“The history of cancer research has been a history of curing cancer in the mouse. We have cured mice of cancer for decades—and it simply didn’t work in humans.”

This statement was made by Richard Klausner, M.D., former director of the National Cancer Institute, to the Los Angeles Times(Cimons 1998). It is a statement that applies equally to many other common diseases. A decade ago, researchers reported on the existence of 195 published methods that prevented or delayed the development of type 2 diabetes in mice (Roep et al 2004). Yet none of these “breakthroughs” ever translated to human medicine.

What prevents successes in mice from becoming human cures and treatments? The reason is largely a simple one, as we showed when we recently analyzed the reputed contributions of mouse experiments to human type 2 diabetes research (Chandrasekera and Pippin 2013).

Type 2 diabetes (diabetes mellitus) is the fastest-growing disease in the United States, currently affecting approximately 26 million Americans, and estimated to quadruple in prevalence to affect one-third of Americans by 2050 (CDC 2011a). Type 2 diabetes is the seventh leading cause of death in the United States (CDC 2011b). It is a complex and multifactorial disease, characterized by many years (often decades) of sequential disruptions in glucose homeostasis. The pre-diabetes stage includes impairment of fasting glucose and glucose tolerance, often evolving into a specific metabolic syndrome that includes abdominal obesity, dyslipidemia, hypertension, and elevated fasting blood glucose. The eventual development of full-blown type 2 diabetes is signaled by overt hyperglycemia resulting from a combination of insulin resistance and dysfunction of insulin-producing pancreatic ß-cells. In other words, type 2 diabetes is a systemic disease occurring at several sites in the body.

In contrast to this complexity, mouse models of human diabetes are biologically simpler. Disease occurs rapidly, and typically manifests only one or a few of the factors and symptoms associated with human diabetes. For example, one category of mouse model impedes or eliminates the production of insulin by the pancreas. This is achieved either by partial or complete surgical removal of the pancreas (Islam 2009), or by chemical destruction of pancreatic ß-cell function with cytotoxic drugs such as streptozotocin or alloxan (Lenzen 1979).

But these models do not replicate human type 2 diabetes, in part because they do not account for the risk factors, long latency, related pathologies, co-morbidities, symptoms, and genetic influences (Dupuis and colleagues 2010; DIAGRAM Consortium 2014) of diabetes as found in humans. The same criticism can be made of other mouse models that employ nutritional interventions, spontaneous genetic mutations, and transgenic manipulations. None replicates the complex human disease either biologically or phenotypically (Chandrasekera and Pippin 2013).

Compounding the problem of replicating human type 2 diabetes in mice without knowing its fundamental causes is the difficulty that mice differ in many respects from humans, including in functions related to glucose metabolism. In our paper, we describe in detail decades of type 2 diabetes research primarily using mice, and the many discrepancies compared to human-based research findings (Chandrasekera and Pippin 2013). For example, Vassilopoulos and colleagues 2009 reported that mice do not possess the protein that in humans mediates transport of circulating glucose into cells by means of membrane-bound vesicles. And Cabrera and colleagues 2006 described prominent pancreatic islet anatomic and architectural differences between humans and mice. These differences not only have “functional implications for islet cell function,” but also potentially invalidate decades of mouse experiments to investigate human diabetes.

The importance of interspecies differences in studying human type 2 diabetes in mice is exemplified by examining the rate-limiting step in human glucose metabolism, which is insulin-dependent glucose uptake into skeletal muscle. In humans, this step is facilitated by glucose transporter 4 (GLUT 4) protein and requires the action of the CHC22 protein. The CHC22 protein is highly expressed in human skeletal muscle, but the encoding gene exists only as a nonfunctioning pseudogene in mice (Wakeham 2005; Vassilopoulos 2009). Thus, attempts to apply mouse glucose transport studies to humans are impeded due to species differences in the GLUT 4 trafficking pathway.

Similar immutable barriers to the use of mice to study human type 2 diabetes exist at every level from gene structure and gene regulation to disease manifestations and phenotypes. This explains why neither any single mouse model, nor any combination of mouse models, can replicate human type 2 diabetes or reliably inform its prevention or cure. Thus species-specific differences and contradictory findings have been identified for every biological level of investigation—nucleic acid, protein, glucose pathway, cellular, tissue, organ, organism, and population levels.

All this is detailed in our recent review paper (Chandrasekera and Pippin 2013), in which we also note the paucity of effective human treatments and the absence of cures. We conclude that the strikingly consistent failure to translate mouse research to human type 2 diabetes prevention and treatment cannot be remedied; but can best be addressed by “humanizing” type 2 diabetes research—that is, studying the disease using a combination of human cell cultures and tissues, in vitro and stem cell methods, laboratory and clinical population studies, and other approaches directly relevant for diabetes patients.

Mouse research into other diseases is equally unsuccessful

In a head-turning article last year, researchers from more than a dozen medical institutions investigated the accuracy of mouse experiments to replicate the changes that occur in three severe inflammatory human disorders: trauma, sepsis, and burns (Seok and colleagues 2013).

They found that patterns of gene expression in humans and mice experiencing these disorders did not correlate. The authors, referred to the murine responses as “close to random in matching their human counterparts.” In some instances the genetic responses were opposite for humans and mice, implying that treatments that benefit mice could harm humans. In interpreting their results, the authors noted that about 150 treatments for these inflammatory disorders had been tested in humans after succeeding in mouse experiments. Every one of those treatments has failed. Their research offered a simple but fundamental explanation for those failures, localizing to differential gene expression in mice and humans.

Seok and colleagues wrote that “our study supports higher priority for translational medical research to focus on the more complex human conditions rather than relying on mouse models to study human inflammatory diseases.” But they also concluded that the complete genetic dissociation in their study impugns the appropriateness of studying mice in other areas of medical research and for drug development. And in fact, mice have been discredited in the eyes of many for the study of numerous other human diseases.

For example, applying treatments that were successful in mice, all of ten randomized prospective clinical trials involving more than 2,700 patients, testing spinal cord injury treatments, have failed, as have many other spinal cord injury clinical trials (Tator 2006). In 2012, Tator updated his review and again reported that “no agents that produce major benefit have been proven to date.” (Tator 2012). A detailed review of animal models for spinal cord injury (Akhtar and colleagues 2008) describes their shortcomings as models and the absence of translation to human spinal cord injury.

Other researchers have acknowledged the failure of mouse research for traumatic brain injury to translate into human benefits. This failure is highlighted in the report by Maas and colleagues 2010 of 22 phase III clinical trials and six unpublished trials that all failed to benefit traumatic brain injury patients, despite positive results in animal studies.

The standard superoxide dismutase 1 (SOD1) mouse models for the study of amyotrophic lateral sclerosis (ALS or Lou Gehrig’s disease) employ transgenic mice with mutations in the gene coding for SOD1, a protein that destroys superoxide radicals in humans. Mutated SOD1 is strongly associated with ALS, but these mouse mutants have been an abject failure, in that no phase III human trial in four decades has shown convincing benefit for patients (Rothstein 2004; Traynor 2006; Benatar 2007). Nor have mouse studies contributed to any useful treatments for related motor neuron diseases, prompting the National Institute of Neurological Disorders and Stroke to acknowledge: “There is no cure or standard therapy for MNDs [motor neuron diseases].” (NINDS 2014)

Mice are also the standard models for Alzheimer disease, in particular transgenic mouse models designed to express mutated proteins associated with brain pathologies in Alzheimer disease. In a new review, Cavanaugh and colleagues 2014 reported in detail the failure of two dozen transgenic mouse and rat models and more than three dozen drugs when tested in Alzheimer disease clinical trials. Cavanaugh also proposed improving the reliability and applicability of Alzheimer’s disease research by emphasizing the causative role of lifestyle factors and the potential for humanizing research.

For decades, human xenograft mouse models have formed the foundation of basic science research into human cancers. Yet the translational success rate for this research is lower than for any other area of animal research into human diseases (Kola 2004; Roberts 2004; Hutchinson 2011). Only a small percentage of mouse research into human cancers is even reproducible (Prinz 2011; Begley 2012), and both the U.S. National Cancer Institute (Shoemaker 2006; NCI 2014) and the National Cancer Institute of Canada (Voskoglou-Nomikos 2003) have demonstrated the superiority of human cancer cell lines compared to mouse models for the development of beneficial treatments.

The failing research paradigm

More than forty years after the launch of the War on Cancer by President Richard Nixon, that war is still being lost in laboratories pursuing mouse experiments. Choose almost any area of medical research using mice, and you will see a failed paradigm often spanning several decades. The reasons why such discredited research continues are complex and often unrelated to scientific merit, but medical researchers would be wise to remember that the first thing to do when you’re in a deep hole is to stop digging.

Dr. Pippin is a cardiologist, former animal researcher, and director of academic affairs for the nonprofit Physicians Committee for Responsible Medicine.

References

Akhtar AZ, et al. (2008). Animal models in spinal cord injury: a review. Rev Neurosci 19:47-60.
Begley CG, Ellis LM (2012). Raise standards for preclinical cancer research. Nature 483:531-533.
Benatar M (2007). Lost in translation: treatment trials in the SOD1 mouse and in human ALS. Neurobiol Dis 26:1-13.
Cabrera O, et al. (2006). The unique cytoarchitecture of human pancreatic islets has implications for islet cell function. Proc Natl Acad Sci 103:2334-2339.
Cavanaugh SE, et al. (2014). Animal models of Alzheimer disease: historical pitfalls and a path forward. ALTEX online first (April 10, 2014).
CDC (2011a). National diabetes fact sheet. Accessed May 12, 2014 athttp://www.cdc.gov/DIABETES/pubs/factsheet11.htm.
CDC (2011b). Leading causes of death. Accessed May 12, 2014 athttp://www.cdc.gov/nchs/data/nvsr/nvsr61/nvsr61_06.pdf.
Chandrasekera PC, Pippin JJ (2013). Of rodents and men: species-specific glucose regulation and type 2 diabetes research. ALTEX 31:157-176.
Cimons M, et al. (1998) “Cancer drugs face long road from mice to men.” Los Angeles Times, May 6, 1998, page A1.
DIAGRAM Consortium, et al. Genome-wide trans-ancestry meta-analysis provides insight into the genetic architecture of type 2 diabetes susceptibility. Nature Genetics 46:234-244.
Dupuis J, et al. (2010). New genetic loci implicated in fasting glucose homeostasis and their impact on type 2 diabetes risk. Nature Genetics 42:105–116.
Hutchinson L, Kirk R (2011). High drug attrition rates – where are we going wrong? Nat Rev Clin Oncol 8:189-190.
Islam MS, Loots DT (2009) Experimental rodent models of type 2 diabetes: a review. Methods Find Exp Clin Pharmacol 31:249-61.
Kola I, Landis J (2004). Can the pharmaceutical industry reduce attrition rates? Nat Rev Drug Discov 3:711-715.
Lenzen S (1979). Insulin secretion by isolated rat and mouse pancreas. Am J Physiol 236:E391-E400.
Maas AIR, et al. (2010). Clinical trials in traumatic brain injury: past experience and current developments. Neurotherapeutics 7:115-126.
NCI-60 DTP Human Tumor Cell Line Screen (2014).
NINDS website (2014). Motor neuron diseases fact sheet.
Prinz F, et al. (2011). Believe it or not: how much can we rely on published data on potential drug targets? Nat Rev Drug Discov 10, doi:10.1038/nrd3439-c1
Roberts TG, et al. (2004). Trends in the risks and benefits to patients with cancer participating in phase 1 clinical trials. JAMA 292:2130-2140.
Roep BO, et al. (2004). Satisfaction (not) guaranteed: re-evaluating the use of animal models of type 1 diabetes. Nat Rev Immunol 4:989-997.
Rothstein JD (2004).Preclinical studies: how much can we rely on? Amyotroph Lateral Scler 5 (suppl 1): 22-25.
Seok J, et al. (2013). Genomic responses in mouse models poorly mimic human inflammatory diseases. Proc Natl Acad Sci 110:3507-3512.
Shoemaker RH (2006). The NCI60 human tumour cell line anticancer drug screen. Nat Rev Cancer 6:813-823.
Tator CH (2006). Review of treatment trials in human spinal cord injury: issues, difficulties, and recommendations. Neurosurgery 59:957-987.
Tator CH, et al. (2012). Translational potential of preclinical trials of neuroprotection through pharmacotherapy for spinal cord injury. J Neurosurg: Spine (suppl) 17:157-229.
Traynor BJ, et al. (2006). Neuroprotective agents for clinical trials in ALS: a systematic assessment. Neurology 67:20-27.
Vassilopoulos S, et al. (2009). A role for the CHC22 clathrin heavy-chain isoform in human glucose metabolism. Science 324;1192-1196.
Voskoglou-Nomikos T, et al. (2003). Clinical predictive value of the in vitro cell line, human xenograft, and mouse allograft preclinical cancer models. Clin Cancer Res 9:4227-4239.
Wakeham DE, et al. (2005). Clathrin heavy and light chain isoforms originated by independent mechanisms of gene duplication during chordate evolution. Proc Natl Acad Sci 102:7209-7214.

This article originally appeared on Independent Science News.