Fluorescence Lifetime Imaging Using a Confocal Laser Scanning Microscope

Transcription

1 SCANNING Vol. 14, (1992) OFAMS. Inc. Received: April 1, 1992 Fluorescence Lifetime Imaging Using a Confocal Laser Scanning Microscope E.P. BUURMAN,*? R. SANDERS,*? A. DRAAIJER," H.C. GERRITSEN,? J.J.F. VAN VEEN,* P.M. HOUR,* and Y.K. LEVINEt *TNO Institute for Environmental Sciences, Sensor Group, P.O. Box 601 1,2600 JA Delft;?Department of Molecular Biophysics, University of Utrecht, P.O. Box ,3508 TA Utrecht, The Netherlands Summary: The implementation of a fast fluorescence lifetime imaging method in a confocal laser scanning microscope is described. The set up utilizes a low-power continuous wave (CW) argon ion laser equipped with an electrooptic chopper producing nanosecond pulses with a repetition rate up to 25 MHz. A time-gated detection technique enables the measurement of the lifetime of a pixel in 40 ps. The first confocal fluorescence lifetime contrast images are presented. Application of fluorescence lifetime imaging in multilabelling experiments for discrimination between different labels with overlapping emission bands, for probing the local environment of a fluorescent molecule, and for quantitative fluorescence are discussed. Introduction Fluorescence microscopy is used widely in biological research for the detection and imaging of low levels of labelled material. Considerable technical improvements were achieved due to the recent development of confocal laser scanning microscopes (CLSM). The confocal technique enhances the lateral spatial resolution and simultaneously suppresses flare from out of focus planes in the sample. The higher depth resolution of confocal microscopes results in high contrast fluorescence images and in addition, affords the construction of three-dimensional (3- D) images. In this way fluorescent species can be localized within the sample. However, the fluorescence yield of Address for reprints: E.P. Buurman TNO-IMW Sensor Group P.O. Box JA Delft The Netherlands the species under study is affected by environmental factors such as the local ph, ion concentration, and quenching due to the presence of oxygen. These effects arise from the sensitivity of the fluorescence lifetime of a molecule to its chemical environment. The variation of fluorescence lifetime on chemical factors has been extensively studied using macroscopic samples. However, the combination of fluorescence microscopy and the time-resolved spectroscopy techniques opens the way for similar studies on microscopic volume elements. Indeed, on performing these measurement at different positions in the sample, it is possible to build up images reflecting the spatial variations of the fluorescence lifetime. This recently developed method (Lakowicz et al. 1992, Morgan et al. 1990, 1992, Ni and Melton 1991, Wang et al. 1990, 1991) usually is referred to as fluorescence lifetime imaging (FLI). Until now FLI studies have utilized conventional microscopes. This approach suffers from the drawback that the flare from out-of-focus fluorescence gives rise to a degradation of lifetime contrast. Since the lifetime variation is small in general, this limits the potential of the method for imaging small structures within the sample. These disadvantages can be overcome by utilizing a confocal geometry in lifetime imaging. Now only the fluorescence from the focal spot is detected so that the fluorescence lifetime of molecules localized in a small volume is monitored. The usefulness of such a setup is determined to a large extent by the time needed to determine the fluorescence lifetime of a single pixel. Unfortunately conventional methods for such lifetime measurements as timecorrelated single-photon counting and phase fluorimetry, are far too slow for recording images in a reasonable time. Here we describe the implementation of a fast fluorescence lifetime determination method in a confocal microscope. The setup utilizes a low-power continuous wave (CW) Argon laser equipped with an electro-optical chopper producing nanosecond pulses at a repetition rate of up to 25 MHz. A time-gated detection technique enables the measurement of the lifetime of a pixel in 40 ps. Consequently, an image consisting of 512x512 pixels can be captured in about 10 s.

2 I56 Scanning Vol. 14, 3 (1992) An advantage of our approach to FLI is that any CLSM can be easily modified in the same way to produce fluorescence lifetime images. Fast Lifetime Determination Method The determination of the fluorescence lifetime is based on the fast method described by Woods et al. (1984) and references therein. This technique relies on the acquisition of photons in two time windows at different delay times with respect to the excitation pulse (Fig. 1). For a monoexponential fluorescence decay, the lifetime r is defined by the ratio of the counts in the two windows: T = At log (NAl N, where AT is the time-offset between the two windows, A and B in Figure 1, and NA and NB are the corresponding numbers of counts. Equation (1) is valid for windows of equal width and provided the laser pulse is much shorter than the fluorescence lifetime. The observed signal decay is in fact a convolution of the laser pulse shape and the fluorescence decay. We note, however, that the intrinsic fluorescence decay is observed only at times longer than the width of the laser pulse. This arises simply because no transitions from the ground state to the excited state then take place and the fluorescence emission reflects directly the decay of the excited state population. Consequently, Eq. (1) is valid in the portion of the decay following the laser pulse. The foregoing neglects the distortion of the observed decay by the detector response because of the short transit time spread of our detection system (of the order of 150 ps). The error in the value of the fluorescence lifetime obtained with this technique strongly depends on the number of counts accumulated in the two windows. The error 1.00 A (1) propagation analysis of Ballew and Demas (1989) yields a relative error of 12% for a total count of 100 photons under optimal conditions where the width of each window is 2.5 XT. This relative error of 12% can be reduced to 9% on employing two time windows of unequal widths with the same number of counts (Buurman et al. 1992). This error estimation is obtained from computer simulations, as no simple analytical expression analogous to Eq.( 1) exists for the lifetime T. Rather, T is related to the number of counts NA and NB by a functional relation of the form: The function F as well as the error in T, depend on the setting of the windows. For a typical setting in our experiments and a total count of 100 photons a relative error of 9% in the lifetime is expected. This error falls to 3% if 1000 photons are detected. Instrumentation A commercially available confocal laser scanning microscope (CLSM) (Draaijer and Houpt 1988) was modified for lifetime imaging as shown schematically in Figure 2. Short laser pulses (1 ns) were produced by an electrooptic modulator (EOM) inserted in the laser beam prior to entering the microscope. The EOM consists of a Pockels cell placed between two crossed polarizers. The Pockels cell is driven by a pulse generator with a repetition rate of up to 25 MHz. The peak efficiency of this combination is 8%. The resulting power in the specimen depends on the efficiency of the microscope optics. The signal from a red-sensitive photomultiplier (rise time is 0.8 ns) with low dark count (500 cps) is fed to a discriminator (350 MHz pulse pair resolution) (Fig. 2). Two time-gated counter circuits are used to acquire the pulses from the discriminator. The time gates are synchronized with the pulse generator driving the EOM. All the time-critical electronics were built using ECL logic. The maximum count rate of the system is 35OX1O6 cps. A maximum count rate of 25X lo6 cps is used in order to limit pile-up effects to less than 1%. A 512x512 pixel lifetime image with 1000 laser pulses per pixel is obtained in about 10 s. Results FIG. 1 At time/r Method of rapid lifetime determination using two time win- dows. Window window B Confocal fluorescence lifetime images obtained with our set up are shown in Figures 3 and 4. Figure 3 shows an image of a capillary (1 mm dia.) filled with acridine orange dissolved in demineralized water. The photons were acquired using equal time windows each with a width of 1.8 ns. The number of detected photons per pixel is about 400.

3 E.P. Buurman et al.: Confocal fluorescence lifetime imaging 157 f i 1 b I 1 excitation generator I $5 fluorescence discriminator 1 FIG. 3 Confocal fluorescence images of a capillary (1 mm dia.) filled with acridine orange dissolved in demineralized water. The top left image gives the fluorescence intensity of the early time window which starts 1.5 ns after the laser pulse. The top right image shows the intensity in the second time window which starts 2 ns after the first one. The bottom left is the lifetime image whose greyvalue is proportional to the lifetime. Horizontal field width for each image = 1.6 mm. FIG. 2 Modifications to a confocal laser scanning microscope. The left top image in Figure 3 gives the fluorescence intensity of the early time window which starts 1.5 ns after the laser pulse. The top right image shows the intensity in the second time window which starts 2 ns after the first one. Figure 3 bottom left is the lifetime image whose greyvalue is proportional to the lifetime calculated using Eq.(2). It is noted that the intensity fluctuations in the top images in Figure 3 are canceled out in the lifetime contrast image, indicating a constant lifetime across the capillary. A lifetime of I.8 ns is obtained on averaging of the lifetimes in all the pixels. The standard deviation is 0.1 ns. The value of the lifetime for acridine orange in demineralized water found here is in excellent agreement with the value of 1.85(5) ns found by us using a standard time correlated single photon counting technique. Figure 4 gives confocal fluorescence images of the alga Gymnodinium. The cell wall of the alga was immunolabelled with FITC. Fluorescence imaging of FITC attached to these cells is hampered by the underlying fluorescence from chlorophyll molecules in the photosynthetic apparatus. The excitation light was suppressed by using a highpass filter with a cut-off wavelength of 510 nm, but which transmits the fluorescence of both chlorophyll and FITC. Equal time windows with a width of 1.8 ns were used and their intensities are shown in the top images of Figure 4. In the left image, the early window collects fluorescence light 1 ns after the laser pulse, while in the right one, the image of the second window is delayed by 1 ns relative to the first. The bottom image in Figure 4 is the fluorescence lifetime contrast whose greyvalue is proportional to the lifetime. FIG. 4 Confocal fluorescence images of the alga Gymnodinium. The excitation light was suppressed by using a high-pass filter with a cut-off wavelength of 510 nm, but which transmits the fluorescence of both chlorophyll and of FITC. The top left image is the image of the early window collecting fluorescence light 1 ns after the laser pulse, while the right one is the image of the second window delayed by 1 ns relative to the first. The bottom image in Figure 4 is the fluorescence lifetime image whose greyvalue is proportional to the lifetime. Horizontal field width for each image = 100 pm. The chloroplasts are visible in the lifetime image as darker areas, corresponding to a shorter lifetime. The lifetime, 0.7(1) ns, was determined by averaging over the chloroplasts. Haehnel et al. (1 983) report a triple-exponential decay for the fluorescence of chlorophyll in green algae. Two of the decay components have equal weights and lifetimes ca. 0.1 and 0.6 ns. The third component contributes only 14% to the fluorescence signal and has a

4 158 Scanning Vol. 14, 3 (1992) lifetime ca 1.4 ns. Only the 0.6 ns lifetime component is expected to contribute to the images under our experimental conditions, in good agreement with our findings. The brighter areas in the intensity image (Fig. 4 top left) located at the cell wall represent the FITC. The average lifetime of these FITC molecules is found to be 1.1 ( 1) ns. The lifetime of FITC molecules is very sensitive to their chemical environment and ranges from 1 ns to several ns on binding to proteins (Matko et al. 1992). Applications and Discussion We have shown confocal fluorescence lifetime imaging as a potentially valuable new tool in biological microscopy. This technique can be used in a number of ways not open to conventional microscopy. First, the lifetime contrast images provide a spatial discrimination of molecules with overlapping fluorescence emission bands, but with different fluorescence decays. This has important applications in multilabelling experiments. Alternatively, this can be used to visualise specific fluorescent species against a background of autofluorescence. Second, measurement of the fluorescence lifetime of molecules reveal the extent to which the quantum yield, and hence the emission intensity, is changed by their direct environment. This information can now be utilized to correct the observed fluorescence intensity, so that a quantitative determination of the concentration of the molecule in different regions of the specimen is feasible. These effects can also be used in a novel way to characterise the local environment of the fluorescing molecules. Multilabelling Using Fluorescence Lifetime for Discrimination Between Different Labels with Overlapping Emission Bands The most common technique in fluorescence microscopy for discriminating between distinct fluorescent species relies on differences in their emission spectra. Multilabelling experiments or the suppression of autofluorescence thus utilize optical filters to produce fluorescent images in different spectral regions. The fluorescence lifetime imaging presented here uses the fluorescence lifetime as the discriminating factor. The advantage of this approach is that it can be used to discriminate between fluorescent molecules with overlapping emission bands. The method of lifetime discrimination is illustrated in Figure 5 where the same data are used as in Figure 4. As noted above, the optical filter used to record the data transmitted both the autofluorescence from chlorophyll as well as that from the FITC molecules. The data collected contain information about both the fluorescence intensity and the fluorescence lifetime at every volume element, the voxel. The autofluorescence of chlorophyll and fluorescence of FITC can now be distinquished by making use of FIG. 5 Illustration of lifetime discrimination using the same data as in Fig. 4. The top left image reveals the chlorophyll fluorescence intensity because only pixels with a lifetime shorter than 0.9 ns are shown. The top right image shows the FITC fluorescence by only including those pixels with a lifetime greater than 0.9 ns. For comparison two images obtained with the conventional method of using specific optical filters are shown at the bottom. For the left image an optical filter for chlorophyll was used. For the right image a bandpass filter was used to detect the FITC fluorescence. Horizontal field width for each image = 100 wm. the lifetime data. This is done by imposing a lifetime mask on the intensity image. The result is shown in the top images in Figure 5. The top left image reveals the chlorophyll fluorescence intensity because only pixels with a lifetime shorter than 0.9 ns are shown. The top right image shows the FITC fluorescence by including only those pixels with a lifetime greater than 0.9 ns. For purposes of comparison, two images obtained with the conventional method of using specific optical filters are shown in the bottom of Figure 5. For chlorophyll (left image) an optical high-pass filter with a cut-off wavelength of 580 nm was used. For the right image a band-pass filter with a bandwidth of 10 nm and a center wavelength of 540 nm was used to detect the FITC fluorescence. The excellent agreement between the two sets of images illustrates the power of the lifetime contrast method. Quantitative Fluorescence The intensity of fluorescence is proportional not only to the concentration C of a fluorescent molecule but also to its quantum efficiency Q (Lakowicz 1983): with Q=- 2 (3) (4)

5 E.P. Buurman et al. : Confocal fluorescence lifetime imaging 159 Where Iem is the fluorescence intensity, Iex the excitation intensity, T the actual lifetime, and TO the intrinsic lifetime. A quantitative determination of the spatial concentration profile of fluorescent molecules in a sample is only possible if the quantum efficiency Q is independent of the local environment of the molecule. Consequently, if the molecular fluorescence is quenched, with a concomitant reduction in the intensity, the concentration of the molecules is underestimated. On the other hand, knowledge of both the lifetime and the intensity can be used in a straightforward way to correct for this effect, using Eq. (3). An image in which the greyvalue is proportional to the concentration, can be derived by a pixel by pixel division of the intensity image by the lifetime image. Lifetime as a Probe to Monitor the Environment of a Fluorescent Molecule In the preceding dlscussion, we have noted that the fluorescence emission spectra and lifetimes of molecules are determined by their local chemical environment. Thus once these effects are characterized in an in vitro experiment, they can be used to probe the local environment of a fluorescent molecule in a biological cell and monitor any changes in that environment brought about by cellular activity. A limitation of the instrument described is that the fluorescence decay is assumed to be effectively a single exponential decay. For molecules exhibiting a multiexponential fluorescence decay our method gives an average lifetime whose value depends on the timing of the gates. However, in this case the results can be calibrated against the computer simulations which generate Eq. (2). Nevertheless, the lifetimes and pre-exponential factors characterizing the multiexponential decays must be known. It is important to note that our instrument can in fact be extended to deal with multiexponential decays directly, through the addition of more time windows. Acknowledgments E. Vrieling (Department of Marine Biology, University of Groningen, The Netherlands) is thanked for preparation of the algae. This research is supported by the Technical Foundation (STW) of The Netherlands, grant number UNS References Ballew RM, Demas JN: An error analysis of the rapid lifetime determination method for the evaluation of single exponential decays. Anal Chem 61,30-33 (1989) Buurman EP, Gemtsen HC, Draaijer A: Numerical rapid lifetime determination. Appl Spectrosc (submitted) 1992 Draaijer A, Houpt PM: A standard video-rate confocal laser-scanning reflection and fluorescence microscope. Scanning 10, (1988) Haehnel W, Holzwarth AR, Wendler J: Picosecond fluorescence kinetics and energy transfer in the antenna chlorophylls of green algae. Photochem and Photobiol37, (1983) Lakowicz JR: Principles of Fluorescence Spectroscopy. Plenum Press, New York (1983) Lakowicz JR, Szmacinski H, Nowaczyk K: Fluorescence lifetime imaging. Proc Nut1 Acad Sci (USA) 89, (1992) Matko J, Ohki K, Edidin M: Luminescence quenching by nitoxide spin labels in aqueous solution: Studies of the mechanism of quenching. Biochemistry 31, (1992) Morgan CG, Mitchell AC, Murray JG: Nanosecond time-resolved fluorescence microscopy: Principles and practice. Trans R Microsc Soc 1, (1990) Morgan CG, Mitchell AC, Murray JG: Prospects for confocal imaging based on nanosecond fluorescence decay time. J Microsc 165,4940 (1992) Ni T, Melton LN: Fluorescence lifetime imaging: An approach for fuel equivalence ratio imaging. Appl Spectrosc 45, (1991) Wang XF, Kitajima S, Uchida T, Coleman DM, Minami S: Time resolved fluorescence microscopy using multichannel photon counting. Appl Spectrosc 44,25-30 (1990) Wang XF, Uchida T, Coleman DM, Minami S: A two-dimensional fluorescence lifetime imaging system using a gated image intensifier. AppZ Spectrosc 45, (1991) Woods RJ, Scypinski S, Cline Love LJ, Ashworth HA: Transient digitizer for the determination of microsecond luminescence lifetimes. Anal Chem 56, (1984)