Credit: Getty Images/Jessica Kolman

When proposing a human clinical trials, ethics demand evidence. The International Campaign for Cures of Spinal Cord Injury guidelines state: “A study involving risk to human subjects is not ethically defensible if it is not scientifically defensible”1.

Outi Hovatta, a researcher at the Karolinska Institute in Stockholm, Sweden, well known for her studies of the tumourigencity of human embryonic stem (hES) cells, thinks the science in this field is not yet defensible. “It is dangerous to inject hES cells into the spinal cord,” she warns. Even after a population of hES cells has been differentiated into neural cells, she says, they still form teratomas in the brains and spinal cords of animal receiving transplants.

When U.S. regulators halted plans for the first clinical trial involving cells derived from embryonic stem cells, the reaction of many in the stem cell community smacked more of relief than impatience. ES cells are powerful and unpredictable, went the reasoning, so despite their vast therapeutic promise, their use in human patients should not be rushed. (The sponsoring company, Menlo Park, CA based Geron Inc., says it is addressing the Food and Drug Administration's concerns and will launch its trial in spinal cord injury when it has done so.)

It is difficult to evaluate risk even for interventions involving small molecules or antibodies, but stem cells raise harder questions. Although fetal and adult stem cell isolates aren't usually tumorigenic, lines derived from hES cells or other pluripotent cell types can be. The hope that in vivo tissue-specific signals could instruct hES cells to differentiate into the therapeutic cell type after transplantation hasn't panned out. Instead, Hovatta's work shows a high degree of malignancy in cultured lines. She believes that even a single pluripotent cell can cause cancer. Now, researchers pursuing hES cell–based therapies plan to differentiate cells prior to transplantation, but problems remain. The techniques necessary to expand and specialize the cells can increase the risk of genetic instability or drive cells into unwanted pathways. Worse, cells' potential missteps are hard to spot.

Cancer: the heart of the problem

Animal studies assessing risk have produced muddled results. Though teratomas rarely occur in mice receiving hES cell–derived neural cells, most mice receiving neural cells derived from mouse ES cells develop tumors2. Monkeys that get transplants of haematopoietic cells derived from monkey ES cells suffer from the same problem3. The key to avoiding teratomas in humans, most researchers think, is to use fail-proof techniques to weed out stray hES cells and to monitor transplants over long periods to watch for tumours.

The fact that pluripotent cells love to divide is a “real barrier,” says Andre Terzic, a researcher at the Mayo Clinic in Rochester, Minnesota, and getting around this is a “critical step in clinical translation.” Terzic and other researchers are pursuing dual strategies to exclude pluripotent cells from transplants for heart disease. In one strategy, explains Terzic, “you allow pluripotency to manifest itself. Then you fish out the cells that are predestined to a specific lineage.” Specifically, proteins upregulated during cardiogenesis are used to pluck out the desired cells4.

The other strategy takes the opposite approach, pushing pluriptotent cells down a specific lineage with a cocktail of growth factors that mimic natural signalling pathways. The derived cells show lower oncogenic marker expression compared to their pluripotent counterparts. Once transplanted into mice with injured hearts, they make cardiomyocytes, not tumours, and heart function improves5,6,7. Though his own work focuses on coaxing cardiomyocytes from stem cells, Terzic believes that these twin approaches—guiding and culling—might generalize to other cell types, such as those found in the brain and pancreas.

The four commandments

All these mouse-made assessments share a huge uncertainty, irrespective of transplanted cell type. Even if scientists manage to create a virus-free, genetically stable, clinically pure population of derived cells, the xenograft models of disease or injury may not accurately predict the same response in humans because the mice lack functioning immune systems.

To even begin to assess replacement therapies using preclinical data, Irving Weissman, director of the Stanford Institute for Stem Cells and Regenerative Medicine in California, ticks off four thresholds of clinical effectiveness: “First, you have to show the cells home to the diseased or injured tissue,” he says. “Second, they must engraft — not just fuse with cells that are already there. Third, they have to function. Fourth, they must persist”8.

Weissman's four rules are captured by a lab technique that emerges in nearly every discussion about risk assessment: in vivo fate mapping. Genetic tags can help researchers and bioethicists get a handle on where the human cells go after transplantation and what they do when they get there. The simplest tag uses genetic recombination to introduce a fluorescent marker, such as green fluorescent protein, into a cell. Because stem cells self-renew and differentiate, scientists can track both the stem cells and their progeny. Homing, engraftment, cell fate, persistence and tumour formation can be assessed using a system like this, though signals can be confounded because the tags tend to be cytotoxic.

Grafts must be characterized to see whether the cells trigger endogenous mechanisms of repair or whether they contribute to the repair directly. For complex neuronal diseases where migratory potential and engraftment is uncertain, stem or progenitor cells may be better at protecting functioning neurons rather than replacing non-functional neurons. A trial already underway in Batten's disease tests the ability of fetal neural progenitors to clear out toxins that patients' own brains cannot. A similar approach may work on other neurodegenerative diseases, such as Parkinson's or stroke. Although it isn't necessary to understand the precise mechanism of repair or renewal, information about the pathology of the disease, the probable behaviours of the administered cell type and the effects of the transplants on the host environment (and vice versa) will be useful to predict what might go wrong.

The risk calculation depends on which cells are used and how they are cultured. For the central nervous system, flexible progenitors rather than terminally differentiated neurons will be necessary for long-term repair, but neural progenitors are less predictable and more proliferative than neurons, and likely harder to separate from stray pluripotent stem cells. “I would look for anomalous preclinical results — iPS and hES cells can make any cell type, so presence of the wrong tissues would be disturbing,” says Theo Palmer, a Stanford University neuroscientist and chair of the university's Stem Cell Research Oversight Committee. Palmer emphasizes that culture methods have trade-offs. When it comes to differentiation, more passages of a neural stem cell line mean less of a chance of stray pluripotent cells, but also more time to accrue genetic instability and other problems.

Get on with it!

The first ES cell human studies will be critical; they could kindle suspicion or enthusiasm. Katherine High, a paediatrician and director of the Center for Cellular and Molecular Therapeutics at the Children's Hospital of Philadelphia, has seen much in the tumultuous years since the death of Jesse Gelsinger caused the FDA to shut down gene-therapy trials across the country. Her work uses gene transfer to replace defective copies of factor IX, the clotting agent missing in haemophiliacs. High says the first gene transfer failures and the crackdown on trials produced a mountain of preclinical information — and that they are still shovelling through it. “Our clinical experience lags these overwhelming troves of data,” she says.

High recites a hard-won list of issues that the gene-transfer field tackled over the past fifteen years: gene silencing, transmission through the germline and the environment, plus genotoxity, immunotoxicy and other toxicities. “The stem cell field can use some of these as reference points,” she says. “No new problems [in gene transfer] are arising.”

Trying the unknown

As oversight committees and ad hoc experts wrestle with risk-assessments, consensus is emerging on mandatory lines of evidence. Species-specific interactions between experimental transplants and recipients confound analysis, but some demonstration of efficacy in animals mimicking human disease will be necessary9. In certain cases, approval may require evidence in large animal or primate models. How and where cells are administered must also be considered. If cells are transplanted deep within an organ, even the delivery device may impact safety and efficacy. The International Campaign for Cures of Spinal Cord Injury recommends that the first trials transplant only to thoracic regions of the spinal cord. Cervical or lumbar sites are riskier because an adverse event could affect respiratory, upper limb and lower limb function.

The immortality of stem cells is a double-edged sword for individuals receiving stem cell transplants. The effects of conventional drugs tend to last as long as the drug stays in the body, but cells' effects — good or bad — could last a lifetime. Restoration in bowel function would significantly improve disabled patients' quality of life. On the other hand, an inoperable tumour caused by the transplant may mean a lifetime of peripheral pain.

Can there be too much caution? As long as patients fully understand the dangers they face, how far should our obligations go to protect them? There is a palpable (and necessary) tension among bioethicists, basic stem cell researchers, transplant surgeons, and their patients about how quickly stem cell trials should proceed. Some bench researchers say slow down: understanding the mechanisms of repair and renewal will lead to better clinical outcomes.

It's not just a one-way transmission of knowledge. Last time I checked, the title on my door said director of clinical research.

Clinicians respond that their patients are sick or dying and the evidence seems sufficient. I asked a well-known stem cell transplant surgeon (who asked not to be named) if clinical trials should wait for more evidence from the bench. He said that his basic scientific collaborators have just as much to learn from his clinical research as the other way around. “It's not just a one-way transmission of knowledge. Last time I checked, the title on my door said director of clinical research.”

But improvements could mean that early adopters receive inferior care. Even if a therapy has no adverse effects, subjects who leap into the first clinical trials may find themselves excluded from subsequent trials using improved techniques. Rigorous clinical trials will likely accept only untreated patients, as previous cell treatments would be a confounding variable. “This is high risk, high reward research,” explains University of Wisconsin neuroscientist Clive Svendsen, who is working on a gene-transfer approach for ALS using fetal neural progenitors.

Balancing tradeoffs such as these have long been a part of the larger discussion about when to begin the first human experiments using untested technologies. As we move into this promising frontier, we should take care to be informed by past experiences, but not immobilized by them.