Fragments vs IAPs: resisting the affinity Siren

In 2011, Mike Hann decried “addiction to potency”. Indeed, newcomers to fragment-based methods often have to undergo a psychological shift to work with low affinity binders. But, as we asked last year, how weak is too weak? In a paper just published in J. Med. Chem., Gianni Chessari and colleagues at Astex may have set a new bar.

The researchers wanted to develop leads against inhibitor of apoptosis proteins (IAPs). The BIR3 domains of proteins such as cIAP1 and XIAP bind to and block the action of caspases and other proteins, allowing cancer cells to survive. Several groups have developed molecules that block these protein-protein interactions, usually by starting with an endogenous peptide inhibitor. However, most of these bind preferentially to cIAP1, and some evidence suggests that a balanced antagonist may have advantages.

The BIR3 domain is quite small, just 11.8 kDa, making standard ligand-observed NMR screening difficult. Instead, the researchers looked at line broadening and chemical shift changes of protein protons with δ < 0.4 ppm or 9.8 – 10.4 ppm on addition of fragments. Anticipating very weak binders, they used 0.2 mM protein and 10 mM fragments (in mixtures of two). A total of 1151 fragments were screened, of which 100 had been computationally preselected.

Among the best hits were those containing an alanine residue; one of these had mid-micromolar affinity and reasonable ligand efficiency. However, these bound preferentially to the BIR3 domain of cIAP1, in common with other reported inhibitors that also contain an alanine. In contrast, compound 1 appeared to have more balanced activity. Its affinity was risibly weak, but crystal soaking led to interpretable electron density, and also suggested that adding a suitably positioned methyl group could fill the small hydrophobic pocket normally occupied by the alanine side chain. The resulting compound 5 showed measurable – and balanced – activity against both proteins.

Next, a small virtual library was constructed in which the pyrrolidine was replaced with substituents to both improve affinity and create a scaffold for reaching into the P4 pocket. Thirty compounds were made, but disappointingly none of these had significantly improved affinity. Undeterred, the researchers obtained crystal structures of some of them bound to XIAP-BIR3 and performed careful modeling. The phenyl ring of compound 7 binds in a region of the protein with an electronegative potential, and by simply adding electron withdrawing substituents the affinity could be improved by >50-fold. Further growth ultimately led to compound 21, with nanomolar potency against both XIAP and cIAP1, though with a preference for the latter. This compound also showed on-target activity in cell-based assays as well as activity in mouse xenograft models.

It would have been easy to overlook compound 1; indeed, it took rather strenuous efforts to find it. Yet, comparing the structures of compound 1 and 21 bound to XIAP-BIR3 reveals that the initial fragment maintains its position and binding interactions in the elaborated molecule. This is a clear example that, with persistence and creativity, it is possible to advance even the weakest of fragments. The researchers note in the conclusion (and have reported at conferences) that they were able to optimize this series to low nanomolar inhibitors against both targets. Whether or not this leads to a drug, it does look like another candidate for a useful chemical probe.