Investigating protein-protein interactions in living cells using fluorescence lifetime imaging microscopy

Abstract

Fluorescence lifetime imaging microscopy (FLIM) is now routinely used for dynamic measurements of signaling events inside living cells, including detection of protein-protein interactions. An understanding of the basic physics of fluorescence lifetime measurements is required to use this technique. In this protocol, we describe both the time-correlated single photon counting and the frequency-domain methods for FLIM data acquisition and analysis. We describe calibration of both FLIM systems, and demonstrate how they are used to measure the quenched donor fluorescence lifetime that results from Förster resonance energy transfer (FRET). We then show how the FLIM-FRET methods are used to detect the dimerization of the transcription factor CCAAT/enhancer binding protein-α in live mouse pituitary cell nuclei. Notably, the factors required for accurate determination and reproducibility of lifetime measurements are described. With either method, the entire protocol including specimen preparation, imaging and data analysis takes ∼2 d.

Figure 1

TCSPC FLIM setup. TCSPC FLIM requires a pulsed excitation source. In our system, a multiphoton laser (repetition rate = 78 MHz, T = 12.82 ns, pulse width < 150 fs) is used. The pulsed laser is coupled to the Bio-Rad Radiance 2100 scanning system that is attached to a Nikon TE300 microscope. Thus, spatial information of a specimen is obtained by an XY raster scanning mechanism, and is also available in Z for optical sections. Photons emitted from the specimen pass through the emission (EM) filter and are detected by a fast detector (timing jitter: 25–300 ps). A photomultiplier tube (PMT) with a response time of ~150 ps is placed in the side port of the microscope (nondescanned detection) and is connected to the TCSPC device. The TCSPC device contains PC plug-in boards (commercially available) that function as the time-to-amplitude converter, time-to-digital converter, discriminator and multichannel analyzer. The TCSPC device synchronizes the detector to the excitation pulse and records the arrival time and spatial information for the detected photons. A reference (REF) signal is acquired from a glass cover slip (GC) reflecting 4% of the excitation light, and this reference beam (RB) is passed to a photodiode (PD) that is connected to the TCSPC device. Given a time period (a few seconds or minutes) of accumulating emitted photons for thousands or millions of excitation pulses, a ‘photon counts’ histogram is built for each pixel of an image. The fluorescence lifetime at each pixel can be estimated by fitting the corresponding photon counts (decay) data into a single- or multiexponential model.

Figure 2

Frequency-domain (FD) FLIM setup. FD FLIM uses a modulated light source to excite the specimen. The basic principle of the FD FLIM method is illustrated, showing the phase delay (Φ) and modulation ratio (m = (B/A)/(b/a)) of the emission (Em, dashed sinusoidal curve) relative to the excitation (Ex, solid sinusoidal curve), which are used to estimate the fluorescence lifetime. For cyan fluorescent proteins, a 440-nm pulsed-diode laser is directly modulated by the ISS FastFLIM module at the fundamental frequency of 20 MHz with several sinusoidal harmonics. The modulated laser is coupled to the ISS scanning system attached to an Olympus IX71 microscope to scan specimens. Photons emitted from the specimen travel through the scanning system (descanned detection) and are then routed by a beam splitter (BS) through emission (EM) filters to two identical avalanche photodiodes (APD) detectors. The phase delays and modulation ratios of the emission relative to the excitation are measured at up to seven modulation frequencies (ω = 20–140 MHz) for each XY raster scanning location, and are also available in Z by changing focus. The excitation profile is obtained through a calibration procedure in which a reference fluorophore of known lifetime is used.

Figure 3

Photobleaching affects the accuracy of FLIM measurements. Live cells that only express the donor (here, Cerulean-bZip) are used to establish a suitable excitation power level for the TCSPC FLIM-FRET measurements. The amount of photobleaching can be evaluated by monitoring the intensity changes in selected cell regions over time. (a) Shown is a typical photobleaching level at the excitation power level used in the TCSPC FLIM-FRET study (~1.2 mW at the specimen plane). (b) The increased photobleaching when 250% (2.5-fold) of the excitation power level (~ 3 mW at the specimen plane) is used. (c) The effect of the increased photobleaching on the FLIM measurements is shown, demonstrating that the decay profile for Cerulean-bZip is markedly shifted by photobleaching. The mean lifetime obtained from the FLIM data acquired at the high laser power level (~1.3 ns) is much shorter than the normal value (~2.75 ns).

Figure 4

TCSPC FLIM data representation of the FRET standard. (a) The representative Coumarin 6, C5A (donor-alone control) and C5V (FRET standard) raw decay data and corresponding fitting curves are shown. Fittings were carried out using the measured instrument response function (IRF) with a full width at half maximum of ~300 ps. The fluorescent lifetime decay kinetics for Cerulean in the C5A and C5V constructs was determined through the comparisons of fitting the decay data into both single- and double-exponential decay models. There was little difference observed between the single- and double-exponential fits of the C5A decay data, confirmed by the calculated χ²s—1.08 (single) versus 1.07 (double). However, the C5V FRET standard was better represented by the double-exponential fit (χ² = 1.07) compared with the single-exponential fit (χ² = 2.38). The results clearly show the quenched state of Cerulean (the donor) in the presence of Venus (the acceptor) resulting from FRET. (b) The lifetime distributions in representative Coumarin 6, C5A or C5V lifetime (overlaid with intensity) images are shown (the apparent lifetime for C5V, see equation (2)). Scale bar, 10 μm.

Figure 5

FD FLIM data representation of the FRET standard. (a) The representative HPTS, Coumarin 6, C5A (donor-alone control) and C5V (FRET standard) data measured at the fundamental modulation frequency (20 MHz) are displayed in the phasor plot. Both HPTS and Coumarin 6 are expected to have a single lifetime component, and this is clearly shown by the phasor plot, in which the lifetime distributions are centered on the universal semicircle of the phasor plot. This is also observed for the lifetime distribution for the cell expressing the C5A fusion protein. (b) However, the lifetime distribution for the cell expressing the C5V fusion protein falls inside the semicircle, and is better fitted with a biexponential model than a monoexponential model, as shown. As HPTS (~5.3 ns) has a much longer lifetime than Coumarin 6 (~2.5 ns), the two populations in the phasor plot are well separated. The phasor plot clearly demonstrates the quenching of the donor (Cerulean) in the C5V protein compared with the unquenched donor in C5A. Scale bar, 10 μm.

Figure 6

Localization of dimerized C/EBPα-bZip in living cell nucleus using TCSPC FLIM-FRET microscopy. C/EBPα-bZip was tagged with either Cerulean (C) or Venus (V). The fluorescent lifetime decay kinetics for the C–bZip (FRET donor) in the absence and the presence of V–bZip (FRET acceptor) was determined by fitting the measured decay data into a single or double-exponential decay model, respectively, with the measured instrument response function (IRF). Use of a threshold allowed fitting to the pixels in regions of centromeric heterochromatin of the cell nucleus. The comparison between the representative measured decay data points, the fitting curves and the lifetime (overlaid with intensity) images of the two cases clearly shows that the C in cells expressing both C–bZip and V–bZip decayed faster (or has a shorter lifetime) than that in cells expressing C–bZip alone, indicating that the C attached to bZip was quenched by the V attached to bZip because of FRET. Scale bar, 10 μm.

Figure 7

Investigation of the dimerization of C/EBPα-bZip in living cell nucleus using FD FLIM-FRET microscopy. bZip was tagged with either Cerulean (C) or Venus (V). (a) The intensity and lifetime images of representative cells, which only express C–bZip (donor-alone control) and that co-express C–bZip (FRET donor) and V–bZip (FRET acceptor), are compared. The FD FLIM data acquired at the fundamental modulation frequency (20 MHz) is displayed on the phasor plot. The comparison demonstrates a shorter lifetime of Cerulean in the cell that expresses both C–bZip and V–bZip. The lifetimes of C–bZip in the absence and the presence of V–bZip were estimated by single- and double-exponential fittings, respectively. (b) The average donor lifetime, obtained from ten cells that expressed only C–bZip, was 3.15 ns (indicated by the black dot). The apparent lifetimes for 80 regions of interest (ROIs) identified in ten cells coexpressing C–bZip and V–bZip were then determined, and the range was from 2.5 to 3.05 ns, resulting in a variety of energy transfer efficiencies (E) calculated on the basis of equation (3). To investigate how the quenched C–bZip lifetimes were influenced by the acceptor-to-donor ratio, we roughly determined the ratio using the intensities obtained in the acceptor and donor channels for each ROI. With all 80 ROIs, the lifetime (blue dots with a dark blue trend line: R = 0.35) or E (red triangles with a dark red trend line: R = 0.19) shows a negative or positive dependency on the acceptor-to-donor ratio, respectively (scale bar, 10 μm).