Lineage mapping the pre-implantation mouse embryo by two-photon microscopy, new insights into the segregation of cell fates.

Generation of Histone-2B GFP transgenic mice Transgenic mice expressing Histone-2B

GFP were created in house as described previously ( Vong et al., 2010 ). Embryo harvesting, culture, imaging, and development Histone-2B GFP males were crossed with natural, hormone-primed or super-ovulated CD1 wild-type females approximately 6–8 weeks of age. The presence of a vaginal plug the next day indicating copulation was noted as 0.5 d.p.c. Late in the afternoon of 1.5 d.p.c. 2–4 cell embryos were harvested by oviduct flushing with M2 media (MR-015-D, Millipore) and cultured in a droplet of KSOM media (MR-107-D, Millipore) covered with mineral oil (M8410, Sigma) on a World Precision Instruments 35 mm cover-glass Fluorodish (FD35-100, WPI). Embryos were allowed to develop over-night in a 37°C, 5% CO 2 incubator before being placed on the microscope stage the next morning at an 8-cell stage. Embryos were imaged on a Zeiss LSM 510 Confocor3 inverted microscope with a Chameleon Ti:Sapphire laser (Coherent Inc.). The microscope stage and objectives were enclosed by a cage incubator from In Vivo Scientific, and maintained at 37°C and 5% CO 2 . A 25x LD LCI Plan Apochromat 0.8 W Corr DIC objective used with Immersol (000000-1252-136, Carl Zeiss Microimaging, Inc.) was used at a zoom of x0.7 for all time-lapses. In some experiments multiple-stage positions were utilized to capture a large cluster of embryos. Time-lapses were acquired at 820 nm wavelength at an average laser power of 4–5%, equivalent to 12 mW. Z-stacks were taken at 2 μm intervals, with 51–53 sections for each stack with a scan time of ~981 ms for each section with a line averaging of 2. To determine the effect of imaging on the potential of embryo development, embryos that were imaged for 48 hours as described above were then transferred into a 2.5 d.p.c. pseudo-pregnant CD1 WT female. Pseudo-pregnancy was induced by mating to vasectomy-treated CD1 male mice. Pups were born naturally allowed to develop to age P12 to ensure developmental competency.

Immunohistochemistry and Reconstruction

CD1 wild-type females were mated with CD1 wild-type males and embryos harvested at 2.5 d.p.c. corresponding to the 16-cell stage. Embryos were fixed in 4% paraformaldehyde in PBS for 10–15 min, permeabilized in a solution of 0.5% Triton X-100 in PBS for 30 min and then blocked over-night in 10% FBS, 0.1% Triton X-100 in PBS. Embryos were then placed in primary antibody diluted in blocking solution for 12 hours or over-night. Primary antibodies used were anti-mouse Cdx2 (1:100, BioGenex, Cdx2-88), and anti-rabbit Oct-4 (1:100, Cell Signaling C30A3). After incubation with the primary antibody embryos were washed three times in PBS and placed in fluorescent secondary and DAPI (1:1000) diluted in blocking solution for 2–4 hours and imaged un-mounted in droplets of PBS on an inverted Leica SP5 confocal microscope. For actin phalloidin staining, in place of the primary antibody step embryos were placed in a 1:500 dilution of Alexa Fluor 568 Phalloidin (Invitrogen, A12380) and 1:1000 of DAPI in blocking solution for 30 min to 1 hour and imaged as before. Images were acquired in z-stacks every 1–2 μm for each wavelength, and then exported to IMARIS for reconstruction. Nuclei positions were identified using the Spot Identification Function as described previously, and coordinates of each cell nuclei were then exported to MatLab to determine the RDI. For relative Cdx2 levels, the mean intensity was determined by IMARIS for each nuclei in the embryo and corrected for relative background levels. Relative Cdx2 expression was then determined by comparing a cell’s mean intensity with the mean intensity of the closest, outer-most neighbor in the z-plane. This was to control for any loss of intensity from sectioning through the full depth of the embryo. For phalloidin staining, nuclear positions were obtained as described above, and cells were visualized in 3D in IMARIS to determine cell membrane surface exposure.

Cell-tracking algorithm

To track individual cells accurately, it is necessary to find the correspondence between the cells detected at different time points, according to the positions of their nuclei. More specifically, as the basis of cell lineage analysis, it is necessary to know the history of the trajectory of each cell, which cell was its ancestor, and which cells were its daughters. To achieve this, the positions of all nuclei throughout a movie, each represented by a 3-dimentional vector ( x , y , z ), were first acquired from IMARIS. These vectors were then sorted according to the time of occurrence, i.e. frame number, and automatically checked to make sure that no nuclei were lost over time. For effective tracking, the first frame of the process was selected such that there were exactly 2N nuclei present, where N is an integer. By doing this, all cells were at the same stage in the first frame, and it is convenient to assign the exact cell-stage numbers and initiate the tracking process. The algorithm developed is based on the idea that if no division occurs between two frames (the numbers of nuclei are the same), then cells in frame n +1 are assigned to cells in frame n in a one-to-one fashion so as to minimize the total distance summed over all pairs. When division does occur (noted by an increase in the number of nuclei), the cells of frame n +1 are also designated to those in frame n ; but in this case, in addition to the one-to-one correspondence, it also happens that two daughter cells in frame n +1 are assigned to the same mother cell in frame n . Given all the m nuclei positions in the n th frame p n ,1 = ( x n ,1 , y n ,1 , z n ,1 ), ···, p n , m = ( x n , m , y n , m , z n , m ), and all the k nuclei positions in the ( n +1)th frame p n +1,1 = ( x n +1,1 , y n +1,1 , z n +1,1 ), ···, p n +1, k = ( x n +1, k , y n +1, k , z n +1, k ), with m ≤ k , the optimal correspondence between nuclei of the two frames was found by minimizing the Total Distance Function (TDF): α ∑ i ∈ F n , 1 i 0 ∈ B n + 1 , 1 ∣ p n , i − p n + 1 , i 0 ∣ + β ∑ j ∈ F n , 2 j 1 , j 2 ∈ B n + 1 , 2 ( ∣ p n , j − p n + 1 , j 1 ∣ + ∣ p n , j − p n + 1 , j 2 ∣ ) . In the TDF, F n,1 and F n,2 is a 2-partition of the set of all indices for the m nuclei in the n th frame, i.e., F n ,1 ∪ F n ,2 = {1,2, ···, m }, F n ,1 ∩ F n ,2 = Ø. The partition was determined by the forward relationship from the n th frame to the ( n +1)th frame: for each i ∈ F n ,1 , the cell with nucleus at p n , i does not divide, so that it corresponds to only one nucleus in the ( n +1)th frame; on the other hand, for each j ∈ F n ,2 , the cell with nucleus at p n , j divides into two daughters in the (n+1)th frame. Similarly, B n +1,1 and B n +1,2 is a 2-partition of the set of all indices for the k nuclei in the ( n +1)th frame {1,2, ···, k }, but determined by the backward relationship to the n th frame: for each i 0 ∈ B n +1,1 , the cell with nucleus at p n +1, i 0 corresponds to a cell without division in the previous frame; however, for each pair of j 1 , j 2 , the two cells with nuclei at p n +1, j 1 and p n +1, j 2 are the daughters of the same cell which has divided from the n th to the ( n +1)th frame. A combination of these two 2-partitions then uniquely determines a correspondence between nuclei of the two frames. The constants α and β satisfy α +β = 1, weighing the two types of distances. Since the non-dividing cells usually move much less than the displacements from parents to daughters, in our implementations, α was set to be 0.75 and β to be 0.25 so that more tolerance was allowed in the case of division. The optimization process to find the minimal TDF between each pair of consecutive frames consists of two steps. In the first one, a heuristic search strategy was used to find the initial correspondence between the two groups of points. More specifically, numbers of elements in each subset F n,1 , F n,2 as well as B n,1 , B n,2 were first decided by m and k . The single-cell connections were first decided by finding the nearest one available, and the parent-daughter connections were then decided by finding the nearest two available. Clearly, the results from this strategy will be influence by the order of searching. Thus, in the second step of the algorithm, simulated annealing (SA), a strategy developed in genetic algorithms (GA), was used to improve the results closer to the global optimum. More specifically, two pairs of connections were randomly selected from the population and swapped with each other. If the swap decreased the TDF, the results were updated accordingly; otherwise, the swap was discarded and the original results kept unchanged. The total number of random swaps was decided by the size of populations. In our implementation, it was set to be three times of the product of the numbers of points in the two consecutive frames.

Determination of radial distance of cells and division angle

To determine the change of position of cells within each embryo at different time frames, it is necessary to reconstruct the embryos in three dimensions. First, we modeled all cells as 3-dimensional balls, with center positions given by the nuclei, and radii acquired as follows. The radius of a cell in the first frame was set empirically according to the spatial resolution of the imaging process. Assuming that all cells at the same stage have the same volume, and that all division events preserve the total volume of cells, the radius of each cell throughout the movie was thus uniquely determined. It is important to note that the ball-model is not intended as a representation of cell shape, it instead computes relative positions between cells and cells and the embryo. Next, we reconstructed the surface of the embryo at each time frame as the 3-dimensional convex hull that encompasses all points on the surface of any one of the cells in the same frame. At the same time, the volume of the reconstructed embryo was acquired, and the positions of all points that reside on the outer surface of the convex hull were also recorded ( Supplementary Movie 4 ). The above 3-dimensional reconstruction of embryos allowed us to determine the radial distance index (RDI) defined as a ratio, for each cell in each frame, between the radial distance from the nucleus to the center of the embryo and the radius of a cell in the initial stage. Thus, it is a normalized distance: the denominator serves as a scaling factor to eliminate the differences in spatial resolution during imaging among different embryos. Since the surface of an embryo is reconstructed as a convex hull, the center of the embryo is approximated accordingly as the center of the 3-dimensional space enclosed by the convex hull. The latter is computed as a weighted average of the centers of all the triangles that contribute to the surface, and the weights being their associated areas.

Calculation of Cell Division Angle

Using the 3-dimentional reconstructed embryos, we also determined the cell division angle (CDA), which measures the difference between the two daughter cells in deviation from their parent cell in each division process. To compute CDA, two angles, θ 1 and θ 2 are first decided, each representing the deviation of a daughter cell from the parent. For each daughter, this is calculated as the angle between two rays, both emitting from the position of the parent cell, but one points to the center of the embryo (the control direction), the other to the position of the position of the daughter cell. This angle is always between 0° and 180°. CDA is then computed as the absolute value of θ 1 -θ 2 . If these two deviations are the same, the CDA for this division process is 0º, representing the division angle that is tangential to the embryo surface. On the other hand, if one daughter cell appears in the control direction, while the other is opposite to the control direction, the CDA is 180º, which represents the division that is perpendicular to the embryo surface ( Fig. 4A ).

Supplementary Material 01 Supplementary Movie 1. 3D reconstruction in IMARIS of a mouse embryo expressing Histone-2B GFP starting from an 8-cell stage and ending at a 32-cell stage blastocyst. Each 8-cell stage nuclei (green) is marked with a differently colored spot to allow for easy tracing of each 8-cell lineage by eye. The time interval is every 7 min for 53 2 μm z-sections at 820 nm wavelength for the duration of the movie. 02 Supplementary Movie 2. Demonstration of transient-outer cell movement during blastocyst formation. Starting from a 16-cell stage, the outer-cell marked by a red spot (arrow) resides on the outer edge of the embryo. After the 5 th cleavage division the red cell moves inward at the start of the formation of the blastoceol cavity, and occupies a position in the ICM generally occupied by PE cells. 03 Supplementary Movie 3. Example of asymmetric cell division in an outer-cell lineage during the 16–32 cell division. The nucleus marked with a red spot undergoes asymmetric division during the 5 th cleavage (first arrow), however the two daughter nuclei (also marked by red spots) move outward and occupy a position in the TE after cavitation (final arrows). 04 Supplementary Movie 4. Reconstruction of a mouse embryo using convex hull projections as described in the Materials and Methods. The embryo shown in this movie is the same as that in Supplementary Movie 1 , starting from an 8-cell stage and ending at a 32-cell blastocyst, with the cell surface of each cell modeled as previously described.