Fibroblast-specific loss of TGF-β signaling mediates lipomatous metaplasia in the infarcted heart

Abstract

Background:

In patients surviving myocardial infarction (MI), adipose tissue infiltration within the scar is a common pathologic finding and has been suggested to contribute to dysfunction and arrhythmogenesis. However, the cellular mechanisms for lipomatous metaplasia after infarction remain enigmatic. Our study reveals a novel molecular mechanism that mediates fibroblast-to-adipocyte conversion and causes fatty infiltration in the infarcted heart.

Methods:

Mice with fibroblast-specific disruption of Transforming Growth Factor (TGF)-β or Smad-dependent signaling, and corresponding controls were generated and underwent reperfused and non-reperfused MI protocols. Echocardiography, histological studies, and transcriptomic analysis were used to study effects on cardiac repair and remodeling. Lineage tracing experiments were used to document fibroblast-to-adipocyte conversion. In vitro, the effects of genetic disruption or pharmacologic blockade of TGF-β signaling on mouse cardiac fibroblasts were examined. ScRNA-seq data from human patients with MI were analyzed to assess fibroblast-to-adipocyte conversion.

Results:

Fibroblast-specific abrogation of TGF-β signaling through deletion of the type 2 TGF-β receptor (TbR2) increased the incidence of early post-infarction cardiac rupture, promoting a matrix-degrading fibroblast phenotype. Moreover, fibroblast-specific TbR2 loss resulted in replacement of 30–40% of the mature scar with adipocytes in both reperfused and non-reperfused infarcts. Lineage tracing demonstrated that fibroblast-specific TbR2 loss promotes fibroblast-to-adipocyte conversion through effects that involve Smad-independent cascades. In vitro, TbR2 inhibition induced expression of adipogenesis-associated genes in primary cardiac fibroblasts. Genetic TbR2 loss promoted conversion of infarct fibroblasts to adipocytes upon stimulation with adipogenic medium. A subpopulation of fibroblasts in human fibrotic infarcted hearts acquired expression of adipocyte genes.

Conclusions:

TGF-β signaling plays a central role in maintaining fibroblast cell specification following cardiac injury. Fibroblast-to-adipocyte conversion induced by disrupted TGF-β signaling may underlie scar-associated lipomatous metaplasia and may play an important role in the pathogenesis of heart failure and arrhythmogenesis in patients with ischemic cardiomyopathy.

INTRODUCTION:

The adult mammalian myocardium has very limited endogenous regenerative capacity. Thus, repair of the heart after myocardial infarction (MI) is dependent on timely activation of matrix-secreting reparative fibroblasts and myofibroblasts that deposit a collagenous matrix network, preserving the structural integrity of the ventricle and protecting the heart from catastrophic rupture, or excessive dilation¹,²,³,⁴,⁵,⁶,⁷. Infarct fibroblasts undergo dynamic phenotypic transitions critical for the reparative response⁸. During the proliferative phase of cardiac repair, myocardial fibroblasts expand, secrete large amounts of extracellular matrix proteins, and undergo myofibroblast conversion, expressing contractile proteins such as α-smooth muscle actin (α-SMA)⁹,¹⁰. As the scar matures, myofibroblasts become de-activated and convert to matrifibrocytes, fibroblast-like cells that do not express myofibroblast markers, but synthesize genes encoding specialized matricellular and tendon proteins¹¹. Although these phenotypic transitions highlight the remarkable plasticity of infarct fibroblasts, effective cardiac repair requires preservation of fibroblast identity and cell specification throughout the healing response. Defects in the reparative cellular response, resulting in infiltration of the infarct with fatty¹²,¹³ or calcified tissue¹⁴, have been reported in both experimental models of cardiac injury and in human patients with myocardial infarction and may contribute to adverse post-infarction remodeling and arrhythmogenesis¹²,¹⁵,¹⁶.

Cardiac imaging and biopsies of autopsied hearts have documented that adipose tissue infiltration is a common companion to scar formation in patients with ischemic heart disease. Histological studies have demonstrated that lipomatous metaplasia in fibrotic lesions is present in ~70% of patients with ischemic cardiomyopathy¹⁷. Moreover, clinical data coupled with computational analysis suggest the involvement of fatty infiltration in the pathogenesis of ventricular dysfunction and arrhythmogenesis after infarction¹². Despite the high prevalence and potential clinical significance of scar-associated adipose tissue infiltration, the cellular mechanisms for adipogenesis in the infarcted myocardium remain enigmatic. Our study reveals for the first time the cellular and molecular basis of lipomatous metaplasia in ischemic heart disease, demonstrating that selective disruption of TGF-β signaling in infarct fibroblasts promotes their conversion to adipocytes, and results in extensive infiltration of the scar with adipose tissue. Fibroblast to adipocyte transdifferentiation upon abrogation of TGF-β signaling requires disruption of Smad-independent pathways. ScRNA-seq data from human MI patients supports the notion that a subpopulation of fibroblasts undergoes adipocyte conversion.

METHODS (detailed methodology is provided in the Online Supplement):

Data availability statement:

The data supporting the findings of the study are available from the corresponding author upon reasonable request. All RNA-seq data are available under GEO Accession No: GSE277184.

In vivo and in vitro experiments:

All animal experiments were performed according to the guidelines issued by the Animal Care and Use Committee at Albert Einstein College of Medicine and conform to the Guide for the Care and Use of Laboratory Animals published by the NIH. Mice with fibroblast-specific loss of TbR2 and Smad4 were generated using the Col1a2^CreER driver (Figures S1–S3). Models of reperfused and non-reperfused MI were used. Echocardiography, quantitative morphometry and histology, immunohistochemistry, flow cytometric sorting of fibroblasts, qPCR and RNA-seq were used to study the effects of fibroblast-specific disruption of TGF-β signaling. In vitro, cardiac fibroblasts from control and infarcted hearts were isolated and stimulated to study the role of TGF-β disruption on adipogenic conversion. To assess fibroblast-to-adipocyte conversion in patients with MI, we analyzed scRNA-seq data from Kuppe et al.¹⁸.

Statistical analysis.

Sample size estimates were based on our own experience with comparable experimental studies to achieve 90% power at a significance level of 0.05. For all analyses, normal distribution was tested using the Shapiro-Wilk normality test. For comparisons of 2 groups, an unpaired 2-tailed Student’s t test using (when appropriate) Welch’s correction for unequal variances was performed. The Mann–Whitney test was used for comparisons between 2 groups that did not show Gaussian distribution. For comparisons of multiple groups, 1-way ANOVA was performed followed by Sidak’s multiple comparison test for normal distributions, or the Kruskal-Wallis test followed by Dunn’s multiple comparison test for non-Gaussian distributions. Survival analysis was performed using the Kaplan-Meier method and mortality was compared using the log-rank test. Data are expressed as means±SEM. Statistical significance was set at 0.05. Statistical analysis was performed with Graphpad Prism 9.4.

RESULTS:

Fibroblast-specific TGF-β signaling protects the infarcted heart from rupture.

We generated mice with inducible conditional fibroblast-specific loss of TGF-β receptor 2 (TbR2), the only type 2 TGF-β receptor that transduces TGF-β1/2/3 signaling (FTbR2KO mice). Cardiac fibroblasts harvested from FTbR2KO mice had complete absence of p-TbR2, indicating abrogation of all TGF-β signaling (Figure 1A–B). Fibroblast-specific TbR2 loss had no acute effects on cardiac function and geometry (Figure S4). Fibroblast-specific loss of TGF-β signaling significantly increased mortality after MI (Figure 1C–D, p=0.022). Although our study was not powered for sex-specific analysis, both male and female FTbR2KO mice had trends towards increased mortality after MI (Figure 1E–F). Increased mortality in FTbR2KO animals was due to a higher incidence of rupture (Figure 1G–J). Fibroblast-specific TbR2 loss did not affect left ventricular ejection fraction, LVEDV, and LV mass after infarction, but accentuated left ventricular anterior wall thinning (Figure S5). Fibroblast specific loss of TGF-β signaling did not affect the size of the infarct 7–28 days after MI (Figure S6).

Figure 1. Fibroblast-specific TbR2 KO mice (FTbR2KO) have increased rupture-associated mortality after myocardial infarction (MI).

A-B: Fibroblasts harvested from FTbR2KO hearts have complete loss of p-TbR2 compared to TbR2^fl/fl cardiac fibroblasts. C: Timeline of experiments (TMX, tamoxifen). D-J: FTbR2KO mice exhibit increased rupture-associated mortality post-MI, in comparison to controls. Representative H&E-stained images show rupture at the site of the transmural infarct in FTbR2KO mice that died 4 (G), 5 (H), 5 (I), and 7 (J) days after MI. Arrows show the rupture channels throughout the whole thickness of the left ventricle. Scale bar=100μm. B: *p<0.05; TbR2fl/fl: n=4, FTbR2KO: n=4. Statistical analysis (B) was performed using the Mann–Whitney test. D: TbR2fl/fl: n=33, FTbR2KO: n=26. E: TbR2fl/fl: n=16, FTbR2KO: n=11, F: TbR2fl/fl: n=17, FTbR2KO: n=15. Survival analysis was performed using the Kaplan-Meier method. Mortality was compared using the log-rank test (D, E, F).

Fibroblast-specific loss of TGF-β signaling induces lipomatous metaplasia.

Fibroblast-specific abrogation of TGF-β signaling perturbs scar maturation, resulting in extensive infiltration of the infarct with adipose tissue 28 days after MI (Figure 2A–H). Immunofluorescence for the adipocyte marker perilipin-1 (PLIN1) showed that although adipocytes were virtually absent in TbR2^fl/fl scars, they occupied ~30% of the scar in FTbR2KO animals (Figure 2I–M). To examine the effects of fibroblast-specific TbR2 loss on the transcriptional profile of infarct fibroblasts, we harvested PDGFRα+ fibroblasts from FTbR2KO and TbR2^fl/fl infarcts 7 days after MI and we performed RNA-seq (Figure S7). Fibroblast-specific TbR2 loss increased expression of the adipocyte-associated genes Plin1, Lpl, and Pparg in infarct fibroblasts (Figure 2N–P).

Figure 2. Fibroblast-specific TGF-β signaling abrogation induces infiltration of the scar with adipocytes.

A-H: Picrosirius red staining identifies collagen fibers in TbR2^fl/fl mice and in FTbR2KO mice, 7 and 28 days post-MI. B, D, F and H show high magnification fields from A, C, E and G, respectively. 7 days post-MI, replacement of dead cardiomyocytes with granulation tissue is noted in both TbR2^fl/fl and FTbR2 KOs. (A-D). During the maturation phase (28 days post-MI), FTbR2KO (G-H) but not TbR2^fl/fl (E-F) exhibit extensive adipose infiltration of the scar. I-L: Perilipin-1 staining (red) documents adipocyte infiltration in FTbR2KO infarcts, 28 days post-MI (L, red arrow). M: Quantitative analysis shows that in FTbR2KO mice ~30% of the scar is replaced with adipose tissue, 28 days after MI. In contrast, there is no adipocyte infiltration in TbR2^fl/fl infarcts. Infarct fibroblasts harvested from FTbR2KO mice 7 days post-MI had upregulation of adipocyte genes, such as Plin1, Lpl, and Pparg. Scale bar=100μm for A,C,E,G,I,J,K,L. Scalebar=30μm for B,D,F,H. ***p<0.001. M: TbR2^fl/fl 7d: n=7, FTbR2KO 7d: n=5, TbR2^fl/fl 28d: n=8, FTbR2KO 28d: n=6. N-P: TbR2^fl/fl 7d: n=4, FTbR2KO 7d: n=4. Statistical analysis was performed using Kruskal-Wallis test followed by Dunn’s multiple comparison test (M) and the gene expression differences were assessed using DESeq2, which models normalized count data with a negative binomial distribution and applies Benjamini-Hochberg correction to control the false discovery rate (FDR) (N, O, P).

Fibroblast-specific TbR2 loss promotes conversion of infarct fibroblasts to adipocytes.

Next, we performed lineage tracing experiments of COL1A2+ fibroblasts in infarcted FTbR2KO;ROSA26^tdTom and corresponding Col1a2^CreER control mice. 28 days after infarction, the abundant PLIN1+ adipocytes found in FTbR2KO scars were derived from tdTom+ fibroblasts. In contrast, in Cre+ animals, tdTom+ fibroblasts did not undergo adipocyte conversion. Adipocytes located in the epicardial fat served as a negative control and were not derived from COL1A2+ fibroblasts in both FTbR2KO and Col1a2^CreER control animals (Figure 3).

Figure 3. Fibroblast-specific abrogation of TGF-β signaling promotes fibroblast-to-adipocyte conversion in a Smad-independent manner.

A-F: Lineage tracing of COL1A2+ fibroblasts was performed in FTbR2KO mice and in tamoxifen-treated Cre+ controls. A-C: In Cre+ controls, the abundant tdTom+ infarct fibroblasts 28 days post-MI (arrows) do not express the adipocyte marker perilipin-1 (PLIN1). D-F: In FTbR2KO infarcts (28 days post-MI), numerous tdTom+ fibroblasts express PLIN1 (thin arrows), documenting fibroblast-to-adipocyte conversion in infarct fibroblasts lacking TGF-β signaling. The epicardial PLIN1+ adipocytes located outside the infarct do not originate from COL1A2+ fibroblasts (thick arrows, E-F). G-L: Fibroblast-specific Smad4 KO mice (FSmad4KO) do not exhibit fibroblast-to-adipocyte conversion (arrows show the abundant tdTom+ fibroblasts in 28-day infarcts of FSmad4KO and Cre controls). M-S: Quantitative analysis documents the extensive infiltration of FTbR2KO infarcts with fatty tissue (M), PLIN1+ adipocytes (N) and fibroblast-derived PLIN1+ adipocytes (O), 28 days after MI. No PLIN1+ adipocytes (P-Q) were found in FSmad4KO mouse infarcts. Fibroblast-specific TbR2 loss and Smad4 loss had no significant effects on the number of tdTom+ fibroblast-derived cells in 28-day infarcts (R-S). Scale bar =20μm. ****p<0.0001. M-O,R: Cre+ 28d: n=7, FTbR2KO 28d: n=3. P,Q: Cre+ 28d: n=4, FSmad4KO 28d: n=6. S: Cre+ 28d: n=4, FSmad4KO 28d: n=3. Statistical analysis was performed using the Mann-Whitney test (M-Q) and the unpaired 2-tailed Student’s t test (R,S).

Fibroblast-specific TbR2 loss induces fibroblast to adipocyte transdifferentiation in reperfused infarction.

In a model of reperfused MI (2h ischemia/28 days of reperfusion, Figure 4), fibroblast-specific disruption of TGF-β signaling had no significant effects on mortality, cardiac geometry and function (Figure S8). However, FTbR2KO mice exhibited significant lipomatous metaplasia after 28 days of reperfusion with PLIN1+ adipocytes occupying ~40% of the scar. Lineage tracing showed that all PLIN1+ adipocytes in FTbR2KO reperfused infarcts were derived from tdTom+ fibroblasts (Figure 4). Thus, in both reperfused and non-reperfused infarcts, fibroblast-specific disruption of TGF-β signaling induces lipomatous metaplasia, caused by fibroblast-to-adipocyte conversion.

Figure 4: Fibroblast-specific TbR2 loss triggers fibroblast-to-adipocyte conversion promoting lipomatous metaplasia in reperfused mouse infarcts.

A: Timeline of the reperfused infarction experiments (2h coronary occlusion/28 days reperfusion). (TMX, tamoxifen). B-E: Perilipin-1 (PLIN1) immunofluorescence followed by quantitative analysis showed that fibroblast-specific TbR2 loss was associated with lipomatous metaplasia in the scar (C, arrows) occupying ~40% of the reperfused infarct at the 28-day timepoint. In contrast, no significant adipocyte infiltration was noted in TbR2^fl/fl infarcts (B, arrows). F-H: Lineage tracing of COL1A2+ fibroblasts in reperfused infarcts (labeled by tdTom, red) showed that virtually all perilipin-1+ adipocytes in FTbR2KO infarcts were derived from tdTom+ fibroblasts (arrows). Scale bar=100μm. **p<0.01, n=6/group. Statistical analysis was performed using the Mann–Whitney test (D, E).

Fibroblast-specific Smad signaling disruption does not recapitulate the effects of TbR2 loss on fibroblast-to-adipocyte conversion.

In contrast to the effects of fibroblast-specific TbR2 loss, fibroblast-specific Smad4 disruption (Figure S9) did not result in conversion of infarct fibroblasts to adipocytes (Figure 3G–L). Thus, TGF-β-mediated activation of non-Smad signaling cascades in cardiac fibroblasts is sufficient to prevent their conversion to adipocytes.

TGF-β blockade induces adipogenesis-associated gene expression in isolated cardiac fibroblasts.

To examine whether TGF-β inhibition promotes fibroblast-to-adipocyte conversion in vitro, we performed pharmacologic inhibition experiments in isolated cardiac fibroblasts using LY2109761, a selective TbR1/TbR2 inhibitor, and assessed expression of adipogenesis-associated genes (Figures S10–S13). TGF-β signaling blockade for 24h or 7days increased expression of adipocyte-associated genes and adipogenic mediators by cardiac fibroblasts. Lpl expression increased in fibroblasts cultured for 7 days with the TbR1/TbR2 inhibitor (Figure 5A). TbR blockade also stimulated expression of the adipogenic transcription factors Klf4 and Tcf7l2 at both 24h and 7d timepoints (Figure 5B–C). Moreover, TbR inhibition increased levels of several other adipogenesis-associated mediators, including Mapk14, Ddit3, and Bmp4 (Figure 5D–F) and attenuated expression of the adipogenesis repressors Hes1 and Ncor2 (Figure 5G–H).

Figure 5. TGF-β signaling inhibition in cardiac fibroblasts stimulates expression of adipogenesis-associated genes.

A-H: Mouse cardiac fibroblasts were cultured for 24h or 7 days in the presence or absence of the selective dual TbR1/TbR2 inhibitor LY2109761. A PCR array was used to study effects of TGF-β signaling inhibition on expression of adipogenesis-associated genes. TGF-β signaling blockade resulted in increased expression of the adipocyte-associated gene Lpl (A), and of the adipogenic transcription factors Klf4 (B) and Tcf7l2 (C). Moreover, TGF-β inhibition increased levels of Mapk14 (D), Ddit3 (a member of the C/EBP family of adipogenic transcription factors, E), and of the adipogenic growth factor Bmp4 (F). Blockade of TGF-β signaling also reduced expression of the adipogenesis suppressors Hes1 (G) and Ncor2 (H). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. Control 24h: n=4; TbR2 inh 24h: n=4; control 7d: n=4; TbR2 inh 7d: n=4. Statistical analysis was performed using 1-way ANOVA followed by Sidak’s multiple comparison test.

TbR2 loss promotes adipogenic conversion in cultured infarct fibroblasts.

To examine whether genetic TGF-β signaling disruption promotes adipogenic conversion of infarct fibroblasts in vitro, we isolated and cultured fibroblasts from TbR2 fl/fl and FTbR2KO infarcts 5 days after MI. Infarct fibroblasts were cultured with adipogenic or standard medium, in the presence or absence of TGF-β1. After 7 days of culture, infarct fibroblasts from both FTbR2KO and control mice cultured in standard medium showed no evidence of adipocyte conversion. In contrast, culture with adipogenic medium resulted in emergence of a population of PLIN1+ adipocytes, only in FTbR2KO cells but not in TbR2fl/fl cells, independently of the presence of TGF-β. (Figure 6A–I).

Figure 6. Disruption of TGF-β signaling promotes infarct conversion of cultured infarct fibroblasts to adipocytes.

A-H: Fibroblasts were harvested from 5-day infarcts and were cultured for 7 days. α-smooth muscle actin (α-SMA) fluorescence labels myofibroblasts, whereas Perilipin-1 fluorescence documents adipocyte conversion. TbR2KO, but not TbR2^fl/fl infarct fibroblasts exhibited adipocyte conversion in the presence of adipogenic medium (G-I). Perilipin-1+ fibroblast-derived adipocytes exhibited loss of α-SMA expression. Infarct fibroblasts with functional TbR2 cultured in adipogenic medium expressed less α-SMA (C,D) compared to controls in standard medium (A,B); however, this was not accompanied by adipocyte conversion (no perilipin-1 labeling, C,D,I). TbR2^fl/fl and FTbR2KO fibroblasts cultured in standard medium with or without TGF-β stimulation had no evidence of adipocyte conversion (A-B, E-F, I). J-O: A PCR array showed that only FTbR2KO infarct fibroblasts cultured in the presence of adipogenic medium exhibited upregulation of adipocyte-related genes, including Adig (adipogenin) (J), Adipoq (adiponectin) (K), Lep (leptin) (L), Lpl (lipoprotein lipase) (M), and Fabp4 (fatty acid binding protein 4) (N) and increased expression of the adipogenic nuclear receptor Pparg (peroxisome proliferator-activated receptor gamma) (O). Scale bar=10μm. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. I-O: TbR2fl/fl: n=3, FTbR2KO: n=3. Statistical analysis was performed using 1-way ANOVA followed by Sidak’s multiple comparison test (I-O). C=control, std=standard medium, adip=adipogenic medium.

Gene expression analysis (Figure S14) showed that, when cultured with adipogenic medium, FTbR2KO, but not control infarct fibroblasts, exhibited induction of adipocyte genes, including Adig (adipogenin), Adipoq (adiponectin), Lep (leptin), Lpl (lipoprotein lipase), and Fabp4 (fatty acid binding protein 4) (Figure 6J–N). Moreover, disruption of TGF-β signaling in fibroblasts significantly increased expression of the master regulator of adipogenesis and fat storage Pparg (Figure 6O). Thus, both immunocytochemical and gene expression studies documented that TbR2 disruption promotes fibroblast-to-adipocyte conversion only in the presence of adipogenic mediators.

Fibroblast-specific TbR2 loss reduces scar collagen deposition and decreases collagen fiber thickness.

Quantitative analysis of picrosirius red-stained sections showed that fibroblast-specific loss of TGF-β signaling did not affect collagen content 7 days after infarction but diminished collagen accumulation in the infarct zone at the 28-day timepoint (Figure S15). Polarized light microscopy showed that fibroblast-specific loss of TGF-β signaling reduced accumulation of the thicker orange and red collagen fibers in the infarct zone (Figure S15T), without affecting deposition of the thinner green fibers. In contrast, no significant effects of fibroblast-specific TbR2 loss were noted on collagen fiber composition in the border zones and in the remote remodeling myocardium (Figure S16). To assess collagen fiber thickness, we used the MATLAB software framework package CT-FIRE¹⁹. Collagen thickness was significantly reduced in FTbR2KO infarcts, when compared with corresponding floxed controls, 28 days after MI (Figure S17).

Fibroblast-specific loss of TGF-β signaling does not affect infarct myofibroblast infiltration but attenuates fibroblast accumulation in the mature scar.

To examine whether perturbed repair in FTbR2KO infarcts is associated with perturbations in the phenotypic transitions of infarct fibroblasts, we performed immunofluorescent staining for PDGFRα (to label all fibroblasts), α-SMA (to identify myofibroblasts) and periostin (as a matricellular marker of fibroblast activation). Fibroblast-specific loss of TGF-β signaling did not affect myofibroblast density at the peak of the proliferative phase, but significantly reduced the PDGFRα+ area and the periostin+ area in the mature scar, 28 days after MI (Figure S18). Thus, fibroblast-specific loss of TGF-β signaling induces fibroblast-to-adipocyte conversion (Figures 2–3), reducing the number of scar-preserving matrifibrocytes in the infarct (Figure S18).

Fibroblast-specific loss of TGF-β signaling attenuates collagen 1 synthesis and increases expression of matrix-degrading genes in infarct fibroblasts.

Considering that fibroblast-specific loss of TbR2 did not affect myofibroblast conversion during the proliferative phase of cardiac repair, we postulated that the reparative defects and increased incidence of rupture in FTbR2KO animals may be due to changes in expression of genes involved in matrix deposition and remodeling. RNA-seq showed that FTbR2KO infarct fibroblasts had lower expression of the collagen genes Col1a1, Col1a2, Col16a1, Col9a1 and Col9a2 (Figure S19A–E), and higher levels of several proteases involved in matrix degradation and remodeling, including Mmp3, Adam17, Adamts1, Adamts4, Adamts5, and Adamts12 (Figure S19F–K).

Effects of fibroblast-specific TbR2 loss on infarct macrophage and microvascular density.

Mac2 immunofluorescence showed no statistically significant effects of fibroblast-specific TbR2 loss on macrophage density in infarcted and non-infarcted segments at 3– 7- and 28-day timepoints (Figure S20). CD31 immunofluorescence was used to quantify microvascular profiles in infarcted and remote non-infarcted myocardial segments. Although no significant differences were noted between FTbR2KO mice and corresponding floxed controls at the 7-day timepoint, fibroblast-specific loss of TbR2 was associated with higher microvascular density in the infarct zone 28 days after MI (Figure S21).

Fibroblasts in late-stage human infarcts express adipocyte genes and may undergo adipocyte conversion.

To examine whether fibroblasts in human MI may undergo adipocyte conversion, we analyzed scRNA-seq data from MI patients (Table S1)¹⁸. The data include samples from acute infarcts (IZ, infarct zone), as well as later stages after infarction (FZ, fibrotic zone). Fibroblasts from fibrotic zones had significantly higher expression of several adipocyte-associated genes, including PLIN1, PPARG, ADIPOQ and ZBTB16 (zinc finger and BTB domain 16, a transcription factor involved in adipogenesis²⁰), than fibroblasts from control hearts (Figure 7). Moreover, genes expressed higher in FZ fibroblasts than in controls were enriched for genes involved in fat cell differentiation (padj=0.04, Figure 7B), by Gene Set Enrichment Analysis (GSEA). We also performed GSEA using a customed gene set, containing the top 100 adipocyte markers computed from the human scRNA-seq data (Figures S22–23). FZ fibroblasts were enriched for this adipocyte marker set (vs controls), supporting the notion that some fibroblasts may acquire adipocyte-like characteristics in human myocardial scars (Figure 7B, p=5.37E-10). Next, we calculated module scores for individual fibroblast cells using the top 100 adipocyte markers, referred to as the AdipoMarker Module scores (Figure S24). At the individual cell level, a positive correlation between the AdipoMarker Module scores and the times after infarction was found for both FZ and IZ fibroblasts (Figure 7C), indicating a gradually increased expression of adipocyte genes in infarct fibroblasts. Moreover, at the individual patient level, the mean AdipoMarker module score was significantly higher in FZ in comparison to controls (p=0.0285) (Figure 7D), reflecting acquisition of an adipocyte transcriptional program in fibroblasts infiltrating late-stage infarcts. Thus, in MI patients, a subpopulation of fibroblasts exhibits late upregulation of adipocyte-associated genes.

Figure 7: Expression of adipocyte genes in fibroblasts in human myocardial infarction (MI).

(A) Violin plots show the normalized expression of adipocyte-related genes in fibroblasts at different zones in MI patients. CTRL, control; RZ, remote zone; BZ, border zone; IZ, infarct zone; FZ, fibrotic zone. Fibroblasts in FZ from late-stage infarcts had significantly increased expression of the adipocyte-associated genes PLIN1, PPARG, ADIPOQ and ZBTB16. Bonferroni-corrected p-values (p.adj) (< 0.05; Wilcoxon rank-sum tests between controls and different MI zones) are displayed above comparisons. (B) Gene set enrichment analysis (GSEA) comparing fibroblasts in FZ (or IZ) vs controls for two genet sets: “Regulation of Fat Cell Differentiation” genes and the top 100 adipocyte marker genes (Supplemental Figure S22). (C) Boxplots show the single cell module scores computed for the top 100 adipocyte markers in fibroblasts from different zones in individual patients. In fibroblasts from infarct zone and the fibrotic zone a significant positive correlation was noted between the module score and the time after MI (Pearson’s correlation statistics computed with cells from all patients combined). (D) Boxplots show the mean module score in individual patient samples for each zone. The mean module score was significantly higher in fibrotic zone samples (FZ) in comparison to controls (p=0.0285) reflecting increased expression of adipocyte-associated genes in a subset of fibroblasts infiltrating late-stage infarcts. Statistical analysis was performed using 1-way ANOVA followed by Dunnett’s multiple comparison test (n=3–7).

DISCUSSION:

Adipose tissue infiltration commonly accompanies fibrotic remodeling in both ischemic¹⁷ and non-ischemic²¹ cardiomyopathy patients. A histological study found fatty infiltration in myocardial scar tissue in 68% of end-stage ischemic cardiomyopathy patients¹⁷. Despite its potential functional and electrophysiologic implications, the mechanisms driving lipomatous metaplasia remain unknown. Our study provides the first robust evidence of a cellular and molecular mechanism mediating adipose infiltration in the infarcted mouse heart (Figure 8). Using cell-specific loss-of-function models and lineage tracing, we show that fibroblast-specific TGF-β signaling abrogation induces fibroblast-to-adipocyte conversion in the healing infarct, replacing ~30–40% of the scar with adipose tissue. In isolated cardiac fibroblasts, TGF-β signaling inhibition stimulates adipogenesis-associated genes (Figure 5). Thus, maintenance of fibroblast identity in the infarcted myocardium requires activation of TGF-β signaling.

Figure 8: The cellular mechanism driving lipomatous metaplasia in the infarcted heart.

Loss of TGF-β signaling in fibroblasts upon conditional deletion of the type 2 TGF-β receptor (TbR2) promotes fibroblast-to-adipocyte conversion during the maturation phase of repair, leading to extensive adipose infiltration of the scar. The effects of TbR2 loss are not recapitulated by Smad4 deletion, suggesting that disruption of Smad-independent pathways may be important in fibroblast to adipocyte transdifferentiation. Lipomatous metaplasia may contribute to post-infarction cardiac dysfunction and arrhythmogenesis.

Fibroblasts and myofibroblasts as cellular targets of TGF-β.

TGF-β superfamily members signal by binding to TbR1/TbR2 receptor complexes, subsequently activating downstream Smad-dependent or Smad-independent cascades²². In the healing infarct, TGF-β signaling exerts cell-specific protective and maladaptive actions²³,²⁴. TGF-βs have profound activating effects on fibroblast phenotype and function acting through both Smad-mediated² and Smad-independent pathways²⁵. We have previously demonstrated that myofibroblast-specific Smad3 loss perturbs cardiac repair, increasing the incidence of ventricular rupture². In our current study, fibroblast-specific TGF-β signaling abrogation similarly caused rupture post-MI (Figure 1). Thus, during the proliferative phase, TGF-β/Smad3 mediates formation of a scar by aligned, matrix-producing reparative fibroblasts²,²⁶. As the scar matures, infarct myofibroblasts transition to matrifibrocytes, specialized fibroblast-like cells expressing unique matrix and tendon genes that maintain scar integrity. The signals regulating matrifibrocytes phenotype and function remain enigmatic.

TGF-β preserves fibroblast cell specification inhibiting adipocyte conversion.

Our findings show that TGF-β signaling stabilizes scar-preserving infarct fibroblasts, preventing their conversion to adipocytes (Figures 3–6). Fibroblast-specific loss of Smad signaling did not recapitulate the effects of complete TGF-β signaling disruption, suggesting that fibroblast-to-adipocyte conversion in healing infarcts may require attenuation of TGF-β-driven non-Smad signaling pathways. Among these, Rho GTPases have been implicated in suppression of adipogenesis²⁷. In vitro, ROCK inhibition accentuates adipogenic conversion of 3T3-L1 fibroblast-like cells²⁸, and porcine embryonic fibroblasts adopt an adipocyte phenotype when treated with both ALK5/TbR1 and Rho/ROCK inhibitors²⁹. In contrast, other TGF-β-activated non-Smad pathways (p38, Erk, TAK1, etc) promote adipogenesis³⁰,³¹,³² and are unlikely to contribute to the phenotype observed in TbR2KO fibroblasts. Thus, fibroblast-to-adipocyte conversion following TGF-β signaling loss may predominantly involve inhibition of Smad-independent Rho/ROCK signaling.

Extensive lipomatous metaplasia in FTbR2KO infarcts resulted from complete fibroblast-specific TGF-β signaling loss (Figure 1A–B). Whether partial loss of TGF-β signaling in fibroblasts has similar effects is not known. TGF-β’s role in fibroblast specification may be dose-dependent: low levels may suffice to maintain fibroblast identity, while full loss -alongside adipogenic cues- may be required for adipocyte conversion.

Does TGF-β signaling loss drive adipocyte conversion in a specific fibroblast subpopulation?

Fibroblasts exhibit remarkable heterogeneity³³,³⁴,³⁵. In the mouse heart, 2 subpopulations distinguished by Ly6a/Sca-1 (Stem cell antigen 1) expression have been identified³⁶. Sca-1+ fibroblasts show high plasticity, adopting inflammatory, immunomodulatory, angiogenic or fibrogenic phenotypes in response to microenvironmental cues³⁷,³⁸,³⁹,⁴⁰. They may also possess adipogenic potential and have been proposed as candidate effector cells in the fibrofatty infiltration in arrhythmogenic right ventricular cardiomyopathy (ARVC)⁴¹. Moreover, a population of PDGFRα+/DDR2− fibroblast-like cells expressing the adipogenic transcription factor C/EBP were implicated in fibrofatty infiltration in a genetic ARVC model⁴². Whether fibroblast-to-adipocyte conversion in FTbR2 KO infarcts reflects transdifferentiation of such subpopulations remains unclear.

The cellular origin of lipomatous metaplasia in MI patients.

Although lipomatous metaplasia is documented in patients with old myocardial scars¹⁷, the origin of the infiltrating adipocytes remains enigmatic. Analysis of published scRNA-seq data suggests that fibroblast subpopulations in fibrotic infarct zones express adipocyte genes (Figure 7). One possibility is that lipomatous metaplasia in some patients may reflect impaired TGF-β signaling due to loss-of-function mutations in the TGF-β/TbR system. The role of TGFB1, TGFBR1 and TGFBR2 mutations is well documented in aortic diseases, such as the Loeys-Dietz syndrome and Marfan-related conditions⁴³. While TGFB1 polymorphisms have been linked to MI risk⁴⁴, their role in post-infarction remodeling, fibrosis, and lipomatous metaplasia remains unexplored.

The consequences of lipomatous metaplasia in the clinical setting.

Although lipomatous metaplasia was not associated with increased dysfunction in infarcted FTbR2KO mice, this does not rule out later adverse effects. Follow-up was limited to 28 days post-MI and the higher rupture-associated mortality in FTbR2KO mice, may have introduced a bias, selectively eliminating the most vulnerable animals. Substantial evidence suggests that adipocyte infiltration of the infarct may have clinical implications. In an ovine MI model, intramyocardial adiposity altered electrophysiologic properties linked to arrhythmias¹⁵. Cardiac magnetic resonance imaging (CMRI)-defined conduction channels at deceleration zones often overlapped with adipose tissue deposits, implicating lipomatous metaplasia in post-infarction arrhythmogenesis⁴⁵. In MI patients, lipomatous metaplasia predicted adverse outcome⁴⁶. A combined clinical/computational study suggested a role for adipose infiltration in promoting re-entry and post-infarction arrhythmogenesis¹².

Whether lipomatous metaplasia in patients with ischemic cardiomyopathy is reversible is unknown. Due to the limited plasticity of mature adipocytes, reversing lipomatous metaplasia may prove challenging. However, early targeting of fibroblast-to-adipocyte conversion in susceptible patients with reparative defects due to hypo-active TGF-β cascades may be protective by preventing fatty infiltration while preserving scar structure. Identifying such patients will likely require a combination of clinical criteria, imaging, and biomarkers.

Conclusions

We identify a novel mechanism driving adipose infiltration in MI. In this dynamic environment, fibroblasts rely on continuous TGF-β stimulation to maintain their identity. Disruption of this cascade promotes fibroblast-to-adipocyte conversion, leading to lipomatous metaplasia. Extensive adipose deposition in the scar may contribute to post-MI dysfunction and arrhythmogenesis.

CLINICAL PERSPECTIVE:

What is new?

• In mouse models, fibroblast-specific loss of Transforming Growth Factor (TGF)-β signaling promotes extensive adipose tissue infiltration within the mature infarct, promotes conversion of fibroblasts into adipocytes, and induces the expression of adipogenesis-related genes in primary cardiac fibroblasts.

• Single cell RNA-seq analysis in patients with myocardial infarction reveals that a subset of fibroblasts within the fibrotic zone expresses adipocyte-associated genes.

• The effects of fibroblast-specific loss of TGF-β signaling on fibroblast to adipocyte transdifferentiation involve disruption of Smad-independent pathways.

What are the clinical implications?

• We demonstrate for the first time a molecular mechanism responsible for scar-associated lipomatous metaplasia in myocardial infarction.

• Disrupted TGF-β signaling in infarct fibroblasts may be involved in the pathogenesis of post-infarction lipomatous metaplasia by inducing conversion of fibroblasts to adipocytes.

• Replacement of the myocardial scar with adipose tissue may contribute to the pathogenesis of heart failure and arrhythmogenesis in ischemic cardiomyopathy patients.

NON-STANDARD ABBREVIATIONS AND ACRONYMS:

ALK5

Activin receptor-like kinase 5

BZ

Border Zone

C/EBP

CCAAT/enhancer-binding proteins

COL1A2

Collagen, type I, alpha 2

CMRI

Cardiac Magnetic Resonance Imaging

CreER

Cre recombinase–Estrogen Receptor fusion

CTRL

Control

Cebpd

CCAAT/enhancer-binding protein delta

DAPI

4’,6-diamidino-2-phenylindole

dNTP

Deoxynucleotide triphosphate

DMEM/F12

Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12

DPBS

Dulbecco’s Phosphate Buffered Saline

DDR2

Discoidin domain receptor 2

EDTA

Ethylenediaminetetraacetic acid

EF

Ejection Fraction

Erk

Extracellular signal–regulated kinase

FACS

Fluorescence-activated cell sorting

FBS

Fetal Bovine Serum

FITC

Fluorescein isothiocyanate

FDR

False Discovery Rate

FPKM

Fragments Per Kilobase of transcript per Million mapped reads

FTbR2KO

Fibroblast-specific TβR2 Knockout

GSEA

Gene Set Enrichment Analysis

GEO

Gene Expression Omnibus

IZ

Ischemic Zone

IVSTd

Interventricular Septal Thickness in Diastole

KHB

Krebs–Henseleit Buffer

LV

Left Ventricle

LVEDV

Left Ventricular End-Diastolic Volume

LVESV

Left Ventricular End-Systolic Volume

LVAWTd

LV Anterior Wall Thickness in Diastole

LVPWTd

LV Posterior Wall Thickness in Diastole

MAPK

Mitogen-activated protein kinase

MI

Myocardial Infarction

PCR

Polymerase Chain Reaction

PLIN1

Perilipin-1

Postn

Periostin

p-TbR2

Phosphorylated Transforming Growth Factor (TGF)-β Receptor 2

qPCR

Quantitative PCR

RNA-seq

RNA sequencing

ROCK

Rho-associated coiled-coil kinase

RZ

Remote Zone

Sca-1

Stem cell antigen-1

std

Standard

scRNA-seq

Single-cell RNA sequencing

TAK1

TGF-β-activated kinase 1

tdTom

Tandem dimer Tomato

TGF-β

Transforming Growth Factor-β

TbR1/2

TGF-β receptor types 1 and 2

VSMCs

Vascular Smooth Muscle Cells

ZBTB16

Zinc finger and BTB domain–containing protein 16