate, which ranged between 600 and 650 beats/min.

Echocardiographic Imaging and Image Analysis of the Left Atrium

Beyond the obvious size difference, there are important anatomical differences between human and mouse hearts that need to be considered when imaging the left atrium (LA) using echo. Compared with human anatomy, the mouse chest is dome shaped, the heart is vertically oriented, it is medially rotated along its longitudinal axis, the shape of the LA and the anatomy of the pulmonary veins (PVs) are markedly different (57, 58). Because of these differences, the standard imaging planes adopted from human echo do not provide adequate visualization of the LA and PVs in mice. To overcome these limitations, we developed a modified long-axis image traversing the LA proper through its midsagittal plane by holding the ultrasound transducer in a left lower parasternal position with a caudal angulation. We used the following landmarks to obtain consistent high-quality 2-D images of the LA (Supplemental Fig. S1: see https://doi.org/10.6084/m9.figshare.23058137; Supplemental legends: see https://doi.org/10.6084/m9.figshare.22812788): two-chamber view of the LV with its long axis at an approximate 45° angle relative to the screen (LV), a clear midline sagittal view of the posterior and anterior mitral valve annulus (MA post, MA ant), a cross-sectional view of the coronary sinus posterior to the mitral annulus (CS), a slightly oblique longitudinal view of the proximal main pulmonary artery anterior to the mitral annulus (PA), and a fine adjustment of the imaging plane to obtain an uninterrupted view of the LA wall represented by a continuous line (tracked by the green line, Supplemental Fig. S1F) connecting the posterior and anterior mitral annulus while avoiding interruption of this line by the left atrial appendage anteriorly and by the PV posteriorly.

STE Analysis of LA Size and Function

Quantitative image analysis of the LA was performed using the same speckle-tracking algorithm implemented in the Vevo system that was used for LV strain analysis. Suitable 2-D cine loops of the LA were selected from digitally acquired echocardiographic images based on adequate visualization of the LA wall and absence of image artifacts. For timing purpose of the atrial cycle, the ventricular cycle was used as the point of reference starting at LV end diastole (marked by mitral valve closure and/or a small “bump” in the anatomical M-mode of the basal posterior LV wall during isovolumic contraction, Supplemental Fig. S1A). The time point between early and late diastole was determined by a small notch on the descending slope of the LA volume curve. Two consecutive cardiac cycles were selected for analysis based on image quality. Semiautomated tracing of the endocardial border was performed from the posterior mitral annulus to the anterior mitral annulus and verified across the two cardiac cycles and then corrected as needed to achieve good quality tracking throughout each cine loop. Tracked images were then automatically processed in a frame-by-frame manner to derive the following parameters: maximal and minimal LA volumes calculated by the disk summation method (LA max and LA min); LA volume before atrial contraction (LA vol pre-A); LA stroke volume (LA_max − LA_min), LA ejection fraction (stroke volume divided by LA_max × 100); LA expansion index [(LA_max – LA_min)/LA _min] corresponds to LA reservoir function; LA passive EF [(LA_max – LA pre-A)/LA_max] corresponds to LA conduit function; LA active EF [(LA pre-A − LA_min)/LA pre-A] corresponds to LA booster pump function; peak positive longitudinal strain (εs) corresponds to atrial reservoir function; strain during early and late diastole (εe and εa, respectively) corresponds to conduit and atrial booster function; peak strain rates during ventricular systole, early diastole, and late diastole (SR-S, -E, and -A, respectively) correspond to LA reservoir, conduit, and booster pump functions.

In addition, PV flow velocity was interrogated using the parasternal short-axis image with slight angulation to visualize by color Doppler with the ostium of the right inferior PV entering the LA in its posterior portion (Supplemental Fig. S1). Spectral-Doppler images were used to manually trace the systolic and diastolic portions of the PV velocities envelops.

The full range of echocardiographic parameters to assess diastolic function is in Supplemental Table S1 (see https://doi.org/10.6084/m9.figshare.23059265). All images were obtained by a single operator with expertise in mouse echocardiography and who was blinded to the treatment group.

Hemodynamic Analysis

Under general anesthesia (isoflurane) and full ventilatory support, retrograde catheterization was performed with a 1-Fr high-fidelity micromanometer pressure catheter (SciSense Advantage System, London, ON, Canada). Systolic and diastolic blood pressure (BP) were recorded at 3, 10, and 28 days in the right carotid artery and analyzed with SciSense software. After advancing the catheter into the LV, LV systolic and diastolic pressures were recorded at 28 days and analyzed with SciSense software.

Serum Protein Biomarkers and Cytokines

Serum was collected at the terminal stage of the experiment by collecting arterial blood from the carotid artery.

The Olink Target 96 Mouse Exploratory assay was used to assess changes in serum proteins involved in key biological processes in response to infusion of ANG II/PE. Relative levels of 92 biomarkers per sample (EDTA plasma) were simultaneously measured using the Olink Target 96 Mouse Exploratory Panel (Product Cat. No. 95380, Data file Cat. No. 3801, Olink Proteomics AB, Uppsala, Sweden). Full link to protein assay is as follows: https://olink.com/content/uploads/2021/09/1082-v1.0-mouse-exploratory-panel-content-final.pdf. Briefly, samples were incubated overnight at 4°C for 19 h with Olink antibody reagents. Following incubation, detection extension was completed using a PCRmax Alpha Cycler 4 quad-thermocycler (Item Cat. No. EW-93945-12, Cole-Parmer, Vernon Hills, IL), then stored overnight at −20°C. Sample measurements were detected using real-time qPCR on the Olink Signature Q100 (Product Cat. No. 96000, Olink Proteomics AB, Uppsala, Sweden). Data from the Olink Signature Q100 were imported into the Fluidigm Real-Time PCR Analysis Software (v.4.5.2). Fluidigm-chip QC was performed and a heatmap of threshold cycles (CT) signal was generated and saved as a Microsoft Excel (.csv) file. The (.csv) heatmap file was imported into Olink NPX Manager Software (v.3.1.0.393) where CT values were converted to an arbitrary unit in log2 scale known as normalized protein expression (NPX). Run QC, sample QC, and quality assessment were performed using Olink NPX Manager Software and pass/fail qualifications provided in the Olink Data Analysis User Guide (v.4.0). NPX data were exported in wide format into a Microsoft Excel (.xlsx) file before statistical analysis (see Statistical Analysis).

Cytokine analysis was performed using Invitrogen Cytokine & Chemokine 36-Plex Mouse ProcartaPlex Panel 1A (Invitrogen EPX36026092901). Serum FGF21 and FGF23 levels were measured by ELISA (Mouse/Rat FGF21; FGF23; R&D Systems, 60-6300). NH₂-terminal pro-brain natriuretic peptide (NT-proBNP) was detected by ELISA (Elabscience, E-EL-M0834).

Histology

Assessment of cardiomyocyte cross-sectional area, capillary density, and small vessel smooth muscle actin was performed as previously described on 6-µm paraffin sections stained with wheat germ agglutinin (WGA) (59–61). Quantitative analysis of cardiac fibrosis was performed on Masson’s trichrome-stained histological slides. Images showing regions of interest (ROI) were analyzed using ImageJ color thresholding to measure the blue versus red area. ROIs include whole LV coronal sections at the level of the papillary muscles. Qualitative analysis of renal fibrosis used a semiquantitative histological fibrosis scale applied to four ROIs, containing blood vessels, in the renal cortex. Score 0 was allocated to samples with no fibrosis and scores 1–4 were assigned to samples with the presence of fibrosis on 1 to 4 ROIs, respectively (Supplemental Fig. S2: see https://doi.org/10.6084/m9.figshare.23058515).

Immunofluorescence

Histological sections (6 μm) were prepared from paraffin-embedded tissues. Sections were deparaffinized and rehydrated. Antigen retrieval was performed using a pressure cooker (15 min) and citrate buffer (pH 6.0). Tissues were blocked with 5% goat or donkey serum. For immunofluorescence, the following primary antibodies were used: collagen 1 (Abcam 1:100, ab 138492, Lot GR3370247-5), α-smooth muscle actin (α-SMA, 1:200, Dako North America, M0851), platelet endothelial cell adhesion molecule 1 (PECAM1, CD31, 1:50, Dianova, Dia310), CD45 (1:100, Bioscience, Lot 553380), Mac-2 (1:100, Invitrogen, Bioscience, ref 14-5301-85, Lot 2153405), Ly6g (1:100, Abcam, ab25377; Lot GR3443546-1), and cardiac troponin I (Abcam 1:100, ab 47003, Lot GR3248433-1). Antibodies were added to the blocking buffer and slides were incubated overnight at 4°C. Slides were washed with PBS, and Alexa Fluor-conjugated secondary antibodies (1:200, Thermo Fisher) were added for 30 min at room temperature. Immunofluorescent imaging was performed using a Zeiss Apotome II, and image processing was performed using Zeiss Axioplan software and ImageJ. Digital scanning of whole slides was performed using a Zeiss Axio Scan.Z1.

Cell Death Analysis

Paraffin sections (6 μm) were assayed by DeadEnd Fluorometric TUNEL System (No. G3250, Promega). Slides were mounted with 4′,6-diamidino-2-phenylindole (DAPI) containing VECTASHIELD mounting medium (No. H-1200, Vector) for fluorescent detection.

RNA-seq Analysis

Three days after initiating NaCl or ANG II/PE infusion, elevated systolic BP was confirmed by carotid artery cannulation. Mice were then exsanguinated under general anesthesia followed by cervical dislocation. Hearts were excised, and the whole LV was isolated and rinsed in PBS and then snap-frozen in liquid N₂. Tissue was stored at −80°C until processing. mRNA was isolated using the Ambion PureLink RNA Mini Kit (Thermo Fisher, 12183018, Lot 2379554). Samples were submitted for library preparation and sequencing through the Washington University Genome Technology Access Center (GTAC, https://gtac.wustl.edu/). Total RNA integrity was determined using an Agilent Bioanalyzer. Library preparation was performed with 0.5–1 µg of total RNA. Ribosomal RNA was removed using RiboErase kit (Kapa Biosystems). mRNA was then fragmented in reverse transcriptase buffer and heated to 94°C for 8 min. mRNA was reverse transcribed to yield cDNA using SuperScript III RT enzyme (Thermo Fisher-Life Technologies, per manufacturer’s instructions) and random hexamers. A second strand reaction was performed to yield ds-cDNA. cDNA was blunt-ended and had an A base added to the 3′ ends, followed by ligation of Illumina sequencing adapters. Ligated fragments were amplified for 12–15 cycles using primers incorporating unique dual index tags. Fragments were sequenced on an Illumina NovaSeq-6000 flow cell using paired-end reads extending 150 bases. Reads were processed using an in-house pipeline and open-source R packages. Briefly, raw reads were first trimmed using Cutadapt (v.3.5) to remove low-quality bases and reads. Trimmed reads were then aligned to the mouse genome mm10 with GENCODE annotation vM25 using STAR (v.2.7.9) with default parameters. Transcript quantification was performed using featureCounts from the Subread package (v.2.0.1). Further quality control assessments were made using RSeQC (v.4.0.0) and RSEM (v.1.3.1), and batch correction was performed using edgeR, EDASeq, and RUVSeq (Bioconductor version: release 3.16).

Principal component analysis (PCA) and differential expression analysis for samples were determined using DESeq2 in negative binomial mode using batch-corrected transcripts from feature counts (>2-fold expression change, >1.5 counts/million, and Benjamini corrected P < 0.05). Pairwise comparisons were made between control (CTL) and ANG II/PE infused (treated, TRT) mice to identify differentially expressed genes (DEGs). Gene ontology (GO) analyses were performed using Enrichr (62, 63) for DEGs from treated (TRT) versus control (CTL) groups. DEGs were grouped into different pathways, and their relative expression was plotted using ggplot2 and heatmap.

Differentially expressed genes, downloaded from Hahn et al. (64) and Das et al. (65), were further selected (Benjamini corrected P < 0.05) for KEGG and GO analyses, which were later compared with DEGs from day 3 ANG II/PE data sets.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism 9 software. Data are shown as means ± SE. For comparisons of multiple experimental groups, one-way ANOVA was used followed by post hoc Tukey’s tests for pairwise comparisons. Student’s t tests were used for comparisons of two groups. Data with a P < 0.05 were considered statistically significant.

The Olink NPX data include 92 analytes from 100 samples. One of the samples was labeled as “warning” by Olink QC. PCA and outlier analysis were implemented to verify that sample and detect any other potential outliers. No sample was detected as an outlier. Seven analytes were removed because the observed NPX value was below the limit of detection (BLOD) for more than half of the total samples and no treatment group observed NPX value was within the limit of detection for more than half of the samples within the treatment group. ANOVA, including the combination of treatments and time points as the fixed factor, was applied to the NPX data to test the comparisons of using R software (https://www.R-project.org/). For each comparison, the fold change and P values, as well as the false discovery rate (FDR) (66), were used to adjust for multiple testing across the analytes that were reported.

The following R packages were used for statistical analysis and results visualization of the protein biomarker panel. R software version 4.1.2; library(xlsx), library(stats), library(lsmeans), library(dplyr), library(tidyverse), library(openxlsx), library(ggplot2), library(reshape2), library(PEIP), library(scales), and library(gridExtra).

RESULTS

Features of HFpEF pathology in ANG II/PE infused mice were assessed beginning with clinical presentation and diagnostics followed by advanced echocardiographic analysis and hemodynamics, histopathology, and gene expression profiling as outlined in Fig. 1A.

Effects of Chronic ANG II/PE Infusion on Physical Endurance, Cardiac Hypertrophy, and Clinical Laboratory Measures

Heart failure in human patients is associated with reduced physical endurance. To validate this pathophysiological feature in mice infused with ANG II/PE, physical endurance was assessed by treadmill exercise tolerance. When compared with saline-infused controls, ANG II/PE-infused mice demonstrated a substantial decline in exercise tolerance, reaching stages of exhaustion at a shorter distance (Fig. 1C). An additional criterion of HFpEF in ANG II/PE-infused mice was the increase in heart weight [assessed by heart to body weight ratio (Fig. 1D) and heart to tibial length ratio (Fig. 1E)]. We also assessed concentric hypertrophy on short-axis M-mode echocardiography (Fig. 1F) that showed a relative increase in LV wall thickness (Fig. 1G) in ANG II/PE infused mice. Increased levels of serum NT-proBNP, a common marker for diagnosis and trend of severity of congestive heart failure, was demonstrated at 28 days in ANG II/PE-infused mice (Fig. 1H). Additional serum cytokine levels were measured after 10 and 28 days of ANG II/PE infusion. This analysis showed increased levels of the proinflammatory cytokines GCSF, IL6, and MCP3 at 10 days, and MCP3 and CXCL1 at 28 days (Supplemental Fig. S3A: see https://doi.org/10.6084/m9.figshare.23058659). The endocrine fibroblast growth factor, FGF21, was elevated at 10 and 28 days, and FGF23 was elevated at 3, 10, and 28 days (Supplemental Fig. S3B). Proteomic analysis of serum collected after 3, 10, and 28 days of ANG II/PE infusion revealed perturbed pathophysiological pathways that include: 1) Inflammation and immunomodulation (CCL20, CXCL1, EDA2R, IGSF3, IL10, IL17A, IL17F, IL1B, TNFRSF11B); 2) Mitochondrial homeostasis (ROS neutralizing, aerobic respiration) and myocardial contractility (GCG, ENO2 FLRT2, FSTL3, NADK, PARP1, PLIN1, PRDX5, QDPR, TNNI3, CCN4/WISP1); 3) Vascular homeostasis (FLI1, NOTCH, PLXNA4, VEGFD); and 4) Cell signaling, adhesion, and apoptosis (CASP3, CNTN1, CNTN4, EPO, FAS, FOXO1, NOTCH3, SNAP29) (Supplemental Table S2: see https://doi.org/10.6084/m9.figshare.23258096).

Effects of Chronic ANG II/PE Infusion on Multiple Organ Systems: Pulmonary and Renal Histopathology

To determine the effect of elevated filling pressure originating from myocardial diastolic dysfunction on the pulmonary parenchyma, lung wet and dry weights were measured (Fig. 2, A and B). Both absolute wet and dry lung weights were significantly higher in ANG II/PE-infused mice. Contrary to expectations, the wet-to-dry (w/d) lung weight ratio (a measure of pulmonary edema) was lower in treated compared with control mice (Fig. 2C). This phenomenon may be explained by the relatively higher increase in dry lung weight in treated animals resulting from the accumulation of additional extracellular matrix (ECM) and proteinaceous material in the lung parenchyma and alveoli which is consistent with the observed increased hydroxyproline levels in treated compared with control lungs (Fig. 2D). Histological analysis of lung sections stained with Masson’s trichrome showed increased alveolar membrane thickening, as well as interstitial collagen deposition in the treated mice (Fig. 2, E and F).

Figure 2.

Pulmonary and renal changes following chronic infusion of angiotensin II and phenylephrine (ANG II/PE). A–C: quantitative analysis of lung weight at 28 days showing increased wet (A) and dry (B) lung weight and a decreased wet-to-dry lung weight ratio (C) in ANG II/PE-treated mice compared with control mice. D: quantitative analysis of lung hydroxyproline levels in ANG II/PE-treated mice compared with control mice. E and F: Masson’s trichrome staining of the whole right lung lobe of control (E) and ANG II/PE-treated (F) mice, demonstrating interstitial collagen deposition and alveolar wall thickening in treated animals. Insets: magnification of the left lobe. Scale bars = 1 mm, 250 µm. G and H: Masson’s trichrome staining of the kidney of control (G) and ANG II/PE-treated (H) mice, demonstrating cortical layer perivascular fibrosis (green arrows) and arteriolar wall hypertrophy in treated animals. Insets: magnification of the renal cortical layer. Scale bars = 1 mm, 250 µm. I and J: semiqualitative analysis of renal fibrosis performed as described in methods and histological scoring as shown in Supplemental Fig. S2. Lung and kidney data (lung weight, hydroxyproline, and renal fibrosis analysis) represent n = 4–7 for control and n = 6–15 for ANG II/PE-treated mice. *P < 0.05; ***P < 0.001.

Multisystem organ involvement was substantiated by qualitative assessment of renal parenchyma stained with Masson’s trichrome, which revealed significant renal perivascular fibrosis and arteriolar hypertrophy (Fig. 2, G and H). Semiquantitative analysis of renal histology demonstrated significantly increased fibrosis in treated animals (Fig. 2, I and J, and Supplemental Fig. S2).

Effects of Chronic ANG II/PE Infusion on Echocardiographic and Hemodynamic Measures

Echocardiography was used as a noninvasive method to assess diastolic dysfunction and to confirm the maintenance of normal LV systolic function (Fig. 3, A–K). We combined a comparison of widely used echocardiographic criteria of diastolic function, including E/E′ ratio, Tei index, and E/A ratio (Fig. 3, A–F, and Supplemental Table S1), with more advanced and less common measures such as myocardial speckle tracking-based strain analysis (Fig. 3, G and H) and analysis of the LA (Fig. 3, I–K). We observed echocardiographic criteria of diastolic dysfunction starting as early as 10 days after initiation of ANG II/PE infusion, which was sustained at 28 days of treatment (Fig. 3, A–F). Interestingly, LV myocardial strain analysis after 10 days of ANG II/PE infusion did not demonstrate worsening but rather showed unchanged or increased mean values, which suggests an nitial compensatory adaptation of the myocardium to increased afterload (Fig. 3, G and H, and Supplemental Table S1). Interestingly, LA strain analysis at 10 days of ANG II/PE infusion showed a significant decrease in LA reservoir function, consistent with the notion that impairment of LA function is an early sign of LV diastolic dysfunction (Fig. 3K). By 28 days, ANG II/PE infused mice showed worsening of the LV and LA strain and strain rate, which suggests insufficiency of myocardial compensatory mechanisms (Fig. 3, G–K, Supplemental Table S1, and Supplemental Movies S1 and S2: see https://doi.org/10.6084/m9.figshare.23059262). Invasive methods of BP assessment showed a significant increase in systolic BP as early as 3 days of ANG II/PE infusion that continued to increase, reaching its peak at 28 days of ANG II/PE infusion (Supplemental Fig. S4: see https://doi.org/10.6084/m9.figshare.23058818). To confirm and correlate the parameters obtained by noninvasive measurements, we performed invasive assessment of LV hemodynamics, which is widely accepted as the standard for the diagnosis of diastolic dysfunction. After retrograde LV catheterization, quantitative analysis of LV end-diastolic pressure (LVEDP) demonstrated a significant increase in filling pressures after 28 days of ANG II/PE treatment compared with control mice (Fig. 3L). Analysis of echocardiographic criteria of diastolic dysfunction with respect to sex at 28 days showed no significant differences between male and female mice (Supplemental Table S1B).

Figure 3.

Echocardiographic and hemodynamic changes following chronic infusion of angiotensin II and phenylephrine (ANG II/PE). A–F: conventional echocardiographic evaluation of the left ventricular (LV) in ANG II/PE-treated mice demonstrated increased LV mass index (LVMI; A), decreased LV end-diastolic volume (LVEDV; B), increased LV ejection fraction (LVEF; C), increased E/E′ ratio (D), prolonged isovolumic relaxation time (IVRT; E), and increased Tei index (F). G–K: speckle tracking echocardiography (STE) of the LV in ANG II/PE mice showed normal LV peak longitudinal strain at 10 days but decreased longitudinal strain at 28 days (G). Similarly, LV peak longitudinal diastolic strain rate was maintained at 10 days but decreased at 28 days (H). Novel use of STE of the left atrium (LA) revealed LA enlargement (I) and decreased LA ejection fraction (J) in ANG II/PE-treated mice after 28 days and showed decreased LA reservoir strain at the 10- and 28-day time points (K). L: retrograde LV catheterization after 28 days of ANG II/PE infusion showed increased LV end-diastolic pressure (LVEDP). Data represent n = 6 for control and ANG II/PE-treated mice at 10 days and n = 18 for control and n = 21 for ANG II/PE-treated mice at 28 days. #P < 0.05, ##P < 0.05, ###P < 0.001, NaCl vs. ANG II/PE; *P < 0.05, **P < 0.05, ***P < 0.001, ANG II/PE treated 10 vs. 28 days. Additional echocardiographic parameters may be found in Supplemental Table S2.

Effects of Chronic ANG II/PE Infusion on the Development of Myocardial Fibrosis

Histological assessment of the LV demonstrated diffused myocardial fibrosis at 3, 10, and 28 days of ANG II/PE infusion with Masson’s trichrome staining of transverse cardiac sections (Fig. 4A). Maturation of the collagen extracellular matrix (ECM) was assessed by Picrosirius red staining (Fig. 4, B and C) and immunostaining for Col1 (Fig. 4D). Interestingly, despite demonstrating ECM deposition with Masson’s trichrome staining at days 3 and 10 of ANG II/PE infusion, Picrosirius red staining only showed diffused collagen deposition at 3 days of ANG II/PE and increasing regions of strong staining at 10 and 28 days of ANG II/PE infusion (Fig. 4, B and C). Polarized light microscopy confirmed organized collagen fibers within the heart parenchyma at 10 and 28 days. Immunostaining for collagen 1 showed regions of organized collagen fibers at 10 and 28 days of ANG II/PE infusion within matrix associated with myocytes (identified by WGA staining), as well as in areas of fibrosis that lacked myocytes (Fig. 4D). Increased fibrosis after 28 days of ANG II/PE infusion was confirmed by quantitative measurements of hydroxyproline levels that demonstrated increased collagen deposition in the LV from mice treated with chronic infusion of ANG II/PE for 28 days compared with control mice infused with NaCl (Fig. 5).

Figure 4.

Extracellular matrix deposition and myocardial fibrosis after chronic angiotensin II and phenylephrine (ANG II/PE) infusion. A–C: representative images of myocardial histological sections after 3, 10 and 28 days of ANG II/PE infusion stained with Masson’s trichrome (A) and Picrosirius red (B and C). C: high-magnification images of boxed regions in B. Top: brightfield. Bottom: polarized light microscopy showing organized fibrosis only after 10 or 28 days of chronic ANG II/PE infusion. Histological sections in B and C are from near serial sections of the same heart. D: representative images for collagen 1 (Col1) immunohistochemistry showing mature collagen fibril deposition (red) only after 10 or 28 days of chronic ANG II/PE infusion. Top: wheat germ agglutinin (WGA) and 4′,6-diamidino-2-phenylindole (DAPI). Bottom: Col1 and DAPI. Scale bars = 1 mm (A and B), 200 µm (C), and 50 µm (D).

Figure 5.

Hydroxyproline analysis of whole lung after 28 days of chronic angiotensin II and phenylephrine (ANG II/PE) infusion. n = 12/17 for control/ANG II/PE-treated. *P < 0.05.

Effects of Chronic ANG II/PE Infusion on Perivascular versus Interstitial Myocardial Fibrosis

Histological analysis of the LV after ANG II/PE infusion showed two distinct fibrosis phenotypes: predominant interstitial fibrosis (∼66% of ANG II/PE-treated animals) and predominant perivascular fibrosis (∼33% of ANG II/PE-treated animals). The perivascular phenotype was observed at later (42) but not early (3, 10) days of ANG II/PE infusion and consisted of significant smooth muscle cell hyperplasia surrounded by a region of fibrosis (Fig. 6, A and B). Immunostaining for the endothelial marker CD31 (PECAM1) indicated ectopic staining within the muscularis layer (Fig. 6, B3) compared with control (Fig. 6, A3). We therefore immunostained for the smooth muscle cell marker, αSMA (ACTA2), and CD31, which supported the possibility of endothelial cell contribution to the vascular smooth muscle compartment through an endothelial to mesenchymal transition or to the acquisition of endothelial markers by hyperplastic smooth muscle cells (Fig. 6B, 4–6). In contrast, hearts with the predominately diffused fibrosis phenotype had coronary vasculature that appeared with less muscular hyperplasia (Fig. 6C, 1 and 2) and with minimal perivascular fibrosis and minimal overlap between SMA and CD31 expression in the vascular media or fibrotic regions (Fig. 6C, 3–6). Quantitative assessment of physical endurance and echocardiographic measures did not demonstrate a significant difference with respect to the predominantly interstitial versus perivascular type of fibrosis (Supplemental Table S1C).

Figure 6.

Perivascular vs. interstitial fibrosis following chronic angiotensin II and phenylephrine (ANG II/PE) infusion. Qualitative analysis of perivascular hyperplasia and endothelial cell perivascular migration. A, 1 and 2; B, 1 and 2; and C, 1 and 2: Masson’s trichrome-stained left ventricular (LV) section after 28 days of chronic ANG II/PE infusion at low and high magnification. A, 3; B, 3; and C, 3: immunofluorescence staining for CD31 [platelet endothelial cell adhesion molecule 1 (PECAM1), red], costained with 4′,6-diamidino-2-phenylindole (DAPI, blue). Arrows in B3 show CD31 staining in the muscularis layer in the absence of αSMA staining as a control for potential bleed through from the green channel in B, 4–6. A, 4–6; B, 4–6; and C, 4–6: immunofluorescence staining for CD31 (PECAM1, red) and smooth muscle actin (SMA, green), costained with DAPI (blue). Arrowheads show cells in the muscularis layer that costain with CD31 and SMA. Scale bars = 500 µm, 50 µm, and 25 µm, as indicted.

Effects of Chronic ANG II/PE Infusion on the Development of Myocardial Hypertrophy and Capillary Rarefaction

Cardiomyocyte cross-sectional area was measured in WGA-stained heart tissue as an estimate of myocardial remodeling at the cellular level. Myocyte cross-sectional area was significantly higher in ANG II/PE hearts at all time points compared with hearts from control mice but was increased the most after 3 days of ANG II/PE infusion, with a slight decrease at 10 and 28 days of ANG II/PE infusion (Fig. 7, A and C).

Figure 7.

Analysis of myocyte cross-sectional area and capillary density following chronic angiotensin II and phenylephrine (ANG II/PE) infusion. A: representative images of wheat germ agglutinin (WGA, green)-stained cardiomyocytes in the left ventricle (LV) after 0 (NaCl), 3, 10, and 28 days of infusion of NaCl (control) or ANG II/PE. B: capillary density in the LV shown by immunofluorescence for CD31 (red) and 4′,6-diamidino-2-phenylindole (DAPI; blue). C: quantification of myocyte cross-sectional area near regions of fibrosis showing increased myocyte cross-sectional area peaking at 3 days after chronic ANG II/PE infusion. n = 15, 5, 6, and 13 for control; ANG II/PE 3, 10, and 28 days, respectively. D and E: quantification of capillary density normalized to area (D) and nuclei (E). F: analysis of cell death in regions near areas of myocardial fibrosis. Myocytes were identified by WGA (CF640R, green) staining, and cell death was assessed by terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) assay (red). Inset: positive control for myocardial cell death in regions proximal to areas of acute myocardial infarction. n = 9, 5, 6, and 9 for control/ANG II/PE at 3, 10, and 28 days, respectively. Scale bar = 50 µm. #P < 0.05, NaCl-treated vs. ANG II/PE-treated mice; *P < 0.05; **P < 0.01; ***P < 0.001.

To determine whether ANG II/PE infusion affects capillary remodeling, capillary density was analyzed in all groups of mice (Fig. 7B). Quantitation of capillary density normalized to the region of interest or to the number of nuclei demonstrated capillary rarefaction at all time points of ANG II/PE infusion, but no significant difference was observed between 3, 10, and 28 days (Fig. 7D). However, when capillary density was normalized to number of nuclei, there was a larger reduction at 3 days compared with 28 days of ANG II/PE infusion, possibly due to the presence of nuclei belonging to hematopoietic cell infiltration at the early time point (Fig. 7E).

To determine if ANG II/PE infusion resulted in myocyte cell death, histological sections from control mice and at 1-, 3-, 10-, and 28-day time points were assessed by staining with WGA to outline myocytes and terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) assay to assess cell death. Minimal TUNEL-positive cardiomyocytes were detected in ANG II/PE-treated animals at any time point (Fig. 7F and data not shown). As a positive control, we used LV tissue following ischemia-reperfusion injury, which showed TUNEL-positive myocytes in the peri-infarct region (see Fig. 7F, inset) (61).

To evaluate whether in the absence of significant cardiomyocyte death, there is a change in the expression of cardiac Troponin I (TNNI3), an abundant protein involved in cardiomyocyte contraction, histological sections from control and ANG II/PE mice at the 3-day time point were immunostained for Troponin I. Troponin I was present inside cardiomyocytes encircled by the sarcolemma in control hearts. In hearts from ANG II/PE-treated mice, some myocytes in the proximity of inflammatory infiltrates showed reduced or absent immunostaining for Troponin I, suggesting the potential for cellular loss of this protein (Supplemental Fig. S5: see https://doi.org/10.6084/m9.figshare.23059028).

Effects of Chronic ANG II/PE Infusion on Myocardial Inflammation

To evaluate the immunological response in myocardial tissue to ANG II/PE infusion, transverse sections of the myocardium were assessed for the presence of inflammatory cell types. Hematoxylin-and-eosin staining was used as a roadmap for immunostaining at 3, 10, and 28 days of ANG II/PE infusion (Fig. 8A). The pan-leukocyte marker CD45 was abundantly present at 3 days of ANG II/PE infusion (Fig. 8B). After 10 days, the inflammatory response was markedly reduced and approached baseline levels by 28 days of ANG II/PE infusion. Among assessed hematopoietic cell types, macrophages were the most abundant cell type (Fig. 8C) and neutrophils, representing the myeloid lineage, were present at lower levels (Fig. 8D).

Figure 8.

Inflammatory response following chronic angiotensin II and phenylephrine (ANG II/PE) infusion. A: representative hematoxylin-and-eosin images following 3, 10, and 28 days of ANG II/PE infusion showing regions of matrix deposition and cellular infiltrates within myocardial tissue. B–D: immunostaining for CD45 (green; B), Mac2 (red; C), and Ly6g (yellow; D) identifying inflammatory cell populations that peak at 3 days and resolve by 28 days of ANG II/PE infusion. Scale bar = 50 µm.

Effects of Chronic ANG II/PE Infusion on Myocardial Gene Expression

RNA sequencing was used to identify changes in gene expression in the LV in response to 3 days of chronic ANG II/PE infusion. Principal component analysis (PCA) identified four groups based on treatment and sex (Fig. 9, A and B). PCA showed that most of the variance (86%) was accounted for in PC1 based on treatment and only a small amount of the variance (5%) was due to sex in PC2 (Fig. 9A). DEGs showing greater than twofold change and mean expression levels across all samples greater than 1.5 counts per million identified a large number of differentially expressed genes (DEGs) following 3 days of ANG II/PE infusion with 2,213 genes increased and 1,725 genes decreased in treated (TRT) mice compared with control (CTL) mice (Fig. 9B, Supplemental Table S3: see https://doi.org/10.6084/m9.figshare.22666114).

Figure 9.

Transcriptomic differences between NaCl control and angiotensin II and phenylephrine (ANG II/PE)-treated animals after 3 days. RNA-seq was performed on controls (n = 6) and treated animals (n = 6 at 3 days of ANG II/PE infusion). A: principal component analysis (PCA) using all identified genes for controls (CTL circles) and treated (TRT boxes) (pink, females; blue, males) reveals within-group clusters with minimal overlap. B: volcano plot showing the distribution of gene expression fold change and P values. A total number of 3,939 genes in the core gene category were tested. Blue color changes indicate upregulation in control group, and orange color indicates upregulation in treated group. C–I: hierarchical clustering analysis using genes classified into specific pathways using Pearson correlation and (Euclidean) distance shown as heat maps of variance stabilizing transforms of the reads (scaled between −1 and +2) that also largely separates the groups. Gene lists are derived from gene ontology (GO) biological processes with selective addition of related genes of interest. Subcategories in the heat maps are defined in Tables 2 and 3. Gene list for selected major pathophysiological pathways can be found in Supplemental Table S3.

Gene set enrichment analysis (Enrichr) comparing NaCl and ANG II/PE-infused mice identified changes in gene expression in several of the common major pathways associated with HFpEF (Table 2, and Fig. 9, C–I). Pathways that were predominantly upregulated include extracellular matrix activation (Fig. 9C), depression of aerobic oxidation (Fig. 9D), myocardial mechanical function (Fig. 9E), angiogenesis and vascular homeostasis (Fig. 9F), activation of inflammatory pathways (Fig. 9G), and signal pathway activation and sarcoplasmic reticulum stress (Fig. 9H). Also identified in this RNA seq data set are genes with potential unanticipated roles in HFpEF and genes that have been described in the literature as affecting myocardial homeostasis and stress (Fig. 9I, and Table 3) (67–73). Increased expression trends of several genes were validated by qRT-PCR in an independent cohort of ANG II/PE infused mice (Supplemental Fig. S6; see https://doi.org/10.6084/m9.figshare.23059076).

Table 2.

Comparison of affected biological pathways in mice treated with ANG II/PE for 3 days and ventricular biopsies from patients with HFpEF

Gene Ontology: Biological Process	Mouse LV (3 Days ANG II/PE)		Human RV (Hahn et al.)*		Human LV (Das et al.)#
	Gene Count	P Value†	Gene Count	P Value†	Gene Count	P Value†
Extracellular matrix activation (C)§
Positive regulation of fibroblast proliferation (GO:0048146) (C1)§	12/28	0.0043	14/28	0.32	1/28	0.64
Chondroitin sulfate metabolic process (GO:0030204) (C2)	15/30	1.91e-04	17/30	0.11	1/30	0.67
Extracellular matrix organization (GO:0030198) (C3)	145/300	2.70e-29	60/300	0.014	26/300	4.5e-5
Collagen fibril organization (GO:0030199) (C4)	59/89	1.26e-21	20/89	0.085	11/89	3.9e-4
Supramolecular fiber organization (GO:0097435) (C5)	136/351	7.17e-17	67/351	0.0011	32/351	2.1e-6
Extracellular matrix assembly (GO:0085029) (C6)	13/24	1.83e-04	13/24	0.209	7/24	1.7e-5
Mesenchymal cell differentiation (GO:0048762) (C7)	17/51	0.015	22/51	0.597	3/51	0.28
Depression of aerobic oxidation (D)
Aerobic electron transport chain (GO:0019646) (D1)	52/70	7.86e-23	47/70	7.3e-5	12/70	7.9e-6
Mitochondrial ATP synthesis coupled electron transport (GO:0042775) (D1)	52/71	2.38e-22	44/71	1.2e-4	12/71	9.2e-6
NADH dehydrogenase complex assembly (GO:0010257) (D1)	40/58	5.09e-16	41/58	3.3e-5	9/58	2.3e-4
Mitochondrial respiratory chain complex assembly (GO:0033108) (D1)	48/90	1.25e-12	58/90	6.8e-5	9/90	5.6e-3
Negative regulation of cAMP-mediated signaling (GO:0043951) (D2)	5/10	0.031	4/10	0.7105	2/10	0.049
Glucose metabolic process (GO:0006006) (D3)	21/62	0.006	32/62	0.1369	7/62	0.007
Branched-chain amino acid catabolic process (GO:0009083) (D4)	15/20	1.43e-07	13/20	0.047	3/20	0.03
Fatty acid β-oxidation (GO:0006635) (D5)	27/52	2.15e-07	22/52	1.5e-3	4/52	0.12
Angiogenesis and vascular homeostasis (F)
Positive regulation of angiogenesis (GO:0045766) (F1)	48/116	6.72e-08	60/116	0.055	10/116	0.01
Positive regulation of nitric oxide synthase biosynthetic process (GO:0051770) (F2)	7/11	0.0018	6/11	0.34	1/11	0.63
Regulation of vascular associated smooth muscle cell proliferation (GO:1904705) (F3)	15/37	0.0029	27/37	3.2e-4	2/37	0.39
Activation of the inflammatory pathways (G)
Neutrophil mediated immunity (GO:0002446) (G1)	124/488	0.0011	240/488	0.01	28/488	0.013
Regulation of platelet activation (GO:0010543) (G2)	8/22	0.0518	8/22	0.82	2/22	0.19
Platelet aggregation (GO:0070527) (G2)	20/36	2.01e-06	4/36	0.107	5/36	9.5e-3
Platelet degranulation (GO:0002576) (G2)	45/125	1.49e-05	64/125	0.06	11/125	6.2e-3
Regulation of cytokine production involved in inflammatory response (GO:1900015) (G3)	16/43	0.0057	19/43	0.54	3/43	0.21
Positive regulation of tumor necrosis factor superfamily cytokine production (GO:1903557) (G3)	30/81	2.11e-04	37/81	0.41	3/81	0.57
Macrophage activation (GO:0042116) (G4)	19/36	9.93e-06	13/36	0.86	N/A	N/A
Macrophage activation involved in immune response (GO:0002281) (G4)	8/13	0.0011	3/13	0.96	1/36	0.38
Regulation of inflammatory response (GO:0050727) (G5)	64/206	6.46e-05	31/206	0.44	N/A	N/A
Positive regulation of inflammatory response (GO:0050729) (G5)	31/89	5.80e-04	45/89	0.12	4/89	0.41
Myocardial involvement
Myocardial function (mechanical) (E)
Regulation of cardiac muscle contraction (GO:0055117)	18/36	4.42e-05	20/36	0.11	4/36	0.04
Relaxation of cardiac muscle (GO:0055119)	7/13	0.0064	9/13	0.059	1/13	0.38
Myocardial homeostasis and stress (Table 3)
Signal pathway activation and sarcoplasmic reticulum stress (H)
Cellular response to oxidative stress (GO:0034599) (H1)	36/125	0.0091	74/125	4e-4	8/125	0.08
Positive regulation of intracellular signal transduction (GO:1902533) (H2)	140/546	3.54e-04	263/546	0.02	25/546	0.14
Negative regulation of signal transduction (GO:0009968) (H2)	76/267	3.32e-04	48/267	0.003	22/267	3.5e-4
Positive regulation of transmembrane transport (GO:0034764) (H3)	7/14	0.01039	11/14	0.009	4/14	0.028
Transmembrane receptor protein tyrosine kinase signaling pathway (GO:0007169) (H4)	113/404	3.23e-05	207/404	0.002	18/404	0.23
Downregulated biological pathways in HFpEF (Hahn et al. human RV)
Cellular response to thyroid hormone stimulus (GO:0097067)	23/76	0.017	44/76	9.7e-3	4/76	0.30
Cellular response to peptide hormone stimulus (GO:0071375)	20/106	0.18	63/106	9.1e-4	6/106	0.19
Cellular response to insulin stimulus (GO:0032869)	32/129	0.086	77/129	2.2e-4	8/129	0.1
Endomembrane system organization (GO:0010256)	26/199	0.99	114/199	9.5e-5	12/199	0.062
Cytosolic transport (GO:0016482)	6/116	1.00	82/116	0.001	7/116	0.13
Vesicle organization (GO:0016050)	6/46	0.91	29/46	7e-3	3/46	0.23
Regulation of cellular protein localization (GO:1903827)	12/46	0.17	23/46	0.2462	7/46	0.001
Endoplasmic reticulum to Golgi vesicle-mediated transport (GO:0006888)	17/185	1.00	110/185	1.4e-5	5/185	0.81
Establishment of protein localization to organelle (GO:0072594)	17/76	1	44/76	9.7e-3	4/76	0.3
Protein localization to nucleus (GO:0034504)	12/106	0.99	65/106	2.2e-4	10/106	0.005
Cellular response to organonitrogen compound (GO:0071417)	25/103	0.14	46/103	0.47	5/103	0.32
Cellular response to nitrogen compound (GO:1901699)	2/20	0.9	8/20	0.71	1/20	0.52
Response to insulin (GO:0032868)	18/84	0.3	50/84	0.003	7/84	0.03
Positive regulation of catabolic process (GO:0009896)	6/67	0.2	40/67	0.007	7/67	0.01
Positive regulation of autophagy (GO:0010508)	13/90	0.9	58/90	6.86e-5	9/90	0.005
Positive regulation of cellular catabolic process (GO:0031331)	32/141	0.2	81/141	8.2e-4	13/141	0.002
Blood vessel morphogenesis (GO:0048514)	19/56	0.008	31/56	0.056	6/56	0.016
Positive regulation of angiogenesis (GO:0045766)	47/116	1.7e-7	60/116	0.055	10/116	0.01
Positive regulation of cellular protein localization (GO:1903829)	15/71	0.42	36/71	0.15	5/71	0.1
Macroautophagy (GO:0016236)	14/120	0.99	74/120	6.6e-5	8/120	0.07
Upregulated biological pathways in HFpEF (Hahn et al. human RV)
Mitochondrial respiratory chain complex assembly (GO:0033108)	48/90	1.1e-12	58/90	6.2e-16	9/90	0.0056
Mitochondrial ATP synthesis coupled electron transport (GO:0042775)	52/71	1.9e-22	47/71	1.2e-4	12/71	9.1e-6
Regulation of oxidative phosphorylation (GO:0002082)	7/16	0.02	11/16	0.04	1/16	0.45
Respiratory electron transport chain (GO:0022904)	10/14	3.8e-5	10/14	0.035	1/14	0.41
NADH dehydrogenase complex assembly (GO:0010257)	40/58	4.4e-16	41/58	1.6e-14	9/58	2.3e-4
Mitochondrial respiratory chain complex I assembly (GO:0032981)	40/58	4.4e-16	41/58	3.3e-5	9/58	2.3e-4
Mitochondrial electron transport, NADH to ubiquinone (GO:0006120)	33/39	3.8e-18	28/39	3.9e-4	7/39	4.7e-7
Electron transport chain (GO:0022900)	10/15	4.5e-6	9/15	0.16	1/15	0.43
Cellular respiration (GO:0045333)	24/42	8.5e-8	27/42	6e-3	5/42	0.018
Nucleobase-containing small molecule catabolic process (GO:0034656)	5/16	0.056	6/15	0.71	N/A	N/A
Organic acid catabolic process (GO:0016054)	2/9	0.55	3/9	0.83	2/9	0.04
Carboxylic acid catabolic process (GO:0046395)	2/12	0.71	4/12	0.84	1/12	0.36

Italics indicates upregulated pathways. Boldface indicates downregulated pathways. ANG II/PE, angiotensin II and phenylephrine; GO, gene ontology; HFpEF, heart failure with preserved ejection fraction; LV, left ventricular; RV, right ventricular.

^†P value for GO biological processes (Enrichr analysis); *human RV [Hahn et al. (64)]; #human LV [Das et al. (65)]; §()reference to Fig. 9 hierarchical clustering analysis.

Table 3.

Differentially expressed genes in control and ANG II/PE-infused mice

Gene	Gene Name (DAVID Annotation)	Log2-FC	Padj (FDR)	Function
Myocardial homeostasis and stress (Fig. 9I1)
Thbs4	Thrombospondin 4	9.61	4.16e-35	Adaptive responses of the heart to pressure overload and in myocardial function and remodeling
C1qtnf6	C1q and tumor necrosis factor related protein 6	5.18	1.13e-221	Protective through activation of AKT signaling pathway in doxorubicin-related heart failure
Rcan1	Regulator of calcineurin 1	2.64	1.13e-57	Protective role of RCAN1 in myocardial ischemia/reperfusion injury
Fstl3	Follistatin-like 3	5.60	8.97e-76	Sensitive biomarker of changes in HF status
Aplnr	Apelin receptor	−1.03	1.44e-06	Myocardial homeostasis and stress
Emp1	Epidermal growth factor-containing fibulin-like extracellular matrix protein 1	−3.94	1.46e-30	Myocardial homeostasis and stress
Prss23	Protease, serine, 23	3.57	2.40e-18	Prss23 inhibited the endothelial-to-mesenchymal transition
Mylk4	Myosin light chain kinase 4	−3.71	7.46e-29	Myocardial homeostasis and stress
Atp1b1	ATPase, Na+/K+ transporting, beta 1 polypeptide	1.15	5.88e-26	Myocardial homeostasis and stress
Slc8a1	Solute carrier family 8 (sodium/calcium exchanger), member 1	1.42	2.13e-43	Myocardial homeostasis and stress
Myh7	Myosin, heavy polypeptide 7, cardiac muscle, beta	1.62	2.07e-05	Myocardial homeostasis and stress
Tnnt3	Troponin T3, skeletal, fast	6.34	4.66e-22	Myocardial homeostasis and stress
Selected other significant genes (Fig. 9I2)
Postn	Periostin	11.29	0	ECM activation
Cthrc1,	Collagen triple helix repeat containing 1	15.22	1.67e-84	ECM activation
Ddah1	Dimethylarginine dimethylaminohydrolase 1	4.71	1.24e-75	Attenuating LV remodeling after AMI
Sparc	Secreted acidic cysteine rich glycoprotein	3.83	1.46e-223	Enhancement of post synthetic collagen processing and myocardial fibrosis
Fhl1	Four and a half LIM domains 1	3.57	1.55e-184	Beneficial functional response to pressure overload through Gaq (Gq) signaling pathway
Nos2	Nitric oxide synthase 2, inducible, macrophage	−2.19	2.05e-19	Modulate effects on contractile myocardial function and mitochondrial homeostasis
A530016L24Rik	Nrac	−4.54	4.45e-119	Novel protein to diagnose and evaluate HFpEF progression

ECM, extracellular matrix; HFpEF, heart failure with preserved ejection fraction; LV, left ventricular; AMI, acute myocardial ischemia.

To evaluate how changes in gene expression in ANG II/PE-infused mice compare with human HFpEF, we examined DEGs in two available human studies. Das et al. evaluated LV myocardial biopsies from five patients undergoing coronary bypass grafting (CABG) who had HF and an EF > 45% (65). Hahn et al. (64) evaluated right ventricular (RV) septal endomyocardial biopsies from 41 patients meeting consensus criteria for HFpEF. Our evaluation of the Das et al. data set identified 707 genes downregulated and 24 genes that were increased in HFpEF compared with normal heart samples. Hahn et al. identified 4,936 downregulated genes and 3,847 upregulated genes in HFpEF compared with normal heart samples. Comparison of these data with DEGs identified in the mouse LV following 3 days of chronic ANG II/PE infusion (1,726 downregulated and 1,220 upregulated) identified one gene, Lumican (LUM), that was increased in all three data sets, 220 genes that were increased in both the Hahn et al. data set and mouse day 3 data set, and five genes that were increased in both the Das et al. and Hahn et al. data sets (Supplemental Fig. S7A; see https://doi.org/10.6084/m9.figshare.23059187). There were 17 genes downregulated in all three data sets, and 292 genes decreased in both the Hahn et al. data and mouse day 3 data sets, 63 genes decreased in both the Das et al. data and mouse day 3 data sets, and 208 genes decreased in both the Das et al. data and Hahn et al. data sets (Supplemental Fig. S7B; https://doi.org/10.6084/m9.figshare.23059187). The gene lists represented in up- and downregulated Venn diagrams are shown in Supplemental Table S4 (see https://doi.org/10.6084/m9.figshare.22666207).

Comparison of GO biological pathways regulated in the Hahn et al. data set (Table 2) and in the mouse following 3 days of ANG II/PE infusion (Table 2) with the Das et al. data set revealed common biological processes that were differently regulated among the three data sets. Extracellular matrix activation was significantly downregulated in both the RV (Hahn et al.) and the LV (Das et al.) data sets, which contrasted with the upregulation seen in the ANG II/PE model of HFpEF. Interestingly, biological pathways responsible for maintenance of metabolic and mitochondrial homeostasis were significantly downregulated in the mouse following 3 days of ANG II/PE infusion and in the Das et al. data set but upregulated in the Hahn et al. data set. As anticipated, inflammatory pathways were increased in the mouse LV given the early time point in the pathogenesis of this model and were relatively inactive in the human heart samples (Table 2).

DISCUSSION

Heterogeneity of HFpEF phenotypes both clinically and at the molecular level originates from distinct pathophysiological pathways making it difficult to mimic human disease with animal models. Multiorgan/system involvement (including heart, lungs, vasculature, and kidneys) and the progressive nature of the onset of disease further contribute to the inability to model the major features of human HFpEF reliably and reproducibly in rodents. Despite the complexity of the problem, a murine model would provide a valuable instrument for the identification of pathophysiological mechanisms and for the investigation of factors that could contribute to a “multiple hit” model satisfying key HFpEF features. Furthermore, with our better understanding of HFpEF pathophysiology and comorbidities, criteria that satisfy human HFpEF features in a murine model can be significantly expanded to include multisystem and organ cross talk beyond the cardiovascular system (12, 18, 74). The goal of this study was to characterize a mouse model that exhibits key features of HFpEF, demonstrate reproducibility, and document pathophysiological factors, including fibrosis, inflammation, myocardial hypertrophy, diastolic dysfunction, capillary rarefaction, and metabolic derangements that contribute to the progression of HFpEF.

Previously described in elderly woman with systolic hypertension, HFpEF is now associated with epidemics of obesity, metabolic syndrome, and inflammatory dysregulation. A mouse model of HFpEF should therefore mirror the aforementioned clinical paradigms, prioritizing hypertension, inflammatory response, and metabolic derangements as causative factors. We implemented a flow chart scheme for identifying the major features of HFpEF previously described (12), and we expanded it to cover additional criteria that includes multiorgan involvement and analysis of serum cytokine levels which can be easily correlated with patient data (Fig. 1A). The choice to administer ANG II/PE was primarily to create an environment of systemic hypertension (both drugs are involved in increasing systemic vascular resistance) that would lead to cardiac remodeling (75), as well as incorporate ANG II-related blood pressure independent factors, including its proinflammatory and profibrotic features (76–78). To circumvent the sudden onset of heart failure due to pressure overload (TAC model) or other drug interventions (21), we infused ANG II/PE over a 28-day period. Interestingly, in contrast to results demonstrated in other studies that used the same model of ANG II/PE infusion for 28 days (35), where Aghajanian et al. demonstrated an echocardiographic picture of combined systolic and diastolic dysfunction, in the current study we did not observe the development of a HFrEF phenotype. On the contrary, we observed an increase in EF in ANG II/PE-treated mice as early as 10 days, which remained elevated by the end of 28 days of ANG II/PE infusion. Explanation of interobserver variability may be related to the ability of different strains of mice to respond to the same stressor, and perhaps more importantly, the sedation used during echocardiographic analysis (subcutaneous and moderate sedation versus deeper anesthesia during isoflurane administration). In addition, the mouse ANG II/PE infusion model combines all known general pathophysiological features of HFpEF, thus we consider it as a useful small animal model. We also acknowledge that specific clinical situations do exist that could represent particular scenarios of HFpEF-like conditions; for example, HFpEF that develops in dialysis patients with chronic kidney disease (CKD) or patients with diabetes (DM). In these particular situations, a nephrectomized mouse model of HFpEF or a model that uses mice that are susceptible to DM (db/db mouse) would probably be more relevant than the ANG II/PE model of HFpEF (22–25). We also observed that the majority of animal model studies limited their evaluation of HFpEF by only describing myocardial changes and specifically diastolic dysfunction criteria, which are by all means major criteria for HFpEF, while omitting the systemic nature of the syndrome and specifically impairment of other systems and organs (28, 30–32, 34). An additional important point that is often undervalued is the attrition rate (<5% in the ANG II/PE model), which can be high in some animal models (26, 27, 33). In this article, we discussed multiple murine models (Table 1). We believe that investigation of HFpEF mechanism, genetics, pathogenesis, and potential therapies should be initiated in rodent models because of the relatively rapid experimental time course and cost-effectiveness and then advanced to large animal models (e.g., swine) for HFpEF that are clearly superior to any rodent model of HFpEF in the context of human translatability. In terms of clinical relevance, mouse models are amenable to both invasive and noninvasive hemodynamic measurements. Notably, in both swine and mouse models, investigators have used identical stress factors that include pressure overload (aortic banding and ANG II/PE infusion), and metabolic stress. Despite several advantages, large animal models also present numerous challenges that often include not meeting all criteria for HFpEF, significant attrition, less amenable genetic modification, long experimental and generational time, requirement for surgical manipulation, and resulting high cost (29, 33, 34). One model, using Göttingen miniswine, does not require surgical manipulation and does fulfill all criteria for HFpEF (Table 1); however, the model begins with 14-mo-old animals that are then followed for 20 wk. Tan et al. (31), Ishikawa et al. (32), and Liao et al. (27) used a model with aortic banding, that also demonstrated narrow margin on transitioning between HFpEF and HFrEF. Weil et al. (30) used a repetitive phenylephrine infusion through an indwelling catheter to induce pressure overload), Ishikawa et al. (32) used aortic banding, and Olver et al. (34) used aortic banding and metabolic stress. However, these models have not demonstrated multiorgan involvement (Table 1).

Multiorgan involvement, though often not evaluated, appears to be an etiological factor for HFpEF. The lungs and kidneys of ANG II/PE-infused mice showed histological fibrosis and for the lung, we showed increased levels of hydroxyproline (as a quantitative measure of collagen content). Pulmonary edema (a comorbidity associated with HFpEF due to LA hypertension) has been associated with increased lung weight and an elevated wet/dry lung weight ratio in mice (79, 80). Interestingly, although lung weight was increased, the wet/dry lung weight ratio in ANG II/PE infused mice was decreased at the 28-day time point. This discrepancy may be related to the greater relative increase in dry lung weight (because of ECM deposition) along with a modest increase in the wet lung weight caused by developing pulmonary edema. A longer experimental time course could resolve this discrepancy. Renal involvement in ANG II/PE-treated mice was suggested by the presence of sparse cortical fibrosis primarily surrounding the renal vasculature. Renal involvement makes this model potentially suitable for studying cardiorenal syndrome, a frequent comorbidity in HFpEF (81). Of note, analysis of RNA seq data from the LV of a mouse model for cardiorenal syndrome showed increased expression of fibronectin (Fn1) and periostin (Postn) (82), both of which were increased at 3 days in the LV of ANG II/PE-treated mice (15- and 51-fold, respectively, Supplemental Table S3).

Echocardiography has become the cornerstone of the noninvasive evaluation of diastolic function in patients with HFpEF. Numerous human studies over decades of investigations have established the clinical utility of a number of echocardiographic parameters in categorizing the severity of diastolic dysfunction, among which mitral annular tissue Doppler, E/E′ ratio, LA size, and tricuspid regurgitation velocity are key components of an algorithmic assessment of these clinical grades (83). Unfortunately, mouse physiology and cardiovascular hemodynamics differ significantly from the human condition. Notably, the high heart rate and the near-simultaneous occurrence of early and late diastolic LV filling limits the direct adaptation of the human parameters in evaluating diastolic function in mice (84, 85). Although E/E′ ratio has been extensively validated in patients as a correlate of LA pressure, its utility in mouse models is far less established. Nevertheless, empirical evidence from small sample size studies would support its use in the evaluation of HFpEF models, and our data showing increased E/E′ ratio early in the progression of the ANG II/PE model provides further evidence for its role as a sensitive parameter of diastolic dysfunction. LA size and function have emerged as an important surrogate of diastolic dysfunction in clinical studies. Although left atrial volume is widely measured by two-dimensional (2-D) or three-dimensional (3-D) echocardiography, the most sensitive parameters for the diagnosis of HFpEF appears to be related to LA function assessed by speckle tracking analysis (54). In mouse models, there is a paucity of data regarding the changes in LA size in HFpEF models, and LA deformation analysis by STE has not been reported previously. Therefore, we developed a technique using a modified imaging plane to consider the unique anatomy of the mouse LA and surrounding structures and subjected these images to commercially available STE software (VevoStrain) to quantitate LA volumes, chamber function, and wall deformation. Despite no significant changes in LA volumetric measurements after 10 days of ANG II/PE infusion, we demonstrated a significant decrease in LA reservoir strain, which may provide added value not only for noninvasive diagnosis of LV diastolic dysfunction but also for grading its severity and monitoring the effects of treatment. Our assessment of the progressive changes in LA size and function correlates well with the degree of LV diastolic dysfunction following chronic ANG II/PE infusion. Finally, analysis of the pulmonary vein velocity profile provides additional information about LA/LV hemodynamics, lending further support to the notion that comprehensive evaluation of the HFpEF phenotype requires the use of multiple lines of evidence, which in total indicates the presence of altered LV filling.

For diseases with multiple stress-induced etiologies (hypertension, obesity, metabolic syndrome) and with a clinically broad range of phenotypes (minimal echocardiographic changes vs. overt signs of heart failure and decompensation), it is always a challenge to find a mouse model that encompasses all pathophysiological disease features. Based on the existing available data for the cellular and molecular pathobiology of HFpEF (12, 74), we prioritized the evaluation of cardiac fibrosis, myocardial hypertrophy, diastolic dysfunction, capillary rarefaction, and the local and systemic inflammatory response. As discussed earlier, all of these pathological features will ultimately contribute to altered myocardial relaxation and/or stiffness, and all demonstrated either a histopathological or gene level change following ANG II/PE infusion. Notably, no significant difference between male and female mice with respect to the levels of myocardial fibrosis (perivascular and interstitial) or echocardiographic parameters was observed.

Myocardial Fibrosis

Myocardial fibrosis has multiple etiologies and precedes stiffening of heart tissue and a decrease in passive chamber compliance and end-diastolic relaxation (86). Myocardial fibrosis is considered a primary cause of diastolic dysfunction in HFpEF (87). As observed on postmortem and endomyocardial human biopsy samples, systemic hypertension alters the quality of the myocardium, changing its composition and leading to increased deposition of connective tissue matrix (fibrosis) (88–91). Distribution of fibrosis was reported to increase from the outer to the inner left ventricular free wall, reflecting transmural gradients of myocardial wall stress (89). Indeed, in mice infused with ANG II/PE, fibrotic tissue was predominantly closer to the endocardium and often affected the papillary muscles. The gradual onset of myocardial fibrosis was also modeled with chronic ANG II/PE infusion. Even though there was no overt organized fibrosis at early stages of ANG II/PE infusion (absence of or sparse collagen deposition on Picrosirius red staining and Col1 immunohistochemistry), we demonstrated widespread activation of ECM gene expression on RNA seq analysis as early as 3 days and increasing histological evidence of organized fibrosis at 10 and 28 days of ANG II/PE infusion.

Although systemic hypertension is considered a major factor in the progression of HFpEF, nonhemodynamic factors also contribute to fibrosis. Clinical studies showed that the ability for fibrosis to regress in patients with hypertension is independent of antihypertensive treatments (88, 92). Of particular importance, activation of the renin-angiotensin-aldosterone system (RAAS) may explain the perivascular fibrosis phenotype in the heart, lung, and kidney and the correlation of severity of systolic hypertension and myocardial fibrosis. Importantly, angiotensin as a part of RAAS may independently worsen fibrosis and alter the balance between profibrotic and antifibrotic molecules (91, 93). Consistent with activation of profibrotic pathways (94–97), RNA seq showed increased expression of osteopontin (Spp1), plasminogen activator inhibitor-1 (Serpine1), and TGFβ (Tgfb1, Tgfb2, Tgfb3) in ANG II/PE infused mice. Whether these effects are reversible will require additional studies that assess the effect of ANG II/PE termination on the clinical (treadmill, echocardiography) and histopathological phenotypes.

In mice infused with ANG II/PE for 3 days, expression of Postn, Cthrc1, and Ddah1 were highly increased, 51-, 174-, and 5.6-fold, respectively. Following myocardial infarction (MI) in mice, cardiac fibroblasts convert into periostin (Postn)-expressing activated fibroblasts with properties of myofibroblasts (39). Single-cell RNA sequencing of the fibroblasts that emerge following MI show expression of the myofibroblast signature genes, Postn, Cthrc1, and Ddah1 (98, 99). This suggests that ANG II/PE infusion is activating a similar population of fibroblasts to those activated following acute MI.

Difficulty in obtaining human LV biopsy samples (due to risks associated with LV prograde vs. retrograde access) led to a paucity of transcriptomic analysis from human patients with HFpEF. However, two recent studies attempted to identify changes in myocardial gene expression in human HFpEF (64, 65). Comparison of affected biological pathways following 3 days of ANG II/PE infusion in mice to LV and RV biopsy tissue from patients with HFpEF in Das et al. (65) and Hahn et al. (64), respectively, identified a small number of common DEGs, especially with the Hahn et al. data. These results are encouraging, especially considering the limitations of this analysis which compared a relatively acute mouse model (3 days of ANG II/PE infusion) with data from patients at late-stage disease that aimed to identify genetic consequences of HFpEF stressors.

Considering the duration of disease, pathways related to acute inflammatory response and extracellular matrix activation were not as pronounced in human compared with this early pathogenic stage in mice. Notably, the histological analysis clearly demonstrated a fading inflammatory response by the 28th day of ANG II/PE infusion. Interestingly, pathways involved in metabolism, mitochondrial homeostasis, and signaling were commonly affected and downregulated in the ANG II/PE model and in the Das et al. LV biopsy samples from patients with HFpEF, whereas upregulated in RV biopsy samples from patients with HFpEF in the Hahn et al. data set (Table 2). A potentially more valid comparison will require RNA-seq data from later time points in the mouse ANG II/PE infusion model.

Interestingly, this comparative RNA-seq analysis identified only one gene, Lumican (Lum), that was increased in the two human studies and in the mouse. Lumican is a small leucine-rich proteoglycan that binds fibrillar collagens and can promote fibrotic disease, possibly through activation of TGFβ signaling (100, 101). Lumican expression is increased in human hypertrophic cardiomyopathy and HF, and in pressure overload-induced HF in mice (101, 102). Mice that are heterozygous for Lum show less fibrosis following aortic banding (103). However, mice completely lacking Lum showed increased mortality following isoproterenol infusion or aortic banding (104, 105). Following aortic banding, hearts lacking Lum showed reduced collagen expression and cross-linking, and increased LV dilatation, hypertrophy, and systolic dysfunction (105).

Perivascular versus Interstitial Cardiac Fibrosis Phenotypes and Its Significance

In chronic ANG II/PE-infused mice two types of myocardial fibrosis were observed; predominantly perivascular fibrosis and predominantly interstitial fibrosis, consistent with previous observations (86). Perivascular fibrosis was associated with significant perivascular smooth muscle hyperplasia that could eventually result in narrowing of the coronary vessels, decreased coronary blood flow, and tissue oxygen deprivation (106, 107). Despite these histological differences, clinical (treadmill), as well as echocardiographic values did not demonstrate significant differences with respect to cardiac histology (Supplemental Table S2C). However, these histological phenotypes could reflect different pathological pathways to the development of diastolic dysfunction.

In the case of predominantly perivascular fibrosis, contributions to the origin of the increased vascular smooth muscle and surrounding perivascular fibrosis could come from endothelial cells. In several pathological conditions (e.g., pulmonary hypertension and atherosclerosis), the contribution of endothelial-to-mesenchymal transition (EndMT) to vascular smooth muscle has been demonstrated (108–110). Although the contribution of endothelial cells to fibrosis has been demonstrated, this result remains controversial (111–114). In ANG II/PE-induced HFpEF, coexpression of CD31 and SMA was minimal and confined to the vascular smooth muscle compartment. This suggests that EndMT does not significantly contribute to the vascular or fibrotic phenotype in this model. A definitive assessment of EndMT will require inducible genetic lineage tracing studies. Future studies will be needed to determine if these histological differences predict phenotype reversibility upon the removal of ANG II/PE.

Myocardial Hypertrophy

Ventricular hypertrophy can develop because of genetic (hypertrophic cardiomyopathy), metabolic (glycogen storage and lysosomal diseases, mitochondrial disease), or hemodynamic (pressure overload, systemic hypertension) disease. Ventricular hypertrophy is an etiological factor for the development of delayed diastolic relaxation and reduction in LV compliance, and as a result, diastolic dysfunction and HFpEF progression (115, 116). Despite the clinical presentation of LVH in fewer than 60% of patients with HFpEF, this particular myocardial phenotype has been well described in various in vivo models of diastolic dysfunction (117, 118). Indeed, our model demonstrated consistent cardiomyocyte hypertrophy that developed as early as after 3 days of chronic ANG II/PE infusion. Interestingly, the hypertrophic response in our model was more pronounced at the very beginning with a significant reduction of cardiomyocyte cross-sectional area size after 10 and 28 days of chronic ANG II/PE infusion. Consistent with reduced myocyte hypertrophy but continued myocardial stress, serum levels of NT-proBNP were increased after 28 days of ANG II/PE infusion. NT-proBNP is a cleavage fragment of the prohormone B-type natriuretic peptide (BNP), which in response to myocardial stretch is produced and secreted from the cardiac ventricles and later cleaved into vasoactive BNP and inactive NT-proBNP. These natriuretic peptides have been directly correlated with several hemodynamic measures including LVEDP, LV end-systolic and end-diastolic volumes, and pulmonary capillary wedge pressure (119).

In previous work, we established that activation of fibroblast growth factor receptor 1 (FGFR1) signaling in cardiomyocytes leads to rapid myocyte hypertrophy, increased myocyte contractility, and worsened cardiomyocyte relaxation, and after long-term FGFR1 activation, diastolic dysfunction (120). More recently, it was shown that inhibition of FGF21-FGFR4 signaling reduced myocardial hypertrophy in a mouse model of diabetes (121). However, another other study showed a protective effect of FGF21 on cardiac hypertrophy in response to isoproterenol infusion (122, 123). In this study, it was concluded that the heart was responsive to both locally and systemically released FGF21, which acts as a cardiomyokine, owing to the expression in cardiomyocytes of β-klotho, an essential coreceptor rendering specific responsiveness to FGF21 action (123). In ANG II/PE-infused mice, we found elevated serum levels of FGF21 and FGF23, at 3, 10, and 28 days of ANG II/PE infusion suggesting the potential involvement of endocrine FGF signaling in HFpEF development.

Capillary Rarefaction and Functional Capillary Impairment

HFpEF is known to be associated with diseases with the primary pathophysiology related to microvascular dysfunction, leading to impaired microcirculation in the heart and other organs (16, 124). Reduced coronary flow reserve, oxygen demand/supply imbalance, and impaired angiogenesis may promote cardiomyocyte hypoxia and myocardial fibrosis culminating in diastolic dysfunction and HFpEF (125–127). In chronic ANG II/PE-infused mice, we demonstrated decreased capillary density beginning as early as 3 days. Interestingly, the capillary-to-nuclear ratio was significantly lower at early stages of disease, which is likely related to the influx of hematopoietic cells. Further studies are needed to determine whether capillary rarefaction is a primary cause of HFpEF or a consequence of cardiac remodeling.

Cardiomyocyte Sarcomere Function and Myocardial Death

One of the major histopathological differences between HFpEF and HFrEF is the decreased contractility and systolic dysfunction in the context of myocardial cell loss (128). That said, HFpEF is better known for impaired Ca²⁺-dependent sarcomeric properties and abnormal cGMP-dependent protein kinase activity (PKG) (18, 129–132). In chronic ANG II/PE-infused mice, we confirmed the absence of significant cardiomyocyte death based on the absence of TUNEL staining at 1, 3, 10, and 28 days of ANG II/PE infusion. Further studies to determine mechanisms of impaired cardiomyocyte contractility and relaxation, as well as contribution of NO-cGMP-PKG signaling contribution are warranted.

Immune and Inflammatory Response

Chronic low-grade local and systemic inflammation is considered a causal link to the development of worsening myocardial compliance (2, 16, 133, 134). Underlying mechanisms are thought to involve cardiac tissue edema, noncollagenous proteinaceous ECM deposition, as well as hematopoietic cell cross talk with myocardial fibroblasts and the stimulation of mature collagen production (16, 18). Evidence of a systemic inflammatory state in patients with HFpEF comes from increased levels of cardiac biomarkers commonly identified in serum, including IL-6, C-reactive protein (CRP), IL-10, TNFα, and myeloperoxidase (MPO) (119, 135, 136). In both humans and mice with diastolic dysfunction, cardiac macrophages are increased (136, 137). Consistent with the established inflammatory response observed in human HFpEF, in chronic ANG II/PE-infused mice we found an immune cell infiltration that was predominantly composed of macrophages at early stages (3 days) and that subsided by 28 days. This is consistent with high levels of CD68 expression in the LV after 4 days of ANG II/PE infusion (138). Similarly, a smaller myeloid hematopoietic cell response peaked at 3 days and resolved by 28 days.

Transcriptional Profiling of the Myocardium in the ANG II/PE Mouse Model of HFpEF

An unbiased assessment of transcriptomic changes in mice infused with ANG II/PE for 3 days mirrored previously described histopathological changes. DEGs identified in the ANG II/PE model were consistent with different animal models of HFpEF and with human HFpEF (64, 139, 140). Although there is substantial heterogeneity in the clinical HFpEF phenotype, histological and transcriptomic changes are relatively consistent. ANG II/PE-infused mice showed reduced expression of genes involved in the aerobic oxidation of both fatty acids and glucose, whereas studies of human heart biopsies demonstrated increased fatty acid oxidation and decreased glycolytic processes (64, 141). These differences may be related to the prevalence of obesity in patients with HFpEF and consequent higher cardiac output required to perfuse more tissue (fat), necessitating increased expression of fatty acid energy-related genes.

In addition to changes in common histopathological, immune, and metabolic pathways known to be associated with HFpEF, we also identified genes that have been described in the literature as affecting myocardial homeostasis and stress and cardiomyocyte physiology (67–73).

For example, expression of the cardiac troponin genes, Tnnt1, Tnni3, and Tnni3k, were decreased 1.5- to 2-fold (that was confirmed by immunohistochemical analysis at 3 days of ANG II/PE infusion), whereas the fast skeletal muscle troponin, Tnnt3, was increased ninefold. Mutations in TNNT3 are etiologies of distal arthrogryposis type 2B and nemaline myopathy (142, 143). Similar to the LV of ANG II/PE-infused mice, Tnnt3 was increased in the ophthalmic artery following acute treatment with ANG II (144). Overexpression of Tnnt3 in mouse cardiac muscle results in hypertrophy and degeneration (145). Further investigation will be needed to determine if endogenous Tnnt3 is increased in cardiomyocytes or vascular smooth muscle and whether this represents a phenotypic switch in muscle function.

We also identified DEGs that could provide unanticipated clues to pathophysiological mechanisms. Here we highlight a few such examples.

Sparc (Secreted protein acidic and rich in cysteine) is increased 3.9-fold following 3 days of ANG II/PE infusion. SPARC is a matricellular protein that is required for insoluble collagen deposition and increased myocardial stiffness in response to pressure overload in mice (146). Interestingly, cardiac macrophages produce SPARC and are thought to be important for the observed increase in SPARC 3 days after transverse aortic constriction (TAC) surgery (147).

C1qtnf3 and C1qtnf6 (CTRP3 and 6, C1q and tumor necrosis factor-related protein 3 and 6) are increased 180- and 6-fold, respectively, following 3 days of ANG II/PE infusion. C1qtnf3 is increased in the heart in response to pressure overload (TAC) in mice and in dilated cardiomyopathy in humans (148). Mice lacking C1qtnf3 showed an exacerbated response to pressure overload whereas overexpression reduced cardiac hypertrophy following TAC (149). Similarly, overexpression of C1qtnf6 in the heart improved cardiac function (decreased hypertrophy, fibrosis, and myofibroblast differentiation) after myocardial infarction (73). A possible mechanism may be through inhibition of p38/CREB signaling or activation of AKT/PKB or TAK-1/JNK signaling (148–150).

Fhl1 (four and a half LIM domains 1) is increased 3.7-fold following 3 days of ANG II/PE infusion. Fhl1 encodes a scaffolding protein within the cardiomyocyte sarcomere that senses biomechanical stress and mediates the hypertrophic response. FHL1 is induced in human hypertrophic cardiomyopathy and in multiple mouse models with cardiac hypertrophy (151). Mice lacking Fhl1 showed a reduced hypertrophic response and a beneficial functional response to pressure overload through Gaq (Gq) signaling pathway (152).

A530016L24Rik (nutritionally regulated adipose and cardiac enriched, Nrac) is decreased fourfold following 3 days of ANG II/PE infusion. Nrac is abundantly expressed in the plasma membrane of adipose tissue and heart but has not been functionally characterized (72). Expression of Nrac may serve as a marker of the metabolic state of the heart under physiological stress.

Nos2 (nitric oxide synthase 2, inducible Nos2, iNos) is decreased twofold following 3 days of ANG II/PE infusion. Inflammation and TNFα induces expression of iNos in the heart. Cardiac function is known to be regulated through both vascular-dependent and -independent effects. The former mechanism involves regulation of the coronary vessel tone, thrombogenicity, and inflammatory response (51). The later modulate effects on contractile myocardial function and mitochondrial homeostasis. Interestingly, other models of HFpEF in mice demonstrated upregulation of Nos2 and potential benefits from pharmacological or genetic suppression of iNOS (21). Decreased expression of Nos2 in ANG II/PE-infused mice may suggest early stages of disease and needs to be compared at later points.

Translation to Clinical Research and Patient-Oriented Care

To enhance the utility of animal models for assessing disease diagnosis and progression, it is important to include readily available bedside biological specimens (like blood). So far, major diagnostic screening tools to differentiate patients with heart failure include serum NT-proBNP and cardiac troponin I levels. In our model, in addition to elevated NT-proBNP that is known to correlate with heart failure progression, we also demonstrated elevation of the nonconventional serum circulating markers including FGF21 (153) and FGF23 (154) (Supplemental Fig. S3).

Consistent with immunobiological and RNA-seq data, serum proteomic analysis demonstrated a predominance of upregulated inflammatory pathways at early stages of ANG II/PE treatment (3 days) that correlated with clinical inflammatory markers seen in HFpEF pathogenesis (119). On the other hand, upregulation markers of serum mitochondrial aerobic oxidation and activation of the ROS scavenging system were evident later in the course of ANG II/PE treatment (28 days). This also correlates with the complex metabolic changes in observed in HFpEF (141). Serum cardiac Troponin I (TNNI3) was significantly lower at 3 and 10 days of ANG II/PE infusion, which corresponded to immunohistochemical analysis (Supplemental Fig. S5) and RNA-seq analysis.

Study Limitations

This study has several limitations. We did not define the relative contribution of phenylephrine and ANG II to the development of the HFpEF phenotype or whether the contribution of the two drugs is largely synergistic. Based on multiple other studies, as well as known multifactorial actions of angiotensin, including its prohypertensive, proinflammatory, and profibrotic effects (48, 91, 93, 155, 156), we presume that angiotensin and activation of the RAAS system is the predominant driver of the HFpEF phenotype. This model is also limited to 28 days of ANG II/PE infusion and HFpEF phenotype investigation. Future studies will be needed to determine if continuation of ANG II/PE infusion beyond 28 days will maintain the HFpEF phenotype without converting to a HFrEF phenotype (35). We attempted to treat this model as a multisystem organ disease by assessing other organs, including lungs and kidneys. However, not all pathological situations (e.g., wound healing) and organ systems were evaluated, e.g., skin and skeletal muscles (clearly involved in this model that demonstrated physical intolerance and more likely contributed to deranged V̇o₂ based on RNA-seq downregulated metabolic homeostasis). Mitochondrial electron transport chain (ETC) and respiratory complex analysis will warrant further study. Although transcriptome data indicated that metabolic gene programs were disrupted, ANG II/PE-infused mice were not exposed to a metabolic stressor that leads to obesity, one of the clinical criteria of HFpEF (29). Finally, we did use mice in their reproductive age, whereas the majority of human patients with HFpEF syndrome are elderly (157).

The histological stratification into perivascular versus interstitial fibrosis phenotypes will require additional delineation of potentially unique mechanisms of HFpEF development or progression. We also used whole LV rather than single-cell RNA-seq, thus, cell-specific expression and lineage relationships remain to be clarified. Comparison of male and female mice did not reveal differences in echocardiographic and histopathological data, demonstrating that the ANG II/PE-induced phenotype is not dependent on sex. Finally, this study did not focus on treatment strategies, though it did demonstrate multiple pathophysiological pathways, as well as implementation of the novel noninvasive methods of HFpEF evaluation that may provide a framework to understand pathophysiological mechanisms and evaluate potential therapies.

Conclusion

The ANG II/PE infusion model demonstrated integration of major histopathological and genetic signatures of HFpEF phenotype, as well as illustrated the utility of novel noninvasive methods for HFpEF evaluation. The presence of important pathophysiological pathways makes this model an attractive tool to investigate novel features of HFpEF, which may allow for the discovery and evaluation of therapeutic targets for treatment and potential reversal of this disease.

DATA AVAILABILITY

Data will be made available upon reasonable request.

SUPPLEMENTAL DATA

Supplemental figure, table, and movie legends: https://doi.org/10.6084/m9.figshare.22812788.

Supplemental Fig. S1: https://doi.org/10.6084/m9.figshare.23058137.

Supplemental Fig. S2: https://doi.org/10.6084/m9.figshare.23058515.

Supplemental Fig. S3: https://doi.org/10.6084/m9.figshare.23058659.

Supplemental Fig. S4: https://doi.org/10.6084/m9.figshare.23058818.

Supplemental Fig. S5: https://doi.org/10.6084/m9.figshare.23059028.

Supplemental Fig. S6: https://doi.org/10.6084/m9.figshare.23059076.

Supplemental Fig. S7: https://doi.org/10.6084/m9.figshare.23059187.

Supplemental Table S1: https://doi.org/10.6084/m9.figshare.23059265.

Supplemental Table S2: https://doi.org/10.6084/m9.figshare.23258096.

Supplemental Table S3: https://doi.org/10.6084/m9.figshare.22666114.

Supplemental Table S4: https://doi.org/10.6084/m9.figshare.22666207.

Supplemental Movies S1 and S2: https://doi.org/10.6084/m9.figshare.23059262.

GRANTS

This work was funded by the Lilly Research Award Program, the Departments of Pediatrics and Developmental Biology at Washington University School of Medicine, National Center for Advancing Translational Sciences Grant UL1TR002345 [to Washington University Institute of Clinical and Translational Sciences (ICTS)], and ICTS Grant JIT857.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

D.M., S.S.O., and D.M.O. conceived and designed research; D.M. and A.K. performed experiments; D.M., T.L., S.J.M., S.S.O., and D.M.O. analyzed data; D.M., A.K., K.K.-S., S.J.M., S.S.O., and D.M.O. interpreted results of experiments; D.M., A.K., and D.M.O. prepared figures; D.M. and D.M.O. drafted manuscript; D.M., A.K., T.L., K.K.-S., S.J.M., S.S.O., and D.M.O. edited and revised manuscript; D.M., A.K., T.L., K.K.-S., S.J.M., S.S.O., and D.M.O. approved final version of manuscript.