Transgenic line with gal4 insertion useful to study morphogenesis of craniofacial perichondrium, vascular endothelium-associated cells, floor plate, and dorsal midline radial glia during zebrafish development


  • Sohei Nakayama,

    1. Graduate School of Life Science, University of Hyogo, 3-2-1 Koto, Kamigori, Akou-gun, Hyogo 678-1297
    Search for more papers by this author
  • Takanori Ikenaga,

    1. Graduate School of Life Science, University of Hyogo, 3-2-1 Koto, Kamigori, Akou-gun, Hyogo 678-1297
    Search for more papers by this author
  • Koichi Kawakami,

    1. Division of Molecular and Developmental Biology, National Institute of Genetics, Department of Genetics, The Graduate University for Advanced Studies (SOKENDAI), 1111 Yata, Mishima, Shizuoka 411-8540, Japan
    Search for more papers by this author
  • Fumihito Ono,

    1. Laboratory of Molecular Physiology, National Institute on Alcohol Abuse and Alcoholism, National Institutes of Health, Bethesda, MD 20892, USA
    Search for more papers by this author
  • Kohei Hatta

    Corresponding author
    1. Graduate School of Life Science, University of Hyogo, 3-2-1 Koto, Kamigori, Akou-gun, Hyogo 678-1297
      Author to whom all correspondence should be addressed.
    Search for more papers by this author

Author to whom all correspondence should be addressed.


Zebrafish is a good model for studying vertebrate development because of the availability of powerful genetic tools. We are interested in the study of the craniofacial skeletal structure of the zebrafish. For this purpose, we performed a gene trap screen and identified a Gal4 gene trap line, SAGFF(LF)134A. We then analyzed the expression pattern of SAGFF(LF)134A;Tg(UAS:GFP) and found that green fluorescent protein (GFP) was expressed not only in craniofacial skeletal elements but also in the vascular system, as well as in the nervous system. In craniofacial skeletal elements, strong GFP expression was detected not only in chondrocytes but also in the perichondrium. In the vascular system, GFP was expressed in endothelium-associated cells. In the spinal cord, strong GFP expression was found in the floor plate, and later in the dorsal radial glia located on the midline. Taking advantage of this transgenic line, which drives Gal4 expression in specific tissues, we crossed SAGFF(LF)134A with several UAS reporter lines. In particular, time-lapse imaging of photoconverted floor-plate cells of SAGFF(LF)134A;Tg(UAS:KikGR) revealed that the floor-plate cells changed their shape within 36 h from cuboidal/trapezoidal to wine glass shaped. Moreover, we identified a novel mode of association between axons and glia. The putative paths for the commissural axons, including pax8-positive CoBL interneurons, were identified as small openings in the basal endfoot of each floor plate. Our results indicate that the transgenic line would be useful for studying the morphogenesis of less-well-characterized tissues of interest, including the perichondrium, dorsal midline radial glia, late-stage floor plate, and vascular endothelium-associated cells.


The zebrafish is a good model for studying vertebrate development because of its transparency during embryogenesis, its rapid development, and the availability of powerful genetic tools. We are interested in the craniofacial skeletal structures of the zebrafish. For this purpose, it is important to visualize the structure. Tol2-mediated gene trap and enhancer trap methods have been developed to study zebrafish development (Kawakami et al. 2004; Asakawa & Kawakami 2008). Also, the Gal4-UAS system has been developed to study the function of the Gal4-expressing cells or to observe them in detail (Scheera et al. 2001; Aramaki & Hatta 2006; Hatta et al. 2006; Scott et al. 2007; Asakawa et al. 2008). Gal4 transgenic fish are crossed with UAS-reporter lines, such as Tg(UAS:EGFP), Tg(UAS:RFP) (Asakawa et al. 2008), or Tg(UAS:KikGR). KikGR is a fluorescent protein that is photoconvertible from green to red by irradiation with ultraviolet light (UV) (Tsutsui et al. 2005; Hatta et al. 2006). Gal4-VP16 that contains the DNA-binding domain of Gal4 and transcriptional activator domain of VP16 is sometimes used to enhance transcriptional activity of Gal4. However, VP16 has nonspecific toxicity in vertebrate cells (Gill & Ptashne 1988; Sadowski et al. 1988). Thus Gal4FF that contains the DNA-binding domain of Gal4 and two short transcriptional activation modules from VP16 was used to minimize the toxicity of VP16 (Asakawa & Kawakami 2008; Asakawa et al. 2008).

SAGFF(LF)134A was a line identified in the course of a Gal4FF-gene trap screen. We here show the reporter expression pattern in SAGFF(LF)134A. Our results indicate that it provides a unique and novel tool for studying relatively unexplored subpopulations of tissues in several organs, such as the perichondrium, which envelops chondrocytes in the craniofacial cartilage; cells associated with the endothelium in the vascular system; and radial glia located on the midline in the spinal cord. We also demonstrate its use to reveal morphological changes in the floor plate during development, as well as the relationships between the floor plate and commissural axons or between dorsal midline radial glia and Rohon–Beard neuronal soma.

Materials and methods

Fish husbandry

Embryos were produced by pair-wise mating and kept at 28.5°C. The embryos were staged by hours postfertilization (hpf) or days postfertilization (dpf) according to Kimmel et al. (1995). The embryos used for live imaging and immunocytochemistry were treated in 0.003%N-phenylthiourea (P7629; Sigma) in egg water to block pigmentation. The experiments conducted in this study used the following strains of zebrafish: AB, SAGFF(LF)134A, Tg(-4.9sox10:EGFP) (abbreviated as Tg(sox10:EGFP) in this article) (Wada et al. 2005; Carney et al. 2006), Tg(huc:mcherry) (Won et al. 2011), Tg(UAS:GFP), Tg(UAS:KikGR) (Hatta et al. 2006), Tg(UAS:RFP) (Asakawa et al. 2008), Tg(hsp70l:Gal4) (Scheera et al. 2001), Tg(pax8:DsRed) (Ikenaga et al. 2011), Tg(isl2b:EGFP) (Pittman et al. 2008), and Tg(flk1:mRFP) (A. Kawahara, pers. comm., 2011).


Embryos and larvae were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) overnight at 4°C. They were washed with PBS three times and then incubated in 100% ethanol followed by acetone at −20°C. After washing in PBS with 0.5% Triton-X (TPBS) twice for 10 min each, the samples were incubated in 2% normal goat serum in TPBS for 3 h at room temperature and then were incubated with primary antibody overnight at 4°C. For fluorescent detection of antibody labeling, we used Alexa Fluor 488, Alexa Fluor 568-conjugated goat anti-mouse IgG or goat anti-rabbit IgG (1/500, Molecular Probes). The primary antibodies were mouse Zrf1 (1/5,000, ZIRC), mouse Zn8 (1/500, DSHB), and rabbit anti-green fluorescent protein (GFP)(1/5,000, Medical Biological Laboratories). To increase antibody penetration in spinal cord staining, 3-dpf larvae were treated with 1% collagenase P (1213857; Roche), which allowed the removal of trunk muscle cells with a fine tungsten needle before fixation (Ikenaga et al. 2011).

Infrared laser-evoked gene operator

An infrared laser-evoked gene operator (IR-LEGO) optical unit (FBGLD-1462-200-C; Sigma-Koki, Japan) was used in this experiment (Deguchi et al. 2009). A mono-coated objective lens (UApo340 20×/0.75 UV, Olympus, Japan) was used to irradiate the targets with the IR laser. During the irradiation trials, fish were embedded in a well-cooled 1% low melting temperature (LMP) agarose (A9045; Sigma) for stable positioning of the targets. Zebrafish embryos were anesthetized with 0.02% Tricaine (M0387; Sigma) before embedding. After IR irradiation, target organisms were kept under normal culture conditions for 3 days before the samples were observed. Embryos were mounted in LMP agarose for confocal imaging (FV300, Olympus).


Overall expression pattern of GFP in SAGFF(LF)134A;Tg(UAS:GFP) during development

The insertion of gal4ff was located at chromosome 13, 32052847-32053050 (Fig. 1A) (from the zTrap database at the website of the Kawakami laboratory; Kawakami et al. 2010). The insertion site was determined by inverse polymerase chain reaction (PCR) (Urasaki & Kawakami 2009). To visualize the Gal4FF expression pattern in SAGFF(LF)134A, double transgenic fish SAGFF(LF)134A;Tg(UAS:GFP) were generated. They displayed broad GFP expression. In particular, at 1–3 dpf, GFP expression was strong in the craniofacial region and partial midline structures including the floor plate and the posterior hypochord, but excluding the notochord (Fig. 1B-E). In addition, GFP was expressed in the dorsal part of the spinal cord, gut, and cartilage in the pectoral fins after 2 dpf (Fig. 1D,E).

Figure 1.

 The insertion site in the enhancer trap zebrafish line SAGFF(LF)134A, and green fluorescent protein (GFP) expression patterns during development in SAGFF(LF)134A;Tg(UAS:GFP). (A) The site of gal4ff insertion in SAGFF(LF)134A based on the zTrap genome browser. gal4ff was inserted at 32052847-32053050 into chromosome 13. The flanking regions of the gal4ff contain two genes encoding synaptotagmin XIVa and a SERTA-domain-containing protein. The directions of the arrows and arrowhead indicate the directions of the genes and the insert, respectively. (B–D) SAGFF(LF)134A;Tg(UAS:GFP) immunostained with anti-GFP antibody at 1–3 days postfertilization (dpf). (B) At 1 dpf, GFP was expressed most prominently at the craniofacial region and in the floor plate. (C) A higher-magnification image of the area indicated by the open box in B. GFP was detected in the floor plate and the posterior hypochord, but not in the notochord. (D) At 2 dpf, GFP was detected in the dorsal spinal cord in addition to the floor plate, and in the gut. (E) At 3 dpf, GFP was detected in cartilage in the craniofacial and pectoral fin bud areas. ch, ceratohyal; f, pectoral fin; fp, floor plate; g, gut; hc, hypochord; hm, hyomandibular region of the hyosymplectic; no, notochord; sp, spinal cord. Lateral views. Anterior, to the left. Dorsal, to the top. Scale bars: (B, D, E), 200 μm; (C), 20 μm.

Expression in the craniofacial skeletal structures

First, we examined the craniofacial regions of SAGFF(LF)134A;Tg(UAS:GFP) in detail. GFP was detected in the peripheral craniofacial region at 1–2 dpf (Fig. 1B,D; GFP expression in the rostral regions) and also in the ventral craniofacial cartilage elements at 3 dpf (Fig. 1E). To examine whether or not the GFP-expressing cells in SAGFF(LF)134A;Tg(UAS:GFP) at early stages are derived from the neural crest (NC), we compared the GFP distribution in SAGFF(LF)134A;Tg(UAS:GFP) with that of Tg(sox10:EGFP), which labels migratory NC cells during early development (Wada et al. 2005; Carney et al. 2006). We also used Zn8 antibody that strongly labels the cell–cell contact sites in pharyngeal pouches (Trevarrow et al. 1990), and weakly in epidermis (Fig. 2C,D; white arrows). In SAGFF(LF)134A;Tg(UAS:GFP) at 30 hpf, GFP-expressing cells were present in pharyngeal arches (PAs: a1–a3 in Fig. 2A). We observed each optical section of the Z-stacked images, and found GFP expression in the surface epidermis that was weakly Zn8 immunoreactive, but neither in the center of PAs nor in pharyngeal pouches that were strongly Zn8 immunoreactive (Fig. 2C). GFP expression was also observed in the cells under the brain and around the otic vesicles. On the other hand, in Tg(sox10:EGFP) at 30 hpf, GFP-expressing cells were accumulated at the center of each pharyngeal arch, but neither in the surface epidermis nor in the pharyngeal pouches (Fig. 2B,D). This comparison suggested that GFP-expressing cells in SAGFF(LF)134A;Tg(UAS:GFP) at 30 hpf were not NC-derived mesenchyme.

Figure 2.

 Comparison of green fluorescent protein (GFP) expression in the pharyngeal arches in SAGFF(LF)134A;Tg(UAS:GFP) (A, C) and Tg(sox10:EGFP) (B, D) at 30 h postfertilization (hpf). Both embryos were immunostained with anti-GFP (green) and Zn8 (magenta) antibodies. (A) Lateral view of SAGFF(LF)134A;Tg(UAS:GFP). Anterior, to the left. Dorsal, to the top. (B) Lateral view of Tg(sox10:EGFP). (C) A horizontal optical section image along the dotted line in A. Anterior, to the left. Midline, to the top. Lateral surface of the embryo, to the bottom. GFP-positive cells were localized at the surface region compared with Tg(sox10:EGFP). (D) A horizontal optical section image along the dotted line in B. GFP-positive, neural-crest-derived cells were distributed at high density in each pharyngeal arch sandwiched between pharyngeal pouches. The Zn8 antibody also recognizes cranial sensory ganglia (asterisks). White arrows indicate the surface epidermis that had Zn8 immunoreactivity. a1–3, pharyngeal arches 1–3, respectively; ot, otolith; p1–3, pharyngeal pouches 1–3, respectively. Scale bars: 50 μm.

Next we observed the craniofacial skeletal structures in detail at later stages and found strong GFP expression in the anterior-ventral skeletal structures at 3–4 dpf; for example, in Meckel’s cartilage (m indicated in Fig. 3A,B) and the palatoquadrate (pq in Fig. 3A,B), which was derived from PA1. GFP expression was also detected in PA2-derived skeletal structures, such as ceratohyal (ch in Fig. 3A,B) and hyosymplectic (hs in Fig. 3B). More posterior structures, such as ceratobranchial 1–5 (cb1-5 in Fig. 3B), which were derived from PA3-7, respectively, expressed GFP more weakly than PA1-2-derived cartilage (Fig. 3B). We further focused on the cartilage elements and found that these skeletal structures expressed GFP not only in developing chondrocytes but also in the perichondrium (Fig. 3C). This is in contrast to Tg(sox10:EGFP), which showed weaker GFP expression in the perichondrium (Fig. 3D). We also found that GFP expression of SAGFF(LF)134A;Tg(UAS:GFP) in the chondrocytes was weaker in the central part of the ceratohyal than at either end (Fig. 3C, asterisk).

Figure 3.

 Green fluorescent protein (GFP) expression in chondrocytes and perichondrium at the craniofacial region in SAGFF(LF)134A;Tg(UAS:GFP) (A–C) and in Tg(sox10:GFP) (D). (A) Lateral view of SAGFF(LF)134A;Tg(UAS:GFP) in the head at 3.25 days postfertilization (dpf). GFP was detected in craniofacial skeletal elements. In particular, cartilage derived from the first and second pharyngeal arches, exhibited strong GFP expression. Anterior, to the left. Dorsal, to the top. (B) Ventral view of the head at 4 dpf. Additional expression was observed in the ceratobranchial 1–5. Anterior, to the top. (C) A higher-magnification image of the open boxed area in B. GFP was detected in both chondrocytes and perichondrium. Some chondrocytes in the central part of the ceratohyal exhibited weaker GFP expression (asterisk) than those near either of the ends. (D) Ceratohyal in Tg(sox10:EGFP) at 4 dpf. GFP expression was strongly detected in chondrocytes but weakly in the perichondrium. cb, ceratobranchial; cc, chondrocyte; ch, ceratohyal, hs, hyosymplectic; m, Meckel’s; pe, perichondrium; pq, palatoquadrate. Scale bars: (A, B) 100 μm, (C, D) 20 μm.

To explore additional methods for visualizing perichondrial cells, we used the IR-LEGO system. In this method, local expression of a gene of interest is induced by irradiation with IR laser in a transgenic animal that carries the gene under the control of a heat-shock promoter (Deguchi et al. 2009; Kamei et al. 2009). To induce RFP expression in PAs, we crossed Tg(hsp70l:Gal4);Tg(UAS:RFP) with Tg(sox10:EGFP). Tg(sox10:EGFP) was used to visualize the position of each PA for focused IR irradiation. We embedded the triple transgenic embryos at 30–35 hpf and targeted a group of GFP-positive cells in PA2 on the left side with a single focused irradiation of IR laser (40 mW for 1 s, 20× objective) to induce local heat-shock (Fig. 4A). Three days after IR irradiation, RFP was expressed in the ceratohyal, specifically in the irradiated (left) side but not in the right side (Fig. 4B, arrowhead). Some epithelial and mesenchymal cells were also labeled with RFP in a restricted area that might indicate a PA2 compartment (Fig. 4B, black arrow). A higher-magnification image revealed RFP expression not only in chondrocytes but also in perichondrial cells (Fig. 4C, white arrow). This result indicated that the IR-LEGO system could be used to label perichondrial cells in PA2.

Figure 4.

 Visualization of the skeletal tissues including perichondrial cells by infrared (IR) laser-mediated gene operator system. (A) Live confocal image of the lateral view of an embryonic head at 35 h postfertilization (hpf). Anterior, to the left. The IR laser was irradiated in the second pharyngeal arch (indicated by the red circle) of Tg(sox10:EGFP);Tg(hsp70l:Gal4);Tg(UAS:RFP). Pharyngeal arches were visualized by GFP and are indicated in green. Inserted schematic drawing indicates the irradiation point by the red circle. (B) Live confocal image of the same animal 3 days after irradiation. Ventral view. Anterior, to the top. Green fluorescence from Tg(sox10:EGFP) was used as a marker for craniofacial cartilage. Red fluorescent protein (RFP) (magenta) expression induced by the local heat shock by IR-laser irradiation was found in the cartilage derived from the second pharyngeal arch (arrowhead) and on the irradiated side. RFP expression was also observed in some epithelial and mesenchymal cells in a restricted area (black arrow). (C) High-magnification image of the open box in B. RFP expression was observed in perichondrial cells (white arrow) and chondrocytes in ceratohyal. RFP expression in lenses (l) was due to the endogenous hsp70l promoter activity, and the magenta signals over the eyes (e) were due to auto-fluorescence. Some RFP-expressing epidermal and mesenchymal cells were also observed (asterisks). Scale bars: (A) and (B) 100 μm, (C) 20 μm.

Expression in the hypochord and in the vascular system

The hypochord is a single row of cells known to be present beneath the ventral midline of the notochord and on the dorsal midline of the dorsal aorta endothelium during early development (Hatta & Kimmel 1993). One of the most prominent expressions of GFP in SAGFF(LF)134A;Tg(UAS:GFP) at 1 dpf appeared in the posterior hypochord (Fig. 1C). Another prominent expression was found in cells associated with the vascular system. To investigate the relationship between GFP-positive cells and the vascular system, we crossed SAGFF(LF)134A;Tg(UAS:GFP) with Tg(flk1:mRFP), in which mRFP is expressed in the vascular endothelium. At 1 dpf, GFP-positive cells were detected in the posterior hypochord adjacent to the ventral side of the notochord (Fig. 1C). At 2.25 dpf, GFP expression was detected not only in the posterior hypochord (arrow in Fig. 5A) but also at the ventral part of the dorsal aorta in the trunk (Fig. 5B). At 3.25 dpf, more GFP-expressing cells appeared around the dorsal aorta (Fig. 5C), the posterior cardinal vein (Fig. 5C), and the endothelial cells of segmental vessels (Fig. 5D). Interestingly, these GFP-positive cells associated with the blood vessels were clearly distinct from mRFP-positive endothelial cells but were directly in contact with mRFP-positive endothelial cells at the outer side (Fig. 5B, open box).

Figure 5.

 Green fluorescent protein (GFP) expression in cells contacting the blood vessel endothelium in SAGFF(LF)134A;Tg(UAS:GFP). Confocal images (optical sections in A and B; projections in C and D) of SAGFF(LF)134A;Tg(UAS:GFP);Tg(flk1:mRFP). mRFP is expressed in the vascular endothelium. Lateral views. Anterior, to the left. Dorsal, to the top. (A and B) At 2.25 days postfertilization (dpf). (A) In the tail, ventral to the notochord. GFP expression was detected in the cells closely associated with the endothelium of the dorsal aorta (double arrowhead) as well as in the hypochord (arrow) that is also in contact with the dorsal aorta. (B) In the trunk, ventral to the notochord. Some GFP-positive cells (white arrowheads) formed the outer perimeter of mRFP-positive endothelial cells (black arrowheads). (C and D) At 3.25 dpf. (C) In the tail region, ventral to the notochord. The number of cells that expressed GFP and surrounding the dorsal aorta had increased. (D) In the trunk region, lateral to the notochord. Some GFP-expressing cells ensheathed the endothelium of the segmental vessels (black arrows). da, dorsal aorta; no, notochord; pcv, posterior cardinal vein; sv, segmental vessel. Scale bars: 20 μm.

Expression in the spinal cord

Green fluorescent protein (GFP) expression was found not only in the floor plate but also in other parts of the spinal cord from 2 dpf (Fig. 1D). To examine whether GFP-positive cells in the spinal cord are neurons or not, we crossed SAGFF(LF)134A;Tg(UAS:GFP) with Tg(huc:mCherry). In Tg(huc:mCherry), mCherry expression is driven by a 2771 bp segment including huc promoter and is expressed in all neurons (Won et al. 2011). At 3.25 dpf, the GFP-positive cells were distributed throughout the spinal cord (Fig. 6A). GFP-positive cells, that were located in the dorsal side of the spinal cord and extended from the central canal to the pial surface, were all negative for mCherry (Fig. 6B). We also found weakly GFP-positive neurons that were positive for mCherry and scattered in the spinal cord (Fig. 6C). Some showed axons extending ventrally, however, we were not able to characterize them further. The GFP expression in the dorsal elongated cells and the neurons became weaker after 3–4 dpf and were hard to detect in live animals at 10 dpf (data not shown). On closer examination of the dorsal elongated GFP-positive cells, we observed their contact with the dorsal-most mCherry-positive large neurons in SAGFF(LF)134A;Tg(UAS:GFP);Tg(huc:mCherry) (data not shown). Their locations and cell body sizes suggested they might be Rohon–Beard (RB) neurons. We therefore crossed SAGFF(LF)134A;Tg(UAS:RFP) with Tg(isl2b:EGFP), which labels sensory neurons, including RB neurons (Won et al. 2011). We confirmed that RB neurons contacted the dorsal elongated GFP-positive cells (Fig. 6D).

Figure 6.

 Green fluorescent protein (GFP) expression in the spinal cord in SAGFF(LF)134A;Tg(UAS:GFP);Tg(huc:mCherry) and the relationship with RB neurons in SAGFF(LF)134A;Tg(UAS:RFP);Tg(isl2b:GFP). (A–C) GFP and mCherry expression patterns in the spinal cord of Live confocal image of SAGFF(LF)134A;Tg(UAS:GFP);Tg(huc:mCherry) at 3.25 days postfertilization (dpf). In Tg(huc:mCherry), mCherry (magenta) was expressed in neurons. (A) In the spinal cord of the trunk. A projection of confocal optical slices at 1 μm intervals. GFP was expressed in the elongated cells from the roof to the central canal, in the floor plate, and in scattered round cells. Anterior, to the left. Dorsal, to the top. (B) An optical slice image of the open box in A at higher magnification near the midline of the spinal cord. All the elongated cells were mCherry-negative. (C) A single optical slice image of the open box in A. This optical slice image was taken at a region closer to the surface than B. Round cells were mCherry-positive (asterisks), indicating that these GFP-positive cells were neurons. (D) An optical slice image near the midline of the spinal cord in SAGFF(LF)134A;Tg(UAS:RFP);Tg(isl2b:GFP). RFP-positive cells were in close contact with Rohon–Beard neurons (RB), the GFP-positive dorsal-most large neurons. cc, central canal; fp, floor plate; sp, spinal cord. Scale bars: (A) 20 μm, (B–D) 10 μm.

These elongated GFP-positive cells may represent a subpopulation of radial glial cells because they were negative for mCherry in SAGFF(LF)134A;Tg(UAS;GFP);Tg(huc:mCherry) and they were extended from the central canal to the pial surface. We therefore performed a double immunostaining of 3.25-dpf SAGFF(LF)134A;Tg(UAS:GFP) fish with Zrf1 antibody, which is used as a marker for radial glia (Trevarrow et al. 1990). We found that both dorsal elongated GFP-positive cells and floor-plate cells were positive for Zrf1 immunoreactivity at 3.25 dpf (Fig. 7A–C). We also found that the dorsal GFP-positive cells were located at the midline of the spinal cord, and contact to the central canal and the pial surface (Fig. 7A–C, white arrow), showing the characteristic structure of radial glia.

Figure 7.

 Green fluorescent protein (GFP) expression in SAGFF(LF)134A;Tg(UAS:GFP) observed in radial glia located at the dorsal midline of the spinal cord as well as in the floor plate. A transverse section of the spinal cord in SAGFF(LF)134A;Tg(UAS:GFP) immunostained with anti-GFP (green) and Zrf1 antibody (magenta). Dotted circles mark the perimeters of the spinal cord and the central canal. (A) A Dorsal GFP-positive elongated cell (white arrow) was located along the midline, extending from the central canal to the dorsal top of the spinal cord. GFP was also expressed in the floor-plate cell (black arrow). (B) Widely distributed Zrf1-positive radial glia extending from the central canal to the pial surface of the spinal cord. (C) Merged view. The GFP-positive midline cells located dorsally (white arrow) and floor-plate cells (black arrow) were positive for Zrf1. cc, central canal. Scale bar: 20 μm.

Morphological change of a single floor-plate cell

In SAGFF(LF)134A;Tg(UAS:GFP), floor-plate cells exhibited strong GFP expression during embryogenesis. Floor-plate cells are single rows of cells located at the ventralmost part of the spinal cord (Hatta et al. 1991). They are aligned without any intercellular space at 1 dpf (see Fig. 1C), but intercellular space appears at 3.25 dpf (Fig. 6A). To confirm that this change is actually a morphological transformation of each floor-plate cell, not the development of a new cell in the vicinity, we traced a single floor-plate cell by using a photoconversion technique with SAGFF(LF)134A;Tg(UAS:KikGR). Some SAGFF(LF)134A;Tg(UAS:KikGR) embryos were found to exhibit mosaic KikGR expression in floor-plate cells, so it was easy to irradiate UV light to a single floor-plate cell. When we irradiated a floor-plate cell with UV light for 15 s at 2 dpf in SAGFF(LF)134A;Tg(UAS:KikGR), it was successfully labeled with red fluorescence (Fig. 8A). The photoconverted floor-plate cell showed cuboidal/trapezoidal shapes in the ventral region immediately after UV irradiation (Fig. 8B, 0 h post-photoconversion (hpp)). For the next 12 h, the cell became constricted in the middle (Fig. 8B, 12 hpp) and elongated in the dorso-ventral direction (Fig. 8B, 24 hpp (3 dpf)). The cell became wine glass shaped at 36 hpp (Fig. 8B, 36 hpp). At 48 hpp, no further change in its morphology was observed.

Figure 8.

 Tracing the morphogenesis of a single floor-plate cell in SAGFF(LF)134A;Tg(UAS:KikGR) by using green-to-red photoconversion of KikGR. (A) Schematic drawing of the photoconversion experiment. A floor-plate cell turned red after irradiation with a small spot of UV light for 15 s at 2 days postfertilization (dpf). (B) Morphological change in a single floor-plate cell observed from 2 to 4 dpf for 48 h post-photoconversion (hpp). Lateral view. Anterior, to the left. Dorsal, to the top. Immediately after photoconversion, the floor-plate cell was trapezoidal (0 hpp, 2 dpf). In the next 12 h, it became constricted in the middle (12 hpp, 2.5 dpf) and elongated in the dorso-ventral direction (24 hpp, 3 dpf). At 36 hpp, the constriction at the middle of the floor-plate cell was further enhanced (36 hpp, 3.5 dpf). At 48 hpp, its shape was unchanged (48 hpp, 4 dpf). cc, central canal; fp, floor plate; no, notochord; sp, spinal cord. Scale bar: 10 μm.

Location of paths for axons of commissural neurons in the floor plate

Floor-plate cells exhibited small openings in the basal endfoot during development (Fig. 9A). To investigate whether commissural neurons cross the midline through these small openings or through the larger space created between adjacent floor-plate cells, we examined axons extending from two types of commissural neurons. First, we observed an axon path in the floor plate of SAGFF(LF)134A;Tg(UAS:GFP);Tg(pax8:DsRed) at 3.25 dpf. Pax8-positive neurons in the spinal cord include commissural bifurcating longitudinal interneurons (CoBLs), whose axons extend contralaterally before bifurcating longitudinally (Ikenaga et al. 2011). Optical section images focusing on the medial portion of the spinal cord (Fig. 9B) showed pax8-positive CoBLs extending their axons across the midline (Fig. 9C). We examined the relationship between the axons of pax8-positive CoBLs and floor-plate cells, and found that the axons cross the midline in the ventral portion of the wine glass shaped floor-plate cells and that most floor-plate cells contained more than one axon path of pax8-positive CoBLs at the basal endfoot (Fig. 9D). To observe the axons of another type of commissural neuron, we crossed SAGFF(LF)134A;Tg(UAS:RFP) with the Tol-056 enhancer trap line, which expresses GFP in Mauthner cells, commissural local interneurons (CoLo), and Kolmer–Agduhr (KA) neurons (Satou et al. 2009). In SAGFF(LF)134A;Tg(UAS:RFP);Tol-056 at 3 dpf, we could trace each commissural axon from CoLo neurons by using the series of Z-stack images. In Figure 9, we could observe two CoLo axons crossing the midline of the spinal cord; one from the back (right side of the fish) and the other from the front (left side of the fish). In the first case, the CoLo axon perforated the shaft of the wine glass shaped floor-plate cell, adjacent to but not in the basal endfoot (Fig. 9E–G, black arrow). In the second case, the axon penetrated the large opening between two floor-plate cells (Fig. 9E,F, black arrowhead). Both modes were observed in CoLo axons regardless of their right or left origin (data not shown). We also found that some of the large holes between floor-plate cells were occupied with the KA soma (Fig. 9F, asterisk).

Figure 9.

 Putative channels for the axons CoBLs and CoLos in the floor plate visualized by SAGFF(LF)134A. (A–D) Optical section images of the trunk spinal cord, focused on the midline, at 3.25-days postfertilization (dpf) SAGFF(LF)134A;Tg(UAS:GFP);Tg(pax8:DsRed) in a live larva. Lateral view. Anterior, to the left. Dorsal, to the top. (A) Green fluorescent protein (GFP) expression in the floor plate at 2 dpf. Two or three small openings were observed in the basal endfoot of each floor-plate cell (arrowhead). (B) GFP expression in the other floor-plate cells. (C) DsRed expression in pax8-positive neuronal axons. Pax8-positive axons belonged to CoBLs, whose commissural axons were visible in the ventral spinal cord (arrows). (D) Merged images of B and C indicated that most of the axons of pax8-positive CoBLs passed through the basal endfeet of floor-plate cells. Most of the floor-plate cells had at least one path for pax8-positive CoBLs. (E–G) An optical section image of the spinal cord, at the midline, in 3-dpf SAGFF(LF)134A;Tg(UAS:GFP);Tol-056. (E) Red fluorescent protein (RFP) expression in floor-plate cells. (F) GFP expression in the commissural axons from the CoLos from the back (right side of the fish) (black arrow) and from the front (left side of fish) (black arrowhead), and Kolmer-Agduhr neuronal soma (asterisk). (G) Merged image of E and F indicates that CoLo axons ran either through the shaft of the floor-plate cell (black arrow) or through the space created by two floor-plate cells (black arrowhead). KA soma was located at the large space between floor-plate cells (asterisk). Scale bar: 10 μm.


GFP-expressing cells in the craniofacial skeleton

SAGFF(LF)134A;Tg(UAS:GFP) is suitable for visualizing the craniofacial skeletal structure after 3 dpf (Fig.1E and 3A–C). Craniofacial cartilage arises from the NC-derived mesenchyme of PAs, and cranial NC cells above the primordium of the midbrain or hindbrain migrate down to the PAs, which are located at the ventral portion of the head in the zebrafish (Schilling & Kimmel 1994; Olesnicky Killian et al. 2009). The behavior of migratory NC cells is visualized in Tg(sox10:EGFP) (Wada et al. 2005; Olesnicky Killian et al. 2009). Although GFP-expressing cells in SAGFF(LF)134A;Tg(UAS:GFP) were also distributed in the early developing craniofacial region of PAs, our results indicated that their distribution is distinct from that of migrating NC cells visualized in Tg(sox10:EGFP) (Fig. 2). Rather, GFP was expressed dominantly in the surface epidermis and seemed not present in the NC-derived mesenchyme at early stages. Thus, in addition to the epidermis, a novel GFP expression in the cartilages is likely to start during their formation by 3 dpf (Fig. 1D,E).

Visualization of perichondrium

Perichondrium and chondrocytes play important roles in skeleton ossification (Olsen et al. 2000; Felber et al. 2011). Chondrocytes secrete cartilage matrix and undergo proliferation, maturation into hypertrophic chondrocytes, and finally apoptosis. Perichondrium, a sheath of cells that envelops the cartilage, is directed to differentiate into osteoblasts by adjacent hypertrophic chondrocytes. Osteoblasts secrete a characteristic matrix and form a bone collar (Kronenberg 2003). We found that SAGFF(LF)134A;Tg(UAS:GFP) had strong GFP expression in the perichondrium at 3.25 dpf (Fig. 3A–C). While several transgenic lines including Tg(sox10:EGFP) are currently used to visualize craniofacial skeletal structures (Lawson & Weinstein 2002; Wada et al. 2005; Schilling et al. 2010), SAGFF(LF)134A is the only transgenic line that allows a clear visualization of craniofacial skeletal structures composed of both chondrocytes and perichondrial cells (compared with Fig. 3C,D). This Gal4-inserted transgenic line may contribute to detailed analyses of the perichondrium and the later ossification mechanism.

Gene expression patterns in the perichondrium and its function during ossification in the zebrafish have been under intense investigation recently (Eames et al. 2010; Felber et al. 2011). Although the origin of perichondrium has not been discussed in zebrafish, a transplantation experiment reported previously appears to have indicated that the perichondrium is derived from NC. In the study of Walker et al. (2007), rhodamine-labeled cells at the animal pole corresponding to the fate map of the NC localized at the shield stage were isotopically transplanted to non-labeled embryos at the shield stage, and they showed a photograph that indicated labeled perichondrial cells (Fig. 9D in Walker et al. 2007). We found here that Tg(sox10:EGFP) expressed GFP weakly in the perichondrium (pe in Fig. 3D), which may be residual GFP expressed during the NC migration. In our local gene induction experiment by using the IR-LEGO system, we irradiated IR focusing on the GFP-positive cell cluster in PA2 in Tg(sox10:GFP);SAGFF(LF)134A;Tg(UAS:RFP), and obtained labeling in perichondrium as well as in chondrocytes. These observations are all consistent with the hypothesis that the perichondrium is derived from NC. However, identification of the perichondrium precursors as NC remains to be tested further in the future.

GFP-expressing cells in the vascular system

For the vascular system, flk1/vegfr2 has been used as an endothelial-cell-specific gene (Liao et al. 1997) and its promoter is one of the earliest acting promoters in the endothelial lineage (Jin et al. 2005). Tg(flk1:mRFP) is used to visualize early vascular development, including the dorsal aorta and posterior cardinal vein, and vascular network formation (A. Kawahara, pers. comm., 2011). SAGFF(LF)134A;Tg(UAS:GFP);Tg(flk1:mRFP) indicated that GFP-positive cells were not part of the endothelium but in contact with the developing endothelium of the dorsal aorta, posterior cardinal vein, and segmental vessels (Fig. 5). They may be smooth muscles that eventually make tunica media in blood vessels. The fate of these cells should be investigated in the future.

Midline glial structures in the spinal cord

To identify radial glia, transgenic fish lines, antibodies, and an in situ probe have been used. Tg(gfap:EGFP) labels radial glia (Bernardos & Raymond 2006). Tg(olig2:EGFP) labels the oligodendrocyte-producing radial glia, localized in a part of the ventral spinal cord dorsal to the floor plate (Park et al. 2007). gfap and fabp7a mRNA are used as markers for radial glial precursors in the spinal cord, but their expression patterns are mutually exclusive in the ventral region of the spinal cord: gfap is expressed in the primary motor neuron (pMN) domain and medial floor-plate cells, while fabp7a is expressed in the lateral floor-plate cells but not in the medial floor-plate cells or in the pMN domain (Kim et al. 2008). Zrf1 monoclonal antibody directed against zebrafish GFAP is also used and labels radial glia uniformly in the spinal cord (Trevarrow et al. 1990; Marcus & Easter 1995; Roberts & Appel 2009). On the other hand, in the spinal cord of SAGFF(LF)134A;Tg(UAS:GFP), GFP was expressed in the dorsal midline radial glia and floor plate (Figs 6–8), in a restricted manner different from other markers described above. We showed that the dorsal midline radial glia and the floor plate shared two features characteristic to radial glia. First, they have elongated shapes traversing the entire radius of the spinal cord. Each cell is facing with the pial surface at one end and the central canal at the other end. Secondly they are immunoreactive with the Zrf1 monoclonal antibody that recognizes GFAP and labels radial glia. Thus, we suggest that the floor plate may be considered as a part of the radial glia as well. On the other hand, the unique location of the dorsal midline radial glia at the dorsal top of the spinal cord suggests that they may be considered as the roof plate. Their origin and developmental process should be investigated in the future.

The olig2-positive radial glial cells are the progenitors of oligodendrocytes and retain the ability to divide even in adult zebrafish (Park et al. 2007). It has been also suggested that radial glial cells are neuronal progenitors in mice (Doetsch 2003; Anthony et al. 2004; Kriegstein & Alvarez-Buylla 2009). We performed time-lapse imaging of the dorsal midline radial glia from 2.5 to 3.5 dpf and observed no newly produced neurons or glia (data not shown). However, their ability to produce neurons or glia at different stages needs to be also tested in future studies.

We also found that some dorsal midline radial glial cells are in contact with RB neurons, which are the most-dorsally located sensory neurons and that have large somata in the spinal cord (Fig. 6D). (Bernhardt et al. 1990; Metcalfe et al. 1990). The midline radial glial cells might be involved in the function of RB neurons. For example, the number of RB neurons peaks by 34 hpf and then gradually decreases due to programmed cell death (Williams et al. 2000; Svoboda et al. 2001; Reyes et al. 2004). It will be interesting to test whether or not dorsal midline radial glial cells may play a role in this phenomenon.

We have previously found that the shape of each floor-plate cell changes from cuboidal/trapezoidal to wine glass shaped in embryos expressing Kaede in the whole body by injecting with its mRNA, in which the Kaede fluorescence was photoconverted from green to red by UV or 405 nm laser in the subset of floor-plate cells (K .Hatta, pers, obs., 2003; Hatta et al. 2006). In the present study, by combining Tg(UAS:KikGR) with SAGFF(LF)134A, we demonstrated that the photoconverted single floor-plate cell with red fluorescence changed its morphology. The change in shape was not caused by cell replacement, since cells that start to express KikGR after photoconversion would be fluorescent green. Such morphological change may be due to a change in adhesion between adjacent floor-plate cells. The adhesion may weaken as the spinal cord grows except for the apical and basal ends. As a result, floor-plate cells may be stretched along the dorsoventral axis and acquire the shape of a radial-glial cell, thereby creating large spaces between floor-plate cells. These large spaces appear to be at least partially associated with the ventral-most neurons, KA neurons (Fig. 9F,G, asterisk). KA neurons are GABAnergic interneurons, contact to the central canal and flank the floor-plate cells in 20–24-hpf embryos (Martin et al. 1998). KA axon ascends ipsilaterally in ventral longitudinal fasciculus (Bernhardt et al. 1992). In cyc-1 mutant fish, which does not have floor plate, the abnormality in KA axon trajectory is increased and the number of KA neurons is decreased (Bernhardt et al. 1992). The possible significance of lasting association between floor-plate cells and KA neurons needs to be tested in future studies.

We identified a novel mode of association between axons and floor plates. We found that commissural axons of pax8-positive CoBLs cross the midline through the floor-plate cells in their basal endfeet at 3.25 dpf (Fig. 9B–D), likely through the small openings (Fig. 9A). On the other hand, CoLo axons did not run through the basal endfeet of floor-plate cells at 3 dpf. Instead, they crossed the midline through either the shaft of the floor-plate cell or the large space formed between floor-plate cells (Fig. 9E–G). CoLos are glycinergic interneurons, present at one cell per hemi-segment, active only during escape behaviors. They are activated by the ipsilateral Mauthner-axon, and shut down the activity of contralateral motoneurons (Liao & Fetcho 2008; Satou et al. 2009). The CoLo axon extends ventrally and, after encountering the ipsilateral Mauthner axon on its dorsal side, crosses the midline of the spinal cord by a route dorsal to the Mauthner axon (Satou et al. 2009). It is also possible that CoLo axons penetrate the endfoot of the floor-plate cell or on the basal membrane at early stages, and that their paths in the floor plate shift dorsally. Axons of several other types of commissural interneuron types, including CoLA, CoSA, MCoD (Bernhardt et al. 1990; Liao & Fetcho 2008), and McCoLA, contact both Mauthner-cell somata and axons in the spinal cord (Moly & Hatta 2011). Different types of neurons may take distinct routes in the floor plate. It would be interesting to test whether or not more than one commissural axon goes through the same path, and whether or not the axon paths of these neurons in the floor plate shift during embryogenesis. The growth cone and axonal development of CoBL has been visualized in live animals (Aramaki & Hatta 2006). The possibility of dynamic interactions between the growth cones of commissural neurons and the floor plate during axonal crossing across the midline should be tested.

Genes near the insertion site

SAGFF(LF)134A;Tg(UAS:GFP) expressed GFP in differentiating cartilage, floor plate, and hypochord. Initially we thought that its expression pattern might be similar to that of collagen type IIA (col2a) mRNA (Yan et al. 1995), since the expression of both was observed in the floor plate, hypochord, and craniofacial cartilage. However, SAGFF(LF)134A;Tg(UAS:GFP) did not express GFP in the notochord where col2a mRNA expresses strongly. In addition, col2a mRNA was not expressed in radial glia. In fact, the gal4ff insertion is not located at the franking region of col2a loci.

SAGFF(LF)134A was identified from a gene trap screening, but the gal4ff was not inserted at a gene-coding region (Fig. 1A). In the region within 30 kb of the gal4ff insertion, however, there are two genes; one encodes a protein of the synaptotagmin family and the other encodes a SERTA-domain-containing protein (Fig. 1A). The expression patterns of these genes and that of the gal4ff in SAGFF(LF)134A need to be compared in order to understand the enhancing activity of the region of the gal4ff insertion.


We thank Dr A. Kawahara (RIKEN QBiC), Dr R. N. Kelsh (University of Bath), Dr C. B. Chien (University of Utah), Dr H. Takeda (The University of Tokyo), and Zebrafish National BioResource Project for their kind gifts of Tg(flk1:mRFP), Tg(-4.9sox10:EGFP), Tg(isl2b:EGFP)zc7, Tol-056 enhancer trap line, and AB line, respectively. The Zn8 antibody developed by Dr B. Trevarrow was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the National Institute of Child Health and Human Development (NICHD) and maintained by The University of Iowa, Department of Biology, Iowa City, IA 52242. This work was supported by a JSPS Research Fellowship to S.N.; by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan; and by a Grant-in-Aid for Special Education and Research from the University of Hyogo to K.H.