Since the first cloned mammal, scientists have obsessed over techniques to reset specialized cells from the terminally differentiated state to a primal, early stage. Recent dogma-crumbling discoveries, published this month in Nature, call the need for such backtracking into question.

The birth of Dolly the sheep in 1997 showed that a mammalian egg could convert a specialized cell's nucleus into one that could become all cell types1. Three years of painstaking work later, the first embryonic stem (ES) cells were cloned, showing that oocytes could reprogram somatic cell nuclei to confer the properties of pluripotency and proliferation typical of ES cells2; not much later, ES cells were cloned from neurons and blood cells, leaving little room for doubt that eggs can reset the nuclei of terminally differentiated cells3,4. The idea that a similar feat could be achieved without the egg's reprogramming machinery was unthinkable back then, but it has since been accomplished in spades. With the introduction of genes encoding a handful of transcription factors, multiple mouse cell types as well as human fibroblasts have been converted to an embryonic-like state5,6,7,8.

Conrad Waddington likened cell fate to a marble rolling downhill into one of several troughs representing fully differentiated cell types. Nuclear transfer and reprogramming showed cells could be rolled back to the top of the hill. Now, tunnelling through the landscape seems possible too.

Inevitably, scientists in search of specific cell types for regenerative medicine began to wonder whether the pluripotent state is in fact desirable. The wait times are prohibitive: turning a pluripotent cell into a neuron can take months in the lab9, and some specialized human cell types can take much longer. Might it be possible to save time and reagents by transforming mature cells directly into other cell types capable of warding off diabetes, heart attacks or Parkinson's disease? Can a differentiated cell of one type be reprogrammed to become a differentiated cell of another type?

Shortcuts to beta cells

At a meeting of the International Society for Stem Cell Research (ISSCR) in early June, Harvard Stem Cell Institute director Doug Melton reported findings that strongly suggest this can be done. Using techniques akin to those that generated the first induced pluripotent stem (iPS) cells, a team spearheaded by Melton's postdoctoral associate Qiao Zhou turned digestive enzyme–producing pancreatic exocrine cells into insulin-secreting beta cells. The latter are fairly rare to begin with and are in especially short supply in patients with diabetes, whereas the former comprise 95% of the cells of the pancreas. So, converting one into the other is roughly equivalent to turning copper into gold.

The Harvard team scoured the literature and conducted studies on transcription factors documented in mice, eventually winnowing the 1,400 or so down to 9 that looked important for beta-cell status. After using retroviruses to insert genes encoding these transcription factors into the pancreatic exocrine cells of live mice, says Melton, “we found new, pancreatic insulin-producing cells outside of their normal position in the islets.”

“These cells didn't 'sort of' look like and 'sort of' smell like beta cells. They had been entirely converted to pancreatic beta cells,” says Melton. Ultimately, it proved possible to get this result by introducing just three transcription factor genes: Pdx1, Ngn3 and MafA. Clinical applications are still a long ways off, and not just because of the triple-gene therapy: these cells do not appear to respond to iglucose. Nonetheless, when mice rendered diabetic using a drug that kills beta cells were subsequently infected with the transcription factor–carrying viruses, their pancreases began to secrete insulin and levels of glucose in their blood dropped.

Switching cell fates

The Melton lab's experiments aren't the first to show that manipulating a differentiated cell can alter its state to that of another sort of differentiated cell; there were even some previous hints that exocrine cells could switch to endocrine cells. Perhaps the most famous example occurred in the 1980s, when Harold Weintraub at the Fred Hutchinson Cancer Center in Seattle, Washington, showed that transfecting fibroblast-like cells in tissue culture with complementary DNA encoding the muscle-cell transcription factor MyoD caused them to take on the characteristics of muscle cells10. More recently, other investigators have shown that by deleting or depleting a transcription factor, B lymphocytes can be converted into T lymphocytes and macrophages, respectively11,12. Nor are blood and muscle the only examples. Overexpression of a gene for yet another transcription factor can redirect neuronal progenitor cells toward an oligodendrocytic lineage13.

These cells didn't 'sort of' look like and 'sort of' smell like beta cells. They had been entirely converted to pancreatic beta cells. Doug Melton

But those conversions all involve single transcription factors whose powerful influence on cell state was already known from mutant examples; the Melton lab's pancreatic exocrine-to-beta-cell conversion was more of a long shot. It required a minimum of three payloads — not an outcome that could be easily foreseen from mutant studies. “You don't mutate three genes at a time,” Melton says. It took some guesswork and, he adds, a lot of luck.

Those familiar with Melton's findings (to be published in an upcoming issue of Nature)14 call them a significant step forward. “Being able to get the combination to drive differentiation very far down the beta-cell pathway and restore function the way they did — it was, I think, very impressive,” says Ken Zaret, who leads the epigenetics and progenitor cells programme at the Fox Chase Cancer Center in Philadelphia, Pennsylvania.

It is, however, in the words of Robert Tjian, director of the University of California, Berkeley's Stem Cell Center, “a black box experiment. You take this black box, you beat it over the head with a hammer on one side and something comes out on the other. But in between you have no idea what happened. How many steps did you miss in between?”

Cell-specific transcription

The low efficiency of converting fibroblasts to iPS cells — mere hundredths to thousandths of a percent — and the fact that it takes them close to a month to appear suggests other actors besides transcription factors in setting cell fate. Some of these probable regulators are barely recognized as yet. For instance, the vast majority of all RNA transcripts in mammals may be non-coding15, suggesting that there's a lot more to learn about regulatory RNAs.

Tjian has identified another poorly understood potential player in cell-fate determination, at least for the muscle lineage16. He has found that as cells specialize, there are concomitant switches in the composition of their core transcription machinery — the massive, multi-subunit, complex of RNA polymerase and other proteins that parts the double strands of DNA, reads one of them and makes an RNA copy.

This came as a complete surprise even to Tjian, as the core transcription machinery has long been thought to be invariant across species and cell types. At the annual Cold Spring Harbor Laboratory meeting in Cold Spring Harbor, New York earlier this year, Tjian reported his lab's observation that, as a developing skeletal-muscle cell matures from a rapidly dividing myoblast into a terminally differentiated myotube, there are changes in the subunit composition of the core transcription machinery. In other words, as a cell makes the transition into a new state in which new batteries of genes will need to become active and formerly active ones silenced, some subunits of the transcribing machinery are swapped out. This switch, Tjian conjectures, could be an important — and ordinarily irreversible — influence on gene expression, stabilizing the cell type. That could, in turn, partly explain why reprogramming a skin cell to an iPS cell is so inefficient: reprogramming techniques introduce the appropriate transcription factors but not the matching transcription machinery.

Better elucidated, although far from entirely mapped out, is the network of enzymes — chiefly DNA methylases and histone-modifying enzymes — that, by depositing or removing chemical groups (epigenetic marks) on chromatin, stably repress or activate gene transcription. For example, two huge sets of protein complexes called Polycomb and Trithorax battle it out with one another by methylating different amino acid positions on the histone proteins that structure DNA in the chromosomes, as well as by cancelling out one another's previous histone methylations.

Semipermanent markers

In the absence of an overriding signal (from transcription factors, for example), epigenetic marks are passed down through mitotic divisions. Without this persistence of epigenetic memory, every dividing cell would be courting chaos, unable to take on the role it should inherit from its parent.

Some epigenetic marks survive even the vigorous reprogramming visited upon a somatic nucleus by somatic cell nuclear transfer (SCNT). In experiments published early this year in Nature Cell Biology, Ray Kit Ng and John Gurdon at the University of Cambridge, U.K., found that a particular genomic region associated with a gene expressed early on in muscle-cell development (and expressed exclusively in muscle cells) retained its 'active' epigenetic setting17. They also saw expression of the gene in question in embryos generated through two serial SCNT passages — even in cells of non-muscle lineage.

It appears that epigenetic patterns are both stabilizing and stable, but not unalterable. Whereas epigenetic marks maintain cell states, transcription factors, whose DNA-specific liaisons with chromatin tell epigenetic modifiers where to perform their work, can determine those cells' fates.

“Modifying the chromatin state can enhance the reprogramming process, but the specificity looks like it's coming from the DNA-binding transcription factors,” says molecular biologist Rick Young at the Whitehead Institute in Cambridge, Massachusetts. Young says, “We've got about 1,500 of these things encoded in our genome — something like 7% of all our genes.” Clearly, there will be less guesswork involved in trying to convert one differentiated cell type to another once the constellations of transcription factors involved in effecting and sustaining cell differentiation are mapped.

But even normal differentiation is a dimly lit area. There are not many well-defined programs for guiding cells in vitro all the way down the development path from ES cells to those types that might be useful for cell therapy, Young says. However, the transcription factors necessary to do so have been pretty well worked out for at least one lineage — motor neurons18 — and Young's, Zaret's and Tjian's labs are among those trying to find which other transcription factors are key to lineages that culminate in several cell types, including hepatocytes, dopaminergic neurons and skeletal and cardiac muscle.

More than identification

There may be more to cell-fate analysis than merely identifying key transcription factors. Knowing how much of a factor to toss into the mix, and when during the cell cycle, can be nontrivial. Ihor Lemischka, director of the Black Family Stem Cell Institute, affiliated with the Mount Sinai School of Medicine, both in New York, notes that adding three different amounts of the same transcription factor has, in some cases, led to the generation of three different cell types19.

Are there any hard limits to cell reprogramming? “I don't know how common it will be,” Lemischka says. “So far, in the few cases of walking differentiated Cell A to Cell B that have been achieved — although it's exciting — the two cells are not necessarily that removed from a common predecessor.”

Whereas the backward leap from a skin cell to an iPS cell occurs at low yields and requires weeks of culture, it's a shorter jump from a pancreatic exocrine cell to a beta cell. In less than a week, 20% of the infected cells studied by Melton's team became functioning beta cells. Even though virally introduced gene expression shuts down within a month, the effect has lasted for well over six months and may be permanent — it's too early to tell yet, Melton says.

His success with pancreatic-cell conversions leads Melton to wonder about the medical potential of similar conversions among relatively closely related cells in other organs, such as the brain. In some neurodegenerative disorders, he says, “neurons have degenerated but nearby support cells — astrocyte and glial cells — are perfectly healthy.” In addition, Melton says, “human liver can regenerate. So if we could get a liver cell to become a beta cell, it would be a significant advance.”

Meanwhile, Melton's team and many other investigators are attacking a barrier that will have to be overcome before any cell-type conversion can be put to use in regenerative medicine: the possibility that virally introduced genes may integrate randomly into the genome, creating chimaeric and potentially oncogenic stretches of DNA. These investigators are almost unanimously optimistic that the solution to this problem — in the form of small-molecule mimics of transcription factors or the extracellular signals that influence their activity — will arrive sooner rather than later.