Building a pancreas from scratch.

Doug Melton is upfront about his seemingly futuristic goals. He ultimately wants to take laboratory-grown embryonic cells and induce them to form a pancreas, the organ that secretes crucial enzymes and hormones, including insulin. Melton, a biologist at Harvard University, is equally frank about his progress. "We're practically nowhere on the problem," he says.

The path leading from the earliest embryonic cells to a pancreas has many important milestones. In one of the earliest stage , embryos differentiate into three layers: the endoderm, ectoderm, and mesoderm. From the endodermis, the lungs, thymus, liver, stomach, and pancreas eventually arise. Although investigators have identified chemical signals that trigger formation of ectodermal and mesodermal cells, they've remained largely ignorant of the molecules prompting endodermal development.

Melton believes he has found one of those molecules. He and his colleagues have been looking for genes that prompt embryonic stem cells to turn on known endodermal-specific genes but not genes associated with the mesoderm or ectoderm. They found one gene, named mixer, that fits the bill. The gene encodes a protein that regulates the activity of other genes. This so-called transcription factor "can induce cells of the embryo to become endoderm," says Melton.

Once an embryonic cell becomes endodermal, it must still decide whether to be part of the pancreas or of some other organ. The molecules guiding that decision are also under investigation by Melton and his colleagues. They've found that the notochord, another area in a developing embryo, instructs a portion of the endoderm, to form a pancreas. The command may come in the form of secreted proteins such as activin or fibroblast growth factor. Each can induce endodermal cells to become pancreatic, says Melton.