Multi-Dimensional Fluorescence Imaging

3D multi-dimensional fluorescence imaging (MDFI) is the integration of optically-sectioning fluorescence microscopes with spectral, temporal and polarisation resolution.

Fluorescence microscopy is perhaps the most powerful functional imaging modality available to study biological function. The development of genetically expressed fluorophore labels such as green fluorescence protein (GFP) offers an unparalleled opportunity to follow the location of specific proteins inside cells by imaging fluorescence intensity distributions. To obtain quantitative images, however, and to study cell (and protein) function, it is often necessary to do more than just image fluorescence intensity. One established way to gain quantitative information concerning fluorophore distributions is ratiometric fluorescence imaging, applying spectral or temporal resolution. This can make the measurements more robust against changes in quantum efficiency or scattering losses. It is therefore important that the fluorescence microscope system for this project is capable of temporal and spectral resolution.

For many situations, it is preferable to avoid the necessity of labelling live cell components with an extrinsic fluorophore, since these can significantly perturb or damage the cell in question. Multi-dimensional fluorescence imaging of autofluorescence offers the opportunity to distinguish different cell components without exogenous fluorophores, i.e. it can potentially provide label-free contrast. Under our current DTI Beacon project, this approach is being applied to tissue imaging for research and clinical diagnosis application and is also exciting significant interest for molecular biology, since it could provide important new information both in vivo and in vitro. Label-free contrast in live cells has received relatively little attention to date, however, because autofluorescence typically requires u.v. excitation and because intensity imaging at the commonly used excitation wavelengths available from ion lasers has not provided much useful contrast. Applying temporal, spectral and polarisation contrast to autofluorescence can greatly increase the scope for intrinsic contrast, as discussed below.

Contrast in fluorescence imaging
In principle, straightforward fluorescence intensity imaging should report the distribution of a fluorophore, with the intensity being a function of the local fluorophore concentration. In practice the observed fluorescence intensity may originate from multiple fluorophores (with different quantum efficiencies), the quantum efficiency of an individual fluorophore population may change across a sample and the detected intensity may be modified by optical scattering and absorption in the sample. Thus it is often necessary to eliminate these uncertainties using ratiometric techniques or to address them using more sophisticated measurements to obtain more information. Further information can be gained from the quantum efficiency, which is a function of the radiative (g) and non-radiative (k) decay rates and so is associated the state or local environment of a fluorophore molecule. The radiative decay rate is essentially a function of the electronic energy level structure of the fluorophore molecule (although it does depend weakly on the local refractive index) while the non-radiative decay rate can be a sensitive function of the local fluorophore environment. Thus by observing changes in quantum efficiency, one can, in principle, probe local variations in e.g. viscosity, temperature, refractive index, pH, [Ca], [O2], electric field, etc, depending on the nature of the particular fluorophore molecule. Unfortunately determination of the quantum efficiency from intensity measurements requires knowledge of the absorbed and emitted radiation fluxes and of the fluorophore concentration. This is difficult or impossible to achieve in heterogeneous media such as biological tissues, which are also strongly scattering, and is not possible with autofluorescence.

Spectral resolution of fluorescence can provide functional information when the spectral profile changes as a function of the molecular state or environment but it is not straightforward to engineer fluorescent labels whose spectral profile changes in a predictable manner to permit quantitative imaging and nature is not always so kind as to provide them. Spectral discrimination is important, however, when contrasting different fluorophores – both in excitation and emission. This is particularly important when imaging in heterogeneous biological tissue where multiple fluorophores can present a significant background to measurements on a specific fluorophore.

Fluorescence lifetime imaging
Just as spectral discrimination adds a new dimension to fluorescence data, so temporal resolution adds a further dimension with opportunities for contrast and functional imaging. It will be seen from Figure 1 that the fluorescence lifetime (?) also depends on both the radiative and non-radiative decay rates and it can be used to contrast different fluorophore species (via ?) and different local fluorophore environments (via k). Unlike the quantum efficiency, ? can be determined from relative intensity measurements with no requirement for knowledge of the concentration of fluorophores or the excitation and emission fluxes. FLIM therefore offers a “functional” imaging modality that can provide molecular contrast or imaging of distributions of fluorophore environments. This may be applied to imaging biological samples (cells or bulk tissue), for correlation with morphological information, or it may be applied to arrays of samples for high-speed microanalysis.

Figure 2 shows examples of FLIM of arrays of samples in multi-well plates. The contrast parameter of these lifetime maps is fluorescence lifetime, which is represented on a false colour scale. Figure 2(a) illustrates chemically specific imaging that contrasts two different dyes with very different lifetimes (140 ps and 3.5 ns), illustrating the excellent temporal dynamic range of this technique. Figure 2(b) shows how the fluorescence lifetime of the dye DASPI varies as a function of the solvent viscosity, illustrating how FLIM can non-invasively image variations in physical (or chemical) parameters that impact the local fluorophore environment. Figure 2(c) shows how the local solvent refractive index can be reported by the fluorescence lifetime of enhanced green fluorescence protein (EGFP). This recent observation is interesting for cellular imaging where EGFP is widely used and where its fluorescence lifetime is usually (erroneously) assumed to be constant throughout the cell. Figure 3 shows an immune synapse between an NK cell and a human B cell in which the MHC protein has been labelled with EGFP. There is clearly a change in lifetime at the immune synapse compared to the rest of the membrane – as confirmed by a new quantitative analysis we have developed at Imperial .

Polarisation-resolved fluorescence imaging (anisotropy) imaging can also provide information concerning the fluorophore environment. We have recently demonstrated a new technique to image the linear dichroism of membrane fluorophore probes to report on the degree of lipid order . Combining FLIM with polarisation resolution can provide further information concerning the fluorophore rotational mobility .

Autofluorescence can be used to provide intrinsic contrast between different types or states of tissue, as illustrated in figure 4, which shows an autofluorescence image and FLIM map of unstained rat tissue samples . The collagen is taken from rat tail and the elastin samples are aorta strips that have boiled in 0.1 M NaOH for 30 minutes to remove impurities. This figure illustrates how FLIM is able to contrast collagen from elastin and aorta, also showing how fluorescence lifetime contrast is independent of intensity contrast (and so is independent of intensity artefacts). There is little spectral contrast between these tissues for this excitation wavelength of 415 nm. We believe that it is the cross-linkages of these tissue proteins that are responsible for the fluorescence. It is particularly exciting that the aorta and elastin samples are contrasted, since this suggests that FLIM can provide a non-invasive quantitative indication of the state of the protein cross-linkages that have perhaps been compromised by the process of hydrolysis.

Combining FLIM with spectral discrimination provides further scope to enhance intrinsic contrast (since spectral and temporal information are complementary). We have demonstrated wide-field FLIM combined with multi-spectral imaging applied to biological tissue and seen the potential to further enhance intrinsic contrast, as indicated in the sample of rabbit artery of figure 5.

Recent research achievements
The complexity of picosecond-resolved time-gating has hitherto prevented most groups from exploiting this source of fluorescence information. Our programmes to develop robust and widely deployable FLIM instrumentation at Imperial have led to a number of new instruments in our laboratories and elsewhere (through sales from our partner, Kentech Instruments Ltd). This instrumentation programme based in Physics at Imperial College London has evolved into a major interdisciplinary cluster of projects spanning all four Faculties and the Business School. The Physics team, working with colleagues from Bioengineering, Biology, Chemistry and the Faculty of Medicine have developed a world-leading novel time-domain FLIM technology platform. This work, which builds on the pioneering development by David Phillips’ group of fluorescence lifetime measurement and FLIM and their application to the fundamentals of fluorescence contrast in biomedicine, has been supported by EPSRC and BBSRC, as well as the DTI (in the form of a Teaching Company Scheme award with Kentech Instruments Ltd and recently a DTI Beacon Award) and a Wellcome Trust Showcase Award. Early highlights included the world-leading lifetime discrimination in a wide-field FLIM system and the first application of 415 nm excitation FLIM to tissue autofluorescence 2. We also demonstrated the first time-domain wide-field FLIM system to use a diode laser . This portable, low-cost source provides ps pulses at ~ 400 nm and has been successfully applied to tissue autofluorescence and to multiwell plate imaging.

Under the BBSRC Bioimaging Initiative we developed the world’s first wide-field 3-D FLIM microscope system , in collaboration with Tony Wilson’s group at Oxford, implementing this using custom-built diode-pumped Cr:LiSAF laser technology, developed in our laboratory to demonstrate the potential for compact portable systems . We extended this to rapid multi-spectral FLIM and 5D fluorescence imaging 6, resolving 3 spatial dimensions with lifetime determination and spectral resolution simultaneously Supported by BBSRC, we applied this microscope to EGFP expressed in cells to study NK cell interactions and demonstrated that the fluorescence lifetime of EGFP reports variations in the local refractive index 2. We also demonstrated FLAIM, the world’s first time-domain wide-field microscopy Fluorescence Lifetime and Anisotropy Imaging technique, which provides both FLIM maps and maps of the polarisation anisotropy and rotational correlation time 4 and applied this to multi-well plate and cellular imaging . Under the auspices of our DTI Beacon project, we have developed video rate FLIM microscopy with both single and multi-photon excitation. The latter has been implemented in a commercial multibeam multiphoton microscope, which has also been modified to provide multi-spectral (2 channel) and hyperspectral imaging, as well as polarisation-resolved imaging. This multiphoton hyperspectral imaging has been combined with FLIM microscopy to obtain the images shown in figure 5 and polarisation-resolved multiphoton imaging was used to develop our novel linear dichroism technique. The latter was applied to time-lapse studies of lipid order following the addition of cyclodextrin to remove cholesterol from the membrane.

Under the EPSRC Physics for Healthcare initiative we demonstrated an endoscopic FLIM system, which has been further refined throughout our DTI Beacon project. Recently we have demonstrated real-time endoscopic FLIM of multiwell plate dye and unstained tissue samples . We have also worked extensively on the analysis of fluorescence lifetime data, with particular emphasis on the complex exponential decay profiles usually found for tissue autofluorescence. We have shown that the stretched exponential decay function may be used to fit FLIM data of heterogeneous samples exhibiting distributions of fluorescence lifetime . Figure 7 shows tissue sections fitted to stretched exponential decays and also shows an inverse Laplace transform of the lifetime data, confirming that there is indeed a distribution of lifetimes in tissue fluorescence. Figure 7 also illustrates that this approach provides an additional contrast for tissue imaging through the heterogeneity parameter. We have investigated intrinsic FLIM contrast between normal breast tissue and tumours and investigated intrinsic contrast within tendon and ligament samples. Our most recent tissue imaging has demonstrated intrinsic contrast in cartilage, atherosclerotic plaque and pancreatic tissues.

Selected references