Fishing for protein-protein interactions

In connection with fishing for proteins interacting with a known protein, I want to mention a paper entitled "Screening for in vivo protein-protein interactions", F.G. Germino, Z.X. Wang and S.M. Weissman, Proc. Nat. Acad. Sci. 90:933-937 (1993). The basic idea is to use molecular biology to fuse one "reporter" segment onto a gene for a protein in whose interactions one is interested, and another onto cDNA for all (as best you can) the proteins in the cell.

They fuse a fragment, ≈ 350 bp = 117 amino acids, of the biotin carboxylase carrier protein – the BCCP sequence - onto the protein of interest (in frame) - they use the leucine zipper protein c-Jun. This adds a sequence which will be biotinylated in vivo, and will therefore stick to immobilized avidin. They transformed E. coli cells with this construct, in the plasmid pMALcr-1, and showed that most of the construct is biotinylated in the cells and sticks to columns with avidin immobilized on them. If the BCCP sequence was attached to the b-galactosidase gene and put into a l phage vector, so that transfected cells will give plaques whose released protein can stick to a filter laid over the plate, they duly found b-galactosidase activity on an avidin-coated filter laid on the plate (but not if the b-galactosidase gene did not have the BCCP attached).

They then cloned the leucine zipper region of c-Fos, known to interact with c-Jun, at the 3' end of the lacZ (b-galactosidase) gene, in a l phage vector, and transfected this phage into cells containing the plasmid coding for the c-jun.BCCP. construct. They did find b-galactosidase activity on an avidin-coated filter laid on the plate. Controls giving negative results included: cells with BCCP attached to other proteins which don't interact with c-fos, or with BCCP attached directly to the MalE gene of the vector; filters without avidin; cells with c-jun without BCCP attached. Filters coated with streptavidin or polyclonal anti-biotin antibody also worked, but monoclonal anti-biotin antibody didn't.

This does have limitations: post-translational modifications generally aren't done in bacterial cells, so if they are necessary for the interaction, it won't be seen. If a third protein or nucleic acid is necessary for the interaction, it won't be seen. If the protein isn't stable, or isn't folded right, when attached to BCCP or b-galactosidase, this won't work. In a cDNA library (which they haven't tried yet), only one-sixth of the DNA sequences will be in frame to give the proteins being screened. One could use monoclonal antibody to b-galactosidase, instead of its substrate, and pick up parts of the b-galactosidase sequence, but this takes longer and the alkaline phosphatase assay is at pH 10, so the interaction might break down during the assay.

The point of attaching a reporter sequence to cDNAs is of course that different cDNAs will be in different colonies or l plaques on the plate, you will be looking for the colony or plaque that gives a spot of b-galactosidase activity on an avidin-coated filter. It occurs to me that if you attach the b-galactosidase sequence to cDNA for all proteins being synthesized in the cell, some of them will themselves naturally be biotinylated, and thus would give you b-galactosidase activity without having to interact with the known protein. I don’t know how they get around this.