3.1.

Multigate Detection

In the simplest case of time-domain lifetime imaging, a gated microchannel plate image intensifier is used in conjunction with a CCD imaging camera.^{23 24} By gating the image intensifier at different intervals (with a specified time window) along the exponential fluorescence decay profile, one can obtain a set of images of the actual decay of fluorescence during the excited-state lifetime [Fig. 1(b)]. It has been found that the photon utilization and time resolution of the multigate detection approach are still limited by the detector’s performance. One also cannot avoid background noise arising from scattered fluorescence emission when an imaging (area) detector is used in multigate detection. Regardless of these limitations, the multigate detection FLIM methodology has been found to be very successful with single-photon excitation in recent years.^{25 26} For multiphoton applications, it is imperative to use point-scanning detectors to eliminate background noise.

3.2.

Single-Photon Counting

Photon-counting detectors are well suited for low-light level detection as well as for providing quantized pulses for every photon event.^{27 28} This makes the measurement of lifetimes more accurate. However, these detectors suffer from poor dynamic range compared with their area detector counterparts. Regardless of this minor disadvantage, photon-counting devices have been found to be very reliable in terms of photon economy, rapid lifetime determination, and high temporal resolution. The limiting factor in achieving good temporal resolution is the transition-time spread (TTS) of the fast photomultiplier tubes (PMTs) used in these systems.

3.3.

Streak-FLIM Imaging System

The essential feature of the streak camera is that it converts an optical 2-D image with spatial axes (x,y) into a streak image with temporal information and with the axes (x,t). A simplified picture of the streak imaging principle is demonstrated in (Fig. 2 and Fig. 3) and a brief description of the system is given in the following sections. For the complete technical details, see Ref. 30.

Figure 2

The streak-FLIM imaging system is composed of the laser system (Coherent Mira 900 and Coherent Pulsepicker 9200), optics (Olympus IX-70 inverted microscope, FLIM optics), and the streak system (trigger unit, Streakscope, HiSCA Argus camera).

Figure 3

A schematic of the principle of streak imaging. (a) A single line in the optical image with intensity information at every point is converted by the Streakscope into a streak image (b) with spatial information as the horizontal axis and time as the vertical axis. Every point (containing fluorescent molecules) along this line has an exponential decay profile as shown in (c). By scanning along the entire optical image (x,y plane), similar exponential decay profiles are produced that correspond to every point in the entire plane. These decay profiles are stored and then numerically processed to give a lifetime image as shown in (d). See text for more technical details.

3.4.

System Calibration

Standard solutions of various fluorescent dyes were used for calibration of the system and for optimizing the relevant parameters for a typical measurement. Rhodamine 6G and rose bengal solutions were prepared in different solvents as detailed in Table 1. These dyes are reported to have monoexponential fluorescent decays and therefore display a single lifetime.²⁹ The choice of standard dyes also allowed calibration of the system for measurement of lifetime values ranging from a few hundred picoseconds to a few nanoseconds. Table 1 gives the calculated lifetime values for these solutions obtained with the FLIM streakscope system and compares these values with values reported in the literature for the same fluorophore solutions. There is a very good agreement between the calculated lifetimes and those in the literature, demonstrating the accuracy of lifetime determination of the streak-FLIM system. Accuracy of the lifetime measurements is largely governed by the uniformity of the FLIM images. Lifetime images of the standard solutions that we employed exhibited a high degree of uniformity and thus provided high accuracy in the calculations of lifetime. Reproducibility between two measurements in the same specimen was found to be >99 in standard solutions. However, for cellular specimens, the variability in cellular expressions as well as the inevitable photobleaching artifacts can restrict the reproducibility. A more detailed description of the instrument’s performance and signal and noise issues can be found in Ref. 30.

Table 1

3.5.

Biological Applications

Cyan and yellow fluorescent proteins (CFP/YFP) are mutated forms of the wild-type green fluorescent protein and are employed as a donor-acceptor pair in many FRET experiments.^{31 32} Energy transfer can be manifested in many ways, such as a reduction in donor emission fluorescence, sensitized fluorescence emission of the acceptor, or a reduction in donor lifetime. Of all the different FRET methods, measuring FRET by measuring the changes in donor lifetime has been reported to be more reliable owing to its relative insensitivity to the spectroscopic artifacts commonly encountered in intensity-based measurements. In this section we give a typical cellular application of the streak-FLIM system in studying caspase activity in single cells. Baby hamster kidney (BHK) cells were transiently transfected with a mitochondrially targeted DNA construct (mC2Y) that links CFP and YFP by an amino acid encoding a caspase-2 (VDVAD) recognition sequence.³³ Under identical conditions of growth and transfection, a batch of cells was exposed to the broad-range caspase activator tert-butyl hydrogen peroxide (tBOOH: 50 μM) for 4 h at 37 °C in order to induce oxidative stress in mitochondria. Experiments were carried out under identical imaging conditions for all of the specimens. It is expected that the close proximity of CFP and YFP molecules in the intact caspase-2 substrate would lead to significant energy transfer in the mC2Y specimens. However, induction of caspase-2 activity in tBOOH -treated mC2Y specimens should lead to cleavage of the substrate, thereby altering the spatial relationship between the CFP and YFP molecules and decreasing the extent of energy transfer in these specimens. We measured lifetimes of the CFP in both the tBOOH -treated and untreated cells and calculated the change in energy transfer efficiency in these two situations.

For measurements of CFP fluorescence lifetime, the femtosecond-pulsed laser was tuned to 840 nm. To select for emission from CFP only, the fluorescent signal passed through a 480/30 emission filter (Chroma Inc.) placed before the Streakscope. Donor lifetime images were calculated using AquaCosmos software. Figure 4 shows the (CFP) lifetime images of the mC2Y and mC2Y+tBOOH specimens. The mC2Y specimen shows punctate mitochondrial expression of CFP, whereas there is an observable morphological change in the tBOOH -treated specimen. This morphological change (the cell becomes more rounded), together with the diffuse fluorescence in Fig. 4(b), can be attributed to oxidative stress induced by tBOOH and the consequent induction of apoptosis in the cell. As can be seen from the lifetime histogram, the mean lifetime of the mC2Y specimen is about 2 ns, whereas that of the tBOOH -treated mC2Y specimen is about 2.28 ns. This increase in lifetime is a clear manifestation of loss of energy transfer between CFP and YFP molecules in the tBOOH -treated mC2Yspecimen compared with the untreated mC2Y specimen, where the molecules are held intact by the intervening caspase-2 peptide sequence. This indicates that a significant mitochondrial caspase activity is induced by oxidative stress. As a positive control for FRET measurements, we used DNA constructs encoding mitochondrially targeted CFP-polyglycine-YFP (mCGY) in BHK fibroblast cells and imaged under conditions identical to those for our experimental studies (mC2Y). Both these specimens exhibited average lifetimes of ∼2.0 ns. Upon treatment with caspase activator, tBOOH, only the mC2Y specimen showed a 15 increase in average lifetime, whereas the mCGY specimen showed less than a 5 change in lifetime. This proved that what we see as a change in lifetime in the mC2Y specimen upon tBOOH treatment is essentially due to caspase activity. It is also intriguing to note that the lifetime distribution in the tBOOH -treated specimen is much broader than that of the untreated mC2Y specimen. This can be interpreted as representing nonuniform caspase cleavage activity within the cell.

Figure 4

Fluorescence lifetime images of baby hamster kidney cells transfected with a mitochondrially targeted DNA construct (mC2Y) that links CFP and YFP via an amino acid sequence encoding a caspase-2 recognition site. (a) untreated mC2Y; (b) mC2Y treated with 50 M tBOOH for 4 h; (c) lifetime histogram for both specimens. The laser output from Mira 900 (76-MHz repetition rate; 840-nm wavelength; ∼200-fs pulse width) is rendered to the pulse picker, which steps down the repetition rate to 500 kHz optimized for triggering the Streakscope. The average power at the entrance of the FLIM optics is 8 mW. The 12-bit streak images were obtained in the 1×32 binning condition. Exposure times for imaging (∼195 ms per single line) were optimized to minimize photobleaching during data acquisition. The lifetime images were calculated from the raw streak images by AquaCosmos software (Hamamatsu). An intensity threshold criterion of about 10 of the maximum dynamic range was used in calculation of lifetime images, and the final images were median filtered. The lifetime histograms shown were plotted for the entire cells. The mean lifetime of the mC2Y specimen was 1.98 ns, whereas that of tBOOH -treated mC2Y specimen was 2.28 ns.

Some of the caspase-2 substrates were cleaved when the lifetime measurement was made, while other substrates remained intact. This would cause a wide distribution in the extent of energy transfer spatially within the cell and manifest itself as a wider distribution in the measured lifetime in the tBOOH -treated specimen. Although it would be interesting to analyze these histograms by multicomponent lifetime models, only a single-component Gaussian model was adopted in the present analysis. Furthermore, at 4 h post- tBOOH treatment, only about 50 of the cells showed changes in lifetime upon tBOOH treatment. Different cells in an ensemble would be at different stages of apoptosis at any one time. Since the fixation of the cells was done at a single time point after the tBOOH treatment, it is expected that not all the cells would exhibit apoptotic changes in lifetime measurement. It is also possible that there exists a substantial population of uncleaved caspase-2 substrate, even in the tBOOH -treated specimen because this substrate is being constitutively expressed.

The energy transfer efficiency can be calculated from the expression E=1−(τ _DA /τ _D ), where τ _DA and τ _D are the lifetimes of the donor in the presence and in the absence of acceptor, respectively. In the present case, τ _DA and τ _D represent the mean lifetimes of CFP in the untreated and tBOOH -treated specimens, respectively. From the fitting of mean lifetime histograms in both the specimens, the change in energy transfer efficiency in the above example was calculated to be ∼15.

We have taken care to ensure that the calculated lifetime values are not affected by photobleaching artifacts or autofluorescence contributions. Imaging conditions were optimized for minimal photobleaching so as to get reproducible lifetime values from the same region of the cell in consecutive scans. It was observed earlier that autofluorescence from fibroblast cells has typical lifetime values of about 2.2 ns.¹⁵ We measured the streak images of nontransfected cells and found that the maximum contribution to fluorescence intensity from autofluorescence is less than 5 of that observed in transfected cells. We have thresholded the streak-FLIM image intensity values above this value in our lifetime calculations in order to remove the ambiguity of a possible contribution of autofluorescence to the observed lifetime. It is also worth mentioning that autofluorescence contributions have been found to be less using multiphoton excitation.³⁴

The demonstration of reliable FRET measurement using multiphoton lifetime imaging combined with a streak camera opens up new possibilities for quantitative imaging of protein–protein interactions in living cellular systems. Since the requirement is to measure only donor lifetime in all the specimens, the problem of spectral crosstalk is overcome. However, it is advisable to measure the lifetime of a donor in the absence of an acceptor under identical imaging conditions that can then serve as a control for non-FRET conditions.

In recent years there has been an upsurge in the quantitative imaging of molecular interactions in addition to the intensity-dependent visualization of intracellular probes. Improvements in fluorescence techniques such as confocal, multiphoton microscopy and other energy transfer methods have yielded valuable spatial and spectral information pertinent to submicron molecular interactions. Lifetime imaging methods offer improved temporal resolution in fluorescence imaging, thereby incorporating an additional experimental parameter. Although single-photon FLIM techniques have been in use for many years, it is expected that the combination of multiphoton excitation and FLIM methods will yield an enhancement in spatiotemporal resolution and reduce overall phototoxicity and thereby improve our understanding of molecular interactions.