Green light for quantitative live-cell imaging in plants

Since Robert Hooke's observation of the first cells in the 17th century, plants have been of central importance for numerous key discoveries in cell biology. This is exemplified by the cell theory that was originally formulated in 1838 (‘Beiträge zur Phytogenesis’) by the botanist Matthias Jakob Schleiden, who suggested that every structural element of plants is composed of cells or their products. The same conclusion was reached a year later by the zoologist Theodor Schwann (Mazzarello, 1999). Modern plant cell biology has greatly benefited from genetic and molecular tools that have enabled visualization of dynamic subcellular processes in living plant cells and helped to shed light on the plastic development of plants.

Plants exhibit an outstanding flexibility of adapting their morphology, optimizing their metabolic activity or the timing of developmental programs to fluctuating environmental conditions, such as the initiation of reproductive organs (De Storme and Geelen, 2014). This plasticity depends on intricate sensing and signaling networks that we are only beginning to unravel. Fluorescence-based microscopy has revolutionized various fields of plant biology, such as large-scale organ morphogenesis, Ca²⁺ signaling during sexual reproduction or membrane protein dynamics during pathogen attack (Barbier de Reuille et al., 2015; Fernandez et al., 2010; Shaw and Ehrhardt, 2013). The need to image live plant cells in intact tissues raised the demand for developing novel experimental tools and setups (Shaw and Ehrhardt, 2013). The greatest challenge is the inherent autofluorescence of most plant cells, which is largely caused by the presence of chlorophyll and carotenoids in plastids, as well as by lignin and other phenolic compounds in cell walls (Shaw and Ehrhardt, 2013). This limits high-resolution approaches, such as single-particle tracking or live-cell imaging of plant samples with low fluorescence, and imaging deep within the tissue. Another challenge is to establish growth systems that allow non-invasive, high-quality imaging of intact and growing plant organs, such as roots, or entire plants.

In this Review, we focus on a number of selected techniques and tools that have recently enabled a number of discoveries in plant cell biology (Fig. 1). We first illustrate several of the key challenges that are inherent to live plant imaging, followed by an introduction into how recent technological developments have allowed plant cell biologists to image deep inside tissues for extended time and in defined micro-environments. Finally, we examine state-of-the-art technical solutions and their limitations for imaging and quantifying the distribution of metabolites, proteins or proteins complexes in live plant tissues. Some additional key technologies, such as super-resolution and imaged-based phenotyping have been recently discussed elsewhere (Komis et al., 2015; Rellán-Álvarez et al., 2015; Schubert, 2017), and will not be covered here.

Differentiated plants cells feature a vacuole that occupies the bulk space of the cell. Consequently, each plant cell behaves like a lens, and this affects light propagation (Berthet and Maizel, 2016). In addition, plants produce a plethora of metabolic compounds that exhibit autofluorescence characteristics upon excitation. This autofluorescence is most often strongest in the blue, but when investigating photosynthetic tissues, it also masks a large proportion of the red spectrum (Jamme et al., 2013). The simplest option, even though it is often not feasible, is to avoid fluorophores with overlapping excitation or emission spectra. However, in modern confocal systems, there are several ways to minimize autofluorescence cross talk. One approach is spectral unmixing – a common feature of modern confocal microscopes – to eliminate the contribution of autofluorescence to a confocal image (see Glossary). Where tunable lasers are available, for example, in systems for two-photon-excitation microscopy (TPEM), it is highly recommended to scan different excitation wavelengths to select the one with the best signal-to-noise ratio (SNR). A more recent development is the use of flexible pulsed lasers together with detectors that have a photon-counting ability (Kodama, 2016). This allows time-resolved fluorescence detection and gating based on the fluorescence lifetime of the respective molecule(s). As autofluorescence typically has a very short lifetime (picosecond range), selecting for the longer lifetimes of fluorescent proteins (nanosecond range) can eliminate a large proportion of the autofluorescence (Kodama, 2016). Besides advances on the instrumentation side, clearing techniques that diminishes chlorophyll autofluorescence while maintaining fluorescent protein stability, such as CLEARSEE (Kurihara et al., 2015), have facilitated imaging-based approaches. Although these protocols are not compatible with living tissue, they can be extremely useful to improve imaging of thick specimens or tissues that normally exhibit strong autofluorescence.

In addition, tropic responses to light and gravity have to be considered in plants. With a few exceptions (see light sheet microscopy below), conventional microscopes are built to hold specimens on horizontal stages. This horizontal mounting, however, conflicts with plant gravitropism, which leads to shoots bending up, and roots bending downwards (Muday, 2001). To allow vertical mounting, a recent study presented a custom-made tilted confocal laser-scanning microscope that does not constrain the gravitropic response of the plant (von Wangenheim et al., 2017). This combined hardware and software solution enables tracking of the root tip while it grows along the gravity vector, and therefore opens up new avenues for studying undisturbed root growth with cellular resolution (von Wangenheim et al., 2017). However, this solution relies on the customization of a microscope, which may limit its widespread adoption in the community. In summary, although live imaging of plants poses specific challenges (autofluorescence and the natural tropic responses of plants), technical solutions have been found to diminish these and allowed for the emergence of specific imaging techniques (Table 1).

Over the past years, several techniques have been developed that enable non-invasive imaging of roots down to the cellular level. Specific imaging setups allow growth of plants either between cover glass and a mesh separating roots from soil (Froelich et al., 2011), on sterile medium with roots being covered with gas-permeable, transparent plastic film (Fournier et al., 2008), or imaging and perfusion of seedlings within self-made chambers (Kirchhelle and Moore, 2017; Krebs et al., 2012). One milestone was the development of microchannel platforms for Arabidopsis roots (Grossmann et al., 2011; Meier et al., 2010; Parashar and Pandey, 2011) that took advantage of the compact design, versatility and cost-effective fabrication of polydimethylsiloxane (PDMS)-based microfluidic devices. Here, primary roots grow into observation chambers; thereby, specimen mounting occurs without direct specimen handling. The possibility to cultivate roots in protected chambers inside devices that match the needs that are specific to the experiment has substantially facilitated quantitative and dynamic measurements and provided the ability to precisely control the plant microenvironment. Lab-on-a-chip devices are typically characterized by miniature channel systems that guide the flow of liquids and enable parallel and multiplexed analyses, while consuming minimal fluid volumes (Sia and Whitesides, 2003). Although cell biologists have already utilized microfluidic devices for some time to allow long-term cultivation of microbes or eukaryotic cell cultures (Whitesides, 2006), this technology has found its way into organismal biology only more recently and is now applied to multicellular organisms, such as nematodes, insects, fish and mammalian embryos and plants (Stanley et al., 2016). The adoption of the microfluidic lab-on-a-chip technology for plant cell biology has enabled novel approaches to explore environmental sensing, cellular and organ growth mechanics, and nutrient uptake kinetics.

As an example, the RootChip (Fig. 2A) was developed to host growing primary roots of Arabidopsis seedlings that can be subjected to pulsed treatments (Grossmann et al., 2011). Often used in combination with genetically encoded fluorescent sensors (see below), the RootChip technique has been applied to various imaging-based applications. Examples are quantitative analyses of small-molecule dynamics during nutrient transport (Grossmann et al., 2011), heavy-metal homeostasis (Lanquar et al., 2014), hormone (Jones et al., 2014) and Ca²⁺ signaling (Denninger et al., 2014; Keinath et al., 2015), as well as quantitative root phenotyping (Grossmann et al., 2012; Xing et al., 2017). As a high-throughput root-observation platform, the RootArray has enabled quantitative analyses of cell fate-specific gene expression of 64 roots in parallel (Busch et al., 2012). To facilitate high-resolution phenotyping of even larger numbers of Arabidopsis seedlings, a microfluidic device with automatic trapping of seeds in micro-wells has been developed (Jiang et al., 2014).

A benefit of long-term cultivation of live specimens in microfluidic perfusion systems is their potential to aid studies on interorganismic interactions: one of the earliest devices was designed to observe the feeding of nematodes on Arabidopsis roots, as well as the infection of roots with bacterial plant pathogens (Parashar and Pandey, 2011). Recent work has expanded this approach to investigate the formation of microbial biofilms along specific root zones (Massalha et al., 2017) (Fig. 2B). These examples highlight the possibilities of microfluidic devices to open up new avenues to gain a better understanding of plant development in complex biological environments (Stanley et al., 2016). Microfluidic devices will therefore likely push the boundaries of research on plant–microbial communities (Stanley and van der Heijden, 2017).

A technical challenge in any perfusion system is the targeted application of treatments to selected regions of the specimen. In all the microfluidic systems that are mentioned above, plant organs are subject to global treatments. However, environmental conditions are often highly heterogeneous, with local differences in nutrient availability, the chemical milieu or microbial abundance. Cellular Ca²⁺ responses have been recorded in experiments using micromanipulators to apply localized force to single epidermal cells in roots (Monshausen et al., 2009). A first approach to apply a local stream of liquid within a microfluidic device with a resolution close to that of a micromanipulator probe was developed by the use of a focused laminar flow that was directed perpendicular to a mounted Arabidopsis root (Meier et al., 2010). Here, the localized treatment with the phytohormone auxin resulted in arrested cell elongation and root hair development at the site of stimulation (Meier et al., 2010). Such a local application to roots had previously been approached by the application of auxin-loaded sephadex beads along the root, which led to the finding that auxin gradients play a role in planar polarity, defining the regular positioning of root hairs in the root epidermis (Fischer et al., 2006). Recently, the accurate positioning of external auxin gradients was achieved by employing organic electron ion pumps (OEIPs), which can deliver charged compounds at high spatial resolution to roots grown on gelled media (Poxson et al., 2017). In addition, a novel adaptation of the RootChip concept now allows the generation of asymmetric root environments by guiding root growth through a micropillar array and perfusing the root with different liquids on either side (Stanley et al., 2017) (Fig. 2C). This approach enables the simulation of environmental heterogeneity and the investigation of cell autonomous and systemic mechanisms of root development.

Microfluidic devices have also substantially advanced experimental access to other plant systems. The moss Physcomitrella patens, an emerging model to study plant evolution, features a filamentous network of tissues that makes long-term imaging of growth and development challenging when conventional mounting techniques are being used (Bascom et al., 2016). The first microfluidic device for moss solved this issue by guiding growing filaments through an intermittent barrier into an observation chamber, where they can be studied outside the dense protonemal meshwork (Bascom et al., 2016). For another tip-growing plant cell type, pollen tubes, several micro-devices have already been developed to investigate diverse cellular properties, such as their growth dynamics (Nezhad et al., 2013), their chemo-attraction by the female gametophyte (Horade et al., 2013) and their ability to penetrate narrow gaps (Yanagisawa et al., 2017) (Fig. 2D).

Taken together, microfluidic devices have substantially improved access for imaging plant tissues at high resolution and over extended periods of time. The application of microfluidics will, for some time, remain limited to young seedlings or isolated parts of plants. However, owing to their design flexibility, microfluidic devices bear great potential for cell biological studies in combination with new imaging techniques that require non-invasive specimen immobilization, environmental control and time-lapse imaging with observation times of several days.

Although the combination of advanced growth systems and conventional confocal microscopy has revealed many new biological insights, it is less suited to imaging of deep-lying tissues and thick specimens. Owing to substantial advances in instrumentation and user-friendliness, multi-photon microscopy has become a powerful technique to visualize plant development in 3D.

The most easily implemented and used excitation mode for multi-photon microscopy is the above-mentioned TPEM. Here, two low-energy (longer, red-shifted wavelength) photons combine in the focal plane to excite the target fluorophore. TPEM makes use of red-shifted wavelengths for excitation without changing the emission spectra. Since the two photons only come together in the focal plane, the energy level that provides excitation is only sufficient in the focal plane. This also reduces photobleaching of regions of the sample that are out of focus, which is an issue in confocal microscopy. As all emitted fluorescence originates from a single focal plane, there is no need for a pinhole (Brakenhoff et al., 1996; Denk et al., 1990; Feijo and Moreno, 2004). Hence, the detectors in a TPEM system can be placed in close proximity to the objective in order to reduce the light path of the emitted fluorescence (i.e. the fluorescence does not pass the scan box again, as is the case when using internal detectors). These ‘non-descanned’ detectors are required to maintain the full potential of TPEM and are therefore strongly recommended (Ustione and Piston, 2011). Two-photon excitation spectra of fluorescent proteins often differ (slightly) from the one-photon excitation spectra; a property that can be useful to enable simultaneous imaging of two or more fluorescent proteins with a different single excitation wavelength (Drobizhev et al., 2011, 2009). This accelerates image acquisition and hence reduces phototoxicity. Theoretically, by applying the Abbe–Raleigh criteria (see Glossary), the use of short wavelengths in a well-aligned confocal microscope is half of the resolution of a TPE microscope. However, the strong increase in SNR of TPEM provides a spatial resolution that is very similar to that of standard confocal microscopes (Ustione and Piston, 2011).

In addition to these technical developments, the availability of a large set of cell type-specific promoters in Arabidopsis thaliana has been instrumental to improve imaging of deep-lying tissues (Birnbaum et al., 2003; Gooh et al., 2015). The expression of a fluorescent protein fusion in the tissue of interest makes it more straightforward to interpret generated images, as no signal that can obstruct the view is derived from surrounding tissues. Recently, the combined use of cell type-specific promoters and TPEM has been used to perform, for the first time, live-cell imaging and cell lineage tracing in developing Arabidopsis embryos with unprecedented temporal and spatial resolution (Gooh et al., 2015). Moreover, a related study from the same group revealed the differential contributions of the microtubule and actin cytoskeleton during the first asymmetric division in the Arabidopsis zygote. By comparing confocal microcopy to TPEM, they clearly showed that the latter resulted in better images (Kimata et al., 2016). One potential drawback of TPEM is the use of high-power lasers for excitation. Therefore, users need to take care to not overheat the samples. However, this feature can be turned into an advantage by using the high-energy TPE microscope to specifically ablate deep-lying cells without disturbing the surrounding cells. The suitability of such an approach was recently shown, when TPEM was applied to specifically ablate single cells in the Arabidopsis embryo (Gooh et al., 2015).

In addition, a very recent exciting development is the use of adaptive optics in TPEM, which resulted in a significant increase in resolution; however, this technique has not yet been used on plant specimens (Zheng et al., 2017). Overall, when the goal is deep-tissue imaging with reduced photobleaching, TPEM is clearly the best approach. However, when imaging cell layers at surface or cell cultures, conventional confocal microscopy will be more suited.

The holy grail for many cell and developmental biologists is the ability to capture and analyze cell dynamics within the tissues of an intact organism. Such high-resolution imaging in 3D, deep in tissues, over prolonged periods of time that can range from minutes to hours or days, has already been achieved (De Rybel et al., 2010; Goh et al., 2016); however, maintaining the specimen in physiological conditions currently represents a major challenge. Over the last 10 years, light sheet fluorescence microscopy (LSFM) has emerged as a powerful technology to tackle this challenge (Höckendorf et al., 2012). LSFM differs from traditional fluorescence microscopy in that two optical axes are used. The first axis solely takes care of the illumination of the sample; it collimates a sheet of laser light that is a few micrometers thick. Thereby, it only excites fluorophores that are located in a small volume of the specimen, whereas the rest is kept in the dark. Photons that are emitted in this volume are collected by the second optical axis, which is orthogonal to the illumination axis. This detection axis captures the whole field of view with a camera, without the need of scanning, as it is the case in laser-scanning confocal microscopy. This massive parallelization of detection allows for fast imaging (∼20–30 frames per second), thus reducing the duration of exposure of the specimen and the risk of photobleaching and phototoxic effects. In addition, the quantum yield of modern electron-multiplying charge-coupled device (EM-CCD) cameras allows the use of less incoming laser energy to obtain the same SNR, which contributes further to minimize the deleterious effects of exposing the specimen to intense laser light for a long time (Stelzer, 2015) (see Glossary). Beyond this, the major advantage of LSFM becomes evident when one considers the amount of energy that is encountered by the entire specimen when it is imaged iteratively to generate a z-stack. In LSFM, individual optical planes are consecutively illuminated, whereas standard confocal microscopes rely on whole-specimen illumination and subsequent elimination of out-of-focus light by using a pinhole in front of the detector. Consequently, and in contrast to LSFM, the amount of energy that is received by the sample in confocal microscopy is proportional to the number of images in the stack. Thus, phototoxic effects are drastically minimized in LSFM (Stelzer, 2015). Despite these clear advantages, LSFM is not perfect. The most notable problem is the scattering of both excitation and emission light, especially for thick specimens. Modern LSFMs combine several technological improvements to minimize these adverse effects of scattering (de Medeiros et al., 2015; Krzic et al., 2012). They are designed around four orthogonal optical axes – two for detection and two for illumination – which allows the simultaneous acquisition of four images and their merging into a single one. In addition, modern LSFMs implement a slit detection mode to minimize the detection of scattered light by the camera: a scanned Gaussian beam is used to generate the light sheet and is synchronized with the rolling shutter camera. With this, only photons that originate from the sharpest area of the image reach the camera (de Medeiros et al., 2015). This feature is reminiscent of the pinhole of a confocal microscope, but with a much-reduced penalty on imaging speed. Of particular importance for long-term imaging (hours to days) is to keep the plant in a close-to-physiological environment. Since its implementation for Arabidopsis imaging (Maizel et al., 2011), efforts have focused on emulating the conditions that plants usually encounter in a laboratory set up into the sample chamber of the microscope. Most LSFMs have a design where the sample is held vertically, which is an advantage over most other setups that impose plants to grow horizontally against their natural tropisms. Furthermore, the presence of a light source that emulates the day and night cycle contributes significantly to the survival of the plant during long-term imaging; this guarantees that any observations made have a physiological relevance.

The application of LSFM to plant cell and developmental biology has essentially been focused on the root of Arabidopsis thaliana because of its small diameter and good optical properties. The growth dynamics of the primary roots (de Luis Balaguer et al., 2016; Maizel et al., 2011; Sena et al., 2011) and of lateral roots (Lucas et al., 2013; von Wangenheim et al., 2016) have been documented over the course of hours to days at cellular resolution, which revealed new insights on the dynamics of these processes. At the subcellular levels, LSFM has been used to observe the dynamics of the secretory machinery (Berson et al., 2014) and Ca²⁺ dynamics (Costa et al., 2013).

High-resolution imaging of membrane- or sub-membrane-resident proteins or metabolites is often restricted by the limited resolution of confocal systems along the z-axis, which results in comparably high SNRs. A potent method to overcome these limitations is total internal reflection (TIRF) microscopy. Different to confocal and epifluorescence microscopes, the angle of the laser beam here is adjusted to a super-critical angle (65–67°), where all light is reflected at the coverslip. This generates an evanescent wave that penetrates the sample for less than 400 nm (Konopka and Bednarek, 2008). Even though this penetration depth is far more than the 6 nm thickness of a plasma membrane, TIRF microscopy provides a viable method to greatly improve the SNR of such samples to discriminate (sub-)membrane protein populations and to perform single-particle tracking (Wang et al., 2015b). Best results are obtained when using a number of specifications that have been adapted to the requirements of imaging plant tissues (Jaqaman et al., 2008; Wang et al., 2015b). When applying TIRF microscopy on plant cells, the existence of the cell wall with its thickness between 0.1–1 μm needs to be considered. As a consequence, and in contrast to cell cultures or fresh protoplast preparations, plant cell plasma membranes are never in direct contact with the coverslip when imaged in a multi-cellular tissue context. This may be seen as an inherent advantage, as it further reduces signals that derive from the deeper cytoplasm, but it is equally a limitation with respect to sample preparation. In contrast to cell wall-free cultured mammalian cells, which can be grown directly on the coverslip, or cell wall-containing unicellular organisms such as yeast, which sediment on the glass, the surface of intact plant tissues is comparably uneven (Koch et al., 2008). As a consequence, most regions will be out of focus and thus dramatically limit the field of view during TIRF image acquisition. This is even further pronounced when imaging intact tissues rather than cultured cells, as specimen thickness and shape result in major regions of the tissue being impossible to image simultaneously. In these cases, and in order to avoid lateral drift of the samples that could be caused by evaporation of the immersion medium from the edges of the coverslip, researchers are well advised to seal their samples prior to image acquisition. The comparably great distance from the cover glass can be partially compensated for by applying limited forces on the specimen to gently lower the plant tissue towards the incident light beam (Wan et al., 2011). To further overcome the limitation of distance from the cover glass, variable angle epifluorescence microscopy (VAEM) has been introduced for plant cell imaging (Konopka and Bednarek, 2008) and further plant-specific adaptations in laboratory protocols have been evaluated (Wan et al., 2011). In VAEM, sub-critical angles (59–61°) are used for the incident beam that result in the refraction of the light and a narrow field of illumination.

In plants, VAEM and/or TIRF microscopy have been used, among other approaches, for the imaging of the segregation of proteins into plasma membrane nanodomains (Bücherl et al., 2017; Gronnier et al., 2017; Hao et al., 2014; Hutten et al., 2017; Jarsch et al., 2014; Li et al., 2012). In addition, VAEM was used for single-particle tracking of low-abundant receptors such as BRASSINOSTERIOD INSENSITIVE 1 (BRI1), which was found to segregate into two distinct subpopulations of membrane nanodomains with different motion ranges (Wang et al., 2015a) and to interact in discrete nanodomains with its co-receptor (Hutten et al., 2017). Such observations are clearly facilitated by the fact that most TIRF systems use EM-CCD or sCMOS cameras (see Glossary), which have significantly higher quantum yields in comparison to other detectors, such as photo-multipliers (Shaw and Ehrhardt, 2013), as mentioned above. However, a key restriction that applies to TIRF microscopy – as to any other widefield or standard confocal technique – is the resolution limit along the xy axis. Although a new class of detector (Huff, 2015) or refined deconvolution approaches (Borlinghaus and Kappel, 2016) allow lateral resolution of structures in the range of ∼140 nm, other approaches are required to precisely described closely associated protein assemblies of physical protein–protein interactions.

In summary, VAEM and/or TIRFM is applicable in plants despite the presence of a plant cell wall. Considering the dimensions of a plant cell, the use of VAEM or TIRFM allows a significant reduction in background fluorescence when imaging membrane-resident and objective-facing proteins, even though the field of view is limited compared to confocal or widefield setups.

In order to investigate physiological adaptations and dynamics of plant cells on a molecular level, fluorescent indicators are powerful tools that enable non-invasive measurements of various biochemical and biophysical cellular parameters with high spatio-temporal resolution (Table 2). Based on their make-up, small-molecule dye-based fluorescent indicators are distinct from genetically encoded indicators that are based on contain fluorescent proteins and peptides as structural elements. To date, a great variety of fluorescent indicators exist that cover a broad spectrum of applications from the detection of numerous kinds of ions and metabolites to the assessment of cellular redox state, hydrostatic pressure, molecular crowding and membrane potential (Germond et al., 2016; Sanford and Palmer, 2017; Ueno and Nagano, 2011; Uslu and Grossmann, 2016).

Derivatives of small-molecule fluorescent indicators (SMFIs) that permeate the membrane can be easily loaded into cells. However, owing to impregnations of the plant cell wall, such as cuticle waxes or suberin depositions, the loading of SMFIs into intact plant cells is restricted to non-cutinized and non-suberized tissues. Enzymatic cell wall digestion and mechanical opening still provide alternative strategies to make these tissues accessible (Gilroy et al., 1986; Kuchitsu et al., 2002), but certainly represent invasive approaches that need to be tightly evaluated through the use of appropriate controls. The subcellular distribution of SMFIs can also differ between model organisms: an example are the widely-used pH indicators of the fluorescein family accumulate in the cytoplasm of mammalian cells, whereas in plants and fungi, they are efficiently sequestered to vacuoles (Slayman et al., 1994). In Arabidopsis, these dyes have therefore become valuable tools to measure vacuolar pH in intact roots (Krebs et al., 2010), but have also been established as vacuolar lumen stains to study vacuole biogenesis and to obtain detailed three-dimensional reconstructions of plant vacuoles (Kriegel et al., 2015; Viotti et al., 2013). A summary of challenges and solutions when using fluorescent indicators in plant cells can be found in Table 1.

A better spatial control than for SMFIs is achieved by using genetically encoded fluorescent indicators (GEFIs). Their spatio-temporal expression and cellular localization can be precisely controlled by the use of appropriate promoters and targeting sequences, which enable selective analyses of different tissues, cell types and compartments (Table 2). The majority of available GEFIs have been developed and enhanced in vitro and functionally tested in mammalian cells (Miesenböck et al., 1998; Tantama et al., 2013; Thestrup et al., 2014). The transfer of GEFI applications to plant model organisms worked in many cases without further engineering, because critical improvements of fluorescent protein properties, such as removal of the cryptic intron, protein stability and codon optimization, had been established early on (Davis and Vierstra, 1998; Haseloff et al., 1997; Siemering et al., 1996). The use of the strong cauliflower mosaic virus 35S promoter for high-level GEFI expression in plants has repeatedly led to complications owing to post-transcriptional gene silencing (Daxinger et al., 2008). This issue has been addressed by the use of alternative promoters (e.g. UBQ10), or by introducing GEFIs in a silencing-deficient genetic background (Deuschle et al., 2006; Krebs et al., 2012). The cell type-specific expression of sensors can further improve sensitivity in cases where signals of surrounding cells could interfere with measurements in deeper tissue layers. This approach was successfully applied in Arabidopsis for the recording of Ca²⁺ signaling events during the double fertilization of egg cell and central cell (Denninger et al., 2014; Hamamura et al., 2014), which are hidden inside the ovule and enclosed by layers of integuments.

With regard to the applications of fluorescent indicators, these are intimately linked to the development of appropriate technical accessories that guarantee stable environmental conditions over prolonged periods. As mentioned above, different types of perfusion chambers, setups for vertical-stage microscopy, custom-made software and techniques for local stimulus applications are prerequisites for the successful application of fluorescent indicators in intact seedlings (Grossmann et al., 2011; Krebs and Schumacher, 2013; Poxson et al., 2017; von Wangenheim et al., 2017).

Fluorescent indicator applications in plants have allowed researchers to gain new insights into different aspects of plant physiology and cell biology. An example that exploits the full potential of the fluorescent indicator technology is the analysis of intracellular ATP levels (De Col et al., 2017). The applications range from ex situ functional assays on isolated mitochondria, to in vivo cytoplasmic ATP mapping on a whole-seedling scale, which provides us with an integrated view on ATP-related energy metabolism in Arabidopsis (De Col et al., 2017).

Other fluorescent indicator applications have shed new light on long-standing questions in plant biology: the acid-growth hypothesis, proposed in the 1970s (Hager et al., 1971; Rayle and Cleland, 1970), postulated that cell expansion is enabled through apoplastic acidification that is triggered by auxin. The application of different pH-sensitive fluorescent indicators in the apoplast of Arabidopsis roots and hypocotyls has now significantly strengthened this hypothesis (Barbez et al., 2017; Fendrych et al., 2016). Such targeting of fluorescent indicators to cellular compartments, such as the apoplast, underlines the importance of assessing local biophysical and biochemical cellular environments. With the discovery of the organization of protein complexes in microdomains and nanodomains, this is becoming increasingly important (Bücherl et al., 2017). A future challenge will be the precise spatial analysis (mainly along the z-axis) of cellular micro-environments that are in close proximity to the plasma membrane, as shown for a sub-population of mitochondria during plant cell infection that exhibit an altered redox state compared to more distant organelles (Fuchs et al., 2016).

Beside the requirement for increased spatial resolution, future fluorescent indicator applications will have to cover the simultaneous measurement of multiple parameters. For example, long-distance signal propagation during salt stress involves at least two types of signaling molecules: Ca²⁺ and reactive oxygen species (ROS) (Choi et al., 2014; Evans et al., 2016). It is still not clear whether salt-induced long-distance signals are solely driven by Ca²⁺ and ROS, or whether electrical signals could be involved as well (Choi et al., 2017). To gain insights into the order of events and the interdependent behavior of different signaling molecules, multiparameter imaging is required, which has been recently established in Arabidopsis to simultaneously monitor Ca²⁺ and abscisic acid (Waadt et al., 2017).

Finally, it will be important to convert qualitative fluorescent indicator readouts into quantitative data that can be linked to the biochemical properties of the molecular players that are involved in a particular signaling pathway or a physiological process. The calibration of fluorescent indicators in a multi-cellular plant context is certainly not trivial, but has been successfully reported for indicators to monitor pH, Ca²⁺ and redox state (Jiang et al., 2006; Krebs et al., 2010; Legue et al., 1997; Martinière et al., 2013b; Moseyko and Feldman, 2001; Schwarzländer et al., 2008; Waadt et al., 2017) (Table 2).

In summary, fluorescent indicator applications have become essential and established tools to study physiological and cellular processes in living plant cells. Future efforts in this field will have to focus on the advancement of imaging techniques to be able to study cellular environments at the nanoscale level. Furthermore, implementation of calibration techniques and multiplex imaging will be required to advance our understanding of complex signaling and metabolic networks in plant cells.

Förster resonance energy transfer (FRET) measurements provide a very precise and powerful way to measure protein–protein interactions in vivo with highest resolution in 3D (Förster, 1948; Lakowicz, 2006). For FRET, the two proteins of interest have to be expressed in fusion with appropriate, spectrally distinct fluorescent proteins, so-called FRET pairs (e.g. CFP/YFP, GFP/RFP or GFP/mCherry). For the necessary energy transfer from the excited donor fluorescent protein to the acceptor protein, the donor has to have a substantial overlap of its emission spectrum with the excitation spectrum of the acceptor. FRET depends on the close distance of the two fluorescent proteins to each other, which is usually in the range of 5–10 nm; this is well below the diffraction limit of a confocal microscope, and can therefore be used as an indicator for physical protein–protein interactions (Förster, 1948; Lakowicz, 2006).

One robust, minimally-invasive way to quantitatively measure FRET (apart from intensity-based measurements, such as acceptor photobleaching or ratiometric imaging) is to determine the fluorescence lifetime of the donor by employing fluorescence lifetime imaging microscopy (FLIM). The mean fluorescence lifetime τ of the molecule is defined by its average time that it remains in the excited state before it returns to its ground state, thereby emitting a photon; this usually lies in the nanosecond range. In the case of FRET, the mean τ of the donor is reduced, because energy is transferred to the acceptor, which then emits fluorescent photons. The recent overview by Weidtkamp-Peters and Stahl provides more theoretical and practical details on FRET and FRET-FLIM measurements in plants (Weidtkamp-Peters and Stahl, 2017).

Protein–protein interaction measurements through FRET-FLIM with endogenous levels of fluorescently labeled proteins are desirable and can be achieved by expressing fluorescently labeled, functional proteins by their endogenous promoters in null mutant backgrounds (e.g. in stably transformed Arabidopsis thaliana). This provides the grounds for gaining important insights into complex formations inside the relevant tissues, and even in subcellular compartments in native conditions. These studies can reveal tissue- and cell-type specific differences for protein–protein interactions and unravel the partitioning of specific protein-protein interactions and complex compositions that are necessary for the consequent biological output, such as cell specification, growth and development (Bücherl et al., 2013; Laursen et al., 2016; Long et al., 2017). Even transient expression systems, such as the leaf epidermal cells of Nicotiana benthamiana, can provide fast means of FRET measurements in plants at near-endogenous expression levels by using inducible promoters (Bleckmann et al., 2010; Somssich et al., 2015; Stahl et al., 2013).

Moreover, the use of multiparameter fluorescence imaging spectroscopy (MFIS) allows the differentiation between homomeric and heteromeric protein complexes (Weidtkamp-Peters et al., 2009) (Fig. 3). Here, the fluorescent lifetime τ and fluorescent anisotropy r are measured at the same time in every pixel of the acquired image. Thus, even dynamic measurements of the same cells are possible over time as FRET-FLIM and anisotropy measurements only require non-invasive, low-excitation powers of a pulsed laser source due to the highly sensitive single-photon counting detectors that are used (Somssich et al., 2015; Stahl et al., 2013; Weidtkamp-Peters et al., 2009).

Fluorescence anisotropy r is a fundamental property (as is the fluorescence lifetime τ) of a given fluorescent protein and describes its depolarization. The fluorescence anisotropy changes depending on the rotational freedom of the molecule. In the case of hetero-FRET (FRET between donor and acceptor, which indicates heteromeric complex formation) (Fig. 3A), the labelled proteins usually have less rotational freedom than non-interacting proteins, so the anisotropy increases (polarization). In the case of homo-FRET (FRET between donors only, which indicates homomeric complex formation), the anisotropy decreases, because of the transfer of energy to slightly differently oriented donor molecules in close proximity (depolarization) (Fig. 3B) (Bader et al., 2011; Borst and Visser, 2010). Of course, both heteromeric and homomeric complexes can occur at the same time, which leads to reduced fluorescent lifetimes and reduced fluorescent anisotropy values. This can be deduced pixel-wise and utilized to unravel differential protein complexes, and visualized with the use of 2D plots (Fig. 3B) (Somssich et al., 2015; Stahl et al., 2013). Other fluorescence correlation spectroscopy (FCS)-based techniques, such as fluorescence cross correlation spectroscopy (FCCS) or scanning FCS in combination with brightness analyses, have recently been used to investigate the mobility, oligomeric state and stoichiometry of protein complexes in plants (Clark et al., 2016; Laursen et al., 2016; Wang et al., 2015a) (see Glossary).

In summary, FRET-FLIM, anisotropy and FCS measurements have been successfully used to uncover protein–protein interactions and complex compositions in a dynamic and minimally invasive way, both in transient plant expression systems and in stable transgenic lines.

Constant efforts of multiple research groups in adapting novel imaging techniques or developing custom devices that allow non-invasive plant growth on microscope-compatible supports have greatly accelerated plant cell imaging over the last decade (Shaw and Ehrhardt, 2013). These methods now allow to explore challenging biological questions in living cells and multicellular tissues, such as spatio-temporal dynamics and organization of cellular subcompartments, the stoichiometry of multi-component protein complexes, cell and tissue plasticity on the level of proteins, metabolites and physiological parameters, image-based flux analyses and others. Whereas these tasks are not only methodologically demanding, such approaches will also help to further develop and refine existing and upcoming technologies with the long-term goal to make them available to a broad scientific community. However, with the rapid advancements in modern fluorescence microscopy, user-friendly imaging setups, and the resulting increased SNR, standard confocal systems are commonly purchased and frequently used without the necessary caution during image acquisition. This can result in a tremendous increase in published cell biological data with sometimes alarming over interpretation or misinterpretation of the presented data. Therefore, statistically sound image quantifications, together with a detailed description of any digital image processing (Jarsch and Ott, 2015) need to become an inevitable standard in plant cell biology. In addition, detailed technical descriptions and drawings for engineering, as deposited in freely accessible repositories, or as recently published (von Wangenheim et al., 2017), are essential steps to drive innovations in this rapidly evolving field.