A TR-FRET reading tells you that a lanthanide donor and an acceptor dye ended up close together in the well. It does not tell you how far apart they were, it does not by itself tell you the two labelled molecules are bound to one another, and it does not guarantee the number came from energy transfer at all. Three separate pieces of machinery turn a proximity event into something you can read out of a crude, unwashed, compound-containing well: a donor whose emission outlives everything else in the sample, a collection window that opens after the flash has died away, and a two-channel ratio. Each fails in its own characteristic way, and knowing which one has failed is most of the work.
What the lanthanide donor buys
Selvin and Hearst built the case with a terbium chelate donor and tetramethylrhodamine acceptor, and reported that the results were consistent with a Förster theory of energy transfer provided the appropriate parameters are used. Four properties of that pair carried the improvement: the distance at which 50% energy transfer occurs is large, 65 Å; the donor lifetime is a single exponential and long, on the millisecond scale, making lifetime measurements facile and accurate; uncertainty in the orientation factor, which creates uncertainty in measured distances, is minimised by the donor's multiple electronic transitions and long lifetime; and the sensitised emission of the acceptor can be measured with little or no interfering background, yielding a greater than 25-fold improvement in signal-to-background over standard donor–acceptor pairs [1]. Selvin's later survey generalises the list: relative to conventional fluorophores, lanthanide donors offer farther measurable distances of approximately 100 Å with greater accuracy, insensitivity to incomplete labelling, and the ability to use generic relatively large labels when necessary [2].
Note what the orientation-factor argument does and does not say. It says the κ2 uncertainty that plagues organic donor–acceptor pairs is reduced, not eliminated, and the reason given is photophysical — multiple electronic transitions and a long lifetime — not geometric [1]. It is an argument about the donor. Nothing about it constrains how the acceptor is oriented on your peptide.
Donor choice is not neutral. In the HTRF implementation, europium cryptate was the first-generation donor, extended in 2008 by a second-generation terbium cryptate donor; terbium possesses different photophysical properties compared with europium, including increased quantum yield and a higher molar extinction coefficient. It is compatible with the same acceptor fluorophores used with europium and can additionally serve as donor to green-emitting fluors, because it has multiple emission peaks including one at 490 nm, and terbium assays read on the same instruments as europium assays [3]. The practical consequence is that a terbium donor widens the acceptor palette rather than replacing an existing pair.
The time gate, and the background it does not touch
Every immunoassay carries a signal in the absence of analyte, and that background limits both sensitivity and dynamic range. This is especially so for homogeneous formats and for surface measurements on tissue sections and membranes, which typically show high background because of sample autofluorescence. Lanthanide chelate labels attack this directly: their exceedingly long-lived luminescence, in comparison with conventional fluorophores, allows the short-lived background interferences to be removed by time-gated acquisition, delivering greater assay sensitivity and a broader dynamic range [4]. Mathis frames the same architecture as a double discrimination of the emitted signal, through spectral and temporal selectivity [5] — two independent filters, one on colour and one on time.
The gate removes emission that has already decayed when the window opens. It does nothing about anything that survives the delay, and nothing at all about interference that acts on the donor itself. Hanley and colleagues found that salicylic acids, an extremely common compound subclass in screening libraries, interfere with TR-FRET assays; their data suggest that interaction of the salicylate with the cryptand-ligated europium FRET donor is responsible for the change in assay signal, although the precise mechanism of interference was not established [6]. That is a donor-directed artefact, and it is invisible to the gate by construction.
The broader interference catalogue for this whole assay class is worth holding in mind when a hit list comes back. Homogeneous proximity assays are susceptible to signal attenuation through quenching, inner-filter effects and light scattering; to signal emission through auto-fluorescence; and to disruption of affinity capture components such as affinity tags and antibodies. They are also susceptible to more generalised compound-mediated interferences such as nonspecific reactivity and aggregation, and the response is counter-screens designed to classify which of these is operating [7].
The ratio, and what it is licensed to correct
The plate reader collects two emission channels from the same flash in the same well and reports their ratio. Mathis attributes the media tolerance of the cryptate assays jointly to the cage structure of the complex, ion pairing around europium, and double-wavelength detection, which together fully shield the assay from perturbations of media [5]. Degorce and colleagues put the same claim in operational terms: buffer and media interference is dramatically reduced by dual-wavelength detection, and the final signal is proportional to the extent of product formation [3].
Read the scope of that carefully. The ratio cancels influences that scale both channels by the same factor — lamp output drift, meniscus and volume error, broadly absorbing coloured content. It cannot cancel anything wavelength-selective, and it cannot cancel anything acting on only one of the two labels, which is exactly the class the salicylate result falls into [6]. This is why the donor channel is diagnostic rather than merely a denominator: a compound that drops the raw donor counts and the raw acceptor counts together is doing something optical, while one that moves the ratio without moving the donor channel is doing something to the acceptor or to the complex. Donor-only and acceptor-only wells at the same compound concentration are the arms that separate the cases.
Why r−6 makes this proximity, not distance
Transfer efficiency falls away with the sixth power of the distance between the donor and acceptor [8], and the phenomenon generates signals sensitive to molecular conformation, association and separation in the 1–10 nm range [9]. The Förster distance sets where in that range you are working, and lanthanide pairs sit at the far end of it. For the terbium chelate and tetramethylrhodamine pair the 50% point is 65 Å [1]; Bazin and colleagues report an exceptionally large Förster radius of 9 nm for a lanthanide cryptate donor with a cross-linked allophycocyanin acceptor [10].
Run the arithmetic on a 9 nm Förster radius and the reason TR-FRET is a proximity assay becomes obvious. At half the Förster distance, roughly 4.5 nm, efficiency is about 98%; at twice it, 18 nm, efficiency is under 2%. The observable is steep only across a narrow band centred on R0, and it is flat on both sides. Two antibodies bridged by a short peptide, a cryptate on a SNAP tag and a dye on a bound ligand, a donor and acceptor on the two ends of a flexible linker — these can each sit inside the flat, near-saturated region, though where any particular construct falls has to be established rather than assumed. A change in the reading reports that the pair crossed into or out of that band. It does not report a displacement within it.
Where a genuine distance is wanted, the measurement is done differently and the error bars are stated. Burmeister Getz and colleagues measured an energy transfer efficiency of 21.2 ± 1.9% between Cys707 in the catalytic domain and Cys108 in the regulatory light chain domain of the myosin head in the complex without nucleotide or actin, corresponding to a distance of 73 Å and consistent with the crystal structure. Binding to actin decreased the efficiency by 4.5 ± 1.0%, indicating a conformational change involving relative rotation and/or translation of Cys707 with respect to the light chain domain [11]. The same study contains the warning that matters for every proximity reading: addition of ADP also altered the energy transfer efficiency, likely through a rotation of the probe attached to Cys707 [11]. A probe that reorients changes the signal without anything moving.
Labelling: stoichiometry, chemistry, and which observable inherits the tolerance
The insensitivity to incomplete labelling that makes lanthanide donors attractive is attached to a specific observable. Selvin and Hearst reported measurement of the sensitised emission lifetime, and it is that measurement which is completely insensitive to total concentration and incomplete labelling [1]; Selvin's survey lists insensitivity to incomplete labelling among the general advantages of lanthanide donors [2]. A plate-reader intensity ratio is not a lifetime measurement, and the tolerance does not transfer to it without being demonstrated. In a two-channel intensity format, donors that never acquire a partner still contribute to the denominator, and the labelled fraction of each reagent is therefore a parameter you have to control, not one the method absorbs for you.
The conjugation chemistry can also change the donor. Chen and Selvin synthesised thiol-reactive lanthanide complexes built from a DTPA chelate joined through one amide bond to a carbostyril 124 chromophore and through a second to a maleimide, bromoacetamide or pyridyldithio group, and attached them site-specifically to protein and DNA targets. They report that the compounds coordinate a surprisingly large number of ligation sites of terbium when a hydrazide spacer is used between the chelate and the thiol-reactive moiety, although this extra ligation can cause quenching when europium is used [12]. The linker is part of the photophysics, and its effect there was lanthanide-specific — a spacer validated on terbium is not thereby validated on europium.
The usual escape from labelling the analyte at all is to label something generic that binds it. Bazin and colleagues built their labelling strategies around fully indirect cassette formats, with the emphasis on that indirection for high-throughput work [10]. In receptor formats the same logic appears as tag chemistry: SNAP-tag-labelled receptors, or fluorescently labelled antibodies directed against an epitope engineered into a Flag-labelled receptor's N terminus [15].
Sandwich formats and the hook
A sandwich TR-FRET assay puts the donor and acceptor on two binders that recognise separate epitopes of one analyte, so that only the ternary complex generates transfer. Albrecht and colleagues built such an assay for the Aβ peptide using a biotinylated monoclonal antibody against residues 17–24 complexed with a europium cryptate–streptavidin conjugate as donor arm, and a polyclonal antibody against the N terminus complexed with an allophycocyanin–anti-rabbit conjugate as acceptor arm; binding of both complexes brought the pair into close proximity and produced a FRET signal on excitation at 337 nm, read at 665 nm. Detection was possible at approximately 1 nM, with signal-to-background ratios exceeding 5 and Z′ factors around 0.7 achieved at optimised conditions in a 384-well format, and 56,913 compounds were screened on it [13].
The characteristic failure of this geometry is the hook. Fernando and Wilson studied it in the one-step sandwich immunoassay using three model analytes chosen to differ in epitope structure — a monomeric protein, its dimer carrying a discrete number of repeating epitopes, and ferritin with multiple epitopes — and found that the one-step format suffers the hook effect irrespective of the analyte characteristics, and that the effect is dependent primarily on the analyte concentration. Both their experimental and theoretical data showed several instances where high capacity solid-phase antibodies can effectively shift the hook to relatively higher analyte concentrations [14]. That study is a solid-phase, one-step heterogeneous assay, so the capacity remedy is stated in terms of solid-phase antibody and is not a measured result about homogeneous TR-FRET; what transfers is the shape of the problem, which follows from the one-step architecture rather than from the detection chemistry.
Operationally, the hook means a single ratio can correspond to two analyte concentrations, one on each limb of the curve, and no amount of replicate reading resolves the ambiguity. A dilution series does: on the ascending limb the signal falls with dilution, on the descending limb it rises.
Labelled-ligand binding assays
For peptide ligands the useful configuration is a labelled ligand against a labelled receptor. Emami-Nemini and colleagues give a general procedure using terbium-labelled fluorescent receptor ligands and TR-FRET, worked through for a class B GPCR and its fluorescently labelled 34-amino-acid peptide ligand, and note that radioligand experiments yield excellent quantitative data but have low temporal and spatial resolution. They describe three options: a time-resolved fluorescence separation option that works on native receptors but requires separation of bound and unbound ligand; a TR-FRET option using SNAP-tag-labelled receptors for high-throughput screening; and a TR-FRET option using fluorescently labelled antibodies against an engineered N-terminal epitope. The options are presented as standard procedures with very high signal-to-background ratios for characterising ligands and receptors in living cells and in cell membranes by plate-reader measurement [15].
The specificity controls are the assay, not an appendix to it. Leifert and colleagues built a cell-free, homogeneous mix-and-read TR-FRET assay with a terbium cryptate chelate donor spectrally matched to an Alexa546 acceptor, each conjugated to a separate binding partner, and observed high-affinity interactions as a rapid increase of more than fivefold in the FRET signal. The specificity of those interactions was demonstrated using denaturing or competitive conditions, which caused reductions of 50–85% in fluorescence, indicating that the labelled proteins were no longer in close proximity [16]. A fivefold window with no denaturing and no competition arm is an uninterpreted number.
Where BRET and polarization sit relative to this
BRET replaces the optically excited donor with a bioluminescent luciferase genetically fused to one candidate protein, with a green fluorescent protein mutant fused to the other; interaction can bring the two close enough for resonance energy transfer, changing the colour of the bioluminescent emission. Xu and colleagues used it to show that the cyanobacterial clock protein KaiB forms homodimers, and proposed it as particularly useful for testing protein interactions within native cells, especially with integral membrane proteins or proteins targeted to specific organelles [17]. The trade is structural: BRET as described requires both partners to be genetically encoded fusions, where the TR-FRET cassette and tag formats let one partner stay chemically labelled or unmodified [10, 15].
Fluorescence polarization, covered separately in this library, sits on a different physical variable. It rests on the observation that a fluorescently labelled molecule excited by polarized light emits light with a degree of polarization inversely proportional to the rate of molecular rotation, a property used to measure the interaction of a small labelled ligand with a larger protein [18]. That is one label and a required size contrast between free and bound tracer. TR-FRET is two labels and no size requirement at all, since the variable is separation rather than tumbling rate. When the tracer is not much smaller than its partner, polarization loses the rotational contrast it depends on and TR-FRET does not; when only one binding partner can be labelled without wrecking it, the reverse is true.