Infarct Collagen Topography Regulates Fibroblast Fate Via p38-Yap-TEAD Signals

Abstract

Rationale

Myocardial infarction (MI) causes spatial variation in collagen organization and phenotypic diversity in fibroblasts, which regulate the heart’s extracellular matrix (ECM). The relationship between collagen structure and fibroblast phenotype is poorly understood but could provide insights regarding the mechanistic basis for myofibroblast heterogeneity in the injured heart.

Objective

To investigate the role of collagen organization in cardiac fibroblast fate determination.

Methods and Results

Biomimetic topographies were nanofabricated to recapitulate differential collagen organization in the infarcted mouse heart. Here adult cardiac fibroblasts were freshly isolated and cultured on ECM topographical mimetics for 72 hours. Aligned mimetics caused cardiac fibroblasts to elongate while randomly organized topographies induced circular morphology similar to the disparate myofibroblast morphologies measured in vivo. Alignment cues also induced myofibroblast differentiation, as more than 60% of fibroblasts formed α-smooth muscle actin (αSMA) stress fibers and expressed myofibroblast-specific ECM genes like periostin. By contrast, random organization caused 38% of cardiac fibroblasts to express αSMA albeit with down-regulated myofibroblast-specific ECM genes. Coupling topographical cues with the profibrotic agonist, TGFb, additively upregulated myofibroblast-specific ECM genes independent of topography, but only fibroblasts on flat and randomly oriented mimetics had increased percentages of fibroblasts with αSMA stress fibers. Increased tension sensation at focal adhesions induced myofibroblast differentiation on aligned mimetics. These signals were transduced by p38-YAP-TEAD interactions, in which both p38 and YAP-TEAD binding were required for myofibroblast differentiation. By contrast randomly oriented mimetics did not change focal adhesion tension sensation or enrich for p38-YAP-TEAD interactions, which explains the topography-dependent diversity in fibroblast phenotypes observed here.

Conclusions

Spatial variations in collagen organization regulate cardiac fibroblast phenotype through mechanical activation of p38-YAP-TEAD signaling, which likely contribute to myofibroblast heterogeneity in the infarcted myocardium.

Graphical Abstract

INTRODUCTION

The structural organization of the heart’s extracellular matrix (ECM) undergoes extensive remodeling after myocardial infarction (MI), which continues for weeks after injury until the scar matures. Implicit in the remodeling process are changes in ECM composition but also its mechanical properties. These changes are vital to patient outcomes as structurally unstable scars can result in cardiac rupture^{1, 2}. The association between scar properties and cardiac function has been rigorously investigated^1–3, but the regulatory mechanisms underlying persistent scar remodeling are still poorly understood.

There are a growing number of studies demonstrating that the ECM’s physical properties regulate cell state transitions^4–7. Recent findings indicate that the differentiation of resident cardiac fibroblasts into myofibroblasts, which are defined by de novo expression of α-smooth muscle actin (αSMA) and ECM proteins like Periostin (Postn), is essential for proper regulation of post-infarct scarring and ECM remodeling^8–11. Single-cell RNAseq studies indicate there is considerable heterogeneity in cardiac fibroblast and myofibroblast populations following MI^{12, 13}. Because this approach eliminates spatial information it is unclear how local environmental cues impact this heterogeneity, although previous findings suggest there are temporal and spatial signals that regulate myofibroblast formation following MI^{5, 14–16}. Indeed, histologic evidence suggests that myofibroblast density is low in the scar’s core but high in the border zone region^{14, 17}. Coincidently, regional differences in ECM content and organization are also observed in the remodeled heart after MI, suggesting there is a fundamental relationship between collagen fiber topography and the conversion of cardiac fibroblasts into activated myofibroblasts^{1, 18, 19}. How these spatial variations in matrix topography influence cardiac fibroblast cell state have not been previously explored but could explain (at least in part) some of the fibroblast/myofibroblast heterogeneity and regional variations in cardiac fibroblast phenotype following MI. Moreover, regional ECM structure might underlie bi-directional positive feedback between fibroblasts and the ECM that maintain fibroblasts in an activated myofibroblast state that could in turn exacerbate maladaptive fibrotic scarring and dysfunction.

The goal of this study was to examine how structural alignment cues from the ECM influence cardiac fibroblast fate. Studying the direct effects of the ECM’s structural organization on fibroblast phenotypes in vivo is extremely complicated because the cardiac environment is dynamically changing in both space and time throughout post-MI wound healing and remodeling. Variations in pressure, volume, inflammation, compensatory feedback mechanisms, and alterations in collagen content and structure are happening simultaneously during this process, making it impossible to isolate the effects of collagen organization on cardiac fibroblast biology. In order to discriminate how spatial heterogeneity in ECM topography in the infarcted heart directly influences cardiac fibroblast phenotype, we engineered biomimetics of collagen alignment cues associated with the border zone and core of the infarct scar. These in vitro mimetics were engineered using nanofabrication techniques and ultraviolet light (UV)-curable polymers that recapitulate the stiffness of the infarcted rodent myocardium. By coupling these nanoengineered topographical mimetics with targeted genetic silencing of key signaling pathways, we uncovered that aligned topographical cues increase tension sensation by focal adhesions in cardiac fibroblasts, which in turn induce myofibroblast differentiation programs through boosting p38-YAP-TEAD signaling. By contrast, disorganized topographic mimetics of the infarct scar induced some myofibroblast-like cytoskeletal changes but with reduced ECM gene expression that occurred independent of YAP-TEAD signaling. Taken together these data suggest collagen organization differentially regulates cytoskeletal and ECM components of the myofibroblast phenotype, which we postulate contributes to myofibroblast diversity in the infarcted myocardium.

METHODS

The Supplemental Materials Expanded Methods and Major Resource Sections describe all of the experimental approaches and materials used in this study.

All animal experimentation was approved by the University of Washington’s Institutional Animal Care and Use Committee. Postn-lineage reporter mice contain a gene targeted tamoxifen (Tam)-inducible Cre cassette in the Postn locus (Postn^iCre) and a conditional membrane-targeted dual fluorescent reporter (mT/mG) in the Rosa26 locus^{9, 10, 20}. Cardiac fibroblast-specific p38a knockout mice were generated by crossing LoxP-targeted Mapk14 mice (p38^F/F) and mice with a Tam-inducible Cre cassette knocked into the Tcf21 locus (p38^F/F-Tcf21^iCre) ¹⁰. Cardiac fibroblasts were freshly isolated from the hearts of adult p38^F/F mice by enzymatic digestion, expanded, and seeded onto gelatin coated biomimetics for experimentation. Representative images were selected to depict the mean characteristic of each experimental group.

Statistics

Data are presented as mean ± SEM (standard error of the mean) and p-values for each comparison are shown within the figure or in Online Table I. GraphPad Prism versions 7 & 8 and R-statistical package were used for statistical testing, which is described for each experiment in the figure legends and Online Table I. Expanded details are provided in the Supplemental Materials.

RESULTS

Regions of Aligned Collagen Are Associated with Higher Myofibroblast Density

After myocardial infarction, there is widespread spatial heterogeneity in collagen alignment that occurs across species^{18, 21, 22}. Here spatial variation in collagen fiber orientation was quantified in mice subjected to a surgical model of MI. This procedure elicited infarct sizes of 30–40% of the heart and robust scarring of similar magnitude as demonstrated by PSR staining of transverse myocardial sections (Figure 1A). At 1-week post MI there were regions of highly aligned collagen fibers that run in parallel as well as areas in which collagen fibers were disorganized and randomly oriented. This observation was confirmed by collagen 1 (Col1a) immunofluorescent imaging (Figure 1B). Here collagen alignment was analyzed using CurveAlign software^{23, 24}, which calculated collagen alignment coefficients ranging from 0 to 1, where 1 represents fibers that are oriented in parallel and those approaching 0 have no directionality. CurveAlign analysis revealed that in the border zone of infarcted mouse hearts collagen fibers were more highly aligned; whereas, fibers in the core of the scar were randomly oriented (Figure 1C). Lineage tracing evidence has demonstrated that myofibroblasts are identified by the expression of Postn and αSMA⁹, and studies prior to the development of Postn-lineage reporter mice had demonstrated higher myofibroblast density in the border zone versus the scar’s core^{14, 17}. To examine whether there are regional differences in myofibroblast density associated with collagen organization, the infarcts described above were performed in Postn lineage reporter mice (Postn^iCre-mT/mG). To calculate regional myofibroblast cell densities any area with an average collagen alignment coefficient greater than 0.4 was considered aligned and those less than 0.4 considered randomly oriented. Within each area the number of cells that were single positive for Postn (Postn⁺) and double positive for αSMA and Postn (αSMA⁺,Postn⁺) were counted and normalized to the tissue area encompassing the aligned versus randomly oriented collagen fibers. Both Postn⁺ and αSMA⁺,Postn⁺ cells were scored as recent single cell RNAseq studies have shown transcriptional heterogeneity within the cardiac myofibroblast populations after MI^{12, 13}. Figures 1D–E demonstrate that the density of both Postn⁺ and αSMA⁺,Postn⁺ myofibroblasts are significantly reduced in regions of randomly aligned collagen fibers. Linear regression showed a significant relationship between collagen alignment and αSMA⁺,Postn⁺ myofibroblast density, albeit with a small but reliable effect size (Figure 1F). In addition, Postn⁺ myofibroblast morphology was also significantly different between the two regions (Figure 1G) in which areas with aligned collagen had thin elongated myofibroblasts (Figure 1B, bottom right panel) versus areas of random collagen alignment that had highly circular cells (Figure 1B, top right panel). Together these data suggest that in vivo collagen alignment in the infarcted heart is associated with both myofibroblast density and morphology.

Figure 1. Regions of the infarcted heart with highly aligned collagen are associated with increased myofibroblast density.

Representative images of A Sirius Red-Fast Green and B Collagen I immunofluorescent staining of cardiac sections from infarcted periostin-lineage reporter mice (Postn^iCre-mT/mG). Boxed insets show examples of regions with collagen fibers aligned in parallel versus disorganized (random) orientation. Scale bars=100mm. C Quantification of the average collagen alignment coefficient in the border zone versus the infarct core. Mann-Whitney test was used, n=3 mice per group with 2–4 regions assessed per mouse. D Quantification of the average number of Postn⁺ cells (stained by eGFP, Postn^iCre- mG in B ) and E cells positive for both αSMA and Postn in regions of aligned (coefficients ≥ 0.4 in C, n=8) versus randomly (coefficients < 0.4 in C, n=10) oriented collagen. Unpaired t test used. F Linear regression analysis demonstrating the relationship between αSMA⁺, Postn⁺ myofibroblast density and collagen organization where coefficients ≥ 0.4 equate to a high frequency of aligned collagen fibers. Dashed line = 95% confidence interval. G Violin plot of the shape of Postn⁺ cells in regions of aligned (coefficients ≥ 0.4 in C) versus randomly (coefficients < 0.4 in C) oriented collagen. Dashed line represents the mean and dotted lines represent the 1st and 3rd quartile. Mann-Whitney test used. Graphs represent Mean ± SEM, and filled circles denote biological replicates.

Nanoengineered Topographical Mimetics Alter Fibroblast Phenotype

Given the observed relationship between collagen alignment and αSMA⁺,Postn⁺ myofibroblast density and morphology, we hypothesized that topographical cues in the ECM regulates cardiac fibroblast phenotype. To test this hypothesis biomimetics of the aligned and randomly oriented collagen topography were engineered using previously described nanofabrication techniques schematically depicted in Figure 2A²⁵. Norland optical adhesive-76 (NOA76), a polyurethane acrylate material, was used as the biomaterial because its properties robustly recapitulate the elasticity of the infarcted myocardium, which has been quantified in ranges from 50–800 kPa in rodents^{26, 27}. A master pattern film was placed on top of the UV-curing polymer and cured to print the aligned pattern. A randomly oriented collagen pattern was generated by subjecting the aligned pattern to biaxial stretch, ozone treatment, and then shrinking as previously described²⁵. The UV-curing polymer was then used to manufacture biomimetic devices of the random collagen topography. Flat (unpatterned) surfaces were developed by using a smooth silicon surface as a master. Flat devices were used to recapitulate standard cell culture surfaces and control for the effects of the polymer’s mechanical properties which has a lower elastic modulus than standard tissue culture plastic or glass²⁸. Figure 2B & D show scanning electron microscopy (SEM) images of a region of the aligned and randomly oriented collagen fibers from sections used for collagen organization analysis (Figure 1), but at the nanoscale. Figures 2C & E show images of the corresponding biomimetic devices. These images were used to evaluate how closely the engineered mimetics of collagen topography recapitulate the average spatial variations in collagen orientation observed in infarcted mice (Figure 2B–E, bottom graphs). SEM analysis of synthetic and in vivo collagen fiber nanotopographies showed that 95.5% of collagen fibers in the aligned regions in vivo were within 20° of the major orientation of cells and fibers in that area (Figure 2B). Similarly, 100% of the alignment cues cured into the aligned biomimetic devices were in perfect parallel alignment (Figure 2C). For regions of randomly oriented collagen fibers, 39.8% were within 20° of the major orientation of cells and fibers in that region (Figure 2D). The random biomimetic devices also recapitulated the in vivo frequency of alignment in which only 48.1% of the engineered topographic cues were in parallel alignment (Figure 2E).

Figure 2. Nanoengineered mimetics of collagen topography alter cardiac fibroblast proliferation and morphology.

A Schematics depicting the nanofabrication of each biomimetic. Transmission electron microscopy (TEM) image of B (Top) aligned and D (Top) randomly oriented collagen fibers in regions of aligned collagen (coefficients ≥ 0.4 in Figure 1C) in infarcted mouse hearts, and (Bottom) the corresponding histogram of the percent collagen fibers aligned within 20° of the major axis. Scanning electron microscopy (SEM) image of nanoengineered C (Top) aligned and E (Top) randomly oriented collagen biomimetic and (Bottom) the corresponding histogram of the percent collagen fiber aligned within 20° of the major axis. F Quantification of EdU⁺ cardiac fibroblasts cultured on flat, aligned, or random biomimetics in proliferation conditions (10% FBS, n=3 mice, 123–395 cells counted per mouse on Flat, 454–904 cells counted per mouse on Aligned, & 263–1099 cells counted per mouse on Random) and differentiation conditions (2% FBS, n=3 mice, 201–409 cells counted per mouse on Flat, 309–480 cells counted per mouse on Aligned, & 552–1560 cells counted per mouse on Random). Kruskal-Wallis test was performed independently for proliferation and differentiation conditions followed by uncorrected Dunn’s multiple comparison test between patterns. G Quantification of actin stress fiber alignment (n=3 biological replicates per pattern). H Quantification of cell eccentricity (0 = circular and 1 = a line segment) and I cell extent (higher values = more protrusions) in cardiac fibroblasts on flat (94–284 cells studied per mouse), aligned (83–767 cells studied per mouse), or random (171–844 cells studied per mouse) biomimetics. Kruskal-Wallis test followed by uncorrected Dunn’s multiple comparison tests were made between patterns, n=6 biological replicates per group. Graphs represent Mean±SEM.

To ensure that fibroblasts can be cultured on these biomaterials, MEFs were seeded on flat, aligned, and random devices, and attachment, viability, cell area, and alignment were analyzed 3 days later. MEFs attached with equal affinity to all 3 biomimetics and live-dead staining demonstrated there were no differences in viability between groups (Online Figure IA–B). Primary cardiac fibroblasts had similar attachment and viability (data not shown), but proliferation was significantly reduced independent of serum concentration on the aligned and random mimetics when compared to flat (Figure 2F). Cellular sensations of local topographic cues have been shown to align cells in parallel to the collagen fibers^29–31. Here, there were significant topography-dependent effects on MEF morphology and cell alignment (Online Figure IC–D), in which MEFs cultured on aligned mimetics had significantly increased cell areas relative to those on flat surfaces and aligned in parallel with the major orientation of the grooves on the aligned mimetics (Online Figure ID). By contrast, MEFs cultured on random topographies had significantly reduced size and no directionality (Online Figure IC–D). Similar to MEFs, the actin cytoskeleton of adult cardiac fibroblasts oriented in parallel with the aligned patterns (Figure 2G), and their morphology became elongated and spindly as measured by the increased cell eccentricity (Figure 2H) and smaller geometric extent (Figure 2I). These aligned mimetic-dependent changes in cardiac fibroblast morphology are similar to the morphologic changes observed in the Postn⁺ myofibroblasts in regions of aligned collagen 7 days post MI (Figure 1G). While we did not observe the same circular morphology on the random patterns that was observed in vivo (Figure 1G), cardiac fibroblast morphology failed to elongate and spread and were more comparable to fibroblasts on flat mimetics. Together these data highlight that regional variations in topographical cues alters cardiac fibroblast organization and morphology.

Given the significant aligned topography-dependent alterations in cardiac fibroblast morphology and previous evidence demonstrating that collagen organization regulates stem cell phenotypes^29–31, we postulated that cardiac fibroblast cell fate is differentially regulated by aligned and random patterns. Adult cardiac fibroblasts were seeded onto biomimetic platforms for 3 days and then analyzed by immunofluorescent imaging of αSMA⁺ stress fibers (Figure 3A–B), which is a signature of programmed fibroblast to myofibroblast differentiation. The percentage of fibroblasts with αSMA⁺ stress fibers was significantly higher on aligned mimetics in comparison to those with flat and random topographies (Figure 3A–B). Notably, random mimetics also had a significantly higher percentage of the cardiac fibroblast population positive for αSMA stress fibers than those identified on flat surfaces (Figure 3A–B). Additional transcriptional markers of the myofibroblast phenotype were analyzed including ECM genes like Postn (Figure 3C), Collagen 1a1 (Col1a1, Figure 3D), and the EDa splice variant of fibronectin (FnEDa, Figure 3E). All 3 ECM genes were significantly upregulated in cardiac fibroblasts cultured on aligned patterns when compared to those on flat and random surfaces, but the random mimetic induced no alterations to some ECM genes and a decrease in the expression levels of others, suggesting that collagen organization can differentially regulate myofibroblast-associated cytoskeletal and ECM markers. These analyses were also performed on MEFs and yielded similar findings although the degree of activation was globally depressed (Online Figure II).

Figure 3. Aligned topographical mimetics promote cardiac fibroblast to myofibroblast differentiation.

A Immunofluorescent images and B quantification of the percentage of cardiac fibroblasts positive for αSMA stress fibers on flat (177–399 cells studied per mouse), aligned (157–344 cells studied per mouse), and random (220–470 cells studied per mouse) biomimetics with (n=9 mice) and without TGFb (n=3 mice). Scale bars=100mm. Mixed effects logistic regression followed by Bonferroni corrected post hoc comparisons were made between groups. Fold change in C periostin (Postn), D collagen 1a (Col1a1), and E fibronectin EDa splice variant (FnEDa) gene expression in fibroblasts on flat, aligned, and random biomimetics with and without TGFb treatment. Values are calculated using the 2^-DDCt method and expressed relative to the flat,control condition, n=3 biological replicates per group. Log transformed outcome and two-way ANOVA with the natural log of the variable of interest as the outcome variable and Bonferroni corrected pairwise comparisons were used for analysis. Graphs represent Mean±SEM, 1 vs Control,Flat, 2 vs Control,Aligned, 3 vs Control,Random, 4 vs TGFb,Flat, 5 vs TGFb,Aligned.

After MI both chemical and physical signals contribute to the regulation of cardiac fibroblast and myofibroblast phenotypes, and so we examined the combinatorial effects of topographical cues and recombinant TGFb (10ng/ml), a key cytokine responsible for fibroblast to myofibroblast differentiation and cardiac fibrosis⁴. As expected TGFb activated cardiac fibroblasts on flat mimetics as evidenced by the significant increase in cells with αSMA⁺ stress fibers (Figure 3A–B) and upregulation of Postn and FnEDa gene expression (Figure 3 C & E). Interestingly, the combination of aligned surfaces and TGFb decreased the percentage of cardiac fibroblasts positive for αSMA stress fibers (Figure 3A–B), but significantly increased of Postn, Col1a1, and FnEDa gene expression (Figure 3C–E). On both flat and random patterns TGFb treatment additively increased the percentage of cardiac fibroblasts with αSMA⁺ stress fibers and ECM gene expression in cardiac fibroblasts (Figure 3A–E). While TGFb increased ECM gene expression in cardiac fibroblasts on random patterns, overall the response was blunted in comparison to the other mimetics (Figure 3A–E). Taken together these data suggest that TGFb and ECM topography regulate αSMA stress fiber formation using common intracellular signaling pathways, but independent signals operating in parallel appear to regulate the ECM phenotype.

Focal Adhesions Sense and Transduce Alignment Cues

Previous work has shown that focal adhesions transduce physical signals from topographical cues in the ECM into changes in cellular migration³². Hence, we reasoned that collagen alignment mimetics are regulating topography-dependent transitions of fibroblasts into myofibroblasts through focal adhesion signaling. Indeed, immunofluorescent imaging of focal adhesion protein, Tensin 1, showed that focal adhesion number and size were increased in MEFs cultured on aligned mimetics (Figure 4A–C). Similarly, alignment cues increased the expression of tenascin C (TnC), which regulates focal adhesion signaling (Figure 4D)³³. To quantify topography-dependent changes in tension sensation at focal adhesions, cardiac fibroblasts were lentivirally transduced with a Förster Resonance Energy Transfer (FRET)-based tension sensor encoded into the carboxy-tail of vinculin^{34, 35}. Briefly, the sensor was engineered by separating a Clover-mRuby FRET pair with an elastic element in which FRET efficiency and tension are inversely related. Hence, FRET pairs in close proximity permit high FRET efficiency while tension pulls the FRET pairs apart lowering FRET. Using the sensor’s fluorescent output, we observed an increased focal adhesion alignment in cardiac fibroblasts on the aligned biomimetics (Figure 4E & F, Left). Static FRET measurements in fixed cardiac fibroblasts were significantly decreased in cardiac fibroblasts on aligned mimetics, suggesting that aligned topographical cues cause increased tension sensation at the focal adhesion (Figure 4F). Interestingly, fibroblasts on the random mimetics tended to have higher FRET efficiencies and thus lower tension sensation versus those on flat patterns albeit these values were not statistically different. To determine if focal adhesion tension sensation is required for mechanically transducing collagen alignment cues into myofibroblast differentiation, shRNA was used to silence Tensin 1 expression (shTensin 1) and the percentage of fibroblasts with αSMA⁺ stress fibers analyzed. Western blot analysis demonstrated a significant knockdown of Tensin 1 in MEFs when compared to controls (Supplemental Figure 3A). Loss of Tensin 1 significantly lowered baseline fibroblast activation on flat patterns as demonstrated by the reduced percentage of cells with αSMA⁺ stress fiber expression on flat patterns (Figure 4G). In addition, Tensin 1 knockdown blocked aligned pattern-dependent increases myofibroblast differentiation (Figure 4G), indicating that fibroblasts require focal adhesions to sense and transduce aligned structural cues from the ECM into myofibroblast differentiation.

Figure 4. Alignment-induced focal adhesion and cytoskeletal remodeling is mediated by heightened tension sensation.

A Representative immunofluorescent images of MEFs cultured on flat, aligned and random biomimetics stained with DAPI (blue) and Tensin 1 (red). Scale bar = 30mm. Quantification of Tensin 1 positive focal adhesion B number and C size in MEFs cultured on flat, aligned and random biomimetics. D Fold change in Tenascin C (TnC) gene expression in cardiac fibroblasts on flat, aligned and random biomimetics. Values are calculated using the 2^-DDCt method and expressed relative to the flat condition. Quantification of (E) focal adhesion (FA) alignment index and (F) average FRET efficiency in cardiac fibroblasts expressing a vinculin tension sensor. F,Left Representative image of the acceptor intensity and FRET efficiency of fibroblasts on a flat mimetic. Scale bar = 100mm. Graphs represent Mean±SEM and filled circles denote biological replicates. 1-way ANOVA followed by Tukey post hoc comparisons between groups. G Quantification of the percentage of cardiac fibroblasts positive for αSMA stress fibers when cultured on flat, aligned, and random biomimetics with and without Tensin 1 knockdown (siTensin1). Graphs represent Mean±SEM and filled circles denote biological replicates. Mixed effects logistic regression followed by Bonferroni corrected post hoc comparisons were used for analysis, 1 vs Control,Flat, 2 vs Control,Aligned.

p38 Transduces Alignment Cues Into Programmed Myofibroblast Differentiation

Previous work in our lab has demonstrated that p38 is a nodal point in the signaling network regulating programmed myofibroblast differentiation, as several chemical and mechanical inducers of fibrosis including TGFb and cyclic stretch require p38 to induce myofibroblast gene transcription¹⁰. Hence, the role of p38 was examined as an intracellular transducer of topographic cues into programmed myofibroblast differentiation. Adult cardiac fibroblasts were isolated from conditional p38a knockout mice (p38^F/F) and cultured on flat, aligned, and random topographical mimetics with or without recombinant Cre recombinase. Western blot analysis demonstrated that both aligned and random topographies increase p38 expression, which was significantly reduced with Cre-mediated recombination that excises p38α (Figure 5A–B). Similarly, aligned mimetics increased p38 activity in MEFs as measured by Western blot for phosphorylated p38 (Online Figure IIIB). Additional evidence of p38 activity was identified by the enhanced localization of p38 in the nucleus of cardiac fibroblasts cultured on the aligned mimetics (Figure 5C). To examine if p38 is required for pattern-dependent induction of fibroblast to myofibroblast differentiation, p38^F/F cardiac fibroblasts were cultured on the topographical mimetics with and without recombinant Cre treatment, and the percentage of fibroblasts with αSMA⁺ stress fibers calculated. On both aligned and random patterns loss of p38 function blocked pattern-dependent activation (Figure 5D). Candidate pathways that might work in parallel or synergistically with p38 were also examined as transducers of these topographical cues into myofibroblast gene expression. Previous findings have shown that dysregulated yes-associated protein (YAP) activity causes spontaneous cardiac fibroblast activation and fibrosis^{36, 37}. These findings in conjunction with (1) the known role of YAP in regulating the actin cytoskeleton and transmission of topographic cues and (2) data demonstrating that in cancer cells stress induces interactions between YAP’s transcriptional activator TEAD1 and p38 provided rationale for investigating whether YAP-TEAD-p38 interactions might play a role in this model system^{38, 39}. Here, cardiac fibroblasts on aligned patterns had increased nuclear colocalization of p38 and YAP as well as p38 and TEAD (Figure 5E–F). To verify the interaction between p38, YAP, and TEAD, lysates were collected from cardiac fibroblasts cultured on flat, aligned, and random mimetics and a phosphorylated-p38 antibody used to immunoprecipitate proteins bound to activated p38. Western blot analysis of immunoprecipitates demonstrated that on all 3 patterns p38, YAP, and TEAD are complexed together (Figure 5G), but there is an enrichment of p38-YAP-TEAD binding in lysates from fibroblasts on aligned mimetics when compared to flat or random mimetics (Figure 5H–I). Similar results were seen in MEFs in which YAP was immunoprecipitated by a phosphorylated p38 antibody only in lysates from the aligned pattern (Online Figure IIIC). On the random mimetics phosphorylated p38 still immunoprecipitated YAP and TEAD in cardiac fibroblasts lysates, but the level of binding was significantly reduced when normalized to the input (Figure 5G–I). These data suggest that collagen organization differentially regulates p38-YAP-TEAD interactions.

Figure 5. p38 is required for alignment-induced myofibroblast differentiation and co-localization of YAP and TEAD.

A Representative western blot and B densitometry analysis of p38 and GAPDH expression in cardiac fibroblasts cultured on flat, aligned and random biomimetics with (+) and without (−) Cre. Graph represent Mean±SEM, n=3 biological replicates, log transformed outcome and two-way ANOVA with the natural log of the variable of interest as the outcome variable and Bonferroni corrected pairwise comparisons were used for analysis: 1 vs p38^F/F,Flat, 2 vs p38^F/F,Aligned, 3 vs p38^F/F,Random. C Quantification of intensity of p38 staining in the nucleus in cardiac fibroblasts cultured on flat and aligned biomimetics. Graph represent Mean±SEM, n=36 biological replicates per group. Mann-Whitney test used. D Quantification of the percentage of p38^F/F cardiac fibroblasts positive for αSMA stress fibers on flat (231–451 cells studied per mouse within each condition), aligned (186–279 cells studied per mouse within each condition), and random biomimetics (253–380 cells studied per mouse within each condition) with and without Cre, n=3 biological replicates, and mixed effects logistic regression followed by Bonferroni corrected post hoc comparisons were made between groups: 1 vs p38^F/F,Flat, 2 vs p38^F/F,Aligned, 3 vs p38^F/F,Random. E Quantification of the nuclear colocalization of p38 and YAP and F p38 and TEAD in cardiac fibroblasts on flat and aligned biomimetics. Graph represent Mean±SEM, n=36 biological replicates per group. Mann-Whitney test used. G Representative western blot of lysates (Input) and phosphorylated-p38 immunoprecipated proteins (IP: Phos-p38) from these inputs, which were derived from cardiac fibroblasts on flat, aligned (align), and random (Rnd) biomimetics. H Quantification of the ratio of phospho-p38 immunoprecipitated TEAD to TEAD-input and I YAP to YAP-input. Here, log transformed outcome analysis and one-way ANOVA with the natural log of the variable of interest as the outcome variable and Bonferroni corrected pairwise comparisons were used for analysis. All graphs represent Mean±SEM, filled circles denote biological replicates, and ns = not significant.

p38-YAP-TEAD Signaling Underlies Topography-Dependent Regulation Of Fibroblast Phenotype

To determine if there are both p38 and topography-dependent effects on YAP activity, Western blot was used to analyze YAP expression and phosphorylation status in p38^F/F cardiac fibroblasts with and without Cre (Figure 6A). This experiment revealed a pattern-dependent increase in YAP protein expression that was significantly reduced in conditional p38 knockout cardiac fibroblasts (+ Cre lanes, Figure 6A–B). No changes in YAP gene expression were detected with loss of p38 (Figure 6C), suggesting that p38 is required for post-translational YAP stabilization. YAP nuclear-cytoplasmic shuttling was also examined in this context by 2 independent assays: Western blot for phosphorylated-YAP (Figure 6A & D), which keeps YAP cytosolic, and immunofluorescent imaging of YAP’s subcellular localization (Figure 6E). While aligned mimetics increased YAP protein expression, the levels of phosphorylated to total YAP were not statistically different from flat controls (Figure 6A & D). On the random patterns levels of phosphorylated to total YAP were significantly elevated suggesting that random collagen alignment promotes cytosolic YAP retention (Figure 6A & D). Notably, there was a p38-mediated decrease in phosphorylated YAP in all groups, which we believe is simply a function of the reduction in total YAP protein expression (Figure 6A). However, on aligned surfaces the ratio of phosphorylated to total YAP was higher in conditional p38 knockouts (p38^F/F + Cre, Figure 6D) suggesting that the limited amount of YAP in p38 knockouts remains cytosolic. Consistent with phosphorylated to total YAP ratios, nuclear YAP was increased in cardiac fibroblasts on aligned mimetics, which was significantly decreased with loss of p38 (Figure 6E). Deletion of p38 also decreased nuclear YAP on the flat mimetic likely because YAP requires p38 function for protein stability (Figure 6A–C). Since TEAD was also in complex with YAP and p38, nuclear TEAD levels were assessed, and unlike YAP there were no significant effects of p38 loss of function on nuclear TEAD (Figure 6F). There was, however, a significant enhancement in nuclear TEAD localization on aligned patterns (Figure 6F). To determine if collagen alignment and p38 alter YAP nuclear localization in vivo, cardiac tissue sections from wild-type (p38^F/F) and cardiac fibroblast-specific p38α knockout mice (p38^F/F + Tcf21^iCre) were examined using immunofluorescent imaging of cardiac sections 1 week after MI. Here the intensity of nuclear YAP was measured in resident cardiac fibroblasts that were identified by PDGFRα staining (Figure 6G). Using collagen alignment coefficients described in Figure 1C, nuclear YAP intensity was quantified in regions of aligned versus random collagen orientation (Figure 6G–H). There were no significant differences in nuclear YAP intensity in regions of aligned versus randomly oriented collagen, but consistent with in vitro findings loss of p38 significantly reduced nuclear YAP intensity (Figure 6G–H).

Figure 6. p38-YAP-TEAD interactions are necessary for alignment-induced myofibroblast differentiation.

A Representative Western blot of phosphorylated YAP (phos-YAP), total YAP, pan TEAD, and GAPDH expression. Here GAPDH served as a loading control. B Densitometry analysis of YAP protein levels normalized to GAPDH. C Change in YAP gene expression in cardiac fibroblasts on flat, aligned and random biomimetics with (+) and without (−) Cre. Values are calculated using the 2^-DDCt method and expressed relative to flat,p38^F/F condition. D Quantification of the ratio of phosphorylated YAP to total YAP protein levels in cardiac fibroblasts on flat, aligned and random biomimetics with (+) and without (−) Cre. For B, C, and D, graphs represent Mean±SEM, n=3 biological replicates,and log transformed outcome and two-way ANOVA with the natural log of the variable of interest as the outcome variable and Bonferroni corrected pairwise comparisons were used for analysis where: 1 vs p38^F/F,Flat, 2 vs p38^F/F,Aligned, 3 vs p38^F/F,Random, 4 vs p38^F/F-Cre,Flat, and 5 vs p38^F/F-Cre,Aligned. Quantification of intensity of E YAP and F TEAD staining in the nucleus of p38^F/F cardiac fibroblasts cultured on flat and aligned biomimetics with and without Cre. Graphs represent Mean±SEM, n=18 except for the Aligned-p38^F/F-Cre group in E where the n=14. 2-Way ANOVA and Tukey post hoc comparisons were made between groups, 1 vs flat,p38^F/F, 2 vs flat,p38^F/F-Cre, 3 vs aligned,p38^F/F. G Representative images and H quantification of nuclear YAP staining intensity (purple) in PDGFRa⁺ cardiac fibroblasts (green) in cardiac tissue sections from infarcted p38^F/F and p38^F/F-Tcf21^iCre mice. Nuclei are stained with Hoechst and staining intensity was subclassified by regions of aligned (coefficients ≥ 0.4 in Figure 1C) versus random (coefficients < 0.4 in Figure 1C) collagen organization. Scale bars=50mm. Graph represent Mean±SEM, n=3–5 biological replicates as denoted by the circles, 2-Way ANOVA and Tukey post hoc comparisons were made between groups, 1 vs flat,p38^F/F, 2 vs flat,p38^F/F-Cre, 3 vs aligned,p38^F/F. I Quantification of the percentage of fibroblasts positive for αSMA stress fibers on flat (177–504 cells studied per mouse), aligned (310–604 cells studied per mouse), and random biomimetics (320–358 cells studied per mouse) with and without dominant negative YAP (YAP^S94A) expression. Mixed effects logistic regression followed by Bonferroni corrected post hoc comparisons were made between groups. Fold change in J periostin (Postn) and K fibronectin EDa splice variant (FnEDa) gene expression in cardiac fibroblasts on flat, aligned, and random biomimetics with and without dominant negative YAP (YAP^S94A) expression. Values are calculated using the 2^-DDCt method and expressed relative to the flat,control condition. Here log transformed outcome and two-way ANOVA with the natural log of the variable of interest as the outcome variable and Bonferroni corrected pairwise comparisons were used for analysis. For I-K, graphs represent Mean±SEM, and statistically significant comparisons defined as: 1 vs flat,control, 2 vs aligned,control, and 3 vs random,control. L Quantification of the percentage of fibroblasts positive for αSMA stress fibers on flat (202–279 cells studied per mouse), aligned (310–604 cells studied per mouse), and random (204–402 cells studied per mouse) biomimetics with and without constitutively active YAP (YAP^S127A) expression. Graph represent Mean±SEM, n=3 biological replicates, mixed effects logistic regression followed by Bonferroni corrected post hoc comparisons were made between groups where: 1 vs flat,p38^F/F, 2 vs flat,p38^F/F-Cre, 3 vs flat,p38^F/F,YAP^S127A, 4 vs flat,p38^F/F-Cre,YAP^S127A, 5 vs aligned,p38^F/F, 6 vs aligned,p38^F/F-Cre, 7 vs aligned,p38^F/F,YAP^S127A, 8 vs aligned,p38^F/F-Cre,YAP^S127A, 9 vs random,p38^F/F-Cre, only significant relationships are shown.

To determine if YAP is required for alignment-dependent myofibroblast differentiation, cardiac fibroblasts were transfected with a dominant negative YAP (YAP^S94A) that is deficient in the TEAD binding-site⁴⁰. Expression of dominant negative YAP^S94A blocked myofibroblast differentiation induced by the aligned mimetics as evidenced by the drop in cardiac fibroblasts with αSMA⁺ stress fibers (Figure 6I) and upregulated myofibroblast ECM genes like Postn and FnEDa (Figure 6J–K). Cardiac fibroblasts cultured on random mimetics, which induce a mild but significant αSMA⁺ stress fiber phenotype and no change in ECM gene expression, were also rescued by YAP^S94A expression. In the opposing gain of function experiments overexpression of constitutively active YAP (YAP^S127A) in cardiac fibroblasts demonstrated that nuclear YAP activity is sufficient to induce αSMA⁺ stress fiber phenotype on flat and random patterns, but it failed to additively increase the number of myofibroblasts on the aligned mimetics (Figure 6L). Interestingly, a significant percentage of conditional p38 knockout cardiac fibroblasts, which are refractory to alignment-induced myofibroblast differentiation (p38^F/F + Cre, Figure 6L), spontaneously developed an αSMA⁺ stress fiber phenotype with the expression of constitutively active YAP independent of alignment cues, suggesting that YAP signaling mechanisms are downstream of p38 activity (Figure 6L). Notably, the random mimetic tempered YAP^S127A - mediated differentiation (Figure 6L) suggesting that disorganized topographical cues trigger alternative signaling pathways that partially repress the myofibroblast phenotype.

DISCUSSION

The spatial heterogeneity of collagen fiber organization in the infarcted myocardium broadly impacts myocardial biomechanics^{18, 41}, but also provides extracellular mechanical cues that locally regulate cardiac fibroblast fate⁴². Since fibroblasts in their differentiated myofibroblast state deposit and remodel the ECM during cardiac disease and injury, we aimed to understand how fibroblasts sense and integrate physical signals from the microenvironment into their fate decisions. Figure 7 models the signaling and functional gene expression changes in response to differential topographical cues. The data presented support our conclusion that aligned collagen organization induces myofibroblast differentiation, given ~60% of the population had αSMA⁺ stress fibers and a concomitant upregulation in myofibroblast-specific ECM genes like Postn and FnEDa^{4, 6}. This differentiation program was put into motion by increased tension sensation at focal adhesions (Figure 4F), which was transduced into myofibroblast gene expression via an enrichment of p38-YAP-TEAD interactions and activity. p38 stood out as the nodal transducer of the aligned topographical cues, because it was required to (1) elicit myofibroblast differentiation (Figure 5) and (2) post-translationally stabilize and enrich for nuclear YAP (Figure 6). Notably, YAP on its own was both necessary and sufficient for inducing myofibroblast gene programs (Figure 6). By contrast, disorganized ECM, which was mimicked by the random patterns, induced a mild cytoskeletal phenotype akin to myofibroblasts but with downregulated myofibroblast specific ECM genes (Figure 7). The mild cytoskeletal phenotype was not induced by a change in focal adhesion tension sensation, but it did require p38 signaling (Figure 5). In addition, loss of YAP’s ability to bind TEAD failed to block the mild cytoskeletal phenotype suggesting that p38-YAP-TEAD interactions aren’t transducing signals from disorganized ECM topography. Yet unknown is the identity of receptors and ligands upstream of p38 that contribute to these outcomes and how this pathway is down-regulating matrix gene expression.

Figure 7. Signaling model of ECM topography-dependent regulation of programmed cardiac myofibroblast differentiation.

Dashed arrow represents previously identified interactions between signaling nodes.

The coupling of both physical and chemical cues was also studied by treating cardiac fibroblasts cultured on our topographical mimetics with TGFb. Here, TGFb failed to additively increase the cytoskeletal myofibroblast phenotype suggesting that aligned structural cues and TGFb converge on a common intracellular pathway that regulates the cytoskeletal phenotype. Given previous work demonstrating that TGFb requires p38 to induce myofibroblast differentiation^{8, 10}, we postulated in our model that both alignment-signaling and TGFb are converging on p38 to elicit the cytoskeletal remodeling (Figure 7). TGFb additively increased the expression of myofibroblast-specific ECM genes suggesting there is a TGFb-sensitive pathway not regulated by p38-YAP-TEAD signaling that boosts the ECM component of the myofibroblast phenotype in parallel. Possible candidate factors are SMAD 2 and 3, which were not studied here but shown previously to regulate cardiac fibroblast to myofibroblast differentiation in vivo ^{43, 44}. Interestingly, cardiac fibroblasts on random mimetics responded to TGFb in which both myofibroblast-linked cytoskeletal and ECM components were upregulated, but the magnitude of the overall response was dampened relative to the levels achieved on both flat and aligned topographical mimetics (Figure 3).

The transition from ordered to disorganized collagen structure is a hallmark of disease progression and wound healing that transcends tissues and disease etiologies as these organizational changes are also observed in tumor microenvironments^{29, 45–48}. Here, we observed that the infarcted mouse myocardium has regional variation in collagen organization in which the scar’s core has more disorganized and randomly oriented collagen fibers versus the border zone which has more fibers aligned in parallel (Figure 1A–C). Similar regional variations in collagen topography have been shown in rat MI models^{41, 46, 49}. Interestingly, chronic MI in pig, dog, and human also display regional variation in collagen topography but with disordered fibers in the border zone and highly aligned collagen structure in the core of the infarct scar^{41, 46, 49}. Despite these differences in the spatial location of aligned collagen fibers, higher myofibroblast density has been observed in regions where collagen fibers are in parallel alignment in several species^{14, 17}, as observed and quantified here in our mouse MI model (Figures 1D–F). These in vivo findings further support that aligned ECM topography is a positive regulator of programmed myofibroblast differentiation. This regional variation in collagen alignment persists for weeks and even months after infarct and is consistently prevalent in an injured or diseased tissue^{41, 46}, suggesting that aligned collagen organization positively reinforces the maintenance of an activated fibroblast state. A recent study using single-cell RNAseq and lineage tracing approaches demonstrated that fibroblasts undergo two cell state changes as the heart repairs and remodels over a 4-week period after MI⁵⁰. It is unclear how regional heterogeneity in collagen structure impacts the second osteogeneic/cartilage cell state transition, but these topographical mimetics could be utilized in long-term cultures in combination with profibrotic molecules to further investigate time-dependent regulation of fibroblast fate and function. Moreover, the sensitivity of cardiac fibroblasts to these topographic cues may help explain in part fibroblast heterogeneity within an injured tissue^{12, 13, 45}, highlighting the need to examine relationships between spatial location and cell state, particularly in organs with anisotropic mechanical properties like the heart^{30, 51, 52}.

Fibroblasts have been established as mechanically sensitive cells but how different physical cues are sensed and transduced into cell state changes are poorly defined^{6, 53, 54}. Tension-sensation measured in vinculin demonstrated that there was higher tension at focal adhesions in cardiac fibroblasts cultured on aligned mimetics when compared to other surfaces which we believe underlie an adaptive growth in focal adhesion number and size (Figure 4). The intracellular signal transducers uncovered the unexpected finding that aligned topographical cues initiated p38-YAP-TEAD interactions (Figures 6). Independently, p38 and YAP were required for alignment-dependent myofibroblast differentiation (Figures 5 & 6) and because p38 is required for YAP protein stabilization (Figure 6) the pathway was modeled with p38-YAP-TEAD complex as the transducer of aligned ECM cues (Figure 7). Notably, after MI cardiac fibroblast-specific p38 knockouts had reduced YAP staining in resident cardiac fibroblasts suggesting that p38-dependent regulation of YAP protein stability holds true in vivo (Figure 6G–H). Previous studies have identified YAP’s involvement in transducing mechanical cues into cell fate change and found that unrestrained YAP transcriptional activity by loss of Lats1/2-mediated repression spontaneously induces cardiac fibroblast to myofibroblast differentiation and fulminant fibrosis in mice^{36, 55, 56}. Our findings that overexpression of a constitutively nuclear YAP is sufficient to induce a myofibroblast phenotype further support these findings (Figure 6L). The novel findings presented here include: (1) YAP-mediated myofibroblast transcriptional activity acts downstream of p38, and (2) YAP is post-translationally regulated by p38 independent of the Lats1/2-Hippo pathway³⁶. Moreover, these topography-induced interactions between p38-YAP-TEAD extend the network of chemical and mechanical signals that converge on p38 MAPK to induce a myofibroblast phenotype^{4, 10}. The requirement for p38 MAPK signaling in programmed myofibroblast differentiation and scar formation has been verified in vivo in a mouse model with cardiac fibroblast specific loss of p38a¹⁰, but until now it was unclear whether p38 MAPK signaling also transduces structural cues from the ECM. In cancer cells, hyperosmotic stress, which compresses a cell, induces p38-TEAD interactions that pulls TEAD out of the nucleus causing cytosolic retention of YAP and inhibition of YAP-TEAD transcription³⁸. While tension sensation at the focal adhesions in fibroblasts on the random mimetics was not statistically different from flat, the lower FRET efficiency and cellular morphology indicate disorganized topographical cues in vitro and in vivo compress the cell similar to hyperosmolarity (Figures 1G & 2H–I). Indeed, YAP phosphorylation was increased by random mimetics suggesting there is more YAP in the cytosol (Figure 6D) but with reduced TEAD-p38 interactions, which should reduce myofibroblast gene transcription (Figure 5G–I). The exact opposite occurred in cardiac fibroblasts on the aligned mimetics in which there was increased tension at the focal adhesions and enhanced colocalization of p38-YAP-TEAD in the nucleus (Figure 5). Taken together these data suggest that p38-YAP-TEAD interactions are highly sensitive to physical signals that stretch or compress a cell. Unfortunately, we were unable to resolve collagen topography-dependent effects on YAP subcellular localization in cardiac fibroblasts in infarcted mouse hearts, possibly due to resolution limitations our imaging approach (Figure 6G–H).

Collectively, this study provides new insights into the mechanisms by which fibroblasts respond to regional heterogeneity in collagen fiber organization associated with myocardial infarction. By engineering biomimetics that recapitulate the heterogenous organization of the heart’s ECM and combining them with genetic engineering, this study added new nodes to the network regulating cardiac fibroblast fate during post-MI remodeling. These results stress the importance of investigating context-dependent fibroblast function especially pertaining to spatial location and local physical cues within the extracellular environment, and they suggest that redesigning matrix architecture could be used as therapeutic strategy to prevent the maintenance of maladaptive fibrotic cell states.

NOVELTY AND SIGNIFICANCE.

What Is Known?

• Myocardial infarction creates spatially distinct differences in collagen.

• Collagen organization can modulate cell state.

• Heterogeneous cardiac fibroblast cell state changes contribute to myocardial-infarction induced remodeling of the extracellular matrix.

What New Information Does This Article Contribute?

• Using nanofabricated biomimetics of myocardial infarction-induced changes in collagen alignment, we demonstrated that aligned topographical cues induce cardiac fibroblast to myofibroblast state transitions, a maladaptive response.

• Aligned collagen biomimetics increase cardiac fibroblast focal adhesion tension sensation, which is required for topography-dependent cell state transitions.

• Aligned collagen cues also enrich for p38-YAP-TEAD interactions and these signals are required for cardiac fibroblast to myofibroblast state transitions.

Cardiac fibroblast to myofibroblast state transitions are essential for maladaptive extracellular matrix remodeling and scarring after a myocardial infarction. Understanding the signals that regulate this state transition is essential for developing approaches for scar regression, a process that still lacks a tractable therapeutic despite its role in accelerating heart failure. Heterogeneous changes in collagen organization is a key event following myocardial infarction, yet previous studies failed to explore how local environmental cues impact these cardiac fibroblast fate changes. This study provides new mechanistic insight into how structural alignment cues from the extracellular matrix regulate cardiac fibroblast fate and function. Through engineering biomimetics of myocardial infarction-dependent differences in regional collagen alignment, we uncovered the novel findings that aligned collagen topography initiates cardiac fibroblast to myofibroblast differentiation and that interactions between p38, YAP and TEAD are essential for transducing these alignment cues into programmed myofibroblast differentiation. Also, we provide the first measurement of topography-dependent focal adhesion tension sensation, which transduces topographical cues upstream of p38, YAP, and TEAD. These result underscore the importance of local physical cues within the extracellular environment in regulating cardiac fibroblast function and suggest that matrix architecture needs to be considered when designing therapeutic strategies.

Nonstandard Abbreviations and Acronyms

MI

Myocardial infarction

ECM

extracellular matrix

αSMA

α-smooth muscle actin

UV

ultraviolet light

Tam

tamoxifen

NOA76

Norland optical adhesive-76

SEM

scanning electron microscopy

FnEDa

EDa splice variant of fibronectin

FRET

Förster Resonance Energy Transfer