
Emerging Tools for Single-Cell Analysis
.pdf
232 |
Probing Deep-Tissue Structures by Two-Photon Fluorescence Microscopy |
A second approach, which has been termed multiphoton multifocal microscopy, is analogous to Nipkow disk-based confocal systems (Bewersdorf et al., 1998). This approach is based on a custom-fabricated scan lens consisting of a specially designed lenslet array that focuses the incident laser into multiple focal spots at the field aperture plane. The lenslet array is arranged similarly to the traditional Nipkow design. Upon the rotation of the scan lens, the projected focal spots of the lenslet array will uniformly cover the field aperture plane. The CCD cameras are used to register the spatial distribution of the resulting fluorescent spots and integrate them into a coherent image. The ability to image multiple sample regions simultaneously reduces total data acquisition time. The technique suffers much less from resolution degradation and has the added advantage of being extremely robust.
A High-Speed, Video-Rate Point Scanning System
A third method optimized for high-speed deep-tissue imaging is based on raster scanning of a single diffraction-limited spot utilizing a high-speed polygonal mirror. This instrument is adapted from a very successful two-photon microscope designed for deep-tissue imaging (So et al., 1998). Since a single spot is scanned, this system retains the diffraction-limited resolution of a standard two-photon microscope. Furthermore, since fluorescence is generated at only one location at any time, a large-area single-point detector can be used. The spatial information is encoded by the timing of the raster scan pattern, as in a typical confocal microscopy. The image resolution can also be further improved by replacing the CCD camera with a large single-pixel detector, to remove the dependence on the emission psf. This is particularly important in turbid specimens where the scattered fluorescence signal is not confined in a single pixel of the CCD camera and results in degradation of the image resolution.
In our implementation, a fast rotating polygonal mirror (Lincoln Laser, Phoenix, AZ) is used for high-speed line scanning along the x-axis and a slower galvanometer-
F i g . 10.6. Collagen/elastin fibers imaged in the dermal layer of an ex vivo human skin sample. Each picture was acquired in 90 ms. Images at depths of (a) 80, (b) 100, and (c) 120 m are shown. (Reprinted with permission from SPIE. Proc SPIE 3604:60–66, 1999.)

Two-Photon Fluorescence Spectroscopy in Deep Tissue |
233 |
driven scanner with 500-Hz bandwidth (Cambridge Technology, Watertown, MA) correspondingly deflects the line-scanning beam along the y-axis. The spinning disc of the polygonal mirror comprises 50 aluminum-coated facets (2 2 mm) arranged contiguously around the disc’s perimeter. The facets repetitively deflect the laser beam over a specific angular range, correspondingly scanning a line 50 times per revolution. Rotation speeds of either 10,000, 15,000, 20,000, or 30,000 rpm are available.
The fluorescence is recorded by an intensified, frame transfer CCD camera (Pentamax; Princeton Instrument, Trenton, NJ). The 12-bit data of the 512 512 pixel CCD chip can be read out at 5 MHz. The maximum achievable image transfer rate is about 10 frames per second for 256 256 pixel images (2 2 pixel binning). Consequently, the polygonal mirror spinning speed is adjusted to 10,000 rpm, and the CCD exposure time is correspondingly set to 90 ms. In order to perform 3D volume scans, the objective is mounted on a computer-controlled piezoelectric objective translator (P-721.00, Physik Instrumente, Waldbronn, Germany).
A laser diode (1 mW at 632 nm; Thorlab, Newton, NJ) and a photodiode detector (Thorlab, Newton, NJ) are used to encode the polygonal mirror position and to generate a reference signal to synchronize the x, y scanners, the objective translator, and the CCD camera.
To demonstrate the acquisition of clinically relevant images using a video-rate two-photon microscope, we have studied the dermal structures in ex vivo human skin. The collagen/elastin fiber structures in the dermal layer were imaged (Fig. 10.6). One hundred images were taken at depths between 80 and 120 m below the skin surface. The frame rate was 90 ms and the entire stack of images was acquired in 9 s. The filamentous structures of the collagen/elastin fibers can be clearly resolved from the figure. Representative images of the fiber structures were shown in Figure 10.5. Due to the lower fluorescence intensity of the NAD(P)H components in the epidermal keratinocytes, they are more difficult to observe. The basal cells can be visualized; however, their high melanin content makes thermal damage due to one-photon absorption a major concern. Thermal damage to the basal layer was occasionally observed. We are investigating a method to mitigate this problem (Masters et al., submitted).
TWO-PHOTON FLUORESCENCE SPECTROSCOPY IN DEEP TISSUE
Importance of Spectroscopic Information in Deep-Tissue Imaging
In addition to structural information, two-photon fluorescence microscopy can be used to assess tissue functional state by spectroscopic study of the induced fluorescence. Information such as emission wavelength, fluorescence lifetime, and emission polarization can be used to monitor the tissue microenvironment. Fluorescence spectroscopic techniques can be applied in a number of ways. First, different structures in the tissue can be distinguished either by their endogenous spectroscopic properties or by extrinsic fluorophores labeling the tissue. Second, fluorescence spectroscopy can

234 |
Probing Deep-Tissue Structures by Two-Photon Fluorescence Microscopy |
be used to study tissue biochemical states and the concentration of metabolites such as calcium and oxygen. For example, cellular metabolism can be monitored optically and non-invasively by the technique of redox fluorometry (Masters, 1990; Chance and Thorell, 1959; Masters et al., 1993). A class of common endogenous probes is the pyridine nucleotides. The relative concentration of NAD (nonfluorescent) versus NAD(P)H (fluorescent) in the tissue varies depending on cellular metabolic conditions. Therefore, tissue metabolism can be monitored by tracking fluorescence intensity changes in cells and tissues.
Biomedical Applications of Fluorescent Spectroscopic Information in
Two-Photon Microscopy
A study was performed to quantify the biochemical species responsible for the fluorescence observed in the cornea of a freshly excised rabbit eye. Sample preparation procedures were similar to those used in Piston et al. (1995). The emission spectra of the various cell layers in the ex vivo rabbit cornea (wing cells, basal cells, and stro-
F i g . 10.7. Representative images and emission spectra of three cell layers in the rabbit cornea. (a) wing cells, (b) basal cells, and (c) stroma keratocytes. (Reprinted with permission from IEEE. IEEE Engineering in Medicine and Biology Magazine 18:23–30, (1999).)

References |
235 |
mal keratocytes) with two-photon excitation at 730 nm are shown in Figure 10.7. These spectra show a peak in the region of 450 nm and an FWHM of about 100 nm. The shape and wavelengths of the emission spectra are consistent with those of the NAD(P)H fluorescence.
The maximum intensities of the fluorescence emission occur at 450 nm and show a large variation between the various cell layers. The highest fluorescent intensity was recorded for the basal epithelial cells (35,000 AU). The peak intensity recorded from the wing cells (3000 AU) was more than an order of magnitude less than that from the basal epithelial cell. The lowest fluorescent intensity (600 AU) was recorded from the stromal keratocytes. These results support the fact that the cells in the basal region are the most metabolically active.
CONCLUSION
Two-photon fluorescence microscopy is a powerful technique for studying deeptissue physiology. In this chapter, we discussed three recent complementary developments in the field with significant impacts in advancing two-photon, deep-tissue technology. We have shown that a blind deconvolution algorithm can not only further improve quality of two-photon images but also provide psf information in inhomogeneous specimen. The additional psf information can help to understand two-photon-induced photonic interaction in deep tissue. We have also seen the importance of video-rate microscopy in extending two-photon technology to potential clinical applications. In the section on video-rate, two-photon technology, we saw that video-rate, two-photon microscopy can provide fast structural information in deep tissue. Future development clearly will extend the faster imaging technique to clinical diagnosis of diseases. In addition to structural information, two-photon spectroscopic data can be acquired in conjunction with images. The biochemical information acquired can be complemented to structural images, providing a new dimension to the two-photon technique.
These developments represent significant advances in two-photon, deep-tissue microscopy. In the future, it can be envisioned that they will act in unison to address important physiological problems in deep-tissue studies.
REFERENCES
Albota M, Beljonne D, Bredas JL, Ehrlich JE, Fu JY, Heikal AA, Hess SE, Kogej T, Levin MD, Marder SR, McCord-Maughon D, Perry JW, Rockel H, Rumi M, Subramaniam G, Webb WW, Wu XL, Xu C (1998): Design of organic molecules with large two-photon absorption cross sections. Science 281:1653–6.
Bennett BD, Jetton TL, Ying G, Magnuson MA, Piston DW (1996): Quantitative subcellular imaging of glucose metabolism within intact pancreatic islets. J Biol Chem 271:3647–3651.
Bewersdorf J, Pick R, Hell SW (1998): Mulitfocal multiphoton microscopy. Opt Lett 23:655.
Booth ML, Hell SW (1998): Continuous wave excitation two-photon fluorescence microscopy exemplified with the 647-nm ArKr laser line. J Microsc 190:298–304.
236 |
Probing Deep-Tissue Structures by Two-Photon Fluorescence Microscopy |
Brakenhoff GJ, Squier J, Norris T, Bliton AC, Wade WH, Athey B (1996): Real-time two-photon confocal microscopy using a femtosecond, amplified Ti:sapphire system. J Microsc 181( Pt 3):253.
Centonze VE, White JG (1998): Multiphoton excitation provides optical sections from deeper within scattering specimens than confocal imaging. Biophys J 75(4):2015–2024.
Chance B (1976): Pyridine nucleotide as an indicator of the oxygen requirements for energy-linked functions of mitochondria. Circ Res Suppl 1 38:I-31–I-38.
Chance B, Schoener B, Oshino R, Itshak F, Nakase Y (1979): Oxidation reduction ratio studies of mitochondria in freeze-trapped samples. J Biol Chem 254:4764–4711.
Chance B, Thorell B (1959): Localization and kinetics of reduced pyridine nucleotide in living cells by microfluorometry. J Biol Chem 234:3044–3050.
Corcuff P, Bertrand C, Leveque L (1993): Morphometry of human epidermis in vivo by real-time confocal microscopy. Arch Dermatol Res 285:475–481.
Dabbous MK (1966): Interand intramolecula cross-linking in tyrosinase-treated tropocollagen. J Bio Chem 241:5307–5312.
Denk W, Strickler JH, Webb WW (1990): Two-photon laser scanning fluorescence microscopy. Science 248:73–76.
Göppert-Mayer M (1931): Über Elementarake mit zwei Quantensprungen. Ann Phys (Leipzig) 5:273–294.
Guild JB, Webb WW (1995): Line scanning microscopy with two-photon fluorescence excitation. Biophys J 68:290a.
Hoerman KC, Balekjian AY (1966): Some quantum aspects of collagen. Fed Proc 25:1016–1021.
Holmes T (1992): Blind devoncolution of quantum-limited incoherent imagery: maximum-likelihood approach. J Opt Soc Am A 9(7):1052–1061.
Jackson D (1975): Classical Electrodynamics. New York: Wiley.
LeBella FS (1961): Studies on the soluble products released from purified elastic fibers by pancreatic elastase. Arch Biochm Biophys 93:72–79.
LaBella FS, Gerald P (1965): Structure of collagen from human tendon as influence by age and sex. J Gerontol 20:54–59.
LaBella FS, Lindsay WG (1963): The structure of human aortic elastin as influence by age. J Gerontol 18:111–118.
Maiti S, Shear JB, Williams RM, Zipfel WR, Webb WW (1997): Measuring serotonin distribution in live cells with three-photon excitation. Science 275:530–532.
Masters BR (1996): Three-dimensional confocal microscopy of human skin in vivo: Autofluorescence of normal skin. Bioimages 4:13–19.
Masters BR (1990): In vivo corneal redox fluorometry. In Masters BR (ed). Noninvasive Diagnostic Techniques in Ophthalmology. New York: Springer.
Masters BR, Chance B (1993): Redox confocal imaging: intrinsic fluorescent probes of cellular metabolism. In Mason WT (ed). Fluorescent and Luminescent Probes for Biological Activity, London: Academic Press.
Masters B, So PT, Dong CY, Buehler C, Gratton E (submitted): The use of a laser pulse picker to mitigate two-photon excitation damage to living specimen. J Rev. Sci. Instr.
Masters BR, So PTC, Gratton E (1997a): Multiphoton excitation fluorescence microscopy and spectroscopy of in vivo human skin. Biophys J 72:2405–2412.
Masters BR, So PTC, Gratton E (1997b): Multiphoton excitation fluorescence microscopy and spectroscopy of in vivo human skin. Biophys J 72:2405–2412.
Masters BR, Gonnord G, Corcuff P (1997c): Three-dimensional microscopic biopsy of in vivo human skin: a new technique based on a flexible confocal microscope. J Microsc 185:329–338.
Masters BR, Kriete A, Kukulies J (1993): Ultraviolet confocal fluorescence microcopy of the in vitro cornea: redox metabolic imaging. Appl Opt 32:592–596.
Pawley JB (1995): Handbook of Biological Confocal Microscopy. New York: Plenum Press.
References |
237 |
Piston DW, Masters BR, Webb WW (1995): Three-dimensionally resolved NAD(P)H cellular metabolic redox imaging of the in situ cornea with two-photon excitation laser scanning microscopy. J Microsc 178:20–27.
Raijadhyaksha M, Grossman M, Esterowitz D, Webb RH, Anderson RR (1995): In vivo confocal scanning laser microscopy of human skin: melanin provides strong contrast. J Invest Dermat 6:946–954.
Rummelt V, Gardner LMG, Folberg R, Beck S, Knosp B, Moninger TO, Moore KC (1994): Three-dimen- sional relationship between tumor cells and microcirculation with double cyanine immunolabelling, laser scanning confocal microscopy, and computer assisted reconstruction: An alternative to cast corrosion preparation. J Histochem Cytochem 42:681–686.
So PT, Kim H, Kochevar IE (1998): Two-photon deep tissue ex vivo imaging of mouse dermal and subcutaneous structures. Opt Exp 3:339.
So PTC, French T, Yu WM, Berland KM, Dong CY, Gratton E (1995): Time-resolved fluorescence microscopy using two-photon excitation. Bioimaging 3:49–63.
Thomas J, Elsden DF, Partridge SM (1963): Degradation products from elastin. Nature 200:651–652.
Wilson T, Sheppard C (1984): Theory and Practice of Scanning Optical Microscopy. New York: Academic Press).
Zipfel W (1997): Multi-photon Excitation of Intrinsic Fluorescence in Cells and Intact Tissue. Presented in Application of multi-photon excitation imaging, Pre-Microscope Society of America Symposium, Cleveland, OH, Aug 9–10.

Emerging Tools for Single-Cell Analysis: Advances in Optical Measurement Technologies
Edited by Gary Durack, J. Paul Robinson Copyright © 2000 Wiley-Liss, Inc.
ISBNs: 0-471-31575-3 (Hardback); 0-471-22484-7 (Electronic)
11
Limits of Confocal Imaging
James B. Pawley
University of Wisconsin, Madison, Wisconsin
INTRODUCTION
What Limits?
Perhaps a measure of the utility of the confocal imaging principle is the fact that it seems to have been independently invented at least seven times since the late 1950s (Minsky, 1988; Egger and Petran, 1967; Slomba et al., 1972; Sheppard and Choudhury, 1977; Brakenhoff et al., 1979; Carlsson et al., 1985; White et al., 1987). The optical principle is not complex: focus a point light source into or onto the specimen and arrange for light emitted or scattered from this point on the specimen to be focused onto a point detector [usually an aperture in front of a photomultiplier tube (PMT)]. Photons emerging from features in the specimen that are not in the plane of focus will not be focused into a point at the plane of the detector aperture and, consequently, most of them are excluded. The exclusion of out-of-focus light from the data stream gives the confocal its most characteristic feature: the ability to make optical sections. The price for this sectioning ability is that most laser confocal microscopes at any one time image only a single point on the specimen, and a twoor three-dimensional (2D, 3D) image can be produced only by scanning the focused spot over the specimen or vice versa. This sampling approach allows an image to be built up from a number of individual measurements that reflect optical properties within specific regions of the sample. Any such measurements involve not only the optical limitations inherent in
239
240 |
Limits of Confocal Imaging |
focusing and imaging systems but also counting photons, a process that implies important limitations on data rate and statistical accuracy.
The task of the confocal light microscope is to measure optical or fluorescent properties within a number of small, contiguous subvolumes of the specimen (Fig. 11.1; Pawley and Centonze, 1994). Because confocal microscopy is probably the most sensitive method for imaging living cells and because imaging such cells places the greatest demands on the instrumentation and technique, we will generally consider the specimen to be a living cell. The fundamental limits on this process are related to the quantitative accuracy with which these measurements can be made, a factor that depends on the number of photons that pass into, n1, and out of, n2, the subvolume; its size ( x, y, z); and its position (x, y, z). Additional limitations are imposed on the rate at which these measurements can be made by the effects of photodamage to the specimen, source brightness, and fluorescence saturation. Finally, limitations are imposed by the fact that the continuous specimen must be measured in terms of discrete volume elements called voxels (a voxel is the 3D equivalent of a pixel, which is the smallest element of a 2D image). Although, for simplicity in discussion, these factors are usually treated separately, in practical microscopy they almost always interact. The discussion that follows will often highlight the interactions.
This chapter will outline the factors that ultimately limit the accuracy with which these measurements can be made. Although most of the discussion should be applicable to any form of confocal microscope, we will assume that scanning is accomplished by moving a single beam of laser light.
The data recorded from a confocal laser scanning microscope (CLSM) are usually a set of intensity values matched to every pixel of a 2D optical section or to every voxel throughout a 3D volume in the specimen. Ideally, these values represent the concentration of fluorophore as a function of position. In fact, many other factors can produce unintended contrast. In fluorescence confocal microscopy, the rate at which data can be produced is limited by fluorescence saturation. As a result, the statistical quality of the data is often related inversely to the scanning speed, the instability of the fluorophore, and the size of the raster.
Counting Statistics: The Importance of n
The accuracy of any particular measurement involving fundamental, quantum interactions (such as counting photons) is limited by Poisson statistics, which ensure that the 1 error in a measurement of n photons is n. While similar considerations limit the performance of all types of microscopical measurements, they are more explicit in their effect on confocal microscopy, where peak signal levels often produce signal levels of only about 8 photons per 2- s pixel.
The uncertainty associated with counting quantum-mechanical events is the source of intrinsic statistical noise. Confocal data sets can also contain extrinsic noise introduced by detector dark current, electronic noise, or interference or produced by stray or out-of-focus light. Unlike intrinsic noise, extrinsic noise can be reduced by careful technique and technological improvements. Indeed, the first test of a confocal microscope should be to collect signal from a “blank field.” To do this, form an image

Introduction |
241 |
F i g . 11.1. Parameters related to the fundamental limitations of confocal microscopy.
of a nonfluorescent object such as pure water and set the PMT gain as you would for “normal” operation for a weak specimen and the dc offset so that the average recorded signal is about 10 digital units (assuming 8-bit, 256-level digitization). A statistical analysis of a single-scan data set should allow you to differentiate pixels with one dark count from those with none. A count rate of more than 10,000 “bright” pixels per 1-s scan should be a cause for concern. Perhaps the PMT photocathode is
242 |
Limits of Confocal Imaging |
being accidentally warmed by nearby equipment. Alternatively, you may have a light leak. To localize the problem, try comparing one detector channel with another, turning off the room lights, or obscuring the laser.
To check for “fixed-pattern” noise, accumulate data from a large number of dark frames by Kalman averaging. The appearance of any sort of pattern on the screen as more data are acquired may also indicate a problem. However, one must realize that fixed-pattern noise always exists at some level. It is only a problem if it is similar in size to the signal level, and even then its effect can sometimes be reduced by subtracting a “noise-only” pattern from a “signal-plus-noise” pattern.
In a well-designed CLSM, the major noise source is intrinsic noise. This fact highlights the importance of making sure that as many as possible of the available photons are recorded as part of the signal.
Specimen Response: Dye Saturation. In “normal microscopy” it is safe to assume that photons interact with the specimen in a manner that is independent of the intensity of the illumination and that output signal is always directly proportional to excitation. However, this linear response is not characteristic of any laser-based fluorescence confocal microscopes operated with 1 mW in a 0.5- m-diameter spot in the specimen. Not only may absorption in the specimen cause sufficient warming to produce damage or motion, but the electric field strength of the focused light may also become sufficient to produce a variety of nonlinear responses such as fluorescence saturation.
Saturation occurs when fluorescent molecules are excited by a flux of exciting illumination so intense that, at any instant, a significant fraction of the fluorescent molecules is in the excited state. As excited molecules no longer absorb light at the usual wavelength ( ), this has the effect of lowering the effective dye concentration. This saturation threshold can easily be reached at flux levels around 106 W/cm2. Saturation is related to absorption cross section and fluorescence lifetime. As can be seen from Equations (1) and (2), the problem is more severe when using dye molecules with long fluorescent decay times f:
I
(11.1) h
where is the excitation rate in reciprocal seconds at low intensity, is excitation cross section in centimeters squared, I is the excitation intensity in watts per centimeters squared, and h is the energy of the absorbed excitation photon. Then
N |
|
(11.2) |
1 |
1/ f |
|
where N1 is the fraction of dye molecules in the first singlet state, near saturation. With most dyes, singlet-state saturation occurs at power levels of about 1 mW if focused into a diffraction-limited spot by a numerical aperture (NA) 1.3 lens. As lower NA produces a larger spot area, an NA 0.65 would not saturate until about four times