Special Reviews

Visualizing Signals Moving in Cells

See allHide authors and affiliations

Science  04 Apr 2003:
Vol. 300, Issue 5616, pp. 96-100
DOI: 10.1126/science.1082830


Cells display a highly complex spatiotemporal organization, required to exert a wide variety of different functions, for example, detection, processing, and propagation of nerve impulses by neurons; contraction and relaxation by muscle cells; movement by leukocytes; and adsorption and secretion of nutrients and metabolites by epithelial cells lining the gut. Successful execution of these complex processes requires highly dynamic information transfer between different regions and compartments within cells. Through the development of fluorescent sensors for intracellular signaling molecules coupled with improved microscopic imaging techniques, it has now become possible to investigate signal propagation in cells with high spatial and temporal resolution.

Advances in molecular genetics and biochemistry have led to the identification of many new signaling molecules and interactions between them, as documented in the elaborate signaling maps that are currently under development (1). These maps consist of boxes indicating molecules connected by arrows that delineate the possible flow of information (signals) between them to result in specific cellular actions such as gene expression, movement, cell division, etc. These maps, however, do not take into account the spatial and structural aspects of these signaling pathways, which in real cells are very important. Understanding these pathways and mechanisms of signal propagation in cells will require the measurement of many signaling reactions, with high spatial and temporal resolution. Most cells are small and the concentration of signaling molecules is generally low; therefore, these measurements require both considerable magnification and sensitivity. The most widely used detection methods are, therefore, based on fluorescent microscopic imaging techniques.

Microscopy and detection techniques have improved considerably in sensitivity over the last decades, and it is now possible to take fluorescence images in the μs to ms range. Through the use of confocal and deconvolution microscopy, it has also become possible to measure several fluorescence signals simultaneously in the same cell with high three-dimensional spatial and temporal resolution (2, 3). Using total internal reflection microscopy it is now possible to image single fluorescent molecules in living cells (4). Data analysis requires the development of advanced visualization and analytical techniques. Furthermore, because many of the signaling reactions taking place in cells involve complex positive and negative nonlinear feedback as well as transport, their dynamics can give rise to a wide variety of nonintuitive behaviors. To interpret and understand these data it is becoming increasingly necessary to model and analyze them using qualitative and quantitative mathematical models (5–7).

Widely Different Mechanisms for Signal Movements

Cells respond to signals from the outside world. In many cases, these signals are detected by plasma membrane–bound receptors. Activation of a cell surface receptor typically triggers several intracellular signaling pathways, resulting in an information transfer between the membrane and other cellular locations and compartments, which involves the physical movement of signaling molecules through the cell. Examples are small molecule second messengers that may spread by diffusion or can actively propagate with the aid of the local regeneration of the messenger, as in the case of propagation of calcium waves within cells. However, much larger molecules such as proteins or even protein complexes can act as signals by moving among different cellular locations. Examples of the latter are cytosolic proteins that bind at sites at the plasma membrane during the formation of multiprotein signaling complexes or the movement of transcription factors upon activation from the cytosol to the nucleus. The movement of these larger molecules may result from diffusion but can also involve active transport, especially between different cellular compartments. Here we will highlight how the measurement of the spatiotemporal dynamics of these signaling molecules and the response of their targets yields important information on the signaling mechanisms.

Regenerative Signal Propagation

An efficient way to transport information between two points in space is to actively transmit and propagate a signal, which normally requires the local amplification of the signal during its movement through space. The best studied example of active signal propagation is without doubt the propagation of the action potential along an axon. In nonmyelinated axons, these signals reach speeds well in excess of 10 m/s, and much effort has gone toward the development of optical methods to image action potentials in single neurons and networks of neurons with the use of voltage-sensitive dyes. These methods hold great promise for the analysis of information processing in the nervous system, but they fall outside the scope of this article.

Calcium waves. Other signals that have been shown to propagate in cells are calcium waves. Calcium waves typically cover speeds of 0.1 to 10 μm/s, depending on the mechanisms underlying their propagation (8). Propagation calcium waves have been studied extensively in large cells such as eggs and were first experimentally demonstrated to occur upon fertilization of the Medaka egg (9). The calcium transients were imaged by recording the patterns of photons emitted from the luminescent protein aequorin upon binding of calcium ions. The signal initiated at the site of sperm entry and moved rapidly through the egg. Calcium waves have since been shown to be important during the early development of eggs of all invertebrate and vertebrate species studied (10). These calcium waves are initiated by the sperm-mediated increase of inositol 1,4,5-trisphosphate (IP3), which is possibly the result of injection of phospholipase C by the sperm (11). The local increase in IP3 results in a local release of calcium from intracellular stores in the endoplasmic reticulum (ER), which is mediated by IP3 receptors on the ER (11, 12). The local increase in calcium results in calcium-induced calcium release from intracellular stores, which then results in the propagation of the signal as a calcium wave. The increase in calcium, during the passage of the calcium wave, triggers the fusion of the cortical granules with the plasma membrane and the exocytosis of their content, resulting in the formation of the fertilization membrane, an important defense against polyspermy. In addition, the calcium wave restarts the cell cycle of the egg.

Calcium transients such as puffs and spikes have been observed in many cells and may affect functions as diverse as growth, apoptosis, and secretion. The mechanisms underlying these phenomena are complex and diverse and have been reviewed recently (12, 13). The mechanisms of calcium spike and wave formation and propagation have been the subject of extensive mathematical modeling, and it has been possible to explain and understand many of the observed types of signal dynamics in considerable detail (14).

There have been many technological advances in the development of calcium-specific sensors. Originally, the sensors were based on fluorescent organic molecules that changed their spectral properties upon binding of calcium; however, more recently calcium sensors have been developed that are based on conformational changes of calcium binding proteins, which are measured by fluorescence resonance energy transfer (FRET) (15). The advantage of these indicators is that they can be expressed in specific compartments in cells, which allows the measurement of calcium signaling between these compartments, such as signaling between the cytoplasm and the nucleus. These indicators can also be expressed under the control of cell type–specific promoters, which should allow the measurement of calcium signaling dynamics in individual cells in tissues or even in whole organisms, especially in embryos.

Calcium waves occur in many different excitable and nonexcitable cells, many of which are smaller than eggs. Small cells like polarized neutrophils can exhibit regular calcium spikes at around 20-s intervals (16). The frequency of these spikes is increased upon stimulation with the chemoattractant formyl-methionyl-leucyl-phenylalanine (fMLP). With the use of very fast imaging techniques, it has now been shown that these calcium spikes are not global events but represent very fast propagating waves that were previously not properly resolved (17). The calcium waves start in the leading lamellipod of the moving neutrophil and propagate round the cortex until they come back to the lamellipod. These waves propagate at speeds of 180 μm/s and, thus, propagate around the cell in a fraction of a second, after which the cell is at rest for 10 to 20 s before the next wave is initiated in the lamellipod. The mechanisms underlying these very fast waves are not yet known, but they appear to be ryanodine sensitive, suggesting that they may involve the activation of ryanodine-sensitive Ca2+channels. Fast calcium waves also propagate around forming phagosomes; however, the biological importance of these waves has yet to be established (17) .

NADPH waves. Similar fast imaging techniques have revealed propagating waves of the metabolite NADPH [the reduced form of nicotinamide adenine dinucleotide phosphate (NADP+)] in polarized neutrophils (18). The waves can be detected by monitoring the endogenous fluorescence of NADPH. They originate in the uropod of polarized cells traveling in the direction of the leading lamellipod. On arrival in the lamellipod, they cause the rapid localized release of reactive oxygen species (ROS). It has been suggested that this can result in the fast, localized killing of a target cell at the leading edge. In this case, the propagation of the waves seems to direct the secretion of the ROS at the leading edge. The mechanisms underlying NADPH wave propagation and initiation in the uropod are still unresolved, though models for the oscillations have been proposed (19). These fast imaging experiments on the propagation of NADPH and calcium waves in neutrophils suggest that there may exist another world of fast reaction dynamics, which are only beginning to be explored.

Propagation by Diffusion

Cyclic nucleotides. Many signaling systems use the classical second messengers adenosine 3′,5′-monophosphate (cAMP) and guanosine 3′,5′-monophosphate (cGMP). It is expected that these cyclic nucleotides will distribute rapidly and uniformly through the cytoplasm of cells due to fast diffusional equilibration. However, not many experimental data are available. Cyclic nucleotide levels are regulated through their synthesis, which involves several different cyclases, and their degradation, which is controlled by a family of phosphodiesterases. Many of these enzymes show highly localized cellular distributions. In the last decade, sensitive sensors for these nucleotides have been developed that can be used to image temporal and spatial changes in the concentrations of these second messengers (20–22). These sensors are based on cyclic nucleotide–dependent protein kinases that undergo conformational changes in response to cyclic nucleotide binding that can be measured using FRET. These methods are fairly new and not trivial to use, but local activation of some signaling pathways has been observed to result in localized elevation of cAMP (23). This is an exciting area, given the central importance of these cyclic nucleotide second messengers; these findings will undoubtedly stimulate many further investigations.

Translocation of Signaling Proteins

Translocation from the cytoplasm to the membrane. Cells are often presented with spatially graded information from their environment, which they need to translate into localized action. Examples are the chemotactic movement of a cell up a signal gradient or the localized growth of an axon growth cone in response to a guidance factor. Both cases require the detection of a signal gradient, followed by extension of the cell in the direction of the signal. In the case of chemotactic cell movement up a gradient, extension of the leading edge needs to be coordinated with the retraction of the back of the cell, which requires signaling between the front and the back of the cell (24). The mechanisms underlying polarization and its translation in directed cell movement have been extensively investigated in the amoeba Dictyostelium discoideum, neutrophils, and fibroblasts (25, 26). Detection of a gradient of a chemoattractant results in the rapid localized production of the signaling lipid phosphatidylinositide 3,4,5-trisphosphate (PIP3) at the leading edge of the membrane (Fig. 1). PIP3 acts as a binding site for proteins that contain PIP3 binding domains known as pleckstrin homology (PH) domains such as kinase-like Akt/PKB, or members of the family of guanine nucleotide exchange factors (GEFs) that activate small G proteins (heterotrimeric GTP-binding proteins) of the Ras, Rho, and Arf families. Translocation of these proteins to the membrane results in their activation through interaction with other membrane-bound proteins, often through phosphorylation. The localized activation of members of the Rac and Rho family members is essential for the signal transduction to actin polymerization (27,28).

Figure 1

Membrane localization of the PIP3-specific PH domains of cytosolic regulator of adenylyl cyclase (CRAC) and general receptor for phosphoinositides (GRP1) during chemotaxis (Movie S1). (A) Membrane localization of the PH domain of CRAC to the leading edge of a cell moving chemotactically toward a micropipette releasing a steady signal of cAMP (localized at the top of the cell, outside the field of view). Notice the strong localization of the binding of the GFP-tagged PH domain. (B) Membrane localization of the GFP-tagged PH domain of GRP1 to the leading edge of a cell moving chemotactically toward a micropipette releasing a steady signal of cAMP (top). Notice that GRP1 binding to the membrane appears to spread more posterior than CRAC (Fig. 1A) and is not confined to the leading edge. (C) Accumulation of the GFP-tagged CRAC-PH domain in aggregating cells at their leading edges. The cells move from right to left in response to waves of cAMP traveling through the aggregation stream from left to right. The GFP-tagged CRAC-PH accumulates strongly at the leading edge. (D) Accumulation of the GFP-tagged GRP1-PH domain in aggregating cells at their leading edges. The cells move from right to left in response to waves of cAMP traveling through the aggregation stream from left to right. Although there is still an accumulation of GRP1-PH domain binding at the leading edges of the cells, the GRP1-PH domain is distributed much more evenly than the CRAC-PH (C). This is a result of the much slower release time of the GRP1 domain from PIP3, which results in a considerable diffusion away of the PIP3-GRP1 PH domain complex from the site of PIP3 production at the leading edge of the cell. This shows that one signal (PIP3) can give very spatial and temporal readouts and downstream signaling reactions.

PIP3 production is autocatalytic (29), and theoretical studies have shown that in order to limit this autocatalytic PIP3 synthesis locally, it has to be counteracted by a long-range inhibition, which prevents autocatalytic PIP3 amplification from propagating in the plane of the membrane. Furthermore, to make detection of the gradient both sensitive and robust there must exist a global negative feedback (30, 31). The molecular mechanisms underlying these inhibitions are unknown, but recent data suggest that it may involve a selective dissociation of the PIP3-degrading phosphatase PTEN from the leading edge of the cell (25). PH domains tagged with the green fluorescent protein (GFP) can be used to investigate the involvement of PIP3 signaling in moving cells in multicellular organisms (32). FRET probes have been developed that are able to monitor the activation state of small G proteins of the Ras and Rac family, and Rac is activated as expected in the leading and trailing edges of chemotactically moving cells, sites of active remodeling of the actin cytoskeleton (33,34).

Many PH domain–containing proteins have been cloned and the binding specificities of their PH domain characterized biochemically; these can now be used as selective sensors for different PIPs (35, 36). Several proteins bind selectivity to the same phosphoinositides, but differences in their binding and/or dissociation rates will result in different temporal and spatial readouts of the same signal, stressing the importance of in situ measurements (37). These different PH domains are rapidly becoming valuable probes for the study of the in situ dynamics of many processes that involve phosphoinositidesignaling, such as insulin and growth factor signaling, phagocytosis, and macropinocytosis (38).

Many signaling pathways involve translocation of adaptor proteins to the membrane, and the ability to tag these proteins with GFP presents an excellent measure of the activation of these pathways in vivo. The Wnt signaling pathway is instrumental in setting up the planar polarity in epithelial structures in imaginal discs and eyes inDrosophila, as well as during the polarization and alignment of cells during convergent extension movements of mesoderm cells during gastrulation in vertebrate embryos (39, 40). Activation of the Wnt planar polarity pathway during convergent extension of mesoderm cells results in translocation of one of the downstream components, dishevelled, from the cytoplasm to the membrane and, thus, gives an in vivo readout of the activation of this signaling pathway (41).

Translocation from Cytoplasm to Nucleus

The output of a large class of other signaling pathways is the turning on or off of the transcription of target genes. This process often involves the modification and activation of a transcription factor, which then results in its enrichment in the nucleus. One of the most direct signaling pathways is the Jak/Stat pathway (42). In this pathway, activation of Janus kinase receptors by a cytokine results in binding and phosphorylation of Stat transcription factors on critical tyrosine residues. Stats then dimerize and accumulate in the nucleus, where they regulate a variety of target genes that may control for instance movement (43). Stats are highly conserved and have even been found in simple metazoans such as Dictyostelium discoideum (44). In Dictyostelium, StatA accumulates in the nucleus in response to stimulation of the serpentine cAMP receptor (45). Within minutes of stimulation of an aggregation-competent cell, statA-GFP translocates from the cytosol into the nucleus (Fig. 2). This response can even be studied in the intact organism, where it has been shown to act as a marker for cAMP signaling (46).

Figure 2

StatA accumulation in the nucleus of aDictyostelium cell upon cAMP stimulation (Movie S2). (A) Succesive images of a Dictyostelium cell starved for 4 hours, taken at the indicated times after stimulation with 10–6 M cAMP. Accumulation of the statA-GFP fusion protein in the nucleus is readily recognizable after 200 s and reaches equilibrium after 500 s. (B) Measurement of the ratio of nuclear to cytosolic StatA during the experiment shown in (A). Solid line, moving average calculated over three successive data points.

Other well-known signaling pathways involve the translocation of components of the Wnt signaling pathways APC and β-catenin from the cytosol to the nucleus; interestingly, these proteins are also involved in the formation of junctional contacts and shuttle between different compartments, the nucleus, and the membrane, and are modulated by distinct signaling pathways (47). Shuttling between the cytoskeleton and the nucleus may represent a generally important signaling route, the detailed dynamics and spatial organization of which has yet to be studied in vivo in considerably more detail.

Dynamics of the Cytoskeleton

Many cellular structures are in continuous flux; i.e., they are continually broken down and resynthesized. Examples of this are vesicle transport and breakdown and reformation of the nucleus and its internal structures during the cell cycle, and these events have to be controlled by specific signals. Imaging is playing an important role in elucidating the mechanisms involved. However, a detailed discussion of these systems is not possible in the context of this paper. I will briefly discuss the analysis of cytoskeltal dynamics, because many of the behavioral changes in cells, such as movement, endocytosis, phagocytosis, and cell division, involve local remodeling of the cytoskeleton, which are controlled and integrated by complex signaling networks. Since the development of protocols to tag actin, tubulin, and keratin monomers with GFP, there has been an enormous increase in the information obtained on the role and dynamics of the cytoskeleton during these processes. The actin cytoskeleton can be visualized by transfection of GFP-tagged actin-binding proteins in cells (Fig. 3), and their local synthesis and breakdown can now be followed in great detail (48, 49). By varying the ratio of labeled to unlabeled monomers in filaments of microtubuli and actin microfilaments, it is now possible to visualize them as beads on string. This method, known as speckle microscopy, allows the measurement of filament dynamics with high temporal and spatial resolution, which enables the quantitative analyses of local rates of polymerization and depolymerization during the formation of complex structures, such as extending lamellipods of a moving cell or microtubule spindle formation in a dividing cell. It will become an important tool in the investigation of the signals controlling these polymerization reactions in vivo (50, 51).

Figure 3

Visualization of actin dynamics in cells moving in aDictyostelium slug (Movies S3 and S4).Dictyostelium cells were transfected with the GFP-tagged actin-binding domain of ABP120 (56). Three successive images taken 10 s apart are overlaid in different colors: first image, red; second, green; third, blue. Different-colored dots indicate contact sites with the substrate, which are dynamically made and destroyed. The actin filaments are white, indicating that they do not move with respect to the substrate. Cells moved from left to right of composite image.

Polymerization reactions and the reaction controlling them are highly nonlinear, and, as such, their dynamics allow for the occurrence of polymerization wavefronts that can propagate or percolate through cells, depending on the local density of reactants. Actin polymerization wavefronts occur in Dictyostelium cells, and it will be interesting to see what role they play in the organization of the cytoskeleton (52). The dynamics of the cytoskeleton cannot be understood without taking the activity of the motor proteins into account. The conformational changes in motor proteins, especially myosins, have been analyzed in considerable detail using in vitro systems and imaging methods such as FRET (53). It is to be expected that these measurements will be extended to measurements of motor protein activation in intact cells.

Concluding Remarks and Outlook

The number of systems where the analysis of the spatiotemporal dynamics of biochemical reactions in single cells is increasing rapidly and is expected to expand in a number of directions. A key step will be the development of more reliable and sensitive indicators that can be used to measure signaling at the single cell level. Because many signaling reactions involve proteins that undergo reversible conformational changes and/or formation of complexes with other proteins, it is to be expected that FRET-based measurements will become increasingly important in the analysis of signaling pathways (54). An important requirement will be to show that signaling indicators do not interfere with signaling reactions to be measured. The design of good in vivo FRET indicators will depend heavily on the availability of structural information on conformational changes in proteins as well as on information of complex formation (55). It is to be expected that the number of signaling reactions that can be measured simultaneously in single cells will increase through the use of novel fluorophores and cellular targeting techniques. The increase in the amount of quantitative data on spatiotemporal dynamics of signaling reactions will require the development of new analysis methods, and the assembly of the data in conceptual frameworks will require mathematical modeling (7). It is to be expected that the ability to measure signaling in individual cells, which up to now has mostly taken place in isolated cells in tissue culture, will become increasingly important in the analysis of signaling reactions in tissues and embryos. Such measurement could revolutionize the study and understanding of signaling during development.

Supporting Online Material


Movies S1 to S4


View Abstract

Stay Connected to Science

Navigate This Article