Mechanical regulation of fibroblast migration and collagen remodelling in healing myocardial infarcts

After myocardial infarction, fibroblasts infiltrate the necrotic myocardium, deposit collagen, and remodel collagen, forming scar tissue. Mechanical, structural, and chemical guidance cues have all been shown to regulate alignment of fibroblasts and collagen in vitro. We developed a computational model of infarct healing with parameters closely tied to published data, ran simulations with different combinations of guidance cues, and successfully predicted measured collagen structures in rat infarct scars. We determined that a mechanical guidance cue was the most important determinant of collagen alignment in healing myocardial infarcts. Since anisotropy of infarct scar tissue is an important determinant of cardiac function, our model represents a potentially powerful tool for testing the effects of post-infarction therapies on scar anisotropy and designing new therapies that modulate infarct healing by controlling the mechanical stimuli acting on the infarct region or the response of fibroblasts to those stimuli. Abstract Effective management of healing and remodelling after myocardial infarction is an important problem in modern cardiology practice. We have recently shown that the level of infarct anisotropy is a critical determinant of heart function following a large anterior infarction, which suggests that therapeutic gains may be realized by controlling infarct anisotropy. However, factors regulating infarct anisotropy are not well understood. Mechanical, structural and chemical guidance cues have all been shown to regulate alignment of fibroblasts and collagen in vitro, and prior studies have proposed that each of these cues could regulate anisotropy of infarct scar tissue, but understanding of fibroblast behaviour in the complex environment of a healing infarct is lacking. We developed an agent-based model of infarct healing that accounted for the combined influence of these cues on fibroblast alignment, collagen deposition and collagen remodelling. We pooled published experimental data from several sources in order to determine parameter values, then used the model to test the importance of each cue for predicting collagen alignment measurements from a set of recent cryoinfarction experiments. We found that although chemokine gradients and pre-existing matrix structures had important effects on collagen organization, a response of fibroblasts to mechanical cues was critical for correctly predicting collagen alignment in infarct scar. Many proposed therapies for myocardial infarction, such as injection of cells or polymers, alter the mechanics of the infarct region. Our modelling results suggest that such therapies could change the anisotropy of the healing infarct, which could have important functional consequences. This model is therefore a potentially important tool for predicting how such interventions change healing outcomes. Each year, approximately 600,000 Americans experience a new myocardial infarction and 300,000 Americans experience a recurrent infarction (Lloyd-Jones et al. 2010). Most patients survive the initial event, making management of post-infarction healing and remodelling a high priority (Lloyd-Jones et al. 2010). Current revascularization procedures and pharmacological therapies seek to limit infarct extension, infarct expansion and adverse remodelling of remote myocardium (Sutton & Sharpe, 2000). Additionally, researchers are investigating regenerative therapies that may one day be able to restore contractile function to the infarct region (Forrester et al. 2003). These treatments are all motivated, at least in part, by the understanding that mechanical dysfunction is a key driver of the progression toward heart failure. Treatments intended to reduce mechanical dysfunction of the heart or regenerate myocardium may be improved by controlling the anisotropy of infarct tissue. Anisotropy is an important feature of normal myocardium (Costa et al. 2001), which led us to study its importance in healing infarcts. We have recently shown that the level of infarct anisotropy is a critical determinant of heart function following a large anterior infarction (Fomovsky et al. 2011). However, factors regulating infarct anisotropy are not well understood. Mechanical, structural and chemical environmental cues have all been shown to regulate alignment of fibroblasts and collagen in vitro (Dickinson et al. 1994; Lee et al. 2008; Melvin et al. 2011), but understanding of fibroblast behaviour in the complex environment of a healing infarct is lacking. A better understanding of how fibroblasts integrate these cues as they deposit and remodel extracellular matrix in a healing infarct is needed in order to develop interventions that modify infarct scar anisotropy for therapeutic benefit. Fibroblasts are the cell type primarily responsible for creating the microstructure, that is the density, orientation, thickness and crosslinking of collagen fibre bundles, that explains the mechanical properties of infarct scar tissue (Sacks, 2003; Souders et al. 2009). Fibroblasts assemble collagen fibres, secrete proteolytic and crosslinking enzymes, and apply traction forces that physically deform collagen fibre networks (Thomopoulos et al. 2005; Kadler et al. 2008). Fibroblasts can deposit collagen fibres oriented along their long axis and rotate nearby collagen fibres toward co-alignment with their long axis (Petroll et al. 2003; Canty et al. 2004). Consequently, any cue for fibroblast alignment may cause collagen alignment. In vitro studies have shown that fibroblasts align in response to uniaxial stretch (Neidlinger-Wilke et al. 2001; Pang et al. 2011), micropatterned ridges or aligned fibres (Dickinson et al. 1994; Wang et al. 2003), and chemokine gradients (Knapp et al. 1999; Melvin et al. 2011). These environmental guidance cues have been shown to cause fibroblasts to deposit anisotropic extracellular matrices or remodel isotropic matrices into anisotropic matrices (Wang et al. 2003; Thomopoulos et al. 2005). Several studies have reported differing anisotropic properties of infarcts (Whittaker et al. 1989; Gupta et al. 1994; Holmes et al. 1997; Fomovsky & Holmes, 2010). It is difficult to make inferences about the determinants of infarct scar anisotropy due to the many variables that change from study to study, such as animal model, infarct location, infarct size and infarct shape, but there is some evidence supporting a role for chemical, structural and mechanical cues. In a computational model of skin wound healing, McDougall et al. (2006) showed that wound geometry can regulate collagen alignment by determining the spatial pattern of chemokine gradients. Zimmerman et al. (2000) studied the development of collagen alignment in healing pig infarcts and concluded that the initial matrix alignment in native myocardium determines the alignment of collagen in replacement scar. Recently, we performed a cryoinfarction study that independently varied the location and shape of infarcts in rats and found that infarct scar collagen alignment was correlated with the deformation pattern of the infarct, which depended on infarct location but not shape (Fomovsky et al. 2012). On the anterior, mid-ventricular wall of the heart, infarcts stretched primarily in the circumferential direction and developed scar tissue with circumferentially aligned collagen fibres. On the apex of the heart, infarcts stretched in the circumferential and longitudinal directions and developed scar tissue with a random collagen fibre orientation. This suggested that mechanical deformations regulate collagen alignment in infarct scar. In view of the experimental evidence suggesting a role for mechanical, structural and chemical cues in determining the structural properties of healing infarcts and our limited insight into how fibroblasts integrate and respond to combinations of these cues, we developed an agent-based model of infarct healing in order to test hypotheses about how environmental cues guide fibroblasts to align collagen in healing infarcts. Agent-based modelling is well-suited to addressing problems of this kind, which investigate how the behaviours of individual agents within a system give rise to emergent properties of the whole system (Simpson et al. 2007; Chavali et al. 2008; Groh & Wagner, 2011). The work presented in this paper was primarily inspired by the dermal wound healing model published by McDougall et al. (2006). Our strategy was to build an agent-based model that incorporated experimentally measured fibroblast behaviours, environmental conditions and parameter values in order to predict the structural properties of the collagen matrix formed during infarct healing. We then independently perturbed the mechanical, structural and chemical guidance cues in the model and determined the importance of each cue for predicting measured collagen fibre structures in healing infarcts. Understanding the role of each cue is important for predicting how therapies that intentionally or unintentionally modify one or some combination of these cues will affect infarct scar and cardiac and matrix The agent-based model of infarct healing a of myocardium to the as a and This was into or each a collagen fibre and a fibre the tissue The fibre all of myocardium that collagen normal myocardium is we important to the influence of the not the collagen fibres, on fibroblast into from to The initial fibre to the of normal The within a or in the the infarct region. of collagen fibre deposition and was the model of fibre was for within the infarct, which the influence of the on fibroblast alignment. of the agent-based model of infarct healing of a scar after of healing, collagen fibres of a model the cryoinfarction collagen and collagen fibre orientation. of fibroblast agents that have the infarct region by and of fibre Each in the and of of the fibre Fibroblasts Fibroblasts as of Fibroblasts to environmental to on experimental of fibroblast chemokine and fibre alignment. collagen fibre collagen fibre and matrix for each cell of the the cell and the of the Fibroblasts could within the model they not to was determined by the combined influence of the cell orientation, chemokine anisotropy and fibre alignment. The model the fibroblast making model performed The model was by and at each all of the the order of the fibroblasts was a fibroblast its was from the a fibroblast its cell a new fibroblast was there was to the a fibroblast its for a there cells its A fibroblast then remodel the matrix within its by matrix fibres, collagen fibres, and collagen fibres. making fibroblast these and of determining fibroblast each guidance cue was as a The and in order to a the combined influence of all of the cues on the fibroblast orientation. the and of the the shape of a normal The new fibroblast by the in the was by a random from a set of this The model healing of a infarct in order to and of which approximately to model important of this model to the of the cells to the of the chemokine and the of by a for a cell of we that at least cells along the of the chemokine the chemokine we a of as the of the model infarct to the experimental infarct we that at least within the of the cell and a of Since we the model the fibroblast which was the parameter parameter with that are some combination of and of fibroblasts into the infarct, was by a to as the of the of the model infarct to the of the experimental The model a of parameters The of these the model environment and initial and they therefore by the conditions of the to be the which determined the behaviour of the model, parameters from data and and parameters by published and model results to the published data, as in the of which important for the model results and factors for the mechanical and chemical guidance cues set to the structural cue which was from experimental The cue structural cue and by behaviour to published et al. measured the of fibroblasts in collagen with and high of collagen fibre alignment, for each the data to the random model of cell and reported random for and We used our agent-based model to of fibroblasts in or aligned collagen performed the by et al. and and to the random from the simulations to the published with different mechanical conditions or uniaxial and different oriented oriented in order to the of conditions in the cryoinfarction study performed by our (Fomovsky et al. 2012). In order to test the of the mechanical guidance cue on the predicted collagen a of simulations with the by the for the mechanical cue to In order to test the of the structural guidance cue on the predicted collagen simulations with the fibres of the initial matrix oriented a level of circumferential alignment to that of native myocardium, or aligned in the circumferential In order to test the of matrix during the initial infarction, simulations with the initial matrix set to or the of normal in order to the structural and mechanical cues at infarct with the fibre of the initial matrix from to to the circumferential to the fibre the thickness of normal to of healing of of collagen alignment, as different model infarct collagen at the to in measurements after cryoinfarction in rats (Fomovsky et al. 2012). of collagen fibre and of collagen alignment to in measurements after of an of the in Holmes et al. of infarct healing within the infarct fibroblast and and collagen deposition The are and after The and structural cue factors and the and the collagen and and determined by published experimental et al. for fibroblasts in or aligned collagen The behaviour was determined by the and such that could be The values and the to the published We measured collagen and following in the rat (Fomovsky & Holmes, 2010). The values and the to the published collagen of model to published fibroblast behaviour in collagen et al. measured random for and of fibroblasts in collagen with or aligned collagen fibres. We of the model parameters and by behaviour to the published of model to published infarct collagen Fomovsky & Holmes measured collagen at several after creating infarcts in rats We of the model parameters and by a to the published the parameter values, the agent-based model was able to the data The chemical guidance cue had on the predicted collagen alignment for the of in our study, which was with the experimental simulations oriented or circumferentially oriented circumferentially oriented initial as in normal myocardium at and uniaxial in the circumferential as measured in anterior, mid-ventricular The different different spatial of the chemokine However, the model predicted in collagen alignment that well within the of the experimental measurements from the mid-ventricular with the of an of infarct shape on collagen alignment in our study We then for conditions the model predict a in the of collagen alignment circumferentially and infarcts. The infarcts an of at least or the fibroblasts to be at least to the chemical cue in order for the of the infarct to affect the of collagen alignment of and infarcts uniaxial with initial matrix alignment as found in native myocardium the model predicted in the of collagen alignment as the of the collagen fibre of a oriented infarct infarct and a circumferentially oriented infarct This was with experimental data from our cryoinfarction study (Fomovsky et al. 2012). by the of the infarcts to or the chemical cue the model predicted a of the of the infarcts on the of collagen alignment. The initial matrix had on the predicted collagen alignment after several of healing, which was with data from experimental of model simulations the initial matrix aligned in the circumferential direction or and uniaxial in the circumferential as measured in mid-ventricular The initial matrix alignment had an on the alignment of collagen at but the for the different initial structures the predicted collagen fibre within a of the experimental of the mid-ventricular data, of the of alignment of the initial matrix the initial matrix was to the of normal myocardium, the model predicted alignment of collagen in the infarct at but this after of healing the model was used to infarct healing at different by different initial fibre from to the predicted infarct collagen structures and the predicted collagen fibre the thickness of the infarct was the initial the of collagen alignment was at at the and These with the in pig infarcts and which uniaxial stretch in the circumferential of infarcts uniaxial with different initial matrix alignment or the model predicted that infarcts with initial matrix that was or aligned have different collagen structures at but to the of collagen alignment infarcts with initial matrix that was to or the of normal myocardium had different collagen alignment in the and then by of the of alignment of the initial matrix the predicted collagen fibre within of the experimental for mid-ventricular (Fomovsky et al. 2012). of infarcts uniaxial with fibre of the initial matrix from at the to at the to a of in a healing infarct the model predicted that collagen fibre on the circumferential the model collagen from rat therefore predicted a fibre in infarcts in myocardium a was in fibre measured and after infarction in Holmes et al. the model suggested that fibre after of healing, the model predicted alignment at the direction of stretch and the initial matrix and alignment at the initial matrix most from the direction of stretch a in was reported in pig infarcts after A mechanical guidance cue was to alignment of collagen in infarct which was with the experimental cryoinfarction circumferentially oriented initial and uniaxial in the circumferential direction or each mechanical model simulations with on or the pattern of stretch was the aligned matrix was by oriented collagen, the pattern of stretch in the infarct was scar formed with aligned collagen fibres The predicted collagen fibre after of healing with those for that stretch and mid-ventricular that uniaxial stretch In was although the model results for the of stretch the model to the collagen fibre alignment in the stretched of infarcts with uniaxial or and with initial matrix alignment as found in native myocardium the model predicted that long alignment of collagen in a healing infarct a uniaxial mechanical which was with experimental data from our cryoinfarction study and at with on the model predicted collagen fibre to those measured in or uniaxial the aligned collagen for the of uniaxial could not be studies have proposed that mechanical, structural, or chemical cues could regulate the anisotropy of infarct scar tissue. We developed an agent-based model that accounted for the combined influence of these cues on fibroblast alignment, collagen deposition and collagen remodelling. We pooled published experimental data from several sources in order to determine parameter values in the model and then model collagen alignment measurements from a set of recent cryoinfarction that independently varied infarct shape and pattern of stretch in a animal model (Fomovsky et al. 2012). The model predicted collagen structures in with the experimental measurements all We then independently perturbed the mechanical, structural, and chemical cues in the model in order to determine how each cue the predicted collagen chemokine gradients and pre-existing matrix structures spatial and of collagen in a mechanical cue was critical to the collagen alignment experimentally in healing infarcts. McDougall et al. (2006) proposed that the pattern of collagen alignment in dermal could be by alignment of fibroblasts in the direction of the chemical by within the to this infarcts have different chemokine and different of collagen alignment. We used our model to healing of a oriented infarct, infarct and circumferentially oriented infarct, and found a of infarct shape on the collagen alignment this was in with the model results of McDougall et was with our cryoinfarction The model some insight as to this is the The predicted chemokine was the of the infarct and at the infarct Consequently, the influence of the chemical cue on collagen fibre alignment was limited to the region the infarct the the of the chemical guidance in model infarcts with high this influence the collagen fibre alignment we this study to the results that chemokine gradients some in the collagen of which to be the of in this Zimmerman et al. (2000) proposed that the of collagen fibres in healing infarcts could be by the of the matrix native myocardium, some of which and as a for alignment of fibroblasts and collagen during scar We used our model to healing of an infarct with an initial matrix that was aligned or and found that the influence of the initial matrix on the predicted collagen alignment the of healing We used our model to healing of an infarct with and found that an aligned matrix was to collagen alignment These results with our cryoinfarction which found that alignment of collagen fibres in infarct scar was in infarcts that uniaxial The model some insight as to the aligned matrix to long collagen alignment in healing infarcts. In the model, the of fibroblast had an of which that fibroblasts not in the direction of the aligned matrix fibres. Since fibroblasts collagen fibres along and collagen fibres toward their orientation, the of alignment of the matrix was therefore by the aligned fibroblasts have an to align with the initial structural In fibroblasts can change the matrix then be a guidance we recently reported that collagen fibre in that stretch random by after in the rat (Fomovsky & Holmes, suggesting that our model the influence of pre-existing matrix on collagen fibre in the scar. This could the that some of the pre-existing matrix is in the following infarction et al. with this we infarcts with a initial matrix density, the model predicted a in the of collagen alignment in the after infarction the of structural cues could be high in our However, this the of experimental data with the predicted of infarct collagen alignment, which depended on the of the structural cue In our cryoinfarction study (Fomovsky et al. we found that the location of the infarct determined its pattern of and this deformation pattern was correlated with collagen fibre alignment. we healing of an infarct with we found that the mechanical environment or determined the healing infarct formed an aligned or This was with the experimental that uniaxial stretch developed with collagen fibres aligned in the direction of that stretch developed with aligned collagen fibres. was the collagen alignment in the mid-ventricular could not be The model some insight as to the mechanical cue was the most important determinant of infarct collagen alignment. the chemical which its influence to the infarct the chemokine is the mechanical cue influence the of the the structural which is in the of any supporting cues, the mechanical cue is by the acting on the infarct shown to be for after in rats (Fomovsky & Holmes, 2010). Our of data in this paper was a set of rat cryoinfarction performed in our the model of infarction many such as of infarct shape and location, tissue by is not to by and there may be important in the scar We therefore performed to published data on collagen fibre in healing myocardial infarcts by the published data well with our model infarcts from of an of the in stretched circumferentially and developed circumferentially aligned collagen fibres et al. infarcts from in rats stretched and developed aligned collagen fibres (Fomovsky & Holmes, 2010). published data on infarct collagen fibre we that as in our prior studies in et al. have reported of the collagen fibre in healing infarcts in normal myocardium (Whittaker et al. 1989; et al. We therefore infarct healing at different and the predicted of fibre and to our experimental data from infarcts in in which the initial fibre from to but the infarct uniaxial stretch of at all the that many of the model parameter values from experimental studies of the model the in the pig data well and an for the the of the initial matrix and the direction of stretch the healing infarcts developed aligned collagen oriented in the circumferential the and the initial matrix and mechanical environment cues, the of the initial matrix that by the collagen fibre in these to circumferential but the of alignment was at The of these infarction studies are with the model and the cryoinfarction study, and the of the from the The agent-based model presented in this work is a in the of computational modelling to infarct healing. of infarct healing have used of to predict the of chemokine and collagen in the infarct (Wang et al. et al. 2011). wound healing have used agent-based or to predict the spatial and of cell and extracellular matrix et al. 2007; et al. 2009). Several of these developed to predict structural remodelling of the extracellular which our modelling et al. et al. 2008; Groh & Wagner, 2011). However, our model is the to the combined influence of mechanical, structural, and chemical guidance cues on fibroblast behaviour to predict collagen fibre alignment in scar tissue formed after myocardial we tied behaviours and parameter values in our model to experimental which the model to be to in Several of the model of infarct healing that are and our model represents a the of the Consequently, in the model, fibroblasts can the infarct by in this in cells infiltrate the infarct from the and from cells that the infarct the or survive the initial Our model a of cell that is to in for cells to the of the necrotic which our but studies on the pattern of fibroblast into the infarct are the model the of and fibre in normal we the of the of initial fibre gradients in not they have been shown to within after infarction et al. and at least of healing et al. Consequently, we the infarct as a of each with a different initial matrix but all with the mechanical our model represents infarct as In the due to and infarct and due to collagen deposition will determine how infarct Our model was on the that we change in in our study of healing rat infarcts (Fomovsky & Holmes, 2010). our model represents the chemokine as a In many are at the of an infarct in a complex spatial and Since is not how all of these factors regulate fibroblast we to model a the of fibroblast and collagen remodelling. the of chemokine is to be important for making about infarct healing at we not to be important after fibroblasts have the our model represents the matrix traction forces that fibroblasts as of collagen fibres toward co-alignment with the The which is an important determinant of the predicted of collagen alignment, is an parameter in the We a of a and found of the model with experimental data after of healing, but data at will be needed in order to better a for the Our results suggest that although chemokine gradients and pre-existing matrix structures have important effects on collagen organization, a mechanical cue is critical for the development of collagen alignment in infarct scar. It is important to the effects of these cues interventions are to be that seek to cardiac function after infarction by controlling the anisotropy of infarct scar. many proposed therapies for myocardial infarction, such as injection of cells or et al. of for structural et al. and et al. alter the mechanics of the infarct region. Our modelling results suggest that such therapies could unintentionally change the anisotropy of the healing infarct, which could have important functional consequences. This model represents a potentially powerful tool for predicting how such interventions change healing outcomes. This work is for tissue and for of any tissue anisotropy is an important determinant of tissue function, such as and The work was performed in the at the of The on all of the The for and for the experimental data used for in this was by an and of