Quantitative imaging using genetically encoded sensors for small molecules in plants


( e-mail sokumoto@vt.edu).


Quantitative imaging in live cells is a powerful method for monitoring the dynamics of biomolecules at an excellent spatio-temporal resolution. Such an approach, initially limited to a small number of substrates for which specific dyes were available, has become possible for a large number of biomolecules due to the development of genetically encoded, protein-based sensors. These sensors, which can be introduced into live cells through a transgenic approach, offer the benefits of quantitative imaging, with an extra advantage of non-invasiveness. In the past decade there has been a drastic expansion in the number of biomolecules for which genetically encoded sensors are available, and the functional properties of existing sensors are being improved at a dramatic pace. A number of technical improvements have now made the application of genetically encoded sensors in plants rather straightforward, and some of the sensors such as calcium indicator proteins have become standard analytical tools in many plant laboratories. The use of a handful of probes has already revealed an amazing specificity of cellular biomolecule dynamics in plants, which leads us to believe that there are many more discoveries to be made using genetically encoded sensors. In this short review, we will summarize the progress made in the past 15 years in the development in genetically encoded sensors, and highlight significant discoveries made in plant biology.


Multicellular organisms consist of specialized cell types that perform distinct functions to sustain the organism as a whole. Accordingly, it is indisputable that each cell type performs unique biochemical reactions, and that analysis of cellular machineries with high spatial and temporal resolutions is indispensable for understanding the functions of a specific cell type.

Techniques to analyze transcriptome regulation at high spatio-temporal resolutions have improved greatly in the past decade. The advent of genetically encoded marker proteins such as GFP, along with genetic tools available in Arabidopsis, for example, has enabled profiling of transcripts in a single cell type (Birnbaum et al., 2003; Brady et al., 2007; Dinneny et al., 2008) and resolving the periodic expression of genes within a single tissue at sub-hour resolution in this model plant (Moreno-Risueno et al., 2010). Likewise, single-cell proteomics studies are starting to reveal cell-type-specific expression of proteins in plants (Wienkoop et al., 2004; Wan et al., 2005; Zhao et al., 2008). Undoubtedly such regulations will result in cell specificities of biochemical pathways, which will be reflected in differential metabolite and ion profiles. Indeed, techniques such as metabolic flux analysis (MFA) have provided a quantitative view of specific metabolism in plants at the tissue level in some cases (Schwender et al., 2006; Junker et al., 2007).

Analysis of small molecules such as metabolites and ions at the cellular level, however, poses distinct challenges, since many of these are regulated at temporal scales (sub-second for signaling molecules such as calcium) that cannot be resolved by conventional extraction techniques. Moreover, metabolism in eukaryotic cells is highly compartmentalized. Some techniques such as single-cell sampling (Roy et al., 2003) and non-aqueous fractionation (Farréet al., 2001) offer higher spatial resolution. However, these techniques are not compatible with real-time analysis.

Optical imaging of biomolecules has revolutionized our view of live cells in the past two decades. The availability of chemical dyes for certain key molecules such as calcium and reactive oxygen species (ROS) has allowed researchers to monitor the spatio-temporal dynamics of these molecules in living cells at an unprecedented resolution: subcellular domains with distinct dynamics can be observed, and the temporal resolution is only limited by the time necessary for the acquisition of optical signals (usually in the range of tens of milliseconds). Optical imaging of such biomolecules has led to the discovery of distinct signatures that control various signaling events such as neurotransmission in animals (Higley and Sabatini, 2008) and cell expansion and stress response in plants (Kader and Lindberg, 2010; Kim et al., 2010; Swanson et al., 2011).

Although it is very powerful, the optical approach to analyze small molecules in live cells used to be possible for only a few small molecules for which chemical dyes were available. However, since the late 1990s, the development of genetically encoded sensors for small molecules has changed the situation completely. Following the first genetically encoded sensors for calcium (Miyawaki et al., 1997; Persechini et al., 1997; Romoser et al., 1997), sensors for a number of important biological molecules such as nucleotides, hormones, sugars, amino acids and ions have been developed (reviewed in Okumoto et al., 2008; and Newman et al., 2011). Genetically encoded sensors offer time resolution that is essentially equivalent to that of the optical dyes. Moreover, genetically encoded probes can be introduced into cells using a transgenic approach, hence eliminating the loading process. The number of available sensors for different biomolecules keeps expanding, offering the possibility of applying optical imaging approaches to an ever increasing number of biomolecules.

The use of genetically encoded probes in plant biology was initially slow compared with other fields, not least because of technical problems associated with the imaging of plant cells. However, continuous efforts from multiple laboratories in the past decade have overcome a number of initial difficulties, and some genetically encoded sensors have become a standard laboratory tool. In this review, we would like to summarize the advancement of optical imaging techniques using genetically encoded sensors for small molecules. Due to space limitations, for other types of genetically encoded sensors (e.g. reporters of enzymatic activities, translocation of proteins, and reporters of protein–protein interaction), readers are referred to a recent excellent review (Newman et al., 2011). Likewise, technical considerations associated with the development of genetically encoded sensors will not be discussed. Rather, the focus will be on specific considerations necessary for the use of optical sensors in plants, and recent discoveries made through the use of genetically encoded sensors in plant biology. A list of currently available genetically encoded sensors for small molecules can be found in Table S1.

High-resolution imaging of cellular molecules using biosensors

The list of genetically encoded sensors has kept expanding in the past decade. As of 2011, sensors for more than 70 different ligands are available (Okumoto et al., Annual Review of Plant Biology, in press), and for some molecules such as calcium ions, multiple sensors with different modes of action and properties are available. Before selecting genetically encoded sensors for one’s experiment, it is important to know the properties of the sensors, their compatibility with the experimental system, and potential pitfalls. Some of the important points to consider are stated below.

Mechanisms of sensor functions

Genetically encoded sensors for small molecules can be roughly divided into two categories: single-fluorescent protein (FP)-based and fluorescence resonance energy transfer (FRET)-based sensors (Figure 1). Single-FP sensors take advantage of the fact that the chromophore of FPs, usually well protected by the β-barrel structure around it, is sensitive to the change in oxidative status and the hydrogen bond network within its environment. Due to this fact, either an FP or FP–peptide chimera that responds to a ligand by a change in the structure of the β-barrel (and hence the microenvironment of the chromophore) can change its fluorescent property drastically, and therefore function as a sensor for the ligand. For example, some FPs are naturally sensitive to pH (Kneen et al., 1998; Llopis et al., 1998; Awaji et al., 2001) and/or halide ions (Jayaraman et al., 2000; Galietta et al., 2001), and have been used as sensors for the change in those environmental parameters (Intrinsic sensor, Figure 1a(I)). In other cases, a protein module that changes its conformation upon binding of a substrate, e.g. CaM-M13 peptide modules in GCaMP calcium sensors (Baird et al., 1999; Nagai et al., 2001; Nakai et al., 2001; Souslova et al., 2007; Tian et al., 2009) or the H2O2-sensitive domain of Escherichia coli OxyR protein in peptide in redox sensors Hyper (Belousov et al., 2006; Markvicheva et al., 2011), was fused to circular-permutated versions of FPs to create single-FP sensors (extrinsic sensor, Figure 1a(ii)). Single-FP sensors can be further divided into intensity-based sensors that respond to the ligand by changing the intensity at its emission maxima, or ratiometric single-FP sensors that change their excitation or emission spectrum according to the ligand concentration. In this case, the ratio between emission intensities observed at two different excitation/emission wavelength combinations is used to deduce the ligand concentration.

Figure 1.

 Types of genetically encoded sensors for small molecules.
(a) Single-fluorescent protein (FP) sensors. (i) Intrinsic sensors consist of variants of FPs that change fluorescent intensity and/or spectra in the presence of specific ions. (ii) Extrinsic sensors consist of FPs (often circularly permutated) fused to a ligand-binding domain (tan), and in some cases, another domain that recognizes the ligand-bound form of the binding domain (blue). The conformational change in presence of the ligand cause changes in fluorescent intensity and/or spectra.
(b) Föster resonance energy transfer (FRET)-based sensors. (i) Unimolecular sensors consist of a FRET pair and a binding module similar to extrinsic single-FP sensor. (ii) In a bimolecular FRET sensor, the donor and acceptor are attached to separate proteins. The distance and/or orientation between the proteins attached to the donor and acceptor changes in the presence of the ligand. Figure modified from Okumoto, 2010.

One of the biggest advantages of single-FP sensors is their ease of use. In the case of sensors that respond to the ligand by changing the fluorescence intensity, data need to be acquired at only one excitation/emission wavelength. Hence, the time resolution of intensity-based single FP sensors is usually better than FRET-based sensors or ratiometric single-FP sensors. Also, the requirement for hardware is minimal, since measurements do not require fast switching between multiple excitation/emission settings. In addition, the dynamic range of a single-FP sensor is often larger compared with FRET sensors. The dynamic range of an intensity-based sensor is expressed as Isat/Iapo: Isat is the signal intensity observed at saturating concentration, Iapo is signal intensity in the absence of substrate. For FRET sensors, I is replaced by R = DonorexAcceptorem/DonorexDonorem (ex and em stand for excitation and emission filters respectively. [DonorexAcceptorem]: fluorescent intensity observed using donor excitation and acceptor emission filters). The signal-noise ratio of an assay is linear to the dynamic range of the sensor, hence a greater dynamic range means improved sensitivity to a smaller change in ligand concentrations. Kon/off also tends to be faster with single-FP sensors due to the smaller molecular size. Better time resolution and faster kinetics of single-FP sensors is a desirable trait for ligands with rapid oscillation, such as for calcium (Tian et al., 2009). The disadvantage of single-FP sensors, on the other hand, is that the readout is more prone to artifacts since changes in FP intensity caused by factors such as pH shift or changes in physical dimension (e.g. thickness of the cells) of the specimen are indistinguishable from the intensity change caused by ligand binding. Therefore, the quantification of absolute concentration is less reliable with intensity-based single-FP sensors. This problem can be partially circumvented by expressing another FP in the same cellular compartment that is not sensitive to the ligand, and using the intensity from the second FP as an internal reference (Awaji et al., 2001). In the case of ratiometric sensors, the intensity obtained with another excitation/emission can serve as an internal reference. However, use of the second wavelength requires hardware settings that are comparable with the FRET imaging.

Usually, FRET sensors consist of a FRET donor–acceptor pair and a ligand-binding domain. The conformational change in the ligand-binding domain is revealed by the change of FRET efficiency between the attached donor and acceptor molecules (Figure 1b(i)). Since FRET is extremely sensitive to both the distance and dipole–dipole orientation between the fluorophores, change in FRET efficiency can sometimes be observed in a configuration in which conformational change in the ligand-binding domain does not seem to induce any distance change between the donor and acceptor (Deuschle et al., 2005; Okumoto et al., 2005; Gu et al., 2006). Due to this fact, a surprisingly large number of ligand-binding domains have been successfully converted into FRET sensors (Okumoto et al., 2008).

We can further divide FRET sensors into unimolecular sensors and multimolecular sensors. In a unimolecular sensor, the donor, acceptor, and the binding module are encoded in one continuous chimeric protein. For this type of sensor, a simple ratio between donor and FRET channels (R = DonorexAcceptorem/DonorexDonorem) is used as a proxy for FRET efficiency, since the molecular ratio between the donor and acceptor can be assumed to be 1:1. In multimolecular sensors, the donor and acceptor are encoded in separate proteins, and the ligand concentration affects the physical proximity between the donor and acceptor (e.g. phosphatidylinositol 4,5-bisphosphate sensor; van der Wal et al., 2001) (Figure 1a(ii)). On one hand, multimolecular configurations increase the change in the donor–acceptor distance between apo-states and bound states due to the physical separation of FRET pairs in the apo-state, therefore the signal-to-noise ratio (ΔR/R0) tends to be larger than for unimolecular sensors. On the other hand, the molecular ratio between the donor and acceptor is unknown in multimolecular sensors. In such a case, FRET efficiency calculations becomes more complicated due to the signals from free donor and acceptor molecules, and require additional techniques such as acceptor photobleaching, which increases the time required for each acquisition. Therefore, unimolecular sensors are typically used for time-course experiments. Recently a dual-expression system with a fixed ratio between two separate proteins using an internal ribosome entry site (IRES) and a viral 2A peptide has been reported (Goedhart et al., 2011). Such a system would facilitate the development of bimolecular sensors in future.

Selection of chromophores for imaging in plant tissue

Any plant cell biologist can testify to the problem of high autofluorescence in plant tissue. Plant tissues are filled with autofluorescent compounds such as lignin, phenolic compounds, callose, and chlorophyll (Chapman et al., 2005), hence it is important to determine whether the FPs used in a biosensor are compatible with the tissue analyzed. For example, GFP fluorescence might not be easily detectable in highly lignified tissues such as xylem, whereas red fluorophores would not work when one intends to image chloroplasts. The best practice would be, therefore, to invest some time in observing the tissue of interest using the exact settings of the microscope that would be used for the imaging. So far, blue fluorescent protein (BFP)–GFP, cyan fluorescent protein (CFP)–yellow fluorescent protein (YFP), GFP–red fluorescent protein (RFP) pairs and their respective variants have been successfully used in plant cells (reviewed in Bhat, 2009).

One of the ways to eliminate problems with autofluorescence is to use bioluminescent proteins (BP) instead of FPs. Although there are certain disadvantages to the use of BPs, such as the need for loading of the substrate, and low signal (hence lower spatial and temporal resolution), BP-based sensors are worth considering when having a lower background is more important than a higher spatio-temporal resolution. For example, BP-based sensors are widely used to identify biologically active compounds in drug screening (Bacart et al., 2008; De, 2011). Analogous to such an approach, aquorin, a bioluminescent calcium sensor protein, has been successfully used to identify a peptide that induces the elevation of cytosolic calcium in Arabidopsis (Haruta et al., 2008).

Consideration of post-transcriptional gene silencing

Another problem that plant biologists continuously encounter is a weak signal due to gene silencing. Initial efforts from multiple laboratories to express CFP–YFP FRET sensors in plant tissue resulted in non-fluorescent or weakly fluorescent plants. A notable exception to this phenomenon was found in the guard cells, where an earlier version of the calcium sensor Yellow Cameleon was successfully utilized merely a year after the development of the sensor (Allen et al., 1999).

The fact that guard cells were immune to the suppression of fluorescence suggests that post-transcriptional gene silencing (PTGS) might be playing a role, since simplasmically isolated guard cells do not receive gene silencing signals from their neighbors through plasmodesmata (Wille and Lucas, 1984; Himber et al., 2003). This hypothesis was further supported by experiments from the Frommer lab, where FRET glucose sensors based on the CFP–YFP pair are expressed at much higher levels in Arabidopsis rdr6 and sgs3 mutants defective in PTGS (Deuschle et al., 2006). While one has to keep in mind that working in such a genetic background might have an intrinsic effect on the biological pathway of interest, PTGS mutants offer an attractive method to circumvent the problem with weak signals. The mechanism of why FRET sensors induce such strong PTGS in plants is still not very clear, because it was shown that a tandem repeat of a similar sequence on its own is not sufficient to induce PTGS (Lechtenberg et al., 2003; Schubert et al., 2004). Rather, it is hypothesized that the ‘dose’ of transgenes is related to the probability of occurrence of PTGS. Those FRET pairs that derive from the same protein, such as CFP and YFP, would essentially contribute two doses per construct, hence it might induce stronger PTGS. To circumvent this problem, one of the fluorophores can be replaced by a FP derived from another organism (e.g. mTFP1 from Clavularia coral; Ai et al., 2006). Also, the CaMV 35S promoter, often used for overexpression in plants, has been associated with gene silencing and trans-inactivation of genes (Daxinger et al., 2008). The use of the UBQ10 promoter instead of CaMV 35S was reported to be effective in alleviating problems with gene silencing (Krebs et al., 2011). Also, placing two terminator sequences after the coding region might reduce PTGS by reducing inappropriately unpolyadenylated mRNA (Luo and Chen, 2007).

Biosensor imaging in plants – what did we learn?

Many of the sensors created for mammalian cells are not extensively used in plants. This is partly because of the differences in the biochemical machineries in plants and animals: some cellular processes for which genetically encoded sensors are extensively used in mammalian cells do not operate the same way in plants (e.g. protein relocalization with the production of specific phosphatidylinositol species and extensive G-protein signaling). However, there are a few exceptions where biosensors provided new insights into as yet undiscovered cellular processes in plant cells. The most notable of such discoveries are summarized in this section.

Metabolite imaging

Metabolomics is a rapidly growing area in biology. Studies so far have revealed robust yet dynamic regulation of central metabolic pathways. In addition, an in-depth analysis of transcriptome and network-based modeling revealed that the tissue-specific expression of enzymes, transporters, and regulator proteins can lead to drastically different metabolic fluxes between different organs or cell types (Duarte et al., 2007; Shlomi et al., 2008).

Measuring metabolites at the cellular or subcellular level is a challenging task. Many central metabolites are synthesized, catabolized, and transported through multiple pathways, making it near impossible to capture the concentration under a certain condition without introducing an artifact. Moreover, metabolites are highly compartmentalized, making it challenging to isolate an individual compartment. For these reasons, there are few studies that report metabolite concentrations at a subcellular resolution.

In the past decade, a large array of genetically encoded sensors for metabolites has been developed. Fluorescent indicator protein (FLIP) family sensors, pioneered by the Frommer lab, utilized periplasmic binding proteins (PBPs) from Gram-negative bacteria. The PBP family contains members recognizing a wide variety of substrates including central metabolites such as sugars and amino acids, and various ions such as phosphate and sulfate (Fukami-Kobayashi et al., 1999). The PBPs undergo a large conformational change when bound to their respective ligands, making lucrative scaffolds for FRET-based sensors. Various metabolite sensors for various substrates including maltose (Fehr et al., 2002), glucose (Fehr et al., 2003; Deuschle et al., 2006; Takanaga et al., 2008), ribose (Lager et al., 2003), sucrose (Lager et al., 2006), arabinose (Kaper et al., 2008), glutamate (Deuschle et al., 2005; Okumoto et al., 2005; Hires et al., 2008), arginine (Bogner and Ludewig, 2007), glutamine (Yang et al., 2010), histidine, and leucine/isoleucine/valine (Okada et al., 2009) have been developed using PBPs as the ligand-recognition domains.

One of the notable insights delivered by the use of metabolite sensors is the amazing heterogeneity of the cellular concentration of metabolites. For example, using FRET glucose and sucrose sensors, it was shown that the cytosolic levels of these sugars are subject to large fluctuations, following the extracellular concentration of these substrates very closely (Deuschle et al., 2006; Chaudhuri et al., 2008). Although it is not clear whether plant cells experience such high concentrations of extracellular sugar under physiological conditions, it implies that plant cells are capable of taking up a large amount of sugars. Moreover, the cytosolic concentrations of sugars under a steady-state condition could be estimated from responses of sensors with different affinities (Figure 2). Using sensors with four different affinities, it was shown that steady-state concentrations of sugars were largely different between tissues: sugars in the root drops below 90 nm in the absence of an external glucose supply, while in leaf mesophyll and guard cells the concentrations remained in a much higher range (>340 nm, estimated by the lack of response using a sensor with high affinity). Presumably such a difference reflects the differences in metabolism between the two tissue types. Therefore, tissue-specific measurement of steady-state metabolite concentrations would be valuable tools for validating tissue-specific metabolite network models (Shlomi et al., 2008).

Figure 2.

 A schematic representation of two model cases where the steady-state levels of a substrate are altered, and the response of sensors in each scenario.
(a) The changes in cytosolic concentration of a substrate. Two cell types with different steady-state levels of the substrate are represented as black and blue traces. The box above the trace indicates the time period when the substrate was externally supplied. Three shaded areas represent the working ranges of sensors with different affinities. Note that the sensors with higher affinity have the smaller absolute working range.
(b) The response of cytosolic sensors at higher (upper panel) and lower (lower panel) steady-state substrate concentration. Figure modified from Okumoto 2010.

Another notable discovery was that of transport activities in plant roots, namely Arabidopsis, which cannot be explained by previously characterized transporter systems. For example, the uptake of glucose in Arabidopsis roots, analyzed by expressing FRET glucose sensors in the cytosol, was sensitive to neither extracellular pH nor protonophores such as carbonyl cyanide 3-chlorophenylhydrazone (CCCP) or 2,4-dinitrophenol (2,4-DNP) (Chaudhuri et al., 2008). This result suggested that there are transporter systems that have not been previously characterized, since all hexose transporters discovered at the time were proton symporters (Büttner, 2007). Indeed, with targeted screening of Arabidopsis transporters using FRET glucose sensors it was discovered that a family of transporters, named SWEETs, are capable of bidirectional transport of glucose that is not dependent on a proton gradient (Chen et al., 2010), and plants carrying T-DNA insertions in both AtSWEET11 and -12 genes were found to be defective in sugar export from phloem parenchyma cells (Chen et al., 2011). Likewise, proton-gradient-insensitive transport of glutamine was also reported using a FRET glutamine sensor (Yang et al., 2010). Earlier studies using pea seed-coat suggest the existence of a transport system that is capable of bidirectional transport for amino acids (deJong et al., 1997; van Dongen et al., 2001). It is possible that, in analogy to sucrose, such a facilitative system exists for amino acids. FRET sensors might provide a platform for the discovery of new transporter systems for other important metabolites in the future.

Imaging of ROS and redox potential in plants

Reactive oxygen species in plant cells constitute an important component in the transduction of a number of signals such as biotic and abiotic stress (Moller et al., 2007), stomatal opening (Kim et al., 2010), cell elongation (Swanson et al., 2011), and sugar sensing (Bolouri-Moghaddam et al., 2010). Traditionally, various dyes such as dihydrodichlorofluorescein diacetate (H2DCF-DA, for H2O2) and nitroblue tetrazolium (NBT, for the superoxide radical) have been used to visualize different species of ROS in plant cells (reviewed in Swanson et al., 2011). However many of these chemicals need to be loaded into the cells, making the measurement difficult, and/or have some problems with photo-oxidation. Therefore, genetically encoded sensors that eliminate the loading process have gained some support amongst plant scientists in recent years.

Hyper, consisting of a bacterial peroxide sensor (OxyR) and cpYFP is a H2O2-specific genetically encoded sensor (Belousov et al., 2006; Markvicheva et al., 2011). The two redox-active cysteine residues in Hyper are located in a hydrophobic pocket of the OxyR domain, hence they are accessible only to amphiphilic molecules such as H2O2 (i.e. unlike roGFPs described below, Hyper does not respond to Oxidized form of glutathione (GSSG) in vitro; Belousov et al., 2006). Hyper was successfully used to monitor H2O2 levels in the cytosol and peroxisome of tobacco and Arabidopsis. Interestingly, it was revealed that peroxisomal Ca2+ concentrations followed cytosolic concentrations closely, and that H2O2 catabolism in the peroxisome is enhanced by Ca2+, probably due to the Ca2+-dependent activation of peroxisomal catalase, CAT3 (Costa et al., 2010).

RoGFP and its variants (Østergaard et al., 2001; Dooley et al., 2004; Cannon and Remington, 2006; Lohman and Remington, 2008), genetically encoded sensors for cellular redox potential, contain a disulfide bond between two surface-exposed cysteines that respond to thiol/disulfide equilibria in the local environment. (Jiang et al., 2006; Meyer et al., 2007; Brach et al., 2009; Schwarzlander et al., 2009). Interestingly, the reaction of roGFPs to H2O2in vivo is much faster (within a few minutes) compared to the situation in vitro (∼30 min) in both animal cells (Dooley et al., 2004; Hanson et al., 2004) and plants (Jiang et al., 2006; Meyer et al., 2007), suggesting that cellular enzymes such as glutaredoxin and thioredoxin mediate thiol-disulfide exchange reactions between GSSG and roGFPs. Indeed, it was proven that in both yeast and plant cells, roGFPs readily equilibrate with glutathione redox buffer through the action of glutaredoxin (Østergaard et al., 2001; Meyer et al., 2007). Hence, roGFPs provide a powerful method to investigate the processes that influence cellular redox balance, in particular the ratio between reduced (GSH) and oxidized (GSSG) glutathione (Schafer and Buettner, 2001). In addition to its use in determining redox potential in subcellular organelles, Brach et al. reported the use of roGFP to predict the topology of a membrane protein in the secretory pathway, taking advantage of different redox potentials across the endoplasmic reticulum (ER) membrane (Brach et al., 2009).

Both roGFP and Hyper are based on circular permutated FPs, and are sensitive to pH change. (Belousov et al., 2006; Jiang et al., 2006). Therefore, it is important that the local pH is monitored separately using a pH-sensitive dye or a genetically encoded pH indicator. The methodology for monitoring pH using these optical probes is covered in an excellent review by Swanson et al. (2011).

Sensors for signaling molecules in plants

The development of genetically encoded sensors has been tremendously useful in spatio-temporal analysis of signaling molecules. In particular, the availability of probes for multiple interconnected signaling molecules such as ROS and calcium is starting to make it possible to analyze the spatio-temporal relationship between them. For example, cytosolic calcium is known to be necessary for the production of ROS in plants in response to pathogens (Grant et al., 2000), and the response of guard cells to ROS (McAinsh et al., 1996). In a recent studies using the calcium sensor YC3.6, the ROS-sensitive dye Oxyburst, and the pH-sensitive probes GFP-H148D and fluorescein, Monshausen et al. revealed that the levels of all of these molecules oscillate during the growth of Arabidopsis root hair cells. Curiously, growth oscillations of root hair cells were temporally correlated to the oscillation of cytosolic calcium, extracellular ROS and pH; the peaks in growth were followed by calcium (∼5 sec delay), then by ROS and pH peaks (∼8 sec delay) (Monshausen et al., 2007, 2008). These results suggested that the cytosolic calcium peaks, potentially caused by the calcium influx through mechanosensitive calcium channels on the plasma membrane, trigger bursts of extracellular ROS and pH and rigidified the cell wall temporarily. In another experiment, the same group demonstrated that the increases in the extracellular ROS and pH are dependent on cytosolic elevation of calcium upon mechanical stimulation of root hair cells, while increases in ROS and pH are independent of each other (Monshausen et al., 2009).

Although these experiments revealed causal relationships between calcium and ROS/pH oscillations in growing root hairs, many questions remain. For example, the results described above suggested that temporal elevations of extracellular ROS and pH, triggered by cytosolic calcium peaks, prevent uncontrolled growth of root hairs. On the other hand, abolishing the tip-focused calcium gradient is known to inhibit the growth of root hairs (Felle and Hepler, 1997; Wymer et al., 1997) and pollen tubes (Pierson et al., 1994), and an artificially generated calcium gradient re-orients the tip growth to the new high point of calcium (Bibikova et al., 1997). Therefore, exactly how cytosolic calcium regulates growth is still in question. Analogous to the situation in the guard cells, in which rapid calcium reactive responses are distinct from long-term calcium-induced inhibition of stomatal opening (Mori et al., 2006; Cho et al., 2009; Kim et al., 2010), two separate modes of calcium signature could co-exist; a low basal level that promotes actin filament reorientation and growth, and high concentrations that prevent growth by rigidifying cell walls. Analyses of the precise signatures of calcium, including absolute concentrations, would help to decode the seemingly multiple roles of this messenger molecule. Also, are internal calcium sources such as the ER involved in propagating calcium peaks throughout the cells? If so, what are their contributions to the cytosolic calcium increase? The lag time between the growth peak and cytosolic calcium peak might be caused by secondary calcium release from internal stores (Holdaway-Clarke et al., 1997). Versions of sensors expressed at the plasma membrane (Krebs et al., 2011) or in the ER might answer such questions.

Future applications

One of the recent exciting trends is that measurements in specific organelles, suggested by researchers for many years now, are indeed becoming possible. The calcium cameleon YC3.6 sensor has been targeted successfully to the cytosol, nucleus, plasma, and vacuolar membranes (Krebs et al., 2011). The results using Arabidopsis root cells expressing YC3.6 attached to the cytosolic face of the plasma membrane showed that different membrane regions in the same cells do react differently to the external stimulant ATP, indicating that there is indeed a local gradient of calcium in a single cell. Likewise, genetically encoded sensors have been targeted to the peroxisome, ER, and mitochondria of Arabidopsis and tobacco plants, and has been proven to function in these organelles (Brach et al., 2009; Schwarzlander et al., 2009; Costa et al., 2010).

The advance in high-throughput imaging technologies, in combination with genetically encoded sensors targeted to specific organelles, can be utilized for genome-wide screens. Using a calcium sensor targeted to mitochondria in of Drosophila S2 cells, and a genome-wide RNA interference (RNAi) library, Jiang et al. (2009) successfully identified a mitochondrial Ca2+/H+ antiporter. Although such genome-wide, untargeted screens using sensors have not yet been reported in plants, genetic tools such as Gateway-compatible cDNA libraries (Lalonde et al., 2010) are expected to promote sensor-guided screens in future.

Dual imaging using multiple genetically encoded sensors will be beneficial to resolve the temporal correlation between two interdependent signaling molecules, similar to the case of calcium and ROS described above. Spectrally orthogonal FRET pairs such as eCFP/eYFP and mOrange/mCherry (Piljic and Schultz, 2008), mTFP/Citrine and mAmetrine/tdTomato (Ai et al., 2008), eCFP/mVenus and TagRFP/mPlum (Grant et al., 2008), and CFP/YFP and Sapphire/RFP (Niino et al., 2009) have recently been reported. In addition, the development of single-FP sensors would alleviate the problem of spectral overlapping and hence make dual-imaging more feasible. Most of the currently available single-FP sensors are based on GFP- and YFP-derived proteins that are not compatible with the most common CFP/YFP FRET pair. Recently, a circularly permutated red-FP has become available (Li et al., 2008), paving the way for red-FP based single-FP sensors (Gautam et al., 2009). Moreover, the Campbell lab recently reported a high-throughput screen of hue-shifted single-FP calcium sensors, which resulted in blue, improved green, and red single-FP versions of calcium sensors (Zhao et al., 2011). The availability of permutated FPs and high-throughput screening methods will greatly increase the efficiency of single-FP sensor development, and make multiplex imaging available for an increased number of biomolecules. Alternatively, the use of a dark FRET acceptor reduces the problem of spectral overlapping (Ganesan et al., 2006). Using a dark FRET acceptor, three-way measurement of cGMP, cAMP and calcium in a single cell has been reported (Niino et al., 2010). Along with techniques such as fluorescence lifetime imaging (FLIM)-FRET, which does not rely on the measurement of acceptor fluorescence, we expect multiplex imaging will become available for a large number of biomolecules in the future.


The work conducted in the Okumoto lab was supported by the National Institute of Health (1R21NS064412) and National Science Foundation (1052048).