Diaphragm muscle fibrosis involves changes in collagen organization with mechanical implications in Duchenne muscular dystrophy

Abstract

In Duchenne muscular dystrophy (DMD), diaphragm muscle dysfunction results in respiratory insufficiency, a leading cause of death in patients. Increased muscle stiffness occurs with buildup of fibrotic tissue, characterized by excessive accumulation of extracellular matrix (ECM) components such as collagen, and prevents the diaphragm from achieving the excursion lengths required for respiration. However, changes in mechanical properties are not explained by collagen amount alone and we must consider the complex structure and mechanics of fibrotic tissue. The goals of our study were to 1) determine if and how collagen organization changes with the progression of DMD in diaphragm muscle tissue and 2) predict how collagen organization influences the mechanical properties of the ECM. We first visualized collagen structure with scanning electron microscopy (SEM) images and then developed an analysis framework to quantify collagen organization and generate image-based finite-element models. Image analysis revealed increased collagen fiber straightness and alignment in mdx over wild type (WT) at 3 mo (straightness: mdx = 0.976 ± 0.0108, WT = 0.887 ± 0.0309, alignment: mdx = 0.876 ± 0.0333, WT = 0.759 ± 0.0416) and 6 mo (straightness: mdx = 0.942 ± 0.0182, WT = 0.881 ± 0.0163, alignment: mdx = 0.840 ± 0.0315, WT = 0.759 ± 0.0368). Collagen fibers retained a transverse orientation relative to muscle fibers (70°–90°) in all groups. Mechanical models predicted an increase in the transverse relative to longitudinal (muscle fiber direction) stiffness, with stiffness ratio (transverse/longitudinal) increased in mdx over WT at 3 mo (mdx = 5.45 ± 2.04, WT = 1.97 ± 0.670) and 6 mo (mdx = 4.05 ± 0.985, WT = 1.96 ± 0.506). This study revealed changes in diaphragm ECM structure and mechanics during disease progression in the mdx muscular dystrophy mouse phenotype, highlighting the need to consider the role of collagen organization on diaphragm muscle function.

NEW & NOTEWORTHY Scanning electron microscopy images of decellularized diaphragm muscle from WT and mdx, Duchenne muscular dystrophy model, mice revealed that collagen fibers in the epimysium are oriented transverse to muscle fibers, with age- and disease-dependent changes in collagen arrangement. Finite-element models generated from these images predicted that changes in collagen arrangement during disease progression influence the mechanical properties of the extracellular matrix. Thus, changes in collagen fiber-level structure are implicated on tissue-level properties during fibrosis.

INTRODUCTION

Duchenne Muscular Dystrophy is a Devastating Disease with No Cure and Diaphragm Muscle Weakness Leads to Death

Duchenne muscular dystrophy (DMD) is a fatal genetic disease, with devastating impacts from the subcellular to whole muscle levels (1–3). Muscle degeneration results due to a lack in expression of the protein dystrophin, responsible for maintaining the linkage between the intracellular cytoskeleton and extracellular matrix (ECM) (4). Current therapies are targeted toward either replacing the dystrophin protein or treating the secondary and downstream pathological mechanisms, yet there remains no cure (5). Drug therapies (i.e., genetic therapies and antifibrotics) show promise for the treatment of DMD, but are only effective in a subset of patients and show variable functional benefits (5). The absence of dystrophin at the muscle fiber membrane increases susceptibility to mechanical stress from everyday muscle contractions. Regenerative capacity of muscle is decreased, with contraction-induced damage leading to a chronic state of inflammation and subsequent fibrosis (6–10). A cycle of dysfunction and disuse results in progressive muscle wasting, with differences in severity and progression across muscles (11). Lower limb muscle is impacted in the earlier stages of DMD, leading to a loss of ambulation at 10–14 yr of age. At later stages of the disease, the diaphragm, the main inspiratory muscle (12), is severely impacted. Diaphragm muscle weakness progresses with age in DMD and contributes to respiratory insufficiency, a leading cause of death in the early- to mid-20s (13, 14).

Fibrosis Limits Muscle Function, but the Structure and Mechanics of Fibrotic Tissue Are Not Well Characterized in Diaphragm Muscle

One of the primary sources for progressive muscle dysfunction in DMD is the development of fibrosis, characterized by excessive accumulation of extracellular matrix (ECM) components such as collagen. Indeed, many studies have shown that dystrophic muscles have increased amounts of collagen (15, 16), but collagen amount does not correlate with stiffness in diaphragm muscle tissue from mdx mice, the most common animal model used to study DMD (17). Mdx lower limb muscle shows an increase in collagen fiber alignment (18). Although collagen fiber amount does not predict tissue stiffness, collagen fiber alignment is reported to be a significant predictor of passive lower limb muscle stiffness (18). However, the mdx lower limb muscle does not mimic the severity of the human phenotype nearly as well as the diaphragm does. Similar to the human condition, mdx mice exhibit impairment in respiratory function (19, 20), decrease in diaphragm muscle fiber cross-sectional area, and increased diaphragm muscle fibrosis (19, 21–23). Beyond skeletal muscle, changes in collagen organization with fibrosis are implicated in additional tissue systems. Increased collagen fiber alignment is reported in pulmonary fibrosis (24) and increased collagen fiber straightness is reported in cancerous pancreatic tissue (25). Changes in collagen organization, such as collagen fiber direction, alignment, and straightness, remain unknown in mdx diaphragm muscle, but are needed to understand how and why diaphragm muscle mechanics change during the progression of DMD.

Finite-Element Models Allow us to Study Structure-Function Relationships in Biological Tissues

In prior studies, collagen organization is related to passive properties of skeletal muscle measured by mechanical testing of intact muscle (17, 18). The contribution of the ECM to the passive properties of skeletal muscle is shown by indirect measurements, comparing properties of single muscle fibers and muscle fiber bundles with intact ECM (26–28), and direct measurements of properties of decellularized muscle fiber bundles (29). Although it is difficult to isolate the influence of organizational parameters on ECM properties with traditional mechanical testing, finite-element (FE) models allow us to isolate the impact of specific structural variations on mechanical properties. Whole muscle-level FE models reveal the influence of macroscopic muscle architecture on mechanical properties (30) but have been limited in their representation of the ECM. These models typically lump together connective tissue, muscle fibers, and muscle fascicles into one transversely isotropic material, without accounting for changes in ECM structure during fibrosis (30–33). Micromechanical muscle models at the fascicle-level reveal transversely anisotropic behavior, with macroscopic properties such as shear moduli dependent on fiber and fascicle shapes (34). When changes in fascicle microstructure seen in DMD such as variation in muscle fiber cross-sectional area were simulated in these models, the influence of these changes in microstructure on macroscopic properties was dependent on the relative stiffness between the ECM and muscle fibers (35). However, these muscle fascicle-level FE models assumed that the ECM was aligned with muscle fibers and did not account for changes in the structure of the ECM during DMD. Therefore, as we apply these modeling techniques to study fibrotic muscle, we must first consider the complex structure and function of the ECM.

We Aim to Quantify Changes in ECM Structure and Mechanics during the Progression of Diaphragm Muscle Fibrosis

The goals of our study were to 1) determine if and how collagen organization changes with the progression of DMD in diaphragm muscle tissue and 2) predict how collagen organization influences the mechanical properties of ECM. We aimed to characterize collagen organization within the epimysium and predict how changes in its structure are implicated on both transverse (cross muscle fiber) and longitudinal (along muscle fiber) tissue properties during disease progression. To do so, we developed an image-based finite-element modeling pipeline to explore the influence of collagen organization on ECM mechanics. We collected scanning electron microscopy (SEM) images of epimysium isolated from mdx and wild-type (WT) control mice at 3, 6, and 12 mo, and quantified collagen fiber direction, alignment, and straightness. We then generated finite-element models and simulated biaxial stretch to determine the implications of our collagen organization measurements on ECM mechanical properties.

METHODS

Animal Protocol

All experiments were approved by the University of Virginia Animal Care and Use Committee. This study was conducted in C57BL/10ScSn-Dmdmdx/J male mice (referred to as mdx), bred inhouse, and C57BL/6J male mice (referred to as WT), purchased from Jackson Laboratories. Our study groups included 3-, 6-, and 12-mo-old mice, both WT and mdx (n = 6 mice/group).

Ex Vivo Sample Collection and Imaging

After humane euthanasia, the diaphragm muscle was excised and samples from the costal region were dissected and placed in phosphate-buffered saline. We followed a standard sodium hydroxide digestion protocol to leave only collagen fibrils, removing the muscle fibers, sarcolemma, basement membrane, and proteins associated with the ECM (36). Although we imaged the outer epimysial layer of the ECM, the digestion process was required for clear imaging, visualization, and quantification of collagen fibers alone (Supplemental Fig. S1; see https://doi.org/10.6084/m9.figshare.17026361.v1). Excised muscle samples were first placed in a fixative for 24 h (8% glutaraldehyde, 16% paraformaldehyde, and 0.2 M sodium cacodylate). Samples were then placed in digestion solution (10% sodium hydroxide) for ∼6 days, and then rinsed in H₂O for 24 h. Since samples were physically unconstrained during enzymatic digestion, this protocol left the tissue in a zero-strain configuration after muscle fibers were digested. Therefore, the arrangement of collagen fibers reflects their position with the tissue in a stress-free state. Samples were prepared for scanning electron microscopy (SEM) imaging following standard dehydration with a graded series of EtOH (10%–100%), mounted on 1/8 in. stubs, and sputter coated in gold. Care was taken to ensure that tissue samples were mounted such that SEM images of the surface plane captured the outer epimysial layer of the ECM. One image per sample was captured at the following magnifications: ×40, ×500, ×1,000, and ×15,000 to visualize collagen organization (Zeiss Sigma VP HD field SEM). Images collected at ×1,000 were then used for measurements in our image analysis. SEM images collected in this study are available at https://doi.org/10.6084/m9.figshare.14398184.v1.

Image Analysis

Muscle fiber direction.

After enzymatic digestion in sodium hydroxide, collagen structure was isolated from samples of diaphragm muscle. Ridges in tissue samples were still evident, indicating the presence of muscle fibers that had been digested from the samples (Fig. 1A). From these features, we measured muscle fiber direction manually from raw images (1,024 × 768 pixels) by tracing three locations along the length of digested muscle fibers and averaging the angles of the traces (Fig. 1A). Images collected at ×1,000 magnification were rotated such that the muscle fiber direction aligned with the horizontal direction, and then the images were cropped to a square (540 × 540 pixels) (Adobe Illustrator) (Fig. 1B). Images collected at ×15,000 were used to confirm successful digestion by visually inspecting that collagen fibers were isolated and other proteins and attachments in the ECM were eliminated (Fig. 1C).

Figure 1.

A: after digestion in sodium hydroxide, collagen structure was isolated from samples of diaphragm muscle. Ridges in tissue samples were still evident, indicating the presence of muscle fibers that had been digested from the samples (as seen in light red bars). From these features we measured muscle fiber direction manually from raw images (1,024 × 768 pixels) by tracing three locations along the length of digested muscle fibers (as seen in red arrows) and averaging the angles of the traces. B: images collected at ×1,000 magnification were then rotated with muscle fiber direction on the horizontal and cropped to a square (540 × 540 pixels) (Adobe Illustrator). C: images collected at ×15,000 were used to confirm successful digestion by visually inspecting that muscle fibers were not present. D: collagen straightness parameter (P_s) was determined by manually tracing the contour path length of a representative collagen fiber (L_f) and the linear end-to-end straight-line length was determined by drawing a straight line connecting the ends of the measured fiber (L_o). Collagen straightness parameter (P_s = L_o/L_f) was then calculated for three fibers representative of the fibers in each image (n=3 replicates of measurement per image) and averaged (ImageJ).

Collagen fiber straightness.

Collagen straightness (P_s) was determined using the following relationship:

$$P_{\text{s}}=L_{\text{o}}/L_{\text{f}}$$ (1)

where L_f, collagen fiber length, was calculated by manually tracing the contour path length of a representative collagen fiber, and L_o, the linear end-to-end straight-line length, was determined by drawing a straight line connecting the ends of the measured fiber. The collagen straightness parameter was then calculated for three representative fibers and averaged (ImageJ) (Fig. 1D).

Image Processing Algorithm

Local collagen measurements.

We developed an image processing algorithm to automatically measure collagen orientation within subregions of each image in MATLAB (MathWorks Inc., Natick, MA). Each image was discretized into 4,096 subregions of 16 × 16 pixels (Fig. 2A). We utilized built-in image processing functions (MATLAB R2018b and Image Processing Toolbox 3.5.8) to measure collagen fiber direction. Each subregion of the image (i) was first thresholded using Otsu’s method (37). Collagen pixel ratio (cpr_i) was calculated by dividing the number of white pixels detected as collagen fibers (n_coll) by the number of total pixels (n_tot) (Fig. 2B).

$${\text{cpr}}_i=\frac{n_{\text{coll}}}{n_{\text{tot}}}.$$ (2)

Figure 2.

A: scanning electron microscopy images were discretized into 64 × 64 image subregions (16 × 16 pixels/subregion). B: each subregion was thresholded (Otsu’s Method) and collagen pixel ratio was measured (n_coll/n_tot). C: fiber boundaries (Canny Edge Detection) were determined. D: the Radon transform was computed at fiber boundaries, with rotation angle at the maximum peak used as a measure of dominant orientation. E: local collagen direction (α_i) was reported per image subregion relative to the horizontal axis (muscle fiber direction) and constrained to the first two quadrants. F: examples of input images for 3-mo-old wild-type (WT) (used above in E and F) and mdx mice. G: local collagen directions measured per image window displayed with a spatial heat map for each image. H: mean collagen fiber direction (cfd) was reported per image by taking the circular mean of local collagen directions (α_i), measured as the acute angle relative to the horizontal axis (muscle fiber direction). Strength of alignment (SA) was quantified as the length of the mean resultant vector and used to measure the circular spread in local orientations per image (0<SA < 1, 1 = high alignment).

Fiber boundaries were determined from thresholded image subregions using Canny edge detection (Fig. 2C). We then computed the Radon transform (38, 39) at fiber boundaries to measure fiber orientation (40–42). The Radon transform, R_θ(a′) provides the predominant angle of fiber alignment in a subregion by computing line integrals along parallel-beam projections oriented at discrete rotation angles (θ) and spaced 1 pixel apart. For a two-dimensional function, f(a,b), a and b are the horizontal and vertical axes and a′ and b′ are the axes of the parallel-beam projections determined by the prescribed rotation angle (θ). Equations 3 and 4 describe the Radon transform:

$$R_{\theta }\left(a\prime \right)=\underset{\mathrm{\infty }}{\overset{\mathrm{\infty }}{{\displaystyle \int }}}f\left(a\prime \text{cosθ}-b\prime \text{sinθ},a\prime \text{sinθ}+b\prime \text{cosθ}\right)\text{d}b\text{′},$$ (3)

$$\left[\begin{array}{c}a\prime \\ b\prime \end{array}\right]=\left[\begin{array}{cc}\text{cosθ}& \text{sinθ}\\ -\text{sinθ}& \text{cosθ}\end{array}\right]\left[\begin{array}{c}a\\ b\end{array}\right].$$ (4)

For each image subregion, we computed the Radon transform R_θ(a′) while varying rotation angle (0° < θ < 180°). The subregion Radon transform reaches a unique maximum value at the angle of greatest pixel alignment. The maximum value of the Radon transform was taken as a measure of collagen alignment within each image subregion.

$$R_{\text{peak}}=\text{ max[}{R}_{\theta }\left(a\prime \right)].$$ (5)

The angle of greatest pixel alignment was taken as the subregion predominant collagen fiber direction (α_i) and confined to the first two quadrants such that (0° <α_i < 180°) (Fig. 2D)

$$\left\{{\text{α}}_i\right\}={\mathrm{ arg\; max}}_{\mathrm{\theta }}[R_{\theta }\left(a\prime \right)].$$ (6)

Predominant collagen fiber direction α_i and peak Radon intensity R_peak were calculated for all 4,096 image subregions (Fig. 2E).

Mean collagen measurements.

We utilized built-in MATLAB circular statistics functions (43) to calculate mean collagen direction and strength of alignment. Subregion collagen directions (α_i) were converted to unit vectors (r_i) and averaged across the total number of image subregions (n) to obtain the image mean resultant vector $(\overline{\mathbf{r}}$)

$${\mathbf{r}}_{\mathbf{i}}=\left(\begin{array}{c}\text{cos}({\text{α}}_i)\\ \text{sin}({\text{α}}_i)\end{array}\right),$$ (7)

$$\overline{\mathbf{r}}=\frac{1}{n}{\displaystyle \sum }{\mathbf{r}}_{\mathbf{i}}.$$ (8)

Next, the mean collagen fiber direction (cfd) was calculated per image as the acute angle of $\overline{\mathbf{r}}$ relative to the horizontal axis, such that (0° < cfd < 90°) (Fig. 2H). The resultant vector length was used to calculate the strength of collagen fiber alignment (0 < SA < 1, 1 = high alignment), capturing the circular spread in local orientations per image

$$\text{SA}=||\overline{\mathbf{r}}||.$$ (9)

The mean in peak Radon intensity was determined per image as a measure of subregion collagen alignment. The mean collagen pixel ratio was also determined per image as a measure of the number of pixels detected as collagen. Our SEM images were collected of the surface of the epimysium and thus, the collagen pixel ratio corresponds to the density of collagen in the imaging plane

$${\text{cpr}}_{\text{image}}=\frac{1}{n}{\displaystyle \sum }_i{\text{cpr}}_i.$$ (10)

Sensitivity and validation.

To validate the image processing algorithm and determine the appropriate subregion size, we first used our image processing algorithm with two manually generated sets of test images of dark lines that approximated collagen fibers. In the first image set, we varied collagen (line) direction over the range 0° < cpd < 90° while holding strength of alignment constant (SA_known = 1) (Supplemental Fig. S2A; see https://doi.org/10.6084/m9.figshare.17026388.v1). In the second set, we varied strength of alignment over the range 0.92 < SA < 0.99 with collagen direction constant (cpd_known = 90°) (Supplemental Fig. S2B). To test the sensitivity of our algorithm to the image subregion size, we also varied the number of image subregions (2 × 2 < n × n < 128 × 128) and calculated the error between the known collagen direction and strength of alignment in our test images versus the values determined by our image processing algorithm. As we increased the number of image subregions, error in collagen direction increased (Supplemental Fig. S2C), with error in strength of alignment minimized at the 64 × 64 subregion size, corresponding to 1.8 × 1.8 μm (Supplemental Fig. S2D). Based on this analysis, we selected the 64 × 64 subregion size, which yielded a collagen direction error of 0.35° and a strength of alignment error of 0.014 (Supplemental Material S2).

In Silico Finite-Element Modeling

Geometry.

We generated finite-element (FE) models in the nonlinear finite element solver, FEBio (Musculoskeletal Research Laboratories, University of Utah, Salt Lake City, UT) (44). FE models corresponded to the height and width of our cropped SEM images (118 μm × 118 μm), with a constant thickness (3 μm). The FE model was meshed into 64 × 64 hex8 elements, with each element corresponding to one subregion (1.8 μm × 1.8 μm), from our image processing algorithm (Fig. 3A).

Figure 3.

A: finite-element (FE) models were generated in the nonlinear finite element analysis software suite, FeBio, corresponding to the height and width of our cropped scanning electron microscopy (SEM) images (118 μm × 118 μm), with a constant thickness (3 μm). The FE model was meshed into 64 × 64 hex8 elements, with each element corresponding to one image subregion from our image processing algorithm. B: we assigned a coupled solid mixture constitutive model to the geometry, with material axes assigned per element based on local collagen directions measured in our image analysis. A toe-linear fiber material was used to represent collagen fibers, and a Mooney–Rivlin material was used to represent the remaining ground substance of the extracellular matrix (ECM). C: boundary conditions were assigned to simulate an equibiaxial 20% engineering strain by prescribing displacements to the +y and +x mesh surfaces corresponding to the top and right edge of the SEM image, respectively. The −y surface was fixed in y, the −x surface was fixed in x, and the +z surface was fixed in z. D: Cauchy stress and Lagrange strain, in the longitudinal/muscle fiber (x) and transverse/cross-muscle fiber (y) directions, were output for each element and averaged to determine the stress-strain curve for each model. Average element stress was measured at 20% strain in the longitudinal (S_long) and transverse (S_trans) directions. Effective stiffness in the longitudinal (k_long) and transverse (k_trans) directions were measured with a linear fit at the 18%, 19%, and 20% strain time points. Effective stiffness ratio was then calculated for each model ($k_{\text{r}}=\frac{k_{\text{trans}}}{k_{\text{long}}}$).

Material law.

Connective tissue is often modeled as a composite of collagen fibers embedded in an isotropic “ground matrix” (45–48). The “ground matrix” is referred to as a gel-like amorphous substance and contains all nonfibrillar components of the ECM (e.g., proteoglycans and glycosaminoglycans) (15). Collagen fibers contribute only to the tensile properties of the ECM. Their stress-strain behavior exhibits distinct toe and linear regions under tensile deformation as collagen fibers uncrimp and straighten (49). To represent the skeletal muscle ECM, we assigned a coupled solid mixture constitutive model to the geometry. Collagen fibers were modeled as a toe-linear fiber (50), where the strain energy density is a function of the fiber stretch λ. A transition from the toe to linear region occurs at λ₀, where β is the power law exponent in the toe region and E is the linear fiber modulus. The remaining ground substance of the ECM was modeled as a Mooney–Rivlin material (51) where c₁ and c₂ are the Mooney–Rivlin material coefficients and K is a bulk modulus-like penalty parameter for the coupled solid mixture. Specific details and constitutive equations of these materials can be found in the FEBio user manual (help.febio.org).

Material parameters.

The toe region of the collagen fiber stress-strain curve is associated with both the straightening of wavy collagen fibers (52, 53) and collagen fiber realignment (54), often modeled with λ₀ = 1.06 (55). We explicitly modeled collagen fiber straightness by discretizing our image into subregions and measuring local collagen directions. Since collagen fiber straightening is already accounted for in our model, we used a constant stretch ratio of λ₀ = 1.01, such that it only accounts for the contribution of collagen fiber realignment within the image subregions. Due to the wide range of values for collagen fiber stiffness and ground matrix stiffness reported in the literature, we varied the ratio of E to c₁ and conducted a sensitivity analysis described in detail in Supplemental Fig. S3 (see https://doi.org/10.6084/m9.figshare.17026457.v1). Based on that analysis, we saw that the influence of collagen fiber stiffness on the effective stiffness ratio began to stabilize when the ratio of E to c₁ was greater than 500. Therefore, we selected a collagen fiber modulus 800 times greater than the ground matrix stiffness (E = 800 MPa, c₁ = 1 MPa). We then selected a bulk modulus to maintain incompressibility by ensuring volume preservation (<2% increase in the average volume across elements) (K = 100,000 MPa) (48). The purpose of our FE models was to isolate the effect of ∼1 μm scale structural changes of collagen fiber organization on ∼100 μm scale tissue mechanical behavior. Therefore, the material parameters shown in Table 1 were held constant for all SEM image-based models. Material axes were then assigned per element to reflect the subregion collagen direction measurements (α_i) obtained from our image processing algorithm, with fiber angles in the x-y plane (Fig. 3B). The x-axis corresponded to the longitudinal (muscle fiber) direction, the y-axis corresponded to the transverse (cross-muscle fiber) direction, and the z axis was orthogonal to the x and y axes.

Table 1.

Material parameters for coupled solid mixture material in SEM-image based mechanical models

Toe-Linear Collagen Fiber
E	Fiber modulus in the linear range	800 MPa
β	Power-law exponent in the toe region	3
λ0	Stretch ratio when toe region transitions to the linear region	1.01

Mooney–Rivlin Ground Matrix
c 1	Mooney–Rivlin c1 parameter	1 MPa
c 2	Mooney–Rivlin c2 parameter	0 MPa
K	Bulk modulus	100,0000 MPa

Boundary conditions.

We assigned boundary conditions to simulate an equibiaxial 20% engineering strain (Fig. 3C), by prescribing displacements to the +y and +x mesh surfaces corresponding to the top and right edges of the SEM image, respectively. The −y surface was fixed in y, the −x surface was fixed in x, and the +z surface was fixed in z.

Model outputs.

Cauchy stress and Lagrange strain, in the longitudinal (x, muscle fiber) and transverse (y, cross-muscle fiber) directions, were output for each element and averaged to determine the stress-strain curve for each model (Fig. 3D). Effective stiffness in the longitudinal (k_long) and transverse (k_trans) directions were measured with a linear fit of the 18%, 19%, and 20% strain points and effective stiffness ratio was then calculated for each model (k_r = k_trans/k_long) (Fig. 3E). A sensitivity analysis was performed to determine the influence of applied strain percentage on model outputs of transverse and longitudinal stiffness. We found that model outputs of stiffness stabilized between 15% and 20% strain. Furthermore, strain percentage did not influence the trends between model outputs and thus our overall conclusions were not sensitive to the strain value. We normalized model outputs of stress and stiffness by collagen fiber modulus (E = 800 MPa) due to uncertainty in estimates for our material parameters.

Structure-Function Relationships between Collagen Organization and ECM Stiffness

To determine if ECM-level properties could be predicted by collagen fiber level organization, we first compared our measurements of collagen fiber organization from the SEM images with mechanical properties output from the corresponding SEM image-based models (Fig. 8). This analysis allowed us to ask questions such as, “Do changes in collagen fiber straightness, collagen fiber direction, or collagen fiber alignment measured in SEM images predict changes in effective stiffness predicted in FE models?” To predict the influence of specific parameters of collagen organization on mechanical properties, we then simulated changes in each parameter alone in “simplified” images that we manually generated. We then used our modeling pipeline to measure the effective stiffness in FE models based on each simplified image. This allowed us to determine “theoretical” structure-function relationships (Fig. 9) and answer questions such as, “How do collagen fiber straightness and collagen fiber direction influence ECM stiffness independently?” By comparing the theoretical relationships with SEM image-based models, we then asked questions such as, “Do SEM image-based models follow theoretical structure-function relationships?.”

Figure 8.

Relationships between collagen fiber organization measured from scanning electron microscopy (SEM) images and tissue level properties predicted from mechanical models. Linear relationships for transverse/cross-muscle fiber direction and longitudinal/muscle-fiber direction stiffness vs. collagen fiber straightness and alignment were calculated with linear regression (A–F). Relationships for stiffness ratio (G–I) were calculated by dividing fits for transverse stiffness (A–C) by fits for longitudinal stiffness (D–F).

Figure 9.

Theoretical relationships between collagen fiber organization from simplified images and tissue level properties predicted from mechanical models (see Supplemental Materials S4). Relationships for transverse and longitudinal stiffness vs. collagen fiber straightness and direction were calculated with nonlinear regression (A–D) and relationships for stiffness ratio (E and F) were calculated by dividing fits for transverse stiffness (A and B) by fits for longitudinal stiffness (C and D). Power law relationships between collagen fiber straightness and model outputs with collagen direction constant at 90° (solid lines) and 70° (dashed lines) are shown in B, D, and F. Exponential relationships between collagen fiber direction and model outputs with collagen fiber straightness constant at 1 (solid lines) and 0.85 (dashed lines) are shown in A, C, and E. R² > 0.95 for all fits.

Simplified image-based models.

First, we created images with varied collagen fiber direction (5° < cfd < 85°), with collagen fiber straightness constant at 1.0 and then 0.85, since the straightness parameters from SEM images fell within this range (Supplemental Fig. S4A; see https://doi.org/10.6084/m9.figshare.17026409.v1). Next, we varied collagen fiber straightness (0.589 < P_s 0.997) while holding fiber direction constant at 90° and then at 70°, since the collagen fiber directions from SEM images fell within this range (Supplemental Fig. 4B). We generated FE models from each simplified image and plotted the key model outputs (transverse/cross-muscle fiber stiffness, longitudinal/muscle-fiber stiffness, and stiffness ratio) versus the parameter that was varied in the simplified image. We then conducted nonlinear regression analysis to fit theoretical curves to the relationships between each key model output and collagen fiber direction (Supplemental Fig. S4C), as well as collagen fiber straightness (Supplemental Fig. S4D). For all theoretical models, we matched the collagen fiber stiffness and ground matrix stiffness from the SEM image-based models (E = 800 MPa, c₁ = 1 MPa). We selected a bulk modulus to ensure incompressibility with the criteria described earlier for the SEM image-based models. For the simplified image-based models, a bulk modulus two orders of magnitude greater than the ground matrix stiffness was sufficient to meet our criteria (K = 100 MPa) (Supplemental Materials S4). The image processing and modeling code are available at https://github.com/ridhisahani/sem-fem.

Statistical Analysis

Comparison of image measurements and model outputs between groups.

A two-way analysis of variance (ANOVA) with age (3, 6, and 12 mo) and group (mdx vs. WT) as factors was performed for the following measurements: 1) collagen pixel ratio, 2) collagen fiber direction relative to muscle fiber direction, 3) collagen fiber straightness parameter, 4) collagen fiber strength of alignment, 5) longitudinal/muscle-fiber direction effective stiffness, 6) transverse/cross-muscle fiber direction effective stiffness, and 7) effective stiffness ratio. Assumptions of random sampling, equal variance, and normality of residuals were confirmed with qq plots and distribution plots. When applicable, Tukey honestly significant difference (HSD) post hoc comparison was performed to determine which groups were significantly different. Alpha was set at 0.05 for all tests.

Structure-function relationships between image measurements and model outputs.

Nonlinear and linear regression were used to fit relationships between model outputs and collagen fiber organization [MATLAB (R2018b) and Curve Fitting Toolbox 3.5.8] (56). For the SEM image-based models, we fit linear relationships between image measurements (collagen fiber direction, collagen fiber straightness, and collagen fiber alignment) and ECM effective stiffness (transverse/cross-muscle fiber direction and longitudinal/muscle-fiber direction), confirming assumptions of linear regression. For our models based on the simplified images, the assumptions for linear regression were no longer met. Therefore, we fit power law and exponential relationships between the parameters varied in the simplified images (collagen fiber direction and collagen fiber straightness) and model outputs of ECM effective stiffness (transverse and longitudinal). Power law relationships were better fits between collagen fiber straightness and ECM effective stiffness (transverse and longitudinal), and exponential relationships were better fits between collagen fiber direction and ECM effective stiffness (transverse and longitudinal). For all models, both SEM image-based and simplified image-based, we determined the stiffness ratio by dividing the curve fits for transverse effective stiffness by the curve fits for longitudinal effective stiffness.

RESULTS

Collagen Fibers Were Straighter and More Highly Aligned in Mdx and Older WT Mice but Retained a Transverse Orientation Relative to Muscle Fibers

Changes in collagen fiber organization can be detected visually in the SEM images (Fig. 4). Collagen fiber straightness was significantly greater in mdx over WT at 3 mo (mdx = 0.976 ± 0.0108, WT = 0.887 ± 0.0309, P = 3.3e-6) and at 6 mo (mdx = 0.942 ± 0.0182, WT = 0.881 ± 0.0163, P = 1.0e-3). Collagen fiber straightness was also significantly greater in 12-mo-old WT (0.931 ± 0.0289) over 3-mo-old WT (0.887 ± 0.0309), (P = 0.027), as well as 12-mo-old WT (0.931 ± 0.0289) over 6-mo-old WT (0.881 ± 0.0163), (P = 0.0090) (Fig. 5A). Collagen fiber strength of alignment was significantly greater in mdx over WT groups at 3 mo (mdx = 0.876 ± 0.0333, WT = 0.759 ± 0.0416, P = 3.0e-5) and at 6 mo (mdx = 0.840 ± 0.0315, WT = 0.759 ± 0.0368, P = 4.5e-3) (Fig. 5B). Collagen pixel ratio ranged from 0.47 to 0.61, with no significant differences between age or disease groups (Fig. 5C). Collagen fiber direction relative to muscle fiber direction ranged from 70° to 90°, with no significant differences between age or disease groups (Fig. 5D).

Figure 4.

Diaphragm muscle samples were dissected from the costal region and collagen structure was isolated after muscle fiber digestion. Samples were imaged with scanning electron microscopy (SEM, Zeiss Sigma VP HD field SEM) and images collected at ×1,000 were used to visualize collagen fiber organization. We assume a stress-free, fully relaxed configuration for our samples from 3-, 6-, and 12-mo old Mdx and wild-type (WT) (n = 6) mice.

Figure 5.

A: collagen fiber straightness (P_s) measured for three fibers per image and averaged. B: collagen strength of alignment (SA) used to measure the circular spread in local orientations per image (0 < SA < 1, 1 = high alignment). C: collagen pixel ratio, quantified per image as the number of collagen pixels divided by the number of total pixels. D: mean collagen orientation relative to the muscle fiber direction, measured as the average of local directions measured per image window. E: peak radon transform value per image, measured as the average of the subregion maximum radon transform intensities. Dashed line representing peak transform value of test images of perfectly aligned and straight fibers. Significance between groups shown with bars, where P < 0.05 and n = 6 mice/group.

The Mechanical Models Predicted That Longitudinal Effective Stiffness Was Greater in WT Mice Compared to Mdx Mice, While Transverse Effective Stiffness and the Ratio of Transverse to Longitudinal Effective Stiffness Was Greater in Mdx Mice Compared to WT Mice

Variations in collagen fiber organization measured in SEM images were reflected qualitatively in the element stresses in the FE models (Fig. 6). Longitudinal (muscle fiber direction) effective stiffness was significantly greater in WT over mdx at 3 mo (mdx = 0.143 ± 0.0464, WT = 0.268 ± 0.0451, P = 1.21e-4) and at 6 mo (mdx = 0.177 ± 0.0330, WT = 0.268 ± 0.0383, P = 5.93e-3) (Fig. 7A). Transverse (cross-muscle fiber direction) effective stiffness was significantly greater in mdx over WT at 3 mo (mdx = 0.646 ± 0.0826, WT = 0.500 ± 0.0699, P = 6.75e-3) and at 6 mo (mdx = 0.633 ± 0.0492, WT = 0.487 ± 0.0692, P = 7.15e-3) (Fig. 7B). For all SEM image-based models, the effective stiffness ratio was greater than 1, indicating greater stiffness in the direction transverse to the muscle fibers. The effective stiffness ratio was also significantly greater in mdx over WT at 3 mo (mdx = 5.45 ± 2.04, WT = 1.97 ± 0.670, P = 1.32e-4) and at 6 mo (mdx = 4.05 ± 0.985, WT = 1.96 ± 0.506, P = 3.5e-2) (Fig. 7C).

Figure 6.

Finite-element models based on images of 3-mo-old mdx and wild-type (WT) mice (seen in Fig. 4). Element stress normalized by collagen fiber stiffness is plotted in the longitudinal/muscle-fiber (top) and transverse/cross-muscle fiber (bottom) directions.

Figure 7.

A: longitudinal/muscle-fiber direction effective stiffness at 20% strain, normalized by collagen fiber stiffness. B: transverse/cross-muscle fiber direction effective stiffness at 20% strain, normalized by collagen fiber stiffness. C: stiffness ratio quantified as the transverse stiffness divided by longitudinal stiffness. Significance between groups shown with bars, where P < 0.05 and n = 6 mice/group. WT, wild type.

Collagen Fiber Straightness and Collagen Fiber Alignment Were Significant Predictors of SEM Image-Based Model Outputs

There were positive linear relationships between collagen fiber straightness and transverse (cross-muscle fiber direction) effective stiffness (k_trans = 1.827 × P_s − 1.122, R² = 0.6) and between collagen fiber alignment and transverse (cross-muscle fiber direction) effective stiffness (k_trans= 1.423 × R − 0.5731, R² = 0.8) (θ and C). There were negative linear relationships between collagen fiber straightness and longitudinal (muscle fiber direction) effective stiffness (k_long = −1.189 × P_s + 1.1323, R² = 0.7) and between collagen fiber alignment and longitudinal (muscle fiber direction) effective stiffness (k_long = −1.09 × R + 1.097, R² = 0.9) (Fig. 8, E and F). The data points follow the effective stiffness ratio determined from the linear fits (k_ratio = k_trans/k_long) (Fig. 8, G–I). There were no significant relationships between collagen fiber direction and longitudinal or transverse effective stiffness (Fig. 8, A, D, and G).

Theoretical Structure-Function Relationships between Collagen Fiber Organization and Tissue Stiffness Were Determined from Simplified Image-Based Models

FE models based on simplified images predicted power-law relationships between collagen fiber straightness and effective stiffness and exponential relationships between collagen direction and effective stiffness (Fig. 9).

Collagen fiber direction versus effective stiffness.

The transverse (cross-muscle fiber direction) stiffness values from the SEM image-based models were better approximated by the theoretical curve with collagen fiber straightness constant at 1 (Fig. 9A). Both theoretical curves predicted lower values for longitudinal (muscle fiber direction) effective stiffness than the SEM image-based models (Fig. 9C). The effective stiffness ratio values from the SEM image-based models were better approximated by the theoretical curve with collagen fiber straightness constant at 85° (Fig. 9E).

Collagen fiber straightness versus effective stiffness.

The transverse (cross-muscle fiber direction) stiffness values from the SEM image-based models were better approximated by the theoretical curve with collagen fiber direction constant at 90° (Fig. 9B). Both theoretical curves predicted lower values for longitudinal (muscle fiber direction) effective stiffness than the SEM image-based models (Fig. 9D). The effective stiffness ratio values from the SEM image-based models were better approximated by the theoretical curve with collagen fiber direction constant at 70° (Fig. 9F).

DISCUSSION

In this study, we tested the hypotheses that collagen structure within the ECM is altered in DMD and that these changes have implications on the mechanical properties of the ECM. We first visualized collagen structure with SEM images and then developed an analysis framework to quantify collagen organization and explore the influence of our measurements on ECM mechanics (Figs. 1, 2, and 3). The image analysis reveals that collagen fibers within the diaphragm muscle epimysium are oriented transversely, with increased collagen fiber straightness and alignment with age and disease (Fig. 5). From the SEM image-based mechanical models, we predict that transverse (cross-muscle fiber direction) effective stiffness is also increased with age and disease. In addition, both healthy and diseased models reveal an increase in transverse (cross-muscle fiber direction) effective stiffness relative to longitudinal (muscle fiber direction) effective stiffness, with the ratio of transverse to longitudinal effective stiffness increased with disease (Fig. 7). Collagen fiber straightness and alignment measured in the SEM images were significant predictors of transverse and longitudinal stiffness output from our models, whereas collagen direction was not (Fig. 8). From the models based on simplified images, we predict theoretical power law relationships between collagen fiber straightness and effective stiffness and theoretical exponential relationships between collagen direction and effective stiffness (Fig. 9).

Our Findings Implicate Changes in ECM Structure and Mechanics on the Mechanical Properties of Diaphragm Muscle

As DMD progresses in patients, pulmonary function declines (57) as the diaphragm muscle weakens with age (13, 58). To understand how tissue level properties lead to changes in muscle function, animal models such as the mdx mouse allow us to measure tissue properties with disease. Mechanical properties of diaphragm muscle are often measured from uniaxial strip tests and show a decrease in elasticity and contractile force in mdx diaphragm (21, 59). However, unlike most skeletal muscles, the diaphragm sustains biaxial loads in vivo and exhibits nonuniform and anisotropic behavior (60–62). From uniaxial tests of rat diaphragm muscle, Boriek et al. (61) report that extensibility of diaphragm tissue is decreased when loaded uniaxially transverse to muscle fibers, than when loaded uniaxially along muscle fibers. From biaxial tests of healthy canine diaphragm muscle, Boriek et al. (62) reported that samples were stiffer and more nonlinear transverse to muscle fibers than along the muscle fiber direction. In our models, we simulated a 20% equibiaxial test and predicted that both healthy and mdx tissue was stiffer transverse to the muscle fiber direction, similar to Boriek et al.’s findings (62). Although our models of healthy murine tissue predicted about a two times increase in transverse relative to longitudinal stress, Boriek et al. reported about a five times increase in transverse to longitudinal stress in healthy canine tissue at 20% longitudinal strain when a load corresponding to 20% transverse strain was applied. Differences in the experimental conditions and animal models may account for the decreased stiffness ratio predicted in our healthy models and future studies in murine diaphragm tissue are needed to make direct comparisons with our model predictions.

In our study we only modeled the ECM, focusing on the effects of collagen organization, and found that WT tissue was stiffer than mdx in the longitudinal direction. Indeed, previous studies report that longitudinal muscle stiffness is greater in mdx tissue than WT and Stedman et al. reported that the dynamic elastic modulus of mdx diaphragm muscle was more than 30 times greater than WT (21). Although we expect that in mdx muscle tissue both longitudinal and transverse stiffness are greater than in WT muscle tissue, we posit that the increase in transverse passive stiffness is due to the organization of collagen fibers. Surprisingly, Smith and Barton (17) did not find any significant differences in longitudinal passive stiffness in mdx and healthy diaphragm muscle tissue although collagen area fraction was approximately four times greater in mdx relative to WT mice at 12 mo. Based on the findings presented here, the increase in collagen amount may play a larger role on the transverse properties in the diaphragm muscle, requiring biaxial mechanical testing to elucidate the role of collagen on passive mechanics.

The Methods Presented Here Offer a Novel Framework to Explicitly Model the Influence of ECM Structure on Mechanical Properties

In a previous constitutive model of epimysium from rat tibialis anterior muscle, Gao et al. (63) represented collagen fibers with unit cells and assigned unit cell angle from collagen fiber distributions that were measured experimentally (36) but were not specific to the epimysium. In the image-based modeling pipeline presented here, we explicitly modeled collagen organization by assigning fiber directions in each finite element, allowing for a framework that can be easily translated. In the SEM images, we found an increase in collagen fiber straightness in older healthy mice relative to younger healthy mice (Fig. 5A), but this difference was not reflected in the model predictions (Fig. 7). In both images and models, we did not find any differences between diseased and healthy groups at 12 mo, suggesting that the ECM of older healthy mice resembles that of diseased mice. A benefit of our modeling pipeline is that we can relate collagen fiber-level structural parameters to tissue-level mechanical parameters to predict structure-function relationships. In the models based on SEM images, significant relationships were seen between collagen fiber straightness and alignment with effective stiffness, but not collagen fiber direction (Fig. 8). Theoretically, we would expect collagen fiber direction to be an important predictor of tissue properties and the models based on simplified images show an exponential relationship between collagen direction and tissue stiffness (Fig. 9). However, in the SEM image-based models collagen direction was not a dominant factor in distinguishing the properties across samples. This may be due to the fact that collagen fiber direction did not vary greatly between the tissue samples or because collagen alignment and straightness accounted for key differences between samples and had a greater impact on stiffness. For all SEM image-based models, the effective stiffness was greater in the transverse/cross-muscle fiber direction relative to longitudinal/muscle-fiber direction (k_r > 1), and collagen fibers were oriented in the transverse direction. Taken together, these findings suggest that although collagen fiber direction determines the direction in which the ECM will be stiffer in tension, collagen fiber straightness and alignment explain differences in model predictions for mdx and WT mice. Another advantage of the methods presented here is that we can isolate structural parameters using simplified images and predict “theoretical” structure-function relationships from our FE models. By comparing these theoretical curves with the SEM image-based models, we can see that accounting for collagen fiber direction or collagen fiber straightness alone is not sufficient for explaining the changes in tissue level properties we predicted with age and disease in our FE models. This suggests that there is isotropy beyond changes in collagen fiber straightness or direction that play a role in distributing stresses throughout the ECM.

There are some limitations of this approach that should be mentioned. Enzymatic and detergent digestions such as the sodium hydroxide protocol used in this study have been shown to alter the mechanical properties of the ECM (29). Although this is what motivated us to use modeling to explore the mechanical implications of the ECM, changes in the structure of collagen fibers may have been influenced by removing all other ECM components. As chemical fixation has been shown to lead to tissue shrinkage (64), there may have been shrinkage in our diaphragm muscle samples before isolating the ECM, as well as during the dehydration preparation for SEM imaging. This tissue shrinkage may have influenced the straightness of collagen fibers, possibly increasing the waviness we detected. Since all samples underwent the same protocol, we assume that these influences were constant between groups.

All samples were physically unconstrained throughout the duration of our protocol. Our images were captured after muscle fibers and additional ECM components were digested, leaving only the collagen fibers. Although all samples were imaged with the epimysium in a stress-free configuration, the collagen fibers may still have some form of stress and we cannot directly relate this configuration to the in vivo muscle tissue state. In limb skeletal muscles, collagen fiber angle is known to vary with muscle fiber length, with collagen becoming more aligned with muscle fibers at greater sarcomere lengths (65, 66). However, since the tissue deformations in the diaphragm muscle likely differ significantly from limb skeletal muscles (67), it is not clear how the collagen fiber alignment changes during diaphragm contraction. Henry et al. (22) reported a decrease in “resting” sarcomere length in mdx diaphragm muscle relative to WT, suggesting that mdx diaphragm sarcomeres operate on different regions of the force-length curve. This, combined with our results, certainly indicate that significant structural changes in the muscle associated with disease have occurred and future work should examine how these changes influence in vivo function.

In the FE models, we grouped all collagen subtypes in one material and did not account for changes in other ECM components or crosslinking of collagen fibers, although increased collagen crosslinking has been previously reported in mdx mouse diaphragm tissue and muscle from patients with DMD (68). For this reason, it is important to acknowledge that the results presented in this work focus on the effects of collagen fiber organization alone. In addition, we did not measure mechanical properties of our decellularized samples directly. Although simplistic, this allowed us to focus specifically on relating structural parameters at the collagen fiber level to bulk level tissue properties. Thus, the modeling results presented here are theoretical in nature and provide interesting hypotheses for future experiments to examine changes in passive mechanics of the dystrophic diaphragm.

The Skeletal Muscle ECM is a Complex Three-Dimensional Scaffold That is Organized Uniquely across Muscle Groups

Skeletal muscle fibrosis is often characterized with images of muscle cross-sections showing a honeycomb structure of the perimysium, surrounding muscle fascicles, and endomysium, surrounding muscle fibers (36, 69, 70). The structure of the endomysium is similar across skeletal muscles (71), while differences in perimysium and epimysium are reported across muscle groups and with disease. Borg and Caulfield (71) report that perimysium from diaphragm muscle is less developed than other skeletal muscle groups where large bundles of collagen fibers are seen, arranged both parallel and circumferential to muscle fibers. An increase in the number of such “perimysial collagen cables” has been reported with fibrosis (72), but their prevalence in diaphragm muscle fibrosis remains unknown. In our study, we collected images of the diaphragm muscle epimysium, the outermost layer of the ECM surrounding skeletal muscle. Differences in the structure of the epimysium are also reported across skeletal muscles, where a cross-ply arrangement of wavy collagen fibers oriented ∼55° to the muscle fiber direction is reported in long strap-like muscle and a dense layer of collagen fibers aligned parallel to muscle fibers is reported in pennate muscle (69). These muscles sustain uniaxial loads in vivo, and thus collagen fibers aligned with the muscle fiber direction may mainly contribute to the longitudinal, or along muscle fiber properties. Gao et al. (73) report that the outer layer of the epimysium in rat tibialis anterior muscle consists of wavy collagen fibers highly aligned in a “predominant direction” but do not report the direction relative to muscle fiber direction or quantify collagen alignment or straightness. Our SEM images show a similar arrangement to Gao et al. in the diaphragm muscle epimysium, but with collagen fibers oriented transverse to muscle fibers. As the diaphragm sustains biaxial loads in vivo, this suggests that the epimysium may mainly contribute to its transverse or cross-muscle fiber properties. We hypothesize that the arrangement of collagen fibers we measured in the epimysium is unique to the diaphragm muscle and highlights the need to study the arrangement of collagen fibers in each muscle before we can determine how changes during fibrosis affect its mechanical properties.

Compared with other skeletal muscles, the diaphragm is thin and relatively flat with a larger surface to volume ratio. Thus, the epimysium spans a greater area in the diaphragm than in other skeletal muscles and may be responsible for bearing a greater amount of load relative to the peri- and endomysium than in other muscle groups. Griffiths et al. (74) report the presence of an elastin-rich layer of connective tissue on the thoracic surface of sheep diaphragm muscle, which provides elastic recoil and reduces stress on the diaphragm muscle. A similar structure has not been reported in mouse diaphragm muscle, nor did we notice the presence of one during our muscle isolation. However, the epimysium may serve a similar purpose and the increased collagen fiber straightness and alignment we measured in mdx epimysium may increase stress on the diaphragm muscle. Although the epimysium may play a larger role in the diaphragm muscle than in other muscles, future studies are still needed to explore changes in collagen organization within the intramuscular ECM layers. Prior studies report that the endomysium is composed of a planar network of irregularly wavy collagen fibers (36, 69), and thus the changes in alignment and straightness that we measured in the epimysium may also be implicated in these layers in the diaphragm muscle. Relative to the epimysium, the intramuscular ECM spans a larger surface area as it surrounds individual muscle fibers and fascicles. A large variation in the amount and composition of the intramuscular ECM is noted across muscle groups (75) and we must consider the relative amounts of each ECM layer to determine their role on passive muscle properties. Before we can extrapolate our predictions of tissue-level mechanical properties to increased muscle-level stiffness reported in patients with DMD, we must consider the three-dimensional (3-D) arrangement of collagen fibers within each ECM layer (epimysium, endomysium, and perimysium) relative to the muscle fiber microstructure.

The Skeletal Muscle ECM is Essential for Transmitting Forces from Muscle Fibers

Muscle fibers can transmit force both longitudinally along the muscle fiber axis, and laterally to adjacent muscle fibers (76–78). Lateral force transmission occurs through physical linkages between the actin cytoskeleton and ECM at the muscle fiber membrane (79). This notion is supported by the ability of the endomysium to transmit forces between intrafascicularly terminating muscle fibers through shear (78) and the physical continuity of the perimysium from muscle to tendon (80). Huijing (81) describes “epimuscular myofascial force transmission” as the force transmitted between muscle and its surroundings through the epimsuyim. This idea is supported through experiments where force transmission still occurs after tendonotomy (82, 83) and differences in proximal and distal force are measured when the muscle-tendon complex length is held constant (84). Such experiments provide strong evidence of force transmission through pathways other than the myotendinous junctions and highlight the role of the ECM. Damage to connective tissue is shown to hinder force transmission and we must consider how fibrosis affects the ability of the ECM to transmit forces (84). In mdx mice, where the muscle fiber membrane is weakened due to the lack of dystrophin, lateral force transmission is severely impaired (77). The diaphragm muscle has a complex architecture with a majority of intrafascicularly terminating muscle fibers (85), suggesting that the ECM is especially critical for lateral force transmission. The SEM images in this study revealed that collagen fibers are oriented transverse to the muscle fibers, suggesting that they may serve as a direct pathway for transmitting forces laterally between muscle fibers. Thus, changes in the alignment or straightness of these transversely oriented collagen fibers may influence the ability of the ECM to transmit lateral forces between neighboring muscle fibers from muscle fibers to tendon.

We Must Consider the Role of Collagen Organization on Respiratory Insufficiency in DMD

To hypothesize the implications of our findings on respiration, we must first consider the unique architecture of the diaphragm muscle. The diaphragm is a dome shaped, “sheet-like” muscle, with a central tendon connecting to the costal and crural muscle domains (86). In our study, we collected samples from the costal region. In this region, the muscle fibers are arranged radially from the central tendon to insertion at the rib cage. Therefore, if we consider our results at the whole muscle level we expect that collagen fibers in the epimysium are arranged circumferentially to maintain the transverse orientation we measured at high magnifications. Collagen fibers are responsible for generating force when stretched in tension, suggesting that the epimysium limits circumferential expansion of the diaphragm muscle. Therefore, we reason that the epimysium may regulate the ability for the diaphragm muscle to return to its fully relaxed configuration during expiration, preventing it from moving through the proper excursion lengths contributing to insufficiency. To better understand the mechanical role of the epimysium on respiratory insufficiency, we must characterize the changes in biaxial properties of diaphragm muscle with fibrosis and develop methods to quantify in vivo motion of the diaphragm during respiration. We must also consider the mechanical roles of the intramuscular ECM layers on respiratory insufficiency, as there may be reorganization of collagen fibers within these layers during fibrosis as well.

Future Work Should Consider the Unique Structure and Mechanics of Fibrotic Tissue in Therapies for DMD

Despite progress in recent years, DMD remains a fatal condition, with fibrosis a key contributor to muscle dysfunction and hypothesized to decrease the effectiveness of therapeutics. Antifibrotic therapies target inflammatory pathways such as TGF-β, but their effectiveness is measured by decreasing levels of collagen expression, without accounting for changes in ECM structure or mechanics (87, 88). The importance of mechanical and structural properties of the ECM is well documented in the literature, with ECM stiffness and alignment key regulators of cellular behaviors involved in fibrosis, such as fibroblast alignment, migration, and proliferation (24, 89). Our study reveals changes in both ECM structure and mechanics in fibrosis, highlighting the need to study the role of therapeutics on collagen organization. Furthermore, we must account for differences in collagen organization between tissue systems, especially as we aim to alleviate the deleterious impacts of fibrosis in DMD in the diaphragm muscle.

GLOSSARY

Longitudinal/muscle fiber direction: direction of muscle fibers, aligned with the horizontal axes

Transverse/cross muscle fiber direction: direction transverse to muscle fiber direction, aligned with vertical axes

Collagen fiber direction (0° ≤ cfd ≤ 90°): average of collagen fiber directions measured in each image subregion reported relative to muscle fiber direction, automatically output from image processing algorithm

Collagen fiber straightness (0 ≤ P_s≤1): straightness of collagen fibers (P_s = L_o/L_f), where L_o = linear end-to-end straight-line length of collagen fiber and L_f = collagen fiber path length, measured manually in ImageJ

Collagen fiber strength of alignment (0 ≤ SA ≤ 1, 1 = high alignment): alignment of collagen fibers across image subregions calculated from resultant vector length ($\text{SA}=\overline{\mathit{r}})$, automatically output from image processing algorithm

Transverse effective stiffness (k_trans): tissue-level stiffness in the cross-muscle fiber direction calculated from mechanical models at 20% applied biaxial strain

Longitudinal effective stiffness (k_long): tissue-level stiffness in the muscle fiber direction calculated from mechanical models at 20% applied biaxial strain

Effective stiffness ratio: transverse effective stiffness divided by longitudinal effective stiffness (k_ratio= k_trans/k_long

SUPPLEMENTAL DATA

Supplemental Fig. S1: https://doi.org/10.6084/m9.figshare.17026361.v1.

Supplemental Fig. S2: https://doi.org/10.6084/m9.figshare.17026388.v1.

Supplemental Fig. S3: https://doi.org/10.6084/m9.figshare.17026457.v1.

Supplemental Fig. S4: https://doi.org/10.6084/m9.figshare.17026409.v1.

DATA AVAILABILITY

SEM images collected in this study are available at https://doi.org/10.6084/m9.figshare.14398184.v1, and image processing and modeling code is available at https://github.com/ridhisahani/sem-fem.

GRANTS

Funding for this work was provided by Grant U01AR06393 from the National Institutes of Health, and a graduate fellowship from a T32 Biotechnology Training Grant.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

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

ENDNOTE

At the request of the authors, readers are herein alerted to the fact that additional materials related to this manuscript may be found at https://github.com/ridhisahani/sem-fem. These materials are not a part of this manuscript and have not undergone peer review by the American Physiological Society (APS). APS and the journal editors take no responsibility for these materials, for the website address, or for any links to or from it.