Promiscuous protein biotinylation by <em>Escherichia coli</em> biotin protein ligase

Abstract

Biotinylation of proteins has routinely been done by chemical means, usually by modification of protein amino groups with biotin-N-hydroxysuccinimide or similar acylating agents. In contrast, enzymatic biotinylation has been limited to the few proteins that normally carry this modification, which are largely biotin-dependent carboxylases and decarboxylases of central metabolism (Chapman-Smith and Cronan Jr. 1999). We supposed that if enzymatic biotinylation could be made less specific, it might provide a means to detect weak (i.e., having dissociation constants >10 ^{M) protein–protein interactions. In this scenario the biotinylating enzyme physically coupled to one of the interacting proteins (the target protein) would be used to biotinylate and thereby tag proteins that interact with the target protein. The biotinylation reaction should tag only those protein molecules that are close in space to the target protein. That is, unlike chemical acylation, enzymatic biotinylation should be proximity-dependent. The specific and extremely tight binding of biotin (K}_D 10 ^{to 10 M) to streptavidin and avidin would then allow very sensitive detection of biotinylated proteins by a wide variety of robust protocols and low affinity forms of streptavidin and avidin allow efficient purification of biotinylated proteins under mild conditions. However, the biotin protein ligases (BPLs) catalyzing protein biotinylation have exceptional specificities for their protein substrates and thus are not general protein modification enzymes. For example, in vivo, BirA, the BPL of} Escherichia coli, biotinylates only a single protein, the BCCP (AccB) subunit of acetyl-CoA carboxylase and other organisms contain less than five biotinylated protein species, indicating these BPLs are similarly specific (McAllister and Coon 1966; Samols et al. 1988; Chapman-Smith and Cronan Jr. 1999). Therefore, in order to use a BPL as a general biotinylating enzyme, this extraordinary specificity must somehow be overcome. A possible route to this end is based upon the mechanism of the BPL reaction, which proceeds in two steps:

1. Biotin + ATP ⇌ Bio-5′-AMP + PPi

2. Bio-5′-AMP + apo-Protein →Biotinoyl-Protein + AMP

In the first partial reaction, BPLs catalyze the synthesis of biotinoyl-AMP (bio-5′-AMP, which is also called biotinyl-adenylate) from ATP and biotin (McAllister and Coon 1966; Chapman-Smith and Cronan Jr. 1999). The enzyme then sequesters bio-5′-AMP in the active site until the second partial reaction proceeds. In the second partial reaction, the nucleophilic ɛ-amino group of the target lysine residue of a biotin-accepting domain attacks the mixed anhydride of the bio-5′-AMP bound within the BPL active site to form an amide bond between biotin and the lysine side chain that remains intact for the life of the protein (McAllister and Coon 1966; Chapman-Smith and Cronan Jr. 1999).

A possible means to convert a BPL to a promiscuous protein biotinylation enzyme would be to alter the protein such that the mutant enzyme releases bio-5′-AMP from the active site. Bio-5′-AMP is a mixed anhydride and therefore should act as a nonspecific chemical protein biotinylation reagent. A precedent for this scenario is the nonenzymatic acylation with serine of noncogate acceptor proteins by seryl-AMP released from an adenylation domain derived from EntF (Ehmann et al. 2000). Moreover, the instability of acyl-adenylates such as bio-5′-AMP should result in proximity-dependent biotinylation. This is because any acyl-adenylate molecules that diffuse far from the enzyme should be inefficient protein acylation reagents due to their low concentration and the high rate of acyl-adenylate hydrolysis.

The best studied BPL is E. coli BirA, a monomeric protein of 35.3 kDa. BirA is a multifunctional protein that catalyzes biotinylation of apo-BCCP and also acts as the transcriptional repressor that regulates biotin biosynthesis (Cronan Jr. 1989; Beckett and Matthews 1997). The X-ray crystallographic structure of BirA has been determined (Wilson et al. 1992; Weaver et al. 2001b) and mutations that affect biotinligase activity are located in a disordered loop that becomes more ordered when biotin occupies the active site (Weaver et al. 2001b). Two mutant proteins, G115S and R118G, having alterations within the loop, are defective in binding of both biotin and bio-5′AMP whereas ATP is bound normally. The dissociation constants of the G115S and R118G proteins for bio-5′-AMP binding are, respectively, 3000- and 400-fold greater than that of wild type BirA whereas the reported changes in the biotin binding constants are less dramatic (250- and 100-fold greater than wild type, respectively; Kwon and Beckett 2000). BirA Δ1–34, a protein lacking the N-terminal DNA-binding domain, also binds biotin and bio-5′-AMP weakly (dissociation constants 100- and 1000-fold greater than those of the wild type protein, respectively; Xu and Beckett 1996). These data suggested that these mutant proteins might leak bio-5′-AMP into the solvent where it would act as a promiscuous chemical biotinylation reagent in a proximity-dependent manner.

We report that a mutant BirA protein, R118G, acts as a promiscuous biotinylation reagent that shows proximity-dependence in a model system. Indeed, the R118G protein was found to efficiently biotinylate itself as well as a variety of proteins that normally lack this modification.

Acknowledgments

Notes

Article published online ahead of print. Article and publication date are at http://www.proteinscience.org/cgi/doi/10.1110/ps.04911804.