2.1. Mitral valve chordae samples

This study was approved by the Japanese Red Cross Kobe Hospital Research Subjects Review Boards (#40, 10/7/2015) and the SPring-8 Proposal Review Committee (2015B1508, 2016A1165, 2017B1163, 2019B1542, 2021B1449), and informed consent was obtained from all patients. The investigation conformed to the principles outlined in the Declaration of Helsinki. Ruptured mitral chordae (n = 6) were obtained from patients who had undergone elective mitral valve surgery for MR at the Japanese Red Cross Kobe Hospital. Resected mitral chordae were fixed with 10% buffered formalin, and 3D images were obtained within six months using XPCT (named as the rupture group). The RCT were located on the medial side of the posterior mitral leaflet (P2) in five chordae and on the medial side of the anterior mitral leaflet (A2) in one chorda. In addition, 12 formalin-fixed normal chordae tendineae were obtained from the hearts of cadavers without heart valve disease (named as the control group). The mean age of patients in the rupture group was 70.2 ± 3.0 years (range 65–72 years) and that of patients in the control group was 67.2 ± 14.1 years (range 38–83 years). Patient characteristics are presented in Table 1.

2.2. X-ray phase-contrast tomography imaging

The XPCT system used at the SPring-8 synchrotron radiation facility is based on an X-ray Talbot grating interferometer in the bending-magnet synchrotron beamline (BL20B2), and has been described in detail elsewhere (Hoshino et al., 2016, 2017). The main specifications of the grating interferometer used for X-ray phase tomography at the X-ray energy of 18.8 keV are as follows:

(i) G1 as a phase grating – material Ta, pattern thickness 2.1 µm, grating period 10 µm.

(ii) G2 as an absorption grating – material Au, pattern thickness 16.6 µm, grating period 10 µm.

On the other hand, the specifications of another grating interferometer for 20 keV are as follows:

(i) G1 – material Ni, pattern thickness 3.7 µm, grating period 2.4 µm.

(ii) G2 – material Au, pattern thickness 25 µm, grating period 2.4 µm.

Briefly, the system was located 43 m from an X-ray source, and tuned to 18.8 keV or 20 keV by passing it through a silicon double-crystal monochromator. The specimen was embedded in agarose gel and placed in an appropriately sized plastic canister filled with saline solution. Then, the specimen was rotated slowly in the canister using a rotating stage. The examination room was maintained at a constant temperature. A Talbot grating interferometer comprising a phase grating and an absorption grating was placed behind the specimen, and Moiré fringes generated by the interferometer were detected by an X-ray detector. A fringe scanning method was used for phase retrieval where the number of phase stepping was set to five. Phase retrieval was attained using a five-step phase-stepping procedure that involved shifting the absorption grating with a piezo-driven stage. The data were processed by a high-throughput system to create 3D images. The specifications of the system were as follows: field of view, 13.3 mm (H) × 5.0 mm (V) with the voxel size of 6.5 µm or 14.1 mm (H) × 5.0 mm (V) with the voxel size of 3.4 µm; target density range, 0.9–1.2 g cm⁻³; density resolution, 1 mg cm⁻³; exposure time, 100–400 ms per image for static observations. Due to the improvement of the setup of the grating interferometer, the energy used and the specification of the grating interferometer optimized for that differ depending on the timing of the measurement. However, for all measurement conditions, the calculated density values were calibrated by measuring a standard sample of known density, in this case a salt solution with different concentrations. We used saline solutions of three different densities (1.026 g cm⁻³, 1.053 g cm⁻³ and 1.102 g cm⁻³) to determine the ratio of the experimental values to the theoretical values. We have confirmed that the plot of the experimental values compared with the theoretical values is linear when the densities are changed. In cases involving the observation of a chorda tendinea longer than the effective vertical field view, the specimen was scanned step-by-step vertically, with a stepping amount of 4 mm per tomographic scan to observe its whole structure. A conventional filtered back-projection (FPB) method with a Chesler filter was used to calculate the reconstruction of the cross-sectional images. Any filters or image processing to remove the ring artifact were not applied. The phase image was calculated by integrating the differential phase image in one direction. The following relation between the density difference (Δρ) and the refractive index difference (Δδ) from the background was used to calculate the density,

Here, λ is the X-ray wavelength, r_e is the classical electron radius, N_A is Avogadro's constant, and Z/M is the ratio of the number of electrons to the molecular weight of the sample, here set to 0.55. Then, the density was calculated from ρ = ρ_BG + Δρ, where ρ_BG is the density of the background. In X-ray phase tomography, the sample was measured in a dedicated plastic container filled with normal saline. Therefore, the density of the background ρ_BG corresponds to that of normal saline in this study.The reconstructed volumes were converted into 16-bit TIFF images of each tomography slice with custom-made software. Further image processing and analysis of the tomography data was performed with ImageJ (https://rsbweb.nih.gov/ij/index.html) (Momose & Fukuda, 1995; Yagi et al., 1999).

Morphological differences were quantitatively evaluated by estimating the mean mass density within a region of interest in the chordae tendinea for each formalin-fixed sample. Each region of interest was a cross-sectional area × 100 voxels as depth. The mass density of the chordae tendinea was measured at around the papillary muscle attachment site in the normal chordae tendinea as control. In ruptured chordae tendinea, the density was measured at two portions, including the ruptured portion located near the papillary muscle attachment site and the non-ruptured portion near the mitral leaflets.

2.3. Histological examinations

After XPCT imaging was completed, the samples were stored in 10% formalin and sectioned transversely. Thereafter, the sections of each specimen were stained with hematoxylin and eosin, and Sirius red. The Sirius red stains were used to detect elastic fibers and collagen fibers. Tenomodulin was identified by immunostaining with an anti-tenomodulin antibody (Sigma-Aldrich, St Louis, MO, USA) and used to detect the healing process in chorda tendinea (Shukunami et al., 2016). The histological examination focused on the identification restoration process, including inflammation and abnormalities in tissue components such as collagen and elastic fibers. Moreover, to evaluate the contribution of collagen fibers to densitometric changes in the chorda tendinea, the occupation ratio of collagen fibers was determined in the sample using ImageJ. The occupation ratio was defined as the area occupied by collagenous components as a percentage of the selected area.

2.4. Statistical analysis

All statistical tests were conducted using the JMP statistical software (SAS Institute, Cary, NC, USA). Students' t-tests were employed to compare the densities in the chorda tendinea. The significance was set at P < 0.05 for all the statistical analyses. Data are presented as mean ± standard deviation values.

3.1. Representative XPCT Images

Fig. 1(a) clearly demonstrates that XPCT can provide a clear and detailed three-dimensional image of a chordae tendinea from the papillary muscle to the leaflet of a normal mitral valve. After XPCT imaging, the target marginal chorda tendinea was selected in an axial image [Fig. 1(b)], and, as explained in the Methods section, its density was obtained from TIFF images of each tomography slice sample. The tissue density in the chordae tendinea of a normal mitral valve was almost homogeneous, with a single peak in the density histogram at approximately 1.0816 g cm⁻³ [Fig. 1(c)].

Figure 1 (a) Clear and detailed three-dimensional imaging of a chordae tendinea from the papillary muscle to the leaflet of a normal mitral valve obtained by synchrotron-based X-ray phase computed tomography. (b) A selected marginal chorda tendinea (red circle) in an axial image after the imaging process. (c) Calculating density from TIFF images of each tomography slice of the selected samples.

3.2. Quantitative assessment of the chorda tendinea

To quantitatively evaluate morphological differences, the mass densities of the chorda tendinea were estimated in each sample. In the control group (n = 12), the mean density of the chorda tendinea was 1.085 ± 0.015 g cm⁻³, with no significant variation within the chorda. These findings were consistent with the histological findings for the same region, wherein uniform high-density collagenous fibers were observed in the media (Fig. 2). These findings may show that the density of the chorda tendinea is homogeneous, providing the chorda with a stable structure. In the rupture group, the mean density of the ruptured portion of the chorda tendinea was 1.029 ± 0.004 g cm⁻³, significantly lower than the mass density for normal mitral chordae (p < 0.0001) (Fig. 3). The differences in the mass density between the normal and the ruptured chorda were consistent with the distribution of collagenous fibers observed in the associated histological samples (Fig. 2). The distribution of mass density in the ruptured chordae, i.e. not including portions that were not ruptured, was measured by analyzing each ruptured chordae tendinea [Fig. 4(a)] at six points. There was a significant difference in the density of the ruptured portion and non-ruptured portion (1.029 ± 0.004 versus 1.063 ± 0.007 g cm⁻³, respectively; p = 0.005) [Fig. 4(b)]. Figs. 5(a) and 5(b) show formalin-fixed specimens of the ruptured chordae, and XPCT image and mass densities of ruptured and non-ruptured portions. The differences in the mass density were consistent with the histological findings, demonstrating that pathogenesis of RCT might be a local event rather than occurring along the entire length of the chordae tendinea.

Figure 2 An axial image of the XPCT (a, e), HE (b, f), Sirius red (c, g), and polarization of the Sirius red (d, h) in a normal chorda (upper row) and ruptured chorda (lower row). The mass density was 1.092 g cm−3 in the normal chorda (a) and 1.034 g cm−3 in the ruptured chordae (e). The differences in the mass density between the normal and ruptured chorda were consistent with the distribution of collagen fibers observed in the associated histological samples. XPCT – X-ray phase computed tomography; HE – hematoxylin-eosin.

Figure 3 Comparison of the mass density between the chorda rupture group and control group. The mass density based on the synchrotron-based X-ray phase computed tomography in the ruptured mitral chordae tendineae was significantly lower compared with normal chorda tendinea. *p < 0.05.

Figure 4 The distribution of the mass density in the ruptured chordae tendinea. (a) Analyzing the mass density of the ruptured portion and non-ruptured portion in the ruptured chordae tendineae. (b) Comparison of the mass density between the ruptured portion and non-ruptured portion in the ruptured chordae tendineae. There was a significant difference in the density of the ruptured portion and non-ruptured portion. *p < 0.05.

Figure 5 Sagittal section of the ruptured chordae stained with hematoxylin-eosin (HE) and polarization of the Sirius red (SR) staining, and the collagen occupation ratio. (a) There was a non-longitudinal abnormal interstitial tissue layer consistent with the pathological cross-sectional findings (HE). (b) Inflammatory infiltration was not seen in the entire layer, including the abnormal interstitial tissue and longitudinal collagen bundle (HE). (c) SR staining revealed that the non-longitudinal abnormal connective tissue layer included elastic fibers (black color lesion) and collagen fibers (yellow color lesion; type I collagen, green color lesion; type III collagen). (d) The collagen occupation ratio was calculated in the two selected areas, including an area of abundant abnormal connective tissue (yellow rectangle #2) and an area of normal collagen bundle (yellow rectangle #1) (SR staining). (e) The collagen occupation ratio was 61.1% in the area of normal collagen bundle and 17.2% in the area of abundant abnormal connective tissue.

In addition, to assess the effect of aging on the mass density of mitral chorda, a comparison was made of the density of chordae tendinea from normal mitral valve between samples from people aged 75 years or older (aged group; n = 5) and samples from those aged below 75 years (non-aged group; n = 7) in the control group. The mass density of the chordae tendinea was 1.078 ± 0.012 g cm⁻³ in the aged group and 1.089 ± 0.016 g cm⁻³ in the non-aged group, i.e. slightly lower in the aged group. However, there was no statistically significant difference between the two groups (p = 0.1836) (Fig. 6).

Figure 6 Comparison of the mass density between samples aged 75 years or older (aged group; n = 5) and samples less than 75 years old (non-aged group; n = 7) in the control group. The mass density of the chordae tendinea was slightly lower in the aged group, but there was no statistically significant difference between the two groups.

3.3. How the collagen in the obtained CT looks compared with histology

Images from histological examinations of normal and ruptured chordae tendinea are shown in Fig. 2. In the normal chordae, highly dense collagen fibers occupied almost all cross-sectional surfaces and the constituted collagen bundle [Figs. 2(b) and 2(c)]. In the Sirius red staining, black lesions found on subendothelial small interstitial tissues indicated elastic fibers [Fig. 2(c)]. Polarization of the synchrotron radiation staining demonstrated that collagen fiber (type III), stained in green, occupied most of the collagen bundle, along with a small interstitial tissue area beneath the endothelial layer [Fig. 2(d)]. By contrast, ruptured chorda stained with hematoxylin and eosin showed increased abnormal interstitial tissue growth surrounding the collagen bundle [Fig. 2(e)]. Synchrotron radiation staining of the ruptured chordae showed an increase in abnormal interstitial tissue, with more elastic fibers and fewer collagen fibers compared with the adjacent collagen bundle [Figs. 2(f) and 2(h)]. The obtained CT images reflected the differences of constituent elastic and collagen fibers in normal and ruptured chordae tendinea. Normal chorda tendinea showed a uniform high signal indicating collagen bundles, while, in ruptured chorda, core collagen was surrounded by a signal reflecting abnormal interstitial tissue, like annual rings on a tree [Figs. 2(a) and 2(e)].

3.4. The difference of collagen occupation ratio in ruptured chordae tendinea in histology

In the sagittal section of the ruptured chordae stained with hematoxylin and eosin, a non-longitudinal abnormal interstitial tissue layer consistent with the pathological cross-sectional findings [Fig. 5(a)] was observed. Inflammatory infiltration was not seen in the entire layer, including the abnormal interstitial tissue and longitudinal collagen bundle [Fig. 5(b)]. Polarization of the Sirius red staining revealed that the non-longitudinal abnormal connective tissue layer included elastic fibers (black color lesion) and collagenous fibers (yellow color lesion; type I collagen, green color lesion; type III collagen) [Fig. 5(c)]. To quantify the difference in distribution of collagen fibers, the collagen occupation ratio was calculated in two selected areas, including an area of abundant abnormal connective tissue [#2 in Fig. 5(d)] and an area of normal collagen bundle [#1 in Fig. 5(d)]. The collagen occupation ratio was found to be 61.1% in the area with a normal collagen bundle and 17.2% in the area with abundant abnormal connective tissue [Fig. 5(e)]. Moreover, immuno­staining with anti-tenomodulin (Tnmd) antibody was performed to detect the healing process in the chorda tendinea. In the normal chordae, Tnmd positive cells were extremely rare and localized in interstitial tissues beneath the endothelial cell layer [Fig. 7(a)]. In the ruptured chordae, however, Tnmd positive scattered cells were often found in the region of increased abnormal connective tissues [Fig. 7(b)], which was viewed as an explanation for any healing processes that had taken place in the ruptured chordae tendinea.

Figure 7 Immunofluorescence staining for tenomodulin. Tenomodulin positive cells were rare and localized in interstitial tissues under the endothelial cell layer (a), whereas, in the ruptured chordae, tenomodulin positive scattered cells were often found to exist in the region of increased abnormal connective tissues (b).