Three‐dimensional reconstruction of testis cords/seminiferous tubules

Abstract Background Due to the development of novel equipment for the acquisition of two‐dimensional serial images and software capable of displaying three‐dimensional (3D) images from serial images, the accurate 3D reconstruction of organs and tissues has become possible. Methods Based on published studies, this review summarizes techniques for the 3D reconstruction of the testis cords/seminiferous tubules, with special reference to our method using serial paraffin sections and 3D visualization software. Main findings The testes of mice, rats, and hamsters of various ages were 3D reconstructed and species and age differences in the structures of the testis cords/seminiferous tubules were analyzed. Our method is advantageous because conventional paraffin‐embedded normal and pathological specimens may be utilized for the 3D analysis without the need for complicated and expensive equipment. Conclusion By further decreasing the time and labor required for the procedure and adding information on molecular localization, the technique for 3D reconstruction will contribute to the elucidation of not only the structures, but also the functions of various organs, including the testis.


| INTRODUC TI ON
Despite the development of advanced imaging technologies, traditional histology using tissue sections remains the gold standard for morphological tissue assessments in research and clinical practice.
However, tissue sections obtained from three-dimensional (3D) organs and tissues only provide two-dimensional (2D) information. It is very important, but often challenging, to reconstruct missing 3D information in attempts to match the shape observed in a tissue section to the original 3D structure. Manual 3D reconstruction from serial sections is possible, but has seldom been used in morphological studies, possibly due to the extensive amount of labor required. In recent years, with the development of new technologies for the acquisition of 2D serial images and software capable of displaying 3D images from serial images, the accurate 3D reconstruction of organs and tissues has become possible.
Confocal microscopy, 1,2 one of the new imaging technologies, uses a laser beam and confocal pinholes to block out-of-focus light for the acquisition of 2D serial images with high resolution and contrast. The capturing of multiple 2D images at different depths in a tissue containing fluorescent molecules enables the reconstruction of 3D structures without the need for real tissue sections. However, the depth of tissue observable with confocal microscopy is restricted to 250-500 µm due to the attenuation of excitation light and fluorescence. Multiphoton excitation fluorescence microscopy 3 and light sheet fluorescence microscopy 4 increase the observable depth to a certain degree, but require complicated and expensive equipment. Micro-computed tomography (micro-CT) 5,6 is another imaging technology that creates virtual cross sections of organs and tissues with a micrometer-level spatial resolution. It is composed of a rotating X-ray tube and a row of detectors placed in the gantry to measure X-ray attenuation. Multiple X-ray measurements taken from different angles are then processed on a computer using reconstruction algorithms to produce tomographic images. Although the images obtained by micro-CT are currently inferior to real histological sections because they lack color and have lower resolution and contrast, micro-CT may become a useful tool for analyzing 2D and 3D structures in unfixed or fixed normal or pathological organs and tissues without their destruction.
In addition to the development of imaging technology, the advent of software that visualizes a 3D image from serial 2D images has been important for accurate 3D reconstruction. 7 The key features of 3D visualization software are as follows: (1) image registration, which refers to the process of image alignment; (2) segmentation, which is the process of dividing multiple regions of interest in a 2D image; and (3) 3D rendering, which is a 3D computer graphic process that constructs a 3D model and visualizes it in a 2D image. The segmentation step is performed using automatic, semiautomatic, and manual tools. After 3D reconstruction, segmented regions may be used for a number of tasks, such as volumetric, density, and shape analyses.
Due to the development of imaging technology and software, whole organs, such as the brain, 8-10 kidneys, [11][12][13] lungs, 14 uterus, 15 epididymis, 16 ovaries, [17][18][19][20] and testes, 20 have been analyzed in 3D at the micrometer level. This review provides a historical overview of techniques for the 3D reconstruction of the testis cords/seminiferous tubules, with special reference to our method using serial paraffin sections and high-performance 3D visualization software.

| MORPHOLOGYANDFUN C TIONSOF THETE S TIS
Mammalian testes consist of two compartments: the interstitial compartment and seminiferous tubule compartment. [21][22][23] The former contains blood, lymphatic vessels, and nerves. The most frequent cell type in this compartment is the Leydig cell, which secretes testosterone. 24,25 Macrophages are also observed in the interstitial compartment and may play roles in the differentiation of spermatogonial stem cells. 26,27 The seminiferous tubule compartment is the site for spermatogenesis. There are 28 seminiferous tubules per testis for rats and 12 for mice on average, [28][29][30] and they eventually connect to the rete testis while branching. Seminiferous tubules develop from fetal testis cords and consist of a seminiferous epithelium, which is divided into 14 stages in rats and 12 stages in mice according to germ cell associations. [30][31][32][33][34][35] Adjacent stages are aligned along a seminiferous tubule, and the time required for a particular stage to reappear in the same area is called the cycle, while the space occupied by a series of adjacent stages, including all possible types, is called the wave. 31,36,37 A growing body of evidence indicates that region-specific morphological and functional differences exist within the testis. For example, we recently revealed that the sites at which spermatids initially occur in the postnatal mouse testis are preferentially distributed in the upper-medial areas of the testis close to the rete testis, 37 and that the extent of markedly impaired spermatogenesis is significantly greater in tubule areas near the branching points in the mouse model of non-obstructive azoospermia induced by the administration of busulfan. 38 These regional differences may represent regionspecific gene expression that is regulated by particular molecules produced inside or outside the testis. Therefore, obtaining precise 3D information on the structure of the testis as well as cellular and molecular distributions provides a basis for investigations in various fields, including histology, pathology, and developmental biology.

| ME THODFORTHE3D RECON S TRUC TI ONOFTE S TISCORDS/ S EMINIFEROUSTUBULE S
The 3D structure of the whole testis was initially reported in the early 1900s [39][40][41][42] ; however, the majority of studies were on fetal testis cords and limited morphological information was available on the adult seminiferous tubules at that time. Curtis was the first to apply the technique of the 3D reconstruction of serial sections to adult mouse seminiferous tubules in 1918. 43 The testes used were fixed in Flemming's solution, embedded in paraffin, cut into serial 10μm-thick sections, and stained with iron hematoxylin. To analyze a whole tubule, a straight tubule opening into the rete testis was selected in a particular section and continuously followed through a series of sections. With an aid of the optical equipment available at the time, two complete adult mouse seminiferous tubules were manually reconstructed and one testis was estimated to contain 15 highly convoluted seminiferous tubules with three branching points. One and two seminiferous tubules in rabbit and dog testes, respectively, were also reconstructed.
A study by Clermont and Huckins in 1961 on the 3D structure of the rat testis cords/seminiferous tubules was the most impressive among the reconstruction studies using testis serial sections at that time. 29 The testes used were fixed in Zenker's fluid, embedded in paraffin, cut into serial 5μm-thick sections at intervals of 100 μm, and stained with periodic acid-Schiff-hematoxylin (PAS-H). Plane figures on graph papers and 3D models constructed using stacked acrylic plastic sheets were both employed to visualize the structures of the testis cords/seminiferous tubules and elucidate their morphometric parameters. They reconstructed the rat testis cords on embryonic day (E) 17, E19, and postnatal day (P) 0 and analyzed their numbers and distributions. They also reconstructed 2 and 20 seminiferous tubules in P12 and adult rats, respectively. Two types of testis cords/seminiferous tubules were described: "outer cords/ tubules" are in contact with the tunica albuginea, whereas "inner cords/tubules" are not. The next epoch-making study on the 3D structure of mammalian testis cords/seminiferous tubules was not published until 2009, when two groups showed detailed 3D images of the mouse testis cords. 44 To generate 3D reconstruction, images were aligned using Adobe Photoshop CS2 (Adobe Inc.). A custom-written MATLAB program (The Mathworks, Inc.) based on a linear algorithm was used to interpolate between the extracted sections. However, the 3D reconstruction of whole seminiferous tubules in the adult mammalian testis using a confocal microscope has not yet been performed. In the case of mice, the testis is an ellipsoid of 8 × 5 × 5 mm in size; therefore, it is difficult to observe the whole testis, even if made transparent and fluorescence-stained, with confocal microscopy without making real histological sections.

Silva et al. recently investigated seminiferous tubules using
micro-CT. 47 Mouse testes were fixed in 4% PFA, dehydrated, and stained with an alcohol-based iodine agent to enhance soft tissue X-ray contrast and prevent organ shrinkage during imaging. Samples were scanned on a Zeiss Xradia Versa 500 system (Oberkochen, Germany) with the X-ray source operating with an anode voltage at 50 kV and power at 3 W. Images were reconstructed using volume reconstruction software integrated in the Xradia machine and 3D renderings were made using the commercial software VGstudio (Volume Graphics GmbH). Morphometry was performed using the Phong or Scatter HQ algorithm. Using this method on a 4-week-old mouse testis, they obtained micro-CT images that were equivalent to conventional histological sections and a 3D rendering of the surface of seminiferous tubules in contact with the tunica albuginea.
However, the reconstruction of whole seminiferous tubules was not conducted.
We recently reconstructed whole testis cords/seminiferous tubules using serial sections and Amira 6.3.0 (Visage Imaging GmbH), high-performance 3D reconstruction software. 30,37,38,48,49 In mice, the testis and epididymis were dissected out en bloc and fixed in Bouin's solution overnight at room tem- Photoshop 2020 software. Using Amira 6.3.0 software, serial images were automatically aligned with rigid algorithms followed by manual adjustments, and the inside of the outlines of a selected tubule was filled with a particular color using threshold processing and traced from section to section. This procedure, called semiautomatic segmentation, was repeatedly applied to all seminiferous tubules with different colors, and they were then subjected to 3D rendering. To draw the core lines of seminiferous tubules, individual traced tubules in cross sections were shrunk concentrically and reconstructed into a thin tubule, in which the core lines were drawn using the same software. The software sometimes failed to distinguish closely apposed different tubules from the branched portions of a single tubule and drew wrong lines, partly because alignment was slightly inaccurate due to the distortion of sections. This error was corrected manually by tracking the reconstructed core lines in all serial sections. While the method described above was the standard, we also made some modifications to it depending on the specimens and fixatives. For example, small and soft specimens, such as embryonic gonads, were fixed in modified Davidson's fluid or 10% formalin neutral buffer solution, a milder fixative than Bouin's solution, and the basement membranes were immunostained for marker proteins, such as laminin, to enable more accurate semiautomatic segmentation. On the other hand, large specimens, such as the adult hamster testis, were sometimes insufficiently fixed, even with Bouin's solution, making PAS-H staining of the basement membrane not sufficiently clear to allow for semiautomatic segmentation. In such cases, we adopted manual segmentation by marking the outlines of seminiferous tubules from section to section. Using these methods, we reconstructed the testis cords/ seminiferous tubules of mice, rats, and hamsters of various ages using serial paraffin sections. 30 Figure 1 shows the 3D structure of the testis cords soon after the onset of coiling in representative testes of mice, rats, and hamsters. 30,49 In E15.5 mice, the average numbers of testis cords and branching points per testis were 12.3 and 21.0, respectively (n = 3).

| 3DIMAG INGOFTE S TISCORDS/ S EMINIFEROUSTUBULE SUS INGS ERIAL S EC TIONS
In contrast, in E19.5 rats, the average numbers of testis cords and branching points per testis were 29.7 and 3.3 (n = 3), which were significantly larger and smaller, respectively, than those in mice.
In P0 hamsters, the average numbers of seminiferous tubules and branching points per testis were 9.0 and 93.0 (n = 3), which were significantly smaller and larger, respectively, than those in mice.
Therefore, the numbers of testis cords and branching points per testis were inversely related among the 3 species. This is presumably due to differences among the 3 species in the frequencies of the fusion of neighboring testis cords during the earlier process of testis formation, which increases the number of branching points and decreases the number of testis cords/seminiferous tubules. In relation to this, Clermont & Huckins 29 classified the types of testis cords/ seminiferous tubules in rats into "inner" and "outer," with the former having no contact with the tunica albuginea. In contrast, we found that most of the testis cords/seminiferous tubules in mice and hamsters were in contact with the tunica albuginea, similar to the outer cords in rats. Therefore, we considered it inappropriate to separate the outer and inner types in mice and hamsters. This phenomenon is presumably interpreted by the lower frequency of the fusion of testis cords in rats. On the other hand, we found that seminiferous tubules in hamsters may be classified into "shorter" and "dominant" types. Dominant-type cords, 2-4 per testis and accounting for more than 80% of the total cord length, have numerous branching points, suggesting that they are formed by the extensive fusion of preexisting testis cords during the earlier process of testis formation in hamsters. This pattern in the paths of seminiferous tubules was already established by P21, but became clearer in adults. Although testis cords/seminiferous tubules showed marked variations between individual mice, their basic structures were similar and retained from E15.5 to adults.

| PER S PEC TIVE S
Accurate 3D information at the cellular and tissue levels is becoming increasingly important in research and clinical practice. Despite the development of advanced imaging technologies, such as confocal microscopy and micro-CT, an abundant amount of information is still obtained from classical histological sections. In this review, we demonstrated the 3D reconstruction of testis cords/seminiferous tubules using serial histological sections and 3D visualization software. This method is advantageous because conventional paraffin-embedded normal and pathological specimens, the accumulation of which is extensive in ordinary histology and pathology laboratories, may be utilized for 3D analyses without the need for complicated and expensive equipment.
The limitations of this method are that it requires time and labor, preventing its application to the analysis of larger specimens, such as the human testis. The segmentation and tracing of seminiferous tubules in serial sections, even if performed semiautomatically using PAS-or immunolabeled basement membranes, remain the most time and labor-consuming steps. However, the use of emerging technology on artificial intelligence (AI) based on the deep learning of pattern recognition will overcome this issue by automating both the segmentation and tracing steps. We have started analyzing the 3D structure of the human testis using this AI technology.
Furthermore, as already discussed, region-specific morphological and functional differences within the testis may represent region-specific gene expression that is regulated by particular molecules produced inside or outside the testis. The histochemical localization of various bioactive molecules and their receptors may be superimposed on the 3D-reconstructed testis structure, similar to the lectin that recognized acrosomal molecules in our previous study. 37 With the addition of this molecular information, 3D reconstruction will contribute to the elucidation of not only the structures, but also the functions of various organs, including the testis.

ACK N OWLED G M ENTS
This work was supported by JSPS KAKENHI Grant Numbers JP19K16473 and JP21K06730.

CO N FLI C T SO FI NTE R E S T
The authors declare no conflicts of interest. The present animal