Off mice and men

September 25, 2013

Personalized disease models will speed up and open up rare disease drug discovery.

Our collective experiment in government-cultivated biomedical research — the National Institutes of Health, or NIH – began in 1938 with a Congressional angel investment of $464,000 in today’s dollars. The mandate of NIH is two-fold: 1) sow the seeds of discovery, and 2) put promising saplings on a path to sustainable growth. Since its inception, the US biomedical research enterprise has expended half a trillion dollars, subsidized professional training of tens of thousands of scientists, and created a 20M+ article library of scholarly knowledge.

 

since 1938

 

You might imagine that after nearly 80 years of managed research we’d be reaping bountiful harvests in terms of new drugs, and even cures. However, from Eroom’s Law to the writing on the wall, all evidence points to a collapsing Academia-Pharma Complex, as I’ve been calling it. This is especially so for the simplest of diseases: the rare genetic diseases, which are set in motion by a single defective gene.

 

To be clear – simplest doesn’t mean simple!

 

Rare disease research is, in fact, complex. First, the function of the disease-triggering gene needs to elucidated. But remember that proteins are usually built of modular domains that have distinct functions. Second, proteins not only have multiple functions but also interact with multiple dance partners, each with its own tempo and moves. Third, naturally occurring mutations run the gamut: they can selectively hobble one function or interaction without disturbing others, or stop a protein cold in its tracks. All of this complexity must be worked out in a basic research setting if we want a bottom-up explanation of disease that will create the conditions for strong drug candidates down the line.

 

But the current tally of ~450 FDA approved drugs for rare diseases is a drop in the bucket. There are 7,000 rare diseases and counting, 80% of which are caused by recessive mutations in both copies of a single gene — 2 off bits in our 3-billion bit genome. I contend that a big part of the problem are missed opportunities at the embryonic stages of drug discovery, where unbiased genetic screens are deployed to elucidate gene function and illuminate compensatory pathways, bypasses or overrides that are entrees for rational therapeutic intervention.

 

These basic research activities have traditionally taken place in academia, and have exploited simple genetic model organisms, like the budding yeast Saccharomyces cerevisiae. We’ve learned so much about how eukaryotic cells work from genetic screens with yeast. For instance, how intracellular cargo is transported to and fro in tiny sacs called vesicles (Novick & Scheckman); how the gears of the clock that regulates cell division spin (Hartwell); how cells ease stress by recycling aged or damaged parts (Ohsumi).

 

We are, after all, eukaryotes.

 

Alas the one-time polyculture of genetic model organisms, spanning single-celled eukaryotes to multicellular eukaryotic behemoths like mice, has fallen out of favor. In its place a pernicious mouse model monoculture has crowded out the competition. If what I’m saying sounds sacrilege naive, you clearly didn’t get The Mouse Trap memo a few years back. Intrepid #longread reporter Dan Engber described the monopolization of human disease models by laboratory inbred mice in a wonderful three-part series for Slate in 2011.

 

If we take evolutionary conservation at face value, why do we stop at mice? Why do we even stop at mammals? Why aren’t we following the rabbit hole all the way down to shared ancestry that runs like a thread connecting primordial gene functions to their modern analogs in complex organisms? Why are model organisms languishing in academic silos, such that we still talk about “yeast labs,” or “fly labs,” and the like?

 

As I’ve been making the case on my blog this summer, we desperately need to accelerate the innovation cycle of rare disease research with an injection of genetics, and by tactically leveraging evolutionary conservation. My way out of this thicket is to break our addiction to rodent models and restore a sustainable polyculture of patient-matched disease models whose screening outputs are directly coupled to patient-derived cells. I’m putting the finishing touches on my business plan as we speak.

 

This concept is neither new nor radical. Take my ongoing case study of Niemann-Pick Type C, or NPC, which is one of the ~50 lysosomal storage disorders. Loftus et al discovered NPC1, the gene that when mutated is responsible for 95% of NPC. NPC1 gene is an ancient gene conserved over a billion years of evolution. But what does gene-level conservation actually mean? Ideally, validating a model organism for a specific rare disease is contingent on the ancient version of the disease gene being functionally interchangeable with the human version in what’s called a complementation test. For example, if you take the yeast version of NPC1 and insert it in a mammalian cell lacking its own NPC1, what happens? If the genetically modified mammalian cell looks as though it has normal NPC1, it’s a go. Now of course that doesn’t make yeast and humans suddenly interchangeable — that would be absurd. But it would be equally absurd not to start the study of NPC1 function in its most ancestral state, and then work forward. Loftus et al argued just as much in the close to their landmark paper:

 

Loftus_finalsent

 

But did the biomedical researchers who followed heed their call? By analyzing the results of a simple PubMed search using the terms [“Niemann Pick type C” + “model organism X”] you can see that the community has failed to heed Loftus’ exhortation. There are 342 papers that looked at NPC in mice or mouse cell lines. By contrast, only 4 original research papers study a yeast or worm or fly model of NPC. Think about that for a second. Almost a 30-to-1 ratio in favor of mice to non-mammalian models!

 

If evolutionary conservation were really a guide, we might expect a more balanced, step-wise validation of disease-associated phenotypes from simple models to complex models, though some of those phenotypes won’t be conserved throughout evolution. None of this is to say that mice haven’t be tremendously important for rare disease research — they have. To a first approximation they are a great proxy for human physiology. But I’m arguing that without systematic validation across multiple model organisms representing snapshots of gene function over time, leads are likely be fraught, misleading, or just plain weak.

 

So what can flies — the middle of the road in terms of evolutionary complexity — teach us about NPC? The Scott Lab at Stanford happened onto NPC because of serendipity. A part of the NPC1 gene resembles another a functionally distinct gene involved in developmental biology. In 2005, the Scott Lab published a paper on what happens when the fly version of NPC1 is defective. Based on a number of diagnostic criteria, including clinical biomarkers of NPC, NPC1-mutant flies appear sick in ways reminiscent of people with the NPC:

 

Huang 2005_figsummary

 

The mouse model monoculture is insidious for more reasons than just crowding out non-mammalian model organisms. For example, the first mouse model of the lysosomal storage disease Tay-Sachs was created in 1994. The authors noted with some consternation that the Tay-Sachs mouse didn’t exhibit signs of neurodegeneration, a hallmark of the disease in people. The authors concluded that while the underlying physiology may be conserved, the fact that mice have much shorter life spans means that a disease that takes several years to manifest in people doesn’t have a enough time to manifest in mice. (It also turns out that mice do actually have a physiological bypass not conserved in us).

 

The root of the imbalanced pipeline is a biomedical bias that remembers all the times when leads generated in non-mammalian model organisms fall flat in more complex models, but forgets all the times when leads start in the humblest of model organisms and steadily work their way up the evolutionary chain. I’ve blogged a bunch before about Susan Lindquist and others who use yeast to model the neurodegeneration caused by misfolded proteins like alpha-synuclein and huntingtin, and here’s the most recent example of an integrated evolutionary approach in action.

 

The rare disease research renaissance that we’re seeing glimpses of won’t happen overnight. For one thing, rare diseases lose out to common diseases. From a well understood but no less tragic economic perspective, Pharma has been loath to invest in rare disease research because the market won’t support it as profitably as research into common diseases. Academics also neglect rare diseases in favor of common diseases because that’s where the grant money is. With the exception of a generation of gene hunters who were part of the special Genomic Great Race of the 1980s and 1990s to identify the first gene causing a rare disease, it seems as though it takes a personal connection to get a rare disease on an academic’s radar screen.

 

I’m about to wager that a panel of personalized disease model organisms comprised of yeast, worms, flies and fish will reinvigorate the earliest stages of drug discovery. At a minimum, this approach will derisk mouse models, hopefully on the way to weaning ourselves off of them entirely. My evolutionary pharmacology vision is meeting a rising tide of rare disease awareness spurred on by patient advocacy groups (PAGs). To borrow the language of tech startups, there are alphas of patient-driven drug discovery and patient-driven clinical trials in progress right now. What’s more, the Internet is also allowing PAGs to map out connections between rare disease patients and their larger social networks, so the 1 in 10 abstraction of rare disease patient populations suddenly becomes very personal. Of course, I would be remiss if I didn’t mention the achievements of the pioneers who parlayed the rare disease awareness campaigns of the Jerry Lewis telethon era into the Orphan Drug Act, which celebrated its 30th birthday this year.

 

A Noah’s Ark of human disease models once prevailed, imperfect as they were. The one-time diversity of human disease models that has ceded to a maniacal mouse monoculture needs to be revived with a personalized touch for the Open Science Age. As organs on a chip technology merges with 3D printing, and non-invasive nanotechnology sensors emerge, the need to sack mice for science will disappear.

 

  • disqus_WTkiP6CcE7

    Hi Ethan – I think that your approach is fascinating and I’m following your work with great interest. But, I would like to know, what’s the best example of drug discovery that’s been done in yeast? Certainly, tons of basic biological processes have been elucidated in cerevisiae (like the sec and the cdc screens that you cite). But, have we ever found that drug X treats mutation Y in yeast, and then moved from that up the evolutionary ladder all the way to the clinic? The Lindquist lab has done tons of protein folding research in yeast, but so far as I know none of it has lead to a new treatment in humans. Has what you’re proposing been done before?

    • Ethan O. Perlstein

      Thanks for the comment, disqus_WTkiP6CcE7. :)

      Short answer: yes, I’m proposing what has been done before. It’s called genetics.

      As for actual drugs, point taken. But Pharma has targetophilia and doesn’t do genetics so of course they’re not using simple genetic model organisms.

      I do know of one case that’s making its way through the clinical trial process now. But you’re right. Yeast to people is a long road. I like to point out that the road currently traveled completely skips yeast and starts with mice — and it hasn’t been working out so great.

      I would argue that the drug discovery “valley of death” is in part academic discoveries dying on the vine, or not being properly transferred to the private sector. Ironic, since the Bayh-Dole Act was expressly designed by legislators to incentivize the flow of discoveries into the private sector. BDA and I are the same age, so we’ve have over three decades of implementation and metrics.

      Lindquist lab is expressing human proteins in yeast cells. Not the same as dialing mutations into a yeast homolog of a rare disease gene. But we do both like genetic screens! That said, I’m also a bit surprised that her system has borne therapeutic fruit in some capacity. I was doing similar yeast-based screens at the time in grad school. I knew the postdoc who did the original chemical genetic screens.

      I know a company called FoldRx was spun off from the Lindquist lab, and is now a wholly-owned subsidiary of Pfizer. Maybe they’ve got something up their sleeve? Then again I wonder how carefully they vetted the yeast screening hits in primary culture cells (now iPS cells) before moving to whole-animal experiments? And then you have to ask what mice and what clinical biomarkers, not to mention more downstream things like PK and biodistribution.

      In other words, I’m not sure yeast is necessarily the source of the problem there, though I agree that you have to make that call on a case-by-case basis.