Getting real about curing polygenic brain disease
Thanks to the molecular biology revolution, it is conceivable that over the next two decades, humanity will cure most monogenic diseases — those caused by a single faulty gene. If the trend of increasing efficacy and specificity of gene editing technologies continues, the “undruggable” genome may one day become “mostly druggable”. To be sure, there are some monogenic congenital diseases that won’t be curable because of irreparable consequences — an egg can’t be unscrambled. The damage has been done, developmental doors have been closed, and the path down a tragic biological side road can’t be reversed. There are aspects of a healthy functional adult physiology that must be assembled gradually, over time during specific periods of development. That said, replacing a missing gene product, restoring the appropriate amount of gene product, or repressing the production of a malformed and damaging gene product will likely address, if not outright cure, most monogenic diseases.
Such lone single-target approaches, however, are doomed to fail in the quest to cure polygenic neurological diseases such as Alzheimer’s, schizophrenia, Parkinson’s, and autisms — ironically, common neurological diseases of high prevalence. There are two major reasons.
Reason 1: Polygenic diseases have complex origins
The root cause of these common polygenic diseases is the interplay of more than one gene. We know this because if there were single genes solely responsible for a common complex disease, they would have turned up in genome-wide association studies (“GWAS”) studies, which have been wildly successful for discovering single genes responsible for many rare diseases.
We don’t know how these common variant neurological diseases arise. This is the crushingly hard work of disease science. But we can be sure they don’t arise from a single mutated gene. We can’t easily use our GWAS method because the epidemiological detection of multiple causal gene variants working in concert requires massive, possibly unobtainable (we run into the limits of human population size) amounts of data.
We offer up a general hypothesis: brains are redundantly adaptive systems. Like viruses and metastatic cancer, the brain has evolved to maintain its integrity and function in the face of perturbations of many kinds, including pharmacologic, i.e. drugs. If you target one pathway, the system actively adjusts itself back to its pre-drug state; this is the ubiquitous phenomenon of homeostasis. Drug tolerance is the consequence, and it is the norm for brain drugs. Ironically, the homeostatic system that normally keeps the system operating stably within safe parameters acts to preserve the disease state. The brain’s adaptivity is profound; humans can unwittingly compensate for severe neurodegenerative disease for years, and sufferers of severe mental disease have transient periods of cognitive normalcy, but adaptivity can break.
We now appreciate that homeostasis is enforced by multiple overlapping control systems, both for organismal (e.g. caloric balance), and microscopic (e.g. synapse health) homeostasis. Like the wily adaptiveness of a rapidly mutating population of viruses or metastatic cancer cells, breaking one pathway quickly leads to a compensatory response from other parallel, redundant pathways.
We have also learned that what may be clinically classified as a single disease should be more accurately labeled a syndrome — a set of co-occurring symptoms — that stem from multiple possible etiologies. We must develop sophisticated molecular methods of stratifying patients by mechanistic origin, in order to focus drug development on particular etiological subtypes and thereby boost the probability of clinical trial success. Medicine must be made precise.
To make matters yet more complicated, polygenic neurological diseases are often the product of environmental influences on top of genetics, further obscuring our ability to detect reliable genetic drivers. Studies of twins have shown that diseases such as schizophrenia are heavily genetically determined as well as consequences of life (possibly prenatal) history.
There are exceptions: certain “rare variant” patient sub-populations of common diseases. For example, a mutation in the GBA gene leads to early onset Parkinson’s disease for many carriers of the gene. However, attacking monogenic flavors of these diseases at their root genetic cause, say via gene therapy, will not yield widespread cures for the polygenic common variants. To reiterate: we now know that these monogenic flavors of complex disease are the exception, not the norm, otherwise we would have detected these smoking-gun genes via large population genomics studies. There is value in studying these rare variant genetics as robust methods to induce common disease states in experimental research models, but we must always be cautious in betting on the extrapolatory power of mono-targeted approaches.
Even diseases labeled as monogenic aren’t as etiologically simple as the phrase “monogenic” implies; for each patient, the nature of their particular mutation in the gene in question as well as their particular background genetics can be substantial determinants of the course and severity of disease.
We must move beyond the lens of single target effects considered in isolation and develop systematic ways to discover complex etiologies of disease that involve multiple interacting gene products.
Reason 2: Polygenic diseases require poly-target cures
It is clear that we must modulate more than one molecular target if we wish to cure complex disease. But how can we develop drugs in a systematic way that affect more than one target? We have two choices: combine multiple drugs that each address different molecular targets, or develop a drug that modulates multiple targets (known by the jargon word polypharmacological or the more casual term “dirty” drug).
The single target mindset is pervasive in pharma. It forms the basis of a convenient, investor-friendly story which asserts that drugs can be developed by following a deterministic and linear procedure: first, select a target, then screen/develop a modulator of the target, then optimize it for drug-like properties, then test for toxicity, then you have a drug. This mindset betrays the brutal truth of attacking complex disease.
There is an optimistic argument in favor of single-target cures for polygenic diseases — the notion that these diseases can be cured by fixing — turning up or turning down — a single downstream effect. This approach has worked brilliantly to stave off brutal secondary effects of many diseases; immunosuppressants like steroids give profound symptom relief for many diseases… for a time. But they never represent complete cures because these downstream systems are needed for healthy, adversity-resistant life. Nevertheless, hopes for a cure-all spring anew with regularity.
It is unlikely that a toggle of a single downstream pathway component will produce complete cures to widespread polygenic disease for the same reason that severe monogenic diseases are always rare and never common; evolution would have found a way to toggle that component if it were so deleterious to so much of the population and so simple to address. There is a counter-argument which holds that we are in a new era of long-lived humans, and so perhaps complex diseases of senescence have not yet had sufficient evolutionary pressure to be eradicated. Certainly the global shift to older population distributions accounts for a great uptick in senescence-associated disease prevalence — in particular degenerative disease. But long-lived people have been part of humanity for a very long time; we now appreciate that they provide evolutionary fitness to their kin groups. And the simple fact remains that well-powered genomic studies fail to turn up single-target smoking guns for any substantial subset of complex brain disease sufferers.
Are we doomed to suffer? Developing complex cures for diseases of complex etiologies sounds like hard business. But there is reason to be optimistic — even bullish. It turns out we have done it many times.
Multi-target cures are proven for other complex diseases
Comparable challenges of complexity have been met with profound success in the treatment of infectious diseases such as HIV via combination therapies. One viral inhibitor will only cause temporary suppression of viral proliferation; a two-drug cocktail will remove any measurable trace of the disease in the patient. Practically, for those who can afford treatments, HIV has been cured. Metastatic cancer is yielding bit by bit to combination therapies. More directly relevant: in psychiatry, many successful treatments consist of an ad hoc titration of combinations of drugs, often polypharmacological, drugs.
Redundant adaptive systems are everywhere in biology. The immune system is another complex, adaptive system that can be repaired by combination therapies: the author’s own formerly intractable and debilitating allergies, diagnosed in youth as “chronic idiopathic angioedema,” are completely suppressed with an ad hoc off-label combination of two inhibitors of different inflammatory pathways, cetirizine (Zyrtec) and montelukast (Singulair).
Why do combination therapies or polypharmacological drugs often work so well? Because adaptive systems can’t outfox multiple different attacks at once. A mathematical identity at least partly explains this: small, independent probabilities multiply into a vanishingly small joint probability. One inhibitor can kill 99% of a viral population, leaving 1% to survive and recolonize, but two in combination will kill 99.99%.
The question for the industry is: can we come up with a systematic way to develop combination therapies or polypharmacological drugs that will cure polygenic neurological diseases?
How do we find multi-target cures?
The discovery of viral inhibitor mixed cocktails that render HIV a non-lethal disease was made by enterprising, desperate doctors experimenting with mixes of drugs in the clinic. Psychiatrists play with a variety of drugs to hone in on effective treatments. But rather than performing a giant ad hoc experiment on the human population, can we find cures in a more systematic way?
The essential ingredient is a prosecutable (scalable and accessible) biologically rich experimental model — one that contains the multiple biological processes that collectively go wrong in a complex disease state — in which we can search for effects of various combinations of target modulations. Such models have recently come to the scientific fore under the moniker of “complex in vitro systems”. In the past decade, scientists have achieved a major step forward in the ability to model realistic human biology at the microscopic level in a dish in a lab. We can now grow cell cultures in such a way that they self-assemble into tiny little bits of proto-organs, called somewhat ghoulishly, “organoids”. Organoids grown from (human) patient-derived stem cells exhibit microscopic features of the disease and represent a realistically effective way to systematically discover cures for complex diseases.
A scaled-up in vitro human experimental system based on patient-derived organoids requires heavy technological machinery in the form of robotic automation and machine learning for data analysis; fortunately, recent revolutions in these fields give us the needed muscle. It is also to crucial to integrate this scaled experimental human biology approach with clues mined from large-scale genetic databases, detailed clinical records, post-mortem tissue analysis, predictive molecular modeling, and yes, animal models.
But suppose we do find effective, curative combination therapies worth taking into clinical trials? There remains a major practical problem: the FDA drug approval process is not set up to think about drugs in this way. Rightfully so, a credible hypothesis of mechanism of action (MOA) strengthens a drug application, and single-target MOAs are easier stories to tell. Furthermore, trials become more complex and expensive, since combination drug approvals require more safety/toxicity studies than single agents thanks to the FDA “combination rule”.
“Piggyback” strategies — layering on new drug candidates with an already approved drug — can reduce the heavy FDA lift, but they are of course constrained by choice of the first drug out of the existing armamentarium which, for some of these diseases, is meager or nonexistent. This clinical trial strategy is growing in popularity in metastatic cancer and infectious disease, where we have embraced the need for a multi-pronged attack.
All of this is not to say that the conventional approach of developing drugs against single targets is futile. Characterizing molecular target engagement is crucial to developing any therapeutic agent. However, for complex brain disease, we believe that single-target modulators will almost certainly be components of a broader therapeutic strategy rather than a Hail Mary touchdown victory. We are grateful for relief, even in measures.
Despite these structural impediments, narrative norms, and scientific complexity, it is time for the pharma industry to embrace this grand challenge, leveraging the new generation of complex — but prosecutable — biological models to solve the puzzles of complex brain disease.