In the course of delivering the 2013 Cartwright Lecture at Columbia University, molecular biology pioneer Sydney Brenner made a startling prediction. The Nobel Prize winner had done foundational work in genetic regulation of development — and as a side project, laid the groundwork for the development of the C. elegans worm as a model organism for genetic research — so most in the audience expected his lecture to cover some aspect of the evolution of simple nervous systems. Instead, he spent the majority of his time arguing that the age of the model organism was coming to an end. In short, thanks to advances in human genomics and tissue cultivation, the future of biomedical research would not require the necessary evil of animal testing, which, on top of ethical concerns, was a poor predictor of a drug’s efficacy in human subjects. 

This did not go over particularly well with the large group of biologists in the room, most of whom had built their careers conducting research on worms, flies, mice and rats, all standardized organisms designed to provide a controlled model for investigating biological processes. At the time, there was simply no foreseeable alternative to model organisms. Less than a decade later, however, that is starting to change, and Brenner’s prediction is starting to be realized. In the near future, we may be able to conduct all biomedical research directly on lab-grown human cells.   

Mice lie, Monkeys exaggerate

Animals are not always the ideal stand-in when studying potential treatments for diseases that represent the biggest threats to human health. In some cases, model animals may, in fact, present an impediment to crucial research. 

Take the case of Alzheimer’s disease (AD). Drugs designed to treat Alzheimer’s have been remarkably unsuccessful, with a 99% overall rate of failure in clinical trials. This is a huge problem because the US alone is projected to see 12.7 million cases of Alzheimer’s dementia by 2050 if no effective treatment is found. The situation is so desperate that aducanumab, the first AD drug to reach approval in close to two decades, appears to have been approved by the US FDA despite a lack of conclusive evidence that it is effective. 

One might argue that Alzheimer’s disease is just too difficult to treat, but there are other potential explanations for this dismal success rate. These drugs have more than failure in common: they were all tested initially in mouse models. 

Mice lend themselves well to biological research because they reproduce and mature quickly, have all the important complex tissues of a human body, and are amenable to genetic modification and manipulation. This would all be well and good for Alzheimer’s research, except for one big problem: mice don’t get Alzheimer’s. In fact, in all the years of research into the disease, scientists haven’t found any evidence that any species other than humans, cats and dolphins can develop it naturally. 

Alzheimer’s researchers and pharmaceutical developers have found ways to coax mice into developing some forms of the disease, whether through precise genetic alteration, capitalizing on chance mutations, or even “humanizing” the mice themselves — forcing them to produce human forms of the proteins involved in AD in order to create a sort of hybrid animal model of neural degeneration. And yet, despite enormous leaps in understanding of disease progression and etiology, the results on the pharmaceutical side speak for themselves: after over 400 clinical trials, only a handful of drugs have been approved, and none of them have been particularly effective at treating Alzheimer’s disease.

This problem is greater than Alzheimer’s, or neurodegenerative diseases in general. There’s a saying in the virology community, “mice lie and monkeys exaggerate,” a weary way of summarizing a long history in the drug development world of seeing drugs that performed beautifully in preclinical animal studies go on to flame out in human trials. All this has contributed to a growing momentum in the shift away from model organisms in early pharmaceutical development. 

A Human Alternative

The increasing frustration with animal models would be effectively meaningless if it weren’t for the fact that reliable methods of testing early drugs in human tissues are now becoming possible. Over the past decade, advances in the understanding of how specific cell types are formed, and how to reprogram a cell so that it changes from one type to another, have allowed researchers to produce samples of highly differentiated human tissue types for early drug research — none of which involve mice, monkeys, hamsters, or any of the other members of the eclectic menagerie of biomedical research animals. 

Now, a handful of companies have dedicated themselves full-time to the production of pure, tissue-specific cell populations for biological research and pharmaceutical development. However, a few fundamental challenges remain: figuring out the exact genetic code required to make all the different cells that will be needed for a future of drug development based on engineered human tissues, and developing the technology to engineer and execute those codes via DNA to reprogram them into the right cell type. 

One company, bit.bio, believes they are on track to solve these problems. Their machine learning pipeline is steadily working away at identifying the specific set of regulator genes that guide the development of many cell types, ranging from immune cells to central nervous system cells. 

What’s more, they’ve developed a system that can transform undifferentiated cells — pluripotent stem cells, a pool of raw biological potential — into their cell type of choice in as little as five days. Considering that the normal time course for producing differentiated cells can be on the order of weeks or months, with a conversion efficiency that maxes out in the low double digits, this could make things a lot easier for scientists looking to study tissue-specific drug effects or disease processes. 

Drawing on these advances, the scientists at bit.bio now plan to radically transform the way early-stage drugs are tested. They intend to provide a steady supply of pure, reliable, human cells that can be tailored directly to the needs of researchers studying diseases or attempting to test the safety and efficacy of new drugs. 

Removing Technological Barriers

“Biologists have always been at the mercy of technology,” says Dr Tonya Frolov, the product manager responsible for the muscle and central nervous system tissue pipelines at bit.bio. “​​Our mission is to provide reliable, relevant, and reproducible tools so that experts can go and make the discoveries that they would [otherwise] not be capable of making, just because they didn’t have those tools.” 

Frolov has run up against these technological barriers herself. As a graduate student, she spent years studying glioblastoma (GBM), one of the deadliest and most untreatable forms of brain cancer. She found herself questioning the potential of the tools she was using in her research after a number of publications demonstrated the incredible genetic and developmental variation between populations of supposedly standardized GBM cell lines. Genetic drift had set in after years of continuous cultivation in labs scattered across the globe. “There are labs around the world seemingly buying the same cell lines but over the course of 10, 20 years, the cell lines become vastly different,” Frolov says. If a community of dedicated scientists couldn’t even trust their own cell cultures to accurately model the disease, how could they be expected to make any serious headway in developing treatments for GBM?  

This phenomenon is not limited solely to glioblastoma. In some cases, in fact, this type of variability is well-known — and simply accepted as a cost of doing business. 

Consider primary cell cultures, a popular way of studying diseases, and potential treatments thereof, in the lab. Primary cell cultures are made by collecting a sample of diseased tissue from a patient, then growing a culture of the cells in a lab and attempting to squeeze as much useful knowledge out of the resulting cells before they run out, which they very often do. And once they’re gone, they can often be impossible to replace, which can make it extremely difficult to share materials across labs for the purposes of trying to build on — or simply reproduce — the results of other investigators. This is where the team at bit.bio hope to make a big impact with their expertise on cellular reprogramming.

New Models

bit.bio’s goal of elaborating a genetic code for making every individual cell type in the human body remains somewhat far off, though. While they currently have multiple cell types in active development, recent estimates put the number of distinct cell types in the human body in the hundreds. But a thoughtful choice of targets has allowed them to make rapid and significant progress in certain tissue types — especially the central nervous system, where they have already brought considerable genetic/developmental knowledge and biotechnology to bear against some of the hardest targets in neurological disease. For example, by combining the precise genetic alterations enabled by CRISPR/Cas9 with their fast-differentiating cells-on-demand pipeline, they are developing disease-specific models - such as Alzheimer’s or Parkinson’s diseases - for research and high-throughput screening applications.

On some level, of course, even a high quality model is still only a model. You might be able to precisely and reproducibly recreate a diseased cell type at scale, but the specific biology of that disease model will still only be a reflection of the current state of knowledge of the disease itself. In Alzheimer’s disease, for example, one might be able to create a steady supply of genetically uniform human neurons that all reliably produce the diseased beta-amyloid proteins thought to be the cause of Alzheimer’s neurodegeneration, but the field of Alzheimer’s disease research isn’t even entirely settled on whether this is the true (or only) cause of the disease itself. 

Scientifically (and philosophically), this is a serious potential issue, but the world of drug development is in such a state right now that any marginal improvement in the efficiency of rapidly assessing the value of drug candidates could have huge impact. While the outlook for any given experimental drug candidate isn’t quite as bleak as that for drugs targeting Alzheimer’s disease, the numbers are still fairly stark: almost 50% of drug candidates that make it through rigorous clinical testing to phase III, the final hurdle before drug approval, still ultimately fail, and on average it costs close to 3 billion US dollars to develop a successful central nervous system drug. In fact, the failure of drugs to make it from successful preclinical experiments to FDA approval is so common that the process has been likened to passing through the valley of death. 

As a result, more and more pharmaceutical companies are moving toward high-throughput screening models of early drug validation, where thousands of candidate molecules are tested against both healthy and diseased cells in plate after plate of cell cultures. bit.bio’s hope is that, by significantly improving the quality and reliability of the cells in those plates, they can reduce the noise in the system of pharmaceutical research, help drug developers to identify successful — or toxic — compounds much earlier, and bring the overall costs of drug development way down. 

Ultimately, if bit.bio is successful in their quest to uncover the genetic programs for creating any given human cell type, it might spell the end of more than just model organisms as a research tool. With the genetic knowledge and computational power they are accumulating, it’s conceivable that even physical cells could become irrelevant. Sometime in the future, drug screens may be performed entirely in silico, in simulated cells that are so true to life that they can accurately predict whether a drug will be effective against a disease of interest. This would be good for humanity, and even better for lab mice.