Considerations prior to experiments

Note that there is no “single lifetime” that describes the decay of Trp fluorescence. Already a single tryptophan residue in a homogeneous bulk solvent exhibits a multi-exponential fluorescence decay. The “rotamer hypothesis” interpreted this behavior as a superposition of the fluorescence of Trp residues in the three stable rotameric structures around the C _α-C _β bond, which cannot freely interconvert during the nanosecond lifetime (Gudgin et al., Reference Gudgin, Lopez-Delgado and Ware1981; Swaminathan et al., Reference Swaminathan, Krishnamoorthy and Periasamy1994). However, only one rotamer has been recorded for most proteins despite great improvements in the resolution of most methods for structural analysis. This gave rise to the alternative explanation of these two characteristic lifetimes in proteins as a result of heterogeneity of protein structure and, hence, the direct environment of different Trps. In some proteins, a third, usually long lifetime over 6 ns, is commonly accepted to be a consequence of the interaction of Trp with its surroundings. Chen et al. explained four or more lifetimes of proteins in solution in terms of the heterogeneity of the sample or/and a contribution of parallel fluorescence of tyrosine (Chen et al., Reference Chen, Toptygin, Brand and King2008).

The dramatic improvement of the time resolution and data volume achieved by TCSPC instruments over the years has increased the number of lifetimes that can potentially be distinguished upon fitting one or more fluorescence decays in a meaningful manner. At the same time, it should be emphasized that sample heterogeneity and non-exponential decay components will generally cause lower reduced χ ² values as the number of adjustable parameters increases – overparameterization will improve the fit but render individual parameters meaningless. Furthermore, technical artefacts can resemble additional lifetimes. An example is stray light of a very short “apparent lifetime” interpreted as a short lifetime of Trp. Especially if the maximum of the decay-associated spectra (DAS) of the short-lived component is located at the first emission wavelength of the emission spectrum, the contribution of stray light should be checked either by a reduction of the excitation wavelength (increase in difference between excitation and emission wavelength) or/and a filter should be included in the pathway.

While quenching experiments are straightforward, there are still a couple of considerations that have to be taken into account. Usually, time-resolved measurements are less affected by artefacts than steady-state measurements. For example, the otherwise crucial inner filter effect has almost no impact on the lifetime measurement (Amaro et al., Reference Amaro, Šachl, Jurkiewicz, Coutinho, Prieto and Hof2014). Yet, for the highest concentration of the quencher, the fluorescence signal may become so small that it interferes with the very short lifetime of the Raman or Rayleigh light or fluorescence of impurities in the sample. Therefore, it is highly advisable to examine the emission spectra of Trp at high concentrations of the quencher in detail. For an overview of further pitfalls, see Table 2.

Table 2.

Overview of the most common pitfalls of measuring fluorescence lifetimes and amplitudes

Note: See Lakowicz (Reference Lakowicz2006) for background.

Assessing binding from static quenching

Quenching experiments may provide information about the binding of a quenching substance to Trp. Quenching has been described by the Stern-Volmer equation (Stern, Reference Stern1919)

(1) $$ \frac{F_0}{F}=1+{K}_{SV}\left[Q\right] $$

where F ₀ is the intensity in the absence of the quencher, F is the intensity at quencher concentration [Q], and K _SV is the Stern–Volmer quenching constant. The fluorescence intensity, F, is the product of the lifetime τ and the preexponential factor (amplitude) A or, for a multiexponential decay,

(2) $$ F=\sum \limits_i{A}_i{\tau}_i $$

Hence, F ₀/F is affected by changes in amplitude or in lifetime, referred to as static and dynamic quenching.

Purely static quenching is, exclusively, a result of an association of quencher Q and dye D in the ground state, with an association constant K _a defined by the law of mass action:

(3) $$ {K}_a=\frac{\left[D\right]\left[Q\right]}{\left[ DQ\right]} $$

where D and Q represent free dye and free quencher, and DQ dye-quencher associates. Since only the unbound dye fluoresces, binding does not affect the lifetime but only the amplitude A (subscripts 0 refer to the absence of the quencher). In a homogeneous system where all i lifetime components (see Eq. (2)) are quenched by the same factor A_i/A _0i  = const., we obtain

(4) $$ {\left.\frac{F_0}{F}\right|}_{static}=\frac{A_{0i}}{A_i}=\frac{\left[{D}_0\right]}{\left[D\right]} $$

where [D ₀] = [D] + [DQ] is the total dye concentration.

Comparison of Eqs. (1) and (3), (4) reveals that as long as most quencher remains free, [Q] ≈ [Q] + [DQ] ≡ [Q ₀], the Stern–Volmer constant is just the association constant of the complex, K _a = K _SV.

The direct interpretation of K _SV in terms of K _a requires the confirmation that first, lifetimes are not or only little changed (τ ₀/τ < < F ₀/F) (Ding et al., Reference Ding, Liu, Li, Zhang and Sun2010; Palacios-Ortega et al., Reference Palacios-Ortega, Amigot-Sánchez, García-Montoya, Gorše, Heras-Márquez, García-Linares, Martínez-del-Pozo and Slotte2022; Tayeh et al., Reference Tayeh, Rungassamy and Albani2009) and that affinity is not too high to compromise the approximation [Q] ≈ [Q ₀] (see above and Castanho and Prieto, Reference Castanho and Prieto1998). Note that the concept of static quenching envisages the direct binding of the quencher to the fluorophore – this situation is different from the binding of a drug to a protein site at some finite distance from one (or even more than one) fluorophore. This and related problems give rise to numerous serious pitfalls for classic protein binding studies, interpreting K _SV in terms of K _a (van de Weert and Schönbeck, Reference van de Weert and Schönbeck2024). Such systems might better be considered based on an – also quite non-trivial – FRET approach as addressed in the following sub-section.

Monitoring accessibility and dynamics by dynamic quenching

Dynamic quenching occurs upon a collision of excited Trp with the quencher. Since it takes a finite time for the quencher to hit, it does not affect the amplitude of the fluorescence, A = A ₀. Instead, it reduces the lifetime by adding another route to the ground state. The decay, F(t), becomes:

(5) $$ F(t)=\left[\sum \limits_i{A}_i{e}^{-{k}_{Fi}t}\right]{e}^{-{k}_q\left[Q\right]t} $$

where k_Fi = 1/τ_i are the rate constants of the fluorescence (expressed by i exponentials) and k _q the bimolecular quenching rate constant. For dynamic, i.e., collisional quenching by quencher in solution, k _q cannot supersede the rate of a diffusion-controlled reaction, which is of the order of 10 M⁻¹ ns⁻¹ for small molecules in solution at room temperature. That means, significant dynamic quenching requires quencher concentrations of some 100 mM, whereas high-affinity static quenchers may work at much lower concentrations. Dynamic quenching provides information on the exposure to the quencher-containing solution and/or changes in the local diffusion coefficient of fluorophore and quencher, for example, in a membrane.

Whereas dynamic quenching of dyes with monoexponential decay can be handled analogously to the Stern–Volmer equation as

(6) $$ \frac{\tau_0}{\tau }=1+{k}_q{\tau}_0\left[Q\right] $$

this is not really an option for Trp with its intrinsically multiexponential decay. Tracing the changes of each individual lifetime upon the addition of a quencher is impractical since, as the decays change, the three lifetimes of the fit do not represent the same three fractions of dye. Importantly, Eq. (6) does not hold true for average lifetimes. Instead, we find for amplitude-weighted average lifetimes, 〈τ〉, with Eq. (5):

(7) $$ \left\langle \tau \right\rangle =\frac{1}{\sum \limits_i{A}_i}\sum \limits_i\frac{A_i}{k_{Fi}+{k}_q\left[Q\right]} $$

so that 〈τ₀〉/〈τ〉 becomes

(8) $$ \frac{\left\langle {\tau}_0\right\rangle }{\left\langle \tau \right\rangle }=\frac{\sum \limits_i{A}_i}{\sum \limits_i{A}_{0i}}\frac{\sum \limits_i\frac{A_{0i}}{k_{Fi}}}{\sum \limits_i\frac{A_i}{k_{Fi}+{k}_q\left[Q\right]}} $$

Figure 1 illustrates the consequences on the example of acrylamide quenching data for a Trp-analog, L5W, of the lipopeptide viscosin located in a lipid membrane. The data points represent experimentally determined, amplitude-average lifetimes as a function of acrylamide concentration. The solid green curve depicts a fit based on Eq. (8) with a k _q of 0.33 M⁻¹ ns⁻¹. Note that the relationship is slightly non-linear. If one (ab)uses the linear Eq. (6) valid for single lifetimes to fit the average ones, there is still a reasonable apparent fit but with a wrong k _q of 0.47 M⁻¹ ns⁻¹ (blue, dash-dotted line in Figure 1). It should be emphasized that the fact that the data deviate only slightly from linearity does not justify using the linear Eq. (6) – the slope would only serve for an empirical comparison of quenching effects and not reveal k _q.

Figure 1.

Effect of the concentration of a dynamic quencher, [Q], on the amplitude-weighted average of the lifetime 〈τ〉, with 〈τ ₀〉 referring to the absence of quencher. The green solid line describes a fit with k _q = 0.33 M⁻¹ ns⁻¹ according to Eq. (8). The blue, dash-dot line represents an apparent fit using the (for average τ inappropriate) Eq. (6), yielding a wrong value for k _q of 0.47 M⁻¹ ns⁻¹. Experimental data for acrylamide quenching of viscosin L5W in liposomes (Carabadjac et al., Reference Carabadjac, Steigenberger, Geudens, De Roo, Muangkaew, Madder, Martins and Heerklotz2024).

A sophisticated way to fit k _q is to reconvolute Eq. (5) with the IRF to fit the raw decays with fixed A₀₁, τ_0i obtained in the absence of quencher and an adjustable k _q (Hirshfield et al., Reference Hirshfield, Toptygin, Grandhige, Packard and Brand1998; Toptygin et al., Reference Toptygin, Chin and Hilser2015).

Non-linear Stern–Volmer plots

As described in the previous section, Stern–Volmer plots based on average lifetimes of multiexponential decays as for Trp are generally non-linear.

Furthermore, a combination of static and dynamic quenching:

(9) $$ \frac{F_0}{F}=\frac{A_0}{A}\frac{\tau_0}{\tau }=\left(1+{K}_{static}\left[Q\right]\right)\left(1+{K}_{dynamic}\left[Q\right]\right) $$

with quenching constants for static and dynamic quenching, K _static and K _dynamic, introduces a quadratic term in F ₀/F as a function of [Q]. Individual evaluation of amplitudes and lifetimes, respectively, as possible with TR measurements provides separate information on static and dynamic quenching.

Finally, unequal accessibility of Trp residues in the protein by the quencher (also named after Lehrer (Lehrer, Reference Lehrer1971), who first rationalized this observation) or by diffusion limitations in heterogeneous media (downward curvature), heterogeneities in quencher concentration or very high quenching efficiency (upward curvature) and many other phenomena may cause nonlinear Stern–Volmer plots. For a detailed review of the analysis and interpretation of non-linear Stern–Volmer in complex systems, consult a review from Castanho and Prieto (Castanho and Prieto, Reference Castanho and Prieto1998).

Binding characterized by FRET

As discussed in some more detail below, Förster resonance energy transfer (FRET) reduces the lifetime of a donor dye (such as Trp) by transferring the excitation to an acceptor (e.g., a drug) that can be excited at the emission wavelength of the donor. Since the rate of transfer, which competes with the rate of fluorescence of the donor, depends strongly on the donor–acceptor distance, typically on the single-nm scale, it has been used to measure molecular distances (see below). However, it is also relevant for binding studies. If a drug acting as an acceptor binds to a protein site at a given distance from a Trp, it reduces τ of this Trp in a characteristic fashion.

In principle, the extent of binding can hence be derived from the appearance and extent of a shorter-lifetime component. This case has been discussed in detail, considering the binding of ANS, a widely used drug-analog and “site marker” competing for certain drug binding sites on albumin. Specifically, the question was addressed whether large local concentrations of glycocholate (GC) resulting from the dissolution of micellar drug delivery systems would compete with the pickup of the released drug cargo by albumins (Carabadjac et al., Reference Carabadjac, Vormittag, Muszer, Wuth, Ulbrich and Heerklotz2024). As another rate constant depleting the excited state, the rate constant of FRET, k _FRET, fulfills Eq. (5) analogously to k _q. This means that analogously to dynamic quenching, FRET from Trp should be evaluated by reconvolution curve fits according to Eq. (5) or using average lifetimes according to Eq. (8) to obtain correct values for k _FRET.

FRET from Trp in HSA to the model drug ANS was used to study the competition of ANS and GC for HSA binding sites (Carabadjac et al., Reference Carabadjac, Vormittag, Muszer, Wuth, Ulbrich and Heerklotz2024). This is important to ensure that the drug can be safely transferred from a GC-based micellar drug delivery system, which disintegrates after intravenous injection, to HSA. Figure 2 shows the concentration of ANS needed to induce a certain average lifetime of Trp as a function of GC concentration. Submicellar GC enhances FRET by ANS (less ANS needed to induce a certain reduction in τ) and supermicellar GC increases it gradually.

Figure 2.

Partially competitive binding of ANS and glycocholate to HSA: the concentration of ANS needed to collisionally quench the average fluorescence lifetime of Trp in HSA from τ ₀ = 5.5 ns to 1.5 (diamonds), 2 (spheres), and 2.8 ns (squares) as a function of the GC concentration. Decreasing values up to the CMC of GC imply that competition by GC increases FRET from Trp to ANS – a finding that can be explained by the relocation from a remote, high-affinity site for ANS (competed by GC) to a lower affinity, but not GC-binding site closer to the Trp residue. The slope above the CMC of GC reveals the low amount of ANS that is removed from HSA into GC micelles. Reproduced from Carabadjac et al. (Reference Carabadjac, Vormittag, Muszer, Wuth, Ulbrich and Heerklotz2024), copyright ACS 2023.

The counterintuitive finding that the presence of GC as a potential competitor of ANS binding increased FRET from Trp to ANS indicated that GC relocated ANS from competing sites to sites that bind ANS but not GC, which are located more closely to the Trp residue. Furthermore, the weak reduction of FRET in the presence of increasing amounts of glycocholate micelles demonstrated that ANS affinity to albumin is much higher than to the micelles. Hence, ANS-like drugs must be considered to transfer from micellar drug delivery systems to albumin, even if the micelle does not disintegrate.

The binding of the drug hymecromone to BSA reduced the amplitude of Trp fluorescence but not the average lifetime (Chaves et al., Reference Chaves, Cesarin-Sobrinho, Serpa, da Silva, de Lima and Netto-Ferreira2024). While BSA shows strong similarity to HSA in many aspects, it is markedly different with respect to intrinsic fluorescence since it contains two Trp residues instead of one. The lack of a significant effect on lifetime suggests that hymecromone is bound to a site very close to one of the two Trp residues, causing static quenching or very fast FRET, and sufficiently far from the other Trp to render FRET to this Trp negligible (mutual orientations matter as well).

Structure of a protein

A high number of time-resolved fluorescence studies puts the focus on albumins. Not only are albumins the most abundant plasma proteins (Goldwasser and Feldman, Reference Goldwasser and Feldman1997) and involved in the transport of numerous endo- (Czub et al., Reference Czub, Venkataramany, Majorek, Handing, Porebski, Beeram, Suh, Woolfork, Hage, Shabalin and Minor2019; De et al., Reference De, Kaur and Datta2013; van der Vusse, Reference van der Vusse2009) and exogenous substances including drugs (Varshney et al., Reference Varshney, Sen, Ahmad, Rehan, Subbarao and Khan2010; Yamasaki et al., Reference Yamasaki, Chuang, Maruyama and Otagiri2013; Zhivkova, Reference Zhivkova2015) – they are also widely used in drug formulation (Elzoghby et al., Reference Elzoghby, Samy and Elgindy2012; Gradishar et al., Reference Gradishar, Tjulandin, Davidson, Shaw, Desai, Bhar, Hawkins and O’Shaughnessy2005; Langer et al., Reference Langer, Köll-Weber, Holzer, Hantel and Süss2022). Human serum albumin (HSA) has only one Trp residue and is therefore remarkably suited for intrinsic fluorescence studies. Many groups use HSA alone or combine the findings with observation of bovine serum albumin (BSA), which contains two Trp residues.

The different exposure of Trp to the protein surface can in theory be evaluated with the help of reconstructed decay-associated spectra (DAS) for each component of the fluorescence decay with an individual lifetime (see Figure 3 for an example of DAS below). The idea is that the two Trps in different environments show different lifetimes. However, as described above, one Trp residue in a homogeneous solution already exhibits multiple lifetimes. If the lifetime results from overlapping signals of multiple Trp species in different environments, there is no conclusiveness in DAS belonging to the water-exposed or the buried Trp.

Figure 3.

Amplitude spectra, also referred to as decay-associated spectra (DAS, top panel) and corresponding time-resolved emission spectra (TRES, bottom) of Trp in the B1 domain of streptococcal protein G as a linear function of wavenumber, ν (bottom abscissae), and the corresponding wavelength λ (in top abscissae). The DAS represents amplitudes of four characteristic (globally fitted) lifetimes. An amplitude becoming negative beyond a certain wavenumber (as seen for the 73.3 ps component) indicates fluorescence at lower energy to grow at the expense of that at higher energy, for example, by a relaxation mechanism. TRES indicates electrostatic relaxation as a red shift at constant (as shown here) or varying width of the spectrum. Reprinted compilation of Toptygin (Reference Toptygin2014) based on Toptygin et al. (Reference Toptygin, Gronenborn and Brand2006), reproduced with permission, copyright 2006 American Chemical Society.

An elegant way to solve this problem is to use both DAS reconstruction and a quenching experiment with a quencher only soluble in the aqueous phase. In this manner, the DAS of BSA was observed upon increased concentration of the quencher calcofluor. In the presence of high quencher concentration, the species with red-shifted maxima of 335 and 340 nm declined in intensity. In contrast, the emission occurring from the species with the maximum at 330 nm remained constant. The last species was assigned to the interior of BSA, while the two others were assigned to the water-exposed surface (Tayeh et al., Reference Tayeh, Rungassamy and Albani2009).

The accessibility of Trp upon folding can be probed by acrylamide quenching. Acrylamide does not access residues in the interior of a protein. Although acrylamide was also shown to be involved in a weak association with indoles of Trp and induce static quenching, the dynamic part of the process can be selectively accessed by the use of the bimolecular constants k _q.

In contrast to Förster Resonance Energy Transfer (described in the text below and above), k _q reports on more transient and local contact (Lakowicz et al., Reference Lakowicz, Zelent, Kuśba, Gryczynski and Johnson1993). A study of conformational changes of protein OmpA upon contact with model membranes used this method. Bimolecular quenching constants of six single-Trp mutants of OmpA were observed for multiple steady-state folding intermediates. The water-exposed, unfolded, or denatured proteins were quenched the most, Trps inserted in the interface between the lipid head groups and the hydrophobic interior of the membrane were less quenched, and tryptophans buried in the folded protein were the least quenched. In this manner, the evolution of the folding process in dependence on membrane concentration could be described in detail (Asamoto et al., Reference Asamoto, Kozachenko, López-Peña and Kim2021).

Insertion (depth) of the peptide and its effect on the membrane

As portrayed in previous sections, there are different possibilities to distinguish between buried and water-exposed Trp in proteins. Similar approaches are conceivable for the detection of the peptide insertion in a membrane. Acrylamide is highly soluble in water and does not readily penetrate the hydrophobic surface of a membrane. The small amount of acrylamide that could enter the membrane would still lead to less frequent contact with tryptophan due to the high viscosity (low diffusion) in the membrane compared to the buffer. Comparison of bimolecular collision constants allows a demonstration of the binding of a peptide to or into the membrane. For multiple single-Trp mutants of one peptide of similar conformation, a determination of the insertion depth by the accessibility of the Trp to the quencher from the aqueous phase is also possible. This, in turn, allows a description of peptide orientation in the membrane (Carabadjac et al., Reference Carabadjac, Steigenberger, Geudens, De Roo, Muangkaew, Madder, Martins and Heerklotz2024).

Alternatively, Trp buried within the hydrophobic core of the membrane can be quenched by lipid-attached quenchers; for an overview of probes and quenchers, see Kyrychenko and Ladokhin (Reference Kyrychenko and Ladokhin2024).

Note that biologically relevant membrane perturbations or defects induced by low concentrations of membrane-active peptides are typically local and therefore not reported by membrane probes distributed over the whole membrane. This problem is solved by monitoring the properties of Trp included in these peptides – at the expense of information that can be gained only by synthetic membrane dyes optimized for a given experiment, such as Laurdan. One solution to this dilemma is to use FRET-GP: Laurdan is spread out in the membrane but excited only in the immediate neighborhood of the peptide (Thakur et al., Reference Thakur, Uday, Cebecauer, Roos, Cwiklik, Hof, Jurkiewicz and Melcrová2024). This is achieved by excitation via FRET from Trp, with a correction for direct excitation of the S2 state of Laurdan at 280 nm.

Lifetime changes were also described for the incorporation of Trp and Trp-containing di- and tripeptides into hexane-AOT inverse micelles compared to buffer (Gałęcki et al., Reference Gałęcki, Kowalska-Baron, Nowak, Gajda and Kolesińska2023). AOT/hexane caused the appearance of a third, longer lifetime (5―7 ns range) in addition to the biexponential decay seen in buffer. Neighboring amino acids tended to increase the average lifetime.

Conformation

Quite simplistic, yet sufficient qualitative analysis of changes in the conformation of proteins can be accomplished by comparison of lifetimes before and after a conformational change. As the protein unfolds and the surroundings of Trp change, the proximity or accessibility to quenching substances also changes. In this way, the rearrangement of residues of bovine β-lactoglobulin between the native conformation and the refolded state after heating and recooling was examined. The mean lifetime of Trp in the refolded state was significantly longer than the one of the folded state, confirming irreversible changes in conformation (Halder et al., Reference Halder, Chakraborty, Das and Bose2012). In contrast, a similar comparison between the unfolded and native structure in the next example would miss crucial details on intermediate conformational states in between. The examination of two similar luciferases showed that upon unfolding with increasing concentration of a denaturant (urea), the Trps see a very different environment (the lifetime of the first increased and of the second decreased), while the final unfolded structures show almost identical spectroscopical properties (Nemtseva et al., Reference Nemtseva, Gulnov, Gerasimova, Sukovatyi, Burakova, Karuzina, Melnik and Kratasyuk2021).

For a quantitative analysis of conformational changes by quenching, a relationship between the change in distance and the quenching effect has to be established. The spatial proximity on the scale of a protein size or the thickness of a membrane can be probed by Förster Resonance Energy Transfer (Förster, Reference Förster1948), FRET, or short RET to emphasize the non-emissive character of the transfer process. For more abbreviations, consult the literature (van der Meer, Reference van der Meer2002). The requirement for FRET is the overlap of the emission spectrum of the donor molecule with the absorption spectrum of the acceptor molecule. The energy of the photoexcited donor is then transferred to the acceptor without the emission of the photon. The donor molecule returns to the ground state. The acceptor molecule is excited by the energy transfer and emits the photon instead of the donor at a longer wavelength. The rate of FRET is dependent on the extent of the spectral overlap, the quantum yield of the excited donor, the distance between the two molecules, and their mutual orientation to each other. The ultimate goal of such measurements is usually an accurate distance determination. This may require a careful examination of the orientational factor, κ ², by anisotropy measurements of the donor and acceptor rotational diffusion (van der Meer, Reference van der Meer2002).

The transfer of the energy between the molecules can be monitored either by quenching of the donor molecule or by fluorescence increase of the acceptor molecule. Since only the proximity and orientation of the molecules matter, FRET can be either accompanied by dynamic or static quenching, depending on the system. Therefore, the information gained by this kind of experiment is of a great variety. It includes general topology and conformational changes of proteins, distance approximations in macromolecules and between molecules, accessibility studies of Trp in the protein, and hence, structural information and affinity determination of binding.

Studies employing intrinsic FRET of Trp to other structures sometimes use the abbreviation iFRET. Tryptophan shows some degree of spectral overlap between its own absorption and emission, but the so-called “homo-FRET” studies between two Trp residues remain only a small part of all performed studies of energy transfer. So, especially for FRET studies, the use of intrinsic fluorescence of Trp does not exclude the use of extrinsic fluorophores to gain combined information from two sources about the system. Since this review focuses on time-resolved fluorescence of Trp, only examples of dynamic quenching studies of tryptophan are covered in the following. For an extensive overview of applications of steady-state intrinsic fluorescence of Trp with an emphasis on FRET (Ghisaidoobe and Chung, Reference Ghisaidoobe and Chung2014) and for a review of older publications dedicated to conformational changes detected by Trp fluorescence (Engelborghs, Reference Engelborghs2001) consult the literature.

Biological function

The biological relevance of homo-FRET was shown in a study of proteins found in the vertebrate eye lens. The protein human γD-crystallin contains four conserved Trps, which are necessary to attain the required refractive index of the lens. Absorption of the UV light by those Trps leads to excited-state reactions, which cause cataracts; therefore, it is important to return the Trps to the ground state as fast as possible. The goal of the study was to understand the mechanism by which the absorbed energy is dissipated. Comparison in lifetimes of single or double Trp mutants showed that the absorbed energy is internally transferred between the Trps. The mutants containing only the donor tryptophan showed a long lifetime of ~3 ns, while mutants with the acceptor Trp were extremely quenched (~ 0.1 ns) by their surroundings. A part of the transferred energy from UV light can be relaxed in this way by non-radiative mechanisms (Chen et al., Reference Chen, Toptygin, Brand and King2008).

As previously mentioned, tryptophan is believed to act as an anchor for transmembrane domains of membrane proteins. Hence, the correct orientation in the bilayer and with that, the functionality of the protein relies on the preference of Trp for the interface between the aqueous solution and the lipids (de Planque et al., Reference de Planque, Bonev, Demmers, Greathouse, Koeppe, Separovic, Watts and Killian2003; Yau et al., Reference Yau, Wimley, Gawrisch and White1998). The biological function of a membrane is often related to effects of cholesterol (−derivatives). The indole ring of Trp, on the other hand, is involved in stacking interaction with other rigid ring structures (Dougherty, Reference Dougherty1996). So, it is not surprising that multiple groups elucidated interactions between Trp and cholesterol, searching also for consequences of changed organization or distribution of these molecules in membranes for the functionality of the proteins. However, no preferential localization of Trp next to cholesterol was found by FRET measurements between Trp of several peptides and the fluorescent cholesterol derivative DHE. Further, the dynamics of both molecules remained unchanged (Holt et al., Reference Holt, de Almeida, Nyholm, Loura, Daily, Staffhorst, Rijkers, Koeppe, Prieto and Killian2008). In this specific case, small concentrations of cholesterol of <8% were used to ensure a homogenous system, which can be compared to calculated values of random distribution of both molecules in the DMPC membrane. This study does not exclude the possibility of stacking interaction observed by other techniques and other groups at high cholesterol concentrations.

A completely different application of FRET is directed at the distinction of specific effects of isomers. It is known that optical isomers with identical physicochemical properties, such as L- and D-residues incorporated into an enzyme, result in fundamentally different biological activities. Particularly relevant seems the chiral inversion of the L-isomer of Trp into D-isomer upon aging, which may be involved in several pathological conditions including Alzheimer’s and Parkinson’s (Tverdislov et al., Reference Tverdislov, Yakovenko and Zhavoronkov2007). Recent studies use FRET efficiency to describe the differences in interactions of L- and D-Trp with chiral drugs, for example, the non-steroidal anti-inflammatory drug (NSAID) ketoprofen (Ageeva et al., Reference Ageeva, Babenko, Magin, Plyusnin, Kuznetsova, Stepanov, Vasilevsky, Polyakov, Doktorov and Leshina2020). The chiral drug is chemically linked to the Trp residue, producing diastereomers different in their spectroscopic properties. The diastereomers are then compared by their quenching intensity and differentiated by the quenching mechanism. Because these kinds of studies employ photo-induced electron transfer (PET) as well (Ageeva et al., Reference Ageeva, Babenko, Magin, Plyusnin, Kuznetsova, Stepanov, Vasilevsky, Polyakov, Doktorov and Leshina2020, Reference Ageeva, Magin, Doktorov, Plyusnin, Kuznetsova, Stepanov, Alekseev, Polyakov and Leshina2021; Khramtsova et al., Reference Khramtsova, Ageeva, Stepanov, Plyusnin and Leshina2017), it is worth briefly mentioning the differences between FRET and PET. Note first that PET is used mainly in single-molecule fluorescence techniques, which are not covered in this review. In comparison to FRET, only very short distances covered by van der Waals forces of <1 nm can be detected (Chen and Barkley, Reference Chen and Barkley1998). The absorption and emission spectra do not need to overlap in this case and Trp may act as a quencher because of its low oxidation potential. Trp in the ground state transfers an electron to the neighboring excited structure so that radicals are formed instead of emissive return of the neighbor to the ground state. Hence, the conformational flexibility of Trp-containing peptides can be accessed not only with high temporal but also with sub-nanometer spatial resolution (Buschmann et al., Reference Buschmann, Weston and Sauer2003; Doose et al., Reference Doose, Neuweiler and Sauer2005; Neuweiler et al., Reference Neuweiler, Schulz, Böhmer, Enderlein and Sauer2003).

An intriguing, though not technically advanced, approach used intrinsic TR fluorescence of human plasma for diagnostic purposes to detect increased oxidative stress in samples from a total of 59 patients. Among other parameters, the emission of the sample at 350 nm was examined after excitation at 280 nm. The authors of the paper assume that the fluorescence signal is mostly from Trp residues in HSA. The mean lifetime of the tri-exponential decay was then correlated with oxidative markers to identify possible correlations. The findings included, for example, that the level of aldehydes created by the oxidation of lipoproteins influences the mean lifetime, probably due to chemical reactions with the amino acids of HSA (Wybranowski et al., Reference Wybranowski, Ziomkowska, Cyrankiewicz, Bosek, Pyskir, Napiórkowska and Kruszewski2022).

Summary

Observation of the lifetime of Trp leads to very diverse information about the molecule under investigation (Figure 4). The lifetime depends on many different factors, such as proximity of Trp to other structures, the kind of solvent, the flexibility of Trp side chains and others. With that, the lifetime reflects the environment of Trp, but the exact reasons for the lifetime quenching are difficult to assess. Quenching experiments with external quenchers may provide information on the binding of the quenching substance in the neighborhood of Trp, but also on its accessibility from the outside. This may be valuable to describe protein structure or peptide insertion into membranes. Although these techniques are straightforward, some technicalities still must be considered. FRET experiments enable accurate determination of distances on the scale of the protein or membrane, which is exceptionally welcome in studies of conformations or protein folding.

Figure 4.

Examples of molecular information captured by TR quenching of Trp. (a) Depicts a protein with two Trps with different exposure to the aqueous phase. The position of the Trp is marked on the protein in light (surface-exposed) or dark (buried in the interior of the protein) color. The quenching substance is shown as a circle with a cross. The moving substance is marked with an arrow for the direction of the movement. The water-exposed Trp (top) can be accessed by the quencher and therefore is quenched (marked by a red cross). The Trp in the protein interior (bottom) cannot be accessed by the quencher and fluoresces. (b) Shows a protein before (top) and after (bottom) a conformational change. After the conformational change, the non-quenched Trp is nearby of the quenched Trp. Because of energy transfer, the previously quenched Trp (i.e. before the conformational change) is now able to emit light. (c) Shows a protein and a quenching substance before (top) and after (bottom) the binding. After the binding of a small structure nearby Trp lifetimes decrease (dynamic quenching) or the fluorescence intensity decreases and the lifetime remains constant (static quenching). (d) Presents two peptides with different insertion depths in the membrane. The peptide positioned at the membrane surface (top) is accessible for the quencher in the aqueous phase and can be quenched. The Trp buried in the membrane (bottom) is not affected by the quencher.

Considerations prior to experiments

Table 3 lists common issues with TR fluorescence anisotropy measurements. In comparison to other TR techniques, time-resolved anisotropy (TR anisotropy, time-resolved fluorescence anisotropy, TRFA, or TFA) is more susceptible to measurement errors. As the literature on the TR anisotropy of Trp shows, some misconceptions still exist. Therefore, we will first revisit the most common pitfalls of anisotropy measurements and then discuss them with examples.

Table 3.

Overview of the most common pitfalls of time-resolved anisotropy measurements

Usually, the time-dependent value of anisotropy, r(t), in TCSPC measurements is determined by:

(10) $$ r(t)=\frac{F_{\parallel }-{GF}_{\perp }(t)}{F_{\parallel }+2{GF}_{\perp }(t)} $$

where F _∥(t) is the time-dependent intensity (i.e., the photon count in the MCA channel corresponding to time t after excitation) for the vertically polarized fluorescence emission and F _⊥(t) is the time-dependent horizontally polarized equivalent. The subscript symbols for parallel and perpendicular reflect that in these experiments, excitation is vertically polarized. G stands for G-factor and accounts for the polarization dependency of the monochromator (Lakowicz, Reference Lakowicz2006) and potential effects of the deadtimes of the detection procedure (Pierce et al., Reference Pierce, Toptygin and Wendland2013). To obtain the vertical and horizontal intensities, the photons reaching the detector are accumulated for some time at a vertical polarization of the detection filter (VV) and afterward at the horizontal polarization (VH).

Photobleaching is not necessarily a linear process and is usually stronger in the first seconds of the measurement. This means that more of the VV detected signal will be lost than in the following HV configuration. This will, according to the equation above, lead to lower anisotropy values than actually displayed by the sample. So, what some authors label as “minimal photobleaching” of <10%, may indeed result in high measurement error and flat r(t) curves. Some degree of avoidance of this kind of error can be achieved by cyclic change of the detection polarization filter during the photon accumulation so that the photobleaching effects are spread evenly between VV and VH curves. Gentle stirring of the sample during bulk measurements in cuvettes removes the bleached sample from the beam pathway and a general decrease in the intensity of illumination also minimizes the effects of photobleaching.

An underestimate of r(t) can also be measured in samples of high concentrations of the analyte. There, the depolarization process is additionally increased due to homo-transfer or due to radiative reabsorption of the emitted photons. In some cases, this can be taken advantage of to determine the proximity of two Trps, as will be exemplarily shown below.

In comparison, measured values of r(t) being higher than actually exhibited by the sample, can usually be recognized by the value of TR anisotropy at time point zero, r(0), being above the theoretically possible limits. The maximally possible anisotropy, r ₀, is determined at conditions where the dyes are virtually immobile during the lifetime of the excited state. This can be achieved by measurements of anisotropy in propylene glycol at −60 to −70 °C (Valeur and Weber, Reference Valeur and Weber1977). For Trp, r ₀, and therefore r(0), is strongly dependent on the excitation wavelength, so the initial values for anisotropy reported in the literature vary with the measurement settings. As such, values above the limits do not always appear implausible at first glance.

So how does it come to measured r(0) showing incorrectly high values? While incorrect settings on the equipment are always a possibility, the more common issue is the use of an inaccurately determined G-factor or a very high contribution of scattering light to the measured signal. The G-factor is a correction mainly for the unequal ability of a monochromator to transit vertically and horizontally polarized light at a certain wavelength. In systems with a long dead-time of the detector, an underestimation of count numbers of high-intensity samples (VV, not VH) can also contribute to G. If the G factor is chosen poorly, the whole r(t) is shifted and determined r(0) may be above the possible values. The monochromator-related G-factor can be determined by measurement at horizontal polarization of the excitation beam (HV/HH measurement) or by anisotropy measurements of small dyes without specific interaction with the non-polar solvent. In both ways, the sample is displaying zero anisotropy and all differences in vertically and horizontally detected signals are due to the instrument contribution. To avoid detector dead-time effects on G, the frequency of detecting photons should be far below the reciprocal dead time. For example, for a dead time of 85 ns (as in a PQ FT300 used, for example, in Carabadjac et al. (Reference Carabadjac, Steigenberger, Geudens, De Roo, Muangkaew, Madder, Martins and Heerklotz2024)) and a pulse frequency of 20 MHz (one pulse per 50 ns), a detection of one photon per 5 μs (1%) corresponds to the detector being “dead” for 1.7% of the time. Since dead time and pulse interval are similar, this should cause about 1.7% of the counts to be missed. If the counts for F _⊥(t) are tenfold less frequent so that only 0.17% of F _⊥(t) are missed, the G-factor contribution would be 98.3/99.8 = 0.984. That means, depending on the dead time of the TCSPC system and the pulse frequency used, it may be advisable to attenuate the signal below 1% counts per pulse or necessary to quantify the dead-time contribution to G.

Scattering light has the same polarization as the excitation beam; hence, in an extreme scenario, when only the scattered light reaches the detector, the determined r(0) can be as high as “1.” In reality, scattered and fluorescent light both get detected, and r(0) is somewhere between the real value and “1” (Lakowicz, Reference Lakowicz2006).

Note that the determined r(0) is generally lower than the theoretically predicted r ₀ because of the temporal resolution limits of most equipment and hence the loss of the contribution of the very fast depolarization component.

Despite some of the TR anisotropy data containing one or more of these artefacts, usually, the data are obtained comparatively (against changing temperature, concentration, or viscosity) and only relative changes are discussed. In such cases, the conclusions derived from the data are not influenced by r(t) inaccuracy. Following this excursion regarding the most common pitfalls, let us now concentrate on the information about molecular processes that can be governed by this technique.

Size and shape

Most proteins and peptides are nonsymmetrical and have complex shapes. The structure and shape of molecules are connected to their function, to their interactions with other molecules and the environment (Li et al., Reference Li, Mach and Koehl2013). Widely used techniques for macromolecule characterization, such as dynamic light scattering (DLS) or multi-angle laser light scattering (MALS), require a basic idea of the geometric shape of the examined structures. In theory, the shape approximation can be gained by TR anisotropy measurements of such molecules in bulk solvents.

A perfect sphere is expected to have one rotational correlation time only. As the shape of the sphere prolongs or compresses in either direction, the axes begin to differ in length. Each axis has its own rotational rate, which can be determined in TR anisotropy studies. This can be illustrated with a simple example: It is easy to imagine that a disk-like molecule would take less time to perform an in-plane rotation than an out-of-plane rotation, which involves the displacement of solvent molecules (Lakowicz, Reference Lakowicz2006).

In theory, an asymmetric body, such as an ellipsoid with three different axes, can display up to five rotational correlation times. In practice, it is usually not possible to determine more than three because pairs of two correlation times are too close to each other to be resolved (Chen et al., Reference Chen, Toptygin, Brand and King2008, and Lakowicz, Reference Lakowicz2006). In a study that was aimed at confirming the shape approximation of HSA and BSA by static light scattering (SLS) and small-angle X-ray scattering (SAXS), only the prolate ellipsoid (one axis longer than the other two identical axes; imagine a rice grain) could be ruled out by the determined rotational correlation times of Trp. In TR anisotropy measurements, an oblate ellipsoid (one axis shorter than the other two axes; imagine a flatbread) would usually show three rotational correlation times. It can be tricky to resolve all, particularly if they are similar, which can be expected for oblate ellipsoids during a hydrodynamic rotation. Additionally, the theoretical rotational time of the rotation ellipsoid is expected to be only slightly higher than for a sphere. The authors were therefore not able to distinguish between the sphere model and the oblate ellipsoid model (Starosta et al., Reference Starosta, Santos and Almeida2020).

Large sizes

Another study with albumins (cod and single Trp-containing mutant rat parvalbumins) tried to tackle the difficulty of accurately resolving multiple, cross-correlating rotational correlation times, as mentioned in the previous section. Since the depolarization of Trp fluorescence is derived from the counts at a given time after excitation, correlation times can be determined in the range limited by the time resolution of the instrument (related to the width of the IRF) and the loss of a quantifiable signal, which drops to 10⁻³ of its peak value after ~7 lifetimes.

As virtually always, the resolution of the fit for several correlation times could be improved by a global fit. In this case, single Trp residues located in different positions of the protein may have different decays but should represent the same overall shape (i.e., set of up to 5 correlation times) of the protein in solution.

Conformation

The scope of possibilities opened by TR anisotropy in tracking the conformational changes upon unfolding of proteins was shown on three single-Trp mutants of native Apoflavodoxin, containing three Trps. There, a distinction of the folded, unfolded, and intermediate states of the protein was performed by grouping the TCSPC data upon the increase of the denaturizing substance in matrices. Thermodynamic properties of all three populations could be accessed by weighting the anisotropy-derived data (i.e., assignment of the states) by their fluorescence amplitude (that is population of each state involved in depolarization) and a global fit of all data (Laptenok et al., Reference Laptenok, Visser, Engel, Westphal, Van Hoek, Van Mierlo, Van Stokkum, Van Amerongen and Visser2011). Despite an extensive discussion of the TR and SS anisotropy results in this publication, there seems to be no correction for the G-factor in the equations. However, the assignment and comparison of the thermodynamical data of the folding states by their rotational components should not be affected by the changes in the absolute values of the anisotropy decay. Another group studied the denaturant-induced unfolding of Apoflavodoxin by TR anisotropy decrease upon energy migration from Trp to Trp. Single or pair Trp mutants enabled the distinction of rotational correlation times of the whole protein (single-Trp mutant) and the reduction of the apparent rotational correlation time due to the homo-FRET (double-Trp mutants). The changes in homo-FRET rates upon increase in denaturant concentration come from changes in distances between the Trp residues and give therefore more detailed information on the conformation of intermediate states (Visser et al., Reference Visser, Westphal, van Hoek, van Mierlo, Visser and van Amerongen2008).

Another study followed the folding of the transmembrane protein OmpA upon membrane contact by examining the changes in anisotropy decay kinetics of five single-Trp mutants (Kim et al., Reference Kim, Arjara, Richards, Gray and Winkler2006). The r(t) curves are highly uncertain, which is partially discussed in the text. The r(0) exceeds the expected ~0.12 of measurements in cold propylene glycol (Valeur and Weber, Reference Valeur and Weber1977) due to a scattered-light contribution, as mentioned in the manuscript, but the absence of a correction for photobleaching effects, aided by a low signal-to-noise ratio, is not acknowledged. Once again, the uncertainty of the absolute numbers is dealt with by a relative comparison either between the mutants or between the different environments of buffer, micelle, or liposome. The authors interpret the increase of r(t) upon micelle and liposome contact mainly as the differences in the kinetics of the unfolded and folded state of OmpA and do only briefly comment on the contribution of the environment. Keeping the Stokes–Einstein relationship in mind, the assumption of protein kinetics alone dominating TR anisotropy parameters seems a little limited. The CD spectra presented in the same study clearly indicate a folding process of the protein at the membrane contact. Some parts of the changes in the rotational correlation times undoubtedly belong to the new structure of the protein. However, the approximation of the viscosity effect of the DMPC environment in comparison to less viscous OG micelles and even less viscous urea solutions is equivalently needed to quantify real differences in the dynamics of the different folding states.

Binding

The rotational correlation time can be used to see an increase in the size of rotating structures, for example, upon binding of other molecules to a protein. For this, small rotational correlation times are not involved in the analysis to exclude the molecular tumbling movement of Trp itself. Instead, the analysis concentrates on the large, overall motions. In this way, the self-assembly of DNA-bound HIV-1 integrase in solution was studied in the presence of different concentrations of DNA. It was shown that with increasing concentration of DNA, HIV-1-integrase disintegrates and rotates faster (rotational correlation time φ decreasing from 100 ns at DNA absence to 30 ns at the highest concentration). This behavior was confirmed by TR anisotropy from DNA labeled with two different labels (Deprez et al., Reference Deprez, Tauc, Leh, Mouscadet, Auclair, Hawkins and Brochon2001).

Summary

To put the matter in a nutshell, TR anisotropy can provide information on structure, size and shape of a protein or peptide, the viscosity of its environment and interactions with other molecules (Figure 5). Depending on the question investigated, the fast tumbling of Trp or the overall slow motion of the protein may be the best parameter to examine. The current state of research involving TR anisotropy provides the most robust results if performed as a comparative study with at least one changing parameter. This can be concentration, temperature, position of Trp in the protein, or many others. If the determination of accurate absolute values is required, one needs to be more cautious with TR anisotropy measurements than with other TR techniques.

Figure 5.

Examples of molecular information captured by TR anisotropy measurements. (a) Depicts a protein with two unequal axes and therefore two different rotational correlation times. The position of the Trp is marked on the protein in a light color. The different θ can be used for estimation of the oblong shape. (b) Shows two proteins of different sizes with different rotational correlation times of the whole structure. Small protein (top) rotates faster, and big protein (bottom) rotates slower. (c) Shows a protein and a second structure before (top) and after (bottom) the binding. The position of Trp is marked as a red circle. After the binding of a small structure nearby, the rotation of Trp is hindered, and hence slower. (d) Presents a peptide before (top) and after (bottom) the insertion into a membrane. The rotational correlation time of the Trp is longer in the membrane because the environment is more viscous than the buffer above. The rotational correlation time of the whole peptide is slower in the buffer because a full rotation along the long axis is possible. In the membrane, only the rotation along the short axis is not hindered by nearby lipids, and θ of the peptide is faster.

Considerations prior to experiments

As we will see (see also Table 4), a detailed quantitative interpretation of the extent and correlation time of TRES is far from trivial but again, some valuable insight can already be obtained by comparing different systems and conditions on a more qualitative basis. For example, Li and Zhang (Reference Li and Zhang2025) measured TRES of a number of single-Trp mutants of M protein from SARS-COV and SARS-COV-2 at different temperatures (Figure 6). Whereas spectral relaxation-related processes are similar for the two proteins and little temperature dependent at the 31 position, a significant thermotropic enhancement of relaxation at the 218 position is observed for the protein from SARS-COV-2 but not from SARS-COV.

Table 4.

Overview of the most common pitfalls of time-resolved emission spectra measurements

Figure 6.

Time-resolved, relative spectral shifts of single-Trp mutants of M protein in SARS and SARS2 for different Trp positions (plots) and temperatures (see legends in plots). Reproduced with permission from Li and Zhang (Reference Li and Zhang2025), copyright Elsevier, 2024.

This may suffice to raise hypotheses and motivate and guide further experiments, even though a detailed mechanistic interpretation of the effect has to be done with utmost caution. In fact, many other, additional, unknown processes, apart from dipolar relaxation of the solvent, can lead to changes in TRES. To name some examples, proximity to charged amino acids can lead to hydrogen bonding not involving solvent molecules, exciplex formation, acid–base chemistry, charge-transfer interactions, etc (Chen and Barkley, Reference Chen and Barkley1998). Many investigators simplify the readout of TRES experiments in heterogeneous systems and interpret it solely as a level of hydration in the Trp vicinity, disregarding the dynamics of hydrated structures behind them. While this approach is reliable in neat solvents, it is not ubiquitously applicable to complex systems. Others conclude that TRES is not suitable to probe protein hydration dynamics at all, due to the large protein contribution to the relaxation dynamics. Others again jump directly to the conclusion that TRES analysis gives mainly information about the mobility of the environment of the protein/peptide without the knowledge of the exact position of the dye. The rate of time-dependent red shifts may represent the speed of relaxation, but also the change in proportion of the fluorescence of Trps in different static environments that cannot interconvert on the nanosecond scale (see below for more detail). This leads to high ambiguity in the interpretation of data and their analysis, which is not a problem restricted to Trp, but to the overall application/interpretation of TRES.

Additionally, a number of procedures in the analysis of TRES of other fluorophores are not transferable to Trp’s fluorescence directly and have to be adapted. In the following, one relevant example will be covered in detail, but first, the general procedure to obtain this parameter will be described.

One way to obtain TRES is to reconstruct emission spectra with information from fluorescence emission decay histograms for distinct time points after excitation. As already explained previously, these spectra usually show a redshift over time due to the conversion of spectral energy by time-dependent relaxation processes.

The most common analysis of TRES involves mainly the extent of the spectral shift (also known as time-dependent fluorescence shift TDFS, time-dependent red-shift TDRS, time-dependent spectral shift TDSS, also time-dependent Stokes shift TDSS, dynamic Stokes shift, fluorescence Stokes shift FSS, and others) and the time scale of relaxation (relaxation time, dielectric relaxation time, dipolar relaxation time), on which this shift happens (Jurkiewicz et al., Reference Jurkiewicz, Sýkora, Olzyńska, Humpolíčková and Hof2005). Some quantify this shift as a ratio between intensities at two wavelengths (or wavelength regions) in the left and right limbs of the spectrum (Sot et al., Reference Sot, Esnal, Monasterio, León-irra, Niko, Goñi, Klymchenko and Alonso2021, Reference Sot, Gartzia-rivero, Bañuelos, Goñi and Alonso2022), similar to the GP (Parasassi et al., Reference Parasassi, Stasio, Ravagnan, Rusch, Gratton, Sperimentale and Nazionale1991). Most use the center of gravity of a spectrum of intensity versus wavenumber, ν(t), because of the proportionality of the wavenumber to energy. As a result, true relaxation gives rise to an exponential curve proportional to the energy decrease in the system over time.

The inability of the instrumentation to capture the whole relaxation process, including the pico- and sub-picosecond processes, raises the need for an estimation of the amount of lost information. The extrapolation to time zero was shown to largely depend on the time resolution of the equipment. Instead, an estimation of the Frank–Condon state ν(0) is performed by measurements of dyes in non-polar solvents where solvent relaxation is negligible. The subsequent subtraction of the limiting energy after the complete relaxation, ν(∞), in aqueous systems, gives the overall energy change in the sample Δν (Fee and Maroncelli, Reference Fee and Maroncelli1994).

For Trp, this procedure is not exactly suitable, because Trp possesses two different absorption states (¹L_a and ¹L_b) corresponding to the presence or absence of H-bonding to N–H of the indole. Hence, Trp exhibits, depending on the environment, two very different shapes of emission spectra with different energetic states. In non-polar solvents, the ¹L_b state is predominant and does not reflect the ¹L_a state in aqueous systems. Further, for the estimation procedure by Fee and Maroncelli (Reference Fee and Maroncelli1994), the fluorescence lifetime should be much longer than the relaxation time. While this is true in bulk solvents, in some cases, such as in the membrane environment, relaxation takes as long as the lifetime of Trp. Additionally, in peptides and proteins, Trp is connected via an amidic bond to other residues. The “pull and push” of the neighboring electron clouds may influence Trp’s Frank–Condon state. The question arises whether ν(0) of Trp in these different situations remains the same and is allowed to be estimated for a “general” Trp.

Polarity of the environment

Multiple groups estimated ν(0) for Trp in different systems, such as in buffer, proteins, and micelles. The values are in the same range with 31046 cm⁻¹ determined by femtosecond-resolved measurements (Qin et al., Reference Qin, Chang, Wang and Zhong2012), 31160 cm⁻¹ for NATA in water obtained by Maroncelli method (Bose et al., Reference Bose, Adhikary, Mukherjee, Song and Petrich2009), and ~ 30000 cm⁻¹ reconstructed from the time-zero emission spectra using a log-norm distribution function with parameters extrapolated from measurements with the protein Mistic (Frotscher et al., Reference Frotscher, Krainer, Schlierf and Keller2018). Measurements at very low temperatures near absolute zero can bring the relaxation to a negligible level and make a ν(0) estimation possible. This procedure would require suitable equipment and a (protein-)matrix that remains in an amorphous state upon freezing. The value estimated for ν(0) for Trp is 31850 cm⁻¹ at 100 K after excitation at 280 nm. In the same study, it was shown that above 100 K, a substantial part of Trp emits from the ¹L_a state (Scott et al., Reference Scott, Campbell, Cone and Friedman1989). Given the similarity of the estimated values, it may be possible to use these values as an approximation for the calculation of Δν, even if the real overall energy change of the system is dependent on Trp’s substituents and ¹L_a/¹L_b emitting ratio.

Heterogeneity of the environment

Despite the major discrepancies in interpretation, several valuable studies were performed in a variety of pharmaceutically relevant systems. As early as the 2000s, an extensive study of Trp in the protein matrix of CMPK combined different TR measurements to understand the effects of the matrix. They observed the quenching of Trp by water-soluble acrylamide, TRES, and TR anisotropies at several temperatures and in several glycerol/water mixtures in order to disentangle the contribution of different processes to dipolar relaxation (Vincent et al., Reference Vincent, Gilles, Li de la Sierra, Briozzo, Bârzu and Gallay2000). The change in the time-dependent spectral width (time-resolved full spectral width at half maxima, TR-FWHM) was shown to reflect the heterogeneity of a fluorophore’s environment in proteins and membranes. This valuable and easy-to-get parameter makes it possible to evaluate, to what extent the data may be interpreted utilizing the relaxation of only one environment, i.e., hydration level/dynamics. This observation was made by multiple groups and not only for Trp but also for other fluorescent solvent-sensitive dyes, such as Prodan, Laurdan, and Patman (Rieber et al., Reference Rieber, Sýkora, Olzyńska, Jelinek, Cevc and Hof2007; Toptygin et al., Reference Toptygin, Savtchenko, Meadow and Brand2001). By analysis of time-resolved area-normalized fluorescence emission spectra (TRANES), one can additionally unambiguously claim the presence of a single emitting species without and with a continuous solvent relaxation in its vicinity by the absence of isoemissive points. One isoemissive point shows two emissive species (e.g., in the membrane and buffer) and combinations of isoemissive points and continuous shifts of the spectra point to multiple emissive species with different environments (Ira et al., Reference Ira, Koti, Krishnamoorthy and Periasamy2003; Koti et al., Reference Koti, Krishna and Periasamy2001). TRANES analysis was also performed for Trp in proteins, such as HSA. While the publication shows some methodical problems and focuses strongly on trapped-water relaxation dynamics, ignoring the possible contributions from polar residues, TRANES analysis is still shown to be able to reveal two distinct emissive species of Trp (Otosu et al., Reference Otosu, Nishimoto and Yamashita2010).

An interesting perspective on this topic was presented by an MD simulation study predicting the steady-state emission maxima of Trp in 16 proteins based on the change in emission according to the electric field imposed by the protein (internal Stark effect) and the solvent matrix. The main idea is that the maximum of the spectra is primarily determined by the electrical potential differences across the long axis of the indole ring. This means that the same dipole in the vicinity of indoles can induce a red and a blue shift, depending on its position and orientation to the indole ring. Water and residues in proteins contribute to the spectrum in various ratios. If a strong red shift of Trp fluorescence induced by the protein is replaced by a weak red shift by water entering the environment of Trp, water exposure can also induce an overall blue shift (Vivian and Callis, Reference Vivian and Callis2001). This is, of course, transferable to time-resolved studies and has to be kept in mind for peculiarly unexpected shifts, which have to be addressed by additional accessibility experiments. Some examples were described above, such as quenching by acrylamide to determine Trp’s water exposure and the actual possible extent of water molecules reaching the Trp from the protein surface and/or TR anisotropy to determine the degree of restriction of the vector of indole ring in bulk measurements.

H₂O versus D₂O to assess the contribution of water

Comparison of dipolar relaxation times in H₂O and D₂O provides information about the contribution of water as compared to other polar groups to dipolar relaxation. For the fully water-exposed fluorophore in NATA, the ratio between the dipolar relaxation times in D₂O to H₂O, R_D/H, amounts to ≈1.5. The cyclic lipopeptide viscosin with Trp inserted at the 1, 5, and 4 positions localized in a lipid membrane interface or core environment showed a lower ratio of ≈1.2 (Carabadjac et al., Reference Carabadjac, Steigenberger, Geudens, De Roo, Muangkaew, Madder, Martins and Heerklotz2024), interestingly in line with this ratio for laurdan in the carbonyl region of a lipid membrane (Beranová et al., Reference Beranová, Humpolíčková, Sýkora, Benda, Cwiklik, Jurkiewicz, Gröbner and Hof2012). In contrast, viscosin L7W in membranes was argued to relax via interactions within the peptide, R _D/H ≈ 1 (Carabadjac et al., Reference Carabadjac, Steigenberger, Geudens, De Roo, Muangkaew, Madder, Martins and Heerklotz2024). Insensitivity of the lifetime of Trp in phosphofructokinase to deuterated buffer was interpreted in terms of the indole being not accessible to the bulk aqueous solution (Kim et al., Reference Kim, Chowdhury, Stryjewski, Younathan, Russo and Barkley1993).

Partition coefficient

Even for the steady-state measurements of the fluorescence shift of small peptides with surface-exposed Trp, a lot has to be considered during the experiment and the following analysis. In such systems, a simplified analysis in terms of “redshift equals water exposure” may be sufficient to describe the system. When using changes in the spectral maximum of Trp as the degree of binding of a peptide to a model membrane, scattering effects of the latter are contributing to the intensity signal already at low concentrations to a very high degree. While this would not disturb TR measurements, if the binding is described by a steady-state technique, the intensities have to be corrected. For that, additional measurements of a water-soluble Trp derivative, such as N-Acetyl-L-Trp amide (NATA), that does not interact with membranes, have to be performed in the presence of different lipid concentrations (Ladokhin et al., Reference Ladokhin, Jayasinghe and White2000).

It should be noted that interpreting a partitioning-dependent parameter in terms of a combination of fixed values for the free and inserted state to obtain a partition coefficient is appropriate only for linear-response parameters (Toptygin and Brand, Reference Toptygin and Brand1995). This includes steady-state intensity at a fixed wavelength (after corrections for turbidity, inner filter) but not, for example, amplitude-averaged lifetimes. The lines in Figure 7 were simulated using K _d determined from steady-state data – not fitted to the data with adjustable K _d.

Figure 7.

Dynamic parameters of Trp (W) inserted into the membrane-active cyclic lipopeptide viscosin in four different positions (replacing leucine, L, or valine, V, in different positions – see plot) as a function of lipid concentration, c _L. Values at c _L = 0 refer to the peptide in buffer; increasing lipid concentration beyond the dissociation constant K _d (see range marked in plots) causes an increasing fraction of the peptide to insert into the membranes – the shapes of the curves correspond to K _d. The characteristic times extrapolated to total membrane-binding differ markedly for the different positions in the peptide and, hence, in the membrane (see scheme on the right). For example, the deeply membrane-embedded Trp 4 reorients quickly (low φ) but shows a very slow dipolar relaxation over ≈10 ns; a process related to the recruitment of water to Trp-4 within the non-polar region of the membrane. Reproduced with permission from Carabadjac et al. (Reference Carabadjac, Steigenberger, Geudens, De Roo, Muangkaew, Madder, Martins and Heerklotz2024), copyright 2024 Biophysical Society.

Mobility of lipids in the membrane

For the dye Laurdan, the redshift of which is often interpreted to represent the hydration of the membrane, it was shown that the commonly used not-time-resolved generalized polarization (GP) value reflects predominantly the mobility of the hydrated sn-1 carbonyls of lipids and not the amount of solvent in the lipid bilayer (Amaro et al., Reference Amaro, Šachl, Jurkiewicz, Coutinho, Prieto and Hof2014). This conclusion was reached after a careful comparison of steady-state and TR data and is most likely transferable to Trp located in the same region of the membrane. For the correlation between multiple TR parameters of Trp in different positions in a membrane and a discussion with respect to environment, heterogeneity, and mobility, see Figure 7 (Carabadjac et al., Reference Carabadjac, Steigenberger, Geudens, De Roo, Muangkaew, Madder, Martins and Heerklotz2024). The extensive progress in understanding TRES for other fluorophores has led to a higher degree of understanding of such data and can be examined elsewhere (Scollo et al., Reference Scollo, Evci, Amaro, Jurkiewicz, Sykora and Hof2021).

Summary

TRES provides a multitude of information on Trp’s environment. In theory, hydration dynamics, conformational changes in proteins and peptides, binding of ligands to proteins, binding of peptides to membranes, and many more processes can be detected and described (compare Figure 8). In reality, this information is highly interwoven and can be only disentangled by very cautious interpretation. The evaluation should include at least the examination of the heterogeneity of Trp environment by TR-FWHM interpretation, and preferably further experiments regarding the flexibility of Trp by TR anisotropy and accessibility of Trp from aqueous exterior by TR fluorescence quenching. Additionally, one has to keep in mind the large contribution of the protein matrix to the determined values and/or that the determined parameters may reflect not the hydration of the system, but the dynamics of hydrated structures.

Figure 8.

Examples of molecular information captured by TR spectral shift. (a) Depicts a protein with two Trp residues marked as “Trp,” one on the surface (top) and the other buried (bottom) in the protein. The Trp on the surface is surrounded by a solvent shell, shown as blue circles with an arrow, indicating high polarity. Different accessibility to the polar solvent is detectable by TRES. (b) Shows a protein before (top) and after (bottom) a conformational change. A patch of polar amino acids is shown as dark green circles with arrows indicating their high dipoles. After the conformational change, the proximity of polar residues to Trp will change the spectral shift. (c) Shows a protein and a second structure before (top) and after (bottom) binding. After the binding of a small structure nearby, the exposure of Trp to the polar solvent is reduced. (d) Presents a peptide before (top) and after (bottom) the insertion into a membrane. After the insertion into the hydrophobic core of a membrane, the hydration shell of the peptide is mostly stripped off and only a little solvent exposure remains.