Frequency domain time-resolved microscope using a fast-scan CCD camera

Time-resolved fluorescence imaging can enhance the contrast of microscope images and it can also provide important information about the micro-environment in cellular systems. We have developed a fluorescence microscope which can measure fluorescence lifetimes over an entire image. Fluorescence lifetimes are measured by using heterodyne frequency domain techniques. Heterodyning is accomplished by using an intensity modulated laser light source and a fast scan CCD camera coupled with a gain modulated microchannel plate as the detector. The high duty cycle of this method allows us to generate a phase resolved image with about five seconds integration time. Operating in the fast scan mode, the systematic uncertainties in lifetime determination caused by photobleaching are less severe than those of slow-scan cameras. The microchannel plate can be modulated at frequencies up to 300 MHz, which allows us to measure lifetimes as short as 500 ps with resolution of 50 ps. The modulation of the microchannel plate only slightly degrades the spatial resolution of the image from the diffraction limit; 0.8 micron resolution is maintained with 500 nm laser excitation.

Time-resolved fluorescence microscopy offers many unique advantages over conventional techniques and has seen rapid growth in the past few years.1'2 The fluorescence lifetime is a sensitive measure of the micro-environment of the probe molecule. A time-resolved fluorescence microscope can utilize this sensitivity to measure important chemical information such as pH, calcium or magnesium content3, and oxygen concentration4 in living cells. Since fluorescence lifetime is an intrinsic property of the probe molecule, it is insensitive to the unknown local probe concentration in the cell. Furthermore, the lifetime imaging technique can be used to selectively enhance the image contrast in order to highlight cellular features of interest by picking the proper "time window". The development of time-resolved microscopy also makes powerful techniques, such as anisotropy decay and resonance energy transfer5, available for the study of cellular systems. Molecular rotations at picosecond time scale and molecular interactions at the atomic scales can be studied.
We have developed a CCD camera based frequency domain time-resolved microscope. The CCD camera is operated in the fast-scan mode which is essential for minimizing the error in lifetime measurement where significant photobleaching commonly occurs. In this article, we will provide a brief description of the operating principle and mechanics of this time-resolved microscope. Some preliminary results on time-resolved cellular imaging will also be presented.

THEORY AND MECHANTCS OF OPERATION
This time-resolved camera is based on the frequency domain heterodyne technique.6 Heterodyne detection requires that the excitation light source be smusoidally modulated. The resultant sample fluorescence will also be modulated. The lifetime information of the sample is encoded in the phase shift (c1) and the demodulation (M) (Fig. 1). where CL) is the angular modulation frequency and 't is the fluorescence lifetime. Since the fluorescence lifetime is on the order of nanoseconds, the relevant modulation frequency is in the range of 106 i09 Hz. The accurate measurement of the phase shift and demodulation at high frequencies is difficult. The use of heterodyning technique translates this information to a low frequency signal. The fluorescence signal is collected by a detector which is gain modulated at a frequency of (w + Aco) where the crosscorrelation frequency, M, is typically tens of Hz. The fluorescence signal and the gain modulation of the detector are mixed and the output signal is a product of them. The resultant signal can be decomposed into a sum of two sinusoidal components with modulation frequencies, 2o+Aco and Aw. The low frequency signal (Ew) can be easily isolated by low pass filtering. This low frequency signal contains the same phase and demodulation information as the original signal. This low frequency signal can be accurately measured by digital techniques.
The schematic of the time resolved camera is shown on Fig. 2 7,8 The light source used for these experiments is either a Spectra Physics argon-ion laser which is frequency modulated with an acoustic-optical modulator or a Coherent Pyridine I dye laser pumped by a mode-locked YAG laser. The excitation signal is channeled to the sample via the epi-luminescence beam path of a Zeiss Axiovert 35 microscope (Zeiss Inc., Thronwood, NY). The fluorescence signal is detected by a gain modulated dual stage microchannel plate detector (MCP) (ITT' type F4144). The microchannel plate is modulated at a frequency equal to the laser excitation frequency plus a small cross correlation frequency which we have chosen to be 7 Hz. The MCP is the mixing stage in the heterodyning scheme. We have modulated the MCP by providing about 50V sinusoidal signal at its photocathode. We have obtained reasonable modulation up to 300 MHz using this method. The modulation of the photocathode causes some degradation of the spatial resolution of the MCP8. We are currently testing an alternative modulation SPIE Vol. 2137/85 Ziess 35 Axiovert Microscope Fig. 2. Schematics of the frequency domain time-resolved microscope method which varies the high voltage gain across the channels9. Since the response of the MCP output phosphorous screen is slow, it serves as an effective low pass filter. The MCP is fiber-optically coupled to the CCD camera (Cohu Inc., model 4812-2000/ER). The 7 Hz cross-correlated signal is sampled by the CCD operating at a sampling rate of 28 Hz which corresponds to sampling four points per wave form (Fig. 3). The sample rate is twice the rate required by the Shannon's sampling theorem for suppressing the higher harmonic contributions. Successive waveforms are synchronously digitized (at 8 bit resolution), and synchronously averaged (at 16 bit resolution) by a video rate digitizer and DSP (Matrox Electronic System Ltd., Image-1280) which is housed in a PC compatible computer (Gateway 2000, 486/66). This form of synchronous averaging greatly suppresses non-harmonic noise and increases the system's dynamic range. Software has been developed to automate the data acquisition process and to compute the frequency domain images with a Fast Fourier transform.  (1), we can selectivly image either the free or bound population of the dye. At 80 MHz, the nuclei of the onion cells are highlighted (Fig. 6a) since the much longer lived bound dye has a much larger phase shift comparing to the free dye. We can alternatively emphasize the free dye by looking at the modulation resolved picture (Fig. 6b) where the nuclei became invisible as it is demodulated at high frequency. At 2 MHz (Fig. 6 c,d), both the dye population can be imaged as they have similar phase and modulation values. The phase picture appears fuzzy at low frequencies because the phase shift of the dye fluorescence is small comparable to the phase of the scattered light.
We have begun preliminary studies on the ability of this time-resolve microscope to obtain accurate quantitative lifetime data. For this study, we have constructed a two layer sample slide. In one layer, we filled the slide with 15 jim blue FluoSphere (Molecular Probe, Eugene, OR) in water. The second layer was filled with POPOP-ethanol solution which serves as an internal fluorescence lifetime reference (1.3 nsec). We did not mix the two solution together because the dye solution can impregnate the latex spheres and alter their lifetimes. Frequency domain pictures are taken at multiple frequencies from 7 to 120 MHz. Only the phase pictures were used in this measurement since the modulation pictures are typically much noisier (Fig. 6d). The phase differences between the sphere and the dye background are measured in each frequency. Using the known lifetime of the POPOP background, the absolute phase shift of the sphere can be determined (Fig. 7).
These phase values are analyzed with a global analysis routine (Globals Unlimited, Urbana, IL).
The data obtained appears to best fit a double exponential decay instead of a single exponential decay as indicated by a three fold reduction in 2 value. The major component (64%) has a lifetime of 1 .3 nsec which is fixed in the fit. The minor component has a lifetime of 2.3 nsec. The double exponential decay scheme is expected since the signal at the pixels occupied by the spheres are mixed with a significant faction of out of focus fluorescence of the dye solution. The major component lifetime is therefore fixed to the lifetime of the reference POPOP solution. The minor component's lifetime is fitted to be 2.3 nsec SPIE Vol. 2137/89 10 100 which is slightly short comparing to the lifetime measured by a conventional flurometer (3.8 nsec). A large uncertainty in the minor component's lifetime is expected given the lack of good modulation data. The intensity faction ratio of the two lifetime components compares favorably (within 5%) with the measured intensity ratio of the spheres and the dye in the DC intensity picture. The results of these fitting schemes are summarized in Table 1: One of the most important features of the lifetime microscope is its ability to measure minute differences in cellular micro-environments where fluorescence intensity measurements do not have sufficient sensitivity. We have imaged a string of live spirogyra cells, a common pond algae, to test our instrument in a simple cellular system. Spirogyra cells are linked together like a string of pearls. The most prominent feature in each of these cells is the large, double helical chioroplast. The chlorophyll in this organ is extremely fluorescent when illuminated by the 514 nm line of the Argon-ion laser. A string of these spirogyra cells were imaged with the time-resolved microscope integrated for about 300 sec with the laser power on the order of 20 mW. The fluorescence intensity image is shown in Fig. 8a. This image was taken at the junction between two distinct spirogyra cells. The break in the helical structure corresponds to the separating cell walls between the two cells. The cell walls are not visible since they are not fluorescent. Judging from the fluorescence intensity picture alone, these two cells appears to be identical. However, a significant difference in chlorophyll lifetimes in these two cells can be observed in the phase and modulation resolved pictures (Fig. 8c,d). Note that the chioroplast within each individual cell has a uniform lifetime which is distinct from the other cell. This observation shows that the nonuniformity in the phase and modulation pictures truly reflects the environment of the chioroplast in the cells and is not caused by experimental artifacts of the lifetime camera. Comparing the phase and modulation resolved pictures, the cell at the upper right hand corner has a higher phase value and is more demodulated while the other cell has a lower phase value and is better modulated. This is consistent with the hypothesis that the cell at the upper right hand corner has a longer lifetime. Since we have not attempted to precisely control the cellular environment of these cells, we will not speculate as to the cause of this lifetime difference. This experiment demonstrates the promise of the time-resolved microscopy technique in quantitatively monitoring cellular functions. We have successfully adapted the fast-scan time-resolved CCD camera developed in our laboratory to a microscope system. Time-resolved imaging can not only effectively and selectively enhance the contrast of microscope fluorescence images, but can also quantitatively measure lifetimes within cellular compartments to monitor their micro-environment. We are currently working on improving the signal-to-noise ratio of our instrument in order to minimize the data acquisition time which is critical in biological systems where photo-damage and photo-bleaching is significant.