Deep in a lab at the Wyss Institute for Biologically Inspired Engineering at Harvard University, Dr. Donald Ingber has reconstructed a human lung. It absorbs oxygen like a normal human lung. It also transmits that oxygen to blood cells flowing beneath. White blood cells flock to foreign bodies that try to infect its tissue, surrounding the invaders and stamping them out. In many ways, it’s indistinguishable from the lungs that rise and fall inside you and me, with one exception. This lung is on a microchip.
On these microchips smaller than your thumb, Ingber, director of the Wyss Institute, has reconstructed the complicated interface between lungs and their capillaries. The core of the device is a tiny tube created by microfabrication—a technique used to make structures on the micrometer scale—which is divided in two by a flexible, porous membrane. Human lung cells line the top of the membrane, and capillary cells coat the underside. Air flows through the upper chamber, and a liquid containing human blood cells runs through the lower chamber. Graduate students apply suction to compartments on the sides, mechanically stretching the membrane and its tissue to simulate the rise and fall of our own chests.
Ingber’s lung-on-a-chip isn’t just a breakthrough because it mimics a human organ, but because it does so in more ways than one. The lung cells that line the upper chamber stand in for your lung’s alveoli, the microscopic air sacs where gasses pass in and out of the blood stream. As grad students stretch the chamber, it fills with air, passing oxygen through the capillary cells on the other side of the membrane to the blood cells streaming through the lower chamber.
Just as with human lungs, these cells are susceptible to infections. When Ingber’s team added bacteria to the airspace of the lung-on-a-chip, white blood cells swarmed to the bacteria. Again, just as they would in a real lung. As Ingber’s team pumped the airspace full of various foreign bodies, they discovered something more—that breathing increases the absorption of airborne particulates, like those found in pollution and smog, ten-fold.
They have also tested the toxicity of a cancer drug known to fill patient’s lungs with fluid, a condition known as pulmonary edema. When they gave the lung-on-a-chip the same relative dose given to humans over the same timeframe, the drug caused fluid to shift from the blood vessel into the airspace, mimicking a pulmonary edema. The degree to which Ingber’s lung-on-a-chip can emulate a real human organ is uncanny. It’s all possible because the lung-on-a-chip wasn’t designed with just one purpose in mind. It is, Ingber says, “a toxicity model, a drug efficacy model, and a human disease model.”
But perhaps their greatest feat will be the replacement of animal models in research studies. Animals such as chimpanzees, mice, and guinea pigs, have been used in medical research for centuries, and they have taught us much about anatomy and physiology. But in recent decades, we’ve realized the limitations of animal models. In some cases, animal organs and systems serve as passable stand-ins for their human equivalents, but in many cases they do not. Ingber’s lung-on-a-chip is one of many new attempts at replacing animal models with more effective analogs.
The Animal Problem
The pharmaceutical industry has a problem: about 90% of drug trials fail. Treatments often seem promising when tested in animals, encouraging pharmaceutical companies to start clinical trials that test safety and efficacy in human patients. But the majority of the time, the investment doesn’t pay off. Some are pointing their fingers at animal models, saying they don’t accurately represent the human disorder they are designed to mimic. Moreover, their response to treatments doesn’t predict a human’s response and experiments using them often paint an overly optimistic picture.
Ray Greek, president of Americans for Medical Advancement, a group that seeks to restrict the use of animals in medical research, is among those questioning the efficacy of animal models. Often research animals are bred or genetically modified to develop characteristics of a human disease such as diabetes or Alzheimer’s disease. While most drugs used to treat those afflictions don’t act on genes directly, genes do make proteins that drugs bind to, so tiny differences between species could be the difference between a treatment’s success and failure, Greek says. “Really tiny differences can make a gene lethal to you but perfectly fine for a monkey or a chimp or a mouse,” he says. “That’s Evolution 101. Different genes do different things in different species.”
For many diseases, animal models can be challenging to develop. For others, like neurological and neurodegenerative diseases, it’s even more difficult. “The time that a mouse lives—which is about two or three years—is about seventy years less than it requires a human to develop some of these disorders,” says Dr. Christopher Austin, director of the National Center for Advancing Translational Sciences. Austin says researchers try to manipulate mutations to make their effects even more severe, but “because you have to speed it up to make it happen much faster, you think maybe it’s not terribly surprising it would not be predictive.”
Besides genetic models, there are a number of different ways researchers simulate diseases in animals, including by introducing an infection or administering a drug. These, too, can be problematic. Take traumatic brain injury, for example. In a controlled experiment, the procedure is standardized such that all mice receive the same injury. On the football field, however, no two injuries may be the same. A neurologically complex disease like schizophrenia is hard to model in an animal in the first place—you can’t ask a mouse questions, after all—so testing if a drug reduces symptoms is even more difficult.
Not only are induced diseases and conditions often different in animals, but how we treat them doesn’t always translate well to humans. Livers of different species metabolize drugs differently, for example, so toxicity may vary. And in the laboratory, a stroke can be induced and then treated immediately; in real world situations with humans, the time between stroke and treatment is often longer and much more variable.
Austin notes that scientists appreciate these problems. “For all kinds of reasons, scientists would love to have some other way to study most phenomena than using animals,” Austin says. “I don’t know a single scientist who is not trying to reduce, refine, or replace animal use whenever possible. It’s just that for many indications, for many applications, for many diseases—particularly in the neurosciences—there is no alternative. That’s the problem.”
To sidestep the problems with animal models, some scientists use cellular models, which are often derived from human tissues. One way to obtain human cells is to use induced pluripotent stem cells, also known as iPS cells, made from adult tissues such as skin or blood which are reprogrammed into stem cells that can then become any type of cell in the body. This method is useful for studying how a particular drug binds to a receptor, for example, and how the cell responds. But it’s not perfect. “What happens to an isolated cell—even a human cell—growing on a dish, on plastic, is likely to be very different from how that cell is going to react in a tissue surrounded by other cells,” Austin cautions.
This is where Ingber’s organs-on-chips are useful. The chips more closely represent real, live tissue, with various cell types, arranged as in a human, along with their three-dimensional interactions. Importantly, cells grown in a single layer on a dish can’t mimic motion, like the simulated breathing of the lung-on-a-chip. Motion is important in other organs, too. Ingber’s group has modeled an intestine, predictably called the gut-on-a-chip. In addition to hosting the various cell types that make up an intestinal wall, the gut-on-a-chip also pulses in waves, just like our gastrointestinal tracts do to move food along.
Not only does that motion create a more realistic simulation, it also helps the cells thrive. That’s because Ingber’s chips don’t just host human cells. The gut-on-a-chip also contains some of the same microbes found in a living human’s gut. When cells are grown in culture, scientists often try to keep them free of microbes. Any whiff of contamination can kill the other cells. However, in the gut-on-a-chip, fluid flowing through the chamber, along with the peristalsis-like motion, helps gut microbes grow in a healthy, symbiotic way. That allows Ingber’s group to study how microbes contribute to health and disease.
Currently, Ingber’s group is working on replicating nearly every kind of organ and tissue, including kidneys, livers, and bone marrow. Eventually, he even hopes to connect different organs to make what he calls a human-body-on-a-chip. “Imagine delivering a drug by aerosol to the lung chip, watch it flow over to the liver and see if it’s metabolized into different breakdown products—which is what happens in animals and humans—and then see if that’s peed out by the kidney.” You could test which dose would be optimal to slow the heart, for example, or check if a chemotherapy drug kills bone marrow cells. Ingber notes that is “what we do in animals, effectively, but it would be all human.”
Organs-on-chips also could be used to study how genetic variability between people changes how drugs and treatments work. Greek says future treatments need to be based on a person’s individual genome, a concept known as pharmacogenetics. “Right now, a drug that cures you of a particular kind of cancer may not cure me,” Greek says. “And a drug that I can take for hypertension may work well for me…but you may have a gene that causes a very severe side effect and you won’t know until you take the drug.” Human clinical trials do test safety and efficacy, but Greek is concerned that they don’t capture enough of the variability among people. Because of this, he warns, “everybody who takes a drug is a guinea pig.”
Austin speculates that organs-on-chips could be used with pharmacogenetics to address Greek’s concern. “It’s interesting to think about the potential intersection between organoid tissue chip technology and iPS technology and personalized medicine technology,” he says. Austin imagines taking iPS cells from 100 people and putting them on chips, then assessing the variation in responses to drugs. It would be faster, cheaper, and safer than testing in animals or humans. “You could do it very quickly,” he says. “You could do it in an afternoon.”
Or Do We Refine?
Jonathan Kimmelman, a bioethicist at McGill University, doesn’t think we need to do away with animal models entirely, but thinks we can improve them. He believes part of the reason why treatments on animals models don’t always translate well to humans is the way preclinical research—including testing on animal models—is designed. Once a treatment makes it to clinical trials, which test safety and efficacy in human subjects, there are many rules and regulations in place to prevent bias. One is blinding, where researchers don’t know which treatment the subject received, and another is randomization, where treatment groups are randomly assigned. However, these procedures are only used sporadically in preclinical research, which often involves animals. A recent review of 300 animal studies found that only 14% used blinding and just 13% used randomization.
Another issue is publication bias, where studies with positive results touting the success of a new drug are far more likely to be published than studies with negative results. “That means that when results do get published, you only see a slice or a narrow band of the most positive and encouraging findings,” Kimmelman says. “What’s missing is the perhaps the bottom part of the iceberg, which are inconclusive or negative animal findings.”
As decisions are made about clinical trials, this vital information is often missing. For example, in 1980, researchers suspected that lorcainide, an anti-arrhythmic drug, might be a promising treatment following a heart attack. They hypothesized that, since lorcainide decreased abnormal heart rhythms in non-heart attack patients, it would be helpful in for those who suffered from a heart attack, a population more likely to have abnormal heart rhythms. They tested it on a small group of 100 volunteers. Of the 50 people who were treated with lorcainide, ten died. Only one person from the control group given the placebo died. Lorcainide was abandoned as a post-heart attack treatment, but the trial was never published. So when other companies investigated anti-arrhythmic drugs for heart attack patients, they were unaware of lorcainide’s failure. In later trials, when the drugs were again given to people following heart attacks, even more people died.
In the years since, safety regulations for clinical trials have increased dramatically. One current requirement is prospective registration, which requires that details of the study—such as the drug to be tested, the expected enrollment, and the definition of success—are recorded in a publicly accessible database before the study starts. Once the study is completed, researchers can search the database and learn the results of the experiment—even if it was not published in a peer-reviewed journal. Kimmelman would like to see the same registration process used in hypothesis-driven animal studies, such as testing the efficacy of a drug, for example.
“There are no registries out there for prospectively stating the design of an experiment,” Kimmelman says. “Prospective registries are crucial for preventing publication bias.”
The Middle Road
Ingber also acknowledges that although animal models have their flaws, we cannot get rid of them entirely. His organs-on-a-chip and proposed human-on-a-chip offer promising ways to reduce dependence on animal models, but they are not the same thing as a whole, living organism. “There are certain things in animals you are not going to replace on chips, like behavior for example.”
“Sometimes people don’t realize,” Austin says, that animal models “are absolutely essential for what’s going on in medical research now. If animals stop being used, progress in medical research would slow dramatically and probably screech to a halt in some cases. Many of the advances that we take for granted now have happened through the use of animal models.”
Researchers are increasingly moving away from using animals to model an entire disease, but rather to replicate one aspect that could be treated with a particular drug or intervention. For example, mice bred to be obese and hyperglycemic can be used to study type 2 diabetes. The mice aren’t perfect models of type 2 diabetes—they do not have beta cell dysfunction, for instance, which causes insufficient insulin levels in the human disease—but researchers use them to test drugs that improve insulin resistance specifically. “There are many examples of where those particular aspects of a disease or human physiology may actually be reproduced quite faithfully in a particular animal model,” Austin says.
Computational advances may also help extend the insight gained from animal models. The European Union’s Human Brain Project, which aims to build a brain from supercomputers, is just one example. It’s an ambitious project—one that won’t be completed for at least a decade. But if it meets its goals, it will hopefully give us a better appreciation for the neurological differences between animals and humans. Eventually, that may lead to better treatments.
“Animal models are not going away,” Ingber says. “But we hope, over time, one animal at a time, maybe we can replace—maybe we can show this chip can replace what people were measuring for this particular model. Maybe we can use less animals,” he adds. “Everyone knows it’s a problem that needs to be solved.”