Animal handling and experimental procedures were carried out following the Guidelines for Proper Conduct of Animal Experiments by the Science Council of Japan (http://www.scj.go.jp/ja/info/kohyo/pdf/kohyo-20-k16-2e.pdf).

As described previously (39), cumulus-oocyte complexes (COCs) were collected from Japanese Black cows (n = 12; 110.4 ± 34.3-month-old) by OPU using an ultrasound scanner (HS-2100; Honda Electronics, Toyohashi, Japan) and a 7.5-MHz convex array transducer (HCV-4710MV; Honda Electronics) with a 17-gauge stainless steel needle guide (242 COCs, in total). Follicles >2 mm in diameter were aspirated with a vacuum through a disposable aspiration needle (COVA Needle; Misawa Medical, Tokyo, Japan). The aspiration rate was 14 mL/min and the vacuum pressure was 100 mmHg. The IVM medium was 25 mM HEPES-buffered TCM199 (M199; Gibco, Paisley, Scotland, UK), supplemented with 10% newborn calf serum (NCS; 16010159, Gibco) and 0.01 AU/mL of follicle-stimulating hormone from porcine pituitary (Antorin-R10; Kyoritsu Seiyaku, Tokyo, Japan). COCs with two or more granulosa layers were washed three times with IVM medium. Recovered COCs were cultured in 4-well dishes (Non-Treated Multidishes; Nalge Nunc International, Roskilde, Denmark) in 600 μL of IVM medium, covered with mineral oil (M8414; Sigma-Aldrich, St. Louis, MO, USA), and incubated for 22 h at 38.5°C in 5% CO₂, 5% O₂, and 90% N₂ in humidified air. All cultures were maintained under these conditions.

Frozen semen of Japanese Black bulls stored in straws was thawed in water (37°C, 40 sec), and then centrifuged twice in IVF100 (Research Institute for the Functional Peptides, Yamagata, Japan; 600 × g, 5 min). After centrifugation, spermatozoa were removed from the pellet, and added to IVF100 to obtain a suspension with a final sperm concentration of 1.0 × 10⁷/mL. This suspension served as the IVF medium. After 22 h of IVM, the COCs were washed twice with IVF100. Up to 20 COCs were incubated in 100 μL droplets of IVF medium in 35 mm dishes (Falcon 351008; Corning, NY, USA) for 6 h.

After insemination, oocytes were completely denuded from the cumulus cells and spermatozoa by pipetting with a glass pipette in IVC medium: potassium simplex optimized medium with amino acid (KSOMaa Evolve Bovine; Zenith Biotech, Bangkok, Thailand) supplemented with 5% NCS and 0.6 mg/mL of L-carnitine (C0158, Sigma-Aldrich). Subsequently, presumptive zygotes were washed three times with IVC medium and cultured in 100 μL droplets of IVC medium for 48 h. Each droplet contained approximately 20 presumptive zygotes. Average value of cleavage rate was 74.0% (179/242). At 48 h post-insemination (hpi), embryos with more than four cells were transferred from the 35 mm dishes to well-of-the-well (WOW) dishes (LinKID micro25, Dai Nippon Printing Co., Ltd., Tokyo, Japan) as described (12). WOW dishes, which have 25 microwells (5 columns × 5 rows) in a circular wall in the center of a 35 mm dish, can culture up to 25 embryos each with a single drop of medium and track individual embryos throughout the culture. IVC medium and mineral oil were pre-cultured for at least 12 h in glass bottles separately at 38.5°C in 5% CO₂, 5% O₂, and 90% N₂ in humidified air, and the pre-cultured IVC medium (100 μL) was placed within the circular wall and covered with the pre-cultured mineral oil. At 168 to 180 hpi, embryos that had developed to or beyond the blastocyst stage were observed under an inverted microscope. Finally, 123 embryos had developed to the blastocyst stage (50.8%), and 80 blastocysts had been cryopreserved.

IVF embryos were cultured for seven days (by this time, they reached the expanded blastocyst stage) and examined under an inverted microscope at 27–31 hpi (at the 2-cell stage; n = 15) and at 168 to 180 hpi (at the blastocyst stage; n = 30). In blastocysts, only the embryos that were independently classified as Code1 according to the IETS standards by three skilled observers were used. OCT imaging was done as described previously (31, 36). Unstained live embryos were imaged by OCT using the Cell3iMager Estier (SCREEN Holdings Co., Ltd, Kyoto, Japan). The imaging system of Cell3iMager Estier is outlined in Figure 1. The system is equipped with a super luminescent diode (SLD; center wavelength: 890 nm, N.A. = 0.3). The SLD output is coupled to a single-mode optical fiber and split at an optical fiber coupler into the sample and reference arms. The reflections from the two arms are combined at the coupler and detected by the spectrometer. The 3D image data of the blastocysts were constructed from individual 2D ×-z cross-sectional images, which were obtained by a series of longitudinal scans obtained by laterally translating the optical beam position. The data acquisition window was 200–300 × 200–300 × 200–300 μm, and the voxel size was 1 × 1 × 1 μm. The OCT system scans light source positions in the x-axis direction while the shifting scanning line positions on the y-axis to obtain a signal on the x–y plane at a focus position on the z-axis. By repeating this scanning while shifting the z-axis focus positions, 3D images of the embryos were acquired. The lengths of the imaging range on the x-, y-, and z-axis were 300, 300, and 200 μm, respectively. The exposure time was 150 μs, and scanning an entire embryo was completed in a few minutes. OCT provided cross-sectional images with a slice thickness of 1 μm.

As described previously (40), blastocysts imaged by OCT were transferred to a cryoprotective solution [1.8 M of ethylene glycol and 0.1 M of sucrose in Dulbecco's PBS (D-PBS)], which was then placed in a 0.25-mL straw (IMV Technologies, L'Aigle, France) at room temperature. After the blastocysts were equilibrated at room temperature for 15 min, the straws were directly set in a programmable freezer (ET-1N; FUJIHIRA INDUSTRY CO., LTD, Tokyo, Japan) at −7°C where seeding was manually performed. Subsequently, the straws were cooled at a rate of 0.3°C/min to −30°C and then directly transferred to liquid nitrogen for storage until use. The straws were thawed in air for 10 s and then immersed in water (30°C, 20 s) for the ET. Because of the status of our farms, we have chosen a cryopreservation protocol rather than a verification protocol.

We recently reported using OCT to image bovine embryos (36). After the 3D images were captured by OCT (Figure 1), the 3D images were analyzed in an automized way (Figures 2, 3). Process (i) The 3D images of bovine embryos were binarized (Figures 2, 3). For each image, vectors were drawn at equal angles in the elevation direction (−90–90°) and the azimuth direction (−180–180°) from the center of the embryo to the outer surface and the inner surface (the outermost of blastocoel). The thicknesses of the embryo along each vector (T_All) were then measured (Figure 3). Process (ii) To distinguish the ICM from T_All, an appropriate threshold (TH_icm; the cut-off value for distinguishing ICM from T_All) was set by Otsu's method (41, 42) (Figure 3). The parts where the thickness from the outer edge of the embryo region was lower than TH_icm were excluded from the binarized images. The largest object among the remaining objects after the exclusion was defined as ICM. T_ALL was separated into the thickness information of the area corresponding to ICM (T_ICM) and the thickness information areas other than ICM (T_other). Process (iii) Based on the T_other value, the average thickness (TH_m) was evaluated, and the threshold (TH_t) was set by Otsu's method (41, 42) (Figure 3). The threshold (TH_zp) that separates TE and ZP was set by TH_m–(TH_t-TH_m) (in case of TH_t > TH_m) or by TH_m + (TH_m-TH_t) (in case of TH_m ≧ TH_t). The region where the thickness of T_other was lower than TH_zp was defined as ZP, and the remaining region after removing TH_zp from T_other was defined as TE. The unfulfilled region, surrounded by the embryo parts, such as ICM, ZP, and TE, in the binarized image was defined as the blastocoel. The volumes of the defined ICM, TE, ZP, and blastocoel were evaluated. The means, medians, standard deviations, minimum, maximum, and range of the thickness of ICM were evaluated from T_ICM. The thickness of TE was evaluated from the thickness information, which was obtained by subtracting TH_zp from T_other. Summary statistics related to the thickness of ZP was set by TH_zp.

The OCT-imaged embryos were transferred to 30 recipient cows [47.5 ± 25.5-month-old Holstein (18 lactating cows) and Japanese Black cows (10 cows; two cows received ET twice)] from March 2018 to February 2019 at the farms in Tottori prefecture, Japan. The cows were clinically normal with body condition scores (BCS) between 2.75 and 3.0 (BCS scale goes from 1 to 5 with 0.25 increments). Before ET, recipients were estrus-synchronized by the administration of a CIDR device (CIDR 1900; Zoetis Japan, Tokyo, Japan) for 9 days and a treatment with cloprostenol (Dalmazin 150 μg [i.m.]; Kyoritsu Seiyaku Corporation, Tokyo, Japan) 2 days before the CIDR removal. Estrus of recipient cows was monitored, and embryos were transferred seven days after estrus with the confirmation of the presence of corpus luteum (CL; diameter ≧20 mm). The recipients were examined for pregnancy 23 days after ET using ultrasonography (HS101V; Honda Electronics). Pregnancy was confirmed in nine Holstein cows and six Japanese Black cows by observation of a CL ≧ 20 mm in diameter and an embryo with a detectable heartbeat in the intraluminal uterine fluid and an embryonic membrane. Ages of pregnant and non-pregnant cows were 42.3 ± 19.9 and 52.7 ± 29.5-months-old, respectively. Cows with BCS of 2.75 and 3.0 were included equally in both pregnant and non-pregnant cows.

Hierarchical clustering analysis was performed using 13 blastocoel-related and ZP-related parameters of bovine blastocysts. Metrics and linkage criteria for hierarchical clustering were Pearson's correlation and unweighted average linkage. The data was normalized so that the average = 0 and SD = 1 for each of the parameter. Hierarchical clustering analysis was performed using the SciPy (ver.1.3.0) package in Python (ver.3.6.5). Statistical significance was analyzed using the Mann-Whitney U test. A value of p < 0.05 was considered statistically significant. Mann-Whitney U test was performed using the SciPy (ver.1.3.0) package in Python (ver.3.6.5). Principal component analysis (PCA) was performed using 22 parameters of bovine blastocysts. The data was normalized so that the average = 0 and SD = 1 for each of the parameter. PCA was performed using the scikit-learn (ver.0.23.2) package in Python (ver.3.6.5).

With OCT, it was possible to non-invasively obtain live 3D images of a 2-cell embryo. OCT images also showed the nuclei (Figures 4B,C), which made it easy to find binuclear blastomeres (blue and green in Figures 5B,C). Of the 15 embryos examined by OCT, two were binuclear and were not used for transfer.

Before transfer, we obtained images of an embryo with a microscope (Figures 6A, 7A) and with OCT. OCT images of the trophectoderm (TE) and inner cell mass (ICM) of representative embryos resulting in pregnancy (P embryos) and non-pregnancy (NP embryos) are shown in Figures 6B, 7B, respectively. In these figures, TE and ICM were artificially colored green and orange, respectively. The structure of a whole embryo, including ICM (orange), TE (green), and zona pellucida (ZP; gray) were also 3D-visualized (Figures 6C, 7C; Supplementary Movies 1, 5). Each part of the embryo, the ICM (Figures 6D, 7D; Supplementary Movies 2, 6), TE (Figures 6E, 7E; Supplementary Movies 3, 7), and blastocoel (Figures 6F, 7F; Supplementary Movies 4, 8), was also 3D-visualized individually.

Twenty-two parameters were measured for each of the 30 embryos (Table 1). For the embryos that were subjected to ET (n = 30), the average thicknesses of ICM, TE, ZP, and TE + ZP were 55.1 ± 14.1, 4.2 ± 1.2, 15.2 ± 3.4, and 19.5 ± 3.8 μm, respectively (mean ± SD). The average volumes of ICM, TE, ZP, TE + ZP, ICM + TE + ZP, blastocoel, and whole embryo were 3.7 ± 1.3, 2.0 ± 0.9, 16.1 ± 3.5, 18.7 ± 4.0, 22.4 ± 4.7, 13.4 ± 5.5, and 35.8 ± 8.1 × 10⁵ μm³, respectively (mean ± SD), and the blastocoel diameters were 52.3 ± 7.7 μm (mean ± SD). Fifteen of the 30 recipients became pregnant. None of the parameters were significantly different between the embryos that did (P) or did not (NP) lead to pregnancy (Table 1; Figure 8). Twelve of the 15 pregnant cows gave birth (six males and six females), and the remaining three cows experienced a late embryonic death. Newborn calves had typical weights (male: 40.2 kg; female: 35.5 kg, in average), and did not show any congenital defects, neonatal overgrowth, and death.

A PCA identified three principal components, PC1, PC2, and PC3. PC1 was related to the thickness of ZP and TE, which accounted for 41.00% of the variance; PC2 was related to the volumes of parts, which accounted for 29.81% of the variance; and PC3 was related to the thickness of ICM and TE, which accounted for 13.52% of the variance (Figure 9B). Figure 9A shows plots of PC1 vs. PC2 (a), PC1 vs. PC3 (b), and PC3 vs. PC2 (c). None of the plots clearly separated the P and NP embryos.

The hierarchical clustering analysis based on the blastocoel-related and ZP-related parameters (Figure 10) showed two clusters with a threshold of Dissimilarity = 2.0, and these clusters were also found separated into low (Cluster 1) and high (Cluster 2) blastocoel volume clusters on blastocoel volume-TE + ZP thickness plane (Figure 11). In Cluster 1, no significant difference was found in any of the 22 parameters between the P and NP embryos (Figure 12), while in Cluster 2, TE volume was significantly lower in the P embryos (p < 0.05) (Figure 12, 13). No difference between the P and NP embryos in any of the 22 parameters was common to both clusters (Figures 12, 13).