DRUG DEVELOPMENT:

Drugs fail for many reasons, e.g. they do not treat the intended condition or they cause nasty side effects. Predicting whether a drug will likely fail, before a great deal of time and effort are put into development and testing, speeds up and lowers the cost of drug discovery, and is a goal shared by many scientists.

Relevant to this issue, scientists have recently evaluated how small similarities in protein structure may predict whether two different proteins contribute to the same function, relevant to understanding the total biochemical effect of a drug. Complimentary insight can be gained through an analysis of drug action in the face of drug tolerance.

This is not a reference to antibiotic resistance; that's a different topic. Here, I am instead referring to physiological adaptation to a drug.

Feedback loops and drug tolerance.

Given that drugs function by perturbing a biochemical pathway, a successful drug must overcome cells' inherent ability to overcome biochemical pathway fluctuations. Furthermore, this drug-induced feedback alters the biochemical production of the drug targets (i.e. proteins).

Which proteins feature feedback loops that may inhibit drug targeting? The simple answer is that no one knows, but it's nevertheless crucial to take this into consideration when designing and formulating a drug.

Scientists have recently compiled scientific resources for analyzing drug-protein interactions and the effects of drugs on gene expression in cells. This is relevant to predicting the existence of feedback loops.

The limitation of combining these resources to an analysis of drug resistance is that cellular gene expression may be similar in response to unrelated stimuli if the cells are grown at the same time (known as the batch effect). This overestimates the existence of feedback loops across a population of drugs.

Peer Bork (European Molecular Biology Laboratory, Germany) and coworkers have addressed this limitation. Their computational approach suggests that at least 8% of proteins targeted by drugs exhibit drug-induced feedback loops, and have identified several proteins previously unknown to be regulated in this manner.

Predicting feedback loops and protein networks.

The scientists made heavy use of Connectivity Map, a database of gene expression in four types of cells, as a consequence of treatment with small molecules. They also made heavy use of STITCH, their own database of protein-chemical interactions.

They developed a computational approach to eliminate the batch effect, thereby enabling them to link multiple drugs against a single protein target. They checked the validity of their approach by a computational demonstration that drugs linked to the same protein are often chemically similar, a commonly accepted indicator of similarities in drug target and biochemical effect.

The scientists analyzed their data of nearly 1300 drug-protein relations for which gene expression across the entire genome is also available. Thirteen out of 167 unique drug targets (8%) are subject to drug-induced feedback loops; seven of these 13 are supported by experimental evidence.

G protein-coupled receptors are over-represented in this group. These proteins are common intended drug targets.

G protein-coupled receptors perform many physiological roles, e.g. those relevant to scent, vision, inflammation, and mood. Given that the function of these proteins is often critical to life, it makes sense that these proteins are subject to feedback loops (but remember that many drugs are designed to target these proteins).

The scientists further found that 259 out of 466 drugs (56%) target more than one protein, and that 4 of them (1%) act upon proteins which are subject to drug-induced feedback loops. An example is podophyllotoxin, a drug which is commonly used to treat symptoms of herpes, and is under active investigation in chemotherapy.

This drug is known to inhibit two proteins (tubulin beta-2C and DNA topoisomerase 2-alpha). Given that topoisomerase 2-alpha is not inhibited by other molecules which inhibit tubulin proteins, there is no evidence that the feedback loops regulating these two proteins are cross-regulated.

This confirms experimental findings. Perhaps most useful is the scientists' finding of drug-induced feedback loops not yet reported on an experimental level.

One of these is a feedback loop inhibiting calmodulin 1. This is a protein involved in (among other functions) the release and synthesis of neurotransmitters.

Such predictions will aid scientists' search for proteins most likely to be subject to drug-induced feedback loops, and thus most in need of special attention during the drug development stage.

Implications.

Biochemical feedback loops need to be considered early on in drug development. Drugs which cause a protein to be overproduced or underproduced are known to inhibit treatment of various conditions, including cancer and asthma, via the drug tolerance induced by altered protein expression.

The problem of drug tolerance can't be solved by altering drug dosage, as this can lead to nasty side effects (incidentally, scientists are beginning to understand how two seemingly different diseases may be in fact be related based upon similar drug side effects). Given that at least 8% of protein targets are susceptible to drug-induced feedback loops, this is an issue which is begging for greater appreciation in the drug development community.

NOTE: The scientists' research was funded by the Federal Ministry for Education and Research.

for more information:
Iskar, M., Campillos, M., Kuhn, M., Jensen, L. J., van Noort, V., & Bork, P. (2010). Drug-Induced Regulation of Target Expression PLoS Computational Biology, 6 (9) DOI: 10.1371/journal.pcbi.1000925