|Home | About | Journals | Submit | Contact Us | Français|
Advances in microscopy and fluorescent probes provide new insight into the nanometer-scale biochemistry governing the interactions between eukaryotic cells and pathogens. When combined with mathematical modeling, these new technologies hold the promise of qualitative, quantitative and predictive descriptions of these pathways. Using the light microscope to study the spatial and temporal relationships between pathogens, host cells and their respective biochemical machinery requires an appreciation for how fluorescent probes and imaging devices function. This review summarizes how live cell fluorescence microscopy with common instruments can provide quantitative insight into the cellular and molecular functions of hosts and pathogens.
In the nineteenth century, Mechnikov took advantage of light microscopy to peer into the worlds of microorganisms and their interactions with phagocytes. By simply watching the behavior of cells, he discovered that ‘cellular eating’ or phagocytosis was a fundamental mechanism of immune defense. The optical microscope has undergone tremendous changes since then. New instrumentation and techniques for fluorescence microscopy allow observation of living cells with minimal perturbation of their normal function. While conventional molecular biological, genetic and proteomics approaches are important for identifying and characterizing protein-protein interactions, only microscopy has the potential to determine the organization and dynamics of macromolecules in the context of a living cell. Foremost among the improvements of light microscopy are the development of fluorescent protein technologies for labeling individual proteins and new probes for measuring chemical analytes such as calcium. Fluorescence Resonance Energy Transfer (FRET) of fluorescently-tagged proteins has extended the reach of fluorescence microscopes to the macromolecular scale. FRET and ratiometric microscopy allow measurement of the movements of cellular molecules as well as the molecular interactions that organize cell function. These data provide the spatial and dynamic information necessary for building mechanistic mathematical models of the cellular pathways involved in host-pathogen interactions (also see Linderman and Kirschner this issue).
Pathogenic microbes exert multiple strategies to subvert host cell functions and to modulate immune responses in the host organisms. Understanding the fundamental processes involved in these events requires detailed study of localized and transient molecular events within pathogens and host cells. For example, intracellular bacterial pathogens gain entry into phagocytic and non-phagocytic cells by manipulating molecules to alter dynamic endocytic pathways to reach their specific replicative niche. New live-cell imaging techniques hold great promise for explaining the events that control bacterial uptake, endocytic traffic and evasion of immune response. The challenges of studying molecular function by fluorescence microscopy are three-fold: they require sensitive and specific probes, adequate resolution in time and space, and some measure of how well the image reflects a relevant biological truth. This review summarizes fluorescent probe technologies that are specific for particular biochemical activities and fluorescence imaging techniques that can turn these signals into dynamic images of intracellular biochemistry. We highlight studies which have successfully applied live cell fluorescence microscopy to questions of host cell responses to infection, including studies of Listeria, Yersinia, Salmonella and Mycobacteria invasion processes. New technologies for imaging host-pathogen interactions in living tissues and in intact animals are described in recent reviews (Enninga et al., 2007; Mansson et al., 2007; Yeung et al., 2005; Piwnica-Worms et al., 2004) and in the accompanying article by Miller.
Naturally fluorescent proteins (FP) derived from jellyfish and coral have allowed the development of sensitive, genetically encoded reporters of signal transduction networks and protein dynamics in living cells and whole animals. Molecular biology and directed evolution methods have created a large variety of FPs, many of which have been optimized for spectral features and biochemical properties, such as reduced oligomerization (Nguyen and Daugherty, 2005; Shaner et al., 2004), pH-sensitivity (Miesenbock et al., 1998) and photoconversion (Patterson and Lippincott-Schwartz, 2002). FP chimeras are often expressed by transient transfection or stable transformation of cells with plasmids, by retrovirus-based expression or by other methods such as homologous recombination for stable integration of genes encoding FP chimeras (Huh et al., 2003). FP chimeras are now available in many spectral varieties with high quantum yields, including cyan fluorescent protein (CFP), yellow fluorescent protein (YFP), green fluorescent protein (GFP), red fluorescent protein (RFP) and Cherry, among others (Shaner et al., 2004). Most applications use FP chimeras to infer the distributions of their endogenous unlabelled counterparts. For example, GFP-actin chimeras report the distributions of actin in cells. A large number of research groups now use FP chimeras to analyze the spatial organization of pathways for signaling and organelle trafficking.
As with all fluorescent probes, FP chimeras present some potential disadvantages. First, the presence of the FP in the protein structure can interfere with essential functions of the natural protein, such as interactions that regulate localization or activity. Secondly, the intracellular behaviors of low-abundance proteins may be difficult to study using FP chimeras since these chimeras may be present at concentrations approaching single molecules; a situation for which GFP is usually not sufficiently bright relative to cellular autofluorescence. This difficulty is often compensated for by expression of high levels of FP chimera which may create misleading distributions in cells. Thirdly, overexpression of FP chimera expression can interfere with endogenous signaling pathways, by creating unusual conditions for regulatory proteins or target molecules. Therefore, the tremendous experimental advantages afforded by FP chimeras must be exploited with appropriate concern for the potential distortions caused by the probes. For each new FP chimera, it is important to characterize the extent to which it reports signals and perturbs pathways, although accurate methods for this are often hard to come by.
FP chimeras have been used in live-cell microscopy to follow many host-pathogen interactions. Live cell imaging implicated two members of the Ras superfamily of low molecular weight GTPases, Rab22a and Rab14, in phagosomal arrest induced by Mycobacterium tuberculosis var. bovis (BCG) in macrophages. BCG utilizes different strategies to arrest vacuolar maturation at the early endosomal stage, thereby avoiding phagosome-lysosome fusion and creating a vacuolar compartment where bacteria survive and proliferate. FP chimeras of Rab22a and Rab14 (GFP-Rab22a, EGFP-Rab14) persisted on BCG-containing phagosomes, whereas the GFP-Rab5 and GFP-Rab21 were only transiently recruited. Furthermore, Rab22a and Rab14 activities were required to prevent the acquisition of Rab7 and successive fusion with lysosomes (Kyei et al., 2006; Roberts et al., 2006). In these experiments, GFP-chimera recruitment to phagosomes was measured from 3D reconstructions of confocal fluorescence microscope images collected at many planes of focus.
Salmonella enterica serovar Typhimurium proliferates in Salmonella-containing vacuoles (SCVs) inside macrophages and epithelial cells. After SCVs form, they acquire some but not all markers of late endosomes and lysosomes (Meresse et al., 1999). To control its intracellular fate, Salmonella utilizes type III secretion systems (T3SS), which inject protein effectors across the SCV membrane to actively manipulate the host cell chemistry. The T3SS is required for the formation of highly dynamic tubular membrane structures, called Salmonella-induced filaments (SIFs), which originate from the host cell endosomal system within hours after infection. The biogenesis and dynamics of SIFs were followed by live cell imaging in epithelial cells using the lysosomal marker LAMP1-GFP and endocytosed fluorescent dextran. SIFs are unique morphological alterations of endosomes which move and extend along microtubules and are transiently connected with the SCVs (Drecktrah et al., 2008; Rajashekar et al., 2008). The SCV is connected to the endocytic pathway, a feature that may be crucial for addition of membranes for the extension of the vacuole and for bringing nutrients to the bacteria (Drecktrah et al., 2007).
Entry of Listeria monocytogenes into non-phagocytic cells utilizes cholesterol-enriched membrane microdomains and clathrin- and dynamin-dependent internalization processes (Veiga et al., 2007; Seveau et al., 2004). The assembly and disassembly of clathrin and dynamin structures at the forming vacuole containing L. monocytogenes was detected by live-cell imaging of the clathrin-GFP, and dynamin2-mRFP associated with the vacuole (Veiga et al., 2007). Although the duration of bacterial uptake (≥ 2 min) and the size of the vacuoles (1-2 μm) are significantly different from those of the classical clathrin-mediated endocytic events (0.5-3 min and ~0.1μm), these results demonstrated that lipid raft-, clathrin-, and dynamin-mediated internalization routes, which are normally utilized by host cells during receptor-mediated endocytosis, are subverted by the bacteria to access the intracellular compartment of non-phagocytic cells.
The successive steps of the biogenesis of L. monocytogenes-containing vacuoles and bacterial escape from the vacuole were determined by widefield fluorescence microscopy of macrophages expressing FP chimeras. These studies measured the recruitment over time of FP chimeras of endocytic markers, the loss of fluid-phase fluorescent markers ingested together with the bacteria, and the permeation of vacuoles by a YFP-labelled cell wall-binding domain (YFP-CBD) of the Listeria phage endolysin Ply118, as an indicator of bacterial escape (Henry et al., 2006; Loessner et al., 2002). These studies demonstrated that L. monocytogenes are internalized within compartments that rapidly mature into late endosomes (labelled with YFP-Rab7), from which they escape after 15 to 30 min, avoiding vacuolar fusion with lysosomal compartments (labelled with LAMP1-CFP) (Henry et al., 2006). Listeriolysin O (LLO), the cytolysin secreted by L. monocytogenes, plays a key role in escape. Live-cell imaging allowed identification of the compartment and timing of the escape process. A LLO-deficient bacterial strain localized within phagosomes that acquired the late endosomal and lysosomal marker LAMP1-CFP at a higher rate in comparison to the wild-type strain, and remained trapped in a degradative compartment characterized by the accumulation of LAMP1-CFP.
Methods to study molecular mechanisms of pathogenesis in living cells employ a variety of fluorescent reporters of molecular processes in living cells. Fluorescent dyes can measure intracellular ions, such as pH and intracellular free calcium ([Ca++free]) (O’Connor and Silver, 2007). These methods require sensitive detection technologies and an appreciation for how the probes may misrepresent or interfere with the processes they are intended to measure. Probes of [Ca++free] provide a case in point. A fluorescent probe of [Ca++free] must be delivered into the cytosolic space without significantly damaging the plasma membrane or labeling other compartments. [Ca++free] is normally less than 0.1 μM in unstimulated cells, and fluorescent probe concentration and affinity for [Ca++free] must be such that they report physiologically relevant increases in calcium without buffering calcium. Calcium indicators must be sufficiently bright in both the bound and unbound states that they can be detected at low intracellular probe concentrations. Weakly fluorescent probes (such as the early generation probe quin2) must be measured at such high concentrations that they buffer intracellular calcium, dampen signals and cause toxicity. On the other hand, small, soluble, calcium indicators with high affinity for calcium can also misrepresent subcellular localization of [Ca++free] changes, as they may diffuse some distance from their sites of calcium-binding before losing bound calcium and changing their fluorescence (Demuro and Parker, 2006). The present generation of fluorescent calcium indicators is highly developed in these regards, and there is now a variety of commercially available, robust and sensitive probes of [Ca++free] for different applications (Simpson, 2006).
The environment of the probe can alter its responses to analytes. The affinities of the probes for calcium decrease below pH 6.8. Thus, probes that reliably report [Ca++free] in cytoplasm (pH 6.8-7.0) have lower affinities for calcium in acidic vacuolar compartments. Measurements of calcium in lysosomes and acidic organelles could be obtained by coordinated measurements of pH and calcium (Christensen et al., 2002). This method allowed measurements of vacuolar pH and calcium concentrations during escape of Listeria monocytogenes from vacuoles in macrophages. Those studies indicated that LLO in Listeria-containing vacuoles increases pH and depletes calcium shortly after bacterial entry (Shaughnessy et al., 2006). These alterations of transmembrane ion gradients may slow vacuole fusion with lysosomes.
Salmonella responds to the environment inside the SCV of host cells by changes in gene expression. The two-component regulatory system PhoP-PhoQ responds in vitro to low pH and low Mg++ concentrations. Early studies using fluorescein-labeled dextran to measure SCV pH showed that Salmonella delay SCV acidification, relative to rates seen in pinosomes and phagosomes, but require acidification for activation of PhoP-activated genes (Alpuche-Aranda et al., 1992). Later studies demonstrated that decreases in Mg++ regulated the PhoP-PhoQ system in vitro (Garcia Vescovi et al., 1996), and indicated that this was the principal regulatory ion activating PhoP in vivo. However, Martin-Orozco et al. (Martin-Orozco et al., 2006) used microscopy of novel fluorescent reporters of Mg++ called PEBBLES to measure conditions in the SCV and found that Mg++ concentrations were not in a range that would affect the PhoP-PhoQ regulon. These findings highlight the importance of measuring analyte concentrations in the context of the living cell.
For microscopy to explain the molecular mechanisms of host-pathogen interactions, it should determine the localization, activities and interactions of all of the relevant molecules. Accordingly, an ‘ideal microscope’ would record this information for the proteins of the host cell and the pathogen at nanometer spatial resolution and microsecond temporal resolution. Furthermore, this measurement should not perturb the biological system. If this were possible, the scientist’s job would simply be to record and interpret images. In reality, current microscopes can image no more than about five FP-tagged molecules simultaneously in a live cell with millisecond and ~200-500 nm resolution. The researcher is therefore left to study a small sample of intracellular biochemistry while balancing efforts to maximize sensitivity, resolution and to minimize perturbation of the biological system.
We expect images acquired from a microscope to map the relative locations and concentrations of molecules inside the cell. However, fluorescence images carry distortions that limit resolution and alter the correspondence between probe concentration and image intensity. These distortions arise from the fact that the structures being imaged are of the same size as the wavelength of light used to image them. The result is a blurring of boundaries of small objects which limits the microscope’s ability to resolve them. Additionally, this blurring has the subtle effect of averaging fluorescence between adjacent volume elements. This problem is so pervasive in microscopy that essentially all measurements of subcellular biochemistry are influenced by blurring.
Most labs have access to three imaging modalities: widefield (standard fluorescence microscope), confocal and total internal reflection fluorescence (TIRF) microscopes. Each of these microscopes imparts different kinds of distortion on the data, which can be characterized by the point spread function (PSF). The PSF is a three-dimensional image of the fluorescence emanating from an infinitely small point. A PSF can be used to describe the performance of the imaging system and to reverse the blurring distortion using computational methods such as iterative deconvolution (Holmes, 2006; Verveer and Jovin, 1997; Agard and Sedat, 1983). For more detailed descriptions about how each microscope works, the reader is referred to The Handbook of Confocal Microscopy (Pawley, 2006). For the purposes of this review, we simply compare the potential of these instruments for imaging host-pathogen interactions at the cellular level.
Widefield (conventional fluorescence) microscopes create a uniform field of excitation light. Fluorescence emitted from the sample is then imaged directly onto a camera. The resulting image is relatively sharp in the x-y plane (the image one normally sees when looking through the eyepiece) but is significantly blurred along the z-axis. This distortion can be seen in the PSF (Fig. 1A) which has a finite diameter of about 200 nm in the in focus x-y plane but extends infinitely along the z-axis. The impact of this PSF on the data can be seen in the blurred image in Fig. 1A.
Confocal fluorescence microscopes illuminate the sample with light focused to a point through a pinhole (or an array of pinholes). A conjugate pinhole in the imaging plane (or an array of pinholes) rejects out-of-focus fluorescence. The resulting PSF is smaller than that from the widefield microscope (Fig. 1B). Although it is still elongated along the z-axis, it has a finite boundary (unlike the PSF for the widefield microscope). This property makes the confocal microscope particularly well suited for imaging in thick specimens.
TIRF microscopy takes advantage of the refractive index difference between coverslips and cell culture media. TIRF illumination creates a shallow illumination field that decays exponentially away from the coverslip-medium interface. This property allows selective imaging of molecules which are within ~100-200 nm of a coverslip. By excluding fluorophores in other focal planes, TIRF allows fluorescence imaging with very low background fluorescence. This feature makes TIRF microscopy the predominant mode for imaging single fluorescent molecules attached to glass surfaces or inside adherent cells (Cai et al., 2007; Ha et al., 1996). The PSF for the TIRF microscope is similar to the widefield microscope; however, the high numerical apertures used for TIRF objectives (NA = 1.49) and the close proximity of fluorescence to the coverslip make this PSF somewhat smaller than the widefield and confocal PSFs encountered in lower numerical aperture water immersion objectives (NA = 1.2) used for live cell imaging.
Each of these microscopes has strengths and weaknesses for live cell imaging that depend on their speed, resolution and sensitivity. The widefield microscope has the lowest three-dimensional resolution, but is relatively inexpensive. Additionally, widefield microscopes equipped with filter wheels for selection of excitation and emission wavelengths have proven to be sufficiently fast for ratiometric and FRET based analysis in live cells. For three-dimensional imaging, of multiple colors and FRET analysis, more sophisticated configurations are needed to acquire data at rates faster than cellular movements (e.g. complete acquisition of the three-dimensional data set in approximately 1 second). Recently, we devised a widefield microscope that used multiple cameras, fast hardware coordination and image reconstruction to achieve sustained multicolor and FRET of living cells in three dimensions (Hoppe et al., 2008). The confocal microscope has improved resolution over widefield microscopes, but image acquisition can be slow, and much fluorescence is discarded in making an image. This can lead to considerable photobleaching of specimens. The speed limitations have been improved by the development of spinning disc and array-scan microscopes (Pawley, 2006). Additionally, three dimensional imaging with a confocal requires acquisition of more focal sections than a widefield microscope. Lastly, TIRF microscopy offers the best z-axis resolution and exquisite sensitivity, but is limited to imaging fluorophores near coverslips. For example, the bacterium in Fig. 1 does not contact the bottom membrane of the cell and is therefore not observed in the TIRF image. In all cases, the speed and duration of imaging that can be performed in live cells is limited by the stability of the fluorophores and the toxicity generated by the excitation light and the fluorophores themselves.
While imaging an individual FP-chimera with these microscopes is instructive, simultaneous analysis of multiple FP-chimeras tell a deeper story. Three questions can be addressed by using these microscopes to image multiple fluorophores during a cellular process: 1) how similar are the distributions of two probes, 2) how do their localizations change in response to interactions with microbes and 3) how do the molecular interactions between two labeled proteins change? These questions are addressed by colocalization and image correlation microscopies, ratiometric fluorescence microscopy and FRET microscopy.
Colocalization, or the comparison of the distributions of two fluorescent probes, is particularly useful for defining the intracellular compartment in which a particular protein is located. Colocalization studies are routinely used to analyze the trafficking of intracellular microbes, as in the examples described above for Mycobacteria, Listeria and Salmonella. The typical approach for colocalization is to collect one image for each probe and then to compare their patterns by means of overlaying them. Typically, the image of one probe is colored green and the image of the other probe is colored red. When these images are overlayed, subcellular regions occupied by both probes appear yellow. For probes that show nearly identical subcellular distributions, their perfect colocalization can be strong evidence that they belong to the same subcellular compartment. For such measurements to be valid, the microscope must cleanly distinguish the two probes since spectral crossover between probe channels can lead to erroneous colocalization.
Ambiguity also arises in this approach when the probes show only partial colocalization. For example, if one probe is distributed throughout the cell (e.g., endoplasmic reticulum), then occasional overlapping fluorescence from the other probe cannot be taken to indicate colocalization. All too often, the interpretation becomes subjective and the researcher must consider the likelihood that the low frequency of colocalization is caused by a biological mechanism rather than chance. To answer this question wither greater accuracy, new approaches are being developed for quantitative colocalization, including image cross-correlation methods and the Manders coefficients (Oheim, 2007; Comeau et al., 2006). Effectively, these methods quantify how often and in how many pixels two different probes are present. Such methods are particularly useful for comparing changes in colocalization (say before and after treatment with a drug) of a particular probe-pair, but are generally underused.
While colocalization indicates similar distributions to subcellular structures, it cannot be used to demonstrate direct molecular interactions. This limitation arises from the fact that the microscope’s resolution limit of ~200 nm is much larger than the dimensions of molecular interactions (Fig. 2). Hence, probes can accumulate independently on a subcellular structure and report fluorescence from the same unrresolvable point despite significant spatial separation. Additionally, strong accumulation of probes to different structures, but weak colocalization may mask subtle, but important interactions. The development of superresolution methods such as photo-activated localization microscopy (PALM) (Betzig et al., 2006) and stochastic optical reconstruction microscopy (STORM) (Rust et al., 2006) can push the resolution of the optical microscope down to ~10 nm. As these methods approach molecular resolution, colocalization signals become indicative of molecular interactions.
A more quantitative variation on colocalization is ratiometric fluorescence microscopy. Here, the localization dynamics of a labeled fluorescent protein are recorded relative to another fluorescent molecule that marks cytoplasmic volume or cell membranes. For example, in an image in which the FP-chimera is localized to the cytosol, a ratio with a volume-marking fluorophore creates a uniform image. However, if the FP-chimera localizes to a sub-cellular region, the ratio of FP-chimera/FP-volume in that region will increase. This technique distinguishes changes in the distribution of FP-chimeras from changes in cell shape and the movement of probe-displacing organelles. A simulation illustrates the utility of ratiometric imaging for discerning small changes in probe localization (Fig. 1). The widefield microscope is capable of detecting weak signals coming from virions inside a cell and the confocal microscope and TIRF microscope are better at detecting these localized signals due to their improved 3D sectioning capabilities. In addition to blurring, detection noise also affects the quality of the ratio image. For live cell imaging in the confocal microscope, the improved z-axis resolution is somewhat offset by increased noise relative to the widefield microscope.
FRET is an established and powerful method to quantify formation and dissociation of protein complexes (Sekar and Periasamy, 2003), the conformational states of individual doubly tagged proteins and the densities of membrane proteins in intact cells. FRET detects interactions between proteins of interest conjugated to appropriate FP by the transfer of energy from an excited donor FP to an acceptor FP in close proximity (≤ 10 nm). This technique provides the spatial and temporal localization of protein interactions with nanometer-scale resolution. Another unique benefit of FP-based FRET techniques is that formation and dissociation of protein pairs can be monitored in real time, as these interactions bring the FPs within the Förster distance (~5 nm). This property distinguishes FRET from techniques that detect protein interactions based on complementation of FPs or light-emitting enzymes (Luker et al., 2004; Hu and Kerppola, 2003), since these methods require significant lag times (seconds-hours) for detection and, in the case of FPs, leads to the irreversible formation of fluorescent complexes when the two halves of the FP meld together. Currently, FRET is the only technique capable of probing dynamic interactions and conformational changes with spatial resolution limited by the optics of the microscope.
Numerous methods have been developed for FRET-based analysis of molecular interactions and conformational states (Jares-Erijman and Jovin, 2003). In its simplest form, FRET can be used to detect changes in conformational states of protein domains, when donor and acceptor FPs are attached to either end of these domains. In this case, a simple ratio of acceptor fluorescence to donor fluorescence, during preferential illumination of the donor, is sufficient to detect changes in the FRET efficiency (i.e., the fraction of energy transferred from the donor to the acceptor) as the changes in conformation change the distance between donor and acceptor. When examining the binding of two separate FP-labeled proteins by FRET, more sophisticated methods are needed. Here, the goal can be to obtain qualitative measures that simply detect the presence or absence of an interaction by detecting the FRET-induced acceptor fluorescence (a.k.a. sensitized emission) from a background of overlapping non-FRET fluorescence from the donor and acceptor.
FRET microscopic studies of Rac signaling analyzed the effects of the Yersinia pseudotuberculosis invasion proteins YopE and YopT on Rac1 activation dynamics (Wong and Isberg, 2005). YopE and YopT are delivered into the cytoplasm of host cells via a T3SS, where they interfere with signaling via the Rho-family GTPases Rac1 and RhoA (Viboud and Bliska, 2005). To examine the effects of YopE and YopT on the localization and activation of Rac1, Cos1 cells were transfected with plasmids encoding CFP-labeled Rac1 and YFP-labeled p21-binding domain (PBD) of Pak1 which binds to active, GTP-bound Rac1 but not to inactive, GDP-bound Rac1. In its active GTP-bound state, Rac1 forms a bimolecular complex with PBD. The proximity of FPs within the YFP-Rac1/CFP-PBD complex allowed FRET microscopic detection of YFP-Rac1 activation. The YFP-Rac1 fluorescence indicated the distributions of GTPase and the calculated FRET signals indicated the distributions of the active YFP-Rac1. Using mutant strains of bacteria expressing YopE, YopT or both, Wong and Isberg (Wong and Isberg, 2005) determined that the two proteins exerted distinct and counteracting effects on Rac1 activation and distributions. YopE inactivated Rac1 by acting as a GTPase-activating protein (GAP) on plasma membrane-associated Rac1. YopT protease activity released Rac1 from plasma membranes, which led to accumulation of active YFP-Rac1 in the nucleus. When the two Yops were overexpressed in the same cell, the YopT activity led to accumulation of active YFP-Rac1 in the nucleus and the YopE activity inactivated those cytoplasmic YFP-Rac1 molecules which could still associate with the plasma membrane. These distinct effects on Rac1 localization and activation indicated that YopE and YopT can exert complex effects on the spatial organization of Rac1 activity in target cells.
While the methods used in these studies provide relatively simple detection of protein interactions, the measurements do not return the concentrations of interacting FPs or the FRET efficiency, and they cannot be compared between instruments, which precludes full characterization of intracellular biochemical activities. Additionally, all FRET measurements between two ectopically expressed proteins are influenced by the relative expression levels of donor and acceptor FP chimeras. These limitations can be overcome by other quantitative FRET methods. Quantitative FRET microscopy seeks to estimate the concentrations of interacting proteins and the distance between FPs, which is described by the FRET efficiency. While currently no methods exist that can quantify all of these aspects in a single, rapid measurement, a number of techniques have been developed to quantify the relative concentrations and apparent FRET efficiencies of the interacting proteins from calibrating the FRET-induced changes in the fluorescence spectrum (Hoppe et al., 2002; Erickson et al., 2001). Here, the apparent efficiency quantifies the interaction as the fraction of interacting proteins multiplied by the FRET efficiency (Erickson et al.). These techniques do not provide the absolute concentrations of interacting fluorophores, but they do provide a quantitative gauge of the magnitude of an interaction by being proportional to the number of molecules in complex. The use of apparent efficiency is a limitation that arises from the amount of information available in the fluorescence spectrum (Hoppe et al.). Importantly, these apparent efficiency measurements are transferable from one instrument to the next and have been extended to fluorescence polarization (Mattheyses et al.), and acceptor photobleaching (Jares-Erijman and Jovin).
FRET-microscopy-based studies were applied to determine the spatio-temporal activation of Rac1 and phosphatidylinositol 3-kinase (PI3K), upon cellular stimulation by the L. monocytogenes invasion factor InlB. In cells co-transfected with plasmids encoding YFP-Rac1 and CFP-PBD, these experiments determined the ratio of GTP-bound YFP-Rac1 molecules to the total YFP-Rac1 ([YFP-Rac1GTP/CFP-PBD]/[total YFP-Rac1]). FRET microscopy was also used to measure the density of 3′ phosphoinositides in the plasma membrane, such as phosphatidylinositol (3,4,5)-trisphosphate (PIP3). The quantification of FRET between two co-expressed FP chimeras of the Akt pleckstrin homology domain that specifically associate with 3′ phosphoinositides, YFP-AktPH and CFP-AktPH, was used as read-out of PI3K activity. At low concentrations of InlB, PIP3 was not generated on membranes, the FP chimeras were dispersed in the cytosol and did not exhibit FRET. Upon receptor-mediated activation of PI3K and generation of PIP3, the concentration of the fluorescent AktPH chimeras increased at the plasma membrane, consequently leading to an increase in the FRET signal. These methods showed that upon receptor stimulation by InlB, Rac1 was activated downstream from PI3K within 1 min and was down-regulated within 5 min, whereas PIP3 reached its maximum concentration after 3 min and slowly decreased for more than 20 min. The results led to the hypothesis that the spatial distribution of 3′-phosphoinositides within cholesterol-enriched membrane microdomains is critical for the successive activation of Rac1 and consequently for F-actin assembly at bacterial entry site (Seveau et al., 2007).
In the analysis of cellular signaling pathways, the ultimate goal of FRET microscopy is to convert fluorescence signals from FP chimeras into concentrations of bound and free proteins and the FRET efficiencies between those interacting proteins. Recovering this information would allow FRET microscopy to measure the exact quantities of interacting components in genetic model systems in which the FP chimeras are expressed from endogenous genetic loci. Additionally, precise concentration estimates would allow new insight into how multi-subunit complexes are assembled inside cells by allowing measurement of their stoichiometry during assembly.
Two technical limitations prevent FRET microscopy from reaching its potential for analysis of subcellular pathways. First, with two exceptions (Galperin et al., 2004; He et al., 2003), FRET microscopy has only been applied to the analysis of interactions between two labeled proteins. This limitation prevents placement of specific interactions into the temporal and spatial context of other biochemical activities, thereby limiting the potential of FRET for constructing models of pathway organization. Multispectral microscopy platforms have the ability to unmix overlapping fluorescence signals and thus to allow observation of multiple fluorescently tagged molecules (Nadrigny et al., 2006; Neher and Neher, 2004). However, these methods are not currently capable of measuring FRET between multiple fluorophores and therefore are limited to measuring the co-localization of fluorescently tagged molecules. Second, FRET microscopy measurements are significantly impaired by out-of-focus light and detection noise (Hoppe et al., 2008). These limitations lead to reduced sensitivity and accuracy in FRET measurements and limit live cell 3D-FRET microscopy. Until these limitations are overcome, the potential of FRET microscopy for measuring the timing and spatial organization of biochemical reactions inside cells will not be realized.
New developments in microscopes and image processing techniques are extending the capabilities of live cell fluorescence microscopy. Recently, methods were developed for fast imaging of molecular interactions throughout the 3D-space of the living cell (Hoppe et al., 2008). This work included the development of new instrumentation and new algorithms for the deblurring of 3D-FRET microscopy data to achieve sustained imaging of the activation of Cdc42 during FcR-mediated phagocytosis of beads. Additionally, new advances in superresolution microscopy hold the promise of improving protein localization and FRET measurements. While many of these superresolution techniques require extravagant instrumentation, some, such as PALM, STORM and structured illumination microscopy (SIM)(Schermelleh et al., 2008) have the potential to see immediate application to host-pathogen problems. PALM and STORM are exciting developments because they allow imaging of photoactivatable proteins in thin sections of fixed cells with 2D-resolution similar to the scale of single protein interactions (2-25nm range) (Betzig et al., 2006). PALM has also been applied to live cell microscopy with TIRF illumination, but in this configuration it is limited to the analysis of slow moving membrane-associated structures and has reduced resolution over fixed-cell PALM (~60 nm) (Shroff et al., 2008). Recently developed 3D-SIM methods provide a new level of resolution for imaging of the 3D architecture of cells. By patterning excitation light in 3D and then collecting images for various orientations of these patterns, 3D-SIM improves the 3D resolution of the optical microscope to ~100nm × 100nm × 300 nm, a 2-fold resolution improvement over the confocal microscope (Gustafsson et al., 2008; Schermelleh et al., 2008). In addition to improving the resolution of the microscope, this technique holds the possibilities of being performed on live cells and providing quantitative analysis of colocalization, ratiometric imaging and FRET microscopy. This exciting possibility may lead to new views of the architecture of the dynamic chemistries that control cell function.
New intravital imaging techniques are bridging the gap between high resolution analysis of host-pathogen interactions in cultured cells with their analysis inside living animals. Techniques such as single-plane illumination microscopy (SPIM)(Huisken et al., 2004) and scanned sheet microscopy (Keller et al., 2008), in addition to multiphoton microscopy (Miller this issue) provide new tools for imaging host-pathogen interactions in tissue.
Numerous key effectors that participate in host cell subversion by pathogens have been identified and many more will be identified. A major challenge is to determine how these effectors are assembled and coordinate their activities in time and space during invasive processes. Quantitative imaging methods that follow several kinds of molecule in living cells will facilitate discovery of dynamic and sequential interactions as well as sub-cellular localization. In turn, mechanistic mathematical modelling (Linderman and Kirshner this issue) of these data will provide new levels of description of the molecular pathways that control host-pathogen interactions.
Supported by NIH grants AI-035950 and AI-064668 to J.A.S.