Can singlet oxygen label both sides of transmembrane proteins in vivo?

Why proximity labelling usually stays on one side

Proximity labelling tags proteins that sit close to a chosen bait, so that they can be enriched and identified by mass spectrometry. For a transmembrane protein, the interesting partners are split between the outside and the inside of the cell, and the two sets are regulated separately. The paper's starting problem is practical: most reactive species generated by enzymes or photocatalysts used for proximity labelling are membrane-impermeable or poorly permeable, so labelling happens mainly on the same side as the catalyst.

That limitation matters most for receptors whose N- and C-termini both face the extracellular space. Such proteins offer no cytoplasmic domain where an enzyme could be fused, so the intracellular side is hard to reach with conventional proximity labelling. The authors argue that the barrier could be removed if the reactive species itself could cross the lipid bilayer, but note that this had not been thoroughly validated.

Singlet oxygen crosses while the biotin probe enters

tmPhoxID separates two steps that other methods couple. First, a photosensitiser anchored to the target protein is irradiated and generates singlet oxygen, a reactive oxygen species. Singlet oxygen is membrane-permeable, and the paper estimates its diffusion distance at tens to about one hundred nanometres, which is enough to cross a lipid bilayer while remaining short enough to keep labelling local. Second, proteins oxidised by singlet oxygen are tagged by a nucleophilic reagent carrying biotin or a clickable group; this reagent is chosen to be cell-permeable.

The design therefore does not require putting a catalyst or enzyme inside the cell. The authors present this as an advantage: the intracellular environment stays intact, and the labelling reagent can be redesigned independently of the photosensitiser. They also note that the photosensitiser can be attached through noncovalent small-molecule ligands or nanobodies, or through covalent strategies such as ligand-directed chemistry or HaloTag when in vivo residence is short.

A cytoplasmic sensitiser labels the cell surface

The first test reversed the usual geometry to check that singlet oxygen really crosses the membrane. HeLa cells expressed HaloTag on the cytoplasmic side via a fusion to the inner-leaflet protein Lyn, and a cell-permeable acetylated dibromofluorescein-Halo ligand was used to attach the DBF photosensitiser. After irradiation in the presence of a hydrazide biotin reagent, cell-surface biotin was detected with cell-impermeable streptavidin-Alexa Fluor 647 in live cells.

Surface labelling appeared only when DBF was anchored and light was applied; it was absent in unirradiated samples and in cells lacking DBF. Because singlet oxygen generation occurs through short-range Dexter energy transfer, the authors interpret this as evidence that intracellularly generated singlet oxygen reached the outer surface. Under these conditions, lipid oxidation, loss of membrane integrity and reduced cell viability were all reported as negligible, which argues against oxidative membrane damage explaining the signal.

Labelling tracks a rapamycin-induced recruitment

A second model tested whether tmPhoxID reports changes in the intracellular interactome rather than a fixed background. HaloTag was fused to the N-terminus of EGFR to anchor MBF on the extracellular side, and an FRB domain was fused to the cytoplasmic C-terminus. An mCherry-FKBP12 fusion was expressed in the cytosol, and propargylamine was used as the cell-permeable labelling reagent.

In the basal state, labelling of mCherry-FKBP12 was minimal because the protein was not enriched at the plasma membrane. After rapamycin treatment recruited mCherry-FKBP12 to the membrane, its labelling signal rose three- to four-fold relative to the untreated control. This is a quantitative comparison within the same model system, and it supports the claim that the method can follow composition changes at the inner leaflet.

Receptor-specific interactomes in the mouse brain

The in vivo workflow anchored the photosensitiser to a receptor, injected the labelling reagent, irradiated the brain for 10 minutes at 0.96 W/cm2, and enriched labelled proteins on NeutrAvidin beads before NanoLC-MS/MS. Targeting GRID2, GABA-A receptors and GRM1 in the living mouse brain yielded membrane-proximal intracellular interactomes for each receptor. GRID2 and GRM1 are coexpressed but not colocalised in the cerebellum, and the datasets for the two receptors were distinct, which the authors read as evidence of receptor specificity and spatial precision within subsynaptic microenvironments.

For GABA-A receptors, hits were defined as proteins with a fold change of at least 1.5, a P-value below 0.05 and at least two unique peptides, using three mice per experimental condition and two technical replicates. Gene Ontology analysis showed stronger enrichment of the GABA receptor complex term in the GABA-A receptor dataset, while the GRID2 dataset was more enriched for the ionotropic glutamate receptor complex term. Proteins associated with glutamatergic synapses were more enriched in the GRID2 dataset, and GABAergic synapse proteins in the GABA-A receptor dataset. These are comparisons between two receptor-targeted datasets, not a direct measure of absolute abundance.

Transsynaptic nanocolumns and a new partner

Network analysis of the GRID2 dataset indicated that tmPhoxID can characterise native components of transsynaptic nanocolumns at parallel fibre–Purkinje cell synapses in genetically unedited animals. The authors contrast this with classical synaptosome fractionation, citing reduced contamination, minimal perturbation of protein complexes and clearer distinction between synapse types. They also caution that some identified hits are likely previously unannotated components and would need future validation.

The GABA-A receptor work produced the study's most concrete biological claim: a previously uncharacterised GABA-A receptor–CAMKV interaction. The evidence combines co-immunoprecipitation in mouse cortex and hippocampus lysates, co-IP mass spectrometry of hippocampal samples from three mice, proximity ligation assays in mouse hippocampal slices, and proximity ligation assays in postmortem human hippocampus slices. A ColabFold-predicted model places the CAMKV palmitoylation site adjacent to the inner-leaflet-facing side of the GABA-A receptor transmembrane helices, which the authors present as a structural rationale rather than a demonstrated mechanism.

Where the method stops being informative

The authors state that the detailed reaction mechanism is not fully resolved. Their preferred interpretation is that singlet oxygen made outside the cell diffuses across the membrane and oxidises intracellular proteins near the receptor, which are then labelled by the membrane-permeable biotin probe. They argue against lipid oxidation and membrane disruption as major contributors, and against relay through membrane-spanning proteins for neurotransmitter receptors, but these remain interpretations supported by the evidence rather than a complete mechanism.

Two practical limits follow from the design. Because singlet oxygen diffuses only tens to about one hundred nanometres, the method maps membrane-proximal rather than broad intracellular interactomes; wider coverage would require directly targeting the photocatalyst inside the cell, at the cost of engineered systems. Efficiency also depends on the receptor being present at the cell surface for photosensitiser anchoring, so low-abundance or rapidly recycling receptors are expected to give weaker labelling. The authors further note that they cannot determine how much receptor internalisation contributes to the observed results, and that suboptimal light intensity or irradiation duration could compromise data interpretation.

What this proof-of-concept changes

The paper's distinctive contribution is not a new receptor biology result alone but a working geometry: an extracellularly anchored sensitiser can report on the intracellular side without introducing a catalyst into the cytoplasm. That makes tmPhoxID most valuable for transmembrane proteins whose cytoplasmic face is hard to engineer, including receptors with both termini outside the cell. The GABA-A receptor–CAMKV finding is the clearest demonstration that this geometry can surface a genuinely new interaction, and it is supported by co-IP, proximity ligation in mouse and human tissue, and a predicted structural model.

The right way to read the study is as proof of concept with receptor-specific comparisons. The GRID2, GABA-A receptor and GRM1 datasets show that the method can distinguish neighbouring receptor environments, and the rapamycin model shows it can follow a recruitment event. What it does not yet provide is a resolved mechanism or a way to separate surface from internalised receptor pools. Readers weighing tmPhoxID against other proximity labelling approaches should focus on that trade-off: access to the intracellular side of hard-to-engineer receptors, in exchange for a labelling radius limited to the membrane-proximal zone.

Sources used in this answer