Drugs run in families

When you’re sick — whether you just have a mild headache or you’re at risk of a heart attack — it’s likely that the drug that will be used to treat you is either a natural product or a human-made copy of a molecule originally found in nature.  About half of the drugs on the market today were discovered by screening collections of small molecules made by bacteria, fungi, snails, leeches and other such creatures.  Though the pharmaceutical industry has made serious efforts to get away from this reliance on the natural world, attempting to create rationally designed drugs that are perfectly crafted to fit the structure of the target — or even to make their own collections of random molecules through combinatorial chemistry — natural products still represent an important fraction of the new drugs that are being discovered and approved.  And so it’s interesting to ask where these drugs come from, phylogenetically speaking.  The answers, reported in a recent paper (Zhu et al. 2011 Clustered patterns of species origins of nature-derived drugs and clues for future bioprospecting Proc. Natl. Acad. Sci. USA 10.1073/pnas.1107336108), are surprising.

First, why would it occur to anyone to look for phylogenetic patterns of drug production?  Every species produces biologically active molecules (otherwise the species itself would hardly be biological or active) and surely any biologically active molecule has some chance of turning out to be useful as a clinically approved drug?  Well, perhaps not.  Pharmacologists have come to the empirical conclusion that chemical molecules built on a set of specific patterns — called privileged drug-like scaffolds — have a much better chance of becoming actual approved drugs than other compounds.  It turns out that the enzymes required to produce a specific scaffold run rather strongly in families, and many of these privileged drug-like scaffolds are produced by only one, or at most a few, families of species.  So, for example, the 12 natural-product-derived drugs approved by the FDA as kinase inhibitors turn out to be built on just three scaffolds, and each of these scaffolds is made by only a few species. There is a similar pattern when you look at ligands for the nicotinic acetylcholine receptors: the 53 ligands known are built on 29 (wildly varying) scaffolds, and each of these scaffolds is made by a different family, or small set of families.  There are 5 approved drugs within this set, each of which is made by only a single family.

Streptomyces species. From Discovering Biology in a Digital World.

A few families of bacteria and fungi are well known to be especially good at producing drugs.  For example the bacterial species in the Streptomyces family produce about 2/3 of the clinically useful antibiotics that have been discovered in nature.  To me, that has always seemed like an interesting anecdote, but not an observation with predictive power: surely there are other antibiotics out there, from different families, that we’ll find if we look harder? Again, perhaps not.

If you look at a phylogenetic tree of the Bacteria and color the species that produce clinically approved drugs, what will jump out at you is four major clusters of families that produce lots of drugs against a background of families that produce nothing of clinical interest at all — plus a few (but a very few) outliers. The same is true of plants and fungi: there are a few families that produce lots of drugs (roughly speaking, about 10% of the total number of families) and many families that produce nothing.  In other words, Zhu et al. have extended the idea of privileged drug-like scaffolds to the notion of species that have privileged metabolisms: they express, for their own reasons, a set of enzymes that produce metabolites of a specific chemical character, and these metabolites turn out to be unusually likely to produce drugs.  In general, under 1 in 10 families turns out to be “privileged” (or contain at least one privileged species).  Of 289 families of bacteria, 23 have produced one or more clinically approved drugs, and most of these families fall into the 4 clusters I mentioned.  There are 740 known families of plants, but only 66 produce drugs; fungi do much worse, with 16 of 521 fungal families being drug-producing, and metazoa worse yet: of 4,468 known families, just 38 have produced clinical drugs.  In each kingdom of life, the drug-producing families are strongly clustered.

That’s not to say that there are no drugs coming from species outside the privileged classes.  But the authors classified the nature-derived drugs discovered each year for the last 50 years according to whether the species producing the drug came from a previously known drug-productive family (previous to the discovery of the drug, that is), or from a neighbor of a drug-productive family in a previously-known drug-productive cluster.  The numbers go like this: of 886 nature-derived drugs discovered between 1961 and 2010, 783 (88%) were from previously known drug productive families.  A further 41 were from near neighbors of known productive families.  And only 62 came from the dark horses that had no previous form, the species that were completely outside the known clusters.

You may suspect, as I did, that this is partly due to selection bias.  The libraries of molecules we screen may partly be determined by what has been successful in the past; the molecules that get chosen to move forward may be partly selected because they look familiar, or are well covered by patents, or for some other not particularly scientific reason.  Zhu et al. considered this possibility too, but they find the “looking under the lamppost” theory unconvincing.  There are many situations where lots of biologically active compounds have been found to be produced by a given family, and yet that family has never produced a drug.  For example, there’s been a good deal of excitement about drugs that could be produced by marine organisms, and the bioactive molecules produced by many species have been intensively studied.  But it’s still the case that only a few families have turned out to produce drugs.  We can find lots of biologically active compounds wherever we look, but finding drugs is harder.

The surface implication of this work is that if you’re looking for a successful drug, you might want to focus your screening efforts on small molecules made by families that have produced drugs (not just bioactives) in the past, and molecules made by their relatives.  The question is whether new technologies will change the pattern.   For example, it’s now possible to extract biosynthetic gene clusters directly from the DNA of bacteria found in the environment, side-stepping the time-consuming process of attempting to find conditions in which the bacteria will grow.  Will this approach help us expand the list of families that produce drugs more rapidly and more widely than the older screening methods?  And if we do find more families that are drug-producing, will they continue to fall into clearly defined (though larger) clusters? Will this extra information about the scaffolds that can produce successful drugs help us to understand what’s special about a “privileged” scaffold? Only time, and lots of clinical trials, will tell.

Zhu F, Qin C, Tao L, Liu X, Shi Z, Ma X, Jia J, Tan Y, Cui C, Lin J, Tan C, Jiang Y, & Chen Y (2011). Clustered patterns of species origins of nature-derived drugs and clues for future bioprospecting. Proceedings of the National Academy of Sciences of the United States of America PMID: 21768386