Discriminating between individual neurons is essential in any attempt to analyze a neural network, as many axons and dendrites intermingle to fasciculate, making synapses as well as glomeruli. During live imaging in particular, it is often impossible to distinguish between individual neurons when they are labeled in the same fluorescent color and are connected or partially overlapped. Using the green-to-red photoconvertible fluorescent protein Kaede (Ando et al., 2002), it is possible to distinguish individual neurons from each other in the living zebrafish (K. Hatta and H. Tsujii, unpublished observation). The neurons expressing Kaede were initially fluorescent green, but by irradiating a single neuron with (ultra-) violet light, we were able to turn the targeted neuron to fluorescent red in color. This change enabled us to distinguish the morphology of a single neuron from other neurons in a complex network. Only a single neuron was distinguished, however, because this conversion is irreversible.
In this study, we have been able to take advantage of Dronpa (Ando et al., 2004), a green fluorescent protein whose fluorescence can be erased with strong blue light, and reversibly highlighted with ultraviolet (UV) or violet light (Fig. 1A). Dronpa is a genetically engineered version of green fluorescent protein cloned from Pectiniidae (Ando et al., 2004). Its fluorescence can be switched on and off by using two different wavelengths of light. Strong excitation at around 490 nm bleaches Dronpa more efficiently than other fluorescent proteins, and the bleached protein is able to completely regain its green fluorescence with minimal irradiation at around 400 nm. The off state is thermally stable (Ando et al., 2004). Here, we describe the first successful application of the reversible fluorescent protein in vivo as well as novel techniques for analyzing the anatomy of neural networks during imaging: visualization of multiple individual neurons one-by-one in a tangled network, optical antero- and retrograde labeling, time-lapse imaging, and two-photon microscopy to specifically label single neurons buried inside of tissues.
RESULTS AND DISCUSSION
Repeated Elimination and Reactivation of Green Fluorescence of Dronpa In Vivo
We first tested whether Dronpa can be expressed in zebrafish embryos and show the same photochromic behavior as reported in HeLa cells (Ando et al., 2004). We injected synthesized Dronpa mRNA into zebrafish embryos together with mRNA of membrane localized monomeric RFP (mRFP) in an attempt to visualize the cell membrane in red and cytoplasm in green, and detect the shape of the cells expressing the protein even in the absence of Dronpa's green fluorescence. The embryos showed strong green fluorescence by 50% epiboly or 5 hours postfertilization (hpf) and maintained this fluorescence for at least a few days. A 488-nm argon laser, as used for GFP, was used for observation. For photobleaching, we repeatedly scanned the specimen with 488-nm argon laser at the maximum strength, 10–20 times stronger than that used for observation, by conventional confocal microscopy until most of the green fluorescence became nonvisible. For photoactivation, a very brief scanning with 405-nm diode laser or 364-nm UV laser was sufficient. Figure 1B–G shows such an example. In this case, the entire hindbrain in the 19 hpf embryo was photobleached (Fig. 1C), and rhombomere 4 (r4) was specifically scanned and highlighted with the 405-nm laser (Fig. 1D). The entire hindbrain was then photobleached as before, and r5 was specifically scanned and highlighted with the 405-nm laser (Fig. 1E). This process could be repeated several times without showing any significant decay in the brightness of activated Dronpa, thereby allowing us to label rhombomeres 4, 5, 6, 7, and other structures sequentially, one at a time (Supplementary Movie S1a, which can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat).
It was also easy to imprint various patterns on an embryo expressing Dronpa. In the case of Figure 1H–K, we imprinted a radiation pattern on the gastrula in two different ways. The first method was to scan repeatedly with a strong 488-nm laser in a predetermined pattern. This resulted in a non–green-fluorescent pattern becoming visible against the green fluorescence background (Fig. 1I). As mRFP was also coexpressed, the nonfluorescent area appeared as red. In the second method, the entire embryo was photobleached with a strong 488-nm laser (Fig. 1J), then scanned briefly with a 405-nm laser in a predetermined pattern. This strategy resulted in a green fluorescent pattern appearing against a nongreen background (Fig. 1K). These results showed that Dronpa could be expressed efficiently by using the conventional mRNA injection into zebrafish embryos, and its photochromic behavior could be easily achieved with confocal microscopes, in vivo.
Optical Dissection of Neural Networks Expressing Dronpa to the Single Cell Level
We next attempted to use Dronpa to analyze the morphologies of several individual neurons that were overlapped or intermingled with each other (Fig. 2). Here, we were able to take advantage of a genetic mosaic system to express DRONPA in vivo (K. Hatta et al., unpublished observations). We injected UAS-dronpa construct (pUD) into the DeltaD-Gal4 transgenic fish (Tg(deltaD:Gal4; Scheer et al., 2001), in which Dronpa was preferentially expressed in neurons. After selecting areas that contain rather complex wiring patterns (Fig. 2B,F), we optically dissected neural networks to the single cell level using the procedure schematically illustrated in Figure 2A. After complete irradiation with a strong 488-nm laser, a small dot-like fluorescence sometimes remained, especially in the soma of neurons, in which Dronpa was strongly expressed. Even further irradiation had little effect on any attempt in eliminating this (Fig. 2C,D). Although the exact nature of this dot is unknown (it might be a denatured form of Dronpa), it could potentially be used as a marker for locating the soma after the complete elimination of green fluorescence. A brief scanning with 405- or 364-nm laser on the area of the soma was sufficient to re-highlight the green fluorescence. After the diffusion of fluorescent proteins along neurites, the whole morphology of each neuron became apparent (Fig. 2C,D,G; see also Supplementary Movies S2 and S3). By merging these optically isolated neuronal images, we successfully reconstructed the neural networks in which single neurons are distinguished by different pseudo-colors (Fig. 2E,H).
Retrograde Labeling of Neurons from Axons to the Soma With Dronpa
Retrograde labeling is useful in identifying the position of soma from the visible axons, and this becomes possible when the axon is sufficiently illuminated with Dronpa. The entire area of interest was first irradiated with a strong blue light to eliminate the fluorescence of Dronpa (Fig. 3B). The axon of interest was then irradiated with a 405-nm laser (Fig. 3C). It proved necessary to wait for the diffusion of the fluorescent Dronpa throughout the cell, including the soma. The time necessary for diffusion is most likely dependent on the diameter, shape, and length of the axon. In the case of the motoneuron in Figure 3, it took approximately 3 min for Dronpa to diffuse from the axon to the soma (Fig. 3C,D). A single irradiation was sufficient to obtain the selective highlighting of a single neuron in the cluster of three somas. The resulting brightness may also depend on the ratio between the volume of the axon in the area of reactivation and that of the rest of the cell. If the ratio is small, a single irradiation may be insufficient to adequately label the soma. We believe that it may be necessary to repeat the irradiation process at the same position after the diffusion has taken place to a certain extent or completed. For example, after 13 repetitions, we were able to obtain a better contrast for this neuron (Fig. 3E; see also Supplementary Movie S3).
Time-Lapse Imaging With Dronpa
The wavelengths required for eliminating fluorescence and for observation are the same. The relatively fast bleaching rate of Dronpa at 488 nm may be a drawback due to the limited number of images that can be acquired after photoactivation. To obtain multiple bright images, the excitation light must be reasonably attenuated (Ando et al., 2004). In in vivo, in particular, it is often necessary to take several z-slices to obtain three-dimensional information at each time point. As we used a relatively strong laser for observation to obtain sufficient resolution, the fluorescence became considerably weaker after taking a set of z-slices at one time point. This weakening makes it difficult to perform four-dimensional (time-lapse) analysis. We, therefore, decided to irradiate the neuron each time, or after every few times, following a set of scanning for observation, to illuminate it to a sufficient brightness. This technique is especially effective for imaging of neurons that are just starting to extend their neurites. Figure 4 illustrates such an example. In this case, a single neuron strongly expressing Dronpa was targeted (Fig. 4A). The fluorescence of the neuron remained strong by additional repeated irradiation with violet light after every three observations (Fig. 4B–E). It was possible to observe the neuron extending the axon ventrally, crossing under the floor plate at the midline, to reach the lateral wall of the opposite side, then, to bifurcate both anteriorly and posteriorly. During this period, the scan was repeated for 74 times at 15-min intervals without significant decay in brightness (see Supplementary Movie S4).
Activation of Dronpa in Cells Buried Inside of Tissues Using Two-Photon Confocal Microscopy: Visualization of the Early Axonal Trajectory of a Single Mauthner Neuron
As shown above, it was possible to re-highlight a single cell if its target soma is separated from other cells expressing Dronpa, with conventional confocal microscopy. It was not possible, however, to re-highlight exclusively a single cell or a small group of cells buried in the tissue and covered by layers of cells expressing Dronpa due to the difficulty in avoiding unintended re-highlighting of the cells in the light path out of the focal plane (Fig. 5A). If it was possible, however, to use two-photon confocal microscopy for activating Dronpa, in which a laser whose wavelength is approximately two times longer than usual is used, we could highlight cells buried inside of tissues without activating neighboring cells in the light path (Fig. 5A). In an attempt to test this possibility, we chose an area in which Dronpa was expressed in layered cells in a double-hetero transgenic of Tg(deltaD:gal4) and Tg(UAS:dronpa). In this embryo at 21–22 hpf, many neurons and some neuroepithelial cells at the site of neurogenesis were found to be fluorescent (Fig. 5B). After eliminating fluorescence of the visual field by scanning with a strong blue laser of a conventional confocal microscope, we were able to re-highlight fluorescence of the cell of interest by scanning with a 405-nm laser, but this also re-highlighted the cells located dorsally or ventrally along the path of the laser (Fig. 5C). In contrast to these findings, we were able to successfully re-highlight fluorescence in a single or a small group of cells buried inside of the tissue by repetitive scanning at 780 nm using two-photon excitation microscopy equipped with Titan-Sapphire laser, without highlighting the dorsally or ventrally located cells (Fig. 5D). In the case shown here, we were able to label almost exclusively a single neuron belonging to the Mauthner cells, paired interneurons present in the rhombomere 4 near the anterior edge of the otic vesicle. In the specimen observed before eliminating the fluorescence, we could see numerous axons crossing the midline in the hindbrain (Fig. 5E). After bleaching and the two-photon reactivation, we were able to visualize exclusively the earliest axonal trajectory of the Mauthner neuron among others (Fig. 5F). It crossed the midline, gradually turned posteriorly, and started to extend caudally along the contralateral longitudinal pathway at stage 19–21 hpf. As shown here, using two-photon microscopy, we could improve the specificity of the reactivation of Dronpa along the z (depth) -dimension in general. However, it should be noted that, when the cells expressing Dronpa are densely packed in the tissue, it is often challenging to label strongly a single cell without highlighting the cells directly neighboring to it to some degree.
In this study, we showed that it is possible to use reversible fluorescent proteins to optically dissect neuronal networks in vivo. This novel procedure of visualizing individual neurons involves the repetitive process of complete elimination, local reactivation, and imaging of the fluorescence of Dronpa with two wavelengths, and reconstruction of the network in pseudo-colors. It is consequently an operation that is well suited for automation. In the future, it may be possible to automatically visualize all of the neurons in a neural network one-by-one by reconstructing them in multiple pseudo-colors. Dronpa is also useful in studying morphogenesis of multicellular structures. The reversible nature of Dronpa makes the cycle of elimination, imprinting patterns, and subsequent observation of pattern deformation feasible many times on the same embryo, thereby enabling the analysis of cellular movements on a global scale and the sum of the entire process in simulation. As shown in Figure 1 and Supplementary Movies S1a and S1b, it is easy to highlight various structures such as rhombomeres and organs, if they can be defined by light. This highlighting helps in dissecting particular structures or organs manually or in sorting by a cell sorter after dissociation of the tissue for the use of transplantation, culture, and gene-expression profiling. Using the technique described here means that it is not necessary to touch the cells when labeling them. This noninvasive feature is of potential use in visualizing the neuron of interest before electrophysiological recording in vivo or in slice preparation. Our recent results have indicated that a similar technique is also feasible in mammals (K. Hatta and T. Omura, unpublished observation), so that detecting cell movement, axonal guidance, and mapping of the brain will become easier to perform than before and will facilitate single cell labeling and electrophysiological studies.
Co-injection or cotransfection of a Dronpa construct with mRNA or DNA encoding a dominant-negative or constitutively active form of developmentally important molecules or with siRNAs would be an effective means for identifying their roles in cellular morphogenesis of interest. Dronpa is monomeric and was fused to proteins such as ERK to track their movement between the cytoplasm and nucleus (Ando et al., 2004). By expressing such fused proteins in vivo, we expect to analyze their transport and polarized localization in individual neurons or other cell types during embryogenesis.
We also reported the generation of a permanent transgenic zebrafish line carrying UAS:dronpa (Tg(UAS:dronpa)) in this article. Dronpa could be expressed in cell types of interest by crossing this line with appropriate Gal4 transgenic lines (Fig. 5A and unpublished observation). These fish may further facilitate a systematic analysis of neuronal networks.
Wild-type zebrafish used in this article were hybrids of India and Oregon AB lines. Tg(DeltaD:Gal4) lines (Scheer et al., 2001) were obtained from the Zebrafish International Research Center (ZIRC; Eugene, OR). Each embryo was morphologically staged according to Kimmel et al. (1995) and expressed by hours postfertilization (hpf) at 28.5°C.
Injection of DNA and RNA
The UAS-Dronpa (pUD) DNA was constructed by isolating the EcoRI/PvuII-blunt Dronpa fragment from pDG1-S1 (MBL, Nagoya, Japan) and cloning it into the XhoI-blunt/EcoRI fragment of the pUS. pUS was constructed by isolating the EcoRI-blunt/EcoRI-blunt UAS-E1B fragment of pB-UAS-E1B (Koster and Fraser, 2001) and cloning into the HindIII-blunt/SalI-blunt vector fragment of pCS2+ (constructed by M. Takeuchi and K. Hatta). pB-UAS-E1B was a gift from Drs. Koster and Fraser.
For injection, plasmid DNAs were prepared using the Endofree plasmid kit (Qiagen, Tokyo, Japan). Fertilized eggs were dechorionated at the one-cell stage either manually with forceps or by digestion with 0.5 mg/ml pronase (Sigma-Aldrich, St. Louis, MO) for several minutes at room temperature followed by washing three times with fish tank water. For the circular plasmids, pUD was injected into the cytoplasm of single- to four-cell stage embryos at a concentration of 10 pg/embryo (0.5 nl of 20 ng/μl) in 100 mM KCl containing 0.02% phenol red (Sigma-Aldrich) using air pressure (IM-30, Narishige, Japan). In some cases, nondechorionated fertilized eggs were injected with plasmids. Injected embryos were raised in E2 embryo media (EM; 15 mM NaCl, 0.5 mM KCl, 1 mM CaCl2.2H2O, 1 mM MgSO4.7H2O, 0.15 mM KH2PO4, 0.05 mM Na2HPO4.2H2O, and 0.7 mM NaHCO3, pH 7.0–7.5) containing 0.5 ppm methylene blue on 1% agarose-coated culture dishes. To inhibit pigmentation, embryos were transferred to a solution of EM containing 0.003% 1-phenyl-2-thiourea (Sigma-Aldrich) before 24 hpf.
Capped mRNA was prepared from pUD and pCS2+mRFP by in vitro transcription using the mMESSAGE mMACHINE kit (Ambion, Austin, TX) with the SP6 primer. pCS2+mRFP was a gift from Dr. Tsien. The mRNA was injected into the cytoplasm of dechorionated single-cell stage embryos at a concentration of 250 pg/embryo (0.5 nl of 500 ng/μl) in H2O containing phenol red.
Generation of Tg(UAS:dronpa)
A transgenic line carrying UAS:dronpa (Tg(UAS:dronpa)) was generated using Tol2 transposon-mediated gene transfer (Kawakami et al., 2000, 2004). The pTol2UD was constructed by isolating the EcoRV/ApaI UAS:dronpa fragment from pUD (see above) and cloning it into the NruI/ApaI vector fragment of the pT2KSAG. pTol2UD (25 ng/μl) and transposase mRNA (25 ng/μl) synthesized from pCS-TP were co-injected into fertilized eggs as described above. pT2KSAG and pCS-TP were gifts from K. Kawakami.
To permit imaging, embryos were anesthetized with 0.2 mg/ml Tricaine (3-amino benzoic acid ethylester, Sigma-Aldrich) and mounted on a Lab-Tek Chambered Coverglass (Nunc, Naperville, IL) in a drop of 1% low melting point agarose (Sigma-Aldrich) in fish Ringer solution (116 mM NaCl, 2.9 mM KCl, 1.8 mM CaCl2, and 5 mM HEPES, pH 7.2) or EM, or set free in small holes made in a layer of 1% low melting point agarose filled with EM containing 1-phenyl-2-thiourea and methylene blue. Images of z-slices of multiple time points were obtained by laser confocal microscopy, processed to projections or depth coded projections, and pseudo-colored using LSM 510 (Zeiss, Jena, Germany). Elimination of Dronpa's green fluorescence was achieved by repeated scanning with the 488-nm argon laser at the maximum strength. Reactivation of the fluorescence of Dronpa was achieved by a brief scanning of a selected region of interest, using the bleach function of LSM510, with the 405-nm diode laser or the 364-nm UV laser by conventional confocal microscopy, or the 780-nm Titan-Sapphire laser by two-photon confocal microscopy. The fluorescence of Dronpa was observed as for GFP, i.e., with the 488-nm argon laser but at an attenuated strength (1–10% of the maximum depending on the specimen) to minimize photobleaching. The objective lenses used for the elimination, reactivation, and observation of Dronpa by confocal microscopy was Plan-Neofluar 20x/0.5 in general. For the specimen in Figure 1H and Supplementary Movie 1b, Plan-Neofluar 10x/0.3 was used. For the specimen in Figure 2B, 10x/0.3 was only used for the elimination and observation. For observation of the fluorescence of Dronpa with conventional fluorescent microscopes (DMRA2, Leica, Solms, Germany) or fluorescent dissection microscopes (MZFL3, Leica), we used a system involving a mercury lamp passed through a GFP filter set (GFP2 or GFP3). For observation of mRFP, we used either a 543-nm helium-neon laser for confocal microscopy or the light from the mercury lamp passed though a rhodamine filter set (G) for fluorescent microscopes.
We thank S. Aizawa for support; R. Köster, S. Fraser, K. Kawakami, and R. Tsien for plasmids; A. Miyawaki, S. Hayashi, and M. Hibi for encouragement; and M. Royle for comments. K.H. received a Grant-in-aid for Scientific Research from The Ministry of Education, Culture, Sports, Science, and Technology of Japan.