Multiphoton microscopy in life sciences

Authors


K. König. Fax: +49 3641 938552; e-mail: kkoe@mti-n.uni-jena.de

Abstract

Near infrared (NIR) multiphoton microscopy is becoming a novel optical tool of choice for fluorescence imaging with high spatial and temporal resolution, diagnostics, photochemistry and nanoprocessing within living cells and tissues. Three-dimensional fluorescence imaging based on non-resonant two-photon or three-photon fluorophor excitation requires light intensities in the range of MW cm−2 to GW cm−2, which can be derived by diffraction limited focusing of continuous wave and pulsed NIR laser radiation. NIR lasers can be employed as the excitation source for multifluorophor multiphoton excitation and hence multicolour imaging. In combination with fluorescence in situ hybridization (FISH), this novel approach can be used for multi-gene detection (multiphoton multicolour FISH). Owing to the high NIR penetration depth, non-invasive optical biopsies can be obtained from patients and ex vivo tissue by morphological and functional fluorescence imaging of endogenous fluorophores such as NAD(P)H, flavin, lipofuscin, porphyrins, collagen and elastin. Recent botanical applications of multiphoton microscopy include depth-resolved imaging of pigments (chlorophyll) and green fluorescent proteins as well as non-invasive fluorophore loading into single living plant cells.

Non-destructive fluorescence imaging with multiphoton microscopes is limited to an optical window. Above certain intensities, multiphoton laser microscopy leads to impaired cellular reproduction, formation of giant cells, oxidative stress and apoptosis-like cell death. Major intracellular targets of photodamage in animal cells are mitochondria as well as the Golgi apparatus. The damage is most likely based on a two-photon excitation process rather than a one-photon or three-photon event. Picosecond and femtosecond laser microscopes therefore provide approximately the same safe relative optical window for two-photon vital cell studies. In labelled cells, additional phototoxic effects may occur via photodynamic action. This has been demonstrated for aminolevulinic acid-induced protoporphyrin IX and other porphyrin sensitizers in cells. When the light intensity in NIR microscopes is increased to TW cm−2 levels, highly localized optical breakdown and plasma formation do occur. These femtosecond NIR laser microscopes can also be used as novel ultraprecise nanosurgical tools with cut sizes between 100 nm and 300 nm. Using the versatile nanoscalpel, intracellular dissection of chromosomes within living cells can be performed without perturbing the outer cell membrane. Moreover, cells remain alive. Non-invasive NIR laser surgery within a living cell or within an organelle is therefore possible.

1. Introduction

Conventional light microscopes used in life sciences are based on the use of ultraviolet (UV) and visible (VIS) radiation for optical analysis and microsurgery (Minsky, 1988; Greulich & Weber, 1992; Pawley, 1995). The diagnostic methods include (i) 2D fluorescence microscopy using discrete excitation lines of high pressure mercury lamps and (ii) 3D confocal laser scanning fluorescence microscopy with excitation radiation of He–Ne lasers, argon/krypton ion lasers, He–Cd lasers and laser diodes.

Most endogenous fluorophores in cells, such as tryptophan, NAD(P)H and flavins, require UV or blue excitation wavelengths. Important exogenous fluorophores, such as the DNA markers Hoechst 33342 and DAPI (4′,6-diamidino-2-phenylindole) as well as the calcium indicators Fura-2 and Indo-1 possess absorption bands in the UV only, whilst other exogenous fluorophores are designed to match the laser and mercury lamp emission lines. Microsurgery of cells is typically performed with a nitrogen laser at 337 nm or the third harmonic (355 nm) and second harmonic (532 nm) of a Q-switched Nd:YAG laser (Berns, 1998; Schütze et al., 1994). However, the use of UV and short-wavelength VIS radiation in conventional light microscopy has disadvantages mainly because of low light penetration depth and the potential of severe photodamage to living cells (Cunningham et al., 1985; Tyrell & Keyse, 1990). It has been shown that fluorescence microscopy of single cells with the 365 nm radiation from a conventional 50 W high pressure mercury lamp results within seconds in failed cellular reproduction, modifications in intracellular redox state and DNA strand breaks (König et al., 1996a).

A versatile innovation in live cell microscopy is based on the application of near infrared (NIR) radiation in the spectral range 700–1100 nm (Fig. 1). This range and the red region (600–700 nm) is referred to as the ‘optical window’ of cells and tissues owing to the lack of efficient endogenous absorbers in this spectral range and the subsequent high light penetration depth of the order of a few millimetres in most tissues (Fig. 2). Water, with a low absorption coefficient of ≤ 0.1 cm−1 (Hale & Querry, 1973), is considered to be the major absorber in cells without haemoglobin, melanin or chlorophyll. These non-pigmented cells therefore appear to be nearly transparent structures in the case of low intensity NIR radiation.

Figure 1.

Laser microscopy in life sciences. In conventional microscopy, ultraviolet and visible laser radiation are mainly used for confocal laser scanning microscopy (CLSM) and microsurgery. Major applications of NIR microscopy are optical trapping with laser tweezers and multiphoton microscopy. In addition, SHG (second harmonic generation) and THG (triple harmonic generation) microscopy have been used.

Figure 2.

Absorption spectra of major intracellular absorbers. The molecular exctinction coefficients of oxygenated haemoglobin and melanin and the absorption coefficient of water are shown. Most animal cells and tissues are considered to be nearly transparent in the spectral range from about 700 nm to 1100 nm owing to the lack of efficient one-photon absorbers.

There are two major applications of NIR microscopy in life sciences. On one hand are the laser tweezers (optical traps) for optical micromanipulation, cell sorting, picoNewton force measurements (Ashkin & Dziedzik, 1987; Block et al., 1989; Chu, 1991; Perkins et al., 1994a,b; Veigel et al., 1999), and diagnostics of motile cells (König et al., 1998). Laser tweezers are based on the application of highly focused continuous wave (cw) NIR laser radiation with typical mean powers in the range 10–100 mW.

The other important application of NIR microscopy is two-photon and three-photon excitation that enables 3D fluorescence imaging and photoinduced uncaging of compounds. Two-photon excitation of electronic states, based on the simultaneous absorption of photons, was predicted in 1931 (Göppert-Meyer, 1931) and first realized with the availability of lasers in 1961 (Kaiser & Garret, 1961). They detected the blue fluorescence of CaF2 : Eu2+ crystals using red 0.5 ms pulses of a ruby laser. However, the first two-photon excited fluorescence imaging of living specimens using a 100 fs NIR dye laser was published almost 30 years later (Denk et al., 1990). These authors focused the radiation to a diffraction-limited spot using a high-numerical-aperture objective of a laser scanning microscope. Today, tuneable mode-locked Ti : sapphire lasers operating at 70–80 MHz repetition rate and a pulse width of less than 150 fs are typically used as NIR sources in multiphoton microscopy. Compact turn-key solid-state Ti : sapphire lasers, Cr : LiSaF or frequency-doubled fibre lasers are potentially appropriate light sources in multiphoton microscopes. Such microscopes recently became commercially available from leading suppliers of laser scanning microscopes.

This review of multiphoton microscopy focuses on recent applications in life sciences including (i) multiphoton multicolour fluorescence in situ hybridization (FISH), (ii) noninvasive optical biopsy, (iii) two-photon photodynamic therapy (PDT), (iv) in plant biology, (v) nanosurgery as well as (vi) studies directed towards understanding the mechanisms of cellular response to high intensity NIR radiation.

2. Experimental set-up and features of multiphoton microscopes

The major application of multiphoton microscopes in life sciences involves the excitation of routinely used intracellular fluorescent probes by the simultaneous absorption of two NIR photons (Fig. 3) or three NIR photons. Typically, the fluorophore has no one-photon absorption band in the NIR. Owing to the relatively low two-photon and three-photon absorption cross-sections of intracellular absorbers of about 10−48 to 10−50 cm4 s photon−1 and 10−75 to 10−84 cm6 (s photon−1)2 (Xu et al., 1995; Xu & Webb, 1996; Maiti et al., 1997; Xu & Webb, 1997), respectively, non-resonant multiphoton excitation requires photon flux densities of > 1024 photons cm−2 s−1. Using NIR radiation in the spectral range of 700–1100 nm, light intensities in the MW cm−2 to GW cm−2 range are required. These intensities can be obtained by diffraction limited focusing cw (Hänninen et al., 1994; König et al., 1995) and pulsed NIR laser radiation (Denk et al., 1990) of 100 mW mean power or less.

Figure 3.

Principle of non-resonant multiphoton excitation. The simultaneous absorption of two photons at MW cm−2 and GW cm−2 intensities is used in nonlinear fluorescence microscopy. At TW cm−2 intensities, multiphoton effects with four or more photons result in optical breakdown.

Using cw single-frequency and multifrequency NIR lasers with 100 mW power at the sample and a high numerical aperture objective, we have demonstrated two-photon excited fluorescence within a living cell (König et al., 1995, 1996b; König, 1997). Fluorescence occurred only in a sub-femtolitre focal volume at MW cm−2 light intensities. Laser tweezers, which are based on the use of cw NIR lasers, are therefore potential sources of two-photon excitation. Using 800 nm or 1064 nm laser tweezers, cells labelled with exogenous fluorophores remained alive over extended periods of optical trapping. This is consistent with the observations that despite high light intensities in the optical trap, the temperature increase is less than 2 K, thus excluding thermal toxic effects (Liu et al., 1995; Schönle & Hell et al., 1998).

By means of cw lasers (Ti : sapphire laser, krypton ion laser) and scanning units, Hell's group has performed two-photon fluorescence imaging (Hänninen et al., 1994; Booth & Hell, 1998). Two-photon excitation can also be realized with compact and less expensive cw laser diodes (Hänninen et al., 1995; Gu & Day, 1999).

The efficiency of two-photon excitation follows a power squared relation. However, trapping effects and photothermal effects limit the use of high power cw sources for fast fluorescence scanning microscopy. Much more efficient in fast multiphoton fluorescence imaging is the use of high repetition pulsed laser systems with moderate peak power in the watt and kilowatt range but with a low mean power. The efficiency of two-photon excitation and the fluorescence yield follows the following relation (Denk et al., 1990):

image

with n = number of absorbed photon pairs, P = mean power, α = molecular two-photon absorption coefficient, τ = pulse width, f = repetition frequency, NA = numerical aperture, λ = wavelength.

Because the fluorescence yield depends on a P2/τ relation in two-photon microscopy, the efficiency increases for high peak power and low pulse width. This implies that in two-photon microscopy with 1 ps laser pulses, a mean power only three times higher is required compared to 110 fs pulses in order to acquire the same fluorescence images.

In two-photon laser microscopy the use of extremely ultrashort laser pulses (< 50 fs) at the sample has several disadvantages: substantial technical efforts are required to achieve an in situ pulse width (pulse width at the sample) of less than 100 fs because of the pulse broadening during transmission through the whole optical system due to optical dispersion. The major impact is from the objective, which may consist of a variety of different glass types and thicknesses. Typical optical dispersion values D for glasses in objectives at 800 nm are + 251 fs2 cm−1 for CaF2, + 300 fs2 cm−1 for quartz, + 389 fs2 cm−1 for FK-3, + 445 fs2 cm−1 for BK7 glass, + 1030 fs2 cm−1 for SF2, and + 1600 fs2 cm−1 for SF10. At shorter wavelengths, these numbers increase.

In the case of chirp-free entrance pulses with a pulse width τin, the pulse broadening B can be calculated as follows:

image

with S = DL/inline image. A typical DL value of the whole microscope at 800 nm including high NA oil objective is 5000 fs2 (König, 1999). Figure 4 shows the pulse broadening in the case of an assumed ‘CaF2 objective’ (DL = 930 fs2) and of a realistic microscope system with DL = 5450 fs2. A chirp-free laser pulse width, e.g. 10 fs will result in an in situ pulse width of more than 1 ps. In order to obtain in situ pulse widths less than 100 fs, pulse compression units consisting of optical elements of negative dispersion values (Soeller & Cannell, 1996; König, 1999) or laser systems with negative chirp output are required. Very short laser pulses experience the additional problem of large spectral bandwidth (10 fs pulses have about 100 nm bandwidth), which requires efforts to realize transmission without chromatic aberration. As shown in section 8, ultrashort pulses are more destructive than longer pulses at same pulse energy. With a typical laser pulse width of commercially available femtosecond lasers of 50–200 fs, the in situ pulse width is of the order of 150–250 fs. This pulse width is widely used in two-photon fluorescence microscopy (König et al., 1996d). In addition, two-photon ‘picosecond’ systems are commercially available with typical pulse widths of 1–10 ps.

Figure 4.

Pulse broadening in CaF2 optics (d = 32 mm) and in the optical system of a microscope including cell chamber.

Figure 5 shows our experimental set-up that routinely uses the compact solid-state femtosecond laser ‘Vitesse’ (Coherent, Santa Clara, U.S.A.) with 80 MHz repetition frequency, a laser wavelength of 800 nm, an output power of 1 W, and a laser pulse width of 80 fs. This laser system is easy to operate (‘turn-key system’) and requires no extensive water supply for cooling and no high-power supply. After expanding the NIR laser beam by a 1 : 3 Galilean telescope and the transmission through two-piece neutral density attenuators, the beam is coupled into a modified inverted confocal laser scanning microscope (LSM 410, Carl Zeiss Jena GmbH, Jena, Germany) which also incorporates a 633-nm HeNe laser for conventional CLSM. A highly reproducible mirror slider enables switching within 1 s between the visible laser and the NIR laser. The beam is focused to a diffraction-limited spot by conventional high-NA objectives onto the sample. The typical NIR transmission of the objective is 50%. When studying living cells, a closed microchamber with 170 µm thick glass windows was used (‘MiniCeM’, JenLab GmbH, Jena, Germany).

Figure 5.

Set-up of a multiphoton and one-photon confocal microscope with compact femtosecond laser source. Various detectors allow sensitive 3D black and white imaging and true colour imaging as well as spatially resolved microspectrofluorometry and τ-mapping. In addition, nanosurgery can be performed.

The microscope contains one detector for NIR transmission (brightfield, phase, differential interference contrast (DIC)) microscopy, two photomultipliers (PMTs) in the descanned mode with adjustable pinholes (spatial filters) and SP 750 short pass filters in front, and a range of detectors at the baseport. The use of a baseport detector has the advantage of detecting photons with high efficiency without transmission through the scanner and pinhole units. At the baseport, we use different types of PMT and cameras (e.g. colour camera, video camera, slow scan camera), an optical multichannel analyser for spectroscopy, and a single photon counting unit for time-resolved measurements. Time-resolved fluorescence measurements and spectroscopy are typically performed by scanning a region of interest (ROI) or by single point exposure, where the beam is parked at pixels of interest. The high sensitivity of the baseport PMT enables the detection of two-photon excited intracellular fluorophores with high fluorescence quantum yield (e.g. DAPI, Hoechst, Rhodamine) with a mean power at the sample of 100 µW to 1 mW at a frame rate of 1 Hz (512 × 512 pixels) (Fig. 6). More importantly, endogenous fluorophores, such as fluorescent coenzymes, can be excited with a mean power of less than 2 mW. When the power exceeds 10 mW we have detected significant cell damage. Mean powers of about 30–50 mW and µs beam dwell times per pixel are able to induce optical breakdown, which can be used to drill holes and to utilize the set-up as laser scissors, as demonstrated in section 9.

Figure 6.

3D imaging of Hoechst 33342 in cells (right) with 170 fs pulses. Mean powers in the range 100 µW to 3 mW and µs beam dwell times per pixel are sufficient to obtain 3D fluorescence images.

Because < 10 mW are sufficient for two-photon fluorescence excitation, only 1% of the 1 W laser output power is finally used. Instead of wasting most of the laser output, Hell's group (Bewersdorf et al., 1998; Straub & Hell, 1998) suggested splitting the laser beam into beamlets using a microlens array. This multi-beam system enables an enhanced frame rate of 25 Hz (real time) or faster. It should be mentioned that the current maximum frame rate with our commercially available galvanometer scanning mirrors is 3 Hz for 512 × 512 pixels but can eventually be enhanced up to nearly 25 Hz by defining a small ROI. New developments focus on fast resonator scanners (Fan et al., 1999).

The spatial resolution in two-photon fluorescence microscopy is similar to one-photon CLSM despite doubling the excitation wavelength with typical values of 0.3 µm FWHM in lateral and 0.9 µm FWHM in axial direction (Denk et al., 1995). According to Stelzer et al. (1994), the axial resolution can be enhanced by up to 50% by the additional use of a confocal pinhole. Sako et al. (1997) have compared a UV-CLSM vs. two-photon microscope using the fluorophore Indo-1 and found a slightly better resolution when using a pinhole. A comparison of point spread functions in one-photon, two-photon and three-photon microscopy can be found in Schrader et al. (1997). Using 4 Pi two-photon microscopy, axial resolutions of about 100 nm have been demonstrated (Hell & Stelzer, 1992; Hell et al., 1997; Schrader et al., 1998a,b). Recently, elegant high-resolution two-photon near-field scanning microscopy has been realized (Hell et al., 1998; Kirsch et al., 1998; Jenei et al., 1999). Furthermore, single molecule detection with two-photon fluorescence correlation spectroscopy has been demonstrated (So et al., 1998; Schwille et al., 1999).

3. Fluorescence lifetime measurements

Imaging of fluorescence lifetimes (τ-mapping) is a useful contrast enhancing method that provides information related to the microenvironment with spatial resolution. Different cellular probes with similar emission spectra but with different lifetimes can be distinguished. The influence of different metabolites or other environmental conditions can be studied independent of the fluorophor concentration.

The most expensive item in time-resolved microscopy with sub-nanosecond time resolution is the laser source. A femtosecond laser microscope for 3D multiphoton fluorescence imaging can be expanded to perform 3D τ-mapping.

Fluorescence lifetime imaging of living cells with femtosecond laser microscopes has been performed in the frequency domain (So et al., 1995; König et al., 1996c) as well as in the time domain. Time-resolved measurements in the frequency domain are based on the determination of demodulation M and phase shift φ and require modulated light sources (modulation frequency ω). In the case of fluorescence with monoexponential decay the lifetime τ can be deduced from the relation:

image

In the case of multiexponential decay, τ calculated from this relation represents the mean lifetime. The relation implies that the modulation frequency should be in the MHz range for typical ns lifetimes. The repetition frequency f of the mode-locked Ti : sapphire lasers fits the required range. In order to translate the phase and demodulation information from the 80 MHz frequency to a lower kHz frequency dω, a gain modulated detector at (ω + dω) frequency can be used (heterodyne technique) (So et al., 1995). Low pass filtering isolates the cross-correlation signal with dω frequency containing the same phase and demodulation as of the fluorescence signal. Figure 7 demonstrates false-colour coded lifetime imaging of the autofluorescence of Chinese hamster ovary (CHO) cells. The colours in the lifetime image represent mean lifetimes from 0 to 5 ns. Fluorescence arises mainly from mitochondria. An average lifetime of 2.2 ns was calculated from all τ-values of fluorescent ‘pixels’. This lifetime value is typical for the coenzyme β-nicotinamide adenine dinucleotide (NADH) during binding to proteins (König et al., 1996c).

Figure 7.

Two-photon excited autofluorescence intensity image, demodulation image and mean fluorescence lifetime image and lifetime histogram of living CHO cells obtained in the frequency domain. The mean lifetime was deduced from demodulation and phase shift data.

An example of lifetime measurements in the time domain is shown in Fig. 8. Depicted in Fig. 8, left, are fluorescence decay curves of particular regions of fluorophore-labelled cells. The information on decay kinetics was obtained by parking the laser beam (‘spot mode’) in pixel of interest or by line scanning and time-correlated single photon counting (SPC 730, Becker & Hickel GmbH, Berlin, Germany. In addition to single pixel fluorescence kinetics, τ-mapping was performed. The SPC unit with a channel resolution down to 800 fs allows the fast uptake and storage of 128 × 128 = 16384 decay curves at scanning frequency. We performed spatially resolved mapping of fluorescence decay curves (4D microscopy) of intracellular Hoechst 33342 in CHO cells with a typical frame rate of 0.1–0.02 Hz (Fig. 8, right). Furthermore, time-resolved in vivo autofluorescence of chlorophyll in plastids of Arabidopsis during division was detected (Tirlapur & König, recent unpublished results).

Figure 8.

(a) Fluorescence decay kinetics of 800 nm-excited fluorophores within a subcellular subfemtolitre excitation volume. The fluorophore-labelled living cell was first imaged with DIC microscopy (800 nm, 1 µW). Afterwards, the beam was parked in the subcellular area of interest and the time-resolved fluorescence was measured by single-point or line scan exposure and time-correlated single photon counting (time domain). (b) Fluorescence lifetime image of Hoechst 33342 labelled nuclei of CHO cells.

An alternative to the measurement of fluorescence decay curves with SPC units and subsequent calculation of a mean lifetime per pixel is time-gated fluorescence imaging using special image intensifying cameras. These cameras enable imaging at different picosecond and nanosecond time windows at a preselected time delay between excitation and detection. The cameras can be triggered at high frequencies such as 80 MHz (Straub & Hell, 1998).

4. Multiphoton multicolour FISH

An interesting feature of multiphoton NIR microscopy is multifluorophore excitation, where NIR radiation at one fixed wavelength induces simultaneously the visible fluorescence of a wide range of fluorophores with a broad multiphoton excitation spectrum. This method can be used for multicolour gene and chromosome detection, where a variety of spectrally distinct fluorescent dyes are used to label various specific regions of DNA by FISH.

FISH has emerged as an important technique in genomics for the detection of genes and chromosome regions. One of the important advantages of FISH is the possibility of detecting multiple fluorescent targets with different emission wavelengths for multi-gene detection (Nederlof et al., 1989; Ried et al., 1992; Schröck et al., 1996; Speicher et al., 1996; Mackville et al., 1997). Until recently, different excitation wavelengths in the UV, blue and green range have been adopted to induce the visible fluorescence of the most common FISH fluorophores and of the general DNA stains such as DAPI and Hoechst 33342. However, the use of different excitation wavelengths causes problems through different wavelength-dependent focal points (chromatic aberration), the use of expensive UV optics, the separation of fluorescence and excitation photons in the case of simultaneous excitation (overlap of excitation bands and emission bands) and the beam alignment in the case of excitation with a filter wheel. These problems can be circumvented when a single NIR excitation wavelength for multicolour FISH is used. Applying radiation at 800 nm, simultaneous two-photon excitation of the blue-emitting FISH fluorophores Spectrum Aqua and Spectrum Blue (Vysis Corp., Downers Grove, U.S.A.), the green-emitting FISH fluorophores Spectrum Green (Vysis Corp.), FITC (Boehringer, Mannheim, Germany), Cy3 (Oncor, Illkrich, France), and SYBR Green (Molecular Probes, Eugene, U.S.A.) as well as the yellow/red-emitting fluorophores Rhodamine 110 (PE Applied Biosystems Inc. Forster City, U.S.A.), Spectrum Orange, Texas Red (Molecular Probes), JenFluor (JenLab GmbH), Cy 5 and Cy 5.5 (Oncor) has been realized (Fig. 9a,b). The same NIR wavelength could be used to induce the fluorescence of the general DNA/chromosome stains Hoechst, DAPI, propidium iodide and ethidium homodimer. Interestingly, the DNA stain Giemsa, which is commonly used in transmission microscopy for chromosome analysis, exhibits fluorescence when excited with intense NIR radiation. (Fig. 9c) (König et al., 2000).

Figure 9.

Two-photon excited fluorescence of (a) DAPI-labelled human chromosomes. (b) Six-colour-FISH with the FISH fluorophores Spectrum Orange (white, chromosome 13), R110 (green, chromosome 21), Texas Red (yellow, chromosome 6), Cy 5 (red, chromosome 9), Cy 5.5 (dark red, chromosome 1) and DAPI (blue) (c) Giemsa-labelled chromosomes and (d) 3D reconstruction of the position of centromeric regions of chromosomes 8 (Rhodamine) and 18 (FITC) in the DAPI-labelled nucleus of amniotic fluid cell of a foetus with Edward's syndrome, based on optical sectioning. We applied two SP-750 filters to block IR excitation light.

A further advantage of multiphoton multicolour FISH is the intrinsic possibility of three-dimensional fluorescence imaging. Optical sectioning allows the analysis of the architecture of the human genome. Figure 9(d) demonstrates an example of 3D multiphoton multicolour FISH in prenatal diagnosis. The 3D reconstruction from a set of multicolour images shows the intranuclear localization of the FISH-labelled centromeric regions of chromosomes 8 and 18 in an interphase nucleus of an amniotic fluid cell. The detection of three FITC signals in different planes indicates the presence of three chromosomes 18, suggestive of chromosomal abberation (Edward's syndrome).

5. Non-invasive optical biopsies

Besides multifluorophore excitation and optical sectioning, a further advantage of excitation radiation in the 700–1100 nm spectral range is the high light penetration depth due to the lack of efficient endogenous cellular absorbers and low scattering coefficients (Centonze & White, 1998). In principle, endogenous and exogenous fluorophores in thick samples, such as biopsies or living tissue, can be investigated to depths of more than 100 µm. Major depth limitations are the working distance of the objective, aberration phenomena, and multiple scattering. In turbid media with anisotropic scatterers, such as human tissues, multiple scattering of incident photons will induce a spatial and temporal dispersion of the excitation pulses and a subsequent signal reduction. Using a large-area detector (e.g. at the baseport), the effect of multiple scattered non-absorbed fluorescence photons on signal reduction can be minimized.

In order to get information of the spatial resolution and the signal intensity, we measured two-photon excited point spread functions (PSF) and signal-to-noise ratios (SNR) in fresh biopsies of human skin and of solid Ehrlich carcinoma mounted in phosphate buffered saline (PBS) using yellow-green fluorescent microbeads at various tissue depths. Using a 40× oil immersion objective, the PSF increased from values of 0.34 µm and 0.90 µm for radial and axial direction at 0 µm depth to values of 0.45 µm and 1.53 µm at 50 µm skin depth (Fig. 10a). As seen from Fig. 10(b), a typical increase in laser power by a factor of 10 is required to obtain the same SNR of bead fluorescence in 100 µm tumour tissue depth compared to beads at the surface. Therefore, fluorescence images with sufficient resolution can be obtained in deep tissue with moderate excitation powers.

Figure 10.

(a) Two-photon point spread functions obtained from 3D images of 0.17 µm fluorescent latex spheres in human skin tissue at a depth of 50 µm (black) compared to beads on coverslip (white) without tissue. A 40× objective (NA = 1.30, mean working distance: 120 µm) and 170 fs pulses at 800 nm were used. (b) Mean power needed to maintain same SNR vs. tumour tissue depth.

NIR microscopy using cw laser sources and operation in the reflection mode has been used for fast in vivo imaging of human skin. However, fluorophores have not been detected with this system (Rajadhyaksha et al., 1999). By contrast, multiphoton NIR autofluorescence microscopy provides a method to obtain non-invasive optical biopsies with high spatial resolution and to perform morphological as well as functional imaging (Piston et al., 1995; Masters et al., 1997). In addition to coenzyme and elastic fibre imaging, detection of non-linearly excited melanin fluorescence has also been demonstrated (Teuchner et al., 1999).

Figure 11 presents depth-dependent 730 nm excited autofluorescence images of fresh biopsies of human skin (breast) with emission maxima in the blue/green spectral range. Potential endogenous two-photon excitable fluorophores are NADH, NADPH, flavins and collagen. Intracellular structures, cell morphology and location of the nucleus in the different skin layers are clearly seen.

Figure 11.

Two-photon excited autofluorescence of human skin at different tissue depths.

Preliminary in vivo autofluorescence imaging of the human forearm has been performed with a scan rate of 0.25 s−1 (512 × 512 pixels). In addition to the oil immersion objective, measurements were also made with a water immersion objective (Zeiss, C Apochromat 40×, 1.2 W). High-resolution autofluorescence images along the entire 200 µm working distance can be obtained. However, it should be noted that the scan rate is limited due to the low fluorescence quantum yield of the endogenous fluorophores. Nevertheless, stacks of more than a hundred high contrast, motion-free autofluorescence images from different tissue depths have been obtained from patients without anaesthetic treatment. More rapid scanning is possible when exogenous fluorophores with a high fluorescence quantum yield are employed.

An example of in vivo fluorophore imaging is shown in Fig. 12, where the accumulation of protoporphyrin IX (PP IX) in tumour cells of mice with solid Ehrlich carcinoma has been detected. The excitation of the red-fluorescent porphyrin in the cytoplasm at 800 nm results also in excitation of endogenous blue/green-emitting fluorophores (NAD(P)H, flavins) associated with mitochondria. The use of special long-pass or broadband filters enables selective porphyrin imaging.

Figure 12.

Two-photon excited protoporphyrin in tumour tissue of a mouse with solid Ehrlich carcinoma.

To my knowledge as yet no clinical trials have been conducted using multiphoton microscopes.

6. Multiphoton photochemistry in living cells

NIR radiation can induce the formation of reactive oxygen species (ROS) by non-linear excitation of photosensitizing molecules and subsequent type I or type II photo-oxidation processes. ROS-mediated lethal effects are used in photodynamic therapy (PDT) of tumours and other diseases (Doiron & Gomer, 1984). Typically, an exogenous photosensitizer such as the porphyrin mixture Photofrin with a high potential of singlet oxygen formation is applied to the target cells. Another well-known photosensitizer is PP IX, which is synthesized in the mitochondria of the tumour cells following uptake of the porphyrin precursor 5-aminolevulinic acid (ALA) (Kennedy et al., 1990). Light excited porphyrin photosensitizers react with oxygen, mainly by a type II photo-oxidation process. Energy transfer occurs from the metastable triplet state to molecular oxygen and results in the formation of reactive singlet oxygen which induces cytotoxic effects. Although the major absorption band of the porphyrins is at around 400 nm, conventional porphyrin PDT is based on 633 nm radiation due to the higher light penetration depth. However, the absorption coefficient is about a factor of 30 less than the value at the major band at about 400 nm (Soret band, ε = 6 × 104 M−1 cm−1, Photofrin in PBS). The application of intense NIR radiation at about 800 nm is an alternative approach to excite the porphyrin photosensitizers.

A number of efforts to use resonant or nonresonant two-photon excitation of photosensitizers have been reported. Andreoni et al. (1982) demonstrated sequential two-photon photoionization of haematoporphyrin. However, this resonant process requires UV or visible laser light. Experiments with 1.06 µm radiation of a Nd:YAG laser and the photosensitizer Phaeophorbide led also to undesired thermal effects (Yamashita et al., 1991).

Here we present data on the effect of 780 nm femtosecond laser scanning microscopy on the reproduction behaviour − a very sensitive indicator of cell damage – and vitality of Photofrin-labelled and ALA-labelled CHO cells. Cells were incubated with Photofrin for 12 h (5 µg mL−1) and with ALA (1.5 mg mL−1) for 4 h and exposed to 170 fs pulses at a mean power of 2 mW (130 W peak power, about 190 GW cm−2 peak intensity). A frame rate of 0.0625 s−1 (16 s frame−1) was chosen. With a typical cellular area of 700 µm2, the beam dwell time on one cell was ≈ 400 ms per scan. Subsequent scanning of Photofrin-labelled cells reduced cloning efficiency to 50% (only half of the cells show normal cell division) and to 0% (no reproduction) after about 13 scans (≈10 mJ) and 50 scans, respectively. In the case of ALA incubation, about 24 scans and 100 scans were necessary. Live/dead assays revealed cytotoxic effects in most of the cells after 50 scans for Photofrin-labelled cells and 100 scans for ALA-labelled cells. The cell destruction was accompanied by fluorescence fading. Sensitizer-free control cells could be scanned more than 700 times (4 h) without impact on the reproduction behaviour, morphology and vitality (Fig. 13) (König et al., 1999c).

Figure 13.

Cloning efficiency of Photofrin-labelled and ALA-labelled cells vs. scan number after laser exposure with a 200 fs, 2 mW scanning beam at 60 µs pixel dwell time per scan.

Two-photon excitation of intracellular porphyrins with NIR femtosecond pulses therefore seems an appropriate method to induce cell damage by photodynamic reactions. Major advantages of NIR porphyrin PDT are the excitation of higher order singlet states (‘Soret band excitation’), the high light penetration depth of 780 nm radiation and the possibility to induce the desired effect highly localized within the three-dimensional target. Type I photodynamic therapy agents have also been excited by two-photon excitation with femtosecond laser pulses (Fisher et al., 1997). The same group reported also on melanoma treatment using ultrashort NIR pulses (Wachter et al., 1999). Further studies on NIR-exposed tissues and the development of new equipment for intense NIR illumination are necessary to perform PDT of cancer and other pathological disorders by non-resonant two-photon excitation.

Another two-photon induced photochemical effect is the highly localized photorelease of caged compounds with NIR laser radiation. The photorelease (photocleavage, uncaging) of bioactive substances, as well as of photolabile ion (Ca2+ and Mg2+)-specific chelators, have been reported (Denk et al., 1990; Piston et al., 1994; Brown et al., 1999). Using two-photon femtosecond IR laser mediated photolysis of newly developed caged glutamate (Bhc-glu), Furuta et al. (1999) have recently obtained 3D maps of glutamate sensitivity of neurones in intact brain slices from rat cortex and hippocampus. Furthermore, Lipp & Niggli (1998) have used two-photon excitation photorelease of caged Ca2+ to spatially increase Ca2+ rapidly in extremely small volumes (10−15 L) of mammalian cardiac cells. Moreover, two-photon induced photorelease has also been employed for mapping the distribution of ligand-gated receptors in neuronal cells (Denk, 1994) and has been used to track the cell lines during development of sea urchin embryos (Summers et al., 1996). Hence, the possibility to control highly local events in living cells with multiphoton photorelease and to image them with ultraprecise spatial resolution invariably constitutes an entirely new level in the study of cell signalling as well as in developmental biology.

7. Multiphoton microscopy in plant biology

In plants most of the light absorbed by photosynthetic pigments such as chlorophyll and carotenoids is utilized for photosynthetic quantum conversion, while only a small portion is de-excited through emission as red and far-red chlorophyll fluorescence. The red and far-red chlorophyll emission spectrum is characterized by two maxima near 690 nm and 740 nm. The fluorescence induced by a 355 nm Nd:YAG laser has been used for detection of stress-induced changes in the red and far-red fluorescence emission (Lichtenthaler & Miehe, 1997). We have instead successfully employed femtosecond NIR pulses at 800 nm to detect chlorophyll autofluorescence in intact root hypocotyl (Fig. 14a and b) and in leaf epidermis (Fig. 14c and d). Hence, with multiphoton fluorescence microscopy one can assess the photosynthetic performance of a single cell deep in the green plant tissue (Tirlapur & König, recent unpublished results). It is equally interesting to note that in the hypocotyl, in addition to the chloroplasts, the cell walls also show autofluorescence when excited at 800 nm. Similar cell wall associated autofluorescence can be seen following UV excitation, and it is generally believed to be due to blue-green fluorescing ferulic acid covalently bound to cell walls (Buschmann & Lichtenthaler, 1998). While chlorophyll can potentially be excited by far-red wavelengths by a one-photon process, 800 nm excited chlorophyll fluorescence followed a squared relation on laser power, indicating a two-photon process.

Figure 14.

Two-photon excited chlorophyll autofluorescence in hypocotyl (a and b) and leaf epidermal cells (c and d) of Arabidopsis induced with 800 nm laser pulses. Image a is generated by emboss filtering from fluorescence data.

The green fluorescent protein (GFP), a novel marker protein derived from the jellyfish Aequorea victoria, with its autofluorescence characteristics, has proved to be invaluable in the study of gene expression, trafficking and organelle structure within living cells. GFP has been targeted to various cell organelles such as the nucleus, endoplasmic reticulum, Golgi and plastids (Haseloff & Siemering, 1998; Llopis et al., 1998; Tirlapur et al., 1999). In most of these studies the subcellular localization of the fusion protein in living cells has been monitored by a conventional confocal laser scanning microscope using the 488 nm laser line. Wild type GFP emits green (maximum 508 nm). Recently, GFP-expressing neuronal cells have been imaged by two-photon microscopy (Mainen et al., 1999). Similarly, using NIR femtosecond laser radiation at 780 nm and 800 nm we have been able to image endoplasmic reticulum (ER) as well as plastid targeted GFP in intact living root (Fig. 15a and b) and in the stomatal guard cells (Fig. 15c and d) of an etiolated leaf in Arabidopsis.

Figure 15.

Imaging of two-photon excited GFP fluorescence in plant cells expressing ER- (a and b) and plastid- (c and d) targeted GFP. Time-lapse imaging of dye coupling in Arabidopsis root tip (e: 0 s, f: 8 s, g: 2 min) following NIR laser mediated dye permeation in a single central cell. Images (a) and (c) were obtained by emboss filtering.

Besides the use of multiphoton microscopy for fluorescence imaging it can aslo be used for non-invasive introduction of various probes into living plant cells. Intracellular transport of macromolecules between plant cells occurs symplastically through plasmodesmata. Knowledge pertaining to symplastic communication in plants is often analysed by visualizing the cell-to-cell movement of plasma membrane-impermeable fluorescent dyes. Usually the dye is introduced into a single cell by application of invasive microinjection methods (Goodwin et al., 1990). However, the microinjection technique cannot be used to introduce fluorescent dye in a single cell that is deep in the plant tissue. In complete contrast to physical intrusion with a microinjection needle, a novel non-invasive technique based on NIR femtosecond laser pulses for dye permeation and 3D imaging of dye coupling (Fig. 15e–g) between different cell types in the root meristem of Arabidopsis has been developed (Tirlapur & König, 1999). Because the NIR-laser based approach is highly efficient it has a number of potential applications for non-contact, non-invasive loading of fluorescent probes in a wide range of cell types as well as in studies addressing cell–cell communication.

8. Cell damage in multiphoton microscopes

As described in section 6, femtosecond laser microscopy of living specimens can be performed with a peak intensity of about 200 GW cm−2 for hours without impact on cellular reproduction and vitality. Squirrel et al. (1999) exposed hamster embryos for 24 h to NIR femtosecond laser pulses of a multiphoton fluorescence microscope without impact on embryo development, in contrast to conventional one-photon laser scanning microscopy.

However, non-destructive multiphoton imaging of living specimens is confined to a particular optical window. The lower limitation is determined by the optical properties of the fluorophore (e.g. molecular multiphoton absorption cross-section and fluorescence quantum yield) and by the detector sensitivity, the upper by the onset of cell damage.

We investigated the influence of cw and ultrashort laser pulses at different NIR wavelength, power and pulse width on the metabolism, reproduction behaviour and vitality of CHO cells. Cells were exposed 10 times to a highly focused scanning with 80 µs beam dwell time per pixel (König et al., 1996b,c, 1997, 1999).

For the same pulse energy, photodamage was found to be more pronounced at shorter pulses. At 780 nm, cw scanning beams did not harm cells up to 35 mW power, whereas 2 ps pulses induced loss of viability at 22 mW mean power. Impaired cell division was found for half of the 2 ps exposed cells at 11 mW mean power (P50). By contrast, 240 fs pulses led to 50% cloning efficiency and cell death at mean powers of 3 and 7 mW, respectively (Fig. 16). Using laser powers of 2 mW or less, CHO cells could be scanned more than 700 times without impact on the reproduction. There is therefore no simple relation between photodamage and energy density. Below certain power (intensity) thresholds, ‘safe’ multiphoton microscopy appears possible.

Figure 16.

Photodamage resulting from exposure to femtosecond and picosecond laser pulses. The mean powers of 780 nm pulses to induce failed cellular reproduction of half the laser-exposed CHO cells (50% cloning efficiency) vs. pulse width is shown. In the smaller included image, the relative mean power vs. squared root of pulse broadening M is shown. The damage process follows approximately a P2/τ relation.

According to the smaller image in Fig. 16, where the P50 values vs. the root of the pulse broadening are depicted, the photodamage process is probably based on a two-photon excitation process rather than a one-photon or three-photon event. This is consistent with the findings that the cell damage at 920 nm was less pronounced than at 780 nm where water absorbs less. Photodamage is therefore not based on linear water heating. This was also shown in cw 1064 nm microbeam studies where CHO cells, labelled with a thermosensitive fluorophore, were exposed to MW cm−2 intensities and GJ cm−2 fluence, as well as by temperature calculations in the case of exposure to fs and ps pulses (Liu et al., 1995; Schönle & Hell, 1998).

Using 780 nm pulses, three-photon absorption may result in nuclear DNA excitation due to the major absorption band of nucleic acids at 260 nm. However, photoinduced cell damage such as failed reproduction was also obtained when light exposure was restricted to intracellular regions outside the nucleus.

The findings that photodamage seems to be based on a two-photon excitation process has the consequence that picosecond as well as femtosecond pulses are appropriate sources of two-photon microscopy. Because of the same P2/τ dependence of damage and fluorescence, the ratio (typically 1–30) of minimum powers to induce cell damage and to excite fluorescence via a two-photon absorption process is the same for picosecond and femtosecond laser pulses. It makes therefore no sense to use high-cost pulse compression units. By contrast, in three-photon imaging, which requires higher photon flux densities due to a P32 relation, the relative window will be narrower and will decrease with increasing pulse length. In this case, fs pulses are appropriate sources in the case of efficient non-destructive three-photon imaging.

In order to compare the photodamage in NIR multiphoton microscopes vs. UV one photon microscopes, the same assay was performed on CHO cells using a 364-nm scanning beam of a conventional laser scanning microscope with an Ar+ laser. A 50% cloning efficiency was obtained at the low power of 4 µW. Therefore, UV-induced damage occurs at about three orders lower powers (4 µW) compared to the mean values of NIR femtosecond laser pulses (2–8 mW).

With respect to the excitation wavelength, the photodamage can be based on two-photon absorption by endogenous cellular absorbers, such as by the coenzymes NAD(P)H and flavins as well as by porphyrins. It is well-known that excitation of these endogenous photosensitizers can result in the formation of ROS (oxidative stress) and indirect DNA damage (Tyrell & Keyse, 1990; Cunningham et al., 1985). The major localization site of the coenzymes is the mitochondria. Our ultrastructural studies reveal that these organelles are the major targets of NIR irradiation. Laser exposure resulted in swollen mitochondria accompanied by loss of christae combined with the formation of electron-dense bodies in the mitochondrial matrix (Oehring et al., 2000).

ROS play an important role in apoptosis. We found NIR-induced apoptosis-like cell death. In particular, changes in cytoplasmic and nuclear calcium levels, fragmentation of F-actin cytoskeleton, disruption of plasma membrane integrity and morphological deformations of the nuclei and nuclear envelope following exposure with 170 fs laser pulses with powers higher than 7 mW have been detected (Tirlapur & König, recent unpublished results).

On increasing the power levels to values higher than 10 mW, another destructive effect may occur. The use of very high light intensities in the TW cm−2 region induces intracellular optical breakdown and plasma formation accompanied by intense white luminescence and complex sub-nanosecond emission decay kinetics (König et al., 1996c). Scanning at these light intensities results in severe shockwave-induced morphological damage, such as cell fragmentation. Typically, cells die immediately.

In conclusion, multiphoton fluorescence imaging appears to be a safe novel optical sectioning technique within a particular intensity window. Above the threshold, failed reproduction and apoptosis-like lethal effects probably based on two-photon photochemical photodamage as well as on immediate cell damage via optical breakdown may occur.

9. Nanosurgery

If the mean power in 80 MHz femtosecond laser microscopes with high numerical aperture objectives is increased to 30–50 mW, light intensities of TW cm−2 can be used for material processing based on plasma-induced ablation. The threshold for optical breakdown is lower for biological material than for water and cell culture medium and was found to be lowest in the perinuclear region of cells. We used the modes of ROI scanning, line scan and single spot illumination to knock out intracellular material, to perform dissections (König et al., 1999b) and to ‘drill’ holes.

By fine tuning of the laser power only the central part of the beam spot can provide sufficient intensity for material removal. It is therefore possible to drill holes and to cut structures with a precision below the diffraction-limited submicrometre beam spot. We used the multiphoton-induced ablation process to cut human chromosomes. Measurements with the force scanning microscope reveal chromosome dissection with a cut size below 300 nm. In addition, partial removal (incisions) of chromosome material with a precision of 110 nm was obtained (Fig. 17). The figure demonstrates probably the smallest laser cuts in biological materials. The femtosecond laser microscopy opens the possibility to provide a non-contact nanoscalpel for surgery inside the cell, cell nucleus or organelle without affecting other cellular compartments. We were able to cut chromosomes within a living cell (Fig. 18). The cells remained alive and completed cell division after laser surgery.

Figure 17.

Nanodissection of chromosomes with 170 fs pulses at 800 nm and 80 MHz. A cut size of 110–290 nm was measured with the force scanning microscope.

Figure 18.

Chromosome dissection within living cells using the femtosecond NIR laser microscope. The cell remained alive.

In conclusion, femtosecond NIR laser pulses with their advantages of high penetration depth, no out-of-focus absorption, efficient induction of multiphoton processes, post-pulse material removal, absence of plasma shielding effects and of significant heat transfer appear the appropriate choice for intracellular surgery. In contrast to laser microsurgery formerly performed in living cells with UV and visible lasers at typically nanosecond pulse width (Bessis et al., 1962; Berns et al., 1969; Berns, 1974; Greulich et al., 1989; Greulich & Weber, 1992; Schütze et al., 1994; Berns, 1998; Greulich, 1999), NIR nanosurgery with ultrashort laser pulses provides higher precision, and the possibility to treat structures deep within the tissue and to avoid severe damage to surrounding structures.

10. Conclusions

NIR multiphoton excitation laser scanning microscopes provide attractive advantages over conventional fluorescence microscopes and can be employed as novel non-contact biomedical tools for three-dimensionally resolved fluorescence imaging, optical diagnostics, photochemistry and nanoprocessing within submicrometre domains of biomolecules and living cells. Because multiphoton excitation at high intensities occurs only in the minute focal volume of a high numerical aperture objective it inherently provides three-dimensional resolution, while photobleaching and photodamage are minimized in out-of-focus regions. These advantages allow experiments on thick living samples for 3D fluorescence diagnostics with sub-nanosecond and sub-micrometre resolution, DNA and live cell studies, and to acquire optical biopsies. Furthermore, these novel microscopes have versatile applications for photodynamic inactivation of tumour cells, release of photolabile caged compounds and intracellular nanosurgery. It is therefore conceivable that future technical advances will continue to increase the range of applicability of multiphoton laser scanning microscopes in biomedical and biotechnological sciences.

Acknowledgements

The author thanks Professor K-J. Halbhuber, Dr I. Riemann, Dr P. Fischer (Institute of Anatomy II, University of Jena) and Dr W. Fritzsche (IPHT Jena) for scientific support and Dr U. K. Tirlapur for constructive comments and for critically reading this manuscript. The financial support from Carl Zeiss Jena GmbH, the German Science Foundation (DFG) and the Federal Ministry of Education, Science, Research and Technology (BMBF) is gratefully acknowledged.

Ancillary