Surgical reinforcement alters collagen alignment and turnover in healing myocardial infarcts

Abstract

Previous studies have suggested that the composition and global mechanical properties of the scar tissue that forms after a myocardial infarction (MI) are key determinants of long-term survival, and emerging therapies such as biomaterial injection are designed in part to alter those mechanical properties. However, recent evidence suggests that local mechanics regulate scar formation post-MI, so that perturbing infarct mechanics could have unexpected consequences. We therefore tested the effect of changes in local mechanical environment on scar collagen turnover, accumulation, and alignment in 77 Sprague-Dawley rats at 1, 2, 3 and 6 wk post-MI by sewing a Dacron patch to the epicardium to eliminate circumferential strain while permitting continued longitudinal stretching with each heart beat. We found that collagen in healing infarcts aligned parallel to regional strain and perpendicular to the preinfarction muscle and collagen fiber direction, strongly supporting our hypothesis that mechanical environment is the primary determinant of scar collagen alignment. Mechanical reinforcement reduced levels of carboxy-terminal propeptide of type I procollagen (PICP; a biomarker for collagen synthesis) in samples collected by microdialysis significantly, particularly in the first 2 wk. Reinforcement also reduced carboxy-terminal telopeptide of type I collagen (ICTP; a biomarker for collagen degradation), particularly at later time points. These alterations in collagen turnover produced no change in collagen area fraction as measured by histology but significantly reduced wall thickness in the reinforced scars compared with untreated controls. Our findings confirm the importance of regional mechanics in regulating scar formation after infarction and highlight the potential for therapies that reduce stretch to also reduce wall thickness in healing infarcts.

NEW & NOTEWORTHY This study shows that therapies such as surgical reinforcement, which reduce stretch in healing infarcts, can also reduce collagen synthesis and wall thickness and modify collagen alignment in postinfarction scars.

INTRODUCTION

Every year, over one million people in the United States suffer a heart attack, or myocardial infarction (MI). Most patients survive the initial event, but infarction places them at higher risk of heart failure and other serious complications (3). Over the first few weeks after MI, heart muscle that was lost is gradually replaced by stiff, noncontractile collagenous scar (12, 21). The MI often triggers remodeling in the surrounding myocardium, leading to dilation of the left ventricle (LV) and increased risk of heart failure (32). The scar may also continue to stretch and thin, promoting more dilation and further increasing the risk of heart failure (22). The mechanical properties of the healing infarct scar and the geometric changes because of postinfarction remodeling both play critical roles in determining long-term outcomes in patients who survive the initial MI. Thus, a number of therapies such as biomaterial injection, synthetic or tissue-engineered patches, and cardiac restraint devices aim to alter the mechanics of the healing infarct and/or ventricle (5, 11, 20, 21, 30).

Interestingly, emerging evidence suggests that regional mechanics are also an important determinant of both scar formation and LV remodeling. Zhou et al. (34) showed that mechanically unloading infarcted hearts by heterotopic transplantation significantly reduced scar collagen volume while increasing matrix metalloproteinase (MMP)-9 activity. Fomovsky et al. (9) created cryoinfarcts of different sizes and shapes in different locations on the rat LV and showed that collagen alignment in the resulting scars correlated with local strain patterns. Thus, therapies that alter mechanics may trigger unexpected long-term effects by perturbing this mechanical feedback loop.

In the present study, we sought to determine the impact of perturbing local mechanics (stretch/stress) on collagen alignment and turnover in post-MI scar. In particular, we tested two hypotheses suggested by the studies cited above: 1) that mechanical environment regulates collagen alignment in healing infarcts and 2) that mechanical environment regulates collagen degradation in healing infarcts. Through model-driven experiments that altered the direction of stretch imposed on healing infarcts, we show that local tissue mechanics are the primary factor governing the alignment of collagen post-MI. By measuring the levels of peptides associated with collagen synthesis and degradation, we found that contrary to our hypothesis, both collagen synthesis and collagen degradation were significantly altered by changes in regional mechanics. At the tissue level, these changes in turnover translated to alterations in scar thickness rather than in collagen area fraction. The results of these studies may be useful in anticipating the effects on scar healing of therapies that alter regional mechanics, such as local and global restraint and injection of biomaterials.

METHODS

Model-based design of experiments.

To select the mechanical perturbation that would provide the best test of the hypothesis that mechanical environment regulates collagen alignment in the healing postinfarction scar, we simulated various options for mechanical reinforcement using an agent-based model (ABM) of myocardial infarct healing published previously by our group (24); the MATLAB code for this model is available for download at https://simtk.org/projects/infarctabm or http://bme.virginia.edu/holmes/downloads/index.html. The ABM modeled a midwall section of the myocardium as a two-dimensional rectangular space containing a circular infarct, divided into 2.5  × 2.5-µm patches. The model was initially seeded with ~2% (by area) randomly populated fibroblasts and 3% circumferentially aligned collagen as observed in the healthy myocardium (9). After MI, fibroblasts in the circular wound region apoptosed and were gradually replaced by cells infiltrating from the wound margins and proliferating. Fibroblasts deposited and remodeled collagen to generate the postinfarction scar. We prescribed 5% equibiaxial stretch to simulate the baseline mechanics of an anteroapical infarct in the rat (8) and ran additional simulations of mechanical unloading in the circumferential (circumferential strain reduced to 0), longitudinal (longitudinal strain reduced to 0), or both directions (no strain). The predicted orientations of collagen fibers in the wound region were plotted as histograms ranging from −90 to 90° with a bin size of 5°, weighted by the collagen area fraction.

Acute sonomicrometry study.

To validate our surgical reinforcement approach for altering strain in healing rat infarcts, we conducted a pilot study in three animals using sonomicrometry. All experiments were approved by the University of Virginia Institutional Animal Care and Use Committee. Adult male Sprague-Dawley rats weighing 300–350 g (age: 11–12 wk) were anesthetized with 3.0% isoflurane in 100% oxygen, intubated via tracheotomy, and ventilated with oxygen and 2.5–3.0% isoflurane. The chest was opened via midline sternotomy with careful attention to hemostasis. Two pairs of sonomicrometers (Sonometrics, London, ON, Canada) were sewn to the epicardial surface of the anterior wall of the LV, in the region supplied by the left anterior descending coronary artery (LAD). One pair of crystals provided real-time monitoring of the length of a longitudinally oriented segment, whereas the other monitored the length of a circumferentially oriented segment of epicardium. A Millar SP-671 pressure catheter (Millar Instruments, Houston, TX) was inserted into the LV cavity through the apex. After an initial baseline recording, the LAD was ligated using 6-0 suture. Fifteen minutes after the LAD ligation, another recording was taken. A 5 × 15-mm Dacron patch coated in polydimethylsiloxane (PDMS) was sewn under tension over the infarcted area and secured at the lateral and septal edges using 7-0 suture, and a third recording was taken. Plots of pressure versus circumferential or longitudinal segment lengths (see results) were used to assess the impact of patch placement on regional infarct mechanics.

Chronic patch/microdialysis study.

We created large, transmural infarcts in male Sprague-Dawley rats and measured the resulting collagen alignment and turnover at several time points post-MI. Adult male Sprague-Dawley rats weighing 275–300 g (age: 8–9 wk) were ordered from a commercial vendor (Envigo RMS Division, Indianapolis IN) and allowed to acclimate for an average of 1 wk before the initial surgery. Animals were given ad libitum access to food and water throughout the study. On the day of the MI procedure, animals were anesthetized using an intraperitoneal injection of ketamine (60–80 mg/kg ip) and xylazine (5–10 mg/kg ip), intubated, and ventilated with oxygen and supplemental isoflurane (0.5–1%) as needed. Bupivacaine (0.3 ml of a 0.25% solution) was administered locally before incision. A left thoracotomy was performed, and the LAD was permanently ligated with 6-0 suture. Animals in the treatment group received a Dacron patch, as in the acute study. The chest was closed in layers. Subcutaneous buprenorphine (0.5–0.2 mg/kg) was administered immediately, with additional doses every 8–12 h as required for postoperative pain. At 1, 2, 3, and 6 wk, animals were anesthetized with 3.0% isoflurane in 100% oxygen, intubated via tracheotomy, and ventilated with oxygen and 2.5–3.0% isoflurane. The chest was opened via a midline sternotomy, and any tissue adhesions were carefully removed using cotton swabs and cauterization. A 25-gauge syringe needle was inserted into the infarcted region, running parallel to the epicardial surface from the apical edge of the infarct toward its basal edge at a depth of 1–2 mm. The needle was then removed, and a microdialysis catheter (Eicom USA, 3 MDa) was placed in the hole left by the needle and advanced as far as possible. Microdialysis was performed at a flow rate of 3 µl/min for at least 80 min to collect a sufficient amount of interstitial fluid from the LV wall. If the animal had a patch, it was cut in half after microdialysis to test whether the patch was still under tension. If the patch did not spring open when cut, data from that animal were excluded on the logic that we were not confident that the infarct remained mechanically unloaded throughout the study. Whole blood was collected from the LV of control animals using a 1-ml syringe, left to clot at room temperature for 30 min, and centrifuged at 1,000 × g for 10 min, and the serum supernatant stored for analysis. The heart was arrested via retrograde perfusion with cold BDM in PBS, removed, and processed further as described below in Structural analysis.

A total of 106 adult male Sprague-Dawley rats underwent coronary ligation as described above. Twenty-nine of animals (27%) died within 48 h after coronary ligation due to arrhythmia or acute heart failure. The remaining 77 animals (41 animals in the patch group and 36 animals in the nonpatch group) were studied at planned end points of 1, 2, 3, or 6 wk. Five animals in the patch group were excluded because the patch was detached or no longer under tension at the planned end point. By design, after these exclusions we studied at least 8 animals in each group at each time point. In roughly 25% of experiments, we were unable to maintain the open-chest preparation long enough to complete microdialysis; thus, microdialysis results reflect data from at least 6 animals/group at each time point.

Structural analysis.

Scar samples were dissected from the surrounding myocardium, and their thickness was measured at an array of nine evenly spaced locations using a laser displacement sensor (Keyence, Itasca, IL). Scar samples were weighed and their surface area estimated as mass/thickness. After formalin fixation, scar samples were processed via serial dehydration, cleared, and embedded in paraffin. Seven-micrometer sections were cut parallel to the epicardial surface. One slide was selected from the midwall, stained with picrosirius red, and imaged using circularly polarized light. Our method for analysis of collagen content and orientation was adopted from the method developed by Whittaker et al. (33) and has been described in detail by Fomovsky et al. (8). An operator blinded to the test conditions acquired images at ×10 magnification on an Olympus BX51 polarizing microscope using a Lumenera Infinity1-5C camera, subtracted bright-field images from polarized images to isolate collagen fibers from surrounding tissue, and thresholded the subtraction image to differentiate collagen pixels from tissue pixels. Collagen area fraction was computed as the ratio of collagen to tissue pixels. Collagen fiber orientation was measured in the subtraction image with custom MATLAB software (MatFiber, available at http://bme.virginia.edu/holmes/downloads/index.html) using a gradient detection algorithm devised by Karlon et al. (15). Collagen orientation data are presented here using average histograms for each time point, showing the fraction of collagen at each angle weighted by the overall collagen content of each sample. Postmortem analysis showed that patches were oriented close to the circumferential direction but with some animal-to-animal variability; to account for this variability, collagen orientations were expressed relative to the patch axis in all patch animals and relative to the circumferential direction in control (nonpatch) animals.

Benchtop validation of microdialysis system.

To characterize the diffusion properties of the microdialysis catheters, we ran several benchtop experiments and fitted the results to a mathematical model that relates peptide concentrations recovered in the dialysate to the fluid flow rate through the catheter and the peptide concentration surrounding it. Stock solutions containing carboxy-terminal telopeptide of type I collagen (ICTP) and carboxy-terminal propeptide of type I procollagen (PICP) were generated as follows. To generate a solution containing ICTP, we minced rat tail tendons, digested them with 5 mg/ml Liberase (Sigma-Aldrich) at 37°C overnight, centrifuged at 1,000 g for 5 min, and collected the supernatant. To generate a solution containing PICP, we harvested media from primary adult rat fibroblasts cultured on standard tissue culture plastic in medium containing 0.25 mM ascorbic acid for 3 days. For benchtop tests, we placed microdialysis catheters in a 2-ml Eppendorf tube containing known concentrations of ICTP or PICP in PBS and collected 200 µl dialysate at flow rates of 1, 2, 5, and 10 µl/min. The resulting peptide concentrations at each flow rate were measured in duplicate using an ELISA for three independent trials, each using a different microdialysis catheter. We used a custom MATLAB program to predict how recovered peptide concentration should change with flow rate by solving the following transport equation:

$$\frac{\text{dC}}{\text{d}x}=\frac{d}{\text{q}}\left({\text{C}}_{\text{r}}-\text{C}\right)$$ (1)

where dC/dx is the change in peptide concentration with distance as fluid travels through the catheter lumen, d is a lumped transport coefficient proportional to working membrane area and the diffusion coefficient for the peptide across the membrane, q is the flow rate (in µl/min) of fluid through the catheter lumen, C_r is the peptide concentration of the reservoir surrounding the catheter (in ng/µl), and C is the peptide concentration inside the catheter (in ng/µl) at any location and time. The calculated peptide concentration at the outlet of the catheter was accumulated at time steps of 0.2 s to estimate the total peptide concentration collected during each dialysis run. The predicted peptide concentration vs flow rate curve was then fitted to the experimental data by adjusting the effective transport coefficient d.

ELISA analysis.

Two 100-µl aliquots of interstitial fluid collected from each rat were used in ELISAs (MyBioSource) for PICP and ICTP. Samples were first incubated in the ICTP ELISA plate and then transferred to the PICP ELISA plate. Subsequent washes and reagents were conducted according to the kit instructions. Peptide concentrations were quantified using absorbance measurements recorded at a wavelength of 450 nm on a BMG Labtech Fluostar Omega plate reader. The average peptide concentrations from the duplicates of each sample are reported in nanograms per milliliter. Where available, serum samples were analyzed using the same procedure.

Statistical analysis.

Changes in wall thickness, collagen content, and peptide concentration between nonpatch and patch groups were assessed with two-way ANOVA using time and treatment (patch vs. control) as the two factors. Where the overall effect of the patch was significant by ANOVA, Bonferroni-corrected post hoc tests were used to evaluate the effect of patch treatment at individual time points. Collagen alignment was analyzed using circular statistics as previously described by our group (28). Mean vectors having length MVL_j and orientation MA_j were calculated from the fiber distributions in each sample within a group, and the vector components were averaged to produce representative mean vectors with MVL_group and MA_group (2). A group of samples was considered to have significant alignment if the quantity MVL_j × cos[2 × (MA_j – MA_group)] was significantly different from zero by a one-sample t-test with Bonferroni corrections for multiple comparisons at the four time points.

RESULTS

Effect of mechanical reinforcement on collagen alignment.

To predict how local mechanics might affect the strength and direction of collagen alignment post-MI, we simulated mechanical reinforcement of an infarct using a previously published ABM capable of reproducing collagen alignment data measured in infarct scars across a range of animal models (24). We have previously shown that rat infarcts resulting from LAD ligation experience 3–5% equibiaxial strain over the first 6 wk of healing (8). When we simulated infarct healing in unpatched infarcts by imposing 5% equibiaxial strain, the model predicted circumferential alignment of cells and collagen in the midwall at early time points, due to contact guidance from preexisting circumferential collagen (Fig. 1). However, at longer times, the initial structural guidance cue was overwhelmed by the persistent, isotropic mechanical cue, resulting in randomly oriented collagen as previously reported by our group in both standard ligation-induced infarcts (8) and apical cryoinfarcts (9) in rats. Next, we simulated the effects of removing strain in the circumferential direction by reinforcing with a patch under tension in that direction. In this case, the model predicted that persistent longitudinal strain would gradually overwhelm the structural cues from preexisting collagen, resulting in collagen distributions that transitioned from circumferentially aligned at baseline, to random at 2 wk, and to longitudinally aligned at 6 wk. Our simulations of other potential reinforcement strategies (not shown) revealed that they would provide inadequate tests of the hypothesis that mechanical environment is the dominant factor influencing collagen alignment: removing longitudinal strain would align the mechanical stimulus with the preexisting matrix orientation, so that the alignment could be interpreted as resulting from either cue, while removing strain in both directions would eliminate the mechanical stimulus we aimed to study.

Fig. 1.

Histograms showing simulated midwall collagen alignment weighted by collagen content at 1 wk (A), 2 wk (B), 3 wk (C), and 6 wk (D) after an infarction. No patch simulations assumed 5% equibiaxial strain, and patch simulations assumed 5% strain in the longitudinal direction only. A: at 1 wk, preexisting circumferentially aligned collagen dominated model predictions in both groups. B and C: at 2 and 3 wk, isotropic collagen deposition gradually blunted alignment in the no patch group (light gray bars) while deposition of longitudinally oriented collagen fibers gradually overcame the initial alignment in the patch group (black bars). D: by 6 wk, collagen in the patch group was clearly aligned in the longitudinal direction (parallel to the direction of stretch).

To confirm our ability to remove circumferential strain via surgical reinforcement, we measured epicardial segment lengths in the infarct after LAD ligation and reinforcement with a Dacron patch (Fig. 2A). Before reinforcement, circumferential and longitudinal segments traced out counterclockwise pressure-segment length loops reminiscent of pressure-volume loops, indicating active mechanical work in both directions (Fig. 2, B and C). Coronary ligation dramatically reduced the area within the loops, reflecting the loss of active myocardial work within the infarct region. In addition, the shape of the loops changed, with infarct segments stretching and recoiling passively as LV pressure rose and fell (Fig. 2, B and C). Patch reinforcement reduced the systolic strain in the circumferential direction (Fig. 2B) from 10% to ~2%. In contrast, patch reinforcement had little impact on strain in the longitudinal direction, which remained between 10% and 12%. These data confirm that circumferential reinforcement with a Dacron patch converts the biaxial stretch typical for an acute rat infarct to nearly uniaxial longitudinal stretch.

Fig. 2.

Pressure versus segment length loops demonstrating the effect of circumferential reinforcement on regional mechanics. A: schematic of the experimental setup showing the placement of four sonometric crystals as well as a circumferentially aligned patch. B: active baseline pressure versus circumferential segment length loops (gray) converted to passive stretching and recoil during acute ischemia (orange), and patch reinforcement (red) reduced circumferential stretching from ~10% to 2%. C: longitudinal segment length loops also converted to passive stretching and recoil during ischemia and were minimally affected by circumferential reinforcement.

We next investigated how changing the local mechanics of an infarct altered collagen alignment in the evolving scar. As previously reported (8), while individual control scars often showed some preferred direction of alignment, this direction was different in every sample, so that on average there was no significant alignment in any direction at any time point (Figs. 3 and 4). In the patch groups, infarcts showed preferential alignment perpendicular to the direction of reinforcement by 1 wk and maintained this alignment as collagen accumulated (P < 0.01 at all time points; Figs. 3 and 4).

Fig. 3.

Representative images of picrosirius red-stained scar samples 1 wk and 6 wk after myocardial infarction. A and B: in the nonpatch group, the scar contained sparse, wavy collagen fibers at 1 wk (A) and dense, straighter fibers at 6 wk (B). In both cases, fiber alignment varied from region to region even within a single microscopic field, with no clear overall preferred fiber orientation. C and D: in the patch group, collagen density was similar to control at 1 wk (C) and 6 wk (D), but fibers showed clear and consistent longitudinal alignment.

Fig. 4.

Histograms of average experimental collagen alignment weighted by area fraction at 1 wk (A), 2 wk (B), 3 wk (C), and 6 wk (D) after myocardial infarction. The patch group showed significant alignment at all time points (P < 0.01 by Bonferroni-corrected t-tests, n = 8–11 per group, 0° indicates direction of patch reinforcement). In contrast, the no patch group showed no significant alignment at any time point (0° indicates circumferential direction).

Effect of mechanical reinforcement on collagen turnover.

We conducted a series of benchtop tests to assess the repeatability of our microdialysis methods and characterize the transport of PICP and ICTP at different pump speeds. The peptide concentrations measured by microdialysis depend not only on the peptide concentrations surrounding the catheter but also on the speed at which fluid is flowing through the catheter lumen, the working membrane area, and the diffusion coefficient of the peptide of interest through the catheter membrane. The transport model used here (Eq. 1) fitted data from individual trials well, with an average root mean squared error between the model curve and data points of 1.47 ng/ml for PICP and 1.35 ng/ml for ICTP (Fig. 5). Three benchtop trials with three different catheters showed good repeatability, with the range of recovered peptide concentrations representing < 10% of the mean for PICP and 25% of the mean for ICTP at a flow rate of 2 µl/min (Fig. 5).

Fig. 5.

Benchtop validation of experimental microdialysis technique. Three different catheters were used to perform microdialysis in solutions of known carboxy-terminal propeptide of type I procollagen (PICP; A) and carboxy-terminal telopeptide of type I collagen (ICTP; B) concentrations at flow rates of 1, 2, 5, and 10 µl/min, and peptide concentrations were measured in duplicate in the dialysate using an ELISA. Equation 1 was fitted to the data by adjusting the effective transport coefficient (d) and provided a good prediction of how dialysate concentration varied with flow rate. C_r, peptide concentration of the reservoir surrounding the catheter.

We measured the effect of patch reinforcement on peptide levels associated with collagen synthesis and degradation. In nonpatch animals, both PICP (synthesis) and ICTP (degradation) peptide levels rose post-MI and remained elevated through 6 wk (Fig. 6). Overall, patch reinforcement significantly reduced both PICP and ICTP levels by ANOVA (P < 0.0001). However, the timing of the effects on the two peptides differed: reinforcement appeared to affect PICP primarily at early time points and ICTP primarily at later time points (Fig. 6).

Fig. 6.

Measured concentrations of carboxy-terminal propeptide of type I procollagen (PICP; A) and carboxy-terminal telopeptide of type I collagen (ICTP; B) in patch and nonpatch groups compared with baseline levels. Interstitial fluid was collected in vivo at 1, 2, 3, and 6 wk after ligation using microdialysis, and peptide concentrations were measured in duplicate using an ELISA. PICP and ICTP levels rose quickly after ligation and plateaued at 2–3 wk after myocardial infarction. The patch significantly lowered levels of PICP (P < 0.0001) and ICTP (P < 0.0001) as assessed by two-way ANOVA (n = 6–9 per group). *P < 0.05 by Bonferroni-corrected post hoc comparison of patch and no patch groups at the time points indicated.

Histologic analysis of the tissue samples revealed that collagen area fraction rose over time in both groups and was not affected by patch reinforcement (Fig. 7A). We also measured the wall thickness of every sample before fixation and used the measured sample weights to estimate scar surface area. Consistent with previous reports, we found progressive thinning in infarct scars over time, with little change in surface area. Thinning was significantly greater in animals in the patch group (ANOVA, P < 0.0001; Fig. 7B), whereas scar surface area was not significantly different between the groups at any time point (data not shown).

Fig. 7.

Quantification of changes in infarct collagen density and wall thickness. A: the collagen area fraction rose after ligation and plateaued after about 3 wk. The patch had no significant effect on area fraction compared with the nonpatch group. B: wall thickness decreased steadily after ligation in both groups. Wall thickness in the patch group was significantly lower than the nonpatch group (P < 0.0001) by two-way ANOVA (n = 8–11 per group). *P < 0.05 by Bonferroni-corrected post hoc comparison of patch and no patch groups at the time points indicated.

DISCUSSION

A number of therapies such as patch reinforcement and biomaterial injection that aim to alter the regional mechanics of healing myocardial infarcts are currently under development. When designing such therapies, it is essential to anticipate their long-term impact on the structure of the healing infarct. In fact, other promising postinfarction therapies [such as postinfarction administration of steroids (23)] have failed spectacularly because of unanticipated effects on scar healing. Here, we tested how one specific aspect of these therapies (their ability to alter regional strain in the infarct) can influence collagen alignment and turnover in the developing scar. In support of our hypothesis that regional mechanics are the primary determinant of scar collagen alignment, we found that surgically reinforcing the infarct to allow only longitudinal strain resulted in strong longitudinal alignment of midwall collagen fibers, perpendicular to the preinfarction muscle and collagen fiber direction. However, the time course of realignment contradicted model predictions based on prior work on infarct healing in multiple species. Furthermore, the effects of surgical reinforcement on peptides associated with collagen synthesis and degradation were not entirely consistent with expectations based on prior in vitro studies.

Effect of mechanical reinforcement on scar collagen fiber alignment.

The ABM used in this study encapsulates our best current understanding of what factors influence the alignment of collagen fibers in healing infarct scar (9, 24, 25). This model successfully predicts collagen alignment across a range of experiments and animal models, including coronary ligation in the rat, cryoinfarction at various locations on the rat LV, and even transmural distributions of collagen orientation after ligation of a branch of the left circumflex coronary artery in pigs (24). However, the present study represents the first attempt to use the model to prospectively predict the impact of a surgical intervention. In one respect, our data agree well with the model prediction: as predicted, circumferential reinforcement with a Dacron patch produced significant longitudinal alignment, whereas untreated controls displayed no significant alignment. This result strongly supports our hypothesis that mechanical environment is the primary determinant of collagen alignment, because midwall collagen fibers in our patch group oriented parallel to strain but perpendicular to any guidance cues provided by the circumferentially oriented muscle and collagen fibers present initially in the infarcted region.

On the other hand, differences between the model-predicted and observed time course of remodeling suggest that some questions remain about the precise balance between the multiple cues that may influence fibroblast and collagen alignment in healing infarcts. In our published ABM, contact guidance from preexisting matrix and regional mechanics both influence fibroblast and collagen alignment in the healing scar. Thus, we expected a gradual transition from circumferential alignment at 1 wk (Fig. 1A) to either random collagen orientations in untreated controls or longitudinal alignment in animals in the patch group (Fig. 1D). Instead, we saw that these alignment patterns appeared by 1 wk in both groups and then persisted at the remaining time points as collagen accumulated (Fig. 4). One way to reconcile the model and experimental results would be to reduce the impact of contact guidance from surrounding extracellular matrix in the model. However, significant influence from both contact guidance and mechanics appears to be critical to successfully predicting transmural gradients in scar collagen orientation reported previously in large animal models (24). Another possibility would be to assume that some of the large collagen fibers are destroyed by the initial infarction and subsequent inflammatory response. While the original model underpredicted the strength of alignment observed in the patch groups, a modified version with the initial collagen area fraction in the infarct region set to 0% did better, without any additional parameter fitting (Fig. 8). There is certainly evidence that the native collagen fibers are damaged by ischemia (27), but we are not aware of any quantitative data on the fraction of collagen that is damaged or on whether that process differs in different animal models. Thus, more work remains to understand the underlying basis for the time course of alignment reported here.

Fig. 8.

Agent-based model simulations of the strength of collagen alignment (mean vector length, ranging from 0 for randomly oriented collagen to 1 for perfectly aligned). Before the experiments were conducted, the published Agent-based model (light gray line) predicted rapid loss of circumferential alignment and then slow development of longitudinal alignment. However, experimental data showed that the patch group developed fairly strong longitudinal alignment by 1 wk, which persisted throughout the study (filled circles indicate experimental means ± SD). Modifying the model to assume destruction of all preexisting collagen during the initial inflammatory phase (black line) provided a better match to the measured time course but did not capture the strong alignment at 1 wk.

Effect of mechanical reinforcement on collagen turnover and accumulation.

PICP is cleaved during collagen type I assembly, and changes in its tissue level provide a measure of changes in the rate of collagen production (7, 16, 26, 29, 30a). In our control group, levels of PICP measured in dialysate collected at a flow rate of 3 µl/min rose from near the detection limit in control animals to ~15 ng/ml at 1 wk and then ranged from 9 to 12 ng/ml at later times (Fig. 6A). This trend is generally consistent with the expected rapid increase in collagen-producing fibroblasts after infarction, although most reviews of the infarct healing process envision a slower rise to peak collagen production (21). ICTP is generated when collagen is degraded by MMPs and provides an indirect measure of collagen degradation (10, 16, 26). In our control group, ICTP increased over the first 2 wk postinfarction and then remained relatively constant (Fig. 6B). This result may be more surprising given the rapid pulse of MMPs produced by inflammatory cells in the first few days post-MI (4, 6, 19, 31), but it is exactly what we previously predicted (5), due to the fact that the rate of collagen degradation by MMPs depends not only on the concentration of the enzyme (MMPs) but also on the availability of substrate (collagen), which is limited in the first days post-MI.

Based on benchtop tests, we expect tissue levels of PICP to be approximately eightfold higher than the concentrations recovered in dialysate at a flow rate of 3 µl/min; thus, the data shown in Fig. 6 suggest levels in the infarct of 70–120 ng/ml. These values are in line with data from Langberg et al. (16), who used similar microdialysis methods to sample tissue surrounding Achilles tendons in human volunteers and computed PICP tissue levels of 55 ng/ml at rest and 165 ng/ml after exercise. For ICTP, which is smaller, levels measured in the dialysate during benchtop tests were closer to the reservoir concentration, differing by less than fourfold at 3 µl/min. Thus, our dialysate ICTP levels suggest concentrations in the infarct of 20–50 ng/ml. Not surprisingly, these levels are higher than the 5–10 ng/ml reported by Langberg et al. in the tissue surrounding the human Achilles tendon (16). The difference between dialysate and tissue concentrations arises because the fluid flowing through the catheter lumen does not spend enough time in contact with the membrane to fully equilibrate with the surrounding interstitial fluid. Most microdialysis studies have used the loss of a tracer molecule from the dialysate to correct for this incomplete equilibration. Here, we used a different approach, combining a mathematical model of transport in the catheter with benchtop trials across a range of pump speeds to establish the relationship between source (reservoir or tissue) and dialysate concentrations of PICP and ICTP. The transport equation we used makes a number of simplifications, assuming flow through a single straight tube surrounded by an infinite reservoir. Yet it fitted our benchtop data remarkably well and, in our opinion, provides a convenient yet physically realistic basis for interpreting microdialysis data.

In vitro studies in the literature suggest that mechanical unloading should affect collagen degradation by MMPs more than collagen production. Stretching cultured fibroblasts does enhance collagen expression, but in most studies the change was less than twofold (1, 14, 18). In contrast, Huang et al. (13) showed that digestion of collagen fibers by collagenase was at least fourold higher in unstretched versus moderately stretched collagen fibers. Furthermore, Zhou et al. (34) reported that complete mechanical unloading of infarcted rat hearts by heterotopic transplantation increased MMP-9 zymographic activity. Thus, we expected little change in PICP but a substantial increase in ICTP in our patch group relative to control. Instead, we observed that levels of both PICP and ICTP were lower in surgically reinforced scars (Fig. 6). Interestingly, the timing of the effects on the two peptides differed, with reinforcement affecting PICP primarily at early time points and ICTP primarily at later time points. Given that reduced levels of collagen deposition and accumulation should also reduce the rate of collagen degradation, the simplest explanation, but certainly not the only one, consistent with our peptide data is that mechanical unloading by patch reinforcement primarily reduces the rate of collagen deposition, with a secondary effect on collagen degradation at later times due to the reduced availability of substrate (collagen) for cleavage by MMPs.

We expected the substantial reduction in collagen synthesis rates observed in reinforced infarct scars to result in reduced collagen content. However, area fraction was clearly not different in the two groups. This finding highlights the fact that area fraction, along with traditional biochemical measures of collagen content, is really a measure of collagen concentration rather than mass. To estimate the total mass of collagen that has accumulated in a post-infarction scar, we should multiply collagen concentration times scar volume. If we estimate collagen mass in our samples by multiplying area fraction times scar volume, we get a much lower estimate for the patch group, because the scar is so much thinner in that group. This result is particularly interesting given the apparently paradoxical findings from several studies that inhibited or knocked out various MMPs in healing myocardial infarcts but observed no changes in collagen area fraction (for a review, see Ref. 5). It seems plausible that some of these studies might have missed changes in total collagen mass by focusing only on area fraction. Our findings suggest that when testing therapies that modulate collagen turnover, it is important to measure wall thickness or scar volume in addition to collagen concentration or area fraction. Furthermore, the fact that changes in collagen synthesis rates led to significant changes in scar thickness in our study might have important consequences when designing biodegradable patches and injectables, since a temporary support that reduces wall thickness but then degrades could leave the scar vulnerable to rupture and/or promote overall ventricular dilation.

Limitations and sources of error.

The biggest uncertainty in the work presented here relates to the time course of mechanical changes induced by surgical reinforcement. Although we verified that sewing patches to the epicardial surface under tension eliminated nearly all systolic strain in the infarct acutely, both the scar formation process and geometric remodeling could have altered regional infarct mechanics over the several weeks of this study. To guard against this potential uncertainty, we cut the patches at each end point study to assess whether the edges separated, indicating that the patch remained under tension. Thus, while we cannot know the exact strain history of each infarct, we are reasonably confident that circumferential strains remained below normal in the reinforced infarcts, a confidence further reinforced by the significant effects we observed on scar structure and collagen turnover. However, one place where the exact value of strain might be critical to the interpretation of our data is the effect of reinforcement on collagen degradation. In vitro studies have revealed a U-shaped relationship between strain and the rate at which collagenase digests large collagen fibers, with a minimum around 5% (13); thus, whether reduced strain would be expected to increase or decrease collagen degradation depends on both the starting strain value and the extent of the reduction.

In terms of the microdialysis data, benchtop testing allowed us to document low variability in both the dialysis and ELISA methods. The biggest source of error in the in vivo microdialysis studies was therefore likely the variability in the fraction of the catheter length that could be placed within infarct tissue, which in our experience ranged from roughly 70−100%. Due to the long times required to collect sufficient sample volume, we chose to subject each aliquot of dialysate to both ELISA assays in series, which in theory could introduce error. To guard against this possibility, we reversed the order of the assays in some benchtop tests and confirmed that we could not detect any difference in the measured values. Another consequence of the limited sample volume was that we were only able to measure peptides related to synthesis and cleavage of collagen type type I and not to other subtypes, such as collagen type III, that are known to be present in healing infarct scar. Thus, conclusions about collagen turnover in this manuscript are specific to collagen type I. Furthermore, without comprehensive data on the levels and activity of the array of MMPs, tissue inhibitors of metalloproteinase, and other relevant players, we cannot draw conclusions about the specific mechanisms underlying the mechanically induced changes in collagen turnover reported here.

We interpreted the reduction in PICP levels in the patch group as indicating a reduction in collagen synthesis; an alternate potential explanation would be that surgical reinforcement decreased or delayed fibroblast proliferation or migration into the infarcts. In this case, total PICP levels could be reduced even if per-cell collagen synthesis rates were unchanged. To check the likelihood of this alternate explanation, we measured the cellular area fraction in hematoxylin and eosin-stained sections from a subset of scar samples. The cellular area fraction was high at 1 and 2 wk and then dropped, with no apparent differences between patch and nonpatch samples. Manual counts of cells displaying a fibroblast-like morphology also showed no difference between the groups at any time point. Thus, it appears unlikely that differences in fibroblast density could account for the severalfold reduction in PICP concentrations we observed at some time points after surgical reinforcement.

Conclusions.

This study confirms a central role for local mechanics in regulating scar formation after MI. When we perturbed infarct mechanics by surgical reinforcement, collagen in healing infarcts aligned parallel to regional strain and perpendicular to the preinfarction muscle and collagen fiber direction within 1 wk, suggesting that stretch or stress plays an even more dominant role than we hypothesized based on prior experiments and computational modeling of infarct healing. Surgical reinforcement also reduced collagen synthesis early and degradation late and dramatically reduced wall thickness in the infarct region. These effects may be particularly important to consider when designing biodegradable patches or injectables, since reducing wall thickness could lead to increased long-term risk of rupture or ventricular dilation.

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grants R01-HL-116449 and R01-HL-137755.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

L.R.C. and J.W.H. conceived and designed research; L.R.C., J.-J.L., and J.W.H. performed experiments; L.R.C., J.-J.L., and J.W.H. analyzed data; L.R.C., J.-J.L., and J.W.H. interpreted results of experiments; L.R.C. prepared figures; L.R.C., J.-J.L., and J.W.H. drafted manuscript; L.R.C., J.-J.L., and J.W.H. edited and revised manuscript; L.R.C., J.-J.L., and J.W.H. approved final version of manuscript.