The most straightforward way to find out whether a drug or environmental chemical might harm an unborn baby is to test its effect on a pregnant lab animal. In recent years, however, the thousands of chemicals in need of testing—in food, cosmetics, and medicines, for example—have driven researchers in industry and government to search for in vitro alternatives with the hope of reducing the number of animals required.
Increasingly, embryonic and induced pluripotent stem cells from humans and animals alike have been put to work in such toxicology studies. In 1997, scientists introduced the mouse embryonic stem cell test, or EST, which assesses the effects of a chemical on cell differentiation and death. Today, the EST is one of the most widely used in vitro toxicity assays. Several pharmaceutical companies, such as Pfizer and Merck, use the EST for preclinical toxicity screening, and the US Environmental Protection Agency (EPA)...
“One of the reasons [the EST is] so valuable to us is that many of the toxicity pathways critical for embryogenesis are recapitulated in stem cells,” says Sidney Hunter, Chief of the EPA’s Systems Biology Branch.
The EST is far from perfect, however. The classic version, with a 70 to 80% accuracy rate, is not amenable to high-throughput studies and not specific enough to predict toxicity in certain tissues, such as bone and neurons. Plus, it’s technically challenging and takes time to learn. Numerous groups have worked to improve the test in recent years, devising protocols for turning stem cells into neurons and bone, skin, and liver cells, and creating a battery of tests to assess developmental toxicity in those lineages.
The Scientist talked to EST veterans about how to get the most out the test and its variations. Here’s what they said.
The Classic Test
Designed to predict toxicity in both embryo and mother, the EST uses mouse embryonic stem cells (the D3 line) and adult mouse cells (3T3 fibroblasts). In addition to cell death, the inhibition of ?beating heart cells—the first tissue derived from differentiating D3 embryonic stem cells—serves as a readout for general toxicity. (See Nat Protocols, 6:961-78, 2011, for the classic protocol and its newer molecular endpoints).
1. Prepare several different concentrations of the test chemical in assay medium, as well as controls. Prepare single-cell suspensions by trypsin treatment from undifferentiated mouse embryonic stem cells (mESCs). Add mESCs to each test solution.
2. Pipette 20 µL droplets from each suspension onto the lid of a petri dish, carefully invert it, and incubate for 3 days. This so-called “hanging drop” method is critical for creating consistently sized aggregates of embryonic stem cells called embryoid bodies (EBs), which in turn helps synchronize differentiation.
3. Collect the droplets, each of which contains an EB, and put them back into suspension in the appropriate test or control medium for 2 days to allow the EBs to differentiate. Then replate them, one EB per well, into multiwell plates, and incubate for 5 more days. Heart cells start to form by day 7. At day 10, count these differentiated cells, and calculate the compound concentration that impedes differentiation by 50% (ID50) relative to controls.
4. In parallel, assess cell death in undifferentiated embryonic stem cells and 3T3 fibroblast cells by calculating the concentration of a chemical that causes death in 50% of cells (IC50) from each set.
5. Plug the ID50 and IC50 into an established prediction model that categorizes the test compound as nonembryotoxic, moderately embryotoxic, or strongly embryotoxic in vivo.
Sidney Hunter, Chief of the EPA’s Systems Biology Branch, Integrated Systems Toxicology Division, Research Triangle Park, North Carolina
Examining the effect of 309 environmental chemicals on stem cell viability and differentiation in the EPA’s ToxCast™ project, which employs hundreds of assays to screen thousands of chemicals for toxicity
Culturing embryonic stem cells for the EST is laborious. Researchers must use the tricky hanging-drop method to make embryoid bodies, then transfer them to culture the cells for several more days until beating heart cells emerge. “That’s a whole lot of handling, especially if you are running 10 concentrations of a chemical,” Hunter says.
Using a different stem cell culture technique, Hunter’s group created a new spin on the EST—the “adherent cell differentiation and cytotoxicity assay,” or ACDC. Their method skips the hanging drops, instead seeding stem cells straight into multiwell plates, which are gelatin coated to help the cells stick (Methods Mol Biol, 889:181-95, 2012). This eliminates the need for repetitive replating, Hunter says. Although differentiation still takes 9 days or so, as in conventional EST, in the ACDC method both differentiation and cell death can be measured in the same well using a commercially available assay called the In-Cell Western. Depending on the cost of the antibody label, Hunter’s team can run a 96-well plate for around $100–200 for supplies only. “The EST is cheaper in that you don’t need an antibody to do the assay,” says Hunter.
Most ESTs use the D3 mouse ESC line, but Hunter found that a different line, called J1, more readily maintains the normal number of chromosomes in culture compared with the D3. Using a different cell line is fine, but it’s best to test up front for any chromosomal abnormalities that might affect how the cells respond to toxic substances.
Even with cell culture experience, it takes time for a newcomer to become familiar with stem cells and differentiation. In Hunter’s group, postdoc Kelly Chandler spent 2 or 3 months observing the cells to learn what they look like in differentiated and undifferentiated states and under varying conditions.
Aldert Piersma, Institute for Risk Assessment Sciences, Utrecht University, The Netherlands
Studying the toxicity of compounds on neuronal differentiation
The classic EST requires a lot of time—and a trained eye—to scan multiwell plates for beating heart muscle cells. And the readout gives only limited clues about whether neuronal development will be affected by a compound, Piersma says. His team wanted a faster and more objective way to forecast toxicity, specifically in neural cells.
Piersma’s group steered stem cells to differentiate into neural lineages, then identified changes in the expression of 29 genes which, taken together, predict a chemical’s toxicity. This signature was 84% accurate for 10 test compounds at varying concentrations (Toxicology, 300:158-67, 2012). Although it takes longer to derive brain cells than heart cells—14 days instead of 10—gene expression changes occur by day 4, so the team could conduct its analysis faster than with the standard EST test, Piersma says.
Piersma and other groups have applied this technique to other types of differentiated cells as well. He estimates that gene-expression analysis costs approximately $400/compound, though this estimate can vary widely with the number of test compounds. Going to the EST’s more basic measures (such as counting differentiated cells) to save money isn’t always better, though: gene-expression changes provide molecular details that hint at mechanisms of action. That can help you decide whether your finding is relevant to humans, Piersma says.
Nicole ?zur Nieden, assistant professor of cell biology and neuroscience, University of California, Riverside
Investigating the effects of compounds on differentiation of bone-forming cells called osteoblasts
Because some drugs affect skeletal formation, industry scientists needed a test system that could predict toxicity in osteoblasts.
Zur Nieden and her colleagues developed protocols for differentiating mouse ESCs into osteoblasts and assessing the extent to which test chemicals dampen differentiation. Her group previously identified expression of the bone-specific mRNA marker osteocalcin, analyzed by quantitative real-time PCR, as a differentiation endpoint that could be used to measure toxicity. They have now identified two assays—one absorbance- and the other imaging analysis-based—for quantifying calcium deposition in the bone matrix that are comparably reliable but faster and less costly than the PCR test. Results from both mRNA and calcium assays accurately classified two known teratogens (drugs known to cause birth defects) and one non-teratogen (Toxicol Appl Pharmacol, 247:91-97, 2010).
Given that both endpoints seem to work, you may want to go with the calcium absorption-based assay because it’s cheaper than PCR and easiest, zur Nieden says. According to zur Nieden’s calculations, supply costs for gene expression assays run about $7,500 per chemical, whereas the calcium tests both cost about $1,000. This includes performing an assay in triplicate, with five concentrations of the chemical and one nontreated solvent control. The absorbance-based calcium test takes 2 hours, the calcium imaging assay takes 4 to 5, and the qPCR takes 10 hours of labor.
“We find that the differentiation protocol is relatively robust,” zur Nieden says, who has performed the technique in Canada, Germany, and the United States. “It’s variability of the reagents, more so than the differentiation, that can cause problems,” she says. She recommends starting with the suppliers her group used in the published protocol (Methods Mol Biol, 889:147-79, 2012).
Take care not to mistake especially cell-dense regions in the plate for calcification, the hallmark of osteogenesis, zur Nieden says. Both appear as dark spots, but calcification is darker—almost black.