by Sam Zimmerman 
figures by Hannah Zucker

If we were all mice, Alzheimer’s disease, cancer, diabetes, and most inherited disorders would be a thing of the past. We could nibble on as much cheese as we wanted without fear of heart disease and run around our favorite wheel for hours on end without knee pain because all these ailments have been cured in mice. Unfortunately, we are not mice, and most of these cures fail miserably in humans. Out of the hundreds of Alzheimer’s disease treatments that helped mice, none have been beneficial to people. But why do Alzheimer’s treatments that show so much promise in animals never work on human beings? Even though mice and humans look slightly different from each other, they share 92% of their DNA. Additionally, humans and mice have identical genes. One would think a drug that targets and activates a gene in mice would also activate the gene in humans, but this is not always true. 

The same gene in mice and humans are used in different ways

A recent study published in Nature by Hodge et al. showed that even though humans and mice share the same genes, they function differently in the cells of different animals. In order for a cell to perform its job properly, it needs to make specific proteins unique to that cell. These proteins are the worker bees, the ones that actually do the job of a cell, and the directions used to make these proteins are encoded in sections of DNA called genes. A gene is like a very important sentence of a book that a cell can read every time it wants to make the protein encoded by the gene. The more the cell reads the gene, the more the gene is expressed, and the more protein that is made. All of the 20,000 genes in the human genome can be read an infinite number of times, allowing a cell to make many different types of proteins particular to its needs.  

Figure 1. Two cells with the same set of genes can express different proteins by regulating how DNA is “read” into protein, as shown in panel A. In the example shown in B, the neuron on the left panel expresses the gene for a serotonin receptor, whereas the neuron in the right panel has the same gene but does not express the serotonin receptor. Consequently, a drug designed to act on the serotonin receptor will affect the left neuron on the left but not the one on the right.

To compare genes in humans and mice, Hodge et al. used a technique called DNA sequencing to identify the DNA that makes up all the genes in human and mouse brain cells. The researchers learned that almost every type of cell in the mouse brain, including almost all neurons, is also present in the human brain. However, once researchers compared the expression of individual genes within the same cell type, they found vast differences between mice and humans. Two-thirds of all genes shared between mice and humans are expressed differently in the same cell type. The most striking difference was found in neurons, where several genes used to make serotonin receptors are turned on in mice but off in humans. Serotonin, a chemical that regulates mood, sends messages between neurons by binding to its receptor on the receiving neuron’s surface. Without the receptor, serotonin cannot transmit signals to nearby cells. If drugs are made to target receptors that are only present in mice, they will not work in humans. As a result, treatments for depression, schizophrenia, anxiety, Alzheimer’s, and other disorders may be very helpful in mice but would fail to treat the diseases in humans.

Even though genes in mice and human brains are quite different, some parts of the body in humans and mice are more alike. When researchers from the Broad Institute of Harvard and MIT compared the expression of genes in human and mice immune systems, only 169 genes were turned on in one species but off in the other, a much smaller number than the 9,000 genes on in one species but off in the other  in the brain. However, a change to a single gene can be the difference between success and failure in clinical trials. This illustrates that, while mice may model the immune system of humans better than the brain, studies that show drug efficacy in immune disorders should still be treated with caution. 

When differences between human and mice prove fatal

Sometimes these seemingly small differences between humans and mice can have disastrous consequences. Just like a dog cannot eat chocolate because their liver cannot break down the caffeine, humans have also died because their bodies cannot absorb or process the drugs originally tested in mice. In 1993 the drug fialuridine (FIAU) was developed to treat people with hepatitis B and worked amazingly well in mice, rats, dogs, and primates, but once human trials were underway seven people developed liver failure and five died. FIAU was toxic in humans because of a specific protein located on our mitochondria, the structures that generate energy in our cells. This protein transports the drug from empty space in the cell into the mitochondria. Once the drug is let in, it poisons the mitochondria. This turns off the energy supply to our livers where the drug is absorbed. Even though this protein is also present in mice, it does not send the drug into mitochondria because of only 3 differences in the DNA of mice. These 3 DNA mutations change the gene encoding the protein just enough to keep it away from the cell’s mitochondria so the protein cannot transfer the drug into it.

Figure 2. Fialuridine (FIAU) is toxic to humans but not mice because of a difference in protein localization. In both mice and humans, FIAU enters the cell through a transporter protein in the cell membrane and acts against Hepatitis B Virus (HBV). In humans, the transporter protein is also located on the mitochondrial membrane, so FIAU enters the mitochondria and poisons this important energy-generating part of the cell. The mouse transporter is not located on the mitochondria, so FIAU cannot enter the mitochondria and exert its toxic effect.

Research methods may also cause drug trials to fail

Clearly, mice are not the perfect model of every human disease, but scientists still use them to study most ailments out of convenience. To fully replicate complex diseases in mice, researchers would have to know exactly which genes are mutated in humans and make these same mutations to a mouse’s genome. Unfortunately, this knowledge is not usually available. Instead, researchers mutate a small number of genes to replicate the symptoms of a disease. In reality, these mice rarely have all the symptoms of a disease, such as the shaking seen in Huntington’s disease. Additionally, complex diseases such as Alzheimer’s involve many different changes in human genes, and it is unlikely that mice with only some of the changes will have the same underlying illness as humans. So even if a mouse is cured of its illness, that is no guarantee it will cure the human disease the researchers attempt to copy in the mice. 

Researchers are devoted to making every experiment reproducible so other people can check their work. Normally this is a good thing, but to do this in experiments with mice, researchers use mice that, through breeding, have exactly the same DNA. That way, other researchers can use the same mice and get the same result. Curing mice with identical DNA is like devising a cure for only one human out of the 7.5 billion people on Earth. Every person responds differently to medicine and illnesses; what works in one person may not work in another person with different DNA. Some researchers have encouraged scientists to accept that variation in experiments is inevitable. Instead of relying on a single type of mouse with specific DNA, researchers should design experiments that account for this natural variation.

Ironically, even when researchers use the same strains of mice, poor study design and differing methodologies can prevent experiments from being reproducible. In a systematic review of the stroke treatment nimodipine on mice, researchers found the methodologies used in the experiments were poor and there was no consensus on the nimodipine’s effect on stroke. Only 50% of the studies found nimodipine helped relieve symptoms, while the others found no benefit from the drug.

Alternatives to the mouse model

In many ways, mice are ideal for scientific experiments. They are genetically similar to humans, produce many offspring, and mature quickly so that experiments can be done in a short time period. They have helped to develop treatments for numerous diseases, from diabetes to cancer. In 2011, sickle cell disease, a disorder that prevents red blood cells from delivering oxygen to the rest of the body, was cured in mice by decreasing the expression of one gene, allowing for new, healthy red blood cells to form. Just last year, the same strategy was used to create a novel gene therapy that has successfully treated humans as well.

However, in complex neurological disorders where many genes are disrupted, mice are unlikely to model the disease correctly, making a cure for humans less likely. In these cases, some researchers suggest studying human cells in isolation and abandoning mice models. Another alternative to performing research on mice is to use animals such as primates that are more similar to humans. However, as seen in the FIAU example, treatments that work in primates are not guaranteed to work in humans. Ultimately, there is no one-size-fits-all solution, and the best answer may not lie in the type of animal we use at all. Instead, improvements in scientific methods and systematic reviews of animal research that inform the design of clinical trials may be the best way to improve a drug’s success rate. 


Sam Zimmerman is a Ph.D. student in the Biological and Biomedical Sciences program at Harvard University.

Hannah Zucker is a Ph.D. student in the Program in Neuroscience at Harvard University. 

Cover image: “Apodemus sylvaticus bosmuis” by Rasbak is licensed under CC BY-SA 3.0

For More Information:

  • For an overview of how proteins are made from DNA check out this video
  • For more examples of drugs tested in animals that failed in clinical trials take a look at this article
  • For the latest on sickle cell anemia clinical trials see this great piece in the New York Times
  • Learn about amazing new alternatives to the standard mouse model in articles from the Harvard Gazette and Science.
  • For informative commentaries on how studies in mice can cause misleading scientific conclusions look up the twitter hashtag #JustInMice and this story on the twitter handle justsaysinmice.

9 thoughts on “Why Drugs Tested in Mice Fail in Human Clinical Trials

  1. Also new research showing how rodents already have long telemeres and after a generation or two in captivity they lengthen much more. Making cellular damage like that of toxins not apparent because with so long of telemere and high metabolism they heal the damage like wolverine…. the other side of this coin is elongated telemeres allow for greater chances of rogue cells aka tumors/cancer. So a drug that shows cancer potential in rats may not at all cause cancer in humans… such as GW501516/cardarine trials showing tumors everywhere… at 1000s of times the dose per kg effective for both rodents and humans. So basically rats over produce cancer and under produce cellular damage from toxic exposure.. another another reasons not to trust the FDA.

  2. I found this article very interesting, it has a lot of great info i need to know, thanks for your explanation and your effort

  3. I did not read the entire article but it just makes me laugh at how scientists like to make things so complicated. The answer is very simple but admitting it would prove that they not only don’t know it all but they willfully ignore the facts that are right in front of their eyes. The truth is that mice produce their own vitamin c and humans can not. Vitamins c is essential to the functioning of our immune system and we continue to experiment on the why? Convenience? Really? Lives are at stake and they use mice out of convenience? Considering that vitamin c is essential for the proper functioning of our immune systems, perhaps we shouldn’t use mice . It’s very simple!

  4. Research in mice is actually the best way to study new treatments before going in human trials. However, the chance of translation to humans depends also on the accuracy of the experimental design in mice. I think that research in mice has high chance to be translated in humans when:
    1) No pathology is artificially or genetically induced in the mice. We have a lot of pathologies that are similar to humans that develop during aging of mice. It is more challenging to work on these pathologies (also costs are higher) but there is less risk that the model is wrong. We also need to use transgenic animals and other models of induction (especially in the case of human pathologies that are not present in mice like AD) but researchers should be aware of the reduced chance to translate results in humans (it is always better than testing directly in humans).
    2) Studies in mice are performed blinded as in humans. Researchers absolutely must not know who is the treatment and who is the control group. It is well known that researchers can be influenced by the results that they expect to find. This is always considered in clinical trials but rarely in animal studies (and when this is done there is no external monitor to check if this is true).
    3) Sample size in mice is designed carefully. It should be taken into account the real variability of the data (that also in genetically equal mice can be as high as in humans). Conversely, I see that most sample sizes are designed considering the very small error bars shown in “convenient” papers that use few animals.

  5. Hi
    As I run a clinical trial institute in Pune I agree with you that the mice and humans have identical genes. Thanks for sharing

  6. Great article, but I really disagree with some of the points. Drug tests on mice have been used for ages and are definitely a good first step towards actual and more robust testing.

    In many cases, results from mice/small rodents are insightful. In addition these tests are quick to perform and require minimal legal compliance requirements. Indeed in isolation, these tests alone cannot pave the way for production.

    Just my 2 cents!
    Justin

  7. An eye-opening read. As a newbie, i learnt how you cannot use the same models for animals and humans as the results would be disastrous . Thanks for sharing.

Leave a Reply

Your email address will not be published. Required fields are marked *