When something walks like a duck and quacks like a duck, scientists should be wary of decoys.

Bookmarks can help you quickly find just the right page among hundreds or even thousands. But we know better than to confuse a bookmark for the page it points to, and we all know that bookmarks can get misplaced. So it is with stem cells. Cell-surface proteins and other markers may seem a convenient way of identifying stem cells, but they're no substitute for assessing what the cells can really do.

“A high percentage of discoveries that don't hold up fail to hold up because people weren't careful enough about the markers they used to identify the stem cells,” Sean Morrison

Sean Morrison, director of the University of Michigan's Center for Stem Cell Biology, thinks misplaced trust in stem-cell markers is one of the field's biggest problems. “A high percentage of discoveries that don't hold up fail to hold up because people weren't careful enough about the markers they used to identify the stem cells,” he says. That's because many tissues lack good functional assays, or markers, or both. So when researchers hear of a stem-cell marker from another system that just might apply to their own, they tend to latch on.

Label retention routed

Morrison recently undermined the case for one cherished 'magic marker' for stem cells that equated 'stemness' with cells' retention of the nucleotide-analogue label bromodeoxyuridine (BrdU). The longest-running rationale for this marker holds that stem cells divide more slowly than the progenitor cells they generate. So, after the label is withdrawn from the growth medium, the relatively quiescent stem cells hoard the stash of label already deposited in their genomes, whereas label incorporated into progenitor cells' DNA is steadily diluted as those cells proliferate.

But there's at least one fly in that ointment. If a progenitor undergoing a final division to become two long-lived, differentiated cells happens to get labelled, those cells can retain the label for years, confounding the assay.

An alternative notion posits that BrdU is an even more specific marker of stemness. The 'immortal strand' hypothesis proposes that when a stem cell divides into another stem cell plus a differentiation-committed daughter cell, the former retains its original DNA strands with the newly synthesized (and, presumably, imperfectly copied) strands shunted off to the latter1. Thus, you would expect stem-cells' labeled strands to be always retained in those stem cells.

The trouble is that the label-retention hypothesis has not been validated, says Morrison. “A lot of people take it as an article of faith that DNA-label-retaining cells are highly enriched for stem cells,” even in tissues where stem cells cannot be clearly defined by phenotype or function.

Morrison found himself thinking: “If we're going to build a whole field based on this proposition, then somebody should test it.” So he did, in the most extensively characterized system available: haematopoietic stem cells (HSCs).2 His team pulsed mice with BrdU and examined their blood at various times to see where the labeled DNA turned up.

Labeled DNA gradually disappeared from HCSs, Morrison says, and more than 99% of those cells that did retain BrdU were not stem cells. Although at least one well-designed study has shown evidence supporting BrdU retention in another tissue, the hair follicle3, Morrison believes label retention is one of the worst markers that has ever been used in published papers. “It's your textbook example of a terrible stem-cell marker.”

It's also ubiquitous. “Until half a year, or a year, ago I never heard anybody damn or doubt label retention as a stem-cell marker,” says Hans Clevers, director of the Netherlands Institute for Developmental Biology's Hubrecht Laboratory in Utrecht. Gut epithelium is renewed from cells located in the crypts at the base of intestinal villi. In a recent study4, Clevers' group found that the well-known Paneth cells in the crypts had been falsely identified as the gut epithelial stem cells. That widespread assumption had been based largely on observations that the cells retained BrdU.

Clevers's team made transgenic mice in which the gene lacZ, which encodes an enzyme that can yield an intense blue stain in tissue sections, was tethered to the promoter for Lgr5, a gene expressed only, and rarely, in the crypts, making it a possible marker for stem cells. The cells that stained blue were not the Paneth cells, however, but small, sparse cells called crypt-based columnar cells (CBCs).

To visualize the daughter cells of these putative gut stem cells, Clevers created another transgenic mouse in which Lgr5-expressing cells would glow green while alive and would stain blue when fixed in tissue sections. In addition, the transgenic mouse was designed so that the descendants of these cells would permanently inherit a penchant for staining blue, even though they no longer expressed Lgr-5 because they were no longer stem cells. If Clevers's hypothesis that Lgr5-expressing cells were gut stem cells was correct, he would expect to see a gradually elongating blue ribbon of differentiating daughter cells creeping up from the CBCs into the intestinal villi.

“That's what we see happening,” he says. Clevers's lab is getting similar results with Lgr5 in several other tissues — although, having dispatched one 'universal stem-cell marker' to meet its maker, he's in no hurry to coronate another one quite yet.

Oct4 is only for the young

Rudolf Jaenisch at the Whitehead Institute for Biomedical Research in Cambridge, Massachusetts, recently published a paper debunking the ubiquity of another stem-cell marker — Oct4, which is indispensable for self-renewal in embryonic stem (ES) cells5. Oct4 has received a lot of press attention lately as an element of the gene cocktail served to fibroblasts to regress them to an ES-cell-like state. “People would like to think it plays a role in adult stem cells, too,” Jaenisch says.

More than 100 papers have appeared reporting Oct4 expression in adult bone marrow stem cells and in progenitors from pancreas, kidney, peripheral blood, mammary epithelium, uterine endometrium, thyroid, lung, liver, dermis and hair follicles — mainly based on detection systems with messy readouts such as antibody staining and gene transcripts. “We got very suspicious of this data,” Jaenisch says.

His team selectively knocked out Oct4 in one tissue after another in mice and checked to see whether there was any effect on gut, skin, liver, brain or bone marrow. There was none. Then they stressed the mice, causing them to regenerate these tissues. Did deleting Oct4 cramp the self-renewing capacities of irradiated intestinal epithelium or bone marrow, wounded skin or hepatectomized liver? “Zip, zip, nil, zero,” says Jaenisch. “No effect whatsoever.”

Indeed, was Oct4 even expressed in adult stem cells? Examining a number of tissues using PCR, Jaenisch's group did find Oct4 expression — but at levels about 10,000-fold lower than in ES cells.Completely deleting Oct4 did not change the signal a bit. “These low expression values — well, who cares. They're totally irrelevant, just an artefact,” Jaenisch says.

Babies from bone marrow?

Since her postdoc days in the Stanford University laboratory of HSC pioneer Irving Weissman, Amy Wagers, now at the Harvard Stem Cell Institute in Cambridge, Massachusetts, has critically examined numerous claims of the transdifferentiation of HSCs into various tissues6. She uses pairs of histocompatible mice that are surgically joined from the shoulder to the knee and share a common vasculature so that their circulating blood cells mingle. One mouse expresses a transgene for green fluorescent protein (GFP) and the other does not, so the cells from each mouse can be distinguished. Using this system, Wagers and her team7 challenged a report that the supply of oocytes in a female mouse could be replenished throughout adulthood by the transdifferentiation of cells in circulating blood8.

That report was based largely on immunohistochemistry, analysis of mRNA and morphological studies of ovarian cells.8 Had it held up, the consequences would have been momentous, says Morrison. Not only would the biology underlying most work at fertility clinics be up-ended, but the observations implied that women who received bone-marrow transplants might give birth to babies that were genetically unrelated to them.

However, none of the 66 oocytes that Wagers and her colleagues harvested from their non-GFP conjoined mice were green, whereas all of the 80 ovulated oocytes obtained from the GFP-expressing mice were. The researchers did see GFP-expressing cells in the ovaries of non-GFP mice, but these clumps of tissue also expressed definitive haematopoietic markers — they were still blood cells7.

“I spent a lot of time looking for potential conversion of blood cells to cells with other phenotypes. And it was never there,” says Wagers.

Constant hearts

Research by Wagers and others9,10 also casts doubt on assertions11 that bone-marrow-derived cells transdifferentiate to become cardiac stem cells and regenerate damaged mammalian heart tissues. The newer findings call into question a number of current clinical trials of bone-marrow injection into the myocardium.

“The cardiology field has been adopting stem-cell therapy rather quickly, before we even understand the biology of stem cells in the heart,” says Kenneth Chien, director of the Harvard Stem Cell Institute's programme on cardiovascular disease. “Most clinical studies so far have been done either by trying to coax adult cells from elsewhere into transdifferentiation or by harvesting putative resident progenitors from adult heart — even though it might not be clear by lineage tracing that that's what they are.”

As many as three different cell populations have been put forward as resident heart progenitors in adult mammals, but none of these populations has been tested using rigorous modern technologies, Chien says.

Chien used lineage-tracing techniques similar to those used by Clevers to show that single ES-cell-derived cells that were positive for a marker called Islet-1 could differentiate into all key tissue types, including smooth muscle, cardiac muscle, endothelial cells and the conduction system, that are responsible for the architecture of about two-thirds of the heart12. These Islet-1-positive cells are present in a mouse only until the perinatal period, however, so are unlikely to contribute substantially to any cardiac regenerative capacity in adults.

Besides, would an organ that has appeared for so long to be composed entirely of terminally differentiated, non-dividing cells really harbour three or four distinct stem-cell populations? As the bar for inclusion in the winner's circle of stemness is raised progressively higher by sceptical researchers armed with increasingly rigorous lab tests, the question may answer itself before too long.