Technical Highlight - February 2013
Short description: A screen of natural metabolite ligands provides clues to protein function.
Structural genomics projects have yielded myriad structures of proteins with unknown molecular function. Although roles can sometimes be assigned on the basis of sequence or structural similarity, novel methods are needed to determine the functions of the numerous proteins and protein superfamilies that cannot be characterized by bioinformatics alone.
Minor, Shumilin and colleagues (MCSG and JCSG) have developed a new approach centered on ligand identification. Although proteins typically selectively bind their cognate ligand(s) with high affinity, they nevertheless may bind related ligands or ligand fragments with low affinity. The authors inferred that restricting the screen to naturally occurring compounds would reduce nonspecific binding. They created 11 cocktails, each containing 5–11 commercially available natural metabolites.
The screen is carried out in two phases. First, protein crystals are soaked with each of the metabolite cocktails and screened by X-ray crystallography, which enables detection of low-affinity protein-ligand interactions. This necessitates accessibility of the binding site in the crystal and that major structural rearrangements are not required for binding. Second, metabolic compounds structurally related to the hits from the cocktail screening are analyzed by both crystallography and isothermal calorimetry to determine high affinity binders that are likely to be the protein's natural ligands (substrates, effectors, etc).
For the YjeF_N family, which currently consists of just under 2,500 sequences in Pfam, several available crystal structures revealed a two-pocket cavity, but neither the ligands nor the function were known. In this proof-of-principle work, crystallographic screening of cocktail binding by the proteins from this family, Tm0922 and AI-BP, produced two binding hits. These hits were used to design two sets of metabolites, the “NAD” set and “thymine” set, which were utilized to elucidate a putative substrate for the YjeF_N sites. The substrate identification led the authors to propose that this family is involved in transfer of ADP-ribose, despite a lack of similarity to enzymes known to catalyze this reaction. Applying the same approach led to identification of the biochemical function of the PF01256 protein family, which includes almost 3,000 sequences in Pfam. These proteins were found to be responsible for the orphan “metabolite repair” activity that recovered NAD(P)H from a non-physiological adduct spontaneously forming in the cell.
As many apo structures of proteins with unknown function are already available, the bottleneck of determining crystallization conditions is avoided. Furthermore, any available experimental data can be used to narrow down the starting metabolite library. In the absence of protein function hints, expanding library diversity and size would benefit the analysis.
I.A. Shumilin et al. Identification of Unknown Protein Function Using Metabolite Cocktail Screening.
Structure. 20, 1715-25 (2012). doi:10.1016/j.str.2012.07.016