Cardiomyocyte orientation recovery at micrometer scale reveals long-axis fiber continuum in heart walls

Abstract

Coordinated cardiomyocyte contraction drives the mammalian heart to beat and circulate blood. No consensus model of cardiomyocyte geometrical arrangement exists, due to the limited spatial resolution of whole heart imaging methods and the piecemeal nature of studies based on histological sections. By combining microscopy and computer vision, we produced the first-ever three-dimensional cardiomyocyte orientation reconstruction across mouse ventricular walls at the micrometer scale, representing a gain of three orders of magnitude in spatial resolution. We recovered a cardiomyocyte arrangement aligned to the long-axis direction of the outer ventricular walls. This cellular network lies in a thin shell and forms a continuum with longitudinally arranged cardiomyocytes in the inner walls, with a complex geometry at the apex. Our reconstruction methods can be applied at fine spatial scales to further understanding of heart wall electrical function and mechanics, and set the stage for the study of micron-scale fiber remodeling in heart disease.

Keywords: 3D reconstruction; cardiomyocyte geometry; computer vision; fluorescent microscopy; heart wall structure.

Figure 1. Overview of the heart tissue preparation, imaging and analysis pipeline

Top panels from left to right: Representative photographs of a mouse heart before and after applying the CLARITY procedure (Tomer et al, 2014). An illustration of the cleared mouse heart sections analyzed in this study. The long‐axis (LA) sections correspond to longitudinal dissections of the mouse heart through a transverse plane equivalent to the HLA‐4C views of the AHA (or echocardiogram) nomenclature, revealing the right and left ventricular chambers. The short‐axis (SA) sections are dissected at the midventricular plane, equivalent to PSAX‐PML views of the AHA (or echocardiogram) nomenclature (Cerqueira et al, 2002). At least one mouse heart in this study provided four continuous sections, each being approximately 300 μm in thickness, for both the LA and SA analyses (Appendix Table S1; Materials and Methods). After sectioning and WGA staining, the tissue slices were imaged using confocal microcopy spanning the entire length, breadth, and width of the LA and SA sections (Materials and Methods). Using custom‐built algorithms, the individual confocal stacks were preprocessed for contrast enhancement, denoised, and then stitched (Materials and Methods). Bottom panels from right to left: The imaging and preprocessing pipeline results at a 2‐μm voxel isometric resolution of the entire SA and LA sections, up to a depth of about 300 μm in thickness. The estimated structure tensor vectors are visualized as glyphs represented as golden yellow lines for the entire SA and LA sections (Materials and Methods). A representative 3D view (bottom left) of the hand‐segmented cardiomyocytes (gray) from the WGA stain, overlayed with the cell orientations estimated using the true myocyte orientation (purple) and the estimated field orientation (golden yellow) by the structure tensor method (Materials and Methods).

Figure EV1. A comparison of cleared and uncleared mouse heart tissue by imaging

1. A

   A schematic illustration of the CLARITY method. The harvested heart tissues are incubated in a hydrogel/PFA mixture at 4°C for approximately 9 days. Following hydrogel incubation, the solution mixture with the heart tissues is polymerized at 37°C. The heart tissues are then excised from the polymerized hydrogel and shaken at 37°C with a clearing solution until they attain a desirable level of transparency (see Materials and Methods for details).

2. B

   The clarified heart is sectioned along its short‐ or long‐axis. For the short‐axis, a midventricular region that approximates the PSAX‐PML (parasternal short‐axis—papillary muscle level) plane was chosen. For the long‐axis, a transversal plane that represents the A4C (apical four chambers) plane was used.

3. C–E

   A comparison of uncleared and cleared heart tissue sections stained with WGA and the alpha‐actinin antibody.

4. F

   A comparison of 20× (2‐μm isotropic resolution) and 60× (0.663 * 0.663 * 0.79 μm³ x, y, and z resolution, respectively) cleared heart tissue sections stained with WGA. The raw and skeletonized images are shown for each magnification. The visible cell boundaries that are not of cardiomyocytes are marked with arrow heads (red) in the 60× images. The scale bar is 1,000 and 50 μm for the full view and zoomed in regions, respectively, in the WGA uncleared and cleared tissue images. The scale bar for the alpha‐actinin images is 10 μm.

Figure EV2. The structure tensor method for orientation estimation

1. A

   Fractional anisotropy scores are shown using a parula colormap (left) with a value in the range 0–1. The histogram (right) shows the fraction of pixels in each respective FA bin. The majority of the pixels have a high FA score, indicating the presence of dominant local orientations in the tissue stack. The color bar for the FA score is as indicated.

2. B

   Representative 3D views of hand‐segmented cardiomyocytes with randomly assigned colors to each cell (top panel). A representative 3D view of ground truth orientation (purple) based on the second moment matrix for hand‐segmented myocytes (gray) on the left, with the estimated field orientation from the WGA image (golden yellow) using a structure tensor approach (Materials and Methods) in the middle (bottom panel). The magnitude of the difference between the ground truth and the estimated orientation in degrees is illustrated with a graph on the right. The mean difference between two ground truth and estimated orientations is 5.98° ± 2.3°.

3. C

   Helix angle calculation: Masking (left) followed by centroid estimation (middle, top); Masking the short‐axis section and estimating the tangent plane and normal for the penetration axis (middle, bottom). The set of penetration directions is shown on the right, from outer to inner wall.

4. D

   αH plot (left), region of LV wall analyzed (middle) and rate of change of αH calculated over a neighborhood of 15 voxels. The region marked by a dashed box represents the outer wall longitudinal cells, where the rate of change of αH is initially small and then increases sharply as one approaches the middle wall region, after which it plateaus and then increases again.

5. E

   Line fits for αH plots were calculated for the lateral, antero/infero lateral, and anterior/inferior regions, with the extent of the outer wall longitudinal cells shown by length of the first line in each plot.

Figure 2. Cell orientation across a midventricular short‐axis section of the heart

1. A

   An illustration of the angles measured to represent the cell orientations across the ventricular walls. The red box represents a magnified view of the ventricle wall, with the global axes and labels as indicated: C‐Circumferential L‐Longitudinal R‐Radial. The green cylinder and the blue bidirectional arrow represent the cardiomyocyte and its long‐axis, that is, the estimated cell orientation. Φ is the angle between the projection of the estimated cell orientation onto the XY plane and the estimated cell orientation. θ is the angle between the projection of the estimated cell orientation onto the XY plane and the X‐axis direction. αH, the helix angle is the angle between the projection of the cell orientation onto the plane perpendicular to the transmural penetration direction and the circumferential direction.

2. B–E

   A magnified view of the right (green rectangle) and the left (magenta rectangle) ventricular regions for a full view of the WGA stain (middle). The |θ|, |Φ|, and αH angles for the mouse SA sections are shown using a parula colormap. The yellow tones for the |θ| angle represent cell orientation along the global Y‐Axis while the blue tones represent cell orientation along the X‐axis. For |Φ|, the blue and yellow tones represent cells with orientations aligned with the short‐axis section, and orthogonal to it, respectively. The colormaps scale with the angles, as indicated. The scale bar for the full view images is 1,000 μm and for the magnified views is 100 μm.

Figure EV3. WGA staining and angular colormaps of different short‐axis sections

1. A

   A maximum intensity Z‐projection of the WGA‐stained short‐axis section from different mouse hearts as indicated, shown in grayscale, with the |Φ| and |θ| angles for cell orientations shown using parula colormaps, with the color bars as indicated. The scale bar is 1,000 μm.

2. B

   A comparison of αH estimates from raw (red line) and denoised short‐axis section image stacks (black line) are shown for comparison. The region of the LV wall analyzed is shown in the bottom left panel, with a zoom in on the first 200 μm on the bottom right. The transmural penetration direction is from the outer to the inner LV wall. The raw data agree with denoised αH estimates in most of the places; however, we observed that estimation was much more consistent and robust to nonuniform illumination changes with denoised data. The intensity between two fields of view (~ 350 μm in transmural penetration depth) and the αH estimates begin to deviate marking the edges.

Figure EV4. Analysis of a short‐axis section of a rat heart

1. A

   A short‐axis view of the ventricular chambers of a rat heart, sectioned at PSAX‐PML (parasternal short‐axis—papillary muscle level). A maximum intensity projection of the WGA stain is shown in grayscale, with the |Φ| angle shown using a parula colormap. The scale bar is 1,000 μm.

2. B–D

   Zoomed in regions from the left ventricle, septum and right ventricle, with the WGA stain shown in grayscale and the |Φ| angle shown using a parula colormap. The scale bars are 1,000 μm 100 μm for (A) and (B–D), respectively. The color bar is as indicated.

Figure 3. Helix angle plots for left ventricular wall sectors in a short‐axis section

The individual 10° wedge‐shaped sectors are shown in three different groups; anterior/inferior, anterolateral, inferolateral, and lateral regions (top row panel). The wedge sector 3D average αH values were plotted and are highlighted as αH colormaps on a maximum intensity Z‐projection of the WGA‐stained short‐axis section in grayscale. The 3D average αH values* are plotted as a function of the transmural penetration depth walls, depicted as distance in microns along the X‐axis from the outer to the inner walls. The 3D average αH values from micron‐scale and pseudo‐resolution analyses of the selected sectors of the LV region from one representative data set of a short‐axis section (SAS3) are shown in the middle row panel. A zoomed in version of the outer‐ventricular region is provided in the bottom row panels. The region marked with a box in the bottom left panel represents the LV outer wall longitudinal cells. Data information: The values plotted here are derived from combining two sectors, spanning 10° in each, as illustrated.

Figure 4. A detailed view of a long‐axis section

1. A

   WGA stain.

2. B

   Colormaps for the magnitude of the angle with the longitudinal axis (L).

3. C

   Estimated orientations as glyphs.

4. D

   Estimated orientations as streamlines.

Data information: The colors follow a parula colormap, where the blue and yellow colors indicate in and out of the short‐axis plane cell orientations, respectively. Magnified views of the left ventricle (2^nd column), right ventricle (3rd column), septum wall (4^th column) and the apex region (5^th column) are shown for the regions in panel (A). The color bar for the angle with the longitudinal axis is as indicated. The scale bar for the complete view is 1,000 μm and for the zoomed in regions is 100 μm.

Figure 5. Evidence for a long‐axis fiber continuum from the analysis of a long‐axis section

1. A, B

   Estimated myofiber orientations are shown as glyphs and streamlines for the mouse LA sections as indicated. In each panel, a 3D visualization of the orientations is shown using either glyphs or streamlines, with a view obtained by rotation in a clockwise direction shown on the right (Materials and Methods). The colors follow a parula colormap, where the blue and yellow tones indicate cell orientations that are in and out of the short‐axis plane, respectively. These visualizations reveal the continuity of cell orientations across the entire long‐axis section, from base to apex.

Figure EV5. WGA staining and colormaps for the magnitude of the angle with the longitudinal axis, for different long‐axis sections and their connections at the apex

Five mouse heart ventricle long‐axis sections are shown as maximum intensity Z‐projections of the WGA stain, with the magnitude of the angle between the aggregate cell orientation and longitudinal axis shown with a parula colormap marked, with a white dotted box highlighting the apex region under consideration (left column). A zoomed‐in view of the highlighted apex region from the respective long‐axis sections and illustration of long‐axis connections between different ventricular walls is also shown (right column). The continuity of the long‐axis fibers from the septum to the left ventricle outer wall, from the right ventricle inner wall to the left ventricle outer wall, from the right ventricle outer wall to the left ventricle outer wall and from the right ventricle inner wall to the left ventricle inner wall, is highlighted in the first, second, third, and fifth rows, respectively. The colorbar is as indicated, and the scale bar is 1,000 μm.

Figure 6. Analysis of a short‐axis section from the apex region

Representative Z‐planes from four mouse heart apical serial sections are shown with WGA staining (left), and the |θ| (middle) and |Φ| (right) angles related to cell orientation using parula colormaps. The colormap scales are as indicated, with the scale bar being 100 μm.

Figure 7. A model for orthogonal myofiber systems in heart ventricular walls

The composite model is obtained by superposition of the reconstructions of the short‐axis and long‐axis sections from different mouse hearts. The structure tensor‐based orientations are visualized as streamlines (Materials and Methods). The colors follow a parula colormap, where the blue and yellow tones indicate orientations in or out of the short‐axis plane, respectively.