Transmission electron micrograph of 100 μM tetraiodophenolphthalein in 20 mM Tris buffer. © (2002) American Chemical Society.

Many hits from high-throughput and virtual screening programmes are subsequently found to have peculiar, undesirable characteristics — they act non-competitively, show little relationship between structure and activity, and have poor selectivity. Such hits can waste much time and effort, but despite their common occurrence, the underlying reasons for their behaviour have remained unknown. Now, writing in the Journal of Medicinal Chemistry, McGovern et al. provide evidence that these compounds, although structurally unrelated, share the ability to aggregate, which results in them being falsely detected as hits in screening assays.

Initially, the authors investigated 15 diverse compounds described as inhibitors of one or more protein or nucleic-acid targets, and found that the compounds were also micromolar inhibitors of several unrelated enzymes, including β-lactamase, chymotrypsin, dihydrofolate reductase and β-galactosidase. Furthermore, inhibition of various targets was decreased in the presence of bovine serum albumin, a common sign of nonspecific binding. But most intriguingly, the apparent inhibition constants (IC50 values) of all the compounds worsened considerably when the concentration of one of the test targets, β-lactamase, was increased tenfold, in contrast to the behaviour of an established competitive inhibitor.

To account for the extreme sensitivity of the screening hits to the molar ratio of inhibitor to enzyme, the authors considered the hypothesis that the active inhibitor might be an aggregate of many molecules. Dynamic light-scattering experiments (DLS) of aqueous solutions of the hits indicated the presence of particles with apparent diameters ranging from 95 to 400 nm — much larger than the target enzymes, which are 20 nm at most in their longest dimension. The presence of aggregates was confirmed by transmission electron microscopy, and also backed up by further experiments on enzyme kinetics.

So, are aggregate-forming nonspecific inhibitors commonplace in corporate screening collections? To investigate this possibility, the authors tested 30 compounds from the Pharmacia Corporation library, biasing the selection of these compounds towards those that register as hits in multiple screens against different targets. Of these 30, 20 inhibited β-lactamase and chymotrypsin at micromolar concentrations. As with the previous compounds, inhibition worsened as enzyme concentration was increased, and DLS indicated the presence of large particles.

It thus seems that compounds that inhibit enzymes by forming aggregates could often be artificially raising the hit rates in high-throughput screens for new drug leads. Several important questions remain to be answered, such as how the molecules are arranged in the aggregates, and how the aggregates inhibit the enzymes. Increased understanding should lead to computational or screening methods to rapidly eliminate these phoney hits from future screening programmes.