Rare disease diagnosis and drug discovery
Folks in the pharmaceutical drug industry talk about innovation cycles in terms of decades, even in the case of the simplest rare diseases, or autosomal recessive disorders, which are caused by inheriting defective paternal and maternal copies of a single gene. For example, consider the case of cystic fibrosis (CF). In 1989, the precise chromosomal coordinates of disease-causing mutations were triangulated to a gene called CFTR by a team that included present-day NIH director Dr. Francis Collins using the genetics equivalent of an astrolabe.
But then it took 23 years – I repeat, twenty-three years! – to get FDA approval in 2012 of the first drug to correct the underlying physiological defect called Kalydeco/ivacaftor, which as miraculous as it is only rescues around 5% of CF patients who carry a specific, rarest of the rare mutant form of CFTR.
How can we speed up the process of going from diagnosis to drug in the rare disease space? Is it crazy to imagine super-charged innovation cycles on the order of a few years, or even 6 months, like we see in tech and Internet startups? I think recent advances in DNA sequencing — from astrolabe to GPS — and microfluidics — from 96-well plates to nanodroplets — could usher in a renaissance of genetic model-organism screening to the benefit of rare disease research, which to date has been monopolized by well-reasoned but so far narrowly successful attempts at spellchecking mutations in DNA (aka gene therapy), or forcibly colonizing diseased tissue with healthy replacements (aka stem cell transplantation).
Or at least that’s my plan. Accelerating and simplifying rare disease diagnosis and drug discovery will be my major focus as an independent scientist. I will lay out my scientific vision in several parts, starting with this preamble.
There are approximately 7,000 rare diseases. Where do I even begin? Turns out the last 5 years I spent running a small academic lab provide insight into a group of ~50 rare diseases called lysosomal storage disorders (LSDs). The genes responsible for LSDs are known, and many of them are enzymes that live in lysosomes, the vinegary organelles where cellular components go to be scrapped for parts. By contrast, half of the 7,000 rare diseases have yet to be diagnosed at the gene level, and I’ll talk about the challenges associated with them in a future post. (Intrepid biotech reporter Lisa Jarvis at C&E News has a wonderful recent four-part series on rare diseases, including the LSDs called Sanfilippo).
The state of the art treatment for most LSDs is enzyme replacement therapy (ERT), pioneered by the biotech company Genzyme, which manufactured the first ERT for Gaucher disease in the early 1990s. More than a decade on, there are companies like NJ-based Amicus Therapeutics, which is pursuing a complementary approach to resuscitate residual lysosomal enzyme function with small-molecule boosters. However, most drug discoverers focusing on LSDs aren’t pursuing a small-molecule solution, even though that may be the easiest way to cross the blood-brain barrier, a necessity for LSDs that cause neurodegeneration. The reason is simple: besides the fact that we know a specific enzyme is broken, we don’t understand how this deficiency ripples throughout disease-modifying genetic networks.
I’ve spent the last 10 years in academia thinking a lot about pharmacology and evolution, and I’ve come to the conclusion that the only way to develop an integrated theory of a disease is to study it first in simple model organisms with fast-generation times, ease and specificity of genetic manipulation, and appropriate conservation of disease-modifying genetic networks, including obviously the disease-causing node itself. Niemann-Pick type C, or NPC, is a great example of a disease the can be modeled in every single genetic model organism, from brewer’s yeast (Saccharomyces cerevisiae) to nematode worms (Caenorhabditis elegans) to fruit flies (Drosophila melanogaster) to zebrafish (Danio rerio), the last of which, according to the recent genome sequencing paper, shares a whopping 75% of its genome with humans in a pure gene-to-gene accounting. (Whether entire genetic networks are also fully conserved is a safe but not guaranteed bet that must be empirically shown).
If we’re ever going to hasten innovation cycles in rare disease research, we’re going to have to wring out inefficiencies, a process naturally abetted by technology gains, especially in the area of diagnosis. But I think rare disease drug discovery requires fresh conceptual approaches above all else. I’ll treat NPC as an exemplar of the typical historical progression, and where we go from here.
Alfred Niemann and Ludwig Pick first classified the cluster of symptoms that ultimately inherited their names in the interwar period. Only decades later, when powerful electron microscopes allowed us to peer deep inside cells, down to membranes, which are 1/10th the width of a human neuronal synapse, did metabolic indigestion reveal itself as organellar anomalies. What’s characteristic of most if not all lysosomal storage disorders is the accumulation of telltale membranous structures.
Take, for example, this American Journal of Pathology paper from 1973:
What you’re seeing are the insides of NPC patient-derived fibroblasts, which can be yanked out of the body via biopsy and kept alive for weeks on petri dishes. The cells that survive this forced deportation can be viewed under an electron microscope. Orange arrows point to mesmerizing membrane whorls, which are not observed in healthy cells.
The two genes that when mutated cause NPC succumbed to the gene hunters of the post-recombinant DNA Wild West on the eve of the completion of the fabled Human Genome Project. Nowadays gene hunting is done with a next-gen sequencer, but back in the day they used a principle called linkage. In a 1997 paper in Science, a group including longtime NPC researchers Peter Pentchev and William Pavan (and others) finally located the speck of DNA that contained the NPC1 gene, and a quick comparison to the genomes of simple genetic model organisms revealed its deep evolutionary conservation:
We still don’t know exactly what the NPC1 gene does and why aberrant membrane accumulation occurs, though the similarity of NPC1 to proteins that are known to synthesize or respond to cholesterol suggest that it’s involved with the trafficking of cholesterol inside cells, and some cell types, e.g., neurons, seems to be acutely sensitive to its disruption.
As I’ll describe in much greater detail in followup posts, NPC is a disease that doesn’t discriminate evolutionarily.
NPC has been modeled in a yeast cell in work by the Sturley group at Columbia that I’ll highlight. NPC has also been modeled, albeit in fits and starts, in worms starting in 2000. In fact, NPC1 mutant worms die during development, which tells us that the degeneration observed in NPC patients is conserved. Fruit flies got into the action, too. Beautiful work by the Scott lab at Stanford starting in the mid-2000s showed that mutant NPC1 flies not only exhibit developmental growth defects, but also exhibit the telltale ultrastructural marks of aberrant cholesterol accumulation. Finally, in the last few years NPC1 mutant zebrafish have been created, and they too recapitulate the symptoms of the human disease.
Have we extracted every last ounce of value from all those sunken basic research costs using simple genetic model organisms? Are we doing the best job leveraging evolutionary conservation of disease-causing and disease-modifying pathways, even with the caveat that sometimes simple organisms are an atavism and so not always 100% relevant to humans?
That’s what I intend to find out with “Perlstein Lab LLC.”