Linking fragments on DNA, revisited

Three years ago we discussed using DNA-encoded libraries to find and link fragments. In a new open-access Bioorg. Med. Chem. article, Nicolas Winssinger and colleagues at University of Geneva report a different version of this approach.

 

Rather than using DNA, the researchers constructed their libraries with peptide nucleic acids (PNAs), which can be assembled using traditional solid-phase peptide synthesis and will also hybridize to DNA. Each PNA is coupled to a different fragment, and the fragment-PNA molecules are then bound to microarrays of DNA such that two fragment-PNA molecules bind to a single DNA strand. In this case the researchers used 250,000 combinations of fragment pairs.

 

Next, a protein of interest (here the anti-apoptotic cancer target BCL-x_L) was screened at 50 nM. Binding to specific pairs of fragments was assessed by fluorescent detection of the protein at various spots on the microarray.

 

Trying to figure out which of the fragments are best – and how to link them – is “not trivial,” so the researchers took a combinatorial approach. Based on the first screen, they generated a new library of 10,000 molecules in which 10 sulfonamide-containing fragments were linked to 100 heterocycle-containing fragments using 10 different linkers. These compounds were screened using the same microarray technology, and the best binders were then resynthesized with a biotin tag rather than the PNA.

 

The biotinylated molecules were able to pull down recombinant BCL-x_L in solution. Two of them, including compound 80-28, were even able to pull down recombinant protein that was spiked into cell lysate. Importantly, neither fragments 80 nor 28 did this by themselves. The affinity of 80-28 was measured by SPR to be 96 nM. 

 

 

Finally, the researchers tested 80-28 in K562 cells and found that it was cytotoxic with EC₅₀ = 1.7 µM. They compare this favorably to venetoclax, the second fragment-based drug to be approved. However, this is a disingenuous comparison: venetoclax was specifically designed to bind less tightly to BCL-x_L than to the related protein BCL-2. A more appropriate comparison would be a molecule such as navitoclax or the specific BCL-x_Lbinder A-1155463.

 

Like most BCL-family binders, compound 80-28 is also a rather unusual looking molecule, with a high molecular weight of 735. Unlike navitoclax and venetoclax, it also has 7 hydrogen bond donors and many more rotatable bonds. The long floppy linker in particular is something the earlier DNA-based fragment-linking work sought to fix. As we noted then, such linkers may be an inherent liability with the approach. From a technology perspective this is interesting work. But from a drug discovery perspective it still has some way to go to prove itself practical.