Development of a combination spectral/lifetime detector. We are currently developing a combined spectral/lifetime detector that is optimized for low-light level multiphoton imaging. The detector works in photon counting mode and essentially sorts detected photons into spectral and temporal bins. This detector is being developed primarily for the Optical Workstation but will also be used with the high-speed multiphoton imaging system currently under development.

Spectral imaging is the collection and display of the spectral components of a fluorescence image. LOCI is actively engaged in the development of a novel spectral detector that is optimized for multiphoton imaging. The spectral detector will be implemented on the Optical Workstation and the high-speed multiophoton imaging system currently under development.

Benefits of spectral imaging. Most commercial confocal and multiphoton microscopes currently have the ability to collect two or three pre-specified colors simultaneously. However, there is often a need for more complete spectral information to allow the detection of more fluorophores and to facilitate the setting of spectral windows to optimize detection of a specific fluorophore. The main goal of the detection system is to collect the desired signal in the presence of noise (detection noise, system noise, fluorescent background etc.). Background fluorescence from endogenous fluorophores or from another interfering exogenous fluorophores can severely reduce detection &endash; or interpretation -- of the image signal. With multiple labeled samples, the signal from one fluorophore is often much stronger than another and can spill over to an adjacent channel. This problem is exacerbated by fluorophores with extended red tails: DAPI, for example. In these instances it is often better to move the spectral detection windows as far apart as possible to aid discrimination between the two fluorophores being studied rather than choosing spectral windows to give the maximum signal in each channel.

The use of multiple fluorescent labels has long been commonplace in the study of fixed specimens, and is now becoming established for in vivo studies. Not so long ago only three fluorophores were in widespread use (fluorescein, rhodamine and DAPI); now there is a plethora of fluorophores available, each with its own unique spectral characteristics. This has generated a considerable problem for fluorescence microscopists in that many different filter sets are required for double or triple labeled samples. Filter sets use expensive interference filters and dichroic mirrors and are often difficult to interchange. Ideally, filters should be continuously adjustable so that for any particular combination of fluorophores used, an optimal set of band-pass assignments can be selected for each detection channel to minimize signal bleed-through and maximize the signal-to-noise ratio. Perhaps the greatest power of collecting the entire spectrum is this allows fluorophore to be identified and separated computationally (by comparison to reference fluorophore spectra) in the presence of high levels of background.

Most biological tissue is autofluorescent. Molecules such as NAD(P)H, elastin, and chlorophyll act as endogenous fluorophores. Often, these endogenous fluorophores can be identified by their characteristic spectra. A spectral imaging system is of considerable use in identifying endogenous fluorophores and specifying spectral windows that would either maximally accept or reject these signals, depending on the application. Additional information may obtained by comparing spectra obtained at diferent excitation wavelengths.

The use of engineered fluorescent probes as physiological indicators has become a well-established technique. Some probes indicate the presence of a bound ligand by changes in fluorescence intensity (e.g. Calcium Green 1) while others use spectral shifts (e.g. Indo 1). The later are favored because ratio imaging at two different wavelengths may be used to provide measurements that are independent of the concentration of the indicator molecule, measurements that are quantitative. Spectral detection allows an optimum set of spectral windows to be used for ratio imaging.

Fluorescence resonant energy transfer (FRET) is a powerful technique for measuring intermolecular distances in vivo (dos Remedios & Moens, 1995. J. Struct. Biol. 115, 175). This technique also requires custom filter sets that are matched to the donor and acceptor molecule's emission spectra. Ratiometric measurements are used to measure the extent of resonance transfer. FRET is proving to be a valuable technique for the in vivo visualization of the docking of a receptor with its ligand, and it is the basis of operation of a new GFP based calcium indicator, Cameleon (Miyawaki et al., 1997 Nature 388, 882-887).

Fluorescence in situ hybridyzation (FISH) is another very significant area where multiple fluorophores and ratiometric techniques are used (Dauwerse et al., 1992 Hum. Mol. Genet. 1, 593). Often the main requirement in this application is to spectrally resolve as many separate fluorescent probes as possible (Schröck et al., 1996 Science 273:494).

Many standard histological preparations are fluorescent. Often the spectra of the fluorophore differs in a tissue specific manner. This property could be an aid to structural identification and thence to diagnoses in pathological specimens. This is an MP image of a 200µm thick specimen of kidney tubules stained with acid fucsin. Spectral windows: top left, 580 to 630nm; top right 500 to 550nm; bottom left, 390 to 485nm; Bottom right, pseudocolor merge The specimen was prepared by Al Kutchera of Midwest Microtech, Inc

The following list summarizes the main advantages of a spectral imaging system over a conventional, filter-based three-channel detector:

• Dynamic identification of auto-fluorescence and optimization of windows for rejection or imaging as required.
• Dynamic optimization of spectral windows for multiple labels.
• Dynamic background subtraction of reference spectra before the image is even displayed
• Identification of spectral shifts of fluorophores in different environments e.g. bodipy ceramide Golgi marker or acridine orange binding to RNA or DNA.
• Full signal optimization for any given ratiometric indicator.
• Permits fluorophore separation after data collection if full spectral image is taken.
• Permits evaluation of standard histological procedures for MPLSM analysis for identification of tissue-specific spectral shifts of staining.

Fluorescence excited-state lifetime imaging. Time-resolved fluorescence spectroscopy is a well-established technique for studying the emission dynamics of fluorescent molecules i.e. the distribution of times between the electronic excitation of a fluorophore and the radiative decay of the electron from the excited stated producing an emitted photon. The temporal extent of this distribution is referred to as the fluorescence lifetime of the molecule. Lifetime measurements can yield information on the molecular microenvironment of a fluorescent molecule. Factors such as ionic strength, hydrophobicity, oxygen concentration, binding to macromolecules and the proximity of molecules that can deplete the excited state by resonance energy transfer can all modify the lifetime of a fluorophore. Measurements of lifetimes can therefore be used as indicators of these parameters. Furthermore, these measurements are generally absolute, being independent of the concentration of the fluorophore. This can have considerable practical advantages. For example, the intracellular concentrations of a variety of ions can be measured in vivo by fluorescence lifetime techniques (Szmacinski et al., 1994 Methods Enzymol. 240, 723). Many popular, visible wavelength calcium indicators, such as Calcium Green 1, give changes of fluorescence intensity upon binding calcium. The intensity-based calibration of these indicators is difficult and prone to errors. However, many dyes exhibit useful lifetime changes on calcium binding and therefore can be used with lifetime measurements (Lakowicz, et al., 1994 Cell Calcium 15, 7). This gives the considerable advantage that absolute measurements of concentration can be made with no elaborate calibration procedures required. Alternatively, lifetime measurements may be used to calibrate the intensity signals from these indicators when maximum sensitivity is required.

An exciting new development of the field has been the development of the technique of fluorescence lifetime imaging microscopy (Lakowicz et al., 1992 Anal. Biochem 202: 316; Wang et al., 1992. Crit. Rev. Anal. Chem. 23: 369; Gadella et al., 1993. J. Cell Biol. 129, 1543). In this technique lifetimes are measured at each pixel and displayed as contrast. Lifetime imaging systems have been demonstrated using wide-field (Lakowicz et al., 1992 Anal. Biochem 202: 316, confocal (Sanders et al., 1995 Anal. Biochem. 227: 302 and multiphoton (French et al., 1997. J. Microsc. 185: 339) imaging modes. FLIM combines the advantages of lifetime spectroscopy with fluorescence microscopy by revealing the spatial distribution of a fluorescent molecule together with information about its microenvironment. In this way an extra dimension of information is obtained. This extra dimension can be used to discriminate among multiple labels on the basis of lifetime as well as spectra. This would allow more labels to be discriminated simultaneously than by spectra alone in applications where many labels are required such as FISH, for example. There are also promising applications of lifetime imaging in the medical sciences. For example, tumors have been detected in mice sensitized with a hematoporphyrin derivative by lifetime imaging (Cubeddu et al., 1997 Photochem Photobiol 66(2):229).

We are particularly interested in the possibilities that are opened up by multiphoton lifetime imaging of live specimens. In these applications lifetime imaging, in conjunction with spectral imaging should greatly facilitate studies using ion indicator probes and FRET studies of intermolecular distances. For example, a remarkable calcium indicator has recently been described that is a chimeric protein based on two spectrally distinct forms of fluorescent protein (cyan and yellow) and a calmodulin molecule (Miyawaki et al., 1997 Nature 388: 882). Being a naturally fluorescent protein, genetic transformants can be made so that transformed animals will express the indicator in a range of cell types determined by the promoter. The excitation wavelength is chosen to primarily excite the cyan fluorophore. On binding calcium, the calmodulin portion of the molecule changes conformation bringing the two fluorophore regions closer together allowing resonant energy transfer between the cyan and the yellow. This will cause a shift of the emitted spectrum from cyan to yellow. The development of this engineered protein (known as Cameleon) is a remarkable development as it circumvents all the problems associated with loading probes into cells since stable transgenic lines can be used which all express Cameleon. However, one of the problems with Cameleon is that, although ratiometric methods can be used, the signal change on binding calcium is quite small making this indicator less sensitive than other indicators such as Calcium Green. Lifetime measurements are a sensitive indicator of FRET (Godella et al., 1995. J. Cell Biol. 129, 1543) and in combination with spectral measurements, should provide a more sensitive indication of calcium levels.

Techniques for lifetime imaging. Fluorescent lifetimes can be measured either in the frequency domain or in the temporal domain. Three general strategies have been used to measure fluorescence lifetimes:

• Frequency-domain imaging. In this scheme a high-frequency, modulated light source is used for fluorophore excitation. By the use of a gain-modulated detector, the phase shift and amplitude demodulation of the fluorescence signal is determined. From these data the fluorescent life-time of the probe can be calculated. This scheme is robust and has been extensively used (Wang et al., 1992). However for our purposes it suffers from several drawbacks: the detector is only working at 50% of its maximum efficiency because it is gain modulated, several data sets taken at different excitation modulation frequencies have to be taken in order to separate two or more lifetime components and finally, this scheme does not work well with photon counting techniques which we favor for the reasons described in the section describing the spectral detector.
• Time-domain lifetime imaging with gated detectors. In this scheme a gated micro-channel plate image intensifier is used in conjunction with a CCD imaging camera (e.g. Straub and Hell, 1998. Applied Physics Letters 73:1769). Spectral information is obtained by gating the image intensifier on for a narrow time-window at progressively later intervals after the excitation pulse in a succession of data frame captures. This scheme is probably the simplest way of implementing a life-time imaging system. However, it suffers from two major drawbacks for our application: the method has very poor photon utilization as only one temporal interval is detected at a time. If there are 32 intervals for example, 31/32 of the signal is not utilized and 32 separate frames have to be captured. The second reason this scheme is not appropriate for a multiphoton imaging application is that an imaging (i.e. area) detector is used. This means that the deep sectioning advantage of multiphoton imaging are not fully realized because scattered fluorescence emission photons will give rise to background noise rather than contributing to the signal as can be done with a point-scanning multiphoton system.
• Time-domain lifetime imaging with photon counting. For working at low-light levels, photon-counting detectors have considerable advantages in that they can virtually eliminate noise contributions from electronic amplifiers or electron multiplier noise in a photomultiplier. Also, photon-counting systems provide quantized pulses for every detected photon, allowing the lifetimes to be measured directly using electronic circuitry. Because of the very high speeds necessary to obtain sub-nanosecond temporal resolution., time-to-voltage converters are usually used to measure the interval between the fluorophore excitation pulse and the time of detection of the emitted fluorescent photon. Such schemes have been successfully used in practical photon-counting lifetime detectors (Kelly et al., 1997. Rev. Sci. Instrum 66(6):2279). These schemes are attractive because of their efficient utilization of detected photons. However they suffer from dynamic range problems that arise out of limited counting speeds. Typically, a time-to-voltage converter together with an associated analogue (voltage) to digital converter would have a maximum counting rate of around 1Mhz. Also, with this scheme, only one photon can be measured in the interval between laser pulses. These limitations restrict the use of this technique to low light levels when fairly long exposure times are needed in order to obtain sufficient counts for accurate representation of the decay curve. The comparatively large dead-time of this technique can have more insidious consequences. Immediately after the laser pulse, photons will be emitted at the highest rate and therefore more will be preferentially lost at this time because of the dead-time of the detector. This effect can distort the shape of the decay profile.

Applications of lifetime imaging.

At LOCI we are interested in lifetime measurements as a means of providing another dimension of information from fluorescent probes used in vivo. We find that in most applications where probes are viewed in 4-dimensions in vivo, we would benefit from more or better information. Specifically we anticipate that the combined MP spectral and lifetime imaging system will provide the following benefits to our collaborators:

• More accuracy in ratiometric probe measurements. This will be achieved by choosing optimal spectral and lifetime parameters that give the maximum shift with the probe target signal (e.g. Ca++ concentration).
• Lifetime imaging will enable some popular, non-ratiometric probes, such as calcium green, to be used in a way that is concentration independent, thereby facilitating calibration of readings.
• In conjunction with spectral imaging, lifetime imaging will improve the rejection of background fluorescence from endogenous fluorophores by the specification of optimum spectral and temporal windows. This is becoming an increasing important requirement for detecting low levels of GFP probe amid a background of autofluorescence.
• Lifetime measurements add extra information that can be used in conjunction with spectral measurements for fluorophore identification. This will be useful when there is significant spectral overlap between probes. This technique should allow the use of a greater number of probes simultaneously, such as combinations of GFP variants.
• Increasingly cell biological studies are using FRET for studies of protein/protein interactions or physiological parameters in vivo. There is usually a striking change in the lifetime of the donor and acceptor fluorophores undergoing a FRET interaction. Lifetime imaging may well prove superior to spectral ratio imaging for measuring FRET interactions. A combination of lifetime and spectral imaging will probably be better still.