Biology Atlas

Map / Assays

assay · intermediate · partly written

Proximity labelling (BioID, TurboID, APEX)

Proximity labelling fuses an enzyme to a bait protein in living cells so that it stamps a chemical tag on whatever lies within its reach, giving a neighbourhood list rather than a list of binding partners.

Also called: BioID, TurboID, APEX, APEX2, proximity-dependent biotinylation, proximity labeling

What it is

Proximity labelling attaches a short-lived chemical stamp to proteins near a bait inside living cells. Tagged proteins are pulled out and identified. The list says who was nearby, not who binds. 12

Reported, not yet replicated

What sets the radius more

Four things set how far the stamp reaches: the half-life of the reactive species, the diffusion of the probe, quenchers in the cell that destroy it, and time. The authors of a 2022 radius paper note that quenchers and crowding in cytosol and mitochondrial matrix could drastically influence spatial resolution, so a radius measured in one compartment cannot be generalised. Reviews and primary papers disagree on numbers, so treat any single radius as one protocol's contour. 13

Reported, not yet replicated

From BioID to TurboID more

TurboID was evolved over 29 rounds of yeast display from BioID, cutting the labelling window from 18 hours to 10 minutes. The follow-up study found TurboID labels without added biotin from the culture medium and its labelling appears to extend beyond the roughly 35 nm Nup107-160 complex. Their fix was inducible expression plus biotin-depleted serum. 24

Reported, not yet replicated

Question answered

Which proteins were within reach of a labelling enzyme fused to this bait, in a living cell, during the labelling window? It is a neighbourhood readout. 21

Reported, not yet replicated

Readout physical

Biotinylated proteins. An enzyme (or light-driven catalyst) fused to the bait converts a cell-permeable probe into a short-lived reactive species that attaches covalently to nearby proteins. The tagged proteins are pulled out with streptavidin, which grips biotin very tightly, and identified by mass spectrometry, then compared with a control. 12

Reported, not yet replicated

Not measured

Binding, direct contact, affinity or stoichiometry. A hit means "was within reach of the stamp during the window"; a non-hit is not a non-neighbour. The radius is a property of enzyme plus probe plus compartment plus labelling time plus your enrichment cut-off, not of the enzyme alone. In one peroxidase-based experiment the known interactor EGFR ranked 64th of 119 hits. 14

Reported, not yet replicated

Setting

Where in cascade

Sample prep

The bait is expressed as a fusion with the enzyme. Biotin (BioID, TurboID), biotin-phenol plus hydrogen peroxide (APEX) or a probe plus blue light (photocatalytic microMap) is added for the labelling window. Cells are lysed, biotinylated proteins are captured on streptavidin and analysed by mass spectrometry, compared with a control such as the enzyme fused to nothing or localised to the same compartment. 123

Reported, not yet replicated

Conditions

Labelling time: BioID about 15-18 hours; TurboID about 10 minutes in HEK 293T cells with 50 or 500 micromolar biotin. Radius figures conflict by source. A review table gives APEX 20 nm, BioID 10-20 nm, TurboID 5-10 nm and HRP under 200-300 nm. A photocatalytic carbene with a 2 nanosecond half-life has a theoretical radius of about 4 nm (measured 54 +/- 12 nm, equal to the length of the antibody tether of 57 nm), while a peroxidase phenoxyl radical with a low-millisecond half-life gave 269 +/- 41 nm on a protein-coated coverslip model, not in cells. Quenchers and crowding in the cell change the radius by compartment. TurboID labels from the biotin in ordinary culture medium and its labelling can extend beyond the roughly 35 nm Nup107-160 pore complex when biotin or time is added. 2134

Reported, not yet replicated

Throughput

One bait construct and one labelling experiment per sample set; no throughput figure was found in the sources opened. Class set to low as an estimate. 1

Our inference

Time per run

Labelling window: about 15-18 hours for BioID against about 10 minutes for TurboID (HEK 293T). Capture and mass spectrometry time are not given in the sources opened. 2

Reported, not yet replicated

Engineerable variants

BioID (BirA R118G)

A mutant biotin ligase that leaks activated biotin; the mutation weakens its grip on biotin-AMP about 440-fold and the biotin-AMP tags lysines. 4

Reported, not yet replicated

TurboID and miniTurbo

Evolved from BioID over 29 rounds. TurboID is 35 kD with 15 mutations from wild type; miniTurbo is 28 kD with 13 mutations, about 1.5-2-fold less active but with lower background without added biotin. May 2020 report instability of miniTurbo and early generations. 24

Reported, not yet replicated

APEX / APEX2 (engineered peroxidase)

Uses hydrogen peroxide and biotin-phenol to make a phenoxyl radical with a low-millisecond half-life; peroxide limits use in whole organisms. 13

Reported, not yet replicated

Photocatalytic microMap with variable radii

An iridium photocatalyst on an antibody and a diazirine probe under blue light makes a carbene with a 2 nanosecond half-life. Swapping to a PEG24 probe diffused about 2-fold more slowly in the text (1.6-fold in a caption) and shrank labelling clusters from 119 +/- 33 nm to 80 +/- 28 nm on the model surface. 1

Reported, not yet replicated

Use cases

Lamin A as bait: BioID versus TurboID

With lamin A as the bait, BioID gave 58 candidate interactors and TurboID 157 (criteria: found in all three replicates and more than 10-fold above control). TurboID's list had a smaller nuclear share (41% against 62% for BioID). The study is from the lab of the BioID developer, a competitor of the TurboID group. 4

Reported, not yet replicated

TurboID in whole animals

The founding paper showed labelling in living flies and worms, and also that ubiquitous TurboID without added biotin lowered fly viability and size and delayed worm development. 2

Reported, not yet replicated

Before after

Status quo before

BioID needed labelling of about 15-18 hours; in the TurboID paper the benchmark was BioID at 18 hours. 24

Reported, not yet replicated

What forced it

The authors set out to get the biology of BioID, which needs no hydrogen peroxide and can work in whole organisms, at the speed of peroxidase-based labelling. This motive is our reading of the paper's aim; the paper's own wording was not quoted. 2

Our inference

The trick

Directed evolution on yeast: an error-prone PCR library of about 1e7 variants, displayed on yeast and sorted by FACS for self-biotinylation, with the labelling time window cut from 18 hours to 10 minutes over 29 rounds, a cleavage step to exclude mutants that only tag themselves, and negative selections against use of the biotin in yeast medium. 2

Reported, not yet replicated

Why it worked

In HEK 293T cells TurboID made as much biotinylated product in 10 minutes as BioID did in 18 hours. The cost was basal labelling without added biotin and toxicity when expressed everywhere in flies and worms; the competitor study adds that more biotin and time bring in unrelated background proteins. 24

Reported, not yet replicated

Teach this

In one sentence

Proximity labelling paints a short-range chemical stamp on everything near your protein inside a living cell, so a hit means "was nearby", not "binds". 12

Reported, not yet replicated

The five-minute version

Fuse an enzyme to your protein of interest and give it a harmless probe. The enzyme turns the probe into a reactive molecule that diffuses away and sticks to the first things it meets. How far it goes depends on how long the reactive molecule lives and how fast cell chemistry destroys it, so the "radius" is a property of the protocol and the cell compartment as well as the enzyme. Pull out the tagged proteins with streptavidin and identify them by mass spectrometry. TurboID did this in 10 minutes instead of BioID's 18 hours, but it can label before you add biotin and its reach grows with time. 124

Reported, not yet replicated

What a listener will ask

Is a hit a binding partner?

Not necessarily. The radius is nanometres to tens of nanometres depending on source, and many unrelated proteins fall inside it. 13

Reported, not yet replicated

Why use it instead of AP-MS?

It labels in living cells before lysis, so it can capture weak or short-lived neighbours that washes would remove. 1

Our inference

What controls are needed?

A compartment-matched control, a no-biotin control and a check that the fusion localises and functions normally. 34

Reported, not yet replicated

The common wrong idea

Proximity labelling finds the proteins a bait binds.

It finds proteins within the labelling radius during the window; binding needs an orthogonal assay.

The one citation to hand over: 1

Still open here

Not yet written: Cost, Capital, ML role, Data availability.

Linked from

Sources

  1. Oakley JV, et al.. Radius measurement via super-resolution microscopy enables the development of a variable radii proximity labeling platform. PNAS (2022). paper; opened 2026-11-17 all records citing this
  2. Branon TC, et al.. Efficient proximity labeling in living cells and organisms with TurboID. Nature Biotechnology (2018). paper; opened 2026-11-17 all records citing this
  3. Ozmen B, et al.. Proximity labeling techniques for protein-protein interaction mapping in plants. Journal of Biological Chemistry (2025). paper; opened 2026-11-17 all records citing this
  4. May DG, et al.. Comparative Application of BioID and TurboID for Protein-Proximity Biotinylation. Cells (2020). paper; opened 2026-11-17 all records citing this
Where it sits on the map
History of this record (1)

This record has been changed 1 time, on the dates below. The note kept against each change is written for the people who maintain the atlas, in their own shorthand, so it is not shown here; it is in the record itself.

Throughput class
low
Emerging
no
Status
partly written
Last updated
2026-11-17
Record
content/assays/proximity-labelling.json