Structural properties in ruptured mitral chordae tendineae measured by synchrotron-based X-ray phase computed tomography

Tissue density measured by synchrotron-based X-ray phase computed tomography is shown to be associated with the fragility of the ruptured mitral chordae tendineae.


Introduction
Rupture of chorda tendinea (RCT) is the main cause of primary mitral valve regurgitation (MR) (Nishimura et al., 2016), with mitral valve repair chosen as the first-line treatment for MR caused by RCT.However, even when the anatomy of the mitral valve complex inside the heart has been established, the association with the structural properties and the rupturing of chordae tendineae has not been fully investigated.Therefore, there is a possibility of a recurrence of MR following conventional mitral valve repair (Nishida et al., 2018;Murashita et al., 2013).
X-ray phase computed tomography (XPCT) using synchrotron radiation is more sensitive than absorption computed tomography (CT), but it is by no means superior in spatial resolution.Furthermore, the high sensitivity of phasecontrast imaging enables the visualization of small density differences in the biological soft tissue (Margaritondo & Meuli, 2003;Momose & Fukuda, 1995).An XPCT using a grating interferometer developed by the Japan Synchrotron Radiation Research Institute (SPring-8) has been used for several soft biological tissues (Hoshino et al., 2012;Tsukube et al., 2017).A physical quantity about 1000 times larger than that in the attenuation is involved in the phase shift; thus, in principle, X-ray phase tomography has a high sensitivity for soft tissues (Margaritondo & Meuli, 2003;Momose & Fukuda, 1995).The ability to visualize small differences in density and fine details in the structure is attributable to high spatial and density resolution (Tsukube et al., 2017).This suggests that the application of XPCT to obtain densitometric data for mitral chordal pathology could be useful for investigating their morphological change.This study aimed to use XPCT to investigate structural changes in the chorda tendinea in RCT through detailed visualization and structural quantification.

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 formalinfixed 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 AE 3.0 years (range 65-72 years) and that of patients in the control group was 67.2 AE 14.1 years (range 38-83 years).Patient characteristics are presented in Table 1.

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(Hoshino et al., , 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 mm, grating period 10 mm.
On the other hand, the specifications of another grating interferometer for 20 keV are as follows: (i) G1 -material Ni, pattern thickness 3.7 mm, grating period 2.4 mm.
(ii) G2 -material Au, pattern thickness 25 mm, grating period 2.4 mm.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 Moire ´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) Â 5.0 mm (V) with the voxel size of 6.5 mm or 14.1 mm (H) Â 5.0 mm (V) with the voxel size of 3.4 mm; target density range, 0.9-1.2g cm À3 ; density resolution, 1 mg cm À3 ; 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 À3 , 1.053 g cm À3 and 1.102 g cm À3 ) 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 (Á) and the refractive index difference (Á) from the background was used to calculate the density, Here, 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 = BG + Á, where 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 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 Â 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.

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.

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 AE standard deviation values.

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 AE 0.015 g cm À3 , 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 AE 0.004 g cm À3 , 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 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 AE 0.012 g cm À3 in the aged group and 1.089 AE 0.016 g cm À3 in the non-aged group, i.e. slightly lower in the  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 À3 in the normal chorda (a) and 1.034 g cm À3 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.aged group.However, there was no statistically significant difference between the two groups (p = 0.1836) (Fig. 6).

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

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 crosssectional 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.

Discussion
Synchrotron-radiation-based XPCT with Talbot interferometry (Momose et al., 2003) has shown promise as a means of visualizing the microstructure of several biological samples (Momose, 2005) and tissues, including the lungs (Momose & Fukuda, 1995), brain (Yagi et al., 1999), eyes (Hoshino et al., 2011), aorta (Zanette et al., 2012) and heart (Kaneko et al., 2017).In principle, synchrotron-radiation-based XPCT is about 1000 times more sensitive compared with conventional X-ray absorption-contrast techniques (Momose & Fukuda, 1995).It has three major advantages: it provides high-resolution and 3D images without destroying/altering the properties of specimens; it enables the visualization of tissue microstructures and mass density without the need for a contrast agent; and it allows quantitative analyses of the densities of the microstructures of various soft tissues, facilitating quanti- 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.

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).
The RCT is a common cause of primary mitral regurgitation, and mitral valve repair has been an established procedure for MR (Nishimura et al., 2016).However, the causes of RCT and structural properties of the mitral chordae tendinea have not been fully elucidated.In this study, we performed a quantitative comparison among chordae tendinea that revealed differences in the microstructure of the chorda tendinea between specimens from patients with MR due to RCT and those with a normal mitral valve.
The tissue density in the chordae tendinea of the normal mitral valve was almost homogeneous, with a single peak in the density histogram of approximately 1.085 g cm À3 .To clarify the effect of aging on the tissue density of the chordae tendinea, the chordae tendinea of the normal mitral valve was divided into two groups and analyzed.There was no statistically significant difference between the tissue density of the chordae tendinea belonging to the aged group (75 years or older) and non-aged group (below 75 years old), even though the density in the younger group tended to be slightly lower.Although the aged mitral chordae tendinea has been reported to be much stiffer and less extensible than younger chordae (Zuo et al., 2016), our results may explain that aging alone may not be a sufficient cause of RCT.
It is worth noting that the ruptured portion of the ruptured chordae tendineae of patients with MR showed a significantly lower tissue density in the chordae tendineae (1.029 AE 0.004 g cm À3 ) when compared with tissue density of the nonruptured portion of the ruptured chordae tendinea, or with that of the chordae tendinea of the normal mitral valve.The differences in the tissue density between the ruptured portion and non-ruptured portion were consistent with the distribution of collagenous fibers observed in the associated histological samples.
A human mitral chorda tendinea comprises an endothelium cell layer, a small number of elastic fibers, and mostly fibrillary collagens (collagen type I and III) (Akhtar et al., 1998;Stolinski, 1996).Collagen fibers are the main constituent of the collagen bundle and maintain the strength of the chorda tendinea (Millington-Sanders et al., 1998;Icardo & Colvee, 1995).By contrast, elastin fibers surrounding the collagen bundle appear as a mesh-like structure that assist in collagen stretching and contraction, and also act as a barrier to physical stimulation between the collagen and the endothelium (Millington-Sanders et al., 1998;Fenoglio et al., 1972).However, elastin fibers rarely contribute to the mechanical properties of chordae tendinea (Lim & Boughner, 1976).A histological examination of the ruptured chordae tendineae in this study revealed a lower proportion of collagen fibers, a layer of unorganized non-longitudinal tissue between the endothelial cell layer, and an organized longitudinal collagen bundle layer, all of which could be consistent with the connective tissue cushion reported by the previous report (Millington-Sanders et al., 1998).Given that the connective tissue cushion does not contribute to structural strength as effectively as original elastic fibers, our findings suggested that the strength of the chorda should be reduced (Millington-Sanders et al., 1998).The obtained CT images showed a uniform high signal in cross-sectional view in normal chordae tendinea, whereas the ruptured chordae tendinea were characterized by the signal of the core cord-like structure and the concentric structures surrounding it.As a result, the obtained CT images were able to detect structural changes in the core collagen surrounded by abnormal interstitial tissues in the ruptured chordae tendinea, which findings are well accorded with pathological findings.Tenomodulin is a type II transmembrane surface molecule (Shi et al., 2017) mainly expressed in avascular tissues, particularly in dense connective tissue such as tendon/ligament, heart valve/chorda and eye (Oshima et al., 2003;Kimura et al., 2008).As Tnmd possesses antiangiogenesis that maintains avascular tissue -and without inflammatory cells or abnormal connective tissues inducing weakness in the histological strength -Tnmd expression is important in maintaining histological strength in the avascular tendon or ligament tissue.The other roles of Tnmd include maturing tenocyte during the restoration process (Tokunaga et al., 2015), and Tnmd overexpression favors tenogenic lineage differentiation (Shi et al., 2017).In our study, increased Tnmd expression was observed in increased abnormal subendothelial connective tissue in ruptured chordae tendinea compared with normal chordae tendinea.These findings could suggest that increased abnormal subendothelial connective tissue in ruptured chordae tendineae is reflected in the chordae during the healing process as lower histological strength, leading eventually to a tear in the chordae tendinea.
Although mitral valvuloplasty is a well established procedure for MR (Lazam et al., 2017), there have also been reports of recurrent MR requiring repeated mitral valve surgery after an initial mitral valvuloplasty (Nishida et al., 2018;Murashita et al., 2013).In this study, we observed the structural properties of ruptured chorda tendineae by evaluating tissue density.In cases of an advanced modality, which could visualize changes in density or associated changes in collagen fibers during open heart surgery or preoperative evaluations, mitral valvuloplasty might be applied clinically, and prophylactic techniques to prevent chordal elongation or rupture after surgical repair could be introduced.
Further structural and molecular understanding of mitral valve pathology and technological advances is required to improve the outcomes of mitral valve surgery.
The presented study included several limitations.First, at 18 in total, the sample size was relatively small.It should be noted, however, that there are few reports that analyze the structural properties of mitral chordae tendineae in human specimens aside from this study.Second, the RCT group in this study included chorda tendinea with attached anterior and posterior leaflets.Ideally, chorda tendinea should be unified only for anterior or posterior leaflets, although it has been reported that there is no significant viscoelastic difference between chorda tendinea for anterior and posterior leaflets in animal durability tests (Chen et al., 2020).Third, ruptured chordae tendinea investigated in this study were collected from patients with chronic MR due to RCT.Therefore, in terms of outcomes, it is difficult to clarify whether our findings were the cause or effect of RCT of the mitral valve.Further studies using the acute phase RCT model will help facilitate a solution to this problem.

Conclusion
Synchrotron-radiation-based XPCT made it possible to measure tissue density in mitral chordae tendineae.The homogeneous density of the normal chorda tendinea is consistent with highly dense collagen fibers in the collagen bundle, and a localized low density of the ruptured portion of the ruptured chorda may reflect a healing process of a previous injury in the chordae tendinea resulting in structural alterations.Therefore, low density in mitral chordae tendineae is associated with the fragility of the ruptured chordae tendineae.

Fig. 1
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 (a)] at six points.There was a significant difference in the density of the ruptured portion and non-ruptured portion (1.029 AE 0.004 versus 1.063 AE 0.007 g cm À3 , 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 3
Figure 3Comparison 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
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
Figure 5Sagittal 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.

Table 1
Patient characteristics.