Courtesy of Steve Smerdon and Mike Yaffe, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA.

Protein phosphorylation can work in two ways: like lego, allowing other proteins to snap into place; or like a switch, triggering a conformational change. Not long ago, dogma decreed that phosphotyrosine was a lego builder whereas phosphoserine and phosphothreonine were conformational switchers. A paper by Daniel Durocher and colleagues in Molecular Cell provides definitive evidence that serine/threonine phosphorylation can build lego too.

In 1999 the same team discovered that the forkhead-associated (FHA) domain — first identified in forkhead transcription factors but now known to be present in many other proteins — is a phosphopeptide recognition motif. They have now used a peptide library to find the optimal binding sequences for several FHA domains, including the two (FHA1 and FHA2) in yeast Rad53p, which binds phosphorylated Rad9p in response to DNA damage. They found that, just like the phosphotyrosine-binding SH2 domain, FHA domains recognize phosphothreonine in the context of flanking amino acids, so individual FHA domains bind specifically to different peptides. Peptide binding depends absolutely on phosphothreonine (its replacement with phosphoserine blocks binding), but the surrounding residues generate specificity. For example, Rad53p's FHA1 domain strongly prefers aspartate three residues downstream of the phosphothreonine (the +3 position), whereas the FHA2 domain prefers isoleucine. Other residues between the −3 and +3 positions also influence binding affinity but, although there are some exceptions, in general the strongest selection depends on an FHA domain's preference at the +3 position.

The crystal structure of Rad53p's FHA1 domain bound to a longer peptide containing its optimal binding sequence revealed why the +3 residue is so important (see picture). Phosphothreonine and Arg+3 make the largest number of direct contacts with the FHA domain; Another intriguing observation is that the FHA domain's fold is identical to that of the MH2 domain in SMAD transcription factors — proof that you can teach an old fold new tricks. But perhaps the most exciting prospect is that this work allows us to trawl the sequence databases for specific FHA-domain-binding proteins. This should help us to build better models of DNA-damage response pathways and other processes that involve proteins containing FHA domains.