Stem cells

Cell fusion causes confusion

'Transdifferentiation' is a poorly understood process invoked to explain how tissue-specific adult stem cells can generate cells of other tissues. New results challenge its existence.

Most readers have probably heard of stem cells — unspecialized cells that can generate a variety of more specialized cell types. Not so long ago, it was thought that only embryonic stem cells (ES cells) could generate all the different cell types in the mammalian body1, and that adult stem cells are more restricted in their developmental potential. Recently, however, many researchers have observed that adult neural and haematopoietic stem cells are not limited to forming cells of the central nervous system or blood, respectively, but can also generate liver, intestinal cells, and heart and skeletal muscle. It has been proposed that they do so by 'transdifferentiating' (acquiring broader developmental potential). If true, this is enormously important — it implies that tissue-specific adult stem cells harbour much, if not all, of the potential of ES cells. This would remove the need to collect stem cells from human embryos for clinical purposes, thus overcoming many of the political and ethical barriers to stem-cell therapy.

Now, however, Ying et al.2 and Terada et al.3 (see pages 542–548 of this issue) raise doubts about whether transdifferentiation actually occurs. The authors set out to show that ES cells can induce neural stem cells2 or bone-marrow cell populations containing haematopoietic stem cells3 to transdifferentiate into embryonic-like stem cells when cultured together in vitro. Instead, they found that the adult stem cells spontaneously fused with the ES cells and took on their characteristics — an event that may previously have been misinterpreted as transdifferentiation.

One test of transdifferentiation in vitro is based on the idea that the developmental limitations of tissue-specific stem cells are dictated by their environment, and that new signals that relax these restrictions might be provided by cells from a different tissue. In several experiments, for example, neural stem cells were isolated from the central nervous system of mice, labelled with a marker protein such as β-galactosidase or green fluorescent protein (GFP) to make them easy to trace, and cultured with other cell types. Amazingly, when neural stem cells expressing such markers were cultured with myoblasts (precursors of skeletal muscle cells) or embryoid bodies (produced from ES cells), they differentiated into β-galactosidase- or GFP-labelled skeletal myotubes (muscle cells)4,5,6. Moreover, GFP-labelled neural stem cells differentiated into GFP-labelled heart-muscle cells when mixed with heart-muscle cells from newborn rats7. It was concluded that the stem cells achieved this by transdifferentiation5,7, although there are other interpretations (Fig. 1a).

Figure 1: Can adult stem cells transdifferentiate into other cell types?

a, In previous work4,5 neural stem cells, labelled with β-galactosidase, were cultured with myoblast cells or embryoid bodies, producing β-galactosidase-labelled muscle cells. This was interpreted as evidence that the stem cells received signals that caused them to transdifferentiate into muscle cells. But there are other possibilities. The stem cells might have been contaminated with muscle precursor cells; a few stem cells might have mutated; or the stem cells might have fused with myoblasts or embryoid-body cells. b, Ying et al.2 cultured puromycin-resistant neural stem cells that expressed green fluorescent protein (GFP) with hygromycin-resistant embryonic stem (ES) cells. After selection, the surviving cells expressed GFP, were resistant to puromycin and hygromycin, and had double the normal amount of DNA (4n versus 2n). This suggests that the cells arose by fusion. c, Terada et al.3 cultured GFP-labelled, puromycin-resistant bone marrow cells with ES cells. The resulting cells expressed GFP, were resistant to puromycin and had ES-cell properties. They also had double the usual DNA content.

The experiments by Ying et al.2 and Terada et al.3 were different in that both starting cell populations were tagged with distinct markers, enabling the authors to trace the fate of each cell type. Ying et al.2 isolated neural stem cells from mice in which two genes — one encoding GFP and one encoding a protein that confers resistance to the antibiotic puromycin — are linked to a control region from the Oct4 gene, which is expressed only in ES cells8. These cells were cultured with ES cells that had been genetically engineered to include the hygromycin phosphotransferase gene, which confers resistance to the antibiotic hygromycin. The culture medium was an ES-cell growth medium that contained puromycin, to select against these original ES cells.

After two to four weeks, the authors recovered several cell colonies that expressed GFP and were resistant to puromycin, indicating that they were derived from neural stem cells but had become ES-cell-like with respect to their growth characteristics and the Oct4-driven expression of these markers. Taken alone, this might have been interpreted as evidence of transdifferentiation. Surprisingly, however, the cells also expressed hygromycin phosphotransferase (which could only have come from the ES cells); moreover, 18 independently isolated cell lines had twice or nearly twice the normal DNA content. Ying et al. conclude that the ES-like cells arose from fusion of the neural and embryonic stem cells, rather than from transdifferentiation (Fig. 1b).

Terada et al.3 independently took a similar approach (Fig. 1c) by culturing ES cells with bone-marrow cell populations, rich in haematopoietic stem cells, that expressed GFP and the puromycin-resistance gene. Puromycin was added to eliminate the ES cells, and the signalling molecule interleukin-3 was withdrawn to suppress the growth of bone marrow cells. This resulted in proliferating colonies of GFP-labelled, puromycin-resistant cells that mimicked ES cells in their morphology and growth kinetics, expressed ES-cell marker proteins, and differentiated into heart-muscle cells after the removal of growth factors. These cells, too, might have been thought to arise through transdifferentiation. But Terada et al. found that 2 cell lines had 4 sex chromosomes (XXXY, the Y chromosome having come from the ES cells), and 11 had twice the usual amount of DNA, again suggesting that cell fusion had occurred.

Could cell fusion have been misinterpreted as transdifferentiation in previous studies? A complicating factor is that transdifferentiation has been seen to occur in 7–57% of the neural stem-cell population in co-culture assays5,6, whereas Ying et al.2 and Terada et al.3 detected cell fusion at the lower frequency of 1 in 104 to 1 in 5 × 105. The actual rate of fusion may be higher: the stringency of the selection approach and the long culture period required to assay cell fusion (compared with transdifferentiation experiments, which are completed within four to five days5,7) might have resulted in deceptively low levels of detectable cell fusion because of cell death. Alternatively, cell fusion may indeed be exceedingly rare, and thus unable to account for transdifferentiation by adult stem cells; moreover, the stringent approach2,3 may have driven cell fusion artificially, as the only means by which cells could survive and proliferate. It remains to be seen whether cells fuse under less selective conditions.

In addition, cell fusion has been detected only in vitro so far2,3. But both in vitro and in vivo observations have been taken as evidence of transdifferentiation. For instance, neural stem cells and bone marrow cells transplanted into mice have been shown to contribute to the central nervous system, blood, liver and muscle (see refs 2, 3). Nevertheless, it remains formally possible that cell fusion or alternative mechanisms of cell conversion may be significant in vivo, making it important to re-examine the in vivo data. These issues are particularly acute in instances where tissue-specific stem cells convert into multinucleated cells, such as skeletal myotubes and Purkinje neurons5,9.

These questions notwithstanding, it is remarkable that cells can merge at all given how complicated it is to fuse membranes. Intracellular membrane fusion, for example between transport vesicles and the plasma membrane, involves intricate molecular machinery, including proteins known as SNAREs and Rabs10. The mechanism of cell fusion is unknown, but interactions between plasma-membrane molecules in the fusing cells might tether the cells together, partly mimicking the function of SNARE and Rab proteins.

Finally, an interesting feature of the fused cells2,3 was that they had many characteristics of ES cells, suggesting that the embryonic genetic programme may dominate over that of adult stem cells. So cell fusion might afford a means of identifying proteins that make ES cells so versatile. Most important, however, the new results call for a detailed genetic analysis of cells that have been thought to undergo transdifferentiation, and highlight the need to define this process mechanistically so that it can be more rigorously diagnosed in future. Until then, it may be unwise to make generalizations or policies that limit efforts to understand the still-puzzling behaviour of stem cells.


  1. 1

    Gardner, R. L. & Beddington, R. S. J. Cell Sci. 10 (suppl.), 11–27 (1988).

    CAS  Article  Google Scholar 

  2. 2

    Ying, Q.-L, Nichols, J., Evans, E. P. & Smith, A. G. Nature 416, 545–548 (2002); online 13 March 2002 (DOI 10.1038/nature729).

    ADS  CAS  Article  Google Scholar 

  3. 3

    Terada, N. et al. Nature 416, 542–545 (2002); online 13 March 2002 (DOI 10.1038/nature730).

    ADS  CAS  Article  Google Scholar 

  4. 4

    Clarke, D. L. et al. Science 288, 1660–1663 (2000).

    ADS  CAS  Article  Google Scholar 

  5. 5

    Galli, R. et al. Nature Neurosci. 3, 986–991 (2000).

    CAS  Article  Google Scholar 

  6. 6

    Rietze, R. L. et al. Nature 412, 736–739 (2001).

    ADS  CAS  Article  Google Scholar 

  7. 7

    Condorelli, G. et al. Proc. Natl Acad. Sci. USA 98, 10733–10738 (2001).

    ADS  CAS  Article  Google Scholar 

  8. 8

    Yeom, Y. I. et al. Development 122, 881–894 (1996).

    CAS  Google Scholar 

  9. 9

    Priller, J. et al. J. Cell Biol. 155, 733–738 (2001).

    CAS  Article  Google Scholar 

  10. 10

    Clague, M. J. Curr. Biol. 9, R258–R260 (1999).

    CAS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Fred H. Gage.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Wurmser, A., Gage, F. Cell fusion causes confusion. Nature 416, 485–487 (2002).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Sign up for the Nature Briefing newsletter for a daily update on COVID-19 science.
Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing