Multiphoton fluorescence microscopy.

Multiphoton fluorescence microscopy has now become a relatively common tool among biophysicists and biologists. The intrinsic sectioning achievable by multiphoton excitation provides a simple means to excite a small volume inside cells and tissues. Multiphoton microscopes have a simplified optical path in the emission side due to the lack of an emission pinhole, which is necessary with normal confocal microscopes. This article illustrates examples in which this advantage in the simplified optics is exploited to achieve a new type of measurements. First, dual-emission wavelength measurements are used to identify regions of different phase domains in giant vesicles and to perform fluctuation experiments at specific locations in the membrane. Second, we show how dual-wavelength measurements are used in conjunction with scanning fluctuation analysis to measure the changes in the geometry of the domains and the incipient formation of gel domains when the temperature of the giant vesicles is gradually lowered.

In the context of single-point measurements in con-Multiphoton fluorescence microscopy has now become a relatively junction with image analysis, the most striking difference common tool among biophysicists and biologists. The intrinsic secbetween a conventional confocal system and a multiphotioning achievable by multiphoton excitation provides a simple ton microscope is that in the latter case the emitted light means to excite a small volume inside cells and tissues. Multiphoton does not need to travel through the confocal pinhole to microscopes have a simplified optical path in the emission side due to the lack of an emission pinhole, which is necessary with normal reach the detector. This different optical configuration confocal microscopes. This article illustrates examples in which this makes it possible, with relatively simple optics, to use advantage in the simplified optics is exploited to achieve a new type several colors in the emission arm without being affected of measurements. First, dual-emission wavelength measurements by color aberrations of the objective. This alone is a disare used to identify regions of different phase domains in giant tinct advantage. We also note that multiphoton excitation vesicles and to perform fluctuation experiments at specific locations is particularly good at exciting several common fluoroin the membrane. Second, we show how dual-wavelength measurephores using a single excitation wavelength. In addition, Key Words: multiphoton excitation; two-photon microscopy; fluctupare the light losses under typical conditions due to the ation correlation spectroscopy; giant unilamellar vesicles; scanning combined effect of optical path/aberrations and we confluctuation correlation spectroscopy; generalized polarization; cluded that such losses could be quite substantial in a Laurdan.
standard confocal instrument compared with the multiphoton system. One additional advantage of our multiphoton system compared with the confocal is the Several reviews have been published recently about capability to stop the beam at any point during the raster multiphoton microscopy by our group and by others (1scan and to perform single-point measurements on a 8). This article is intended to be a guide to practical small volume of the sample. Of course, this option is not applications of multiphoton microscopy. In particular we unique to multiphoton systems, but very few manufacdiscuss the possibilities offered by this technique for turers of confocal microscopes offer this possibility, mostly quantitative image acquisition and single-or multiplebecause of instabilities in the beam steering mechanism. point fluctuation analysis at specific pixels of the image. We first describe some important features of the instrument setup, which enables this mode of operation. We TYPICAL MULTIPHOTON MICROSCOPY SETUP present, using examples taken from experiments performed in our laboratory, some of the unique features of To acquaint the reader with the basic building blocks multiphoton microscopy. In addition we discuss common of our multiphoton microscope, a schematic of our sysmethodological problems encountered during the course tem is shown in Fig. 1. A unique feature of our microof these measurements.
scope system is that the light detector can operate either in photon mode or in analog mode with a high-on the choice of the operator. Another unique feature mode when high sensitivity is needed. With this microis that we can perform a range of different types of scope it is possible to acquire images as well as fluctuaquantitative measurements just by using different tion data and to program the scanner to perform various parts of the software program but with no changes in types of sub-and full-frame scanning. The possibility the hardware or optical configuration (Fig. 1). Our miof simultaneously acquiring images and fluctuation croscope uses a titanium:sapphire laser as the source data is unique to our multiphoton microscope setup. for two-photon excitation. The laser is tunable from 710 to 990 nm with a single mirror set. The fluorescence is measured directly without a pinhole in the emission TWO-PHOTON CROSS SECTION OF SOME arm. The system, built around the Zeiss Axiovert M100 COMMON PROBES microscope, has a Cambridge Technology galvano-scanner and high-sensitivity photomultipliers (Hamamatsu First we show two-photon cross-section measure-R928) that can operate both in the analog mode for ments we have performed in our laboratory using the lifetime measurements and in the photon counting microscope described on a series of common dyes used in fluorescence microscopy. This collection of two-photon spectra ( Fig. 2) is similar to other measurements performed by other researchers, for example, the methods and results of Xu et al. (9)(10)(11), and illustrates the range of compounds that can be excited by a common excitation wavelength. The experimental system we used for this determination is the same experimental setup illustrated in the previous section, but the titanium:sapphire laser was operated in the picosecond mode to minimize errors due to wavelength-dependent pulse duration due to the dispersion in the optical elements of the microscope. The pulses had a bandwidth of approximately 1 nm and the duration ranged between 1 and 2 ps depending on the wavelength. We sinusoidally modulated the laser intensity at 100 Hz using an electrooptic modulator. The laser wavelength was measured using a wavemeter (IST Rees) FIG. 2. Two-photon excitation cross section (product of two-photon fluorescence quantum efficiency and two-photon absorption cross section). and the pulse duration with an autocorrelator (Fem-section we report some illustrative examples of dualcolor measurements. The first example is from the field tochrome). The excitation light is split and sent to the sample and to a reference detector. In this way, we of membrane studies. It involves the use of the fluorescent probe Laurdan to detect and recognize domain can measure the excitation intensity and extract the quadratic dependence of the fluorescence as a function regions of different phases in a giant unilamellar vesicle composed of two phospholipids that show phase immis-of the excitation power while we scan the laser wavelength. Sample concentrations were determined using cibility in a temperature region close to and below the main phase transition of one of the two phospholipids. manufacturers' or published extinction coefficients.
As previously noted also by others, a single excitation The second example, from the field of scanning fluctuation correlation spectroscopy, is issued to show mea-wavelength in the region 720-800 nm could be used to excite common UV probes, such as Laurdan (6-lauroyl-surements of rapid spatiotemporal fluctuations in the same artificial membrane system obtained with the 2-dimethylaminonaphthalene), and ANS, as well as the visible probes fluorescein and rhodamine. This feature giant unilamellar vesicle method. seems to be unique to the two-photon excitation process. In contrast, the fluorescence emission of all the compounds shown in the figure is the same as that obtained LIPID DOMAINS IN GIANT UNILAMELLAR using one-photon excitation. Also the fluorescence life-VESICLES time is the same. However, fluorescence anisotropy can be different with two-photon excitation compared with To illustrate the quantitation that can be obtained one-photon excitation. For some compounds such as using dual emission we report some recent measure-Laurdan, fluorescein, and rhodamine, the value of the ments of lipid phase coexistence in giant unilamellar initial anisotropy is larger for two-photon excitation vesicles composed of an equimolar preparation of 1,2than for one-photon excitation.

DUAL-COLOR SINGLE-POINT MEASUREMENTS
previously reported (12,13). At temperatures above the AND IMAGES gel-liquid crystal transition of DPPC (42ЊC), a single homogeneous phase is reported for this mixture, while at temperatures below the phase transition of DPPC, We now discuss features and advantages of multiphoton microscopy by way of some examples of the different two separate phases have been reported. Our first aim is to determine if, in this regime, the regions of the measurement modes that the microscope offers. In this vesicles that can be identified as being in the liquid compare this spectral shift with the spectral shift measured in a pure DLPC sample and assess the heterogecrystalline phase, and that are supposed to contain esneity of the lipid phase on the length scale of the point sentially DLPC molecules, have the same properties as spread function by analyzing the GP pixel histogram. vesicles that contain only DLPC.
Laurdan is a membrane probe that shows equal partition in the liquid crystal and gel phases of phosphatidyl-

ANALYSIS OF THE GP HISTOGRAM
choline phospholipids. The emission spectrum maximum of Laurdan changes from about 440 nm in the gel The GP function was originally introduced by Paraphase to 490 nm in the liquid crystalline phase. This sassi et al. in 1990 (12) and used in conjunction with change is due to water penetration in the membrane membrane studies using spectral-dependent lipophilic causing a different membrane packing in the liquidprobes such as Laurdan and Prodan. The GP function crystalline phase compared with the gel phase. Howis defined as ever, the shift toward the red of the emission spectrum of Laurdan in the liquid crystalline phase is a dynamic GP ϭ I B Ϫ I R I B ϩ I R , process. After excitation, the emission spectrum starts to move toward the red due to the changes in orientation where I B is the intensity measured in the blue part of of the water molecules in the immediate surroundings the spectrum, typically at 440 nm for Laurdan, and I R is of the excited Laurdan molecule. This process depends the intensity measured in the red part of the spectrum, on several factors, including temperature and amount typically at 490 nm for Laurdan. The GP function is of water in the membrane. As a consequence, the avermathematically related to the more common ratioage amount of spectral shift during the excited state metric measurement. However, in the field of microslifetime is exquisitely sensitive to the dynamics and copy the additive propriety of the GP function (14) is water content of the membrane.
an advantage. In contrast to cuvette measurements, Figure 3 shows a two-photon (and hence sectioned) where there is only one value of the GP at any given image of a giant unilamellar vesicle. The z section is temperature, in an image of a vesicle it is possible to obtained at the top of the vesicle. This image was obdistinguish regions of different GP values.
The right-hand part of Fig. 3 corresponds to the GP tained at 36ЊC, where relatively large, stable gel doimage. Note that in our instrument we display the GP mains are formed. The left-hand part of Fig. 3 is the image as we acquire the image. The color scale maps fluorescence intensity measured using a bandpass filter the GP between 1 and Ϫ1. Typically, pure gel phase in the detection path centered at 490 nm. We also have has a GP around 0.5-0.6, while in the liquid crystalline another light collection channel with a bandpass filter phase the GP value is between Ϫ0.3 and Ϫ0.4, decentered at 440 nm. The dark regions in the left-hand pending on the temperature and water content of the panel of Fig. 3 correspond to gel domains. In these membrane. The right-hand part of the figure clearly ordered regions, Laurdan molecules are aligned with shows that the GP value is not homogeneous. A typical their dipole transition almost exactly along the memhistogram of GP pixel values obtained in a region corbrane normal (in the more central part of the frame). responding to the liquid crystalline phase ( Fig. 4; region Due to the direction of light propagation, which is along the z direction, no excitation can occur. The bright regions in the left-hand panel of Fig. 3 correspond to disordered liquid-crystal regions. In this case, Laurdan molecules are only partially oriented along the membrane normal. The membrane disorder produces a population of Laurdan molecules with a relatively large component of the transition dipole moment parallel to the membrane normal and then excitation can occur. Although the intensity figure seems to show a relatively homogeneous liquid-crystalline phase, for the purpose of the illustration of the dual-wavelength method, we generalized polarization (GP) function. We can then indicated with square 1 in Fig. 3) gives an average value fluctuations that indicate phase fluctuations of the membrane, as shown in the next section. of Ϫ0.42, with a relatively broad distribution. This average GP value is what should be expected from a liquidcrystalline phase according to cuvette measurements.
Instead, other regions of the image, indicated with SCANNING FCS AND THE FORMATION/ square 2 on the right-hand part of the figure, exhibit MOVEMENT OF LIPID DOMAINS GP values in the range 0.26, also with a relatively broad distribution. This value is definitively lower than the A unique feature of our multiphoton microscope is 0.5-0.6 value expected for a gel phase according to cuthe capability to perform rapid scans of a certain part vette measurements. The GP image analysis clearly of an image. Scanning FCS in conjunction with distinguishes regions of the membrane that have very multiphoton microscopy was first realized by Berland different GP values and unequivocally identifies which et al. (15). For this application, we performed a scan along a circular orbit. The period of the scan for each regions are in the gel and in the liquid crystalline phase.
orbit was 2 ms. By properly aligning the points in time The observation that the border of the field of view in and space, we can have a quasi-simultaneous record of the direction at 20-30Њ from the vertical in the image the changes at different spatial locations. If we consider tends to have larger values of the GP is due to a combithe same spatial location, we can perform fluctuation nation of photoselection and polarization effects since analysis with a time resolution of 2 ms. The most imin this region of the image we start to excite Laurdan portant feature is that we can spatially correlate meamolecules that are normal to the membrane surface surements at different locations. because of the curvature of the vesicle. Figure 6 illustrates a typical situation in which the The distribution of GP values in the different pixels image taken at the top section of a giant vesicle shows of the image implies that on the membrane there is a the existence of gel domains characterized by the abdistribution of gel phase, liquid-crystalline phase, and sence of excitation (dark regions) and by the larger regions of intermediate order. It has been previously value of the GP as shown in the right-hand panel of reported that the GP distribution in the gel phase is Fig. 6. This image was obtained at 39.5ЊC, during the relatively narrow, while it is quite broad in the liquid cooling cycle, when separated gel domains just start to crystalline phase. We confirm this observation and we form. The circle in the figure shows approximately the add that in mixed phases obtained with 1:1 DLPC: scanning path. To present the scanning data, we con-DPPC, the gel segregated phase also has a relatively structed a pseudo-image in which the x axis represents broad GP distribution.
points along the orbit and the y axis represents the time. Figure 7 should not be interpreted as an image, since the vertical axis represents the time evolution. How-FLUCTUATION CORRELATION SPECTROSCOPY ever, using this pseudo-image, we can easily distinguish

(FCS) AT SELECTED POINTS IN AN IMAGE
In our multiphoton microscope it is possible to perform a fluctuation correlation measurement at a specific point of the image. For this illustration, we chose a point in the liquid-crystalline phase at about the center of square 1 in Fig. 3. We recorded synchronously for several minutes the intensity fluctuations in both detection channels. The intensity autocorrelation spectrum of one channel is shown in Fig. 5. measure a definite autocorrelation spectrum of the GP the spatial correlations and the evolution of the do-together into a conventional fluorescence microscope a number of techniques that were used only in specialized mains. Of course, if it were possible to acquire images at a very high rate, we would have even more information. instrumentation before, for example, imaging and fluctuation spectroscopy. A number of other methodologies, However, as far as we know, two-photon images with a time resolution of 2 ms have not been achieved yet also possible in standard confocal microscopy, are basically simplified, for example, dual-emission analysis and the technique we propose provides most of the information without modifying or adding new hardware to and lifetime imaging. One obvious question is whether or not one should get a multiphoton microscope. In this the multiphoton microscope. What is interesting in this experiment is the possibility of measuring relatively review we have discussed only multiphoton microscopy using dual-emission detection and for fluctuation anal-fast fluctuations and spatially correlating the fluctuations at different points in an image. The same experi-ysis. Multiphoton microscopy also has advantages for a single-emission measurement, which has been pre-ment could be performed on a cell and the fluorescence changes could be due to pH or calcium changes. In the viously discussed (4,(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32), in terms of the simplified experiment shown in Fig. 7 we can detect the onset of optical design, photobleaching characteristics, and the formation of gel domains and follow their evolution quality of image formation. The limits of spatial resoluin space and time. Although for this particular experition that can be achieved with multiphoton microscopy, ment the dynamics are relatively slow (on the second also in conjunction with the 4-Pi setup, have been distime scale) data were acquired on the millisecond time cussed (33)(34)(35)(36)(37)(38). The most appropriate laser sources scale. The minimum rotation period for the scanner have also been discussed and schemes using both used in our multiphoton instrument is about 0.2 ms, pulsed and continuous-wave lasers have been described allowing in principle the measurement of submillisec-(39-43). As the complexity of biological experiments ond dynamics.
increases and the number of questions a biologist asks during a microscopy experiment increases, simplification of the optical design seems to be a crucial requirement to make the system capable of multiple measure-DISCUSSION ment modalities (44)(45)(46)(47)(48)(49)(50)(51)(52)(53). For example, in the only commercial system available today (Confocor 2 from The examples we have presented show typical appli-Zeiss) that combines confocal imaging with fluctuation cations of multiphoton microscopy to different biophysianalysis using one-photon excitation, the optical design cal and biological problems. The power of the method is very complex. To change from the imaging mode to is due mainly to the elimination of the confocal emission the FCS mode, the optical path must be changed in the pinhole typical of standard laser confocal microscopes. This simplification of the optical design has brought Confocor 2. Instead, in our multiphoton system, the optical system is the same in both modes of operation. CONCLUSION The only changes are in the way the data stream is treated by the software. In cases in which the biological The differences between standard one-photon confoexperiment requires performing FCS measurements at cal microscopy and multiphoton microscopy are disselected locations in an image (in 3D space) the cussed in the context of multiemission measurements multiphoton system provides distinctive advantages. and fluctuation correlation measurements in cells. The These advantages are augmented if dual-emission measimplified optical design of the multiphoton system which lacks the emission pinhole and has a different surements are also needed. Dual-or multiple-emission emission path permits the use of several measurement measurements produce substantial benefits for the demodalities in the same setup and during the course of tection, interpretation, and quantitation of image feaa single experiment. This is a unique capability, which, tures.
combined with other advantages of multiphoton excita-In the example we presented of GP measurements tion, allows a new class of experiments to be performed in giant unilamellar vesicles, the same measurements on cells. could have been made using a standard confocal microscope. The increase in sensitivity achieved in the twophoton microscope because of the simplified emission ACKNOWLEDGMENTS optical path is crucial to this kind of measurement. Furthermore, in the standard confocal microscope, color This work was conducted under NIH-P41 RR03155 and ICR 122194.
aberrations and color misalignments due to the dualexcitation and dual-emission characteristics of this kind of measurement can produce severe artifacts.