Efficient proximity labeling in living cells and organisms with TurboID

Abstract

Protein interaction networks and protein compartmentalization underlie all signaling and regulatory processes in cells. Enzyme-catalyzed proximity labeling (PL) has emerged as a new approach to study the spatial and interaction characteristics of proteins in living cells. However, current PL methods require over 18 h of labeling time or utilize chemicals with limited cell permeability or high toxicity. We used yeast display-based directed evolution to engineer two promiscuous mutants of biotin ligase, TurboID and miniTurbo, which catalyze PL with much greater efficiency than BioID or BioID2, and enable 10-min PL in cells with non-toxic and easily deliverable biotin. Furthermore, TurboID extends biotin-based PL to flies and worms.

Figure 1. Directed evolution of TurboID

(a) Proximity-dependent biotinylation catalyzed by promiscuous biotin ligases. Ligases catalyze the formation of biotin-5′-AMP anhydride, which diffuses out of the active site to biotinylate proximal endogenous proteins on nucleophilic residues such as lysine. (b) Yeast display-based selection scheme. A >10⁷ library of ligase variants is displayed on the yeast surface as a fusion to mating protein Aga2p. All ligases have a C-terminal myc epitope tag. Biotin and ATP are added to the yeast library for between 10 minutes and 24 hours. Ligase-catalyzed promiscuous biotinylation is detected by staining with streptavidin-phycoerythrin and ligase expression is detected by staining with anti-myc antibody. Two-dimensional FACS sorting enables enrichment of cells displaying a high ratio of streptavidin to myc staining. (c) Tyramide signal amplification (TSA) improves biotin detection sensitivity on the yeast surface. In the top row, the three indicated yeast samples (G1 is the winning ligase mutant from the first generation of evolution) were labeled with exogenous biotin for 18 hours then stained for FACS as in (b). The y-axis shows biotinylation extent, and the x-axis quantifies ligase expression level. In the second row, after 18 hours of biotin incubation, yeast were stained with streptavidin-HRP, reacted with biotin-phenol^, to create additional biotinylation sites, then stained with streptavidin-phycoerythrin and anti-myc antibody before FACS. The third row omits biotin. Percentage of cells in upper right quadrant (Q2/(Q2+Q4)) shown in top right of each graph. This experiment was performed once, but each yeast sample has been analyzed under identical conditions at least twice in separate experiments with similar results. (d) E. coli biotin ligase structure (PDB: 2EWN) with sites mutated in TurboID (left) and miniTurbo (right) shown in red. The N-terminal domain (aa1-63) is also removed from miniTurbo. A non-hydrolyzable analog of biotin-5′-AMP, biotinol-5′-AMP, is shown in yellow stick. (e) FACS plots summarizing progress of directed evolution. G1-G3 are the winning clones from generations 1-3 of directed evolution. G3Δ has its N-terminal domain (aa1-63) deleted. Omit biotin samples were grown in biotin-deficient media (see Methods) for the entire induction period (~18-24 hr). This experiment was performed twice with similar results, except G3Δ omit biotin, which was performed once. (f) Comparison of ligase variants in the HEK cytosol showing that TurboID and miniTurbo are much more active than BioID, as well as the starting template and various intermediate clones from the evolution. Indicated ligases were expressed as NES (nuclear export signal) fusions in the HEK cytosol. 50 μM exogenous biotin was added for 3 hours, then whole cell lysates were analyzed by streptavidin blotting. Ligase expression detected by anti-V5 blotting. U, untransfected. S, BirA-R118S. Asterisks indicate ligase self-biotinylation. BioID labeling for 18 hours (50 μM biotin) shown for comparison in the last lane. This experiment was performed twice with similar results. (g) Quantitation of streptavidin blot data in (f) and from a 30 minute labeling experiment shown in Supplementary Figure 4b. Quantitation excludes self-biotinylation band. Sum intensity of each lane is divided by the sum intensity of the ligase expression band; ratios are normalized to that of BioID/18 hours, which is set to 1.0. Grey dots indicate quantitation of signal intensity from each replicate, colored bars indicate mean signal intensity calculated from the two replicates.

Figure 2. Characterization of TurboID and miniTurbo in mammalian cells

(a) Comparison of TurboID and miniTurbo to three other promiscuous ligases (BioID, BioID2, and BASU) in the cytosol of HEK 293T cells. Here, 500 μM exogenous biotin was used for labeling, whereas 50 μM was used in Supplementary Figure 6c-e. Streptavidin-HRP blotting detects promiscuously biotinylated proteins, and anti-V5 blotting detects ligase expression. U, untransfected. Asterisks denote ligase self-biotinylation bands. This experiment was performed twice with similar results. (b) Quantitation of experiment in (a). For shorter timepoints (-biotin and 10 min), we used a longer-exposure image of the same blot, shown in Supplementary Figure 6a; for longer timepoints (1, 6, 18 hr), we used a shorter-exposure image of the blot in (a), shown in Supplementary Figure 6b. Quantitation performed as in Figure 1g. Grey dots indicate quantitation of signal intensity from each replicate, colored bars indicate mean signal intensity calculated from the two replicates. (c) Comparison of promiscuous ligases in multiple HEK organelles. Each ligase was fused to a peptide targeting sequence (see Supplementary Table 8) directing them to the locations indicated in the scheme at right. BioID samples were treated with 50 μM biotin for 18 hours. TurboID and miniTurbo samples were labeled for 10 minutes with 50 (+) or 500 (++) μM biotin. U, untransfected. Asterisks denote ligase self-biotinylation. This experiment was performed five times for nuclear constructs, three for mitochondrial constructs, four times for ER membrane constructs, and twice for ER lumen constructs with similar results. (d) Mass spectrometry-based proteomic experiment comparing TurboID and BioID on the ER membrane (ERM), facing cytosol. Experimental design and labeling conditions. Ligase fusion constructs were stably expressed in HEK 239T. BioID samples were treated with 50 μM biotin for 18 hours, while TurboID samples were treated with 500 μM biotin for 10 minutes or 1 hr. After labeling, cells were lysed and biotinylated proteins were enriched with streptavidin beads, digested to peptides, and conjugated to TMT (tandem mass tag) labels. All 11 samples were then combined and analyzed by LC-MS/MS. This experiment was performed once with two replicates per condition. (e) Specificity analysis for proteomic datasets derived from experiment in (d). Size of each ERM proteome at top. Bars show percentage of each proteome with prior secretory pathway annotation, according to GOCC, Phobius, human protein atlas, human plasma proteome database, and literature (see Methods and Supplementary Table 2 Tab 4). (f) Same as (e), except for each ERM proteome, we analyze the subset with ER, Golgi, or plasma membrane annotation. Annotations from GOCC were assigned in the priority order: ER>Golgi>plasma membrane (see Methods and Supplementary Table 2 Tab 5). (g) Breakdown of ER proteins enriched by TurboID and BioID, by transmembrane or soluble. Soluble proteins were further divided into luminal or cytosol-facing. Annotations obtained from GOCC, UniProt, TMHMM, and literature (see Methods and Supplementary Table 2 Tab 6). (h) Characterization of nuclear and mitochondrial matrix proteomes obtained via BioID (18 hour), TurboID (10 min), and miniTurbo (10 min)-catalyzed labeling. Proteome sizes across top. Bars show fraction of each nuclear (left) or mitochondrial (right) proteome with prior nuclear or mitochondrial annotation, according to GOCC, MitoCarta, or literature (see Methods and Supplementary Table 3 Tab 1, Supplementary Table 4, Tab 1). Design of proteomic experiment shown in Supplementary Figure 10a, proteomic lists in Supplementary Tables 6-7; further analysis of proteome data in Supplementary Figure 10.

Figure 3. TurboID and miniTurbo in flies, worms, and other species

(a) Comparison of ligases in yeast. EBY100 S. cerevisiae expressing BioID, TurboID, or miniTurbo in the cytosol were treated with 50 μM biotin for 18 hours. Whole cell lysates were blotted with streptavidin-HRP to visualize biotinylated proteins, and anti-V5 antibody to visualize ligase expression. U, untransfected. Asterisks denote ligase self-biotinylation. Bands in untransfected lane are endogenous naturally-biotinylated proteins. This experiment was performed twice with similar results. (b) Comparison of ligases in E. coli. Ligases, fused at their N-terminal ends to His6-maltose binding protein, were expressed in the cytosol of BL21 E. coli and 50 μM exogenous biotin was added for 18 hours. Whole cell lysates were analyzed as in (a). This experiment was performed twice with similar results. (c) – (g) Comparison of ligases in flies. (c) Scheme for tissue-specific expression of ligases in the wing disc of D. melanogaster. ptc-Gal4 induces ligase expression in a strip of cells within the wing imaginal disc that borders the anterior/posterior compartments. (d) Imaging of larval wing discs after 5 days of growth on biotin-containing food. Biotinylated proteins are detected by staining with streptavidin-AlexaFluor555, and ligase expression is detected by anti-V5 staining. Panels show the pouch region of the wing disc, indicated by the dashed line in (c). Scale bar, 40 μm. Each experimental condition has at least three technical replicates; one representative image is shown. This experiment was independently repeated two times with similar results. (e) Quantitation of streptavidin-AlexaFluor555 signal intensities in (d). Error bars, s.e.m. Average fold-change shown as text above bars. Sample size values (n) from left column to right: 5, 6, 3. (f) Scheme for ubiquitous expression of ligases in flies, at all developmental timepoints, via the act-Gal4 driver. (g) Western blotting of fly lysates prepared as in (f). Biotinylated proteins detected by blotting with streptavidin-HRP, ligase expression detected by anti-V5 blotting. In control sample, act-Gal4 drives expression of UAS-luciferase. Bands in control lanes are endogenous naturally-biotinylated proteins. This experiment was performed twice with similar results. (h) – (k) Comparison of ligases in worms. (h) Scheme for tissue-specific expression of ligases in C. elegans intestine via ges-1p promoter. Transgenic strains are fed either biotin-producing E. coli OP50 (biotin+), or biotin-auxotrophic E. coli MG1655bioB:kan (biotin-). Promoter ges-1p drives ligase expression approximately 150 minutes after the first cell cleavage. (i) Adult worms prepared as in (h) were shifted to 25°C for one generation, then lysed and analyzed by Western blotting. Control worms (N2) do not express ligase. Anti-HA antibody detects ligase expression. Streptavidin-IRDye detects biotinylated proteins. This experiment was performed five times (n = 5). In biotin+ conditions, BioID biotinylation activity was undetectable and TurboID gave robust biotinylation signal (n = 5/5). Despite high activity detected by immunofluorescence in embryos, we only detected some low level of biotinylation by miniTurbo in adults (n = 2/5), likely due to its low expression levels. (j) Representative images of bean stage worm embryos (stage 1) from (h). See Supplementary Figure 15a for representative images of comma stage worm embryos (stage 2). Embryos were fixed and stained with streptavidin-AF488 to detect biotinylated proteins, and anti-HA antibody to detect ligase expression. Intestine is outlined by a white dotted line. Scale bar, 10 μm. Quantitation of multiple replicates shown in (k). (k) Quantitation of streptavidin-AF488 signal acquired from IF staining of embryonic stages 1 and 2 shown in (j) and Supplementary Figure 15a. Mean streptavidin pixel intensities for each embryo assessed are plotted for BioID (B), TurboID (T), and miniTurbo (mT). Two independent transgenic lines for BioID and TurboID and one for miniTurbo were assessed. Number of embryos imaged (n) from left to right: 26, 18, 11, 16, 25, 8, 19, 23, 14, 14, 23, 9. Statistical significance via Mann-Whitney U test (two-sided). ***p ≤ 0.0001, **p ≤ 0.001, *p ≤ 0.01. Pink asterisks indicate significance of pairwise comparisons between biotin- and corresponding biotin+ treated embryos. Mean (reported in Supplementary Figure 15b) is shown as a black horizontal line for each condition, and error bars indicate s.e.m. Note that the streptavidin-AF488 pixel intensities for miniTurbo are an underrepresentation of the signal as camera exposure settings were lowered to avoid pixel saturation (see Methods). See Supplementary Figure 15 for more details.