Benchmarking of proximity-dependent biotinylation enzymes across cellular compartments and time windows

Abstract

Proximity-dependent biotinylation has become a powerful approach for mapping protein interactions and subcellular organization in living cells. Although a growing number of engineered biotin ligases have been introduced, their performance has not been systematically evaluated across diverse cellular contexts. Here, we benchmark ten proximity ligases spanning three bacterial lineages using standardized proteomic workflows across multiple labeling durations, subcellular compartments, and two human cell types. While all enzymes efficiently detect proximal associations, they differ in labeling kinetics, background activity, and spatial specificity. TurboID exhibits the highest overall activity but generates substantial background in standard media. miniTurbo and ultraID support rapid, biotin-dependent labeling with low background, making them better suited for dynamic and time-resolved applications. However, miniTurbo showed aberrant mitochondrial localization with two cytoskeletal baits (VASP and PFN1). Across 15 diverse baits, ultraID consistently provides an excellent combination of specificity, efficiency, and spatial compatibility--including unique recovery of Golgi-resident glycosyltransferases. This study serves as a comparative resource, offering guidance for enzyme selection and experimental design in proximity proteomics.