Recovery from Heart Failure: Microvascular Mechanisms

Abstract

Background:

Heart failure (HF) is a significant global health problem. Left-ventricular assist device (LVAD) implantation serves as a bridge for patients awaiting heart transplantation. Intriguingly, LVAD support often improves cardiac histology and function, sometimes enough to avoid transplantation after LVAD removal. However, the cellular programs underlying this recovery remain unclear.

Methods:

Myocardial tissues were obtained from patients with HF at the time of LVAD implantation (pre-LVAD) and explantation (post-LVAD) for histological analysis and single-nucleus RNA sequencing (snRNA-seq). A murine model of HF recovery, combined with lineage tracing studies, was employed to define cellular sources of vascular repair. Cardiac function, fibrosis, and vascular density were assessed using echocardiography, histology, and fluorescent microsphere perfusion. A patient-derived cardiac non-myocyte culture system was established to interrogate mechanisms of cell-fate regulation.

Results:

Post-LVAD myocardial tissues exhibited reduced fibrosis and increased capillary density compared to pre-LVAD samples. Across samples, fibroblast abundance inversely correlated with endothelial cell abundance, consistent with enhanced angiogenesis during recovery. SnRNA-seq identified a fibroblast subset predisposed to undergo mesenchymal-to-endothelial transition, acquiring an endothelial cell identity. Additionally, non-myocytes from pre-LVAD hearts proliferated poorly and failed to form vascular structures, whereas non-myocytes from post-LVAD hearts displayed greater proliferative and angiogenic capacity, forming vessel-like structures, reinforcing the association of HF recovery with angiogenic reprogramming. Mechanistically, knockdown of c-Myc by siRNA shifted post-LVAD non-myocytes to a pre-LVAD-like state, while c-Myc overexpression by mRNA in pre-LVAD cells induced a post-LVAD-like phenotype, implicating c-Myc as one contributor to this fate switch. A model of heart failure recovery in mice mimicked the histological and functional changes in patients, with physiological evidence of increased microvascular perfusion, associated with a fibroblast-to-endothelial transition documented by lineage tracing.

Conclusions:

HF recovery involves reduced fibrosis and enhanced microvascularization, partly driven by fibroblast-to-endothelial cell fate transition. c-Myc functions as one regulator of this transition, offering a mechanistic entry point to develop regenerative therapies in HF.

Introduction

Heart failure (HF) is a major cause of morbidity and mortality, affecting more than 64 million people worldwide^1–3. Myocardial infarction causes loss of functional cardiac muscle and represent a major cause of HF, while approximately 40% of cases are non-ischemic in nature i.e., HF in the absence of significant coronary artery disease^{4, 5}. In HF, there is disseminated interstitial and perivascular fibrosis which may be due in part to endothelial to mesenchymal transition⁶. The accumulation of extracellular matrix (ECM) components, particularly collagen, impairs cardiac function and is associated with adverse outcomes^7–9.

Advances in medical therapy over the last two decades have improved patient outcomes in HF. Nevertheless, there is still a significant number of patients who progress to end-stage HF, necessitating advanced therapeutic strategies including heart transplantation¹⁰. Unfortunately, there is a limited availability of donor hearts with only 3,700 heart transplants performed each year. Therefore, the use of left ventricular assist devices (LVADs) has emerged as a therapeutic strategy, providing a bridge to recovery, remission, or transplantation for patients with advanced HF^11–13.

LVAD support promotes favorable cellular and structural changes within the myocardium¹⁴. We and others have observed that the use of a LVAD device may be associated with improved ventricular compliance¹⁵, reduced cardiac hypertrophy^{16, 17}, reduced cardiac fibrosis¹⁶, increased capillary density¹⁸ and bulk gene expression changes consistent with a mesenchyme to endothelial transition (MEndoT)¹⁹. These alterations may be associated with sufficient improvement in cardiac function that a transplant is no longer necessary. This phenomenon provides a unique opportunity to investigate the mechanisms underlying heart failure recovery (HFR).

To explore the mechanisms of HFR, we employed single-nuclei transcriptomic analysis of pre- and post-LVAD hearts and found a cell fate transition associated with HFR. Furthermore, we validated this cell fate transition in a murine model of HFR combined with a lineage tracing strategy, which suggests HFR may involve microvascular recovery. Specifically, angiogenic transdifferentiation expands the cardiac microvasculature to improve myocardial perfusion and is associated with favorable structural and functional changes of the heart. To define regulatory drivers, we established a patient-derived cardiac non-myocytes (NMs) culture system for mechanistic phenotyping and functional testing. The transcriptional factor c-Myc emerged as one regulator of this transdifferentiation and a potential therapeutic target. Our focus on molecular modulation of the myocardial microvasculature, and the role of angiogenic transdifferentiation^20–22, has generated a novel conceptual framework to understand HFR, and to guide the next generation of HF therapies.

Methods

Human Myocardial Samples and Experiments

Tissue Acquisition

De-identified cardiac tissue samples were collected at the time of the LVAD implant (n=13) or after explant (n=11). Specifically, we sampled the apical core of tissue removed at the time of LVAD implantation (pre-LVAD); and the left ventricle free wall adjacent to the apical region for post-LVAD samples obtained at the time of cardiac transplantation. In both cases, only mid-myocardial region of the tissue was sampled (endocardium and epicardium excised and discarded) (Fig. 1A). Cardiac tissue samples were acquired based on an institutional protocol approved by the Houston Methodist Hospital Institutional Review Board [IRB Pro 00006097] and abide by all ethical regulations corresponding to human tissue research. Tissues were collected and immediately processed for freezing and paraffin embedding for histology. Tissues were flash-frozen at −80°C for protein and RNA assays. For histology, tissues were promptly fixed in 4% paraformaldehyde. Subsequently, the samples were processed following standard techniques for paraffin embedding and sections were cut 5μm in thickness for histology. For snRNA sequencing analysis, frozen samples from five randomly selected patients from the pre- and post-LVAD groups were each pooled and processed.

Figure 1. Post-LVAD cardiac tissue manifests less fibrosis and greater endothelial density by comparison to pre-LVAD myocardium (Heart Failure).

(A) Schematic of sample acquisition from pre-LVAD and post-LVAD human hearts. (B) Masson’s trichrome staining shows reduced fibrosis post-LVAD. (C) Quantification of fibrosis, fibroblast cell count, and endothelial cell count in pre-LVAD and post-LVAD, respectively (n=5 patient tissue/group). Data are mean ± SEM; p values are from unpaired t-test: *<0.05; **<0.01; ****<0.0001. (D) CD144 staining shows an increased number of ECs in post-LVAD myocardium. (E) Pearson correlation analysis reveals a negative correlation between FB and EC count in unpaired pre-LVAD and post-LVAD tissue (n=5 per group). r: Pearson correlation coefficient.

Cardiac Fibrosis

To quantify fibrosis, sections were stained using Masson’s trichrome (Sigma-Aldrich, HT15–1KT), following the manufacturer’s instructions. Slides were examined under 10x objective using an Olympus BX61 microscope. Whole tissue sections of the heart were photographed, and fibrosis was quantified using Olympus cellSens dimension software. A user-independent computerized color cube-based selection criterion, focusing on the color spectrum of the blue dye, was applied to denote positive staining. Stained and unstained areas were measured, expressing the results as the average percentage of tissue (pixels) stained by the dye. The analysis was conducted by a single investigator who was blinded to the samples.

Fibroblast and Endothelial Cell Count

Myocardial fibroblasts and endothelial cells from unpaired pre- and post-LVAD cardiac tissue were quantified by staining for antibodies vimentin AF647 and CD144 AF647, respectively. Five microscope fields (20x) were analyzed for quantification, and the results were expressed as mean ± SD. The analysis was done by a single investigator blinded to the source of the samples. Additionally, von Willebrand factor was used to stain ECs and capillaries. (Antibody details in Expanded Methods)

Cardiac Non-myocyte Cell Isolation

Freshly received cardiac tissue underwent sequential enzymatic tissue digestion using DMEM+Collagenase (10:1) at 37°C. After disaggregation, the solution was filtered and differentially centrifuged to obtain a pellet of cardiac NMs. The resulting cell pellet was resuspended in EGM2 and divided for freezing and cell culture. For future assays, cells were stored in liquid nitrogen. The other half was plated for cell culture and immunofluorescence assay on coverslips placed in 6-well culture plates. Freshly isolated cells are referred to as P0 (passage 0), and the first passage of cells are labeled as P1 once confluent.

In vitro proliferation of NMs

Proliferation was assessed by Electric Cell-substrate Impedance Sensing (ECIS) and CellTrace^™ Violet (CTV) dilution. For ECIS, cryopreserved passage-1 NMs from pre-LVAD and post-LVAD hearts were thawed and plated in duplicate on 8-well ECIS arrays in EGM2. The plates were attached to the ECIS machine, and their normalized resistance was measured until the cells reached a confluent monolayer indicated by plateau on the graph. EGM2 was changed after the first 24 hours and subsequently every 3 days. For CTV assays, NMs were labeled with CellTrace^™ Violet (Thermo Fisher Scientific) according to the manufacturer’s instructions, cultured for 6 days at 37 °C with 5% CO₂, and analyzed by flow cytometry to quantify dye dilution.

Immunofluorescence

The coverslips from culture plates were stained on Days 3, 7 and 14. Coverslips were immune-stained with conjugated primary antibodies for endothelial cells - CD144 AF488 and for fibroblasts - Col1a AF647. The stained coverslips were then inverted and mounted on slides and imaged with a fluorescence microscope at 10x, 20x and 40x objective lens. Z-stacks were also taken using a confocal microscope.

Single Nuclei Isolation and Library Preparation for Sn-RNA Sequencing

We isolated nuclei from frozen human heart tissue samples and constructed 3’ single-cell gene expression libraries (Next GEM v3.1) using the 10x Genomics chromium system, followed by quality control analysis²³. (Details in Expanded Methods)

Sn-RNAseq Differential Expression and Cell Type Annotation

Furthermore, we conducted Differentially Expressed Genes (DEGs) analysis in one cluster versus other clusters using the Find All Markers function^23–26. (Details in Expanded Methods)

Cell Fate Transition, Trajectories, and Cell-Cell Communication with Functional Enrichment Analysis

For cell fate transition, trajectory analysis, and cell rank for directed single-cell transition mapping, we utilized scanpy (version 1.10.2)²⁷, scVelo (version 0.2.5)^28–30, Monocle3 (version 1.4.26)^31–34, CellRank (version 1.5.1)³⁵, CellChat (version 1.5.0)³⁶ and clusterProfiler R package (version 4.18.4)³⁷ with default parameters. (Details in Expanded Methods)

Animal Husbandry

All animal experiments were conducted with approval from the Houston Methodist Research Institute Institutional Animal Care and Use Committee (Houston, TX) and were in accordance with the guidelines for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 85–23, revised 1985). 2-month-old male wildtype C57BL/6 mice were bought from Envigo, Alice, TX and were individually housed under 12-h:12-h light/dark cycle with enrichment and acclimatized for 3 days. Col1a2-cre/ERT and R26RtdTomato mice were purchased from The Jackson Laboratory (Bar Harbor, ME) to generate Col1a2-creERT2: R26R^tdTomato mice³⁸. Experimental groups were identified using sequential numerical cage numbers as placed by husbandry staff.

Murine Model of Heart Failure and Recovery

Details on the murine HF model previously established by our group are described earlier¹⁹. In brief, HF was induced in 3-month-old C57BL/6 or Col1a2-creERT2: R26R^tdTomato mice by adding 0.3mg/mL L-NAME and 1% NaCl to the drinking water for 5 weeks. One week after starting water, angiotensin-II was delivered through subcutaneous osmotic pumps at a rate of 0.7 mg/kg per day for 4 weeks. Control mice did not undergo pump implantation and were fed regular diet and water (Fig. 4A). To induce Col1a2-cre tdTomato labeling, 1mg of (Z)-4-hydroxytamoxifen was administered by intra-peritoneal injection at the end of HF induction (5-week timepoint) for five consecutive days (Fig. 5A). Subsequently, hearts were harvested at 9 weeks (at the time of recovery from HF, i.e., HFR) for immunohistochemistry and flow cytometry analysis.

Figure 4. A mouse model of heart failure and recovery recapitulates human phenotypes.

(A) Schematic of the HF/HFR model and timeline. (B) Representative echocardiography: M-mode at the LV mid-papillary level; pulsed-wave Doppler mitral inflow (E, A); and tissue-Doppler mitral annular velocities (e′, a′). (C) Box plots of Ejection Fraction, heart weight, body weight, and heart-to-body weight ratio at the indicated time points (Age-matched controls, n=15; HF, n=20; HFR, n=20). (D) Schematic of fluorescent microsphere bead perfusion technique to assess vascular density. (E) Representative images and quantification of intravascular blue fluorescent microspheres (VD) and Masson’s trichrome (fibrosis) demonstrating increased VD and reduced fibrosis in HFR versus HF (VD: Control, n=14; HF, n=16; HFR, n=15. Fibrosis: Control, n=10; HF, n=11; HFR, n=10). In C, E, data are mean ± SEM. p values from one-way ANOVA (mixed) with multiple comparisons: ns>0.05; *<0.05; **<0.01; ****<0.0001. Mitral inflow velocity: E = early filling; A = atrial (late) filling. Mitral annulus velocity: e’ = passive LV filling; a’ = atrial contraction. HF: Heart Failure; HFR: Heart Failure Recovery; LV: Left Ventricle; VD: Vessel Density.

Figure 5. Angiogenic transdifferentiation during heart failure recovery in Col1a2-CreERT:R26R tdTomato mice.

(A) Schematic of fibroblast lineage tracing in the HFR mouse model. Col1a2-expressing fibroblast cells and their progenies are marked with tdTomato after tamoxifen (4-OHT) induction. (B) IF staining visualized by confocal imaging in HFR group showed CD31+ (green) tdTomato+ (red) dual staining (yellow) cells representing endothelial cells transdifferentiated from fibroblasts (arrows - dual staining cells). (C) The 3-side view by confocal microscope confirmed the dual staining of the cells that expressed both CD31 and tdTomato. (D, E) Representative plots and quantification of tdTomato+ CD31+ cells during HFR, analyzed by flow cytometry (n=5 mice per group). Data are expressed as mean ± SD. p value for unpaired t-test: ****<0.0001. Control: Non-operated tamoxifen induced age-matched mice; HFR: Heart failure recovery.

Echocardiography

Transthoracic echocardiography was performed as previously described³⁹. In brief, mice were anesthetized with Isoflurane 1.5 % and 98.5% O₂ and placed on a warming pad at 37.5°C. Measurements were performed with the Vevo 1100 imaging system and an MS400 (18–38MHz) transducer. Images were analyzed using Visual Sonics Software Vevo Lab 1.7.1 in a blinded fashion. Systolic function was assessed in the parasternal longitudinal axis (PSLAX) in B-Mode, and ventricular wall dimensions were assessed on papillary muscle level in M-Mode from the parasternal short-axis view. The E/A ratio was assessed in an apical 4-chamber view.

Mouse Microsphere Bead Perfusion Study

A modified mouse perfusion method, as described by Springer et al.⁴⁰ was employed to perfuse mice with Fluospheres carboxylate-modified 0.2 μm blue microsphere beads diluted at 1:20 with phosphate buffered saline (PBS 1X). Briefly, mice were administered deep inhalation anesthesia, the right ventricle was identified, and a small incision was made to allow excess perfusate to exit the vascular space. Using a 26-gauge needle, the left ventricle was accessed via the apex. The mouse was initially perfused with 8ml of PBS at a rate of 10ml/min using a syringe pump. Subsequently, fluorescent microsphere beads (5ml) were used for perfusion. Upon completion of the perfusion, the heart was immediately removed from the chest cavity, weighed, and placed in OCT (optimal cutting temperature) media for freezing at −80°C. Deep cryosections were cut at 10μm thickness using a cryostat for histology and imaging of the beads’ fluorescence (Fig. 4D).

Fluorescent Beads Analysis

Cryosections were imaged under 20X objective using Fluorescence microscope. The DAPI channel was utilized to visualize blue-colored fluorescent beads. Cross-sectional images of the entire heart were captured at the papillary muscle level. A blinded observer analyzed the images for vascular volume representing vessel density using CellSens Dimension Software from Olympus. On each image, a region of interest (ROI) was marked to exclude tissue folds and other artifacts. The results are expressed as the average percentage area (pixels) comprising fluorescent beads. Additionally, CD31 staining was performed on some tissue sections to confirm that beads filled the small capillaries without leaking out of the vessels. (Figure.S17A).

Flow Cytometry

Cells were freshly isolated from mouse hearts according to the manufacturer’s protocol (Miltenyi Biotec, Bergisch Gladbach, Germany). Fc receptors were blocked with anti-mouse CD16/32, and cells were stained with DAPI, CD31-APC, PDGFRa-FITC, and CD45-PE-Cy7. Analysis was performed with assistance from the Houston Methodist Research Institute Flow Cytometry Core on a BD FACS Fortessa machine. Data were analyzed with FlowJo v10.

Immunofluorescence Tissue Preparation and Confocal Microscopy

Freshly isolated mouse hearts were fixed with 4% PFA for one hour at 4°C and subsequently transferred to 30% sucrose overnight at 4°C and then embedded in OCT. Samples were sectioned at 10μM and stained using the following primary antibodies: anti-mouse CD31 rat, RFP Chicken and AF488 donkey anti-rat and AF647 goat anti-chicken as secondary antibodies respectively. The z-stack images were captured using a confocal microscope with a 40X objective lens. The images obtained were analyzed using cellSens dimension software. Merged signals and split signals were used to delineate the signals for single-cell resolution.

Statistical Analysis

The data are presented as mean ± standard deviation (SD). Differences between groups were identified using Student’s t-test (two-tailed) and between multiple groups using one-way analysis of variance (ANOVA) non-parametric or mixed with Tukey’s multiple comparisons test. The Pearson correlation coefficient test was performed to measure linear correlation. A two-tailed p <0.05 was used as the significance cut-off for all tests. Analyses were conducted using Graph Pad Prism 9.3.0 software. Graphically, all results are shown as mean ± standard error (SEM) or mean ± Standard Deviation (SD).

Data availability

All the data supporting the findings in this study are included in the main article and associated files. Source data are provided with this paper. Raw single nuclei sequencing files and processed normalized data have been submitted to the NCBI Gene Expression Omnibus (GSE253535). The GSE226314 and GSE183852 dataset is publicly available.

Code Availability

The code used for the bioinformatic analysis in this study is not publicly available but can be accessed with permission from the corresponding author.

Results

Histological Hallmarks of Heart Failure are attenuated post-LVAD.

Cardiac and myocyte hypertrophy are associated with disseminated fibrosis and reduced capillary density^7–10. We obtained LV myocardial samples from HF patients at the time of implantation of LVAD (pre-LVAD) or at the time of LVAD explantation and heart transplant (post-LVAD) (Fig. 1A). Demographics of patients are presented in Table S1. The patient cohort was composed of 24 patients with end-stage HF, receiving LVAD implantation as bridge-to-transplantation therapy at Houston Methodist Hospital.

Masson’s trichrome stain was used to evaluate and quantify myocardial fibrosis in sample from pre- or post-LVAD hearts (Fig. 1B). Tissue sections from post-LVAD hearts manifested less fibrosis than pre-LVAD hearts (10% vs. 19%, p=0.03, n=5); fewer fibroblasts (FBs) per low power field (LPF) (66±4 vs. 99±7, p<0.0001, n=5) (Fig. 1C, Figure S1A); and increased endothelial cells (ECs) (97±6 vs. 75±8, p<0.01, n=5) (Fig. 1C, D, Figure S1B). Furthermore, Pearson correlation analysis revealed a negative correlation of FB count with EC count (p=0.01) (Fig. 1E).

Taken together, compared to pre-LVAD, post-LVAD heart tissue exhibited reduced fibrosis and increased capillary density.

Single-nucleus Transcriptomics Reveals Cellular Transitions post-LVAD

Guided by our observation that HF myocardium exhibits increased FBs and reduced ECs, a pattern that reverses post-LVAD (Fig. 1), we hypothesized that ECs undergo endothelial-to-mesenchymal transition (EndoMT) in HF and that this process is reversed during recovery¹⁹. To test this hypothesis, we performed single-nuclei RNA sequencing (snRNAseq)^41–44 on pooled left-ventricular tissue from five pre-LVAD and five post-LVAD hearts (Fig. 2A), identified major cell types after integration and batch correction analysis (Figure S2), and interrogated EC to FB dynamics. Pseudo-time trajectory analysis coupled with supervised mapping indicated a transition from global EC to FB phenotype in pre-LVAD hearts, consistent with EndoMT and development of cardiac fibrosis⁴⁵, whereas post-LVAD we observed the reverse transition to an EC phenotype that could promote restoration of the cardiac microvasculature (Fig. 2B). Recognizing compartmental heterogeneity, we subclustered ECs and FBs and identified three FB and five EC subpopulations to localize these transitions (Fig. 2C, D). Cell-composition analysis revealed a significant increase of an EC subtype (EC5) and a concurrent decrease of an FB subtype (FB2) post-LVAD (Fig. 2E left). RNA velocity analysis supported directional flow from EC5 to FB2 pre-LVAD and the reverse FB2 to EC5 post-LVAD (Fig. 2E right, Figures S3A, B). This cell fate transition was accompanied by reduced expression of fibrosis-associated genes in FB2 and increased expression of angiogenesis-associated genes in EC5 post-LVAD (Figures S3C, D). Collectively, these results are consistent with reports of EndoMT in HF; and suggest for the first time that MEndoT may occur post-LVAD in association with heart failure recovery.

Figure 2. Single-nuclei RNA seq reveals a mesenchymal to endothelial transition in post-LVAD cardiac samples.

(A) Schematic of the snRNA-seq workflow for pre-LVAD and post-LVAD cardiac tissue. (B) Pseudotime trajectories linking fibroblast (FB) and endothelial cell (EC) states in pre-LVAD and post-LVAD. (C) UMAP showing EC and FB subcluster heterogeneity. (D) Dot plot of marker genes across EC and FB subtypes. (E) Proportional abundance analysis showing increased EC5 and decreased FB2 in post-LVAD versus pre-LVAD (two-proportion z-test) (left panel). RNA velocity indicates EC5 to FB2 flow in pre-LVAD that reverses toward FB2 to EC5 in post-LVAD (right panel). LVAD: Left Ventricular Assist Device; UMAP: Uniform Manifold Approximation and Projection; VWF: Von Willebrand Factor; PECAM1: Platelet and Endothelial Cell Adhesion Molecule; FLT1: Fms-related receptor tyrosinase kinase 1; EPAS1: Endothelial PAS Domain Protein 1; CDH5: Cadherin 5; LUM: Lumican; FBLN1: Fibulin-1; COL1A2: Collagen type I alpha 2 chain; PDGFRA: Platelet-derived growth factor receptor alpha; NOX4: NADPH oxidase 4.

At the gene-program level, the EC5 subset in pre-LVAD hearts showed elevated expression of fibrosis-associated genes, including connective tissue growth factor (CCN2)⁴⁶, Periostin (POSTN)⁴⁷ and Serpine-1⁴⁸ (Figure S4A). In contrast, the post-LVAD gene expression was associated with downregulation of fibrotic pathways and diminished expression of fibrosis markers like Transforming Growth Factor Beta 1 (TGFβ1) and the Angiotensin II Type 1 receptor (AGTR1) within EC5 (Figure S4B). Concomitantly, genes linked to tissue remodeling and cardiac development, A Disintegrin and Metalloproteinase with Thrombospondin motif (ADAMTSL1)⁴⁹, BMP-binding endothelial regulator (BMPER)⁵⁰ and cell migration inducing hyaluronidase 2 (CEMIP2)⁵¹, were increased in EC5 in the post-LVAD samples (Figure S4A). Cell–cell communication analysis revealed strengthened ligand–receptor interactions pertinent to angiogenesis, vascular repair, and ECM remodeling post-LVAD (Figure S4C). We observed increased FB2 to EC5 signaling through LAMA2 (laminin subunit alpha 2, an ECM protein) -integrin (ITGA/ITGB) pairs, which support cell-ECM adhesion, survival, and migration⁵². Gene Ontology (GO) enrichment analysis further suggested enrichment of angiogenesis and EC migration pathways post-LVAD (Figures S4D, S5–6). Altogether, these single-cell transcriptional findings align with an EC fate transition and attenuation of fibrosis during recovery.

To mitigate potential bias introduced by the pooled samples, we validated these transitions in an independent non-pooled snRNA-seq cohort⁵³. We reanalyzed a publicly available dataset comprising five paired pre- (HF) and post-LVAD (HFR) samples from patients with documented heart failure recovery as assessed by improved ejection fraction. The paired, non-pooled design effectively controls inter-patient heterogeneity and sampling bias. Using the same processing and integration pipeline as for our snRNA-seq dataset, we observed patterns consistent with our findings, with pseudotime trajectory analysis revealing evidence of EndoMT in HF and MEndoT during HFR (Figures S7 and S8A). We further identified six fibroblast subclusters and eight endothelial cell subclusters in this dataset (Figure S8B). Notably, we observed an expansion of a specific endothelial subpopulation (EC3) and a concomitant reduction of a fibroblast subpopulation (FB4) post-LVAD (Figure S8C left panel). RNA velocity further supported a transition from EC3 to FB4 in HF (pre-LVAD) that reversed to FB4 to EC3 after HFR (post-LVAD; Figure S8C right panel). Here, EC3 is a transit cell population (Figure S8D, E).

Furthermore, we analyzed 10 healthy donor (healthy control; HC) samples⁵⁴ from GSE183852 using anchor-based integration followed by label transfer and UMAP visualization. As shown in Figure S9, RNA velocity analysis revealed minimal evidence of state transitions between EC5 and FB2 in HC hearts. This contrasts with the dynamic EC5 to FB2 state transitions observed in pre-LVAD and post-LVAD. These findings indicate that healthy hearts exhibit relatively stable endothelial and fibroblast states, lacking the EC5 to FB2 transition seen in HF or the FB2 to EC5 transition observed during recovery.

Together, these analyses support the notion that fibroblast-endothelial cell transition contributes to restoration of the cardiac microvasculature and functional recovery following LVAD implantation.

Endothelial Cells Transition between Subsets during Recovery.

The partition-based graph abstraction (PAGA) velocity analysis and pseudo-time trajectory mapping suggested a shift of EC5 toward other endothelial states EC2, EC3 and EC4 during post-LVAD and a shift from EC2 towards EC1 subset (Fig. 3A–C, Figure S10). To understand the significance of these transitions, we compared gene expressions of the subsets. Notably, endothelial growth factor and angiogenesis-related pathways involved in vascular remodeling and cardiac repair⁵⁵ were upregulated in all the EC subtypes comparative to EC5 (Fig. 3D). These data were consistent with the notion that the EC5 subset may represent a less mature and transitional EC state, whereas other ECs are more mature states (Figure S11–14). In this regard, it is known that the protein tyrosine phosphatase receptor type M (PTPRM) is expressed in endothelial cells of all major organ systems, regulating barrier integrity and mechanotransduction⁵⁶. Furthermore, it is accepted that PECAM1⁵⁷ (CD31), a protein abundant in ECs, plays a key role in inter-endothelial cell junctions and angiogenesis. With respect to these two proteins, we noted an increase in ligand-receptor signaling in the transition from EC5 to other ECs in post-LVAD for both PTPRM and PECAM1 (Fig. 3E). This observation may be consistent with a transition to a more stable EC state (EC1). In this regard, cadherin 5 (CDH5)⁵⁸ and claudin 5 (CLDN5)⁵⁹ (endothelial cell adhesion molecules maintaining a stable endothelial barrier) were more highly expressed in EC1 vs EC5 or EC2 (Fig. 3F). Each of the subsets (EC 1–5) expressed endothelial identity genes such as histone-lysine N-methyltransferase (MECOM) which is crucial for endothelial cell differentiation^{60, 61} (Fig. 3F, Figure S15).

Figure 3. Tracing the fibroblast-endothelial transition via cellular dynamics and gene expression in endothelial cells in post-LVAD myocardium.

(A) PAGA velocity analysis suggests a transition probability from the EC5 subtype to other EC subtypes post-LVAD. (B) Pseudo-time trajectory outlines a developmental pathway connecting EC5 to other EC subtypes during recovery. (C) PAGA transition confidence score supports a shift from EC5 towards EC2, EC3 and EC4 and from EC2 towards EC1 in post-LVAD myocardium. (D) GO enrichment analysis shows angiogenesis and endothelial migration-related pathways are upregulated in all the EC subtypes relative to EC5 in post-LVAD myocardium. (E) Enhanced ligand-receptor signaling associated with angiogenesis and vascular repair in ECs post-LVAD. CellChat default method for the permutation test to calculate significant communication. (F) An upregulation is observed in the gene expression of angiogenesis and endothelial lineage related pathways in all EC subtypes (EC1 to EC4) compared to EC5 in post-LVAD myocardium. CA4: Carbonic Anhydrase 4; CDH5: Cadherin 5; CLDN5: Claudin 5; DLL4: Delta-Like 4; EC: Endothelial Cell; EFNB2: Ephrin B2; EGFL7: Epidermal Growth Factor-Like Protein 7; EPHB4: Ephrin Type-B Receptor 4; F8: Coagulation Factor VIII; PTPRTM: Protein Tyrosine Phosphatase Receptor Type M; FLT1: Fms related receptor tyrosinase kinase 1; GO: Gene Ontology; LVAD: Left Ventricular Assist Device; MECOM: MDS1 and EVI1 Complex Locus; NR2F2: Nuclear Receptor Subfamily 2 Group F Member 2; PAGA: Partition based Graph Abstraction; PECAM1: Platelet and Endothelial Cell Adhesion Molecule; RGCC: Regulator of Cell Cycle; VEGFC: Vascular Endothelial Growth Factor C; VWF: Von Willebrand Factor.

Recovery from Heart Failure is a Vascular Recovery.

To study the HF and recovery in vivo, our group has established a mouse model to mimic the patient condition¹⁹. Briefly, heart failure (HF) is induced by oral administration of high salt and L-N^G-Nitro arginine methyl ester (L-NAME) together with subcutaneous administration of angiotensin II via an osmotic pump for 5 weeks (Fig. 4A). After cessation of this HF induction protocol, heart failure recovery (HFR) occurs over the ensuing 4 weeks as assessed by echocardiography and histology.

In the current study, the pharmacological induction of HF was manifested at week 5 by a reduction in ejection fraction (from 56 ± 4.8% to 36 ± 11%, p=0.001), and in the E/A ratio (from 1.8 ± 0.3 to 1.6 ± 0.7); together with an increase in isovolumetric relaxation time (IVRT; from 18.8 ± 3.9 to 26.9 ± 1.8ms, p=0.004), left ventricular mass to body weight ratio (from 3.3 ± 0.5 to 4.8 ± 0.7, p=0.01); and heart weight (from 166 ± 11 to 212 ± 42 mg, p=0.03). After cessation of the drugs in week 5, these physiological and echocardiographic features of HF normalized by week 9 (HFR; Figs. 4B–C; Figure S16; Video S1–3).

To investigate whether HFR is associated with vascular recovery, we quantified vascular density in mice using fluorescent microsphere bead perfusion (Fig. 4D). In HF mice, there was a reduction in percentage area of fluorescence (6.9 ± 1%) compared to control mice (10.8 ± 1%), p<0.0001. By contrast HFR mice manifested a restoration of fluorescence area (10.5 ± 0.7%), p<0.0001 (Fig. 4E, Figure S17A). To exclude the possibility that increased capillary density simply reflects reduced cardiomyocyte hypertrophy, we calculated the ratio of vessel density to cardiomyocyte size. Consistently, HF mice showed a reduced vessel-to-cardiomyocyte ratio relative to controls, whereas HFR mice exhibited a significantly higher ratio than HF mice (Figure S17B). Together, these data indicate that vascular density is diminished in HF but restored during HFR. Fibrosis was quantified using Masson’s trichrome staining. Fibrosis increased in HF (control vs HF p<0.0001), whereas during HFR, fibrosis decreased (HF vs HFR p=0.0003) (Fig. 4E). Cardiac tissue fibrosis in HF and HFR mice mirrors changes seen in human pre-LVAD and post-LVAD cardiac tissue samples. The increase in vessel density in HFR mice suggests a vascular component of recovery.

Microvasculature restoration in Heart Failure is partly due to the MEndoT.

To evaluate whether angiogenic transdifferentiation contributes to vascular recovery after HF, we performed a fibroblast lineage tracing study. In Col1a2-creERT: R26R^tdTomato fibroblast lineage tracing mice^62–64, fibroblasts that have ever expressed Col1a2 are permanently labeled with tdTomato following tamoxifen. Mice underwent our HF protocol, tamoxifen was administered at week 5 to label fibroblasts, and hearts were harvested at week 9 (Fig. 5A). Immunohistochemistry revealed a subset of tdTomato+ lineage-traced fibroblasts co-expressing the endothelial marker CD31 (Fig. 5B, C). Flow cytometry analysis corroborated this transdifferentiated population in the HFR group but not in controls (Fig. 5D, E, Figure S18A). Although the observed frequency was low, it is likely underestimated because our tracing labeled only 17–24% of non-myocytes (NMs), whereas fibroblasts comprise ≥35% of the NM compartment (Figure S18B). Together, these data indicate that MEndoT of cardiac fibroblasts occurs during recovery from HF.

A patient-derived cardiac NM culture system reveals Enhanced Proliferation and Plasticity post-LVAD.

Cardiac NMs are primarily ECs, FBs, and mural cells (MCs)⁴². To validate the finding of interplay between ECs and FB during HFR, we established a patient-derived NM culture by dissociating fresh human myocardium and culturing purified NMs in endothelial growth medium (EGM) (Fig. 6A). This system revealed striking, previously unreported differences between pre- and post-LVAD samples: NMs from HF patients expanded poorly and adopted a fibroblastic morphology, whereas post-LVAD NMs reached confluence rapidly and exhibited an angiogenic phenotype, assembling network-like structures (Fig. 6B, C). The vascular-like structures expressed EC markers (CD144) and were surrounded by COL1A2⁺ fibroblastic cells, with dual-positive cells on the edges of the structures (Fig. 6D). Thus, post-LVAD NMs display greater proliferative capacity, cellular plasticity, and angiogenic potential than pre-LVAD NMs.

Figure 6. Patient-derived cardiac non-myocyte culture system reveals a proliferative angiogenic phenotype post-LVAD.

(A) Schematic of isolation of human cardiac NMs from pre-LVAD and post-LVAD tissue using an enzymatic tissue disaggregation protocol. (B) ECIS (Electric Cell-substrate Impedance Sensing) assay showing cardiac NMs from post-LVAD (n=3) proliferate faster compared to those from pre-LVAD (n=3) tissue. (C) IF staining of NMs from pre-LVAD and post-LVAD tissue in culture at day 3, day 7, and day 14 shows differences in growth patterns and cell types. (D) NMs from post-LVAD form vascular-like structures by day 14 in culture, which stain for EC marker (CD144, Green) surrounded by fibroblasts (COL1A, Red) and dual staining (yellow) cells on edges (arrow). EC: Endothelial Cell; ECIS: Electric Cell-Substrate Impedance Sensing; IF: Immunofluorescence; LVAD: Left Ventricular Assist Device; NM: Non-myocyte; NR: Normalized Resistance.

c-Myc is a functional mediator of angiogenic transdifferentiation linked to recovery.

To identify drivers of MEndoT during HFR, we computed pseudobulk expression for EndoMT and MEndoT-related gene sets and correlated these with patient ejection fraction (EF) using five paired pre-/post-LVAD responder samples (GSE226314)⁵³. MYC and SNAI2 were upregulated post-LVAD and positively correlated with EF, whereas TWIST1/2 were downregulated and negatively correlated (Fig. 7). Functionally, in a 14-day small-molecule protocol converting human fibroblasts into induced ECs (iECs)^{21, 65, 66}, c-Myc knockdown significantly decreased the iEC fraction, whereas MYC overexpression increased it (Fig. 8A–B, Figure S19). In the patient-derived NM system, c-Myc knockdown with two siRNAs across six post-LVAD donors reduced proliferation and disrupted vessel-like organization (Fig. 8C, D; Figures S20, 21). Conversely, MYC mRNA overexpression in five pre-LVAD donors enhanced proliferation and network formation and yielded COL1A2⁺/CD144⁺ fibroblast-derived ECs (Fig. 8D, E; Figures S21, 22). Collectively, these perturbations show that reducing c-Myc shifts post-LVAD NMs toward a pre-LVAD-like state, whereas increasing c-Myc drives pre-LVAD NMs toward a post-LVAD-like, pro-angiogenic state.

Figure 7. Candidate drivers of MEndoT during recovery from heart failure.

(A) Pseudobulk expression analysis (snRNA-seq) of upregulated genes (MYC, SNAI2, TWIST1, and TWIST2) in fibroblasts from paired pre- and post-LVAD patient samples (n=5 pairs). p values were calculated using a paired two-tailed t-test: *<0.05; **<0.01. (B) Repeated-measures correlation (rmcorr) was performed to evaluate within-patient relationships between ejection fraction rate and MYC, SNAI2, TWIST1, and TWIST2 expression across pre- and post-LVAD time points. Each line denotes an individual patient. rmcorr coefficients (r) and P values are indicated (n=5 patients).

Figure 8. c-Myc drives MEndoT during recovery from heart failure.

(A) Flow-cytometric quantification of CD31⁺ induced endothelial cells (iECs) after a 14-day transdifferentiation protocol using BJ fibroblasts (n=6 biologically independent replicates per group), normalized to the SiCtrl group. (B) Transdifferentiation efficiency in BJ fibroblasts with MYC-overexpression (OE-MYC) versus those treated with GFP-Overexpression (OE-GFP) control (n=5 biologically independent replicates per group), normalized to the OE-GFP group. (C) Proliferation rates of NMs isolated from post-LVAD sample (#482, #501) treated with SiCtrl or Si-MYC-A/B, measured by ECIS assay and CTV labeling assay. (D) Immunofluorescence images at day 14, illustrating non-myocytes from post-LVAD samples treated with SiCtrl or Si-MYC-A/B, and NMs from pre-LVAD samples treated with GFP or MYC overexpression. Scale bar is 50 μM. (E) Proliferation rates of NMs from pre-LVAD samples (#406, #508) treated with GFP control or MYC overexpression, measured by ECIS assay and CTV labeling assay. In C, E, for ECIS assay (left panel), n=2 independent replicates per group; for CTV labeling assay (middle and right panel, n=3 independent replicates per group. In A, B, C, E, data are mean ± SD. p values are from one-way ANOVA: **<0.01, ***<0.001; ****<0.0001. NMs: non-myocytes; ECIS: Electric Cell-substrate Impedance Sensing; NR: Normalized Resistance; CTV: CellTrace^™ Violet; MFI: Mean Fluorescence Intensity, inversely correlated with cell proliferation. Numerical codes (#482, #501, #406, #508, #455, and #498) represent anonymized patient sample identifiers.

Taken together, our data supports c-Myc as one mediator of MEndoT and microvascular restoration during recovery from HF.

Discussion

The current study indicates that recovery from HF may involve microvascular expansion to improve cardiac perfusion, partly through angiogenic transdifferentiation. We compared histological, molecular, and cellular attributes of human cardiac samples during LVAD implantation (pre-LVAD) and transplantation (post-LVAD). In pre-LVAD specimens, we observed a transcriptional signature and RNA velocity consistent with EndoMT, whereas in post-LVAD hearts, we found a signature consistent with MEndoT. Consistently, in a murine HF model we observed interstitial fibrosis and capillary density reduced with impaired cardiac function; during HFR, interstitial fibrosis was reduced, capillary density increased, and cardiac function improved. Lineage tracing studies demonstrated that the increase in capillary density was associated with a fibroblast-to-endothelial cell fate transition that we have termed angiogenic transdifferentiation⁶⁷. Finally, using a patient-derived cardiac non-myocyte culture system, we found that NMs from HF showed impaired proliferation and a fibroblastic phenotype, whereas NMs from post-LVAD hearts exhibited greater proliferation and plasticity and enhanced vascular-network formation. c-Myc emerged as a functional mediator that promotes angiogenic transdifferentiation linked to recovery.

Current strategies for managing left ventricular failure focus on reducing afterload and preload, controlling arrythmias, and modulating neurohormonal influences. While positive inotropic support is useful during acute decompensation, long-term inotropic activation has not been useful⁶⁸. Methods to increase myocyte proliferation or to enhance myocardial metabolism are experimental^{69, 70}. Our HF patients post-LVAD were all on Guideline Directed Medical Therapy (GDMT) and optimized LVAD flow settings, which management may have contributed to the biological recovery that was observed post-LVAD in this study. Some amount of hemodynamic recovery from heart failure may be observed with medical treatment alone, but it is not known if such recovery is associated with changes in interstitial fibrosis, capillary density and perfusion.

Although HF treatments have primarily targeted myocytes, NMs also play a crucial role. Recent findings indicate that LVAD patients often experience improved cardiac function, reduced fibrosis, and increased capillary density^{71, 72}. There may be an important role for a vascular niche and non-cardiomyocyte signaling in recovery from HF^{55, 73}. Our study highlights a novel interaction between fibroblasts and endothelial cells in the recovering heart, suggesting that understanding these mechanisms could provide insights into cardiac tissue regeneration and recovery.

Our snRNA-seq analyses reveal significant interactions between cellular transitions and gene expression in cardiac tissue. We noted a downregulation of fibrosis markers/mediators, including COL1A1, COL3A1, TGF-β1 and AGTR1^74–76 particularly in the EC5 subset. As each of these genes play a role in the accumulation of ECM, their reduced expression in the post-LVAD specimens may facilitate a recovery from cardiac fibrosis. Interestingly, we found the upregulated expression of inhibin beta A (INHBA) in EC5 subset of post-LVAD tissue, linking it to cardiac regeneration⁷⁷. Following LVAD implantation, angiogenic genes were also upregulated in EC5, suggesting a shift from fibrotic pathology to a more reparative environment. PAGA velocity analysis revealed transitions from EC5 post-LVAD to other EC subtypes particularly from EC2 to the more mature EC1, which expressed higher levels of PECAM1⁵⁷ and PTPRM⁷⁸. Increased expression of endothelial markers- such as vWF⁷⁹, FLT1⁸⁰, and EGFL7⁸¹ in all EC subtypes compared to EC5 post-LVAD further supports this maturation^82–84. Our data indicates that EC5 is a transitional endothelial population that contributes post-LVAD HFR.

We employed a mouse model of HF and recovery closely mirroring our observations in patients, including cardiac function and histological changes. Microsphere bead perfusion confirmed increased vascular volume during HFR. To trace the origin of the transdifferentiating cells, we employed the Col1a2-creERT2: R26R^tdTomato fibroblast lineage tracing strategy. The dual-staining cells during HFR expressed an endothelial marker CD31 and fibroblast tracing marker tdTomato, confirming MEndoT. Whereas only 5.6% of cells exhibited angiogenic transdifferentiation, this is likely an underestimate of the actual number of transdifferentiating cells due to incomplete FB labeling efficiency.

Notably, study context matters. He et al.⁸⁵ reported no MEndoT in their ischemic model, with cardiac FBs expanding post-injury but not contributing to new blood vessel formation. This finding may help explain lower recovery rates post-LVAD in ischemic HF. Conversely, Dong et al.⁸⁶, observed MEndoT during cardiac hypertrophy in a transaortic constriction model, indicating a compensatory response to increased afterload. Our HFR model mirrors the histological and functional changes seen in human hearts post-LVAD and may therefore recapitulate the mechanisms in human HFR.

To further probe mechanisms, we established a patient-derived NM culture system using cardiac tissue from pre- and post-LVAD samples, enabling direct observation of cell fate dynamics and interactions. Post-LVAD NMs displayed greater proliferative capacity and enhanced endothelial network-forming potential. Subsequently, we compared paired pre- and post-LVAD samples from patients that had recovered by echocardiographic measures. Correlating pseudobulk fibroblast expression in these paired clinical responders revealed that MYC, SNAI2, and TWIST1/2 may facilitate MEndoT during HFR.

c-Myc is a transcriptional factor and as one of the Yamanaka factors, can initiate cell-state transitions and stimulate proliferation^87–89. For this reason, we explored its possible role in angiogenic transdifferentiation in our well-established protocol for fibroblast-to-endothelial direct transdifferentiation. In this system, c-Myc overexpression enhanced, while c-Myc inhibition reduced, angiogenic transdifferentiation. Similar observations were made using our NM culture system where c-Myc overexpression induced an angiogenic phenotype, whereas its downregulation attenuated this phenotype.

It is important to note that c-Myc is pleiotropic and its actions are highly context dependent: c-Myc can support angiogenesis and vascular remodeling, whereas sustained activation in cardiomyocytes may exacerbate pathological remodeling and worsen heart failure^90–94. These considerations argue for cell-type–specific and temporally restricted c-Myc modulation ideally targeting fibroblasts rather than global activation. Although MYC is an oncogene in cancer, expression of c-Myc in vivo did not lead to tumor formation⁸⁸. Advances in transient, non-integrating mRNA/LNP delivery and tissue/cell-directed targeting may enable short-term, fibroblast-restricted c-Myc expression to drive microvascular expansion while mitigating oncogenic risk. Overall, our data positions c-Myc as a context-dependent mediator of HFR and a potential therapeutic target.

Several limitations merit attention. First, patient heterogeneity introduces biological variability that can attenuate effect sizes and increase the risk of false negatives; paired sampling would mitigate this source of noise. Accordingly, we confirmed our results in an independent, clinically annotated snRNA seq cohort of paired samples. Second, there are differences between human and mouse models in cardiac physiology and HF pathophysiology, as well as in the degree of genetic and epigenetic variability. While the HF animal model in this study may mimic non-ischemic heart failure, it may not fully capture the clinical diversity of this patient cohort. Furthermore, one might argue that the spontaneous recovery observed with cessation of external noxious agents (as in our mouse model) might have different mechanisms from the recovery achieved with a direct intervention (i.e. LVAD placement). Thus, extrapolations of data from murine models to human disease must be cautious. With these limitations in mind, we propose that recovery from heart failure is in part a microvascular recovery. An angiogenic transdifferentiation contributes to the microvascular recovery. This new paradigm may lead to a novel therapeutic avenue for heart failure treatment.

Supplementary Material

Expanded Methods

Figures S1 – 22

Tables S1

Videos S1–3

Clinical Perspective.

1. What is new? (no more than 100 words, formatted as 2–3 bullets)

   • Recovery from heart failure, as with a left ventricular assist device in humans, or after withdrawal of hypertensive drugs in a murine model, is associated with an increase in capillary density

   • The increase in capillary density is associated with improved cardiac perfusion

   • The increase in capillary density is in part due to transdifferentiation of cardiac fibroblasts to endothelial cells

2. What are the clinical implications? (no more than 100 words, formatted as 2–3 bullets).

   • These data indicate that microvascular rarefaction, disseminated fibrosis, and impaired cardiac perfusion that accompany heart failure can be reversed.

   • The recovery of the microvasculature is associated with a recovery of cardiac function

   • Understanding the molecular mechanisms of microvascular recovery will lead to novel therapeutic approaches to treating heart failure

Non-standard Abbreviations and Acronyms

ADAMTSL1

A Disintegrin and Metalloproteinase with Thrombospondin motif

AGTR1

Angiotensin II Type 1 receptor

BMPER

BMP-binding endothelial regulator

CCN2

connective tissue growth factor

CDH5

Cadherin 5

CEMIP2

Cell migration inducing hyaluronidase 2

CLDN5

Claudin 5

CTV

CellTrace^™ Violet

DEG

Differentially expressed gene

EC

Endothelial Cell

ECIS

Electric cell substrate impedance sensing

ECM

Extracellular Matrix

EndoMT

Endothelial-to-mesenchymal transition

FB

Fibroblast

GDMT

Guideline Directed Medical Therapy

GO

Gene ontology

HC

Healthy Control

HF

Heart Failure

HFR

Heart Failure Recovery

INHBA

Inhibin beta A

ITG

Integrin

IVRT

Isovolumetric relaxation time

LAMA2

Laminin subunit alpha 2

L-NAME

L-N^G-Nitro arginine methyl ester

LVAD

Left Ventricular Assist Device

MEndoT

Mesenchymal-to-endothelial transition

MECOM

Histone-lysine N-methyltransferase

NM

non-myocyte