Thomas Graf, of the Centre for Genomic Regulation in Barcelona, is known for his work on converting one blood cell type to another. Monya Baker spoke to him about how this can be done — and why.

How do you turn one cell type into another?

Thomas Graf, Centre for Genomic Regulation, Barcelona, Spain

It's a matter of learning the changes in transcription factor networks that are instrumental in dictating cell fate. As cells develop away from each other, the transcription factor networks are more and more different, and we have to do more and more to turn one into the other.

As [the cells] go on differentiating they become less plastic, less sensitive to the environment. But you can still push them around if you know the key transcription factors. If you restore the expression of these transcription factors, you can turn [the cells] into an alternative fate.

In your studies of transdifferentiation, what stands out as a particularly exciting day in the lab?

In the early 1990s we had developed chicken cell lines that could be switched from an erythroid [red blood] into a myeloid [white blood] fate by activated Ras and we found that GATA-1 became downregulated. We then asked what would happen if we expressed GATA-1 in the myeloid lines, expecting at best the ectopic activation of some haemoglobin production. To our surprise, the cells completely changed phenotype. They not only showed upregulation of erythroid genes but also the silencing of myeloid markers. I was very excited about this, although I did not immediately realize the full implication of this observation. We later found out that GATA-1 inactivated the myeloid regulator PU.1 and that PU.1 could convert erythroid cells back into myeloid cells. These findings led to the concept of transcription factor antagonisms, a principle now believed to apply generally to lineage decisions.

Are all sorts of conversions possible?

This path [from a specialized cell] back to an embryonic stem (ES) cell may be a unique path because you are basically creating a cell that is at ground state, and you can say that this is where all cells gravitate to. They're very special, proliferative, immortal.

The big question is whether cells can transdifferentiate into very distant cell types if they don't go back to the ground state. We really know very little about how this works — whether it is a cell turning into another in a direct way or whether there is a transient state that resembles the closest common precursor. We also don't know if there are insurmountable barriers to converting, a fibroblast, say, into any cell type you want. I think this is a completely open question.

But haematopoietic stem cells can make so many types of cells. Are they special?

Perhaps. In early haematopoietic tissues such as the yolk sac, blood cells predominantly arise as separate erythroid and myeloid lineages. It appears that later in development, the erythroid and myeloid transcriptional programs fuse in rare mesenchymal precursors, generating myeloid-erythroid progenitors. Curiously, at the same time these cells become essentially quiescent and acquire self-renewal and lymphoid differentiation potential, all hallmarks of haemopoietic stem cells. It will be fascinating to understand how this all happens.

Do you think that the plasticity that you and others have found in blood will be found in other tissues as well?

I'm convinced, yes. Wherever you have options to become one cell or another it is surely a reflection of a difference in key transcription networks. And it looks like the principles that we and others have worked out in the haematopoietic system are acting both synergistically and also, very importantly, in an antagonistic fashion. In the end it all comes down to what the balance is between the key transcription factors. This seems to be true for many differerent cell systems; it happens, for instance, in early decisions to become trophectoderm versus embryo proper.

Why would it be useful to go straight to another cell type rather than going through an ES cell state first?

If we can do this directly it would simplify therapeutic protocols tremendously. It is never going to be trivial for a clinic to generate induced pluripotent stem (iPS) cells for a patient and then coax these cells into the cell type they require for cell therapy in large numbers without the danger of producing contaminating cells that might be tumorigenic. If you had a way to do this directly, it would be an enormous advantage.

We have already seen the first spectacular example that it might be possible in the recent paper by Doug Melton where they converted exocrine into islet cells in an animal. (See Smash the (Cell) State!) So that's even better, you don't even have to do any cell culture. If you can do it in a particular organ, you can repair organs in an almost non-invasive way.

How do you know that the new cell type is really the cell type that you want?

It's an important question, and one that the iPS community is asking itself. I don't think that at the moment we know. In our studies we have made artificial macrophages. We have done a number of assays which have convinced us that they are very, very similar to normal macrophages, but in our system we didn't have the capacity to test the functionality of these cells in vivo. There is no good macrophage-deficient mouse model yet.

However, I would argue that even though this is a very important question, if you are interested in cell therapy and the cells do their job, you don't care if they are to the last dot the spitting image of the real thing.

What's the difference between reprogramming and transdifferentiation?

Cell reprogramming involves the generation of ES cells, or cells that closely resemble them. Induced transdifferentiation/lineage conversions, on the other hand, do not seem to go through an iPS cell state, and by that criterion, the two processes are different. However, it is still possible that, at least in some instances, lineage conversions also involve some kind of retrodifferentiation. It seems premature to coin different terms for processes that we only understand incompletely. This is particularly true for the induced transdifferentiation/lineage conversions.

Can reprogramming be considered a sort of transdifferentiation?

It can, but here we are entering the field of semantics

How do you know what cells should be called stem cells?

By the classical definition, stem cells can both self-renew and differentiate into a variety of cell types. However, in the extreme case of induced iPS cell reprogramming, all somatic cells can be considered as potential stem cells. Admittedly, this is an artificial situation. However, there are also cases where differentiated cells, such as hepatic and pancreatic cells, can be induced to self-renew after injury and form more differentiated progeny. Although these cells are not called stem cells, they are formally not so different from muscle or spermatogonial stem cells that can also be induced to proliferate and to convert into one type of specialized progeny.

Sui Huang has recently suggested that random fluctuations within differentiating cells play a large role in determining their fate. What do you think?

Even though this [population of cells] is not a real stem cell, it's a good model for what might really be happening at the stem cell level. (See Change and the single cell) Probably the hallmark of multipotent cells is that you have fluctuations of transcription factor networks that create continuous heterogeneity, creating extremes that differ from other extremes, a relative abundance of one critical transcription factor over another. However, these states are only metastable, so these extreme states seem to be able to go back to the middle and change back to the other extreme. There seems to be this readiness of these multipotent progenitors to enter into whatever lineage is necessary. We don't know what the oscillator is and exactly how it is regulated, or if it is simply a reflection of what some people call noise. But it may very well be an important mechanism to react to environmental cues.

What questions will fuel stem cell science over the coming decades?

In 2005, the journal Science compiled a list of 125 big questions that stand a reasonable chance of being answered within the next 25 years. They included 'How can a skin cell become a nerve cell', and, 'could scientists come up with a cell-free bath that turned the clock back on already differentiated cells'? Probably by the time the last was formulated, Yamanaka's team had already found the answer. (See Skin cell to stem cell)

In attempts to translate stem cell research into the clinic its actually very hard to predict which aspects will resist solving. Often things move fast where one least expects it.

What advice do you have for scientists entering this field now?

Do not join the iPS craze unless you are pretty sure that you have a unique angle or question. Directed transdifferentiation could become a hot new area; it is not yet too crowded.

What's the most important advice you've received as a scientist?

I have a tendency to compare myself to people that I consider far more brilliant than me, and sometimes I feel really down on myself. I remember that Harold Varmus once told me that I shouldn't worry so much about that. He thinks the most important thing is to develop a good system where you can go step by step and do experiments that can do novel things and maybe answer questions that you weren't able to ask before.