Quantification of the temporal evolution of collagen orientation in mechanically conditioned engineered cardiovascular tissues.

Tissue Culture and Mechanical Conditioning

The methods of scaffold preparation, cell culture, seeding, and mechanical conditioning have been described previously. 35 In brief, human saphenous vein myofibroblasts (p7) were seeded using fibrin as a cell carrier onto rectangular strips (35 × 5 × 1 mm) of non-woven polyglycolic acid scaffolds, coated with a thin layer of poly-4-hydroxybutyrate. 20 In the longitudinal direction, the scaffolds were attached to the flexible membranes of 6-well plates (Flexcell Int., Hillsborough, NC) at the outer 5 mm. After 1 week of culture under these constrained conditions, the engineered tissues were split into two groups (Fig. 1 ). One group ( n = 6) served as constrained control, whereas the other group ( n = 6) was subjected to intermittent uniaxial dynamic straining (4%, 3 h on/off, 1 Hz) using a modified version of a Flexercell FX-4000T straining device 5 (Flexcell Int., Hillsborough, NC). After 2, 3, and 4 weeks of culture the collagen fiber orientation was visualized. Figure 1 Top view of rectangular engineered constructs. The strips were longitudinally constrained at the outer 5 mm at both ends using silicone glue. Samples were either constrained (a) or intermittently loaded (b) in longitudinal direction (d1). Direction d2 represents the direction perpendicular to the long axis. Positive (+90°) and negative (−90°) directions of d2 are indicated Visualization of Collagen Orientation To visualize the collagen fiber orientation, samples were labeled by 15 μ M Cell Tracker Blue CMAC (CTB; Invitrogen, The Netherlands) and 3 μ M CNA35-OG488 as specific vital fluorescent markers for cell cytoplasm and collagen, 26 respectively. CTB and CNA35-OG488 are excitable with multiphoton microscopy and exhibit broad spectra at 466 nm and 520 nm, respectively. An inverted Zeiss Axiovert 200 microscope (Carl Zeiss, Germany) coupled to an LSM 510 Meta (Carl Zeiss, Germany) laser scanning microscope was used to visualize cell and collagen organization. A chameleon ultra 140 fs pulsed Ti-Sapphire laser (Coherent, Santa Clara, CA) was tuned to 760 nm and two photomultiplier tube (PMT) detectors were defined as 435–485 nm for CTB and 500–530 nm for CNA-OG488. Two samples per group were scanned at three different locations: (1) the center point of the tissue, (2) 5 mm left to the center point, and (3) 5 mm right to the center point. At each location a stack of image slices was obtained. The thickness of one image slice was 1 μ m and the maximum thickness of the stack was 100 μ m. Separate images were obtained from each PMT and combined into single images.

Quantification of Collagen Orientation

The collagen fiber orientations were determined by analyzing the individual images of each image stack as described by Daniels et al . 11 In brief, coherence-enhancing diffusion (CED) was applied for denoising to improve the quality of structures in the image without destroying the boundaries of the fibers. 44 Using CED, smoothing occurs along, but not perpendicular, to the preferred orientation of the structures in the image. Subsequently, the local orientations of all collagen fibers were determined by calculating the principal curvature directions from the eigenvalues and the eigenvectors of the Hessian matrix (second order structure). 40 As fibers appear at different widths, the second order derivatives were determined at a scale adaptive to the local width of the fiber. The optimal scale was determined with a contextual confidence measure. 31 At each location a stack of (in-plane) orientation histograms was obtained, representing the statistical distribution of local orientations in each image. To calculate the mean fiber angle and the dispersity of the distributions, circular statistics were required due to the periodicity of the fiber distributions. 23 , 41 , 48 Circular statistics handle periodic data by representing each angle as a unit vector oriented at that angle. To calculate the mean angle, the individual unit vectors are decomposed into vector components on which statistical operations, such as averaging, are permitted (Fig. 2 ). Histograms of collagen orientations were obtained from each multiphoton image, and of each the mean vector was calculated, representing a mean angle α and a mean vector length r. The mean vector length represents a measure for the dispersity of the fiber orientations. A vector length of 1 indicates no variation in fiber orientations (i.e., all fibers perfectly aligned) while a vector length of 0 indicates a random distribution of fiber orientations. Courses of mean angles and vector lengths as a function of tissue depth were obtained per image location. To generalize the results, the vector components at each depth were averaged over all samples per loading condition per time point. Subsequently, averaged courses of mean angles and vector lengths were obtained. It should be noted that when the dispersity of the collagen orientation distribution is large, the distribution becomes essentially random. This diminishes the meaning of the value of the mean angle. Hence, for clarity, orientation angels were not shown when mean vector lengths were consistently lower than 0.2. Figure 2 Example of circular statistics. A circular distribution is represented in a histogram (a) and depicted as unit vectors with corresponding angles of 30 and 60 degrees in a unit circle (b). The mean vector (dashed line) is calculated by averaging the decomposed sine and cosine vector components of the individual vectors. α represents the mean angle and r the length of the mean vector. A vector length of 1 indicates no variation in fiber orientations, while a vector length of 0 indicates a random orientation of fiber orientations Statistics Statistics were performed to determine the effect of mechanical loading on the alignment of the collagen fibers, as described previously. 41 In short, alignment in the longitudinal (d1) direction was tested for significance at each imaging depth in each set of samples per time point and loading condition by performing one-sample t -tests against a reference value of 0 on the cosine components of the mean vectors. A p -value lower than 0.05 was considered statistically significant.