|Home | About | Journals | Submit | Contact Us | Français|
The development of advanced optical methods has played a key role in propelling progress in neurobiology. Genetically-encoded fluorescent molecules found in nature have enabled labeling of individual neurons to study their physiology and anatomy. Here we discuss the recent use of both native and synthetic optical highlighter proteins to address key problems in neurobiology, including questions relevant to synaptic function, neuroanatomy, and the organization of neural circuits.
GFP and other conventional fluorescent proteins have had a tremendous impact on biology, and their subsequent development into optical highlighters (photoactivatable, photoconvertible, and photoswitchable fluorescent proteins) has opened even more avenues of experimentation. Respectively, these molecules are initially synthesized as molecules lacking or having low fluorescence, molecules initially fluorescing at wavelengths other than their "activated" wavelength, and molecules exhibiting fluorescence which can be switched "on" and "off" repeatedly. These features enable experiments not possible through conventional photobleaching of constitutively fluorescent proteins: by definition, photobleaching experiments track the properties of those molecules not subject to the photobleaching itself, whereas optical highlighting enables researchers to track precisely those molecule that have been photostimulated. Recounting the >20 variations of these molecules is outside the purview of this review, but interested readers are directed to reviews detailing their characteristics. [1,2]
Chemical and electrical stimulation protocols induce defined forms of plasticity in vitro whose in vivo correlates are thought to instantiate network-level processes such as learning and memory. These experimental manipulations trigger both local alterations in protein translation, localization and biochemical activity and neuron-wide changes in transcription; together these events respecify the molecular landscape within individual synapses, culminating in altered synapse anatomy and function [3–5].
The use of live imaging techniques has promised to enable visualization of the anatomic and molecular changes coincident with the induction of synaptic plasticity in real-time. However many of the techniques traditionally deployed by researchers — such as the use of GFP-fusion proteins to track single molecules — do not afford sufficient temporal or spatial resolution to effectively address questions about molecular dynamics at single synapses on the timeframes relevant to plasticity. By conferring tight control over molecular labeling both in space and in time, optical highlighters have circumvented many of the technical limitations of conventional approaches, enabling key experiments to characterize synapses and the individual molecules within them.
For example, the observation that isolated synapses on a single neuron exhibit distinct functional properties suggests that the anatomic structure of dendritic spines — whose bulbuous heads are connected to the dendritic shaft by narrow necks only a few hundred nanometers across — may act as electrical or chemical compartments, enabling the spine to act as an isolated computational unit . By using multiphoton techniques to photoactivate PA-GFP within individual spines in hippocampal slice culture and then monitoring the rate of PA-GFP exit into the dendrite, Bloodgood et al  could establish that spines are diffusionally coupled to the dendritic shaft. (Fig. 1) Further, by manipulating levels of electrical activity in the slice and then photoactivating PA-GFP within spines it was shown that degree of effective continuity between spines and dendrites could be dynamically altered .
The observation of regulated diffusional coupling between the spine head and the dendrite suggests that molecular mediators of plasticity may be differentially trapped within or released from spines in response to trans-synaptic cues. As PA-GFP diffusion may not reflect the biophysics of native spine-localized molecules, researchers have measured the diffusion constants of various synaptic molecules fused with optical highlighters. Although the molecular weight of fusion proteins significantly differs from that of the native proteins, diffusion coefficients are only weakly related to total molecular weight (approx. MW1/3), enabling fusion proteins to provide an adequate estimate. Several groups have fused PA-GFP to PSD-95, a core protein within the postsynaptic density (PSD) whose local concentration may scale with synapse size and strength. [9–13] Photoactivation of PSD-95-PA-GFP in spines in vivo has revealed that PSD-95 leaves unstimulated synpases on timescales of 10s of minutes to hours (instead of seconds as measured for free PA-GFP) and that this rate increases in response to activity. Highlighter proteins have also been fused with other adapter proteins, enzymes and ion channels resident in synaptic spines, including Shank2, Shank3, CaMKIIα, CaMKIIβ, GluR2, stargazin, Ras, Rho, and Cdc42 [10,14–17] revealing a diversity of diffusion rates that constrain molecular models of plasticity. These experiments have also shown that the spine concentration of certain molecules (PSD-95, CamKII) scales with spine size. Importantly, while protein diffusion may facilitate the spread of signals from activated spines to adjacent spines (as is the case for PSD-95), in at least some cases diffusion may also promote enzyme inactivation. For example, PA-GFP-tagged cdc42 diffuses rapidly out of activated spines, but (as revealed by 2P-FLIM) its enzymatic activity is disabled upon entry into the dendrite .
Optical highlighters have also been used to probe actin dynamics within spines, where actin plays a critical role in synaptic function by acting both as a tether (through interactions with signaling molecules) and as a strut. Expression and photoactivation within spines of PA-GFP-β-actin fusion proteins in hippocampal neurons revealed two populations of filamentous F-actin: a treadmilling pool of actin that flows from the spine periphery towards the center, and a stable pool of actin at the spine base whose size is proportional to the size of the spine itself [12,13]. Stimulation via single spine uncaging of MNI-glutamate causes formation of a third unlocalized pool of actin that may play a role in structural plasticity . Photoconvertible protein actin chimeras have also been used in a super-resolution microscopy technique, Photoactivatable Localization Microscopy (PALM) . PALM and other molecular localization microscopy techniques, such as Fluorescence-PALM , Stochastic Optical Reconstruction Microscopy (STORM) , PALM with independent running acquisition (PALMIRA)  and many others , rely on the precise localization of single molecules by imaging and fitting their fluorescence signals to two-dimensional Gaussian distributions. These experiments revealed that most actin fibers are short (<300 nm) and that the most dynamic fibers are irregularly distributed within the spine in numbers proportional to spine size [13,24]. As an alternative to direct actin tagging, Izeddin et al  developed an actin binding protein (ABP)-tdEosFP fusion for PALM with which they have visualized actin redistribution after exposure to AMPA agonists in hippocampal neurons .
Like their postsynaptic counterparts, presynaptic boutons have been proposed to serve as compartments that differentially isolate second messengers and specific pools of synaptic vesicles. By expressing and locally photoactivating PA-GFP fused to presynaptic active zone components, an active exchange of presynaptic material between adjacent boutons was observed; the exchange of certain molecules (such as synapsin) was promoted by activity, whereas the exchange of other molecules (such as bassoon) was not [17,26]. Similar synaptophysin-Dendra2 experiments demonstrated that presynaptic vesicles decorated with synaptophysin are also shared in “packets” between adjacent presynaptic boutons . These results indicate that local exchange of multiple components occurs within presynaptic structures, and that optical highlighters afford sufficient temporal and spatial resolution to track this process.
In addition to enabling molecule tracking at the synapse, optical highlighters can enable real-time tracking of de novo protein synthesis. Certain forms of synaptic plasticity likely require the local translation of specific mRNAs into protein via synthetic machinery located within the dendrite [28,29]. Although pharmacological and molecular studies (often utilizing pulse-chase approaches) have strongly suggested that protein translation both occurs locally and may be causally involved in synaptic plasticity, visualization of this process has been difficult. The generation of translational reporters consisting of the cDNAs for two-color photoconvertible proteins fused to the UTRs of translationally-regulated messages has facilitated the identification of both the signaling pathways that impinge upon the local translational machinery and the cis-acting elements that confer translational regulation upon specific messages. Typically, in these experiments a photoconvertible fluor such as Kaede is converted to its active red color, and then the stimulus-induced rate of translation of the reporter is revealed as de novo green fluorescence. This strategy has been used to examine the dendrite-specific translation of Kv1.1, CaMKII and Lypla1, and the axon-specific translation of β-actin [30–37]. Recently Martin and colleagues used a Sensorin-Dendra2 reporter to test synapse-specific mRNA translation in a culture system in which single, defined synapses between Aplysia sensory and motor neurons can be both manipulated and imaged . These experiments demonstrated that stimuli known to initiate facilitation (such as five pulses of serotonin) trigger synapse-specific translation, whereas stimuli that are ineffective (such as a single pulse of serotonin, or a pulse of FMRFamide) fail to generate new protein. It has also recently been observed that certain forms of synaptic plasticity (such as homeostatic plasticity ) may require targeted protein degradation; because photoconversion allows photoconvertible proteins to be used as optical pulse-chase reagents, these molecules can track stimulus-induced protein destruction. For example, degradation reporters have been generated in which PA-GFP is fused to a variety of protein motifs that can target peptides to the proteasome, and used to show that the chronic manipulation of activity can alter dendritic protein turnover [34,35].
Optical highlighters have also been used to reveal mechanisms of new gene expression, which is thought to be essential to the induction of synaptic plasticity over long time scales[36,37]. This process canonically involves recruitment of cellular signal transduction molecules that culminate in the regulation of transcription factors in the nucleus. However it has recently been suggested that some transcription factors are held in abeyance in the dendrite, but that upon stimulation these factors are released and imported into the nucleus where they alter gene expression to promote synaptic modification . For example in the Aplysia system the transcription factor CREB2 has been suggested to translocate from dendrites to the nucleus, where in response to specific types of synaptic stimulation it initiates new gene expression required for long term depression. By tagging CREB2 with Dendra2, Lai et al  demonstrated that CREB2 translocates to the nucleus in response to FMRFamide in an importin-dependent manner . Similarly, tagging of the p65 subunit of NF-kB with PA-GFP has been used to demonstrate that dendritically-localized NF-kB is actively transported from the synapse to the nucleus .
The long distances between the soma and the tips of an axon present challenges to the cell in moving proteins to and from these distal points. In addition to translating mRNAs at peripheral sites (as discussed earlier), neurons deliver proteins to axons via least three other mechanisms, each of which has been clarified through the use of optical highlighter-cargo fusion proteins. Neurofilament fusion proteins, for instance, have been shown to move along microtubules  and rely on myosin Va to keep them on microtubule tracks.  Fusions of at least two soluble cytosolic proteins, synapsin and CamKII, have been found to be transported via a microtubule-based mechanism in which large multimeric complexes interact with motor proteins.  Diffusion can also be a primary mode of transport. For instance, highlightable fusions of the microtubule binding protein tau undergo rapid binding to and unbinding from microtubules during transit to the end of an axon, with transport kinetics indicative of free diffusion to the tip of the neurite. [46,47] Fusions of even slowly diffusing transmembrane proteins, such as the beta-adrenoceptor  and voltage-gated potassium channel isoforms 1.3, 1.4, and 2.1 , have been shown to move by this mode. These studies of diffusive mechanisms have particularly benefited from using highlighting methods since measurements can be made at many distal points in the cell in addition to the irradiation spot. Consequently, mobility fraction and diffusion coefficient maps over large areas of the membrane and various axon segments can be easily established.
Finally, optical highlighters are also well suited for studying trafficking in the secretory pathway. Since new proteins are often continuously produced over the course of an experiment, photoactivated or photoconverted pools effectively provide a time stamp for a given population of molecules. Transport studies of UNC-2, the C. elegans voltage-gated calcium channel, CaV2, found that a newly discovered endoplasmic reticulum (ER) localized chaperone, CALF-1, is required in conjunction with a calcium channel subunit, alpha2 delta, for proper exit of UNC-2 from the ER.  Photoconverted (and thus time stamped) Dendra2-tagged UNC-2 exhibited transport from the ER to synapses, but only in the presence of calf-1, indicating that preexisting protein within the ER exits through interactions with this newly-described chaperone.
Mapping the fates of individual cells or small numbers of cells during development requires the ability to introduce a label into a sparse number of cells at a specific point in time. Traditionally this has been either achieved through laborious single cell methods like DNA electroporation, through the use of genetic chimeras, or through clever genetic approaches such as MARCM and MADM . However these physical or genetic methods are often impractical or impossible (i.e. because of physical access issues or the absence of effective mitotic recombination in the chosen model organism). The ability to generate focal signals with high contrast ratios through precision photoactivation — particularly via multiphoton methods — potentially circumvents these limitations, and has led to the widespread use of optical highlighters to track cell fates during development. This approach was first used to track the fate of neural crest precursors in chick embryos; chicks were electroporated in ovo with plasmids expressing PA-GFP under the control of a ubiquitous promoter, and then single neural crest cells (NCCs, located on the dorsal surface of the neural plate) were photoactivated and imaged .
Since this initial demonstration, a number of different highlighters have been successfully deployed in the chick [52–57], zebrafish [58–62], xenopus , ciona intestinalis  and in mammals [52,54,57,61,63,65,66] to track the fates of neural precursors during development. These experiments have correlated the initial positions of neural precursors with the final positions of differentiated neurons [52,54,57,61], visualized interactions between migrating neurons , assessed proliferation during migration [57,65], followed the fate of cells during metamorphosis  and tested the role of individual signaling molecules and transcription factors in cell migration processes [54,63,66]. By restricting the pattern of optical highlighter expression through the use of cell-type specific promoters researchers can precisely characterize the tracked neuron's function. For example, interneuron subtype-specific promoters driving Kaede expression in zebrafish have been used to address the relative role of CiD and MCoD interneuron types in fast and slow forms of swimming, respectively. . Furthermore, a combinatorial technique called BAPTISM, which combines promoter-specific expression of a photoconvertible fluorescent protein (huc::kaede, for example) with EGFP expression under the control of a second cell-type specific promoter (p2x3b::egfp and trpa1b::egfp, for examples), has enabled researchers to birthdate multiple different cell types was enabled within a specific pool of neurons (Fig. 2) .
Freely-diffusing optical highlighters not targeted to specific subcellular compartments have been used to great advantage in measuring diffusional coupling between spines and dendrites (see above). To a first approximation, highlighted molecules within other subcompartments of the neuron also have free access to the rest of the neuron. Because the anatomy of the neuron is regionally specialized based upon connectivity — with input-receiving dendrites spatially segregated from output-transmitting axons —researchers have used photoactivation methods to focally activate a protein within the dendrite or axon and then utilize its de novo fluorescence to visualize (via diffusion) the contiguous cell bodies, axonal arbors, and dendrites. While labeling is restricted to contiguous structures within a given neuron, the complete morphology of targeted neurons and potentially the direction and nature of information flow (at least as constrained by anatomy) within a neural circuit can be revealed with this approach.
This strategy was first deployed in a paper in which Kaede was expressed under the control of the neural HuC promoter in zebrafish followed by focal photoconversion to reveal the complete anatomy of targeted trigeminal neurons and Rohan-Beard cells in the ventral spinal cord . Aramaki et al  performed a similar experiment in zebrafish, but with a twist: by using the “rewritable” photoswitchable fluor Dronpa (expressed in neurons using the Gal4-UAS system), these researchers were able to interrogate a targeted neuron’s anatomy, erase the fluorescence, and then target and image a second neuron (and so on with multiple neurons in series) . In this iterative manner, structural features of a number different neurons presumed to be within a single neural network were identified.
While these two proof-of-principle papers demonstrated that optical highlighters could be used as long-range neural tracers, subsequent papers in the fruit fly Drosophila melanogaster illustrate the power of this approach to dissect the structure and function of behaviorally-relevant neural circuits. One such circuit is responsible for sexually-dimorphic behaviors in response to the pheromone cis-vaccenyl acetate (cVA), which (in particular contexts) elicits aversive behaviors in males but is attractive to females . To address whether differences in downstream circuits are responsible for sexually dimorphic responses to cVA, Datta et al  expressed PA-GFP in most projection neurons in the antennal lobe, and then took advantage of the spatial specificity afforded by multiphoton techniques to photoactivate the PA-GFP specifically within the DA1 glomerulus; the PA-GFP then diffused, labeling the connected cell bodies and revealing sexual dimorphism in the DA1 projection neuron axonal arbors (Fig. 3) . Furthermore, the induced fluorescence enabled electrophysiological characterization of DA1 projection neurons via cell-attached recordings.
Because this approach depends upon the inherent diffusional characteristics of PA-GFP, the total neurite length that can be traced is limited, and effective tracing requires photoactivation of a significant amount of PA-GFP (through repeated cycles of activation over relatively large target regions). These constraints have been partially circumvented through the development of two variants of PA-GFP (C3PA and SPA) that exhibit improved apparent rates of diffusion in vivo . By expressing C3PA and SPA in neurons potentially connected in a circuit, Ruta et al  traced a single neuron and identified its axonal arbor, and then irradiated the newly-labeled axonal arbor, thereby taking advantage of the spatial overlap with the connected dendritic spine to highlight the downstream neuron. By iteratively repeating this process, a behaviorally-relevant neural circuit was traced across multiple synapses and through multiple neurons .
As work in neural circuits becomes increasingly focused on interrogating the role of individual neurons in driving behaviors, researchers are finding new roles for optical highlighter proteins in disambiguating functionally-relevant neurons from bystander cells. For example, a common difficulty is that the genetic drivers that are used to functionally manipulate the activity of neurons promote gene expression within large numbers of cells, making it difficult to identify the particular neurons responsible for a specific behavior. Claridge-Chang and colleagues used optical highlighters to identify which subset of cells labeled by their driver was relevant to behavior on the basis of connectivity revealed by targeted photoactivation of specific neurons . Similarly, a number of researchers have begun co-expressing various photoconvertible proteins along with optogenetic reagents; optical stimuli then both trigger changes in electrical activity via the light-gated channel, and photoconvert and label the stimulated cells, enabling verification of their identity [72–74]. Photoconvertible fluors are thus proving to be useful adjuncts in optogenetic experiments aimed at testing neural function.
Similar to mapping circuits, the advantages of an optical highlighter become apparent when determining organelle continuities as well as both intracellular and intercellular continuities. Axonal regeneration  has been confirmed by photoconverting freely diffusing protein either in the cell body or distal axon and monitoring the presence of red and green signal mixtures. Similarly, continuity between cells has been established by observing photoconverted KikGR move bidirectionally into neighboring cells within the neural crest of chick embryos via cytoplasmic bridges.  Alternatively, axon degeneration has been confirmed by highlighting fragmented axon segments to ensure that subsequent re-fusion of fragments did not take place. 
Some of the experiments discussed here represent approaches that are difficult or impossible with conventional fluorescent protein imaging, demonstrating that optical highlighters have moved from being neat little fluorescent protein tricks to bona fide neurobiology tools. Yet, the challenges of imaging in the nervous system will likely require further optimization of these molecules, with emphasis on brightness, contrast, photostability, monomeric behavior, red-shifted wavelengths, improved folding efficiency, faster folding kinetics, and higher photoactivation quantum efficiency. Improvements in any and all of these parameters will assist in neurobiology experiments that seem to be delving deeper into tissue, imaging with faster kinetics, and observing over longer timescales. If history provides any reasonable forecast, the rapid development of highlighter molecules over the past 10 years is likely to continue and provide tools to facilitate new and innovative approaches.
This work was funded in part by the Intramural Research Program of the National Institutes of Health including the National Institute of Biomedical Imaging and Bioengineering (GHP), and by grants from the National Institutes of Deafness and Communication Disorders (RO1DC011558), the NIH Office of the Director’s New Innovator Program (DP2OD007109), the McKnight Foundation, the Searle Foundation and from a Career Award in Biomedical Sciences from the Burroughs Wellcome Foundation (SRD).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
* paper of special interest
** paper of outstanding interest