The vertebrate facial skeleton is derived from the cranial neural crest, and patterning of this migratory cell population is a crucial early step in craniofacial development. Each hindbrain segment, or rhombomere (r), expresses a distinct combination of Hox genes, and this 'Hox code' seems to confer positional information on the neural crest. Drew Noden's classic grafting experiments from the early 1980s indicated that rhombomeres and their neural crest derivatives retain their identity if they are transplanted to a different axial level. For example, neural crest precursors that were destined for the first branchial arch still generated first-arch skeletal structures when they were grafted into more posterior regions of the chick hindbrain. However, more recent experiments have shown that r1–6 can be respecified if they are transplanted in a more posterior area of the hindbrain, and that Hox gene expression in the neural crest can be reprogrammed by signals from outside the neural tube. In a new paper, Trainor and colleagues try to reconcile these findings with Noden's data.

One of Noden's results that is often overlooked is that frontonasal neural crest cells also induced first-arch skeletal elements at ectopic sites, even though they do not normally contribute to these structures. Trainor et al. propose that both the anterior hindbrain and frontonasal crest grafts could have included tissue from the isthmus, which is a key signalling centre at the mid–hindbrain boundary. Could the isthmus have induced the skeletal duplications?

To find out whether signals from the isthmus could reprogramme Hox gene expression in the neural crest, Trainor et al. soaked a bead in the isthmic signalling molecule fibroblast growth factor 8 (FGF8), then implanted it adjacent to r4 in a normal embryo. They found that Hoxa2, which is normally expressed in r4 but not in r1, was initially downregulated in the r4 neural crest, but that it was re-activated in crest cells as they migrated away from the bead. So, although FGF8 can cause transient respecification of r4 neural crest cells, additional signals would presumably be required to suppress their identity permanently.

In a variation of Noden's experiment, the authors transplanted r1 tissue with or without the isthmus into r4, then examined the effects on skeletal development. In the absence of isthmic tissue, the second-arch skeleton developed normally. However, if isthmic tissue was included in the graft, it was replaced by first-arch structures, confirming that Noden's result could have been caused by the inclusion of isthmic tissue in his grafts.

In a previous study, Irving and Mason showed that rl tissue can switch to an r4 identity if transplanted into r4, but only if no isthmic tissue is present in the graft. Taken together with the new data, this indicates that r1 and its neural crest do not have an intrinsic ability to generate first-arch skeletal elements, but that additional isthmic signals are required to suppress the expression of more posterior patterning genes, such as Hoxa2. These findings support the increasingly popular idea that rhombomere and neural crest identity is more plastic than was originally thought, and future studies should help to identify the signals that influence this plasticity.