Three-dimensional alignment of microvasculature and cardiomyocytes in the developing ventricle


A toolset to understand coronary microvasculature and cardiomyocytes organization

A solubilized lipophilic dye (DiI) was perfused into the coronary vessels via the aorta to stain the vasculature through the entire thickness of the embryonic heart (Fig. 1a). This procedure was performed at two stages with different maturation in a quail embryo model after coronary circulation had been established (embryonic day 9 and 13)39,40. The hearts were then optically cleared using a new optical clearing technique, termed lipid-preserving index matching for prolonged imaging depth (LIMPID) developed by our group41 and imaged using confocal microscopy (Fig. 1b,c).

Figure 1

Methods summary. (a) Quail hearts were perfused with DiI, (b) optically cleared, (c) and imaged using confocal microscopy to detect both coronary microvasculature and cardiac nuclei. (d) Vessels were segmented, (e) and automatically skeletonized resulting in vessel-nodes, -segments and -points exported for analysis. A vessel-vector (red arrow) was defined for each vessel by connecting the start and end nodes. (f) The helical angle (αH) was calculated for each vessel-vector based on its position with respect to the heart’s surface. (g) Cardiac nuclei were imaged with DAPI. (h) A Fourier transform (FFT) was applied to each image and the resulting spectrum was multiplied by (i) a series of masks with different axial angles. (j) The resulting angular amplitude A(θi) was plotted as a function of angle, (k) and fit to a von Mises distribution.

Following imaging, the coronary vessels were automatically segmented and skeletonized (Fig. 1d,e), and each vessel helical angle (αH) was calculated (Fig. 1f). The helical angle was chosen to describe vessel orientation since it is often used when describing the alignment of cardiac muscle fibers33. To determine each vessel’s helical angle, the reference unit vectors g1, g2, g3 were calculated for each vessel-segment based on the direction normal to the heart surface (g1) the direction running from heart apex to base (g2), and a circumferential direction (g3) orthogonal to g1 and g2 (Fig. 1f). The helical angle (αH) was calculated for all vector segments as the elevation angle from the horizontal (g1–g3) plane (Fig. 1f) as was previously done in other studies of the heart structure32.

The cardiomyocyte nuclei were stained with DAPI to compare microvascular organization with the surrounding cardiac cells during rapid ventricular wall proliferation and thickening. Images of cardiac nuclei (Fig. 1g) obtained in the same hearts as the vascular images were processed using a two-dimensional (2D) Fourier transform (Fig. 1h–j) and fit to a von Mises distribution to determine nuclei orientation (Fig. 1k).

The von Mises distribution characterizes angular distribution of random processes and was used to describe our vascular and nuclei data. A bimodal von Mises probability distribution function (pdf) as a function of angle θ (in radians) is as follows42:

$$pdf(theta ) = frac{1}{{2pi I_{0} (kappa )}}e^{kappa cos (theta – mu )} + frac{1}{{2pi I_{0} (kappa )}}e^{kappa cos (theta + pi – mu )}$$

(1)

where κ measures the concentration of the distribution around its mean (the inverse of the dispersion), μ is the circular mean and I0(κ) is the modified Bessel function of order 0. The von Mises distribution approaches a Gaussian distribution in the limit where κ is large (in which case κ ~ 1/σ2, where σ is the standard deviation), but approaches a uniform distribution in the limit where κ goes to zero. The standard deviation is ill-defined for bimodal von Mises distributions42 and can lead to unintuitive results when used in data with low concentration (κ ~ 0) and poor directionality43. Thus, we did not calculate the standard deviation or the standard error as a measure of the uncertainty on the mean. Instead, we calculated a standard uncertainty u using the von Mises probability density function pdf(θ) so that the probability P(−u < θ < u) = 68%. In the limit where κ is large, the von Mises distribution resembles a Gaussian distribution and u approaches the standard deviation.

Visualizing embryonic coronary microvasculature

Day 9 and 13 embryonic hearts (n = 4 per developmental stage) showed a dense coronary network throughout the wall thickness previously unreported in the quail at these developmental stages (Fig. 2a,b). Our DiI perfusion and clearing technique resulted in high contrast images with continuous, highly connected vessels with no vessel leakage. Based on relative size of the smallest vessels (Fig. 2c) and the diameter of some blood cell nuclei found in the same sample (Fig. 2d), it is likely that the DiI perfusion enables imaging of vessels down to capillaries. LIMPID reduced light scattering which eliminated background signal from surrounding tissue and enabled intact heart imaging obviating sectioning and registration. Overall, the high image quality permitted automatic 3D vessel segmentation, which is essential to analyze volumetric images in experiments with a large number of animals per cohort.

Figure 2
figure2

Highly developed, dense and organized quail coronary vasculature in the day 9 and 13 embryo. Vessels stained with DiI and color-coded based on depth in the 3D confocal image set. (a) Embryonic day 9. (b) Embryonic day 13. Posterior heart wall from the epicardium to the middle of the ventricular lumen. Two representative hearts shown with n = 4 hearts each for ED9 and ED13. Images colored with FIJI based on z-slice depth from confocal image stack. (c) Diameter of a vessel cross section after DiI perfusion. (d) Short diameter of blood cell nuclei as seen in DAPI images.

Embryonic coronary vasculature progressively changed orientation through the transmural depth of the ventricular wall

Vessel images were analyzed to determine vessel orientation through the left ventricular posterior free wall. The helical angle describes how vessels cross in and out of the horizontal plane, with circumferential vessels at 0°, and vessels that go from the apex to the base at 90°.

To assess the orientation of the ventricular wall vessels at embryonic day (ED) 9 and ED13, the mean helical angle was calculated within a small moving window (approx. 136 × 136 × 175 μm), and each vessel was color-coded based on the result (Fig. 3, Supplemental videos 1, 2). As seen on the multiple cross-sections, the surface vessels started with a downward orientation toward the apex (helical angle is negative, color-coded red) before gradually approaching the horizontal plane deeper into the ventricular wall (0°, black), and continuing into the positive (green) toward the endocardium, with even some vessels becoming entirely vertical from apex to base (+ 90°, white) near the ventricular lumen (Fig. 3). This trend was detected in both ED9 and ED13 hearts, even though the overall heart shape matured by ED13 to become more like the adult quail heart.

Figure 3
figure3

Mean coronary microvasculature orientation in the left ventricular posterior wall. Hearts at two developmental stages (left: ED9, right: ED13) with vessels color-coded based on the local mean helical angle αH, as defined by the coordinate vectors g1–g2–g3. In gray, cross-section views for each heart. Two horizontal cross-sections (g1–g3 plane) at different locations between the apex and the base. One vertical cross-section (g1–g2 plane) intersecting the ventricle from the apex to the middle of the left ventricular base. Volumetric heart surface images (orange) created from the DAPI images with the location of each cross-section indicated (black planes). Two representative hearts shown with n = 4 hearts each for ED9 and ED13. Epi Epicardium, Endo Endocardium. Images rendered in Amira with angles calculated in MATLAB.

To understand how individual microvessels contributed to the mean vessel orientation, we focused on a central region throughout the depth of the posterior left ventricular free wall at ED9 (Fig. 4) and ED13 (Fig. 5).

Figure 4
figure4

Quantifying vessel orientation as a function of depth at embryonic day 9. (a) Heart surface created from DAPI images with region-of-interest (ROI) indicated with black box. (b) Mean helical angle (red line, left y-axis) calculated at each depth with a moving window (+ /− 15 μm). Standard uncertainty shown as error bars. The concentration parameter κ was obtained from a von Mises fit performed at each depth (black dashed line, right y-axis). Layers of interest indicated with grey/white shaded regions. Helical angle of 0° indicated with a dashed red line. (c) Vessel images with color coding for helical angle at increasing depths (panels 1–5, depth range indicated in b). White arrows indicate large vessels. (d) Helical angle distribution matched to above panels in c (same color-scale). (e) Transmural cross-section of vessels from the epicardium (Epi) to the endocardium (Endo). Black tick marks separate the five regions in depth shown in b. One representative heart shown with n = 4 hearts for ED9. Images rendered in Amira with angles calculated in MATLAB.

Figure 5
figure5

Quantifying vessel orientation as a function of depth at embryonic day 13. (a) Heart surface created from DAPI images with region-of-interest (ROI) indicated with black box. (b) Mean helical angle (red line, left y-axis) calculated at each depth with a moving window (+ /− 15 μm), with the standard uncertainty as error bars. The concentration parameter κ was obtained from a von Mises fit performed at each depth (black dashed line, right y-axis). Layers of interest indicated with grey/white shaded regions. Helical angle of 0° indicated with a dashed red line. (c) Vessel images with color coding for helical angle at increasing depths (panels 1–5, depth range indicated in b). White arrows indicate large vessels. (d) Helical angle distribution matched to above panels in (c) (same color scale used). (e) Transmural cross-section of vessels from the epicardium (Epi) to the endocardium (Endo). Black tick marks separate the five regions in depth shown in (b). One representative heart shown with n = 4 hearts for ED13. Images rendered in Amira with angles calculated in MATLAB.

For the ROI indicated in Figs. 4a and 5a, the mean helical angle (overline{{alpha_{H} }}) was calculated at all depths (Figs. 4b and 5b, left y-axis, red line). To quantify vessel angle distribution for each depth, the concentration parameter κ was obtained from the von Mises fit. High κ indicates vessel angles concentrated around the mean (i.e., vessels are aligned with each other), while low κ indicates vessels oriented in all directions equally (Figs. 4b and 5b, right y-axis, black dashed line). Five layers of the ventricular wall (Figs. 4b and 5b, gray/white shaded area) were identified for analysis based on the concentration parameter κ.

The most superficial layer, which starts at the epicardium and ends 90 μm into the myocardium for ED9 and 130 μm for ED13, was characterized by highly aligned vessels (high κ) with negative but rapidly changing helical angles (Figs. 4c and 5c, first panel). A second highly aligned layer (90–225 μm in depth for ED9, 130–270 μm for ED13) contained vessels oriented closer to the horizontal plane (color-coded black, Figs. 4c and 5c, second panel). Polar histograms of vessel angle distributions (Figs. 4d, 5d) showed that vessels in the first and second layers are oriented on average at − 27° and − 13° for ED9 (− 23° and − 12° for ED13). A decrease in vessel alignment (concentration κ diminished) marked the beginning of a third layer (ED9: 225–375 μm, ED13: 270–515 μm) where larger blood vessels were observed (Figs. 4c and 5c, white arrows). These larger vessels did not strongly align with the more numerous smaller vessels surrounding them. Finally, the fourth and fifth layers were defined as being between 375–555 μm and > 555 μm in depth at ED9 (515–700 μm and > 700 μm at ED13). Fourth layer vessels started to lose their alignment with each other (sudden decrease in κ) and the mean helical angle became more strongly positive (color-coded green, Figs. 4c and 5c, fourth panel). This organization continued into the fifth layer where vessels became more vertical with low alignment.

Most vessels are not strictly parallel to the heart surface but are often penetrating the ventricular wall, which is not reflected in the helical angle. As seen from the transmural view (Figs. 4e, 5e), individual vessels were more likely to be parallel to the heart surface near the epicardium and endocardium, with vessels in the middle traveling into the myocardium regardless of their helical angle.

In summary, coronary vessels in the left ventricular posterior wall were highly aligned with each other, especially in more superficial layers. The vessels gradually changed their orientation counterclockwise with depth at both ED9 and ED13. This alignment was detected in the microvasculature and was not strongly present in the larger coronary vessels.

Cardiomyocytes in the embryonic heart progressively changed orientation through the transmural depth of the ventricular wall

Coronary vessels develop in close relationship with cardiomyocytes. An embryonic heart immunostained for MF20 confirmed that most DAPI-stained cell nuclei in our images were cardiomyocytes (Supplementary Fig. S1a). Another embryonic heart immunostained for NCAM to delineate the plasma membrane confirmed that embryonic cardiomyocytes contained one oblong nucleus, and the nuclear longitudinal axis corresponded to the cardiomyocyte longitudinal axis (Supplementary Fig. S1b). Thus, it was assumed that nuclei orientation in the ventricular wall represented the cardiomyocytes orientation for our analysis.

Cardiac nuclei images were acquired for all hearts (ED9 and ED13, n = 4 hearts/group) immediately following DiI vascular imaging using a 20X objective. Images were acquired in the same ROI in the left ventricular posterior wall as for vessels. A representative heart is shown in Fig. 6, with the location of the ROI indicated in Fig. 6a.

Figure 6
figure6

Quantifying cardiac nuclei orientation as a function of depth at embryonic day 9. (a) Frontal section of the heart created from DAPI images with region-of-interest (ROI) indicated with the black box. Image colorized in FIJI. (b) Mean cardiac nuclei orientation μ calculated at each depth z (black line) + /− standard uncertainty u with five representative points selected at different depths. Black dashed line indicates 0° (horizontal). (c) Heart ventricle diagram showing the depth axis direction into the left ventricular wall. (d) Five DAPI images corresponding to points in b showing individual cardiac nuclei at different orientations (images used for analysis are 360 × 360 μm). (e) Polar plot showing the corresponding angular distribution of the 2D Fourier transform spectrum (black dots), the von Mises fit (black line), and the mean cardiomyocyte angle (red line) for each depth shown in (d). One representative heart shown with n = 4 hearts for ED9.

The mean cardiomyocytes orientation (overline{{A(theta_{i} )}}) was obtained using the 2D Fourier transform method. A von Mises fit was also performed to obtain the concentration parameter κ, from which the standard uncertainty u was calculated. The mean cardiomyocyte orientation through the ventricular wall depth is shown in Fig. 6b and depth axis orientation indicated in Fig. 6c. To demonstrate how cardiomyocytes orientation at each depth is obtained from individual confocal z-stack images, five images at different depths within the myocardium are presented in Fig. 6d. Nuclei were visibly elongated and aligned with each other, and their orientation gradually changed counterclockwise as depth increased, with an average orientation of − 40° near the epicardium and + 79° near the endocardium. For each image the angle distribution A(θi) was plotted on a polar histogram (Fig. 6e, black dots), with the corresponding von Mises fit (Fig. 6e, black line) and mean cardiomyocyte orientation (Fig. 6e, red line). The von Mises fit shape indicated how aligned the cardiomyocytes were with each other, with highly aligned cells leading to an elongated “figure-eight” shape (Fig. 6e, black line), and low alignment leading to a more circular shape (Fig. 6e, fourth panel).

In summary, cardiomyocytes in the left ventricular posterior free wall already appeared elongated and aligned with each other at these early developmental stages, and their orientation gradually changed counterclockwise with depth at both ED9 and ED13.

Cardiomyocytes were aligned with the coronary microvasculature at all depths throughout the left ventricle

To assess how coronary microvasculature is oriented with respect to the surrounding cardiomyocytes, we compared DiI-stained vascular images (10× magnification) with DAPI-stained nuclei images (20× magnification) for the same ROI in the left ventricular posterior wall. To compare the helical vessel angle (3D analysis) to cardiomyocyte orientation (2D analysis), the helical angle was projected onto the two-dimensional heart surface and became the projection angle αP.

The mean vessel projection angle (overline{{alpha_{P} }}) (red) was plotted alongside the mean cardiomyocyte angle (blue) for all corresponding depths for a representative ED9 (Fig. 7a) and ED13 (Fig. 7b) heart. The two lines overlapped at all depths, demonstrating that cardiomyocytes were aligned with coronary microvasculature.

Figure 7
figure7

Cardiac nuclei and coronary vessels have the same orientation throughout the left ventricular wall. Mean projection angle for the vessels (red line) and mean orientation of cardiomyocyte nuclei (blue line) as a function of depth into the posterior left ventricular wall for (a) ED9 and (b) ED13, with standard uncertainty u as error bars. ρ is the Pearson’s correlation coefficient. (c) Pearson’s correlation coefficient between the mean vessel angle and the mean cardiomyocyte angle as a function of depth for all n = 4 ED9 hearts and all n = 4 ED13 hearts. Red data points indicate the hearts represented in (a) and (b). ρ = 1 is a perfect linear correlation. (d) Confocal microscopy image acquired at 20 × magnification in an ED13 heart with DAPI (blue) and DiI (red) signal. The DAPI signal was acquired at depth z = 496 μm, and the DiI signal was averaged over the range z = 482 μm to z = 511 μm. Image processed in FIJI.

The Pearson’s correlation coefficient ρ was calculated to quantify how well the mean vessel orientation was correlated to the mean cardiomyocyte orientation throughout the ventricular wall, and thus how well these structures were aligned with each other.

Coefficients of ρ = 0.985 and ρ = 0.966 were obtained for the representative ED9 and ED13 hearts (Fig. 7a,b), indicating an almost perfect correlation between vessel and cardiomyocyte orientation. The coefficients of all other ED9 and ED13 hearts were also high (Fig. 7c) indicating that this relationship holds across the two different developmental stages. These results indicated that coronary vessels and cardiomyocytes are aligned even at early developmental stages. A representative image showing aligned coronary vessels and cell nuclei (Fig. 7d) also confirms these results.

Coronary vessels orientation is consistent across individual hearts at both developmental stages

The mean vessel helical angle as a function of depth was compared for all hearts at ED9 (Fig. 8a) and ED13 (Fig. 8b) to determine if coronary vessel organization was consistent across animals. The thickness of the posterior left ventricular wall was normalized from 0 to 1 to facilitate comparison (total wall thickness varied from 500 to 600 μm at ED9, and 800–1,000 μm at ED13).

Figure 8
figure8

Comparison of mean vessel orientation, vessel alignment and vessel density across hearts. Mean coronary vessel helical angle as a function of depth in the posterior left ventricular wall for n = 4 hearts at (a) embryonic day (ED) 9, and (b) ED13. Depth across the ventricle wall is normalized for all hearts from 0 (closest vessels to epicardium) to 1 (closest vessels to endocardium). A.u. is arbitrary units. Black dashed line indicates the horizontal at 0°. (c) Concentration parameter κ averaged over the first 60% of the tissue starting from the epicardium (left) and the last 40% of the tissue until the endocardium (right) for all hearts at ED9 and ED13. Mean for each group indicated by gray line. (d) Vessel volume fraction (left) and vessel length density (right) for all hearts at ED9 and ED13. Mean for each group indicated by gray line. Groups compared using a two-tailed T-test assuming unequal variance.

At ED9, the coronary vessel orientation (mean helical angle) was similar across hearts near the epicardium and for most of the myocardium, but discrepancies across hearts were detected in the inner myocardium. In comparison at ED13, the mean helical angle was similar at all depths across all hearts.

To compare the degree of coronary vessel alignment across individual ED9 and ED13 hearts, the mean concentration parameter κ was calculated (Fig. 8c). As seen previously (Figs. 4b and Fig. 5b), the vessels were more aligned (concentration κ was higher) in more superficial layers of the myocardium, and less aligned (κ decreased) in the two inner most layers adjacent to the endocardium. Vessels were more aligned (higher κ) in ED13 hearts as compared to ED9, suggesting that coronary vessel alignment increased as the heart matured and became more organized. Additionally, we observe that values of κ are approaching zero near the endocardium at ED9 (Fig. 8c) which indicates random vessel orientation and an overall loss of directionality. As a result, the mean helical angle near the endocardium at ED9 (Fig. 8a) behaves erratically and is not consistent from heart to heart. However, these are qualitative observations and a statistical test was not performed to compare the mean concentration parameter κ between ED9 and ED13 hearts, since κ could not be assumed to be normally distributed, and κ was a measure of variance. We could not find a statistical test appropriate for this situation.

In summary, the coronary microvasculature showed strong alignment, with the left ventricular vessels rotating counterclockwise as a function depth. Vessels closer to the epicardium (60% of wall thickness) may have been more strongly aligned. Vessels may have become more aligned with increasing developmental stage (ED13 as compared to ED9).

Vessel density remained the same between two developmental stages

To further quantify vessel organization in the developing ventricular wall, vessel density was calculated at both ED9 and ED13 using two methods: vessel volume fraction (no units) and vessel length density (units of mm−2) (Fig. 8d). The accuracy of the vessel volume fraction can be negatively affected by imaging conditions such as overexposed vessels, while the accuracy of the vessel length density is dependent on proper segmentation and skeletonization of the vessels. Thus, both methods were employed to counterbalance each method’s limitations. Both methods indicated no difference in vessel density as a function of developmental age. A two-tailed T-test assuming unequal variance between groups showed no statistically significant differences (p = 0.47 and p = 0.33 for volume fraction and vessel length density respectively). Thus, coronary vessels did increase in number in the growing ventricular wall, but the overall vessel density remained the same at ED9 and ED13.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *