Three-dimensional understanding of the morphological complexity of the human uterine endometrium

All methods can be found in the accompanying transparent methods supplemental file .

To clear human uterine endometrial tissue, we applied CUBIC protocol IV, which was previously utilized for clearing human brain tissue ( Figure 1 A) ( Tainaka et al., 2018 ). We collected 20 uterine endometrial samples from 16 patients who underwent hysterectomy due to gynecological diseases with no lesions in the endometrium ( Figure 1 B and Table 1 ). With CUBIC protocol IV, we succeeded in substantially clearing all 20 human uterine endometrial tissues ( Figure 1 C). The autofluorescence signal derived from collagen and elastic fibers ( Hong et al., 2020 ; Zhao et al., 2020 ) was useful for observing the intact 3D structure of the uterine endometrial tissue by LSF microscopy ( Figure 1 D, left panel). We added immunostaining with a fluorescently labeled anti-cytokeratin (CK)7 antibody to highlight the endometrial gland structure. Immunohistochemical staining with the anti-CK7 antibody demonstrated selective labeling of the luminal and glandular epithelial cells running through the endometrial stroma with single-cell resolution ( Figures 1 D and S1 ). By 3D reconstitution of the LSF microscopy images of the CK7-stained human endometrium, we succeeded in visualizing the detailed 3D structure of the endometrial glands ( Figure 1 E). Our stereoscopic image made it possible to analyze free-angle images of tissue cross sections. As shown in Figure 1 F, the XY slice of uterine endometrial tissue subjected to the CUBIC protocol retained the characteristic 2D morphology of the endometrial glands for each phase, namely, curving glands in the proliferative phase and serrated glands in the secretory phase. The 3D image reconstituted by Imaris software (Bitplane) enabled us to observe continuous tomographic images of the human endometrium in all directions ( Video S1 ). Figure 1 Tissue clearing and 3D imaging of human uterine tissue using CUBIC (A) Schematic diagram of the clearing and immunostaining protocol for human uterine tissue. (B) Sampling site (yellow box) of human uterine tissue from subject E2. (C) Clearing performance of CUBIC protocol IV for human uterine tissue from subject E2. (D) 3D images of tissue from subject E2 stained with Alexa Fluor 555-conjugated anti-CK7 antibody with clearing by CUBIC. (E) Magnified 3D image of tissue from subject E2 demonstrating numerous glands as well as luminal epithelium and myometrium. (F) Comparison between a microscopic H&E-stained image and the reconstituted XY-plane image after clearing by CUBIC. Upper panels: images of endometrium in the proliferative phase (subject E1). Lower panels: images of endometrium in the secretory phase (subject E8). XY-plane optical slices (subject E1, z = 7.62 μm; subject E8, z = 6.61 μm). (D–F) Images obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. RT, room temperature; Autofluo, autofluorescence; CK7, cytokeratin 7; FFPE, formalin-fixed paraffin-embedded; H&E, hematoxylin and eosin. See also Figure S1 and Video S1 . Table 1 Clinical characteristics of the subjects Subject number Age Clinical diagnosis Menstrual cycle Gravidity Parity Body mass index E1 30 Cervical cancer (ⅠA1) Proliferative phase 1 1 34.1 E2 39 Cervical cancer (ⅠB1) Proliferative phase 3 2 17.9 E3 43 Pelvic organ prolapse Proliferative phase 1 1 23.8 E4 46 Myoma uteri Proliferative phase 0 0 25.0 E5-1, 2 42 Myoma uteri Secretory phase 4 2 20.7 E6 44 Myoma uteri, pelvic organ prolapse Secretory phase 1 1 22.8 E7 44 Myoma uteri Secretory phase 0 0 29.9 E8 46 Ovarian tumor Secretory phase 5 4 22.5 E9 46 Cervical cancer (ⅠB1) Secretory phase 1 1 17.8 E10 49 Myoma uteri, pelvic organ prolapse Secretory phase 3 3 20.9 E11-1, 2 45 Myoma uteri Menstrual phase 3 3 28.7 E12 48 Myoma uteri Menstrual phase 0 0 25.2 E13-1, 2 43 Myoma uteri Menstrual phase 3 3 29.2 E14 45 Cervical intraepithelial neoplasia 3 Menstrual phase 0 0 20.2 E15-1, 2 49 Myoma uteri Menstrual phase 2 2 38.9 E16 50 Myoma uteri Menstrual phase 2 2 23.5 A1 42 Adenomyosis Secretory phase 2 0 21.2 A2 45 Adenomyosis Secretory phase 0 0 23.8 A3 42 Adenomyosis Undergoing GnRH agonist treatment 2 1 26.4 A4 42 Adenomyosis Proliferative phase 1 0 20.0 GnRH, gonadotropin-releasing hormone. Tissue clearing and 3D imaging of human uterine tissue using CUBIC (A) Schematic diagram of the clearing and immunostaining protocol for human uterine tissue. (B) Sampling site (yellow box) of human uterine tissue from subject E2. (C) Clearing performance of CUBIC protocol IV for human uterine tissue from subject E2. (D) 3D images of tissue from subject E2 stained with Alexa Fluor 555-conjugated anti-CK7 antibody with clearing by CUBIC. (E) Magnified 3D image of tissue from subject E2 demonstrating numerous glands as well as luminal epithelium and myometrium. (F) Comparison between a microscopic H&E-stained image and the reconstituted XY-plane image after clearing by CUBIC. Upper panels: images of endometrium in the proliferative phase (subject E1). Lower panels: images of endometrium in the secretory phase (subject E8). XY-plane optical slices (subject E1, z = 7.62 μm; subject E8, z = 6.61 μm). (D–F) Images obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. RT, room temperature; Autofluo, autofluorescence; CK7, cytokeratin 7; FFPE, formalin-fixed paraffin-embedded; H&E, hematoxylin and eosin. See also Figure S1 and Video S1 . Clinical characteristics of the subjects GnRH, gonadotropin-releasing hormone. Video S1. Reconstructed 3D image of proliferative-phase, full-thickness human endometrial tissue (subject E2), related to Figure 1 The sample was stained with CK7 (yellow) and showed autofluorescence (blue). z stack (XY-plane view): 100 μm. The sample was stained with CK7 (yellow) and showed autofluorescence (blue). z stack (XY-plane view): 100 μm. Our 3D imaging of human uterine endometrial tissue revealed some unique 3D morphologies of the endometrial glands that had not been detected by 2D histological observation. First, we sometimes detected cystically dilated epithelial structures by observing continuous tomographic images ( Figure 2 A). The cystically dilated epithelial structure was stained with anti-forkhead box protein A2 (FOXA2), a marker of the glandular epithelium of the human uterus, antibody, as were the surrounding normal glands ( Figure S2 ). Intriguingly, the cystically dilated glands rose from the bottom of the endometrium together with other normal glands but occluded without reaching the luminal epithelium. To visualize and assess the 3D morphology of the “occluded glands,” we added pseudocolor to the glands in the 3D image independently ( Figure 2 B and Video S2 ). Occluded glands were observed in all samples of the proliferative phase and secretory phase ( Figure S3 A). Then, we calculated the number of occluded glands in a 1 mm 2 area on the XZ plane ( Table S1 ). The average number of occluded glands was 2.2 per mm 2 on the XZ plane (95% confidence interval: 1.3 to 3.1). There was a tendency for a positive correlation between age and the number of occluded glands ( Figure 2 C), and there were no significant differences in the number of occluded glands according to the menstrual cycle ( Figure S3 B). Additionally, we measured the height of the top of the head of occluded glands and the endometrial thickness. The top of the head of the occluded glands was widely distributed from near the base to near the surface, regardless of the menstrual cycle ( Figure 2 D). We also measured the volume of the occluded glands. The volume of occluded glands was significantly larger in the secretory phase than in the proliferative phase ( Figure 2 E). Figure 2 Morphology of occluded human endometrial glands (A and B) An occluded gland (subject E8). (A) Reconstructed XY-plane images (z = 99 μm). The red arrow indicates an occluded gland. (B) 3D distribution of an occluded gland that was pseudocolored and separated into a new channel by the Surface module in Imaris. (C) Spearman's correlation analysis was performed to evaluate the association between age and the number of occluded glands. (D) The average thickness of the endometrium and the height of the top of the occluded glands in each case. Error bars show ±SD. In the lower panel, the menstrual cycle is shown. (E) The volume of occluded glands in proliferative-phase samples and secretory-phase samples. Box plots show the median and interquartile range (IQR), with whiskers indicating the 1.5 IQR. Data were statistically compared by t test. Images were obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. A p value less than 0.05 was considered statistically significant. Autofluo, autofluorescence; CK7, cytokeratin 7. See also Figures S2 and S3 and Table S1 and Video S2 . Morphology of occluded human endometrial glands (A and B) An occluded gland (subject E8). (A) Reconstructed XY-plane images (z = 99 μm). The red arrow indicates an occluded gland. (B) 3D distribution of an occluded gland that was pseudocolored and separated into a new channel by the Surface module in Imaris. (C) Spearman's correlation analysis was performed to evaluate the association between age and the number of occluded glands. (D) The average thickness of the endometrium and the height of the top of the occluded glands in each case. Error bars show ±SD. In the lower panel, the menstrual cycle is shown. (E) The volume of occluded glands in proliferative-phase samples and secretory-phase samples. Box plots show the median and interquartile range (IQR), with whiskers indicating the 1.5 IQR. Data were statistically compared by t test. Images were obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. A p value less than 0.05 was considered statistically significant. Autofluo, autofluorescence; CK7, cytokeratin 7. See also Figures S2 and S3 and Table S1 and Video S2 . Video S2. Reconstructed 3D image of an occluded gland (subject E8), related to Figure 2 The sample was stained with CK7 (gray) and showed autofluorescence (blue). The occluded gland was pseudocolored (red). z stack (XY-plane view): 6.61 μm The sample was stained with CK7 (gray) and showed autofluorescence (blue). The occluded gland was pseudocolored (red). z stack (XY-plane view): 6.61 μm Second, many branches of the endometrial glands were detected at the bottom of the endometrium ( Figures 3 A and S4 A). We calculated the proportion of branched glands in each sample ( Table S2 ). The average proportion of branched glands was 68% (±20%). In particular, 90% of the branched glands had a branchpoint in the lower one-third of the endometrium ( Figure 3 B). Although it has been widely assumed that human endometrial glands branch from a single gland and radiate through the endometrial stroma toward the myometrium on the basis of previous 2D histological studies ( Cooke et al., 2013 ; Gray et al., 2001 ), our 3D imaging revealed that a more complex network of endometrial glands is generated in the human endometrium. When we assigned pseudocolors to two endometrial glands lying next to each other in Imaris software, we uncovered a horizontally expanding plexus morphology for the endometrial glands near the bottom of the endometrium. Furthermore, some endometrial glands shared a plexus and rose toward the luminal epithelium ( Figure 3 C and Video S3 ). Then, we calculated the proportion of glands sharing branches with other glands in each sample ( Table S3 ). The average proportion of glands sharing branches with other glands was 37% (±11%). In total, 86% of the glands sharing branches with other glands had a shared branching point in the lower one-third of the endometrium ( Figure 3 D). There was a tendency for a positive correlation between age and the proportion of branched glands per sample ( Figure 3 Ea). The proportion of branched glands per sample was significantly higher in the secretory phase than in the proliferative phase ( Figure S4 B). On the other hand, the proportion of glands sharing branches with other glands showed little correlation with age (( Figure 3 Eb), and there was no significant difference in the proportion of glands sharing branches with other glands according to the menstrual cycle ( Figure S4 C). Figure 3 Morphology of branched human endometrial glands (A) Branches of endometrial glands (subject E8) on reconstructed XY-plane images (z = 198 μm). Red arrows indicate the branches. (B) Classification of endometrial glands based on branch status. (a) Nonbranched gland. (b) Branched gland. The branched glands were subclassified into three categories based on the position of the branchpoint: “Upper 2/3 of the endometrium,” “Lower 1/3 of the endometrium,” or “Both,” (C) Glands sharing branches with the other glands (subject E8). 3D reconstruction of the distribution of branched glands that were pseudocolored and separated as new channels by the Surface module in Imaris. (D) Representative patterns of the glands sharing branches with other glands. These glands were classified into three categories based on the position of the shared branching point: “Upper 2/3 of the endometrium,” “Lower 1/3 of the endometrium,” or “Both.” (E) Spearman's correlation analysis was used to evaluate the association of age with (a) the proportion of branched glands and (b) the proportion of glands sharing branches with other glands. Images were obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. Autofluo, autofluorescence; CK7, cytokeratin 7. See also Figure S4 and Tables S2 and S3 and Video S3 . Morphology of branched human endometrial glands (A) Branches of endometrial glands (subject E8) on reconstructed XY-plane images (z = 198 μm). Red arrows indicate the branches. (B) Classification of endometrial glands based on branch status. (a) Nonbranched gland. (b) Branched gland. The branched glands were subclassified into three categories based on the position of the branchpoint: “Upper 2/3 of the endometrium,” “Lower 1/3 of the endometrium,” or “Both,” (C) Glands sharing branches with the other glands (subject E8). 3D reconstruction of the distribution of branched glands that were pseudocolored and separated as new channels by the Surface module in Imaris. (D) Representative patterns of the glands sharing branches with other glands. These glands were classified into three categories based on the position of the shared branching point: “Upper 2/3 of the endometrium,” “Lower 1/3 of the endometrium,” or “Both.” (E) Spearman's correlation analysis was used to evaluate the association of age with (a) the proportion of branched glands and (b) the proportion of glands sharing branches with other glands. Images were obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. Autofluo, autofluorescence; CK7, cytokeratin 7. See also Figure S4 and Tables S2 and S3 and Video S3 . Video S3. Reconstructed 3D image of the plexus of the endometrial glands (subject E8), related to Figure 3 The sample was stained with CK7 (gray) and showed autofluorescence (blue). The glands sharing the plexus were pseudocolored individually (green or light blue). z stack (XY-plane view): 6.61 μm. The sample was stained with CK7 (gray) and showed autofluorescence (blue). The glands sharing the plexus were pseudocolored individually (green or light blue). z stack (XY-plane view): 6.61 μm. According to the observation of the 3D morphology, a plexus of glands seemed to exist at the lower part of the endometrium and creep along the bottom. To confirm the location of the plexus in the human endometrium and shed light on the horizontal morphology of the plexus, it was necessary to divide an endometrial sample into layers and assess the structure of endometrial glands per layer. As the boundary between the endometrium and the myometrium was not flat, linear XZ sections were unsuitable to evaluate the 3D layer distribution of the endometrial glands. Therefore, we manually traced the border between the endometrium and myometrium of the XY section and reconstituted the bottom surface object of the endometrium in 3D ( Figure 4 A, left and middle panels, Figures S5 A and S5B). Then, new 3D layers of the endometrial glands were created to be the same distance from the bottom layer and have a thickness of 150 μm ( Figures 4 A, right panel, Figures S5 C–S5F). We extracted five 3D layers from the bottom to the lumen of the endometrium (layer 1: 1–150 μm, layer 2: 151–300 μm, layer 3: 501–650 μm, layer 4: 1,001–1,150 μm, layer 5: 1,501–1,650 μm). The plexus structure of the glands was mainly located in and crept along the lower layers (layers 1 and 2), and the morphology of the glands became columnar as the layers approached the lumen of the endometrium (layers 3–5) ( Figure 4 B). Figure 4 3D layer distribution of human endometrial glands (A) Left panel: the 3D tissue image was cropped on the XZ plane to 2.5 mm × 2.5 mm (subject E1). Middle panel: 3D reconstruction of the bottom layer of the endometrium. Right panel: 3D layers of endometrial glands were created at the same distance from the bottom layer and with a thickness of 150 μm by the Surface module in Imaris. Layer 1 (magenta): 1–150 μm; layer 2 (green): 151–300 μm; layer 3 (light blue): 501–650 μm; layer 4 (orange): 1,001–1,150 μm; and layer 5 (yellow): 1,501–1,650 μm. (B) XZ-plane view (y = 150 μm) of the five layers made by the Surface module in Imaris. After surface extraction, each structure was manually curated, and extra surface signals were eliminated. Images were obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. Autofluo, autofluorescence; CK7, cytokeratin 7. See also Figure S5 . 3D layer distribution of human endometrial glands (A) Left panel: the 3D tissue image was cropped on the XZ plane to 2.5 mm × 2.5 mm (subject E1). Middle panel: 3D reconstruction of the bottom layer of the endometrium. Right panel: 3D layers of endometrial glands were created at the same distance from the bottom layer and with a thickness of 150 μm by the Surface module in Imaris. Layer 1 (magenta): 1–150 μm; layer 2 (green): 151–300 μm; layer 3 (light blue): 501–650 μm; layer 4 (orange): 1,001–1,150 μm; and layer 5 (yellow): 1,501–1,650 μm. (B) XZ-plane view (y = 150 μm) of the five layers made by the Surface module in Imaris. After surface extraction, each structure was manually curated, and extra surface signals were eliminated. Images were obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. Autofluo, autofluorescence; CK7, cytokeratin 7. See also Figure S5 . In menstruation, the stratum functionalis exfoliates and the stratum basalis remains. To uncover whether the plexus of the endometrial glands was located in the stratum basalis, we determined the 3D layer distribution of the endometrial glands during menstruation (menstrual cycle day 2) ( Figures 5 A–5C). We reconstituted the bottom surface of the endometrium in a 3D image and divided the endometrium into two 3D layers (layer 1: 1–150 μm, layer 2: 151–300 μm). During menstruation, the plexus structure of the glands remained at the bottom of the endometrium and crept along the muscular layer ( Figure 5 D and Video S4 ). The other eight samples obtained during menstruation (menstrual cycle days 2–4) also had similar plexus structures in their lower layers ( Figure S6 ). These results revealed that the plexus structure of the endometrial glands was mainly located in the stratum basalis regardless of the menstrual cycle phase. Figure 5 3D layer distribution of endometrial glands in a case of menstruation (A) Microscopic H&E-stained image of endometrium during menstruation (subject E11-1). (B) Left panel: the 3D tissue image was cropped on the XZ plane to 2.5 mm × 2.5 mm (subject E11-1). Right panel: 3D reconstruction of the bottom layer of the endometrium and 3D layers of the endometrial glands were created at the same distance from the bottom layer and with a thickness of 150 μm by the Surface module in Imaris. Layer 1 (magenta): 1–150 μm; layer 2 (green): 151–300 μm. (C) XY-plane reconstructions (z = 47.6 μm). Each layer was pseudocolored and separated as new channels by the Surface module in Imaris. (D) XZ-plane view (y = 150 μm) of two layers made by the Surface module in Imaris. After surface extraction, each structure was manually curated, and extra surface signals were eliminated. Images were obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. FFPE, formalin-fixed, paraffin-embedded; H&E, hematoxylin and eosin; Autofluo, autofluorescence; CK7, cytokeratin 7. See also Figure S6 and Video S4 . 3D layer distribution of endometrial glands in a case of menstruation (A) Microscopic H&E-stained image of endometrium during menstruation (subject E11-1). (B) Left panel: the 3D tissue image was cropped on the XZ plane to 2.5 mm × 2.5 mm (subject E11-1). Right panel: 3D reconstruction of the bottom layer of the endometrium and 3D layers of the endometrial glands were created at the same distance from the bottom layer and with a thickness of 150 μm by the Surface module in Imaris. Layer 1 (magenta): 1–150 μm; layer 2 (green): 151–300 μm. (C) XY-plane reconstructions (z = 47.6 μm). Each layer was pseudocolored and separated as new channels by the Surface module in Imaris. (D) XZ-plane view (y = 150 μm) of two layers made by the Surface module in Imaris. After surface extraction, each structure was manually curated, and extra surface signals were eliminated. Images were obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. FFPE, formalin-fixed, paraffin-embedded; H&E, hematoxylin and eosin; Autofluo, autofluorescence; CK7, cytokeratin 7. See also Figure S6 and Video S4 . Video S4. 3D layer distribution of the endometrial glands in cases of menstruation (subject E11-1), related to Figure 5 Yellow: CK7, blue: autofluorescence. z stack (XY-plane view): 52.8 μm. Yellow: CK7, blue: autofluorescence. z stack (XY-plane view): 52.8 μm. Finally, we applied our method for 3D visualization of the endometrial glands to adenomyotic lesions; adenomyosis is a benign gynecological disease that is characterized by the presence of ectopic endometrial tissue within the myometrium. We collected adenomyotic tissue samples from four patients who underwent hysterectomy ( Table 1 ). Three subjects (A1, A2 and A4) did not receive any hormonal therapy within 1 year before the operation. Subject A3 was being treated with a gonadotropin-releasing hormone (GnRH) agonist because she had hypermenorrhea and severe anemia. With the application of CUBIC protocol IV ( Figure 1 A), the adenomyotic tissue samples were successfully cleared and stained with anti-CK7 antibody ( Figures 6 A and S7 ). The reconstituted 3D image of the adenomyotic tissue showed the detailed 3D structures of the ectopic endometrial tissue within the myometrium. Interestingly, the ectopic endometrial glands lengthened thin branches to expand the adenomyotic lesions. The 3D structure of the adenomyotic lesions within the myometrium was similar to that of an ant colony ( Figures 6 B and 6C and Video S5 ). To compare the travel of blood vessels and the ant colony-like network of ectopic endometrial glands within the myometrium, we performed multicolor fluorescence imaging with CK7 and cluster of differentiation (CD)31 in subject A4 as an additional sample ( Figure S8 ). As a result, part of the ant colony-like network of ectopic endometrial glands expanded along the route of large blood vessels ( Video S6 ). This ant colony-like structure was more remarkable in subjects A1, A2, and A4 than in subject A3, who had undergone GnRH agonist therapy ( Figure 6 D). The volume of ectopic endometrial tissue (adenomyotic lesion) per unit volume of myometrium was 1.9 × 10 7 μm 3 /mm 3 , 1.0 × 10 7 μm 3 /mm 3 , 5.0 × 10 6 μm 3 /mm 3 , and 1.5 × 10 6 μm 3 /mm 3 in subjects A1 (secretory phase), A2 (secretory phase), A4 (proliferative phase), and A3 (undergoing GnRH agonist treatment), respectively. On the other hand, it was found that the ectopic endometrium did not completely disappear with the use of GnRH, and residual lesions were scattered in the myometrium. Furthermore, we observed the direct invasion of eutopic endometrial glands into the myometrium, leading to the formation of adenomyotic lesions in subjects A1 and A3 and therefore regardless of GnRH agonist therapy ( Figures 6 B and 6D and Video S5 ). Figure 6 3D morphology of adenomyotic tissue (A) Left panel: microscopic H&E-stained image of adenomyotic tissue in the secretory phase (subject A1). Right panel: reconstructed XY section (z = 10 μm) of the adenomyotic sample after clearing by CUBIC. Black and red arrows indicate adenomyotic lesions. (B–D) 3D distribution of adenomyosis. (B) Subject A1. (C) Subject A2. (D) Subject A3. The sample from subject A2 did not include eutopic endometrium. Red object: 3D structures of the direct invasion of endometrial glands into the myometrium. Yellow object: ectopic endometrial glands in the myometrium excluding direct endometrial gland invasion. Yellow and red objects were made by the Surface module in Imaris. After surface extraction, each structure was manually curated, and extra surface signals were eliminated. Images were obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. FFPE, formalin-fixed paraffin-embedded; H&E, hematoxylin and eosin; Autofluo, autofluorescence; CK7, cytokeratin 7. See also Figures S7 and S8 and Video S5 . 3D morphology of adenomyotic tissue (A) Left panel: microscopic H&E-stained image of adenomyotic tissue in the secretory phase (subject A1). Right panel: reconstructed XY section (z = 10 μm) of the adenomyotic sample after clearing by CUBIC. Black and red arrows indicate adenomyotic lesions. (B–D) 3D distribution of adenomyosis. (B) Subject A1. (C) Subject A2. (D) Subject A3. The sample from subject A2 did not include eutopic endometrium. Red object: 3D structures of the direct invasion of endometrial glands into the myometrium. Yellow object: ectopic endometrial glands in the myometrium excluding direct endometrial gland invasion. Yellow and red objects were made by the Surface module in Imaris. After surface extraction, each structure was manually curated, and extra surface signals were eliminated. Images were obtained by LSF microscopy. Autofluorescence was measured by excitation at 488 nm. CK7-expressing endometrial epithelial cells were measured by excitation at 532 nm. FFPE, formalin-fixed paraffin-embedded; H&E, hematoxylin and eosin; Autofluo, autofluorescence; CK7, cytokeratin 7. See also Figures S7 and S8 and Video S5 . Video S5. Reconstructed 3D image of adenomyotic lesion (subject A1), related to Figure 6 The sample was stained with CK7 (gray) and showed autofluorescence (blue). The adenomyotic lesions were pseudocolored individually (red: direct endometrial gland invasion into the myometrium, yellow: ectopic endometrial glands). z stack (XY-plane view): 10 μm. The sample was stained with CK7 (gray) and showed autofluorescence (blue). The adenomyotic lesions were pseudocolored individually (red: direct endometrial gland invasion into the myometrium, yellow: ectopic endometrial glands). z stack (XY-plane view): 10 μm. Video S6. Reconstructed 3D image of adenomyotic tissue stained with anti-CK7 and anti-CD31 antibodies (subject A4), related to Figure 6 Yellow object: ectopic endometrial glands in the myometrium; green: CD31; blue: autofluorescence. Yellow object: ectopic endometrial glands in the myometrium; green: CD31; blue: autofluorescence.

In this study, we succeeded in rendering full-thickness 3D images of the human endometrium by using an updated CUBIC protocol and LSF microscopy. Our 3D imaging revealed characteristic morphological features of the human endometrial glands, including occluded glands, the plexus of the basal glands, and the gland-sharing plexus with other glands, which were not sufficiently observed by 2D histology alone. The 3D layer analysis clarified that the plexus structure of the glands expanded horizontally along the muscular layer. Furthermore, these morphological features were detected regardless of age or menstrual cycle phase, suggesting that they are basic components of the normal human endometrium. The 2D shape of the endometrial gland, as shown in Figure 7 A, has been described since the early 1900s ( Gray, 1918 ; Lessey and Young, 2019 ; Manconi et al., 2003 ; Padykula et al., 1984 ; Ross and Reith, 1985 ). However, our 3D images of endometrial glands suggest that this conventional 2D shape does not reflect their true morphology. On the basis of our 3D observation, we can provide a new 2D shape of the human endometrial glands, which would be the first in nearly 100 years ( Figure 7 B). Specifically, we referred to the plexus of the stratum basalis as the “rhizome” because of the similarity between the plexus and the rhizome in terms of not only their morphologies but also their functional features; for example, rhizomatous plants, such as grasses, are able to regenerate from erosion ( Yu et al., 2008 ). Figure 7 2D images of the normal human endometrium and adenomyotic tissue (A) Conventional 2D image of the endometrium. (B) New 2D image of the endometrium. (a) Nonbranched gland. (b) Gland sharing the rhizome with other glands. (c) Occluded gland. (C) 2D image of adenomyotic tissue, including direct endometrial gland invasion into the myometrium and an ant colony-like network of ectopic endometrial glands within the myometrium. 2D images of the normal human endometrium and adenomyotic tissue (A) Conventional 2D image of the endometrium. (B) New 2D image of the endometrium. (a) Nonbranched gland. (b) Gland sharing the rhizome with other glands. (c) Occluded gland. (C) 2D image of adenomyotic tissue, including direct endometrial gland invasion into the myometrium and an ant colony-like network of ectopic endometrial glands within the myometrium. In some studies, reconstituted 3D visualization of partial endometrial structures was performed by using computerized two-dimensional binary images of serial sections ( Manconi et al., 2001 ; Simbar et al., 2004 ; Tempest et al., 2020 ) or multiphoton excitation microscopy ( Manconi et al., 2003 ), but these studies had serious limitations in the observable range. The former focused only on parts of the glands, and the latter observed a depth of less than 120 μm because of issues with tissue transparency. Therefore, the whole detailed 3D structure of the endometrial glands could not be detected. Recently, several groups have developed tissue-clearing techniques ( Chung et al., 2013 ; Erturk et al., 2012 ; Hama et al., 2011 ; Ke et al., 2013 ; Susaki et al., 2015 ; Tainaka et al., 2016 ). These techniques make a whole organ or sample transparent so that light can illuminate deep regions of the tissues. To date, only Arora et al. have applied tissue clearing and confocal imaging methods to only one human uterine tissue sample in addition to several mouse uterine tissue samples ( Arora et al., 2016 ). However, they did not clarify any characteristic 3D morphologies of the human endometrial glands. By using an updated CUBIC protocol ( Tainaka et al., 2018 ), we successfully cleared human uterine tissue to a depth of several centimeters. The slight autofluorescence signal that remained after the CUBIC protocol was useful for observing the details of the anatomical structures in the tissue. Furthermore, we performed immunohistochemistry with a fluorescently labeled anti-CK7 antibody to extract the clear tubular structures of the endometrial glands. As a result, we could generate 3D surface renderings of the endometrial glands with Imaris software. Histologically, isolated cystically dilated glands are commonly encountered in the normal endometrium. In this study, according to the 3D visualization results, we proved that the cystically dilated glands were occluded glands that were not continuous with the luminal epithelium. Our results show that the volume of the occluded glands was larger in the secretory phase than in the proliferative phase. Because the occluded glands cannot discharge secretions into the uterine cavity, they accumulate and inspissate within the gland, leading to cystic dilatation of the gland. Our study suggests that the number of occluded glands might be positively correlated with age. Indeed, cystically dilated glands are predominantly detected in the atrophic endometrium of postmenopausal women and in disordered proliferative endometrium, which is also common among perimenopausal women ( Al-Hussaini et al., 2020 ). Glands of irregular shapes and sizes, including glands with cystic dilatation without nuclear atypia, are also characteristic of simple endometrial hyperplasia ( Chandra et al., 2016 ). Endometrial hyperplasia frequently results from chronic estrogen stimulation unopposed by the counterbalancing effects of progesterone ( Sanderson et al., 2017 ). Therefore, gland occlusion might be related to hormone imbalance and/or hormone sensitivity of the gland. In the stratum basalis of the human endometrium, narrow and horizontally running glands are often detected by 2D histological observation ( Garry et al., 2010 ). However, the fact that the branches of glands form a complicated pattern in the stratum basalis has not been noted to date. In this study, we revealed a rhizome structure in the stratum basalis, that is, a horizontally expanding plexus morphology of the basal glands, by 3D observation. At the same time, as in our study, Tempest et al. noted a complex horizontal network of basalis glands by using computerized 2D binary images of 100 serial formalin-fixed paraffin-embedded sections (4 μm/section) ( Tempest et al., 2020 ). Although previous studies have shown the 3D structure of murine endometrial glands ( Arora et al., 2016 ; Vue et al., 2018 ), the bottom of these glands forms a crypt but not a rhizome. This can potentially be explained by the existence of menstruation, which is the crucial difference between the human and murine endometrium. The human endometrium is a dynamically remodeling tissue that undergoes more than 400 cycles of regeneration, differentiation, shedding, and rapid healing during a woman's reproductive years ( McLennan and Rydell, 1965 ). The rhizome structure of the stratum basalis may be advantageous for the reliable and rapid regeneration of the stratum functionalis by protecting the epithelial cells of the basalis in the human endometrium with frequent menstrual cycles. Because the stratum functionalis is shed during menses, the endometrium is believed to regrow and regenerate from endometrial progenitor/stem cells residing in the stratum basalis ( Kyo et al., 2011 ; Padykula, 1991 ; Padykula et al., 1984 ; Prianishnikov, 1978 ). In the field of botany, rhizomatous plants that have stems running underground horizontally are well known for their difficulty in terms of eradication because they can regenerate from a piece of rhizome left behind in the soil after natural or artificial erosion ( Sásik and Elias, 2006 ; Yu et al., 2008 ). The rhizome in the human endometrium will have a functional advantage over the crypt in terms of the conservation of progenitor/stem cells and regeneration. Furthermore, although it has been established that intestinal stem cells exist in the bottom of the crypt ( Sangiorgi and Capecchi, 2008 ), there are still uncertainties about the location of endometrial progenitor/stem cells ( Santamaria et al., 2018 ). Therefore, it may be necessary to take into consideration that the human endometrium has a rhizome in its stratum basalis in future endometrial stem/progenitor cell studies. In this study, 32% glands shared the rhizome with other glands. Some studies have indicated that human endometrial epithelial glands are monoclonal in origin ( Chan et al., 2004 ; Gargett et al., 2016 ; Kyo et al., 2011 ; Tanaka et al., 2003 ), implying that they arise from a single progenitor/stem cell. We recently reported a diversification of cancer-associated mutations in histologically normal endometrial glands ( Suda et al., 2018 ). In our previous study, we sequenced 109 single endometrial glands isolated from the stroma using collagenase and found that two of them shared the same PIK3CA mutation (p.K111N) and the same PPP2R1A mutation (p.S256Y) with high mutant allele frequencies, suggesting that both glands descended from a single common ancestral cell. If an endometrial gland has a monoclonal composition including the rhizome, it follows that some of the glands sharing the rhizome have a common origin and that the endometrium is an aggregate of small clonal segments. Genomic alterations to progenitor/stem cells may be transmitted to several glands through a rhizome. In the most recent study, whole-genome sequencing of the normal human endometrial epithelium showed that six microdissected glands isolated from one section shared over 100 variants; therefore, they were regarded as the same clade ( Moore et al., 2020 ). The authors argued that the clonal evolution of phylogenetically related glands entailed the capture and colonization of extensive zones of the endometrial lining ( Moore et al., 2020 ). Interestingly, the phylogenetically related glands were located at the bottom of the stratum basalis and were interspersed horizontally, suggesting that they formed the rhizomes discovered in our 3D analysis. It is possible that the rhizome of the endometrium is a crucial element for understanding the genetic features of the endometrium. Our 3D imaging method also clarified the 3D morphology of adenomyotic tissue. Adenomyosis is defined as the presence of ectopic endometrial glands and stroma surrounded by hyperplastic smooth muscle within the myometrium ( Vannuccini et al., 2017 ). There are several hypotheses regarding the etiology of adenomyosis, such as endometrial invasion, endometriotic invasion, and de novo metaplasia ( Kishi et al., 2012 ). Among them, it has been generally accepted that uterine adenomyosis results from direct invasion of the endometrium into the myometrium. Although this pathological condition was advocated based on an observation of 2D serial sections ( Benagiano and Brosens, 2006 ), we successfully proved the direct invasion of the endometrium into the myometrium in reconstituted 3D images of CK7-stained adenomyotic tissue. Our 3D image analysis also depicted the manner in which adenomyosis produces growing lesions within the myometrium. In other words, we showed that the ectopic endometrial glands had lengthened, thin branches and expanded lesions with an ant colony-like appearance in patients not undergoing hormone therapy. Leyendecker et al. reported that the ectopic endometrium of adenomyotic lesions mimicked the cyclical hormone receptor expression pattern of the stratum basalis ( Leyendecker et al., 2002 ). In our 3D images of adenomyotic tissue, regardless of the menstrual cycle, fine glands branched and extended intricately within the myometrium, and morphologically, they also seemed to be closer to the rhizome structure of the basalis than the glands of the stratum functionalis. Furthermore, the adenomyotic lesions extended along the large blood vessels in the myometrium, suggesting adenomyotic invasion of the vulnerable areas between the myometrium. Based on our 3D images, we were able to provide a new 2D shape of adenomyotic tissue, as shown in Figure 7 C. At present, no radical treatment for adenomyosis exists aside from hysterectomy. In this study, 3D imaging of samples from the patient undergoing GnRH agonist treatment showed an atrophied adenomyotic region as well as a eutopic endometrium. However, ectopic endometrial glands still survived in spots in the deep part of the myometrium. This suggests that hormone therapy is incapable of eradicating adenomyotic glands forming an ant colony-like structure. If the estrogen level increased again after hormone therapy, the remaining ectopic gland branches would lengthen and reconstitute an ant colony-like structure. A recent study involving genomic analysis of adenomyosis indicated high variant allele frequencies of KRAS hotspot mutations in the microdissected epithelial cells of adenomyotic tissue. This study suggested that adenomyosis may arise from the ectopic proliferation of mutated epithelial cell clones ( Inoue et al., 2019 ). Furthermore, the authors found identical KRAS mutations in adenomyotic and histologically normal endometrium adjacent to adenomyotic lesions and argued that “ KRAS -mutated adenomyotic clones originate from normal endometrium” ( Inoue et al., 2019 ). Our novel findings of the 3D morphology of adenomyotic tissue support this hypothesis from a histological perspective. A combination of genomic and 3D analyses is required to elucidate the etiology of adenomyosis. In conclusion, we successfully obtained 3D, full-thickness images of the human endometrium using an updated CUBIC protocol. This stereoscopic imaging made it possible to analyze free-angle images of cross-sections of tissue. With this procedure, we visualized the 3D morphology of the glands and created a new conceptualization of 2D images of the human endometrium. Furthermore, we revealed the 3D pathological morphology of adenomyotic lesions, which led to clarification of the etiology and pathogenesis of the disease. These findings indicate that this procedure can be a useful tool for analyzing the human endometrium and endometrium-related diseases from a new perspective. The 3D representation of the human endometrium will lead to a better understanding of the human endometrium in various fields, including histology, pathology, pathophysiology, and oncology. We could not perform whole-uterine clearing or whole-uterine sampling because our samples were obtained clinically from patients who underwent hysterectomy. Structural mapping of the whole human uterine endometrium would improve the understanding of the anatomical and histological features of the human uterus. Furthermore, none of our samples were from young women under 29 years old because individuals in that age group seldom undergo hysterectomy. Because we needed uterine tissue samples several centimeters in size containing both endometrium and myometrium, histology samples from younger patients not treated with hysterectomy were insufficient. In addition, it was not possible to use transplant organ donors or postmortem autopsies to obtain uterine tissues from young women. Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Kosuke Yoshihara ( [email protected] ). The data on the 3D histology of the human uterine endometrium and endometrium-related diseases are freely available at TRUE (Three-dimensional Representation of human Uterine Endometrium), the website of our database ( https://true.med.niigata-u.ac.jp/ ). This study did not generate or analyze datasets or code.

The human endometrium is a dynamic tissue that exhibits a high regenerative capability after cyclic shedding, namely, menstruation. Menstruation is a unique biological phenomenon that occurs in a limited number of mammals, such as humans and other higher primates ( Emera et al., 2012 ; Garry et al., 2009 ). Menstruation involves cyclic morphological and functional changes in the uterine endometrium that occur on a monthly basis in response to ovarian hormones ( Noyes et al., 1950 ). The uterine endometrium changes dramatically based on the phases of the menstrual cycle (i.e., the proliferative phase, the secretory phase, and menstruation) and plays a crucial role in the implantation of fertilized eggs. Additionally, “endometrium-related diseases,” such as adenomyosis, endometriosis, endometrial hyperplasia, and endometrial cancer, originate in the uterine endometrium due to its high intrinsic regenerative capacity and affect the lives of women from puberty until after menopause ( Garcia-Solares et al., 2018 ; Koninckx et al., 2019 ; Morice et al., 2016 ). However, the pathogenesis of these endometrium-related diseases remains unclear, and further investigations focusing on the endometrium from the standpoint of disease prevention are required. The conventional morphological theory of endometrial structure was established based on two-dimensional (2D) histological observation ( Johannisson et al., 1982 ; McLennan and Rydell, 1965 ; Noyes et al., 1950 ). Histologically, the endometrium is lined by a simple luminal epithelium and contains tubular glands that radiate through the endometrial stroma toward the myometrium by coiling and branching morphogenesis ( Cooke et al., 2013 ; Gray et al., 2001 ). The human endometrium is stratified into two zones: the stratum functionalis and the stratum basalis. The stratum functionalis is shed during menstruation and regenerates from the underlying stratum basalis during the proliferative period. Therefore, it is widely assumed that regeneration of the stratum functionalis depends on endometrial progenitor/stem cells residing in the stratum basalis ( Kyo et al., 2011 ; Maruyama and Yoshimura, 2012 ; Padykula, 1991 ; Prianishnikov, 1978 ). Despite this well-established understanding, neither the detailed mechanisms of endometrial regeneration during the menstrual cycle nor the location of endometrial progenitor/stem cells have been fully characterized ( Garry et al., 2010 ; Gellersen and Brosens, 2014 ; Santamaria et al., 2018 ). One of the reasons for this knowledge gap is that the fundamental structure of the human endometrium has not been sufficiently clarified. As the human endometrial glands have complicated winding and branching morphologies, it is extremely difficult to assess the whole shape of the glands by 2D histopathological imaging alone. In our previous genomic study, sequence analysis of 109 single endometrial glands revealed that each gland carried distinct somatic mutations in cancer-associated genes, such as PIK3CA , KRAS , and PTEN ( Suda et al., 2018 ). Remarkably, the high mutant allele frequencies of somatic mutations per endometrial gland indicate the monoclonality of each gland. The presence of cancer-associated gene mutations in histologically normal endometrial glands provides important clues regarding the pathogenesis of endometrium-related diseases. Hence, we hypothesized that clonal genomic alterations in histologically normal endometrial glands may change the stereoscopic structure of the endometrial glands, leading to susceptibility to endometrium-related diseases. To this end, we initially needed to evaluate and understand the three-dimensional (3D) morphology of the normal uterine endometrium. Recently, several tissue-clearing methods have been developed to enable 3D imaging of rodent and primate tissue samples ( Chung et al., 2013 ; Erturk et al., 2012 ; Hama et al., 2011 ; Ke et al., 2013 ; Susaki et al., 2015 ; Tainaka et al., 2016 ). The combination of these methods with the use of various types of optical microscopy, including confocal fluorescence microscopy and light-sheet fluorescence (LSF) microscopy, enables us to reconstitute 3D views of the tissues and provide 2D images of free-angle sections without the need for tissue slicing. In this study, we applied the updated clear, unobstructed brain/body imaging cocktails and computational analysis (CUBIC) protocol with fluorescent staining of clear, full-thickness, human uterine endometrial tissue because CUBIC has several advantages that make it suitable for our analysis ( Tainaka et al., 2018 ). First, CUBIC is excellent in terms of its ease of use and safety because it is a hydrophilic tissue-clearing method. Second, the updated CUBIC protocols offer fast and effective clearing of various human tissues ( Tainaka et al., 2018 ). Therefore, we determined that CUBIC had the proper capabilities for clearing human uterine tissue. Here, we aimed to define the “normal” 3D morphology of the endometrial glands. We successfully obtained 3D full-thickness images of the human uterine endometrium by combining the updated CUBIC protocol with LSF microscopy. We also succeeded in constructing the 3D morphology of the endometrial glands and established a baseline for the 3D structure of normal human endometrial glands. Finally, we used our 3D imaging method to reveal the 3D pathological morphology of tissue affected by adenomyosis, a representative “endometrium-related disease,” and compared it with the 3D morphology of the normal endometrial glands. Elucidation of the 3D structure of the human endometrial glands will provide further insights into various fields, including histology, pathology, pathophysiology, and oncology.