Imaging approach for monitoring cellular metabolites and ions using genetically encoded biosensors

Summary

The spatio-temporal patterns of ion and metabolite levels in living cells are important in understanding signal transduction and metabolite flux. Imaging approaches using genetically encoded sensors are ideal for detecting such molecule dynamics, which are hard to capture otherwise. Recent years have seen iterative improvements and evaluations of sensors, which in turn are starting to make applications in more challenging experimental settings possible. In this review, we will introduce recent progress made in the variety and properties of biosensors, and how biosensors are used for measurement of metabolite and ion in live cells. The emerging field of applications, such as parallel imaging of two separate molecules, high-resolution transport studies and high-throughput screening using biosensors, will be discussed.

1. Introduction

The challenge we face in the post genome era is the daunting task of integrating many layers of information (genomic modification, control of transcript and protein levels, post-transcriptional modification, metabolite and ion levels) and understanding how the regulations of these layers ensure the function of the system as a whole. Without a doubt, intricate intra- and intercellular communication is required for the proper function of the higher order units such as tissues and organs. For example, the behavior of a neuronal cell is controlled by the fine balance between excitatory and inhibitory inputs dictated by the network within which the cell is placed and cannot be reproduced in an isolated cell. Therefore, methods to extract information at different levels of regulations from a single cell in its original context are especially relevant in systems biology.

The advancement of cell separation techniques such as Fluorescence Activated Cell Sorting (FACS) and laser dissection, as well as the improvement of amplification and analytical techniques, made it possible to investigate levels of transcripts [1–3] and proteins [4,5] at the cellular level. These studies revealed that even seemingly identical cells could differ in transcriptional and protein profiles, underscoring the importance of high-resolution studies [6,7]. Analyses of metabolites and ions at higher resolution, on the other hand, present a unique challenge. Because these molecules are subject to rapid metabolism and/or transport, accurate determination of concentrations in vivo using lengthy fractionation methods is, in many cases, not appropriate. Rapid sampling and analytical techniques as represented in capillary electrophoretic separation techniques in combination with laser-induced fluorescence (CE-LIF) or mass spectrometric detection (CE-MS) enable detection in very small sample volumes (low nanomolar range for CE-MS) [8]. They are, therefore, promising methodologies for high spatial resolution metabolome analyses. However, while these methods provide an overview of many metabolites, they are not practical for high-resolution time course experiments. Short-lived, temporal modulations of metabolite and ion levels play crucial roles in signal transduction, often involving concerted, sequential modulation of messenger molecules (e.g., neurotransmittor, calcium ion, inositol phosphates, cAMP). Because these transient changes are very short-lived (the typical peak of a neurotransmittor in the synaptic cleft is in the 10 millisecond range) yet physiologically relevant, there is great interest in methods that allow measurements of real-time concentrations in vivo.

Selective chemical dyes for ions and pH that appropriate real-time monitoring of cellular concentrations offer excellent spatial and temporal resolutions, and hence have proven to be revolutionizing tools to study the roles of specific molecules in various cellular events [9]. While an imaging approach to studying the in vivo roles of other cellular molecules with higher spatial and temporal resolution is highly desirable, for the majority of metabolites such specific dyes are not available. A real breakthrough in in vivo-compatible sensors for metabolites came with the availability of Fluorescent Proteins (FPs) derived from organisms in phylum Cnidaria, such as Aequorea jellyfish and corals, and proteins that derive from them [10–17]. FPs have a number of advantageous properties as reporters of cellular events. First, they can be genetically introduced into cells or organisms to function as a fluorescent reporter, offering a large advantage when compared to reporters that need to be externally loaded into the cell. Second, they can be engineered so that a conformational distortion that leads to changes in spectroscopic property is caused under certain conditions, allowing them to report changes in their environment. Finally, it has been proven that two FPs which serve as a Föster Resonance Energy Transfer (FRET) donor and acceptor pair (see below) can function as a reporter of biochemical events in a resolution beyond the limit of optical microscopy. Taking advantage of these properties, it is now possible to use FP-based sensors to observe a number of events in living cells (protein trafficking, ligand-receptor binding, voltage dependent conformational change, protein-protein interaction, enzymatic reactions, and ligand binding to proteins). Here, we review recent advances in metabolite and ion imaging using fluorescence-based sensor proteins. Because of the space limitation, only those types of genetically sensors that detect the concentration of small molecules and ions through fluorescence intensity or spectroscopic properties will be discussed. For other types of sensors that report functions of cellular proteins through protein-protein interactions, protein trafficking and enzymatic activities, and the subtype of metabolite sensors that respond to ligands by changing subcellular localizations, readers are referred to recent excellent reviews [18–22]. New areas of application and perspectives will be discussed.

2. Types of genetically encoded FRET sensors for metabolites and ions

Currently, a wide variety of genetically encoded sensors that recognize metabolites and ions are available. Here, we discuss two subtypes of sensors that report cellular metabolites and ions through their fluorescent spectroscopic properties and intensity. The list of currently available sensors is found in Supplemental Table 1.

2-1 FRET-based fluorescent sensors

FRET is a quantum mechanical effect observed when two chromophores are located in near-field (usually less than 10nm). FRET efficiency is sensitive to both distance and relative dipole-dipole orientation between the donor and acceptor fluorophores, and therefore is an excellent reporter of local protein configuration changes induced by molecular interactions (i.e. protein-ligand binding, Fig. 1). The theoretical details of FRET measurement using fluorescent microscope have been reviewed extensively [20,21] and will not be discussed in this review.

Fig. 1. Various protein modules used for FP-based biosensors.

FRET-based sensors based on different types of ligand-recognition modules. i) Sensors based on a chimeric peptide consisting of ligand recognition domain (green) and a peptide sequence that binds to the ligand-bound form of the recognition domain (blue). ii) Sensors whose binding domain consists of a single protein that binds to the ligand and changes its conformation. B. Single-FP based sensors. i) Variants of FPs that change fluorescent intensity and/or spectra in the presence of specific ions. ii) Circularly permutated FPs fused to an external recognition module similar to the ones in FRET-based sensors.

The first of the FP-based FRET sensors for calcium ions were developed by two laboratories [23,24]. In these prototype sensors, a Calmodulin (CaM)-binding peptide was placed in between FRET donor and acceptor, and subsequent binding of Ca²⁺/CaM to the sensory domain lead to change in FRET efficiency. Their seminal work clearly demonstrated that protein-ligand binding that induces a significant conformational change could be visualized in living cells using FRET, paving the way for a large number of FRET-based sensors.

Typically, an FP-based FRET sensor consists of donor and acceptor fluorophores attached to a naturally occurring ligand recognition module that binds to the substrate of interest, and in many cases, another module that binds to the ligand-bound form of the recognition module. Conformational change caused by ligand binding to this domain induces change in FRET efficiency between the donor and acceptor pair (Fig.1A). Many types of protein modules (e.g. enzymes, membrane receptors, ligand-binding proteins) have successfully been used as substrate recognition modules. Metabolites and ions that function as signaling molecules were major targets for FRET sensor development. Sensors for Zn²⁺ [25–27], glutamate [28,29], cAMP [30–32], cGMP [33–36], phosphoinositides [37], inositol 1,4,5-triphosphate (IP₃) [38,39], diacylglycerol[40], and bacterial quorum-sensing signaling molecules [41] have been developed, and in many cases these sensors were successfully used to achieve sub-second temporal resolution.

Another class of molecules for which many FRET sensors have been developed includes central metabolites such as sugar and amino acids. A family of sensors (FLIP series) pioneered by the Frommer laboratory takes advantage of periplasmic binding proteins (PBPs) from gram-negative bacteria that undergo a Venus-flytrap-like conformational change that is well conserved among this protein family [42,43]. Based on the basic design in which PBPs are sandwiched between FRET donor and acceptor FPs, a number of FRET-based sensors for sugars (maltose [44], ribose [45], glucose/galactose [46], arabinose [47], sucrose [48]), amino acids (glutamate [29]) and ions (phosphate [49]) have been developed. A series of tryptophan sensors that utilizes the dimerization of TrpR, an E.coli Trp operon repressor, has also been developed [50].

Notably, in some configurations of these sensors, conformational change in the binding domain is not likely to lead to distance change between two fluorophores [29,49,51]. Although the exact working mechanisms of this class of FRET sensors are not understood, allosterical hindrance between protein modules created by the conformational change and/or changes in dipole-dipole orientation could explain such phenomena. These cases demonstrate that designs that do not seem to induce large distance change can still function as FRET sensors, making the success rate of this approach surprisingly high thus far.

2-2 Single FP molecule sensors

The chromophore of FP is located in the center of an 11-strand β-barrel structure and usually protected from bulk solvents, making the fluorescence property of FPs stable in a wide range of environments. Conditions that allow solvents to enter the β-barrel and interact with the chromophore, or change the protonation status of the chromphore, however, can change the fluorescence intensity and/or spectroscopic property (Fig. 1B). For example, YFP has several cavities in the vicinity of the chromophore that allow the access of solvent, making the intensity of this fluorophore susceptible to pH and halides [52]. A few mutations in YFP that further increase affinity to halides have also been identified [53,54]. These FPs, either by themselves or as ratiometric sensors in combination with a halide-insensitive FP, can be used as Cl- sensors in vivo [53–56]. Similar single-FP based sensors can be engineered by mutations that allow conditional solvent access to the chromophore. GFP variants that function as sensors for pH [57–59], Hg²⁺ [60], redox status [61,62], and reactive oxygen species (hydrogen peroxide and super oxide) [63], have been developed so far.

Another class of single-FP sensor has a sensory module attached or inserted into the FP sequence. In these cases, ligand-binding-induced conformational changes result in a change in solvent accessibility to the chromophore, shifting the fluorescence intensity and/or excitation and emission spectra [64] (Fig. 1B). The Ca²⁺ binding module, similar to the ones used in FRET-based Ca²⁺ indicators, was successfully used to create single-FP based Ca²⁺ sensors [65–68]. This category includes sensors for ATP:ADP ratio [69] and cGMP [70].

3 Engineering efforts to improve genetically encoded sensors

While the initial success rate in converting a metabolite or ion binding module into a sensor protein was surprisingly high, performing quantitative in vivo imaging requires a higher dynamic range than what is required for in vitro experiments because of factors such as background fluorescence. Therefore, a great deal of research effort has been invested in improving the dynamic range and signal/noise ratio of genetically encoded sensors.

Although algorithms to predict the efficiency of a particular construct have been developed [71], generally it is difficult to predict the dynamic range of a sensor from the structures of FPs and binding modules. The steric effects of conformational change of the binding module on conformation of fluorophore(s) are often unpredictable. This is exemplified by cases where a single mutation in binding modules changes the direction of response (i.e. increase or decrease of FRET efficiency) induced by ligand binding [49]. Therefore, the strategies employed so far mostly relied on empirical exhaustion of possible configurations between the binding domain and FPs. Testing multiple sensory modules as scaffolds [48,49], systematic adjustment of the linker sequence between the binding module and FPs [28,51], replacing FPs with FRET-optimized fluorophore pair [72], and introducing circular permutation into FPs [73,74] and binding modules [75] have proven to be effective strategies for improving the dynamic range.

In a number of cases, decreases in the dynamic range of the sensors were observed in vivo compared to the value obtained from purified sensor proteins ([73,74], Okumoto unpublished results). While in most cases the reason for such a decrease in the dynamic range is unknown, it is reasonable to speculate that complexity in the cellular milieu could restrict the motion of sensor proteins and hence influence the dynamic range. Therefore, at this point the optimization of genetically encoded sensors requires systemic and rigorous tests in the system for which the sensors are intended (i.e. living cells) [28,35].

4. In vivo measurement of metabolite and ion levels

4-1 Steady-state level of substrates in a subcellular compartment

Metabolism in eukaryotic cells is highly compartmentalized; steady-state metabolite concentrations in each compartment are difficult to measure using classic biochemical approaches because of potential cross-contamination. Moreover, cellular and sub-cellular metabolite levels in transient states (e.g. elevated metabolite level in the environment) are near impossible to capture since the turnover rates of these molecules are extremely high. Consequently, a surprisingly small number of studies have reported absolute concentration of metabolites in cellular compartments.

Genetically encoded sensors allow real-time measurement of a substrate at subcellular resolution by targeting sensor proteins to the compartment of interest, offering potential to fill the current knowledge gap. To investigate concentration of a substrate of interest using genetically encoded sensors, a sensor that matches the range of concentration in the tested condition has to be expressed in the compartment of interest. Estimating substrate concentration in a given compartment is generally difficult, since currently there are few technologies that allow such measurement. Therefore it is advisable to simply introduce an array of sensors with different affinities for the cell type of interest. This process assures that the full range of substrate fluctuation is covered, especially for substrates with larger concentration range than the detection range of one sensor (Fig. 2A). In addition, sensors that are out of their dynamic range (i.e. either saturated or do not bind to the substrate at the given concentration, see Fig. 2B) serve as excellent negative controls for a given experimental condition. Since in most cases affinity mutants are created by small number of mutations in the binding domain, artifacts caused by experimental conditions (i.e. fluorescent intensities of FPs) can be examined using these “out-of-range” sensors. Using this principle, Deuschle et al. demonstrated that the steady-state concentration of glucose in epidermal cells and root cells from Arabidopsis plants are drastically different [76].

Fig. 2. Conceptual concentration substrate level change in the cytosol.

A. Two model cases where the steady-state cytosolic levels are altered. The box above the trace indicates the time period when the substrate was externally supplied. The 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.

Typically, to monitor concentration of a specific substrate in a cellular compartment, cells expressing sensor proteins in the desired compartments are perfused in media with incremental increase in the concentration of the substrate (Fig. 3A). FRET efficiency change at a given concentration relative to the maximal FRET efficiency change can be calculated according to Equation 2.

$$\mathrm{S}=(\mathrm{\Delta }{\mathrm{E}}_{\mathrm{x}}-\mathrm{\Delta }{\mathrm{E}}_{\text{apo}})/(\mathrm{\Delta }{\mathrm{E}}_{\text{sat}}-\mathrm{\Delta }{\mathrm{E}}_{\text{apo}})$$ (Eq.1)

where S is saturation, ΔE_x is the FRET efficiency change at the given concentration, ΔE_apo is the FRET efficiency change without the external substrate, and ΔE_sat is the FRET efficiency change in saturating concentration of the substrate.

Fig. 3. Determining cellular concentration using FRET sensors.

Black and red traces represent data from two independent cell types expressing the same sensor. A. Typical time-course of FRET efficiency with step-wise increase in concentration of the substrate. Boxes on top represent the period when the substrate was externally supplied. B. Relative FRET efficiency change (ΔE_x-ΔE_min)/(ΔE_max-ΔE_min) follows a hyperbolic saturation curve. C. The internal concentration in given external concentration can be calculated from the Eq. 3.

The relative FRET efficiency change at a given concentration follows a hyperbolic curve, reflecting the saturation curve of the sensor (Fig. 3B). Therefore, the internal concentration can be calculated from the FRET efficiency by fitting the saturation ΔE_x/ΔE_max with the saturation curve of the sensor;

$$(\mathrm{\Delta }{\mathrm{E}}_{\mathrm{x}}-\mathrm{\Delta }{\mathrm{E}}_{\text{apo}})/(\mathrm{\Delta }{\mathrm{E}}_{\text{sat}}-\mathrm{\Delta }{\mathrm{E}}_{\text{apo}})=[\mathrm{L}]/(K_d+[\mathrm{L}])$$ (Eq.2)

where K_d is the half-saturation point of the sensor calculated from the in vitro titration. The internal concentration at a given extracellular substrate concentration can be calculated accordingly (Fig. 3C).

To perform the calculation above, it is essential to determine the maximal response of the sensor (ΔE_max). For substrates that are rapidly transported into cellular compartments, typically this step is easily achieved since the range of substrate exceeds the dynamic range of the sensor. When the substrate concentration change in the given compartment is smaller than the dynamic range of the sensor protein, absolute concentrations of metabolites can be estimated by in situ titration using drugs that permeate the metabolite of interest [39]. Care must be taken when using such an approach, because the changes caused by drugs can directly affect properties of FPs.

4-2 Characterization of cellular transport mechanisms

For many biologically important metabolites and ions, a single cell expresses multiple transporter proteins that control the accumulation and elimination in a given subcellular compartment. Except for in a limited number of cases, the dominant transporter in a particular subcellular compartment is not known, and even in organisms with sequenced genomes, there are a large number of orphan transporters that could also contribute to the transport processes. Metabolite sensors, in combination with methods that allow inhibition of specific gene product(s) (e.g. chemical inhibitors and RNAi), can be used to examine the contribution of individual transporters to the collective flux of the substrate. For example, investigation of glucose transport kinetics in Arabidopsis roots using FRET glucose sensors demonstrated that the kinetics of cytosolic glucose in roots exposed to external glucose do not rely on a pH gradient across the membrane, suggesting the existence of novel mechanisms for glucose transport that do not rely on H⁺-symport [77]. In other experiments, a combination of chemical inhibitor and RNAi revealed a predominant transporter (i.e. GLUT1) in HepG2 cell line, demonstrating that these sensors can be used as a mean of analyzing transporter activities in a specific cell type [78,79]. Importantly, the substrate elimination process from a cellular compartment, which otherwise requires pre-loading of labeled compounds, can easily be traced using sensors, using either sensor proteins expressed in the compartment of interest [50] or supplied exterior to the cells of interest [80]. Compared to metabolite import into cells, export processes and their components are largely unexplored. Combined with a high-throughput imaging approach, biosensors for metabolites might serve as platforms for discovery of molecular mechanisms for cellular metabolite export, as well as the chemical inhibitors of such processes [79].

Ultimately, transporters and enzymes that are characterized should be integrated into a model that explains the behavior of cells as a whole. Metabolite concentration measured using a biosensor in a given compartment is the summation of uptake, biosynthesis, catabolism and export, and therefore provides a means to examine whether a given model explains the flux of metabolite. The dynamics of metabolite level in the subcellular pool can be predicted using a set of differential equations describing transporter and enzyme reaction kinetics, which can be fitted to the experimental data [81]. Construction of such a model requires that all fluxes around the substrate of interest are identified at the subcellular level, and therefore is not feasible in many cases. Nevertheless, using this approach, Fehr et al. predicted bi-directional glucose transport mechanism on the ER membrane of hepatoma cell line HepG2 [78].

Although the activities of some metabolite transporters are rapidly regulated at the post-transcriptional level, typically cellular metabolite levels change in the course of minutes [46,49,50,77,78]. The temporal-resolutions of genetically encoded sensors are in the order of ten to one hundred milliseconds, and therefore are sufficient in resolving such relatively slow changes. In stark contrast, the temporal signatures of signaling molecules are in the order of tens of milliseconds. Hence the reaction kinetics of detection methods is an important consideration when interpreting the data. Improvement in temporal resolution of genetically encoded sensors has been reported, but even the responses of the fastest sensors reported are still slower than synthetic dyes [82,83]. In situations where head-to-head comparison between genetically encoded sensors and synthetic dyes are possible, an algorithm to resolve a single peak can be developed [84]. Obviously such a comparison is not possible for molecules to which specific dyes are not available, hence the on and off rates of sensors have to be kept in mind when interpreting transients below mid millisecond range.

5. Emerging application of genetically encoded sensors

Due to rigorous improvements in signal/noise ratio and the dynamic range of biosensors, at least for a handful of molecules it is now possible to ask biological questions that require faster and more complex detection systems. Research areas that have witnessed particular progress are summarized as follows. Even though these advanced imaging techniques have been reported for a limited number of metabolites and ions, we predict that these techniques will become applicable for an increasing number of molecules as the repertoire and properties of biosensors improve.

5-1. Dual real-time imaging using genetically encoded sensors

One of the most exciting outcomes of genetically encoded sensors is the possibility to examine correlations between signaling molecules. For example, the signaling pathways of Ca²⁺ and cAMP are interconnected; the enzymes that catalyze the synthesis and degradation of cAMP can be regulated by Ca²⁺ and conversely, cAMP can modulate Ca²⁺ influx through the plasma membrane by modulating the activities of Ca²⁺ channel proteins [85]. The availability of genetically encoded cAMP sensors that can be spectrally separated from calcium dye enabled simultaneous measurement of two molecules in a single cell. Using this system, researchers revealed a close temporal and causal relationship between the cytosolic Ca²⁺ and cAMP fluctuation caused by membrane depolarization. In another study, using a FRET based IP₃ reporter and Ca²⁺ dye, the authors demonstrated that the oscillation of IP₃ is not essential for the oscillation of Ca²⁺[39].

Recent expansion of hues in FPs enabled simultaneous recording of two FP-based FRET pairs. ECFP/EYFP and mOrange/mCherry [86], mTFP/Citrine and mAmetrine/tdTomato [87], ECFP/mVenus and TagRFP/mPlum [88], and CFP/YFP and Sapphire/RFP [89] have been successfully used to monitor two simultaneous events in a single cell. Using four FPs inevitably results in larger problems of spectral bleed-through. This problem can be circumvented by detection methods using Fluorescent Lifetime IMaging-FRET (FLIM-FRET) since the method does not require the measurement of acceptor channels [88]. Fluorescence lifetime is also independent of fluorophore concentrations, and therefore provides more accurate measurement in samples with heterologous fluorophore distribution. Since FLIM-FRET experiments require specific setups, and acquisition of lifetime images from the whole cell takes longer than fluorescent-intensity measurement, the recording of genetically encoded FRET sensors using this technique has not been the common method of choice. However, one can expect that the advancement in acquisition of analytical technologies will make this technique accessible to more researchers in the near future.

5-2 High-throughput applications

Recent progress in automated, high-throughput imaging methods has opened up exciting possibilities for genome-wide functional screening [90–92] and image-guided, high-throughput drug discovery assays [93]. These methods generally rely on quantifiable changes (i.e. fluorescence intensity, altered FP-tagged protein localizations etc.) that can be automatically identified among a large collection of images. Although initially such high-throughput studies were typically performed using fixed cells, recent advances in high-throughput microscope systems are starting to make live cell studies more accessible, opening an avenue for cell biology studies at the genome-wide scale.

Since genetically encoded fluorescent sensors can provide quantitative data about a variety of events in living cells, their potential as reporters for high-throughput, image-guided screening was well recognized by researchers, and in some cases demonstrated on a smaller scale [79]. Recently Jiang et al. successfully identified a mitochondrial Ca²⁺/H⁺ antiporter through genome-wide screening using a mitochondrial calcium sensor as a reporter [94]. In their study, “pericam”, a genetically encoded sensor that responds to Ca²⁺ and pH, was expressed in the mitochondria of Drosophila S2 cells to screen for genes that are responsible for calcium uptake into mitochondria using an RNAi library. This example demonstrates that subcellular events can now be reported using genetically encoded sensors with the robustness that is required for high-throughput studies. The signal/noise ratio of genetically encoded sensors and the image resolution, throughput and data analyses of high-throughput imaging systems are rapidly improving. Thus, genetically encoded sensors will serve as indispensable tools for genome-wide analyses of cellular functions in the future.

5-3 Chronic imaging of a specific set of cells in live animals

One of the ultimate goals of studying signaling molecule at the cellular level is to establish the relationship between cellular signaling network and the resulting complex response at the organism level. To answer such questions, it is mandatory to develop a system where cellular responses can be measured directly while responses at the organism level (e.g. movement) take place. To this end, rapid advancement is currently being made in the development of microscopes that can be mounted on live, moving animals [95–97]. Studies using such setups in combination with in vivo whole cell recording demonstrated its usefulness in examining mathematical models as to how hippocampal place cells encode spatiotemporal information [98].

Synthetic calcium indicator dyes have been successfully used in such setups to monitor neuronal activities of a large population of cells in moving animals [95–97]. While approaches using indicator dyes are suitable for short-term experiments, questions that require chronic (long-term) preparation such as learning paradigms are more challenging because bolus loading of calcium indicator disturbs brain physiology. Since genetically encoded sensors can be stably expressed in the same set of cells, they can be used for chronic imaging experiments. Although the use of genetically encoded sensors in imaging of moving animals has been limited by problems such as relatively poor signal-noise-ratio and protein instability, iterative effort to improve the properties is starting to produce genetically encoded sensors that are amenable to such in vivo chronic studies. Recently, two groups reported chronic imaging of calcium transient in live mouse brain using genetically encoed calcium sensors [68,99]. As the varieties and properties of genetically encoded sensors improve, this exciting new area of research will provide insights into the cellular signaling network underlying long-term physiological changes induced by developmental or conditional (e.g. learning) stimuli.

6. Conclusion

Genetically encoded biosensors became indispensable tools to analyze information encoded in the transients of signaling molecules. It also made it possible to analyze cellular activities such as metabolite export in unprecedented temporal resolution, opening up an exciting possibility to uncover previously unknown molecular machinery. Notable results published in the last few years bear evidence of the extensive effort that has been put into the improvement of genetically encoded biosensors. At least for a few molecules, biosensors have now improved to the point where they can be used in more demanding experimental settings such as chronic studies of specific cells placed in their native context (e.g., specific nerve cells in brain), and high-throuput application. Especially in neurobiology, improved biosensors are expected to serve as reporters of neuronal activity network initiated from localized excitation to decipher neuronal connections [100].