Plenge Lab
Date posted: March 9, 2015 | Author: | No Comments »

Categories: Drug Discovery Human Genetics Precision Medicine


My overly simplistic vision of the way to transform drug discovery is to (1) pick targets based on causal human biology (e.g., experiments of nature, especially human genetics), (2) develop drugs that recapitulate the biology of the human experiments of nature (e.g., therapeutic inhibitors of proteins), (3) develop biomarkers that measure target modulation in humans, and (4) test therapeutic hypotheses in humans as safely and efficiently as possible.

Thus, one of my favorite themes is “causal human biology”. The word “causal” is key: it means that there is clear evidence between the cause-effect relationship of target perturbation in humans and a desired effect on human physiology. Human genetics represent one way to get at causal human biology, and in my last blog I highlighted recent examples outside of human genetics.

I am constantly scanning the literature to find examples that support or refute this model, as I predict that a discipline portfolio of projects based on causal human biology will be more successful than past efforts by the pharmaceutical industry.

This week I have selected two articles on genetics/genomics in drug discovery that provide further support of this model. [Disclaimer: the first study was funded by Merck, my employer.]

Clinical improvement in psoriasis with specific targeting of interleukin-23, Kopp et al Nature (March 2015).

Background: Ustekinumab, an approved therapy for psoriasis, is a monoclonal antibody that binds the IL-12/23p40 (“p40”) subunit shared by both IL-12 and IL-23. The p19 subunit pairs with the p40 subunit to form the heterodimeric cytokine IL-23; the p35 subunit pairs with p40 to form IL-12. Thus, therapeutic modulation of ustekinumab tests both the IL-12 and the IL-23 axis in psoriasis.

Human genetic data implicate the IL12-IL23 axis as a driver of psoriasis (see here), including associations at IL23R, IL23A (p19) and IL12B (p40). Based on these data, a more specific test of the genetic therapeutic hypothesis is to make an inhibitor of p19 or an antibody against IL23R. The investigational drug tildrakizumab specifically targets p19.

Results: The current study was performed in 77 total patients with psoriasis in a randomized, placebo-controlled, rising multiple-dose phase I study in patients with moderate-to-severe psoriasis. After three doses of drug over a 56- or 84-day period, there was a dramatic reduction in symptoms that was sustained out to approximately 200 days (Figure 1b). Tildrakizumab also improved the epithelium in psoriasis patients, including the histopathological psoriasis severity score and immunophenotypical changes (Figure 2).

Why this is important: The most important result is that the drug works in patients with psoriasis. It is also important because it provides another example of human genetics leading to successful therapies (although this drug still requires additional testing in larger populations of patients). Lastly, it is important because it provides an example of how to streamline a clinical trial design to get a meaningful result in a relatively small Phase I clinical study. In the future, I expect that we will see more clinical studies that test therapeutic hypotheses based on causal human biology – just like this one.


In addition to providing direct support for a target, human genetics can also be used to assess targets that are linked to disease states through observational data. That is, human genetics can be used as a tool to assess causality when the cause-effect relationship is ambiguous. The second article of the week in genetics/genomics for drug discovery provides a very interesting example.

Cardiometabolic effects of genetic upregulation of the interleukin 1 receptor antagonist: a Mendelian randomisation analysis, The Interleukin 1 Genetics Consortium, Lancet Diabetes & Endocrinology (February 2015).

Background: Interleukin-1 (IL-1) is a master regulator of inflammation and a target for anti-inflammatory drugs used in rheumatoid arthritis and other autoimmune diseases. Based on the potential link between inflammation and cardiometabolic diseases, there are also ongoing trials of IL-1 inhibitors in preventing heart disease. However, IL-1 plays a role in multiple biological processes and the broader consequences of IL-1 inhibition are not clear. To address this need, a paper published by Dr. John Danesh and colleagues uses Mendelian randomization to assess the impact of natural IL-1 inhibition on the risk of rheumatoid arthritis, cardiometabolic diseases, and number of other phenotypes.

Results: To start, the authors leverage a genetic score that has been robustly associated with serum levels of IL-1Ra, an endogenous antagonist of IL-1 receptors, to test predicted levels of IL-1 inhibition against disease risk using data from large disease mapping studies. Beyond the standard Mendelian randomization caveats (e.g., life-long genetic versus temporary pharmacological inhibition), one key consideration for this study is that IL-1Ra antagonizes both the IL-1α and IL-1β receptors. The results are probably more relevant for dual IL-1α/β inhibitors (e.g., anakinra) than for drugs that selectively target one receptor (e.g., the IL-1β inhibitor canakinumab). Consistent with pharmacological data, the authors find that greater levels of IL-1 receptor inhibition are significantly associated with lower risk of rheumatoid arthritis (OR = 0.97, p-value=9.9×10-4). This is a small effect relative to observations from clinical trials of IL-1 inhibitors, but treatments used in these trials cause much larger (~5-10x) changes in IL-1Ra levels than the genetic score.

Perhaps the most surprising result is that higher IL-1 receptor inhibition is also significantly associated with increased risk of coronary heart disease (OR=1.03, p-value=3.9×10-10) and abdominal aortic aneurysm (OR=1.08, p-value=1.8×10-5). This suggests that drugs that inhibit IL-1 activity may also increase the risk of cardiovascular adverse events. The genetic IL-1 inhibition score was also significantly associated with LDL (0.3%, p-value=9.5×10-5), total cholesterol (0.3%, p-value=5.8×10-7), and triglyceride levels (0.4%, p-value=2.6×10-3). The authors estimate that these lipid effects could explain approximately a third of the association with coronary heart disease.

Why this is important: The results presented in this paper highlight an unexpected connection between IL-1 inhibition and cardiovascular disease that merits a more in depth exploration. Furthermore, this paper demonstrates the power of human genetic data to explore the potential implications of perturbing drug targets without performing randomized clinical trials.




Leave a Reply