At a lecture at the university of Lille in 1854, French chemist Louis Pasteur told the audience: dans les champs d’observation, le hasard ne favourise que l’esprit préparé—in the field of observation, chance only favours the prepared mind.
Until relatively recently, this phrase could adequately summarise the comings-and-goings of drug discovery. Whilst some may rightly baulk at the idea of drug discovery being an observational science, in the past it cannot be ignored that drug discovery was predominantly a passive activity; drugs arose through such innocent activities as leaving poorly prepared bacterial cultures to go on holiday. Comparing this process to the multibillion pound drug industry we find ourselves with today—an industry which clearly doesn’t operate on the offerings of many holidays—which was the first instance of the application of a methodology, rather than Fleming’s luck, to generate new drugs? How does this compare with the industry’s methods today?
One of the first instances of logic being applied to the discovery of drugs can be attributed to the German bacteriologist Gerhard Domagk in the 1930s, whose team posited that if certain dyes could selectively stain bacteria in the process of gram staining, then these same compounds could potentially seek out bacterial colonies in the body. If these dyes could be coupled with a drug, or were drugs in their own right, then it would be possible to selectively target and destroy bacteria within the body. Taking this logic to its natural conclusion led to the discovery of Prontosil, a sulphonamide drug which was tested on Domagk’s own daughter to good effect, curing her from an infection which would otherwise have led to an arm amputation. The glory of the sulphonamide family of drugs would be of limited duration owing to their deleterious side-effects, but this episode showed that drug discovery could be rationalised. A new period was born.
Given the haphazard ways of testing, it might perhaps seem fortuitous that Domagk’s trial was so successful, but not all cases in the drug-industry have proceeded this way. That standards were needed was inevitable, but this wasn’t fully realised until the Thalidomide scandal of the 1960s, after which it was of the utmost importance that manufacturers knew the mechanisms of how the drugs were acting in the body—or what the drug was targeting—and how these drugs were then metabolised. Questions began to be asked, and regulations passed. The partial answer to targeting however only came with the discovery of cellular receptors.
Receptors are proteins present at cell surfaces which elicit chemical responses when they interact with specific molecules. In 1975, Vincent Marchesi and Motowo Tomita of Yale University were the first group to sequence the Glycophorin A receptor, a molecule found in red-blood cell membranes which allows the body to recognise its own cells—the chemical origin behind the idea of a blood group. Another cell-surface species characterised in this period was the insulin receptor in 1971. When blood glucose levels are too high, there is an increase in the amount of insulin binding to the insulin receptor—two complementary structures. The binding process mediates a cascade of chemical reactions —no longer at the cell surface, but inside the cell itself—which serves to reduce blood glucose levels. Drug discovery aims to create molecules, known as ligands, which are structurally complementary to a receptor of interest with one important distinction: they elicit a different level of chemical response than the molecules originally intended to bind, in many cases preventing a chemical response at all.
The above are images (not to scale) taken from scientific journals of the human insulin receptor (left) and monomeric human insulin (right); the type of insulin which binds to the receptor. We can see immediately see that whilst the insulin molecule can fit inside of the middle of the receptor, insulin isn’t a perfect geometric fit. In fact this is an important point: molecules need not be an exact ‘geometric’ complementary shape to the receptors they bind to in order to do anything chemically interesting. As to why millenia of evolution would create insulin molecules which are not perfect geometric complements of their receptor is a telling point, indicating that shape is not the whole story. This is still an open problem.
One of the questions that medicinal chemistry then asks is whether we can design molecules which are better ligands than the natural ones, or whether we can create molecules—called antagonists—which fit the receptor so well they block natural ligands from binding. Hence no chemical response is elicited.
But how can we do this? Predominantly since the early 1990s, computational chemistry has added to the trove of techniques used for drug discovery. One facet of this computer-aided ligand design relies on high resolution structural data of receptors (such as the example of the human insulin receptor above) to computationally ‘grow’ molecules within them, such that the resultant molecules—if they pertain to sensible molecules at all—are known to fit the receptor site exquisitely. Synthesis of these virtual ligands follows if they can be made with current techniques. Such a process is problematic if the ligand in question is very large and has many stereocentres, and so ‘fragment-led discovery’ can be used instead. Fragment-led discovery uses information from molecules already known to elicit receptor responses. Naloxone, structurally similar to heroin and morphine, prevents opiate overdoses based on this principle. Antagonists may have similar structures to the original agonists since they both bind to the same receptor, and conversely
As we mentioned before, complementarity may not always imply activity; the above process might not always give us the desired drug molecules we require. We expand on our previous idea using the method of taking a molecule known to elicit some response, and visually splitting them along bonds to form ‘fragments’ which correspond to smaller, isolable molecules. Each fragment is screened for activity.If the activity is less than the starting molecule we may start the process again. If we find a fragment which leads to increased activity, other groups can be added to the active fragment, and the receptor response analysed again. An iterative process. In this fashion, further addition or removal occurs to alter the response as required. Cimetidine, which prevents peptic ulcers, was discovered in this way , being based on fragments from the natural agonist histamine.
A brief Q and A of how activities of drugs are actually calculated can be found on the Chemistry Stack Exchange, however there has been at least one research group in the world (albeit a very unconventional one) which has utilised the philosophy of these discussions without the need for complex calculation. Based in a small outhouse in the hills near San Francisco, the American chemist Alexander Shulgin (1925-2014) personally synthesised and assayed (i.e. took) hundreds of psychedelic drugs which were new to science. To find new psychedelic substances, Shulgin and his group would start with a compound known to elicit hallucinogenic properties—mescaline for example. He would then synthesise new compounds based on this structure, literally adding or taking away one atom at a time compared to the original molecule. The group would personally ingest this new drug for hallucinogenic properties and rank it on the ‘Shulgin scale of activity’. The intensity of the psychedelic experience thus produced was used as a measure of the drug’s activity.
Mescaline and its derivative alpha ethyl mescaline (above) differ by the addition of an ethyl group to the carbon just next to (the alpha carbon) the nitrogen atom. Shulgin initially synthesised the methyl variety and wrote that:
The extension of the two-carbon chain of mescaline by alpha-methylation to the
three carbon chain of TMA approximately doubled the potency of the compound.
When it came to extending the methyl variation to the ethyl one (above) however, Shulgin and his team found that this molecule had very few hallucinogenic properties at all, merely just by changing one small group.
Bypassing the calculations used in industry, Shulgin and his team showed that for a whole range of psychedelics, adding or removing groups could show implicitly how well these molecules ‘fit’ within the receptor sites responsible; the better the complementarity the more intense the experience. It must be noted however that such methods used to assay activity are extraordinarily dangerous and cannot (and shouldn’t) be used for the majority of molecules out there. The outcomes are always unpredictable.
The nature of Shulgin’s psychedelic research underpins the receptor/ligand approach to drug discovery quite succinctly. Computational methods are anticipated to lend a stronger hand towards discovery of medicinal drugs in the future, improving with better receptor models and increased synthetic prowess. Although, with many disasters to every success, may Fleming have the last laugh yet?
—Le Nouvel Artiste
Photograph of Shulgin in his lab by Scott Houston/Sygma/Corbis, 2001.