Embryonic stem (ES) cells should be able to generate any cell type, but directing their differentiation in vitro has proved to be far from easy. For example, ES cells that are grown in aggregates (embryoid bodies, or EBs) will generate large numbers of neurons, particularly if the EBs are exposed to retinoic acid. However, this process is relatively uncontrolled, and the EBs are highly disorganized, so the neuronal precursors might be exposed to signals that they would not normally encounter in the embryo. Perhaps because of this, neuronal precursors that are derived from EBs can only generate a limited range of neuronal cell types. Now, however, Rathjen et al. seem to have circumvented this problem by developing a new protocol that enables them to bypass the EB stage to generate a virtually pure neuroectoderm from mouse ES cells.

The authors previously generated a conditioned medium called MEDII, which can convert ES cells in adherent culture into a homogeneous population of primitive-ectoderm-like cells. In their new study, they cultured ES cells in suspension in the presence of MEDII. The cells formed aggregates that superficially resembled EBs, although their subsequent differentiation was much more ordered. Initially, the aggregates formed vesicles that were surrounded by a stratified epithelium, but after further culture in a different medium, over 95% of the cells took on a columnar appearance and became organized into a layer that closely resembled neuroectoderm. The cells also expressed neuroectodermal markers, including Sox1 , Sox2 and nestin, and on further differentiation, they gave rise to all the expected neuroectodermal derivatives, including neurons, glia and even neural crest.

The authors speculate that their technique will not only help them to generate pure neuronal populations for therapeutic purposes, but will also provide a valuable model to study neural induction and neuronal differentiation. Importantly, the aggregates generated by this technique lack visceral endoderm and other endodermal and mesodermal lineages, so the cells are not exposed to the signals that direct neuroectodermal patterning and differentiation in vivo. Consequently, the tissue is essentially naive, and it has no positional identity, so it could be used to test putative neural-tube patterning molecules. Also, by identifying the active components of MEDII, it should be possible to precisely define the conditions that are required to direct ES cells into the neural lineage in culture.