Using murine ESCs Correa Cerro et

Using murine ESCs, Correa-Cerro et al. (2011) screened the effect of single overexpressed TFs from a panel of 137 transcription factors and selected for those which had the ability to induce a transcriptome shift towards specific lineages at 48h post induction. They followed that report by testing some of their selected TF by directly differentiating four distinct cell types and verifying some successful differentiation (Yamamizu et al., 2013). The authors suggested that upon identification and expression of a downstream cell-specific gene combination of TF, rapid specific differentiation to mature cell subtypes can be achieved. As this manuscript was in preparation, two groups published important studies showing that a similar type of algorithm applied to identify transcription factors specific to particular cell types profiled and deposited into public databases. Both studies also showed that cell state expression patterns can be used to predict transcription factors able to reprogram cell fate from fibroblasts to retinal pigment epithelial T0901317 (RPE) or keratinocytes (D\'Alessio et al., 2015; Rackham et al., 2016). Both of these studies took advantage of large public databases containing data from at least 100 cell states to compile lists of transcription factors with cell-type specific expression patterns. Rackham et al. also made a tool to allow simple prediction of TFs that could effectively reprogram cell fate, and used it to demonstrate the accuracy of prediction (Mogrify) (Rackham et al., 2016). Together, these studies demonstrated that reprogramming factors can be identified solely on the basis of their gene expression pattern. Some examples for human ESC directed differentiation have been reported as well. Ectopic expression of a cocktail of neurogenin-2 (Ngn2) or NeuroD1 (Zhang et al., 2013) and ASCL1 (Chanda et al., 2014) result in efficient rapid induction of mature specific subtype of neuronal cells. Similarly, the 4 endothelial factors selected by CEMA (ERG–FLI1–HHEX–TAL1) were previously tested individually for their ability to induce hemato-endothelial programs in hPSC (Elcheva et al., 2014), however only induction of ERG was successful in promoting endothelial-like cells. The authors further reported that a single factor induction was not sufficient for a faithful induction of mature endothelial fate as witnessed by the failure of the differentiated cells to turn off their pluripotent gene expression program. In contrast, expression of combination of four endothelial CEMA-selected genes drove homogeneous differentiation toward the endothelial fate. The results presented here point to factors that not only define particular cell lineages such as neural progenitor cells and endothelial cells but also characterize both a temporal (i.e., developmental) and regional patterns of core genes. Lastly, we have provided an interactive web-based tool to allow users to query unique factors of expression, or, simply to obtain the pattern of expression of a particular gene of interest across many human cell types. Our expectation is that the results of the CEMA output will be refined as more cell types are added. We hope that the scientific community will benefit as a result from these analyses that provide information on the types of gene expression patterns that define individual cell states. Beyond reprogramming and direct differentiation, we expect that users will benefit from having the ability to phenotype cells derived from tissue or created from PSCs in their own lab using the new groups of genes defined here. For instance, if one simply uses SOX2 and NESTIN as neural progenitor markers, it is clear from the CEMA output, that a wide array of different types of NPC express these markers. Instead, one could look to CEMA to provide a more complete picture of a cell of interest, from the tissue from which it was derived to the stage of development it might represent.
Materials and methods
Acknowledgments