Hemoglobin inhibits fibroblast-to-cardiomyocyte reprogramming via TLR2/TLR4-dependent chromatin compaction

Abstract

Fibroblast-to-cardiomyocyte reprogramming is influenced by signals present in the injured heart. However, the nature of those signals is unclear. The efficacy of fibroblast-to-cardiomyocyte reprogramming was found to be negatively impacted by extracts derived from the freeze-thawing of normal cardiac tissue. Mass spectrometry identified the adult minor beta chain of hemoglobin (Hbb-b2) as the most abundant protein in the extract, prompting investigation of hemoglobin (Hb) as a candidate inhibitory factor. Indeed, Hb was found to suppress fibroblast-to-cardiomyocyte reprogramming by inhibiting the expression of various cardiomyocyte-specific markers and cardiomyocyte formation. Hb is known to interact with TLR2 and TLR4. While inhibition of either receptor was unable to reverse the effects of Hb, inhibition of both TLR2 and TLR4 reversed the suppressive effects of Hb. To clarify the mechanism, assay for transposase-accessible chromatin using sequencing (ATAC-seq) was performed. ATAC-seq revealed that Hb induced widespread transcription factor network remodeling, whereby stress-associated factor networks were induced and latent CEBP/PAR-bZIP networks were suppressed. These effects were localized to a region 500–750 bp upstream of the transcription start site in cardiomyocyte-specific genes. These findings identify Hb as a potential negative regulator of fibroblast-to-cardiomyocyte reprogramming, with implications for cardiac regeneration strategies.

Graphical abstract

Hodgkinson and colleagues demonstrate that hemoglobin inhibits fibroblast-to-cardiomyocyte reprogramming.

Introduction

Regeneration of the cardiac muscle following injury remains elusive. While research, including our own, has demonstrated that scar fibroblasts can be reprogrammed into cardiac muscle, the critical barrier to progress is low efficacy.^1,2,3,4,5,6

Fibroblast-to-cardiomyocyte reprogramming has been achieved through various methods. We were the first to show fibroblast-to-cardiomyocyte reprogramming through a combination of four miRNAs, collectively known as miR combo. MiR combo comprises four miRs: miR-1, miR-133, miR-208, and miR-499. Differing from transcription factor-based protocols, the same combination of miRNAs is effective in both pig and human cardiac fibroblasts.⁷ Irrespective of the method employed, fibroblast-to-cardiomyocyte reprogramming efficacy is too low for clinical applications. The approach that we, and others, have taken to improve efficacy is to decipher the underlying mechanisms and to apply the knowledge gained. One area that we have focused on is the innate immune response which plays an important role following cardiac injury. Damage-associated molecular patterns (DAMPs) are released by dying cells and activate innate immune receptors, triggering the production of pro-inflammatory cytokines necessary for tissue clearance and healing. DAMPs come in many forms, such as proteins, DNA, and RNA. We were the first to discover that fibroblast-to-cardiomyocyte reprogramming requires RNA-sensing receptor activation.^3,4,8 The information we gained was subsequently used to improve miR combo efficacy by chemically modifying the constituent miRNAs to promote an interaction with the RNA-sensing receptor RIG-I.⁹ As mentioned, RNA molecules are but one component of the DAMPs released by dying cells. However, it is unclear what roles non-RNA DAMPs play with respect to fibroblast-to-cardiomyocyte reprogramming.

The objective of this study was to further clarify how innate immune signaling affects fibroblast-to-cardiomyocyte reprogramming. Heart tissue extracts derived from the freeze-thawing of cardiac tissue negatively impacted fibroblast-to-cardiomyocyte reprogramming. Mass spectrometry identified the adult minor beta chain of hemoglobin (Hbb-b2) as the most abundant protein in the extract, prompting investigation of hemoglobin (Hb) as a candidate inhibitory factor. Subsequent studies showed that Hb suppressed fibroblast-to-cardiomyocyte reprogramming by preventing the expression of sarcomeric components and cardiac ion channels. Gene suppression by Hb was found to be dependent on both TLR2 and TLR4. Assay for transposase-accessible chromatin using sequencing (ATAC-seq) revealed that Hb induced stress-related transcription factor networks and concurrently suppressed networks associated with CEBP/PAR-bZIP factors. These findings identify Hb as a potential negative regulator of fibroblast-to-cardiomyocyte reprogramming, with implications for cardiac regeneration strategies.

Results

The objective of this study was to further our research into the impact of innate immune signaling on fibroblast-to-cardiomyocyte reprogramming. Initially, we focused on the effect of immune cell paracrine signaling following pathogen-associated molecular patterns (PAMPs) versus DAMP stimulation. Macrophages were stimulated with various PAMPs (LPS, PolyI:C, 3p-hpRNA, or Pam3csk4) or DAMPs (heart extract). Once stimulated, conditioned media was collected from these cells and incubated with cardiac fibroblasts undergoing reprogramming. To determine the effects on cardiac reprogramming, the expression of eleven cardiomyocyte-specific genes was evaluated. Notable, but opposing, effects were observed with conditioned media derived from macrophages stimulated with LPS or heart extract (Figure 1A). While conditioned media derived from macrophages stimulated with LPS increased cardiac reprogramming efficacy, conditioned media derived from macrophages stimulated with heart extract was inhibitory (Figure 1A). To determine if these effects were due to differences in cytokines released by macrophages, assays were conducted to determine the expression of the cytokines TNFα, IFNβ1, IL6, and IL1β. Of the five ligands tested, LPS had the greatest effect, significantly increased the expression of four cytokines. In contrast, heart extract had a minimal effect and only increased the expression of IFNβ1 (Figure 1B).

Figure 1.

Cardiac DAMPs inhibit fibroblast-to-cardiomyocyte reprogramming

(A) Raw264.7 macrophages were exposed to the indicated ligands for 24 h. After 24 h, the cells were washed and incubated for a further 24 h in standard growth media. The media was collected, and debris was removed. Once cleaned, macrophage media was added to cardiac fibroblasts immediately after their transfection with a control non-targeting miRNA or miR combo. Complexes were removed one day later. Fourteen days after transfection, expression of the indicated cardiomyocyte-specific genes were determined by qPCR. Expression values were normalized to the house-keeping gene Gapdh. The heatmap shows the average of 3 independent experiments expressed as a Z score. Significances were determined by t test. Significant genes (control macrophage conditioned media versus heart extract stimulate macrophage conditioned media) are shown in bold. Raw data for two representative genes are shown in the left image. (B) Macrophages were exposed to the indicated ligands for 6 h. Expression of the indicated cytokines was determined by qPCR, normalized to Gapdh, and shown on the heatmap as a Z score. N = 3. (C) Cardiac fibroblasts were transfected with either miR combo or a non-targeting control miR. The day after transfection, LPS, heart extract, or vehicle were added to the cells for 24 h. Fourteen days after miR transfection, qPCR was used to measure the expression of the indicated cardiomyocyte genes. Expression values were normalized to Gapdh. The heatmap shows the average of 7 independent experiments as a Z score. One-sample t tests were used to compare groups to the control group. Significant genes (miR combo versus miR combo and heart extract) are shown in bold.

Noting that the macrophage conditioned media influenced cardiac reprogramming only when the cells were stimulated with LPS or heart extract, we asked ourselves if the same ligands would affect reprogramming if exposed directly to fibroblasts. To test this, cardiac fibroblasts were induced to reprogram and incubated with either LPS or heart extract. Fibroblast-to-cardiomyocyte reprogramming efficacy was evaluated by measuring the expression of ten cardiomyocyte-specific genes. Mirroring the conditioned media results, LPS significantly improved cardiac reprogramming efficacy, while reprogramming efficacy was significantly impaired by heart extract. Indeed, expressions of nine of the ten cardiomyocyte-specific genes were decreased to levels indistinguishable from control levels (Figure 1C).

DAMPs are likely to be more prevalent in the post-MI heart than PAMPs. Consequently, we chose to focus on the inhibition caused by heart extract and sought to identify the functional component. To that end, mass spectrometry was employed. Analysis of the mass spectrometry data focused on protein constituents. By molar mass, the largest constituent of heart extract was Hbb-b2 (Figure 2A).

Figure 2.

Hemoglobin inhibits fibroblast-to-cardiomyocyte reprogramming

(A) Macrophages were incubated with either vehicle or heart extract for 24 h. After 24 h, the cells were washed and incubated for a further 24 h in standard growth media. The media was collected and debris removed. The media from three individual experiments were combined and analyzed by mass spectrometry. The relative protein abundance is shown on the right-hand side y axis. Larger proteins produce more peptides, and hence more signal, even if molar abundance is identical. Therefore, to remove size bias, the relative abundance was divided by protein molecular weight (left hand y axis). (B) ELISA measurements of Hb concentration in control media and in five different heart extracts (He). N = 5. (C) Cardiac fibroblasts were transfected with either miR combo or a non-targeting control miR. The day after transfection, the indicated doses of Hb were added to the cells for 24 h. Fourteen days after miR transfection, cells were analyzed for expression of the cardiomyocyte marker Actn2 by qPCR. N = 3. One-sample t tests were used to compare groups to the normalized control group (∗p < 0.05; ns, not significant). (D–G) Cardiac fibroblasts were transfected with either miR combo or a non-targeting control miR. The day after transfection, hemoglobin (Hb, 5 mg/mL) was added to the cells for 24 h. Fourteen days after miR transfection, cells were analyzed for cardiomyocyte gene expression (D and E), cardiomyocyte numbers (F), and functional maturation (G). With respect to gene expression, qPCR was used to evaluate the expression of both cardiomyocyte sarcomere genes (D) and (cardiomyocyte ion channels (E). Expression values were normalized to the housekeeping gene Gapdh. N = 15. One-sample t tests were used to compare groups to the control group (∗∗∗p < 0.001, ∗p < 0.05; ns, not significant). A t test was used to determine the significance between the two miR combo (miRC) groups (###p < 0.001, ##p < 0.01; ns, not significant). (F) Cardiomyocyte numbers were determined by immunostaining for the cardiomyocyte marker Actn2 (red). Nuclei were visualized by DAPI (blue). Representative images are shown (scale bar, 50 microns). N = 4. (G) Functional maturation was determined by counting the number of spontaneous beating cells. N = 4. For (F) and (G), significances were determined by one-way ANOVA with Tukey post-hoc testing. Comparisons were made to the control group (∗∗∗p < 0.001, ∗∗p < 0.01; ns, not significant) as well as to the miR combo + Hb group (###p < 0.001, ##p < 0.01).

Considering that Hbb-b2 was the most abundant protein in heart extracts, we chose to focus on Hb. ELISA was employed to determine the concentration of Hb in our mouse heart extracts. As shown in Figure 2B, the concentration ranged from 1 to 3 mg/mL.

Fibroblasts were induced to reprogram and incubated with increasing concentrations of Hb. Fibroblast-to-cardiomyocyte reprogramming was inhibited by Hb doses of 1 and 5 mg/mL (Figure 2C). The 5 mg/mL dose was used in all subsequent experiments.

To further evaluate the effects on fibroblast-to-cardiomyocyte reprogramming, we measured the effect of Hb on six cardiomyocyte-specific genes: Actn2, Myh6, Tnni3, Cacna1c, Ryr2 and Scn5a. Actn2, Myh6, and Tnni3 play important roles in sarcomere formation. As expected, fibroblast-to-cardiomyocyte reprogramming induced the expression of all three genes (Figure 2D). The effect of reprogramming on sarcomere gene expression was completely ablated by Hb (Figure 2D). Expression of the cardiomyocyte-specific ion channels Cacna1c, Ryr2, and Scn5a followed the same trend. Indeed, Hb was found to suppress expression of all three ion channels in reprogramming cells (Figure 2E). Suppression was significant for Cacna1c and Ryr2. The trend was the same for Scn5a, but the variability between experiments precluded the effect being significant.

To further validate the inhibitory effect of Hb, cardiomyocyte formation was quantified by counting Actn2⁺ cells. Reprogramming significantly increased the number of Actn2⁺ cells (Figure 2F), whereas Hb almost completely abolished the effect of the reprogramming agent (Figure 2F). Functional maturation was assessed by quantifying spontaneously beating cells. Consistent with the Actn2⁺ data, reprogramming markedly increased the number of beating cells, while addition of Hb reduced spontaneous beating to control levels (Figure 2G).

We examined if these effects were restricted to cardiomyocyte-specific genes. Consequently, we measured the levels of the fibroblast-specific genes Acta2, Col1a1, and Postn. Hb significantly inhibited Acta2 expression but had no effect on Col1a1 and Postn expression (Figure 3A).

Figure 3.

Blood mimics the inhibitory effects of Hb

(A) Cardiac fibroblasts were incubated with Hb (5 mg/mL) for 24 h. Expression of the indicated fibroblast-specific genes was determined by qPCR and expression values normalized to the housekeeping gene Gapdh. A t test was used to determine the significance between the two groups (∗∗∗p < 0.001; ns, not significant). (B) Cardiac fibroblasts were transfected with either miR combo or a non-targeting control miR. The day after transfection, blood (10 μL per ml of media) was added to the cells for 24 h. Fourteen days after miR transfection, cells were analyzed for expression of the indicated cardiomyocyte-specific genes by qPCR. N = 4. Expression values were normalized to the housekeeping gene Gapdh. For the heatmap, expression values are shown as Z scores for the four individual experiments. A t test was used to compare the two miR combo groups ∗∗∗p < 0.001).

If Hb was indeed inhibiting gene expression, one would expect the same finding if blood was used instead. Indeed, incubating reprogramming fibroblasts with blood strongly inhibited expression of the cardiomyocyte-specific markers Actn2, Myh6, and Tnni3 (Figure 3B).

Having established the role of Hb in suppressing fibroblast-to-cardiomyocyte reprogramming, further studies were conducted to determine the mechanism. Hb has been described as both a TLR2 and a TLR4 ligand.^10,11 Innate immune signaling downstream TLR2/TLR4 proceeds via MAPK or NF-κB pathways.¹² Hb had no effect on MAPK phosphorylation (Figure 4A). In contrast, Hb induced strong NF-κB phosphorylation (Figure 4A). Application of a TLR4 pharmacological inhibitor completely abrogated NF-κB phosphorylation by Hb (Figure 4B).

Figure 4.

Hemoglobin activates TLR signaling in cardiac fibroblasts

(A) Cardiac fibroblasts were incubated with Hb (5 mg/mL) for the indicated times. Protein extracts were analyzed by immunoblotting for p-NFκB and p-MAPK. Gapdh was used as a loading control. Representative immunoblots are shown on the left-hand side. Quantification was performed by normalizing p-NFκB and p-MAPK band densities with those of the loading control. N = 4. One-sample t tests were used to compare groups to the control group (∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05; ns, not significant). (B) Cardiac fibroblasts were incubated either vehicle or with varying concentrations of a TLR4 pharmacological inhibitor for 3 h. Hb (5 mg/mL) was then added. After 1 h, protein extracts were isolated and subsequently analyzed by immunoblotting for p-NFκB. Gapdh was used as a loading control. Representative immunoblots are shown on the left-hand side. Quantification was performed by normalizing p-NFκB band densities with those of the loading control. N = 4. One-sample t tests were used to compare groups to the control group (∗p < 0.05; ns, not significant). (C) Cardiac fibroblasts were incubated with either LPS or Hb for 24 h, after which expression of the indicated pro-inflammatory cytokines was determined by qPCR. Expression values are shown relative to the housekeeping gene Gapdh. N = 6. ANOVA with Tukey post-hoc tests were used to determine significance (∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05; ns, not significant).

To further understand TLR2/TLR4 pathways downstream of HB, expression of the pro-inflammatory cytokines IL6, TNFα, and IL1β was determined following stimulation of cardiac fibroblasts with either Hb or LPS. LPS is a ligand for both TLR2 and TLR4 and served as the positive control. As expected, LPS induced the expression of all three pro-inflammatory cytokines (Figure 4C). While Hb was able to significantly induce TNFα and IL1β, the level of induction was lower than that observed for LPS (Figure 4C).

Further experiments were conducted to determine if TLR2/TLR4 inhibitors could ameliorate Hb suppression of fibroblast-to-cardiomyocyte reprogramming. Fibroblast-to-cardiomyocyte reprogramming efficacy was again determined by measuring the expression of sarcomere components and cardiomyocyte-specific ion channels. As expected, fibroblast-to-cardiomyocyte-induced expression of sarcomere components (Actn2, Myh6, and Tnni3) and cardiomyocyte-specific ion channels (Cacna1c, Ryr2, and Scn5a) was effectively suppressed by Hb (Figure 5). Hb-mediated suppression of Ryr2 and Scn5a was alleviated by TLR2 inhibition (Figure 5). However, Hb suppression of Actn2, Myh6, Tnni3, and Cacna1c was only alleviated when both TLR2 and TLR4 pharmacological inhibitors were present (Figure 5).

Figure 5.

Hemoglobin mediates gene repression through TLR2 and TLR4

(A and B) A study was conducted to determine the effect of hemoglobin (Hb) on (A) fibroblast-to-cardiomyocyte reprogramming and (B) fibroblast gene expression. With respect to fibroblast-to-cardiomyocyte reprogramming, cardiac fibroblasts were transfected with either miR combo or a non-targeting control miR. 24 h later, the cells were incubated with vehicle, a TLR2 pharmacological inhibitor, a TLR4 pharmacological inhibitor, or a combination of both pharmacological inhibitors for 3 h. After incubation with the indicated pharmacological inhibitors, Hb was added (5 mg/mL) to the media. All media was replaced the next day. Fourteen days after miR transfection, cells were analyzed for expression of the indicated cardiomyocyte specific genes by qPCR. Expression values were normalized to the housekeeping gene Gapdh. N = 6–10. One-sample t tests were used to compare groups to the control group (∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05; ns, not significant). t test was used to determine the significance between the miR combo groups (##p < 0.01, #p < 0.05; ns, not significant). With respect to fibroblast gene expression, cardiac fibroblasts were incubated with vehicle, a TLR2 pharmacological inhibitor, a TLR4 pharmacological inhibitor, or a combination of both pharmacological inhibitors for 3 h. After incubation with the indicated pharmacological inhibitors, Hb was added (5 mg/mL) to the media. All media was replaced the next day. Expression of the indicated fibroblast-specific genes was determined by qPCR and normalized to the housekeeping gene Gapdh. N = 10. One-sample t tests were used to compare groups to the control group (∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05; ns, not significant).

The effect of TLR2/TLR4 inhibition on fibroblast-specific genes was also determined. Again, Hb significantly repressed Acta2 but had no effect on Col1a1 or Postn (Figure 5). Acta2 suppression by Hb was alleviated by either TLR2 or TLR4 inhibition (Figure 5). Unexpectedly, when Hb was present, inhibition of both TLR2 and TLR4 induced the expression of Acta2 and Postn (Figure 5).

To further the mechanistic understanding of Hb-mediated suppression, we carried out ATAC-seq. ATAC-seq is a method to study chromatin accessibility. Analysis of ATAC-seq datasets focused on differentially accessible regions (DARs) and promoter accessibility.

Control fibroblasts displayed the highest number of accessible chromatin regions, with approximately twice as many as miR combo-reprogrammed cells (Figure 6A). These accessible regions were compared to each other to identify DARs. The vast majority of DARs corresponded to regions that lost accessibility during reprogramming. In contrast, DARs representing regions opened by reprogramming were relatively low in number (Figure 6A). Hb resulted in substantial numbers of both gained and lost DARs, indicating a strong chromatin-modulatory effect (Figure 6A).

Figure 6.

Hemoglobin alters chromatin accessibility during cardiac reprogramming

Cardiac fibroblasts were transfected with either miR combo or a non-targeting control miR. The next day, Hb (5 mg/mL) was added for 24 h. Seven days after transfection, ATAC-seq was performed to assess chromatin accessibility (N = 3). (A) Differentially accessible regions (DARs) were identified and classified into four groups: A, regions that became accessible with reprogramming; B, regions that were accessible in control cells but closed with reprogramming; C, regions opened by reprogramming but closed in the presence of hemoglobin; and D, regions that became accessible only in the presence of hemoglobin. The pie charts summarize these data. Top: counts of accessible regions and DARs. Middle: genomic distribution of accessible regions. Bottom: genomic distribution of DARs. (B) Genes nearest to DARs were identified and subjected to gene ontology analysis to reveal enriched biological processes. Bubble size is determined by gene ratio (the number of genes found in the GO term divided by the GO term size). Bubble color is defined by the significance of the gene ontology term, with larger -log10 FDR p values representing greater significance. A, regions that became accessible during reprogramming; B, regions that were closed during reprogramming; C, reprogramming regions that lost accessibility with Hb, D, regions that became accessible with Hb.

Accessible regions and DARs were mapped to their genomic location. Accessible regions were broadly similar across all conditions, with the majority located in promoter regions (<1 kb of the transcription start site) or intergenic regions (Figure 6A). In contrast, DARs were predominantly intergenic or intronic and rarely occurred in promoters (Figure 6A). This pattern of DAR localization was similar across all comparisons tested (Figure 6A).

DARs were annotated to identify their nearest gene. The resulting gene list was analyzed by gene ontology to infer biological function. The most significantly enriched categories included developmental processes, pattern formation, and receptor/intracellular signaling pathways (Figure 6B). This was consistent across promoter-, gene body-, and intergenic-associated DARs (Figure 6B). Similarly, the same pattern was seen irrespective of whether the DARs represented regions becoming accessible and inaccessible with either miR combo or Hb (Figure 5B). Notably, promoter DARs that became accessible specifically with Hb were solely enriched for immune and apoptotic processes (Figure 6B).

Transcription factor motif analysis was used to further infer biological mechanisms. Motifs unique to each group of DARs were further analyzed. Enriched motifs in DARs gained or lost during reprogramming were rare (Figure 7). In contrast, motif enrichment was more common in Hb-associated DARs (Figure 7). To glean more biological insights, motif enrichment was integrated with RNA sequencing (RNA-seq) data. Here, expression levels of transcription factors known to bind to the enriched motifs were determined. In DARs representing reprogramming regions made inaccessible by Hb, Cebp (Cebpd, Cebpb, and Cebpg), and PAR-bZIP (Tef) family transcription factors were notable (Figure 7A). In contrast, DARs representing reprogramming regions made accessible by Hb, transcription factors associated with the stress response (JunD), differentiation (Tead1), and genome integrity (Cftf) were identified (Figure 7A). Further analysis indicated that many of the identified transcription factors are regulated, both positively and negatively, by miR combo (Figure 7B).

Figure 7.

Hemoglobin reshapes transcription factor networks during reprogramming

Motif enrichment analysis was performed on each of the four DAR groups to identify transcription factor-binding motifs (q < 0.01). (A) Motifs are plotted by statistical significance (–log10 q value) and expression level of the corresponding transcription factor. The dotted line represents the expression of the housekeeping gene Gapdh. (B) Motifs are plotted by statistical significance and the effect of miR combo on the expression of their associated transcription factor. N = 3. A DARs, regions that became accessible during reprogramming; B DARs, regions that were closed during reprogramming; C DARs, reprogramming regions that lost accessibility with Hb; D DARs, regions that became accessible with Hb.

Following the aforementioned, analysis was performed on the promoters of genes specific to cardiomyocytes, skeletal muscle, fibroblasts, neuronal cells, and endothelial cells. To enable comparisons between genes, chromatin content at each base in the promoter is normalized to the sum of chromatin content over the entire promoter. To determine how reprogramming and Hb affect chromatin accessibility, normalized chromatin content in the control cells is subtracted from that in cells undergoing reprogramming and incubated with either vehicle or Hb. A positive product indicates reduced chromatin accessibility, whereas a negative result indicates increased chromatin accessibility. Interestingly, fibroblast-to-cardiomyocyte reprogramming did not significantly alter chromatin accessibility in any class of promoter, though there was a greater trend to increased accessibility in the cardiomyocyte-specific gene group (Figure 8A). Nevertheless, Hb significantly reduced chromatin accessibility in the cardiomyocyte-specific, skeletal muscle-specific, and non-muscle gene groups (Figure 8A).

Figure 8.

Hemoglobin induces chromatin compaction in cardiomyocyte-specific genes

Cardiac fibroblasts were transfected with either miR combo or a non-targeting control miR. The day after transfection, Hb (5 mg/mL) was added to the cells for 24 h. Seven days after miR transfection, chromatin was isolated and analyzed via ATAC-seq. Promoters (-3 kb to +1 kb of the transcription start site) were analyzed for chromatin content. In brief, the promoter was divided into 1 base-pair bins and the number of sequences in each bin was calculated. The number in each bin was then normalized to the total number of sequences. The normalized number of sequences in each bin of the control group was subsequently subtracted from the normalized number of sequences in each bin of the test group (miR combo and miR combo plus Hb). In total, 55 gene promoters were analyzed. Of these, 16 belonged to cardiomyocyte-specific genes, 12 to skeletal muscle-specific genes, and 27 to non-muscle (fibroblast-, endothelial-, and neuron-specific) genes. (A) Normalized read counts in each 1 bp bin after control subtraction (Δnormalized read count) were summed for each of the 55 genes. Positive numbers indicate chromatin compaction while negative numbers indicate chromatin opening. One-sample t tests were used to compare groups to 0, which is no change in chromatin state (∗∗∗p < 0.001; ns, not significant). A t test was used to determine the significance between the miR combo groups (###p < 0.001, #p < 0.05). (B) Within each gene group, the average Δnormalized read count at each 1 bp bin was determined. A cumulative sum of read counts was then determined from the beginning of the promoter (3 kb upstream of the TSS). The results are plotted in the heatmaps.

To determine which region of the promoter was being compactified, a cumulative sum of the normalized read count was determined. In this analysis, the objective was to identify regions where the cumulative sum was significantly more positive (more inaccessible) in the miR combo + Hb group than the miR combo + vehicle group. In the cardiomyocyte-gene group, one such region was identified to be −500 to −750 bp upstream of the transcription start site (Figure 8B). Interestingly, the same region was also rendered more inaccessible in the skeletal muscle-specific group (Figure 8B). However, the reverse was found in the non-muscle group of genes. This region was rendered deeply inaccessible in reprogramming cells, a situation alleviated by the addition of Hb (Figure 8B).

Discussion

This study demonstrates that Hb exerts an inhibitory effect on fibroblast-to-cardiomyocyte reprogramming. The original idea to investigate a possible role for innate immunity in fibroblast-to-cardiomyocyte reprogramming arose from the observation that reprogramming efficacy is higher in vivo than in vitro.² This disparity suggested that the post-injury environment was somehow beneficial. Consequently, we began to investigate how cardiac DAMPs impacted the process of fibroblast-to-cardiomyocyte reprogramming. The majority of our research to date has focused on RNA DAMPs. Most notably, we identified roles for the RNA-sensing receptor family members TLR3 and Rig1.^3,4,8 Both receptors play a positive role in fibroblast-to-cardiomyocyte reprogramming, and we have used this knowledge to improve fibroblast-to-cardiomyocyte reprogramming efficacy. However, RNA molecules are but one component of the DAMPs released by cells dying in response to infarction. Thus, to improve our understanding of how cardiac DAMPs impact fibroblast-to-cardiomyocyte reprogramming, we conducted experiments with extracts derived from freeze-thawing of heart tissue. The freeze-thaw method has been used previously to generate a cardiac DAMP mixture.¹³ The cardiac DAMPs generated by this method are a mixture of RNA, DNA, and protein. In Ma et al.,¹³ cardiac DAMPs generated via freeze-thawing of heart tissue induced neutrophils to polarize into a proinflammatory N1 phenotype by activating TLR4.¹³ Similarly, we found that a cardiac DAMP mixture induced macrophages to secrete molecules that negatively impacted fibroblast-to-cardiomyocyte reprogramming. Simplifying the research strategy, we subsequently found that the cardiac DAMP mixture remained inhibitory when applied directly to reprogramming cardiac fibroblasts.

To identify the component of the cardiac DAMP mixture mediating suppression of fibroblast-to-cardiomyocyte reprogramming, mass spectrometry was employed. By molar mass, an Hb subunit was identified as the largest constituent of the cardiac DAMP mixture. Further experiments showed that Hb fully recapitulated the inhibitory effects of heart extract on fibroblast-to-cardiomyocyte reprogramming. Functioning as a DAMP, Hb can activate both TLR2 and TLR4.^10,14,15 Binding to TLR2/4 results in expression of pro-inflammatory cytokines in large part through activation of NF-κB. Indeed, we found that Hb induced NF-κB activity in cardiac fibroblasts. However, we found that Hb effects on NF-κB activation and fibroblast-to-cardiomyocyte reprogramming were de-coupled from each other. While Hb-mediated NF-κB activation was fully suppressed by a TLR4 inhibitor, the effects of Hb on fibroblast-to-cardiomyocyte reprogramming required the inhibition of both TLR4 and TLR2. The decoupling of Hb stimulated NF-κB and fibroblast-to-cardiomyocyte reprogramming suppression was surprising, suggesting the involvement of novel pathway requiring further investigation.

Seeing that NF-κB was not involved in Hb-mediated suppression of fibroblast-to-cardiomyocyte reprogramming, we investigated alternatives. Based on our previous work,^8,16 we focused on chromatin accessibility. To that end, ATAC-seq was employed. Analysis proceeded in two ways. The first approach involved identifying DARs, localizing the DARs to their nearest gene, and investigating the resulting gene lists through gene ontological analysis. With respect to fibroblast-to-cardiomyocyte reprogramming, DARs-representing regions that were closed during reprogramming significantly outnumbered those representing regions that became accessible. This has also been observed in single-cell ATAC-seq studies of fibroblast-to-cardiomyocyte reprogramming driven by transcription factors.^17,18 Similarly, across studies, genes nearest to DARs were broadly associated with development and signaling. Where the studies diverged was in transcription factor motif enrichment. In transcription factor-based reprogramming, motif enrichment analysis revealed extensive remodeling of transcription factor networks.^17,18 In contrast, such motifs were notably scarce in our miRNA-based system. Indeed, by an approximately 10:1 ratio based on reported numbers. This difference is expected given the distinct modes of action of transcription factors versus miRNAs.¹⁹ Unlike miR combo, Hb followed a transcription-factor pattern with significant remodeling of transcription factor networks. Emerging networks (JunD, Myc, and Max) are characteristic of TLR2/4-induced stress responses.^20,21 Disappearing networks included members of the Cebp and PAR-bZIP families. The latter are noteworthy because these networks did not appear in DARs gained or lost during reprogramming. This suggests that miR combo may act by engaging and reconfiguring latent, pre-existing transcription factor networks (e.g., CEBP/PAR-bZIP sites) present in starting fibroblasts, rather than by creating new accessible sites. Such engagement could drive the chromatin changes necessary for successful reprogramming. It is important to note that ATAC-seq can only infer transcription factor binding. Direct assays will be needed to determine whether these proposed networks actively participate in fibroblast-to-cardiomyocyte reprogramming.

The second approach to analyze the ATAC-seq data was one focused on genes that are specific to cardiomyocytes or other non-cardiomyocyte lineages. In this approach, Hb was found to induce chromatin compaction in a region 500–700 bp upstream of the transcription start site in a large panel of cardiomyocyte-specific genes. In addition, Hb was also found to induce chromatin compaction in non-cardiomyocyte genes. Similarly, we also noted that Hb inhibited expression of the fibroblast gene Acta2. Thus, Hb appears to be a general inhibitor of transcription rather than one focused on cardiomyocyte-specific genes.

One other intriguing observation in our study was the finding that conditioned media from LPS-treated macrophages increased cardiac gene expression in reprogrammed cells. LPS signals through the same receptors as Hb, and thus one might expect LPS to be similarly inhibitory. One possible explanation is that conditioned media contains multiple paracrine factors whose net effect is distinct from direct Hb signaling. An alternative possibility is that the differential effects that we observed between LPS and Hb on pro-inflammatory cytokine expression lead to differential effects on cardiac gene expression. Further work will be needed to clarify the role of LPS in fibroblast-to-cardiomyocyte reprogramming.

In summary, we identified Hb as a negative regulator of fibroblast-to-cardiomyocyte reprogramming. The challenges will be identifying the mechanism of suppression and how to translate these findings in vivo. Pharmacological inhibitors of TLR2 and TLR4 suitable for in vivo use do exist, but dissecting their impact on reprogramming versus broader post-injury processes would require careful study.

Materials and methods

Cell isolation and culture

Cardiac fibroblasts were isolated from 1 day old neonatal C57BL/6 mice following the method described by Jayawardena et al.²² Once isolated, these fibroblasts were grown in growth medium consisting of DMEM (ATCC, catalog no. 30–2002) supplemented with 15% v/v fetal bovine serum (Genessee) and 1% v/v penicillin/streptomycin (GIBCO, catalog no. 15140-122). The cells were routinely passaged when they reached 70%–80% confluence, using a 0.05% w/v trypsin solution (GIBCO, catalog no. 25300-054). All experimental procedures involving animals were conducted with the approval of Duke University’s Division of Laboratory Animals (DLAR) and the Duke Institutional Animal Care and Use Committee (IACUC), under the protocol no. A035-22-02.

Macrophages

RAW264.7 cells and primary macrophages were cultured and isolated as previously described.²³

Transfection of cardiac fibroblasts with miR combo

Fibroblasts initially cultured as passage zero were used for experiments at passage two. These cells were seeded at a density of 5,000 cells per square centimeter in growth medium in a 12-well plate. 24 h after seeding, transfection was performed using DharmaFECT 1 (Horizon Discovery, catalog no. T-2001-03). Each well of a 12-well plate received 5 nmol of miRNAs. Transfections were carried out according to the guidelines provided by the manufacturer. After 24 h, the transfection complexes were removed, and the fibroblasts continued to be cultured in growth medium for the remainder of the study with media changes every 48 h.

Small molecules

Hb (Millipore Sigma, H2625), TLR2 signaling inhibitor-TL2-C29 (InvivoGen, catalog no. inh-c29), and TLR4 inhibitor-CLI-095 (InvivoGen, catalog no. tlrl-cli95-4) were used.

Heart extracts

Mouse hearts were excised and immediately frozen at −80°C overnight. The frozen tissue was then minced and passed through a 100-micron filter. Following this, the filtrate was centrifuged at 400 g for 5 min at room temperature. The resulting supernatant was collected as the heart extract and served as a source of DAMPs. A freeze-thaw method to generate cardiac DAMPs has been published previously.¹³

Blood extracts

Whole blood was collected and prevented from coagulation via the addition of EDTA to a final concentration of 20 mM. Aliquots were snap frozen at −80°C prior to use.

qPCR

Total RNA was isolated using the Quick-RNA MiniPrep Kit (Zymo Research, catalog no. R1055) according to the manufacturer’s protocol. RNA was then converted into cDNA using the high-capacity cDNA reverse transcription kit from Applied Biosystems (catalog no. 4368814) in a reaction volume of 40 μL. For qPCR, 4 μL of the synthesized cDNA was used along with FAM-tagged gene-specific primers and TaqMan Gene Expression Master Mix, both from Applied Biosystems, Waltham, MA, USA. The qPCR primers for Gapdh (Mm99999915_m1, catalog no. 4331182), Actn2 (Mm00473657_m1, catalog no. 4331182), Myh6 (Mm00440359_m1, catalog no. 4331182), Tnni3 (Mm00437164_m1, catalog no. 4331182), Tnnc1 (Mm00437111_m1, catalog no. 4331182), Acta2 (Mm00725412_s1, catalog no. 4331182), Postn (Mm01284919_m1, catalog no. 4331182), Col1a1 (Mm00801645_g1, catalog no. 4351372), Scna5(Mm01342518_m1, catalog no. 4331182), Cacna1c (Mm01188822_m1, catalog no. 4331182), and Ryr2 (Mm00465877_m1, catalog no. 4331182) were obtained from Thermo Fisher Scientific. The expression levels were quantified by normalizing to the housekeeping gene Gapdh.

Immunoblotting

As described in previous studies,⁸ cells were washed once with PBS, and proteins were extracted using 150 μL of lysis buffer (62.5 mM Tris pH 8, 1% [v/v] SDS, 1% [v/v] mammalian protease inhibitor cocktail [Sigma, catalog no. P8340-5ML], and 10% [v/v] phosphatase inhibitor cocktail [Roche, catalog no. 4906837001, 1 tablet dissolved in 1 mL of water]) per well in a 6-well plate on ice. The proteins were then separated by SDS-PAGE (Invitrogen, catalog no. NP0335BOX) and transferred onto nitrocellulose membranes (Bio-Rad, catalog no. 1620115). Membranes were incubated overnight at 4°C with a primary antibody diluted in antibody buffer (50 mM Tris-HCl, 150 mM NaCl, 5% w/v BSA, 0.1% v/v Tween 20, pH 7.4). Afterward, the membranes were washed three times with Tris-buffered saline (TBS)-Tween (50 mM Tris-HCl, 150 mM NaCl, 0.1% v/v Tween 20, pH 7.4) at room temperature and incubated for 1.5 h with HRP-conjugated secondary in the antibody buffer. Following three additional washes with TBS-Tween, protein bands were detected using chemiluminescence (Cytiva, catalog no. RPN2235). Band intensities were analyzed with either ImageJ or Syngene software. Primary antibodies used were Gapdh (Cell Signaling, catalog no. 5174), β-actin (Cell Signaling, catalog no. 4970S), phospho-NF-κB (Cell Signaling, catalog no. 3033S), and phospho-p38 MAPK (Cell Signaling, catalog no. 4511S), all diluted 1:1,000. Secondary HRP-conjugated antibodies (Cell Signaling, catalog no. 7074S). Both primary and secondary antibodies were also used at a 1:1,000 dilution. All antibodies were validated by the manufacturer.

Cardiomyocyte counting and beating: cardiomyocyte counting and beating were carried out as described previously.³

Hb ELISA

A Hb ELISA kit was obtained from Novus Biologicals (catalog no. NBP3-18008) and used according to the manufacturer’s instructions.

ATAC-seq

Libraries were prepared from 50,000 cells using a Zymo-Seq ATAC Library Kit (catalog no. D5458). Sequences were aligned to the mouse genome via Bowtie2. Sequencing reads were aligned to the mouse genome with Bowtie2. Aligned reads were filtered to retain true pairs, remove low-quality reads (quality <30), and to exclude mitochondrial reads.

DARs

Peaks representing accessible chromatin were called with MACS2 CallPeak using no shifting model, 147 bp extension, 0 bp shift, q value cutoff 0.05, no broad regions, and allowing duplicate tags at the same position. Peak sets from each group (control, miR combo, and miR combo + Hb) were sorted (bedtools SortBED), concatenated, and intersected with bedtools Intersect Intervals to identify DARs. DARs were annotated with ChIPseeker using a Gencode vM25 GTF. Annotations were grouped as promoter (0–3 kb upstream of transcription), gene body (exons, introns, UTRs), or intergenic (>3 kb from their associated gene). Unique gene lists were analyzed for gene ontology using the Gene Ontology website and summarized with Revigo (0.4 setting).

Transcription factor motif enrichment was performed using MEME/DREME. DARs were converted to FASTA sequences (bedtools getfasta), filtered for uniqueness, and analyzed. Enriched sequences were compared to known transcription factor motifs with TomTom, and motifs with q value <0.01 were selected for further analysis. Expression data was acquired from Wang et al.⁹

Promoter-specific analysis

The approach is described in detail elsewhere.^16,24 In brief, BamCoverage (1 bp window, normalized to effective mouse genome size and in MNase mode) was used to determine read count for each promoter (-3 kb upstream of the transcription start site to + 1 kb downstream of the transcription start site) for the genes in the following list. Read count data were normalized to ensure gene had equal weight via the equation:

For gene x, normalized read count_position-n = read count_position-n/∑read count_genex where position-n is the position of the 1 bp window with respect to the transcription start site.

To determine the effects on chromatin architecture between two conditions the following formula was applied.

For gene x, Δnormalized read count_position-n = (normalized read count_position-n)_condition-a– (normalized read count_position-n)_condition-b.

The cardiomyocyte-specific gene group: Actn2, Cacna1c, Kcna4, Kcnj2, Mb, Mef2C, Myh6, Myl2, Nebl, Ryr2, Scn5a, Sln, Tnni3, Tnnt2, and Ttn.

The skeletal muscle gene group: Cacna1s, Myh1, Myh2, Myh3, Myh4, Myod1, Myog, Neb, Ryr1, Scn4a, Tnni1, and Tnni2.

The non-muscle group: Atl1, Cd34, Cdh1, Col1a1, Col1a2, Dcx, Eno2, Eng, Flt1, Flt4, Map2, Mapt, Ncam1, Neurod1, Nlgn1, Pecam1, Postn, Rbfox3, Scn4a, S100a4, Syn1, Tek, Tcf21, Thy1, Vcam1, Vegfa, and Vwf. This group is larger to ensure sufficient coverage of alternative reprogramming trajectories (fibroblast, endothelial, neuronal).

The ATAC-seq dataset can be found on the NIH Single Read Archive under the BioProject ID no. PRJNA1403564.

Statistics

The data were analyzed by one-sample t tests to determine significance to a control group.

DATA AND CODE AVAILABILITY

All of the data are presented in this article.

Acknowledgments

We thank the Edna and Fred Mandel Jr Foundation for their support. The NHLBI grant R01-HL178582 provides support for I.A., X.W., R.E.P., and C.P.H. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author contributions

I.A., data curation, formal analysis, investigation, and writing – review and editing; X.W., data curation, formal analysis, investigation, and writing – review and editing; R.E.P., formal analysis, methodology, project administration, and writing – review and editing; V.J.D., funding acquisition, methodology, project administration, supervision, and writing – review and editing; C.P.H., conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, validation, visualization, writing – original draft, and writing – review and editing.

Declaration of interests

The authors report no conflicts of interest.