Mickie Bhatia1

Cells derived from embryonic stem cells in culture may be only 'cartoon' versions of the real thing, but that doesn't mean they are not useful.

When embryonic stem (ES) cells fail to perform as we hope, we worry that they are 'only' artefacts of culture1, that they do not reflect in vivo conditions. But we also know that these are artefacts that can, at least in the case of mouse ES cells, generate live mice that run around in a cage. Clearly the ES cells can generate all the types found in the body. While the definitive experiments cannot — must not — be attempted with humans, but it is likely that human ES cells would have a similar capacity. And as these artefacts can clearly generate all the cell types that make up a mouse in vivo, perhaps there are ways to make them do so in vitro. This single hope motivates scientists to devise methods to differentiate human ES cells into specific lineages that can be used to model disease or, maybe, to replace diseased tissues.

Unlike ES cells, many other types of human stem cells can be isolated, clonally defined, and even studied in their natural niches in vivo. Arguably, the only relevant niche of ES cells is not an embryo but a culture dish. The concomitant worry is that the specialized cells derived in vitro from ES cells are unpredictable simulations — caricatures — of in vivo counterparts. Certainly, when we assess these cells' capacity for self-renewal or proliferation or differentiation, we should remember that their performance in a culture dish (where they do not really need to 'function' in a physiologial sense) may not reflect what the cells would or could do in an in vivo setting.

There is a delicate balance between assuming that these cells are the real thing and despairing that they might not be. Instead of succumbing to despair, however, researchers should focus on figuring out how and when these cultured cells or, rather, their differentiated progeny, can serve as workable substitutes for reality.

Laboratory illusions

The variables inherent in culturing pluripotent cells (both ES cells and their reprogrammed counterparts) make it hard for us to assess which techniques can make the 'real thing' in a dish, be it a heart cell, blood cell or motor neuron. Perhaps most discussed are the vagaries of growing and differentiating ES cells. Most of us stem-cell researchers grow our cells in undefined culture systems that rely on feeder layers and serum. These impact how we can interpret experimental results about what factors affect renewal or about how and whether human ES cells can be differentiated into desired specialized cell types. Incremental advances are whittling away at these sources of variability, but addressing other problems may require more fundamental shifts.

Recreating in vivo processes also seems to be a fruitful starting point. Some of the most successful work in getting human ES cells to become functioning pancreatic cells involved transplanting immature cells into the living pancreas, where the cells could respond to cues from the surrounding cells to differentiate from a less specialized state2.

Many new technologies and disciplines are being brought to bear on the successful differentiation of pluripotent stem cells. Importantly, these are focusing on considerations beyond monitoring the proteins that appear on cell surfaces — the markers that we use to follow the cells' fates. More attention is now being given to three-dimensional structures, shear forces, and pH and oxygen balances — properties that are in continual flux during embryonic development and are likely to play complex roles in the differentiation of all cell lineages. But perhaps just as important as identifying the conditions that drive differentiation is identifying the best criteria to assess it.

Laying paths to differentiation

Like an animated cartoon, we envisage the differentiation steps that occur as cultured human ES cells become more specified and mature to a specific cell type. At any given moment we confirm this movement towards maturation by using genotypic or phenotypic markers. These markers, and their sequence of appearance and disappearance during differentiation, give us confidence that lineage specification is progressing as it does in vivo. However, none of the markers can be tested or correlated to the functional regenerative behavior we hope for to provide direct evidence that true lineage-specific stem or progenitor cells have been generated. Ultimately, our checklist is just a series of best guesses drawn from the observational study of developmental biology.

Lineage tracing can be a boon in stem cell differentiation. It shows the parent and subsequent progeny and gives a visual account of who begot whom. But lineage tracing depends on known markers. If I choose the wrong marker, I could have an accurate result but an inaccurate conclusion. Furthermore, some of the best markers for tracking differentiation are surely still unknown, making lineages of cells literally untraceable. Furthermore, unless multiple markers are used, there is no way to tell when cells arrive at a decision point, where one lineage branches off from another. Lineage tracing done wisely can help us check our checklists, but it cannot generate the list of items to check off.

Even worse, interspecies differences can set up guideposts that point in the wrong direction. As the studies on which lineage tracing is based are done in mice, we could well be generating an accurate map of the wrong neighbourhood when we apply results to human cells. For example, cells within the developing mouse heart produce distinct markers for atria and ventricles. However, when Christine Mummery and her colleagues3 studied developing human hearts, they found that these sarcomeric markers could occur together on atria and ventricles, suggesting that efforts to make human heart cells recapitulate murine markers would be misguided.

It's a circular problem. Without reliable markers, it is hard to generate the desired cell types or know whether you have them. Without robust in vivo models to assess the functional capacity of the cells generated in culture, it's hard to assign relevance to the markers used. How can we be sure that our ES cells have become a tissue-specific stem cell or a multipotent progenitor cell when all we see is differentiation in a dish?

Setting guideposts

Interspecies differences are the most obvious reason that mouse ES-cell recipes don't work with human ES cells. Work published last year revealed a less obvious one: mouse and human ES cells represent different stages of development. Two groups of researchers led by David Pedersen4 and Ron McKay5 created stem cell lines from rat and mouse, respectively, but instead of taking the inner cell mass from early blastocysts, they used epiblast cells from slightly older embryos. These lines seemed very much like human ES cells and grew well in culture conditions optimized for human ES cells. If human ES cells are starting from a different developmental state than mouse ES cells, the differentiation protocols may need to follow a different route to reach the desired final state. Perhaps procedures for differentiating human ES cells would work better if we made our analogies with mouse epiblast cells rather than inner mass cells.

One incredibly useful tool for understanding our caricature ES cells are induced pluripotent stem (iPS) cells. Current questions over whether iPS cells are 'really' like ES cells might provide the framework to figure out how differentiation protocols could produce varying results in different sorts of pluripotent cells. Although most published work focuses, and will continue to focus, on comparing ES cells and iPS cells directly, work has begun on comparing the specialized cells derived from these two types of pluripotent stem cells. This will foster comparison between ES cells, iPS cells and tissue-specific stem cells and it may help to explain why culturing tissue-specific stem cells so frequently fails.

Marking the finish line

Whatever the starting material, iPS cells or various sorts of ES cells, the practical questions are the same. Is it possible to efficiently create cells that repair tissues safely and reliably? Can we characterize their properties well enough to mimic them? If so, can we make cartoon human pluripotent stem cells produce 'real cells' in a dish that function biologically? Can we accept these cells as 'real enough'?

An example from our own work illustrates this point. A few years ago, our laboratory used human ES cells to make a fairly accurate cartoon of haematopoietic stem cells. In vitro tests showed that that we could differentiate human ES cells along the blood lineage, eventually producing cells that visually resembled haematopoietic stem cells collected in blood samples and produced many of the same markers (for example, the cell adhesion molecules PECAM-1 and VE-cadherin, and the protein kinase Flk1 ). And although our ESC-derived cells favoured the granulocyte lineage, both these cells and the natural haematopoietic stem cells had similar proclivities to form colonies, and these colonies represented both myeloid (white blood cell) and erythroid (red blood cell) lineages.

In the first experiments introducing our human ESC-derived haematopoietic cells into mice, we transfused the cells straight into the blood stream. They tended to clump together, however, and most of the mice died of embolisms. It was only when we injected the cells directly into the bone marrow that we saw their capacity to take up residence in that niche. In that case, the human ESC-derived cells established themselves at the injection site and subsequently migrated into the marrow of other long bones, albeit at lower rates than endogenous haematopoietic stem cells would. To increase these rates, we adapted a strategy to boost in vivo growth of the human haematopoietic progenitor cells, a strategy that relied on markers identified in both human and mouse studies, and which had delivered promising results in mice. But it did not have the desired effect on the human ESC-derived cells6. In short, most of our success in these experiments depended on how we delivered the cells; the least successful part was based on extrapolations from known human and mouse biology.

So, what happens in culture may only be a caricature of what happens in the body. When scientists start with cells that are artefacts of culture and differentiate them in an artificial environment, the cells that result may not be the 'real thing'. That does not mean that these cells cannot be useful, just that the routes to making them so may not be what we expect. Clearly, using surrogate markers for an in vivo function will give us only the fuzziest of pictures of what the cells we differentiate are capable of. But with sufficient scepticism, better markers and clever functional assays, our cartoon cells can be invaluable for understanding reality — and perhaps even for supplementing real tissues.

References

1. Zwaka, T. P. & Thomson, J. A. A germ cell origin of embryonic stem cells?. Development 132, 227–233 (2005).  | Article | PubMed | ISI | ChemPort |

2.Kroon, E. & Baetge, E. E. Pancreatic endoderm derived from human embryonic stem cells generates glucose-responsive insulin-secreting cells in vivo. Nature Biotechnol. 26, 443–452 (2008).  | Article |

3.Chuva de Sousa Lopes, S.M. Patterning the heart, a template for human cardiomyocyte development. Developmental Dynamics 235, 1994–2002 (2006).

4.Tesar, P. J. et al. New cell lines from mouse epiblast share defining features with human embryonic stem cells. Nature 448, 196–199 (2007).  | Article | PubMed | ISI | ChemPort |

5.Brons, I. G. et al. Derivation of pluripotent epiblast stem cells from mammalian embryos. Nature 448, 191–195 (2007).  | Article | PubMed | ISI | ChemPort |

6.Wang, L. Generation of hematopoietic repopulating cells from human embryonic stem cells independent of ectopic HOXB4 expression. Journal of Experimental Medicine 201, 1603–1614 (2005).

Author affiliations

1.Mickie Bhatia is Director of the Cancer and Stem Cell Biology Research Institute at McMaster University.