Single-cell quantification

We used ImageJ to manually track positions of the nuclei in single cells. Positions were saved and intensities for each fluorescence channel at each position were processed by custom-built Matlab scripts (available upon request). For each of the channels, we have used the mean intensity of the 5 x 5 x 3 voxel as a readout for single cell. To filter out the noise, cells with the Dapi intensity below 1 were removed. To differentiate between TE and ICM cells, for each of the cells, we ranked the intensities of Cdx2, Nanog, and Gata6. Cells where Cdx2 ranked first, were classified as TE cells. We validated that the identified TE cells localize to the periphery of the embryo. We classify the ICM cells as described in Saiz et al. [71]: First, we performed k-means clustering (by Squared Euclidean distance metric, Matlab built in function) on the log(Gata6) and log(Nanog) into 3 clusters with 10 repetitions on all data pooled together. Second, we classified high Gata6 and high Nanog cells as DP cells, high Gata6 and low Nanog as PrE cells, and low Gata6 and high Nanog as EPI cells. See S3 Fig for the results of the clustering.