Macromolecular High-Affinity Binding Probed by Advanced Fluorescence Techniques

Abstract

Due to the extreme sensitivity and the intrinsic selectivity of fluorescence techniques, high-affinity binding can be measured even at extremely low molecule concentrations in the picomolar range. In particular, modern advanced techniques with fluorescence microscopes have provided considerable methodological advancements in recent years. Here, a brief description of the basic physical principles of fluorescence detection and its experimental measurement setups are provided. For interacting biomolecules in solution, confocal fluorescence microscopy enables some very effective approaches to characterize binding in complex sample environments and with small sample consumption. In addition to standard techniques with bulk samples in classical spectrometers, applications with single-molecule Förster resonance energy transfer, two-color coincidence detection, and fluorescence correlation spectroscopy are presented. The strength of the more advanced techniques lies in their broad applicability, ranging from fluorescence-based genetically encoded biosensors for use in living cells to the high controllability in the measurement of binding curves even at very low molecule concentrations. The advantages and limitations of the individual techniques are compared and recent state-of-the-art applications are discussed.

Keywords: FRET; biosensors; fluorescence spectroscopy; protein‐protein interactions; single‐molecule studies.

Figure 1

Scheme illustrating fluorescence principles and introducing corresponding fluorescence parameters. A) Jablonski diagram illustrating electronic states (bold lines) and vibrational states (fine lines) as well as possible transitions. B) Fluorescence detection scheme for the measurement of bulk samples in a spectrometer. Typically, the fluorescence emission is measured under 90° angle with respect to the excitation beam direction. For FA measurements the excitation light is linearly polarized and the emission light is measured with an analyzing polarizer allowing light to pass which is either parallel (I_||) or perpendicular polarized (I_┴) with respect to the excitation beam. C) Absorption and emission spectrum of a typical fluorophore (inset). D) Definitions of relevant fluorescence parameters.

Figure 2

A) Illustration of two different schemes for studying bi‐molecular binding with FRET, (upper panel) a binding‐induced conformational change in one dual‐labeled binding partner (intramolecular FRET) and (lower panel) a binding complex formation with two individually labeled binding partners (intermolecular FRET). The involved dyes are shown in blue color for the donor and in red color for the acceptor. B) Spectra of absorption and emission obtained from donor and acceptor fluorophores as well as the overlap integral J(λ) of donor emission and acceptor absorption (violet colored). The latter is needed to calculate the Förster radius R₀. C) Energy transfer efficiency E plotted as a function of R/R₀ ratio (with the inter‐dye distance R). D) Exemplary data of an intramolecular FRET sensor for the unbound (blue) and the fully bound (red) state. Here the left peak (donor) decreases at the expense of the right peak (acceptor), which increases upon increasing ligand binding.

Figure 3

Scheme illustrating confocal fluorescence detection of diffusing particles and data analysis. A) Upper panel: Illustration of a confocal fluorescence setup with pulsed excitation (laser 1,2) and time‐resolved simultaneous dual‐color detection (avalanche photo diode detectors 1,2). By using a pinhole and a high NA objective a confocal detection volume is created which typically differs in size for different excitation wavelengths (e.g., blue and red excitation light). middle panel: Here, blue and red excitations are carried out alternately (with a time shift of approx. 25 ns) and at a repetition rate of 20 MHz, so‐called pulsed interleaved excitation (PIE). Each repetitive interval is characterized by two detection channels (upper, lower subpanel), an excitation pulse (dark blue & dark red bars) and a potential fluorescence emission (light blue & light red triangles). Furthermore, each interval is subdivided into two time intervals: first one with only blue excitation, second one with only red excitation. As shown in the scheme, different scenarios of dye attachments for particles passing through the detection volumes can be distinguished by using this approach. lower panel: Individual (or some) particles which enter or leave the detection volume at given times generate a more or less strongly fluctuating intensity signal as a function of time. Examples of such intensity time‐traces are shown for the two detection channels (blue and red). A similar time‐trace is also generated for the PIE channel (not shown here). B) Using data of intensity time‐traces as measured with molecules in the nM concentration regime, a correlation analysis has been performed. The calculated auto‐correlation functions from blue and red labeled particles and the corresponding cross‐correlation function (green curve) (according to Equation 9 and 12, respectively) are displayed together with relevant parameters (e.g., τ _D and f_b) as obtained from fits (Equation 10). C) Data measured with molecules in the pM concentration regime are employed for a burst analysis, which gives access to single‐molecule information. Here, the fluorescence intensity of individual bursts (often visible as separated intensity peaks in the time‐trace) can be employed for two different approaches (smFRET and BTCCD). Upper panel: The obtained burst intensities (i.e., number of photons per burst) are related to the donor and the acceptor channel (I_D and I_A) which are then used to calculate FRET efficiencies E (Equation 5). For hundreds to thousands of bursts the individual E‐values are displayed in a FRET histogram. Here the ligand bound (blue) und unbound (green) subpopulations of a FRET sensor are characterized by their respective <E> values and their statistical weights. Lower panel: In a TCCD analysis the number of bursts measured in both detection channels are used to calculate binding fractions, here for example f_BR (see Equation 14). In order to reduce the impact of detection volume mismatch a further brightness threshold is introduced which allows to identify a more reliable f_BR‐value. This value is reached, when the coincidence fraction as a function of a threshold value shows no further increase with increasing threshold values. In the respective graph data from five different calibration samples are shown, which were preset for defined binding fractions (0%, 25%, 50%, 75%, 100%, for details see^{[ 41 ]}).

Figure 4

Two‐color FCCS analysis with a set of calibration samples of the same type as those for which data are shown in Figure 2C. Here three different double‐stranded DNA molecules were labeled with a red and a blue dye to varying degrees. RB: dual labeled with red and blue dye, R: single labeled only with a red dye, B: single labeled only with a blue dye. The three different dsDNA molecules were mixed in a way that they all carry the same total concentration of red and blue dyes, but the individual dyes were distributed differently among the double and the single‐labeled dsDNA. The relative percentages are given for all five different mixtures in the legend within the figure. The corresponding CCFs are shown from 100 % dual labeled to 0% dual labeled, from top (black) to bottom (dark green). In addition, the corresponding auto‐correlation functions are shown for the red and the blue detection channel. Although the molecule concentration of red and blue dyes is within the limit of error the same, a lower ACF amplitude at τ = 0 is observed for the red channel, because the confocal detection volume is larger than the one in the blue channel. This leads to an apparent higher number of molecules inside the detection volume (G(0) ≈ 1/N). The related (uncorrected, see text) binding fractions can be obtained from Equation 13. However, we clearly see the decrease of the cross‐correlation amplitude with more or less equidistant steps from 100%, 75%, 50%, 25% to 0% dual‐labeled species.

Figure 5

Examples of binding studies employing smFRET. A) FRET efficiency histograms of dual‐labeled ProTa molecules and unlabeled H1 with varying concentrations (between 0 and 100 pM). Here the red colored peak represents the unbound population and the blue colored one the bound population. Based on the data shown, a resulting K_D‐value of ≈50 pM was obtained for the given ionic strength. Adapted with permission from ^[12b] Copyright 2018, Springer Nature. B) Binding and unbinding events are observed as anti‐correlated intensity fluctuations between the donor (green) and the acceptor (red) channels (upper panel). In the unbound state the acceptor intensity does not drop to zero, due to the presence of multiple freely diffusing acceptor labeled single strands. From the intensity traces FRET values can be calculated, which exhibit a high and a low FRET state, that is, the bound and unbound state, respectively (middle panel). Crossings between high and low FRET state define a time (dwell time) in which the construct stays in the corresponding state. The measured dwell times are used for fittings with a probability distribution function to obtain association rates (k_a or k_on) and dissociation rates (k_d or k_off), see lower panel. Adapted with permission from ^[63c] Copyright 2013, Elsevier. C) The working principle of a FRET‐based biosensor is illustrated in the upper panel. Upon ligand binding the relative distance between the involved FPs decreases and the related FRET efficiency increases. Two different sensor constructs (no. 1 and no. 2) are compared with respect to their performance. The corresponding FRET histograms display a bound population (blue colored peak fit) and an unbound population (red colored) at different ligand concentrations, from to top to bottom: (I) no ligand, (II) concentration around the K_D value of ≈1 mM, and (III) fully saturated). Adapted with permission from^{[ 36 ]} Copyright 2018, ACS.

Figure 6

Applications of BTCCD for an antigen‐antibody binding study. A) The antigen is represented by an intrinsically fluorescent protein, an EGFP, which is suitable for single‐molecule studies. The corresponding antibody (nanobody) is labeled with a red fluorescent dye. B,D): Depending on the environmental temperature, the equilibrium time for complete binding (at given molecule concentrations) is much shorter for 45 °C as compared to 30 °C. C,E): Single molecule‐based titration measurements were used for fitting a binding curve (hyperbolic fit model, see Box 1) which enables the determination of the corresponding K_D‐values. As expected the binding interaction is slightly weakened at higher temperatures. F) Van't Hoff plot with K_D values from obtained f_BR values as measured at temperatures between 20° and 45 °C. The fit (black line) of data points in the plot results in the given thermodynamic parameters at a reference temperature of 25 °C. Adapted with permission from Schedler et al. ^[13b] Copyright 2023, MDPI.

Figure 7

Experimental realization of a quantitative binding assay. A) Scheme of the titration in droplets. The droplets are produced to set up a concentration gradient of the respective ligand (HumRadA), so that the ratio of HumRadA to the other binding partner (BRC4) increases. B) Schematic view of the fluorescence anisotropy detection system developed for affinity determination. Microdroplets are inserted in an inert tubing (yellow). The 10 – 20 nL droplets encapsulate the aforementioned concentration gradient. The microchannel device is made of a silicone polymer (1) and is bonded to a coverslip bottom (2), while the droplets in the microchannel are imaged by a microscope. C) Titration of BRC4 with three different HumRadA variants (colored data points) at constant 100 nM BRC4. K_D values were extracted from the fits (black lines) Adapted with permission from Gielen et al.^{[ 47 ]} Copyright 2017, ACS.