In vivo selection of intrabodies specifically targeting protein-protein interactions: a general platform for an "undruggable" class of disease targets.

Protein-protein interactions represent a major potential drug target for many human diseases, but these are unanimously considered undruggable with small chemical molecules. We have developed 3-SPLINT, a novel technology for the selection of antibodies that are intrinsically endowed with the ability to interfere with a given protein-protein interaction. The selection procedure exploits the recently described yeast SPLINT libraries of intrabodies, adapting them to a reverse-hybrid system, yielding the selection of recombinant antibodies that are able to disrupt a target protein-protein interaction in vivo. This class of antibodies should therefore perturb an individual protein-protein interaction, without perturbing the scaffolding function of the target protein in that complex, or other protein interactions of that same protein. We provide here a proof of concept of the technology, by the de novo selection of antibodies against two distinct interacting protein pairs: the GABARAP, which interact with the gamma2 subunit of GABA(A) receptor, and the p65 protein dimer, involved in the NF-kappaB-mediated signalling transduction pathway. Intrabodies selected against the latter were functionally validated in cells. Such antibodies, by interfering with the dimerization domain of p65, lead to an activation of the NF-kappaB-mediated transcriptional activity, which is normally inhibited by p65 knock-down RNAi. This provides a clear-cut demonstration that interfering with a protein interaction can be functionally very different from physically removing one of the interacting proteins. The 3-SPLINT approach provides a general and finer tool for the functional validation of selected protein interactions in protein networks, and is ideally applied to protein "hubs", displaying multiple distinct interactions. 3-SPLINT will therefore complement RNAi-based approaches, in the toolkit of target validation strategies, and is amenable to the systematic isolation of comprehensive sets of antibodies against most protein-protein interactions of a given protein network.