The rise of photoresponsive protein technologies applications in vivo: a spotlight on zebrafish developmental and cell biology

Color coding cells

One major challenge in developmental biology is to visualize cellular organization in complex tissues; a related challenge is to follow changes in cell morphology and movement over time and to perform lineage tracing and understand the origin of cells and cell populations. Driving fluorescent proteins under certain promoters can restrict fluorescent protein expression to specific cell subpopulations, but the labeling is often too dense to achieve single-cell resolution when one is studying complex tissues, and sometimes there are no available promoters that are specifically expressed in the cells of interest. A decade ago, researchers came up with a cell-labeling technique to distinguish individual neurons in the mouse brain. The technique, called Brainbow, uses stochastic Cre-loxP recombination to express one of several spectrally distinct fluorescent proteins from a single transgene. Since multiple cassettes are integrated at a single genomic site, and the choice within each cassette is made independently, varied combinations of the different fluorescent proteins label individual cells with distinct fluorescent profiles and generate what is effectively a multicolored Golgi stain³⁰. Brainbow constructs were first tested in zebrafish in 2011³¹, and several transgenic lines incorporating Brainbow technology, including the Zebrabow series³² and Tg(UAS:brainbow)³³, soon followed (Table 1). The Brainbow strategy has been adopted for cell tracking in several zebrafish tissues to great success. For example, Brainbow technology allowed clonal analysis of the zebrafish heart and revealed how the orchestrated division of cardiomyocytes results in the distinct architectures of the ventricle and atrium^34,35 and how endocardial cell movements contribute to cardiac valve formation³⁶. It also allowed the identification and quantification of collective cell behaviors, enabling zebrafish skin homeostatic maintenance and response to injury³⁷ and the precise manner by which zebrafish retinal ganglion cell axons laminate in the tectum³³.

In the original Brainbow design, the expression levels of fluorescent proteins do not follow an even distribution, thus reducing the randomness of fluorescent protein expression and increasing the likelihood that neighboring cells express indistinguishable spectral profiles. Other disadvantages include not only the possibility of changing the color code of the cell because of permanent CRE expression but also the sensitivity of the technique to factors that affect fluorescent protein signal intensity, such as promoter activity, cell depth, or autofluorescence. Recently, there have been efforts to overcome these problems by using independent transgenes for each fluorescent protein, each equipped with a binary ON/OFF switch. In this approach, called Multibow, fluorescent proteins are initially OFF and then probabilistically ON or OFF following Cre recombination³⁸. A zebrafish line generated using this technique takes advantage of our increasing repertoire of fluorescent proteins, expressing no fewer than seven different fluorescent proteins, which have been tagged to three different subcellular localizations (membrane, cytoplasm, and nucleus) to increase variation of labeling³⁸. Although Multibow has not been used for any biological applications yet, it is an interesting new addition to the multicolor labeling toolbox. Given that membrane, cytoplasmic, and nuclear compartments show large variations in fluorescent protein signal intensity, it can be difficult to distinguish the color of one or more of these compartments³⁸. For example, in some cases, the nucleus of the cell is much more brightly labeled than the cytoplasm or membrane, making it difficult to visualize cellular morphology³⁸. Nonetheless, Multibow may prove useful, especially for cell lineage studies where the use of time-lapse imaging is not possible.

A caveat of current multicolor labeling technologies for live embryo imaging is the necessity to image embryos multiple times, including using violet light for the excitation of blue fluorescent proteins, which can cause phototoxicity. This problem is partially elevated by improved microscopy techniques that require fewer excitation photons to achieve the same resolution. A particularly notable milestone in recent years is the adaptation of light-sheet microscopy for the imaging of live embryos^39,40. Light-sheet imaging combines high imaging speed with low light exposure, which, when combined with the optical transparency of zebrafish and standard fluorescent protein labeling techniques, enables many developmental processes to be imaged at unprecedented spatial and temporal resolution. Also significant is the introduction of new point-scanning microscopy setups that apply wavelength mixing to provide simultaneous, efficient, and independent two-photon excitation of more than two spectrally distinct fluorophores, thereby avoiding the use of shorter, more phototoxic wavelengths⁴¹. Another deterrent to using Brainbow technology for cell lineage tracing studies is the time lag between when CRE is first delivered/expressed and when fluorescent proteins mature. An alternative strategy that does not have this limitation is to follow optically highlighted cells, as discussed in the section below.

Highlighting single cells with light

Photoconvertible fluorescent proteins change emission wavelengths in response to irradiation with light of a particular wavelength, whereas photoactivatable fluorescent proteins gain fluorescence after irradiation of light of a particular wavelength (Table 2). Together, these proteins serve as optical highlighters, allowing the rapid and precise labeling of biological structures using laser light. The first fluorescent proteins advertised as being photoconvertible and photoactivatable were Kaede and photoactivatable GFP (PA-GFP), respectively. Kaede’s photoconvertible properties were discovered serendipitously when the researchers left a vial of the protein overnight by the window and discovered the next day that sunlight had converted the sample from green to red; Kaede is named after the leaves of the Japanese maple that change from green to red in the autumn. Further investigation revealed that the red form of Kaede reflects the protein’s ionized state and that this change in emission wavelength can be effectively and irreversibly induced by illuminating the protein with violet 405-nm light⁴². By contrast, PA-GFP was created purposely by modifying the structure of GFP through mutagenesis so that the protein would greatly increase its fluorescence after irradiation using violet 405-nm light⁴³. A few years after the introduction of Kaede and PA-GFP, Dronpa, a fluorescent protein that was not only photoactivatable but photoswitchable, was discovered. The fluorescence of this interesting protein can be switched off by blue 488-nm light and switched on again by violet 405-nm light exposure⁴⁴. Since then, a number of new optical highlighters have been developed, and significant effort has been made to create monomeric versions that perform better in protein fusion studies (reviewed in 45). Optical highlighters have been used in zebrafish to visualize single-cell morphology^46,47; to track cells for cell lineage, migration, or proliferation analyses^48–52; and to mark cells for laser ablation⁵³ (Table 2).

Recently, there have been two exciting developments relating to photoconvertible proteins. The first development is the surprising discovery that the green-to-red photoconvertible protein Dendra2 can be “primed converted”. It turns out that rather than using violet light, one can photoconvert Dendra2 by irradiating the protein first with blue light and then with near-infrared light⁵⁴. The mechanism responsible for this unusual phenomenon is still unknown, but its usefulness soon became evident. Firstly, infrared light and blue light are less toxic than violet light and penetrate deeper into tissue. Secondly, the dual wavelength mechanism overcomes one of the caveats of green-to-red photoconvertible proteins: they cannot be efficiently photoconverted using multiphoton illumination^55,56. Whereas one-photon illumination is not confined in the axial direction (Figure 1A), primed conversion (Figure 1B, 1C) can spatially confine the photoconverted region in all three dimensions by allowing both the priming and the converting beam to meet only at a small axially confined focal volume⁵⁶. Importantly, primed conversion requires much lower peak power than two-photon photomodulation to accomplish confinement⁵⁶. The effectiveness of this technique was demonstrated by highlighting neuronal morphology at single-cell resolution in the densely labeled zebrafish brain⁵⁶. Primed conversion of Dendra2 was also shown to be possible using light from red lasers instead of the more uncommon and expensive near-infrared lasers, albeit at much lower efficiencies⁵⁷. Given that many transgenic lines expressing Dendra2—including PhOTO zebrafish (photoconvertible optical tracking of zebrafish) lines that ubiquitously label nuclear or plasma membrane proteins using Dendra2⁵¹—are readily available, primed conversion promises to be a useful technique for future studies.

Figure 1. Primed photoconversion.

(A) Traditional photoconversion experiments use green-to-red photoconvertible proteins that change from emission after being converted with violet 405-nm light. It is often difficult to specifically photoconvert single cells because the laser beam is not confined in the axial direction. (B) Primed conversion involves first priming green-to-red photoconvertible proteins with blue 488-nm light, before converting the protein with near infra-red 730-nm light. Hence, only cells at the intersection of the two laser beams are photoconverted. (C) Left and middle: Primed photoconversion of Dendra2 in a single neuron in the trigeminal ganglion of a zebrafish at 3 days post-fertilization (maximum-intensity projection of about 82 μm in depth). Right: Higher-magnification axial orientation images of the boxed region in the left frame. Arrowhead indicates the photoconverted cell, and the arrow indicates a neurite extending from the cell body. Asterisks indicate cells that had noticeable signal in the red Dendra2 channel even before photoconversion. Reprinted with permission from Macmillan Publishers Ltd⁵⁶.

The second development is the long-awaited arrival of new photoconvertible proteins with red-shifted excitation and emission wavelengths, such as PSmOrange and its improved version PSmOrange2^58,59. These proteins have many advantages over traditional green-to-red fluorescent proteins, not the least of which is that they can be easily used in conjunction with the many available zebrafish lines that already use the green emission wavelength to label biological structures. Another advantage is that the blue-green light needed for one-photon excitation of the protein is less toxic to cells than violet light. One study suggests that PSmOrange2 can also be efficiently photoconverted using infrared light from a two-photon laser⁵⁹, raising the possibility of spatially confined conversion in tissues using two-photon laser light, although this study was performed in cell culture and the two-photon conversion of PSmOrange2 in vivo remains to be verified. Although PSmOrange2 has not been tested in zebrafish, one study reported the injection of H2B-PSmOrange mRNA into one-cell stage zebrafish embryos to track photoconverted cells during ventral habenulae development⁴⁸. Given the study’s success, we expect PSmOrange and its variants to be much more widely used in the future.

In addition to classic uses of optical highlighters, we expect great applications of single-cell photoconversion and photoactivation in the field of gene expression profiling with the recent development of single-cell genomics. Recent experiments prove its feasibility in vivo⁶⁰ and promise important discoveries in the field of molecular cellular identity and clonal analysis.

Genetically encoded calcium indicators

Genetically encoded calcium indicators (GECIs) have become essential tools to study calcium signaling in a cell type- or cell compartment-specific manner. Since calcium is a second messenger for neurotransmitter reception and membrane depolarization in neurons, GECIs have been especially useful in the study of neuronal function. Most GECIs fall into one of two main groups: Förster resonance energy transfer (FRET)-based indicators, which are composed of two fluorescent proteins linked by a Ca²⁺-responsive element, such as the cameleons, and those based on a single circularly permutated fluorescent protein and a Ca²⁺-responsive element, such as the GCaMPs. Improving the performance of these GECIs has sustained research interest over the last 17 years, and at present the state-of-the-art GCaMP6 series outstrip the performance of the most commonly used synthetic calcium dyes^61–63. Expressed in cultured neurons, the GECIs now possess the sensitivity and kinetics fast enough to detect single action potentials as well as peak fluorescence intensity levels comparable to that of the parental circularly permuted EGFP itself⁶¹.

Zebrafish was the first vertebrate model system where GECI expression was used to visualize neuronal activity in vivo⁶⁴. Today, the use of GECIs has been combined with light-sheet microscopy to visualize neuronal activity in the entire zebrafish brain^7,65–69. Our ability to visualize neuronal activity in whole zebrafish brains has improved dramatically with the arrival of new light-sheet microscope technology, most notably IsoView microscopy⁶⁸, which generates images of greater spatial resolution by performing fluorescence detection along four orthogonal directions, and two-photon light-sheet microscopy⁶⁹, which uses light of wavelength greater than that in the visible spectrum to stimulate GECIs and thus has greater tissue penetration and prevents unwanted visually evoked neuronal activity during imaging. Also significant is the introduction of a new zebrafish pigmentation triple mutant, crystal, that allows unprecedented optical access to the retina and the brain⁷.

Beyond studying neuronal function, GECI expression in zebrafish has been instrumental in studying the diverse roles of calcium signaling during development, particularly during cardiovascular development. For example, GECIs were used to identify and analyze stages of cardiac conduction that correspond to cellular and anatomical changes of the developing zebrafish heart and, in combination with a forward genetic screen, to identify conduction-specific mutations⁷⁰. Recently, GECIs were used to characterize heart conduction defects in a zebrafish model of desminopathy, a type of muscle disorder⁷¹, and to show that calcium signaling is critically involved in cilia-mediated blood flow sensing⁷² and sprouting angiogenesis⁷³ in endothelial cells during early vascular morphogenesis.

Excitingly, the combined improvements in GECI brightness and microscopy have opened up the possibility to study calcium signaling in a cell compartment-specific manner in live zebrafish embryos. For example, SyGCaMP2, a synaptic vesicle calcium reporter that consists of GCaMP2 fused to synaptophysin, has allowed both the identification of synapse localization and the reporting of electrical activity in the zebrafish optic tectum and retina⁷⁴; migrating zebrafish neutrophils have been shown to display enriched leading-edge calcium flux, a finding in contrast to ex vivo studies⁷⁵; zebrafish myocytes were found to show sustained increase in mitochondrial calcium levels during spontaneous muscle contraction⁷⁶; and calcium oscillations within cilia were found to be required for asymmetric Nodal signaling in the zebrafish left-right organizer during left-right symmetry breaking⁷⁷.

Another exciting development is the introduction of red GECIs, which facilitate experiments that require dual-color imaging^78–81. The latest red GECIs—jRCaMP1a,b⁷⁸, jRGECO1a⁷⁸, and R-CaMP2⁷⁹—approach the performance of GCaMP6. The makers of jRCaMP1a,b and jRGECO1a tested the performance of six red GECIs in zebrafish, and their results suggest that jRGECO1a shows the greatest sensitivity in the animal but that jRCaMP1b shows the fastest on- and off-kinetics⁷⁸. The authors also advised that jRCaMP1b may be more suited to many optogenetic experiments than jRGECO1a since it does not photoswitch in response to blue light⁷⁸.

With the advancement in photoactivatable and photoconvertible protein technology (see earlier section), photoactivatable and photoconvertible GECIs are now also available, allowing optical selection of cells/cell compartments for calcium imaging^82–85. A particularly interesting photoconvertible GECI, called CaMPARI (calcium-modulated photoactivatable ratiometric integrator), undergoes efficient green-to-red photoconversion by violet 405-nm light when calcium levels are high⁸⁶ (Figure 2A). Although both red and green forms of CaMPARI respond dynamically to calcium in a manner similar to other GECIs, its uniqueness lies in its ability to retroactively report calcium activity over defined periods of time. For example, by illuminating brains expressing CaMPARI for a few hours with violet 405-nm light, neurons that experience higher levels of activity become red fluorescent while neurons with low levels of activity remain green fluorescent. The permanent labeling alleviates the need to image the right cells at the right time to observe neuronal activity. The expression of CaMPARI in zebrafish has allowed the visualization of integrated neuronal activity in zebrafish subjected to different conditions—anesthetized, swimming freely, treated with seizure-inducing compounds, or put in extreme heat or cold (Figure 2B)—and the technique promises a number of potential uses, ranging from functional connectomics to transcriptional profiling based on cellular calcium signaling levels.

Figure 2. Calcium-modulated photoactivatable ratiometric integrator (CaMPARI).

(A) Schematic of CaMPARI function. Both green and red forms of CaMPARI decrease fluorescence in the presence of calcium. CaMPARI is more susceptible to photoconversion by violet 405-nm light when intracellular calcium levels are high and this is most likely due to conformational change of the protein when bound to calcium. (B) Zebrafish larvae after 10 seconds of 405-nm light applied during exposure to different conditions. Each image represents a maximum-intensity Z-projection of CaMPARI signal in a confocal stack through a 4- to 5-days post-fertilization larval zebrafish brain expressing CaMPARI pan-neuronally under the elavl3 promoter. Reprinted with permission from the American Association for the Advancement of Science⁸⁶.

Protein timers

Traditional molecular biology and biochemistry have provided us with the identity of key proteins and regulatory networks involved in many developmental processes. However, the spatiotemporal dynamics of their regulation remains poorly understood, and imaging the proteome in vivo remains a significant challenge. Methods that have been used to monitor the behavior of proteins in zebrafish embryos include fluorescence recovery after photobleaching (FRAP) and its related technologies (fluorescence loss in photobleaching, fluorescence loss after photoactivation, and fluorescence localization after photobleaching)^87–93 and fluorescence correlation spectroscopy^91,94–98. Recently, a new method has been developed to image protein turnover, based on the use of tandem fluorescent protein timers (tFTs) that change their emission wavelengths as the protein ages⁹⁹.

Using fluorescent proteins that change emission wavelength as the protein ages itself is not a new idea. However, previous versions of fluorescent protein timers consist of a single fluorescent protein that switches color over time as its fluorophore undergoes successive chemical reactions^100–102. These protein timers have low brightness and a tendency to form oligomers, which perturbs the behavior of the tagged proteins, and to our knowledge have not been used for in vivo studies. In this context, Knop et al. created a new type of fluorescent protein timer, the tFT, which involves a tandem fusion of a fast maturing protein of one color and a slower maturing protein of another color¹⁰³. The idea is that by measuring the ratio of one color to the other, one could track protein movement within the cell; by using mathematical models that incorporate the known maturation kinetics of the fluorescent proteins, one could infer the protein turnover rate constants.

The first study featuring tFTs fused superfolder GFP (sfGFP), which reaches its peak intensity very quickly, with mCherry, which matures more slowly and reaches its peak intensity a few hours later. This tFT was then used to uncover the stable nature and asymmetric inheritance of nuclear pore complexes and identify regulators of N-end rule-mediated protein degradation in yeast¹⁰³. tFTs were introduced to the zebrafish model via two elegant studies investigating the dynamics of chemokines and cell-cell adhesion in the zebrafish lateral line primordium. In the first study, the authors tagged Cxcr4b, the receptor of chemokine Cxc112a, with sfGFP and the slower-maturing fluorescent protein TagRFP. The red/green fluorescence ratio at the plasma membrane was used as an indicator of the age of receptor populations, and the rate of ligand-triggered internalization was used to measure Cxc112a chemokine levels. By doing so, the authors effectively showed that a group of cells migrating collectively can self-generate gradients of chemokine activity across its length via polarized receptor-mediated internalization¹⁰⁴. This study has established that tFTs can be sensitive, reliable tools for visualizing protein dynamics over many hours in live zebrafish embryos. Application of protein timers is broad, and other than the zebrafish line primordium^104–106, protein timers have been used to study protein relocalization and stabilization during cardiac development¹⁰⁷ and somite morphogenesis¹⁰⁸.

It must be emphasized that tFTs are not neutral tags, and careful selection of fluorophores for use in tFTs has to be tailored to the expected turnover kinetics of the endogenous protein of interest. Given the wide range of fluorescent proteins with different spectral properties, and different maturation and degradation kinetics available, choosing the best combination of fluorescent proteins may be daunting. Recently, a study compared the performance of different GFPs within tFTs in yeast and provided comprehensive advice on the effects of the stability of the GFP fold on tFT behavior¹⁰⁹. Mathematical models of timer kinetics have been developed and were verified by applying a set of timers to once again investigate chemokine signaling gradients across the zebrafish posterior lateral line primordium. Interestingly, one of their findings was that FRET between the fast- and slow-maturing fluorophores increases timer signal, a point that may be useful for future timer designs. An interactive web application based on the models, called TimerQuant, has been made freely available to guide experimentalists on which fluorescent proteins to choose, to interpret timer readouts, and to detect differences in protein half-life¹¹⁰.

The relative ease of use of tFTs compared with fluorescence correlation spectroscopy and FRAP-based methods—including the ability to capture temporal information in a single snapshot, the possibility to perform experiments with relatively simple microscopy setups, and the availability of open source software—will likely mean that tFTs will become much more widely used in the future.

Protease activity reporters

Proteases are enzymes that perform hydrolysis of peptide bonds and are important in many biological and developmental processes through their ability to activate or inactivate their target proteins. Biosensors for monitoring protease activity typically involve a peptide specific for the protease of interest flanked by a pair of fluorescent proteins that undergo FRET. When the peptide is cleaved, the two fluorescent proteins separate and the FRET signal disappears. However, the signal of FRET reporters is weak because of small fluorescence change in the donor and acceptor fluorophores and the use of these reporters in vivo is limited.

Recently, Shu et al. introduced ZipGFP, a GFP-based protease reporter that achieves about 10-fold fluorescence increase upon caspase activation¹¹¹. To create ZipGFP, the researchers exploited the fact that GFP can be split into two fragments that have sufficient affinity to self-reassemble and form the complete, fluorescent β-barrel protein¹¹²: the two parts of GFP were each flanked with heterodimerizing coiled coils, which prevent the two parts from self-assembling; the consensus protease cleavage sequence for caspase-3 was incorporated into the peptides linking the heterodimerizing coiled coils and the GFP fragments, such that when cleaved by caspase-3, the two parts become uncaged and free to bind. The researchers injected mRNA encoding ZipGFP into live zebrafish to visualize caspase activity and apoptosis of the vertebrate in vivo. They found that the spatial pattern of ZipGFP fluorescence at 2 days post fertilization was consistent with a previous study based on terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining and further showed that apoptosis first occurred at the rostral part of the forebrain before appearing in other parts of the brain and the trunk, a finding that would have been extremely difficult to observe with the limited spatial resolution of TUNEL staining using fixed samples. The design of ZipGFP will likely be used to create reporters of many other proteases, allowing us to study protease dynamics with unprecedented spatiotemporal resolution.