Analysis of Kupffer's vesicle in zebrafish embryos using a cave automated virtual environment

Authors


Abstract

Previous studies show that cilia in Kupffer's vesicle (KV) generate a counterclockwise flow of fluid and provide convincing evidence that this flow regulates left/right asymmetry. We hypothesized that the distribution of cilia in KV leads to this directional flow. However, there are limitations in determining the localization of structures when viewing a three-dimensional (3-D) image on a 2-D computer screen. We analyzed the distribution of KV cilia in the Cave, an immersive virtual environment that displays stacks of confocal images in 3-D. We found 80% of the cilia are located on the dorsal surface and 20% were located on the ventral surface of the vesicle. We confirmed the ventral location of some cilia by electron microscopy. There is an asymmetrical distribution of cilia on the dorsal surface, with the anterior one third containing 50% and the posterior one third containing 20% of the cilia. This dorsal–anterior patch could explain the directionality of the flow, and could drive local differences in flow rate. Developmental Dynamics 236:1963–1969, 2007. © 2007 Wiley-Liss, Inc.

INTRODUCTION

Vertebrates, including humans, have asymmetries in their organ system. In humans the heart, stomach, and spleen are located on the left side of the body, where the liver is located on the right side of the body. Anatomically, asymmetry first becomes apparent during heart development where there is asymmetric looping of the heart tube (Kathiriya and Srivastava,2000). However, left/right asymmetry first becomes evident much earlier in embryonic development, during somitogenesis. At this stage, there is differential expression of several genes, including Lefty-1, Lefty-2, pitx2, and nodal. Expression of these genes is primarily seen on the left side of the body and mutants with either bilateral expression or no expression of these genes show randomization of left/right asymmetry (Harvey,1998; Yost,1999; Capdevila et al.,2000; Hamada et al.,2002; Levin,2005).

Recent research indicates that left/right asymmetry is actually initiated even earlier during development, during gastrulation. The determination of left/right asymmetry in mice involves a transient, ciliated organ that appears during gastrulation called the ventral node. The nodal cilia are different than other cilia in that they move in a clockwise rotation when viewed from the tip to base, instead of a back and forth beating motion seen in other types of cilia (Hirokawa et al.,2006). The clockwise rotation causes fluid in the node to flow toward the left side of the node (Nonaka et al.,1998; Takeda et al.,1999). When nodal cilia are either immobilized or absent, randomization of symmetry is often seen (Supp et al.,1997,1999; Nonaka et al.,1998; Okada et al.,1999; Kawakami et al.,2005).

There are several hypotheses as to how nodal flow can stimulate left/right asymmetry, and there is evidence that several mechanisms may be involved. The chemical gradient model suggests that the leftward flow of the nodal fluid results in the accumulation of a specific morphogen on the left side of the node. This accumulation establishes a gradient of the morphogen across the node that will lead to asymmetric expression of genes involved in left/right asymmetry (Nonaka et al.,1998; Okada et al.,2005). The two-cilia model involves the presence of two distinct types of cilia in the node, one that is involved in establishing the flow, and another which detects the flow and stimulates asymmetric gene expression resulting in left/right asymmetry (Tabin and Vogan,2003). There is also evidence that nodal flow is involved in establishing a prolonged and stable rise in intracellular calcium in cells lining the left side of the node (McGrath et al.,2003; Tanaka et al.,2005). The exact mechanism of how the calcium increase stimulates expression of left-specific pathways is poorly understood.

The homologue to the ventral node in zebrafish embryos is a structure called Kupffer's vesicle (KV). KV is a transient, fluid-filled, ciliated organ that is formed by the dorsal forerunner cells during somitogenesis. It has been suggested that the cilia are distributed equally around the interior of KV (Essner et al.,2002; Kawakami et al.,2005). However, if this were true, the fluid would flow in a continuous circular motion within KV, making it difficult to differentiate between the left and right side of the vesicle either by mechanosensory mechanisms or by accumulation of morphogens. We hypothesized that, to create a flow that allows for distinction between the left and right side of the vesicle, the distribution of cilia within KV must not be uniform. We examined, in detail, the distribution of cilia within KV to determine whether their localization was uniform or patterned.

The visualization of 3-dimensional (3-D) structures in living organisms is complicated by the fact that most 3-D visualization programs present the data in 2-D monoscopic format on a desktop computer screen. This format severely limits the ability of the user to distinguish precise locations and complex patterns. To resolve depth ambiguities, users have to typically rotate the data sets and build up a mental model of the structures from different vantage points. However, in a stereoscopically projected environment the understanding of specific structures is improved due to better depth perception by the user (Ware and Franck,1996). A Cave automated virtual environment offers further advantages in terms of the large display size, wide field-of-view, and letting users take advantage of their natural sense of proprioception to interact with and understand the data set (Cruz-Neira et al.,1993). Here, we provide a detailed description of the distribution of cilia within the KV in zebrafish embryos using a Cave system to analyze confocal data sets. Our data indicate that the cilia are not uniformly distributed, but instead are concentrated on the anterior side of the dorsal surface of the KV. This distribution pattern indicates that KV is more similar to the dorsal roof of the ventral node in mouse embryos than previously believed, providing a strong link across species in vertebrate specification of left/right asymmetry during embryogenesis.

RESULTS

Characterization of 3-D structures using confocal data sets can be a complex task. Looking at conventional 3-D reconstructions of confocal scans on a 2-D desktop-based system is not sufficient to understand and resolve the inherent 3-D complexity in the structures. In this study we used a Cave, which provides for a highly immersive and interactive experience that allows us to visualize our confocal data in 3-D as it is projected in stereo on the walls and floor of an 8 foot × 8 foot × 8 foot room (Fig. 1). This visualization format has helped us to accurately count the number of cilia and determine their location within KV. Multiple users can collaborate effectively by pointing at structures with a virtual pointer (Fig. 1). We use a volume-rendering program VOlume eXplorer (VOX,2006) to render the entire confocal z-stack interactively in real-time. In addition, the program has several features that can be used to emphasize individual channels and manipulate the data set for image analysis.

Figure 1.

Typical three-dimensional (3-D) reconstructions of confocal data sets of Kupffer's vesicle (KV) cilia at the nine somite stage projected in the Cave. Both panels show a person standing in the Cave, an 8 foot × 8 foot × 8 foot room where the 3-D image is projected onto the walls and floor. The user wears glasses that allow him/her to perceive the image in 3-D. The image is interactive in that it will move in relation to the movement of the user, further optimizing the perception. Multiple users can collaborate by using the virtual pointing and marking device, white line and red cones, respectively. A: The image can be rotated to be viewed from any vantage point. This image shows the asymmetric distribution of cilia (green) on the dorsal and ventral surface when viewed along the left/right axis. Dorsal (d) is to the left and ventral (v) to the right. Some of the ventrally located cilia have been marked (red cones). B: A box can be placed around the data set for orientation purposes (yellow). This data set shows the distribution of KV cilia (green); anterior (a) is on top, posterior (p) is on the bottom. Some of the anterior cilia have been marked (red cones). The asymmetric distribution of cilia can be seen as a clustering of cilia on the anterior surface of KV.

We analyzed KV in zebrafish embryos at the six and nine somite stages. Brightfield microscopy at these stages revealed that KV was well developed as early as six somite stage in the zebrafish strain that we used, and is virtually indistinguishable from KV at the nine somite stage (Fig. 2A,B,D,E). Measurements of the anterior/posterior axis and left/right axis of KV revealed that there was no significant difference in size between these two stages, although there is more variation at the six somite stage than at the nine somite stage, possibly due to the fact that KV is still developing at the six somite stage (Table 1).

Figure 2.

Brightfield and three-dimensional (3-D) reconstructions of confocal data sets showing Kupffer's vesicle (KV) in six and nine somite stage embryos. A,B,D,E: Side (A,D) and tail (B,E) views of a six and nine somite stage embryos, respectively. The arrow indicates KV in B and E; the anterior side of KV is on top, posterior on the bottom. The side and tail views at each stage are of the same embryo. C,F: The 3-D reconstructions of confocal data sets showing the distribution of cilia (white) in KV at the six and nine somite stage, respectively. The cilia are labeled with anti-acetylated α-tubulin. Note the greater number of cilia in the anterior one third than in the middle and posterior thirds. a, anterior one third; m, middle one third; p, posterior one third. The 3-D reconstruction of confocal data sets was done using the Leica LCS 2.61 software. Scale bars = 100 μm on brightfield images and 10 μm on confocal images.

Table 1. Comparison of the Distribution of Cilia on the Dorsal and Ventral Surface of Kupffer's Vesiclea
Somite stageA/P diameter of KV (μm ± SE)L/R Diameter of KV (μm ± SE)Average total no. of cilia in KV ± SEPercentage of cilia on dorsal sideb ± SEPercentage of cilia on ventral sideb ± SEP value dorsal–ventral difference
  • a

    A/P anterior/posterior axis; KV, Kupffer's vesicle; SE, standard error; L/R left/right axis.

  • b

    This number was determined by averaging the percent of cilia located in the designated region in each embryo.

661.2 ± 3.4 (n = 13)68.6 ± 3.2 (n = 13)77.1 ± 8.9 (n = 7)79.8 ± 2.520.2 ± 2.51 × 10−9
961.9 ± 1.7 (n = 10)65.8 ± 2.1 (n = 10)73.6 ± 5.9 (n = 10)81.7 ± 1.618.3 ± 1.63 × 10−16

We visualized the cilia lining KV in the Cave by examining confocal image stacks of embryos at the six somite and nine somite stages labeled with an antibody against acetylated α-tubulin. We were able to accurately determine the number and location of the cilia within KV. The total number of cilia in KV was consistent between the six somite and nine somite stages (Tables 1, 2). We found that the cilia were not uniformly distributed within KV. Instead, the majority of cilia were located on the dorsal surface of KV. Approximately 80% of the cilia were located on the dorsal side of KV in embryos at both stages (Table 1; Fig. 1A). The remaining 20% of the cilia were located on the ventral surface.

Table 2. Comparison of the Distribution of Cilia on the Dorsal Surface of Kupffer's Vesiclea
Somite stageAverage total no. of cilia in KV ± SEPercentage of cilia in anterior 1/3b ± SEPercentage of cilia in middle 1/3b ± SEPercentage of cilia in posterior 1/3b ± SEP value anterior–posterior difference
  • a

    KV, Kupffer's vesicle; SE, standard error.

  • b

    This number was determined by averaging the percent of cilia in the designated region found in each embryo.

670.7 ± 9.8 (n = 7)46.7 ± 1.932.0 ± 1.621.3 ± 1.09.2 × 10−7
967.3 ± 5.9 (n = 8)50.7 ± 4.527.2 ± 2.722.2 ± 2.41.6 × 10−4

We examined distribution of cilia on the dorsal surface of KV at the six somite and nine somites stages. The anterior/posterior axis of the dorsal surface was measured and then divided into three equal sections consisting of the anterior third, middle third, and posterior third. The number of cilia was determined for each section. The distribution of the cilia on the dorsal side of KV was not uniform. Approximately half of the cilia were located in the anterior one third, whereas only approximately 20% of the cilia were located in the posterior one third. The remaining 30% of the cilia were in the middle third (Table 2; Figs. 1B, 2C,F).

There is conflicting evidence for the presence of cilia on the ventral surface of KV in embryos of other fish species (Brummett and Dumont,1978; Okada et al.,2005). To confirm the localization of the cilia on the ventral side of KV, embryos were examined by transmission electron microscopy at the six somite and nine somite stages. Sections through KV were examined for cilia location. The dorsal and ventral surfaces of KV could be readily determined by the presence of yolk adjacent to the cells lining the ventral side (Fig. 3A). Closer examination of the ventral surface revealed cilia arising from cells lining the lumen of KV (Fig. 3B,C). Cilia could be distinguished from microvilli by the presence of microtubules running along the length of the cilia and by their straight, rigid appearance (Fig. 3C). Cross-sections of the cilia show that they are the 9+2 configuration (Fig. 3C, insert). There were no cilia found with the 9+0 configuration, as seen in the ventral node in mouse and rabbit, and in KV in medaka (Okada et al.,2005). These results are in agreement with the results of Kramer-Zucker et al. (2005), who also found cilia in the 9+2 configuration in KV in zebrafish embryos.

Figure 3.

Electron micrographs of Kupffer's vesicle (KV) and cilia at the six somite stage. A: An overview of KV. B: A close up of the ventral side of KV. The arrow points to a cilium located on the ventral side. C: A close-up of the cilium in B. The orientation of B and C is the same as A. Insert: A cross-section of a cilium showing the microtubules in the 9+2 configuration. a, anterior; d, dorsal; p, posterior; v, ventral; KV, Kupffer's vesicle; Y, yolk. Scale bar = 10 μm in A, 2 μm in B, 500 nm in C, 100 nm in inset.

DISCUSSION

KV in zebrafish has been reported to be a spherical organ containing monociliated cells on the inner surface (Essner et al.,2002; Kawakami et al.,2005). Using a Cave system, we have examined, in detail, the distribution of cilia within KV in six somite and nine somite stage embryos and have found that this distribution is not uniform. Instead, there is an asymmetrical distribution of the cilia within the vesicle itself. This asymmetrical distribution was difficult to distinguish with desktop-based 3-D reconstruction programs, which included the Leica LCS 2.61 software and Metamorph 6.3 software. Although these programs are very useful for other applications, the quality of the 3-D rendering created from the confocal Z-stacks on a 2-D computer desktop screen was not sufficient to accurately localize all of the cilia in each data set. Therefore, when using these programs, many data sets have to be examined and data from subsets of cilia from many different embryos have to be combined to be fully confident of the localization. This increases experimental variability and error.

The rendering quality of VOX was much better for the confocal data sets. Projecting images in the Cave system helped in determining precise locations and details because the user's interaction and movement was part of the perception in addition to the movement of the data set using the controls. Accurate localization information was obtained for all of the cilia in each confocal data set. Therefore, we could examine fewer confocal data sets to obtain statistically significant information on the localization of all of the cilia in KV. This advantage reduced the experimental variability and error.

We found that the vast majority, approximately 80%, of the cilia were located on the dorsal side of the vesicle, whereas the remaining 20% of the cilia were located on the ventral side. The cilia were not uniformly distributed on the dorsal surface of KV. Instead, there was a concentration of cilia on the anterior side, with approximately 50% of the cilia being located in the anterior one third of KV, and only 20% were located in the posterior one third.

Previous studies have shown a counterclockwise flow (anterior to left flow) in the vesicle when viewed from the dorsal side (Essner et al.,2005; Kramer-Zucker et al.,2005). Figure 4-1 shows a schematic diagram of the distribution of cilia and the direction of flow within KV. The overall direction of flow is anterior to left. The dorsal–anterior patch of cilia would be expected to generate a leftward flow in this region, whereas the remaining cilia in the middle and posterior regions would be expected to generate an overall rightward flow. Figure 4-2 shows a simplified vector map of flow created by counterclockwise rotating cilia with the asymmetric distribution described here. In this simplified map, opposing directions of flow created by adjacent cilia cancel each other out. Based on this model, cilia in the dorsal–anterior patch would generate an overall leftward flow. The more sparse cilia located in the middle and posterior sections will generate an overall rightward flow. Because the flow created by groups of adjacent cilia is equal to the sum of the flow created by each individual cilium (Buceta et al.,2005), the distribution of cilia would be expected to generate a leftward flow in the anterior one third that is 50% faster than the rightward flow generated in the posterior one third (Fig. 4-2). This difference in anterior/posterior flow rates has been reported in KV in zebrafish (Kramer-Zucker et al.,2005). If the velocity of an injected bead is calculated from the travel times reported in Kramer-Zucker et al. (2005) and the KV measurements reported here, a bead moves with an approximate velocity of 24.4 μm/sec from left to right in the anterior region and 16.7μm/sec from right to left in the posterior region of KV.

Figure 4.

1: A schematic showing cilia and fluid flow in Kupffer's vesicle (KV). Previous studies have shown a striking anterior-to-left flow in the vesicle, seen as a counterclockwise flow when viewed from the dorsal side (Essner et al.,2005; Kramer-Zucker et al.,2005). In addition, individual cilia were shown to generate localized counterclockwise vortexes when viewed from the dorsal side. The present study shows that the majority of the cilia are localized in the anterior portion of the dorsal “roof” of the vesicle. Based on the abundance of cilia in the dorsal–anterior region, we hypothesize that the flow will be strong in the anterior region of the vesicle, and will slow down in lateral and posterior regions. 2: Vector-based model showing how cilia distribution in KV can affect fluid flow. The blue vectors represent cilia rotating counterclockwise when viewed from the dorsal side. The green vectors represent lateral and medial flow. Red vectors represent opposing flow directions that effectively cancel each other out, created either by two adjacent cilia or by the rotation of a single cilium. Multiple vectors in the same direction represent a faster flow, because flow vectors are additive (Buceta et al.,2005). In this model, net flow is 50% faster toward the left in the anterior one third than flow toward the right in the posterior one third. A, anterior; P, posterior; D, dorsal (neural plate); V, ventral (yolk); L, left; R, right.

Ventral cilia are likely to turn the same direction as the dorsal cilia when viewed from tip to base. However, viewed from the dorsal side, the ventral cilia would generate clockwise vortexes that are opposite to the overall counterclockwise flow in the vesicle. Movies of vesicles injected with 0.5–2 μm fluorescent beads show striking counterclockwise vortexes, while occasionally, a clockwise vortex can be observed (Essner et al.,2005). We propose that these occasional clockwise vortexes are generated by the ventral cilia.

We found that the cilia in KV have microtubules in the 9+2 configuration. This is consistent with the findings of Kramer-Zucker et al. (2005) who also found that KV cilia in zebrafish embryos have the 9+2 configuration. Cilia in the ventral node in mouse and rabbit and KV of medaka fish lack the central pair of microtubules and have the 9+0 configuration (Bellomo et al.,1996; Okada et al.,2005). This configuration is thought to be important for the rotational movement seen in these cilia (Okada et al.,2005). Rotational movement is believed to occur due to the lack of the central pair of microtubules, which define the plane of movement in conventional cilia that have a back and forth motion (Wargo and Smith,2003). However, KV cilia in zebrafish embryos are in the 9+2 configuration, and they have rotational movement (Essner et al.,2005; Kramer-Zucker et al.,2005). Therefore, the lack of the central pair of microtubules in the nodal cilia in other species does not fully explain the rotational movement seen with these cilia. It is interesting that Buceta et al. (2005) show that bending of the cilia during the recovery stroke from left to right is necessary to generate leftward fluid flow in the mouse ventral node. They show mathematically that the fluid would move in a circular motion, as is seen in KV, if the cilia did not bend. The central pair of microtubules in KV cilia may give enough rigidity to prevent bending, contributing to the circular flow of fluid within KV.

There are two theories on how leftward flow induces left/right asymmetry in mice, and they may not be mutually exclusive. The first involves the presence of two types of cilia in the ventral node in mice: motile and nonmotile. The motile cilia are located in the center of the field of cilia, whereas the nonmotile cilia are located on the periphery. The nonmotile cilia act as mechanosensors to the leftward flow of the nodal fluid, signaling the expression of left-specific genes, such as Lefty-1, Lefty-2, pitx2, and nodal (Tabin and Vogan,2003). Nonmotile cilia have not been identified in KV thus far. The second theory states that leftward nodal flow results in the accumulation of a morphogen on the left side of the node (Nonaka et al.,1998; Okada et al.,2005). The accumulated morphogen stimulates expression of the left-specific genes. If the entire surface of KV was uniformly ciliated then neither of these methods could stimulate the expression of left-specific genes. In this case, the fluid would flow in an uninterrupted circle within KV. Sensory cilia located within the vesicle would be equally activated regardless of their location. There could be no net accumulation of a morphogen as the fluid would flow uninterrupted around the inside of the vesicle. We propose that the biased distribution of the cilia on the anterior side of KV results in an increased rate of anterior to leftward (counterclockwise) flow of KV fluid. The rate of fluid return (posterior to right flow) would be slower, as has been observed (Kramer-Zucker et al.,2005), due to the decreased number of cilia on the posterior surface of the KV. This could result in stimulation of mechanosensory cilia on the left side of KV, and/or an accumulation of a morphogen on the left side.

The fact that 80% of the cilia are located on the dorsal side of KV reveals that the structure of KV is more similar to the ventral node in mouse embryos than previously believed. In the mouse, the ventral node is defined as a field of cilia instead of a vesicle, which is covered with Reichert's membrane (Nonaka et al.,1998). KV in the zebrafish is also an enclosed organ located on the ventral side of the developing embryo near the base of the tail. The distribution of cilia is biased toward the dorsal side of the vesicle, resembling the field of cilia on the dorsal roof of the ventral node in mouse embryos. These data suggest that the mechanisms for determining left/right asymmetry are similar in mouse and zebrafish embryos.

Brummett and Dumont (1978) reported that KV in Fundulus embryos contained cilia on the dorsal surface and that the ventral surface was composed of a syncitial layer that contained no cilia. Our observations in the cave and by electron microscopy show that the ventral side of KV in zebrafish contains cilia. This may represent a fundamental morphological difference between Fundulus and zebrafish.

The use of a Cave system to resolve the precise locations and interactions of structures from confocal data sets is a powerful tool that will have widespread applications in biology and medicine. Resolving detailed structures using a standard 3-D imaging software on a desktop computer screen has many disadvantages. A lot of the inherent detail is lost on a 2-D desktop screen, and there are limits to the extent that the 3-D image can be rotated and retain resolution. As a result, more data sets must be examined to obtain useful information. The ability to view 3-D reconstructions of confocal data sets in the Cave will allow for more detailed assessments of structures using far fewer data sets, which will further enhance the power of confocal microscopy in biological imaging.

EXPERIMENTAL PROCEDURES

Animals

Adult zebrafish (Danio rerio) were purchased from Carolina Biological. The fish were maintained at 28°C in 30-gallon tanks at a density of approximately 50 fish/tank.

Embryo Collection

Adult female fish were induced to lay eggs by maintaining a cycle of 14 hr light/10 hr dark. Fish were transferred to marbled tanks before “dawn” to collect eggs, which were fertilized by males in the same tank. Fish were removed after 1 hr, and the embryos were collected and maintained in spring water at 28°C.

Embryo Fixation

At the appropriate time point (six or nine somite stage), embryos were dechorionated by incubating in 1 mg/ml pronase. Once free of their chorions they were washed several times in Ringer's salt solution (116 mM NaCl, 2.9 mM KCl, 1.8 mM CaCl2, 5 mM HEPES, pH 7.2) and then fixed with BT Fix overnight at 4°C. BT Fix = 4% paraformaldehyde, 4% sucrose, 0.15 mM CaCl2, 0.1 M PO4 buffer pH 7.3 (made by combining 4 parts 0.1 M Na2HPO4 + 1 part 0.1 M NaH2PO4). The fixative was removed by washing the embryos 3 times for 5 min in 0.1 M phosphate buffered saline, pH 7.2 (PBS, Gibco). Embryos not processed immediately were stored in 70% methanol at 4°C.

Antibodies

A monoclonal antibody to acetylated α-tubulin (Covance, MMS-413R) was used to label individual cilia. Goat anti-mouse labeled with Alexa 488 was purchased from Molecular Probes.

Immunocytochemistry

Embryos that were stored in methanol were rehydrated through a series of decreasing methanol concentrations. Samples were incubated for 5 min each in 75% methanol/25% PBS, 50% methanol/50% PBS, 25% methanol/75% PBS, and finally 100% PBS. The embryos were permeablized by incubating in acetone at −20°C for 7 min. To reduce nonspecific binding by antibodies, sites were blocked by incubating the embryos in PBS + 1% bovine serum albumin (BSA) + 2% goat serum + 0.1% Triton X-100 (blocking buffer) for 30 min at room temperature. The embryos were incubated overnight at 4°C in a 1:500 dilution of the primary antibody in blocking buffer. Excess primary antibody was removed by washing for 2 hr in PBS + 0.1% Triton X-100 (PBST), changing the buffer every 15–20 min. Embryos were probed with the fluorescent-labeled secondary antibody diluted 1:500 in blocking buffer overnight at 4°C. Excess secondary antibody was removed by washing for 2 hr in PBST as above.

Confocal Microscopy

Labeled embryos were incubated overnight in Slowfade (Molecular Probes) at 4°C and then mounted in 1% agarose for visualization of KV. All images were collected using a Leica TCS SP2 AOBS confocal microscope equipped with a 63× water objective. The 488-nm laser line was used for excitation of the Alexa 488 label, and the digital emission filter was set to collect light from 525–560 nm.

Analysis of Confocal Data

For each embryo, a series of Z-sections was created through the embryo encompassing the full depth of KV. A 3-D image was reconstructed and analyzed in the Cave (see below). The number and location of the cilia in KV were determined for each embryo. Statistical significance was determined by performing Student's t-test. All calculations were made using Excel software.

Cave

We used a volume rendering program VOX to examine the 3D stack of confocal images. Typical images used in this analysis were 512 × 512 in size, and we used 50 confocal z-sections per stack. The data sets were analyzed in a Cave system at the Center for Computation and Visualization at Brown University. The Cave is an 8 foot × 8 foot × 8 foot cube with support for stereo rendering and 6-DOF (degree of freedom) position tracking. Four Linux machines with nVIDIA 4500G graphics cards were used to drive the display walls of the Cave. We obtained rendering performance of 25 frames-per-second for typical data sets.

Brightfield Microscopy

Images of live embryos were made using a Zeiss Lumar V12 stereomicroscope, equipped with an ApoLumar ×S1.2 lens and an AxioCam MRm digital camera.

Electron Microscopy

Six and nine somite stage embryos in their chorions were fixed as described above. Embryos were postfixed in 2% glutaraldehyde/4% paraformaldehyde for 2 hr, washed in PBS, and dechorionated by hand. Embryos were incubated in 2% osmium tetraoxide for 2 hr, washed, and embedded in specific orientations in agarose cubes. The agarose cubes were dehydrated through a series of ethanol washes (30%, 50%, 80%, and 90%) and incubated overnight in 100% ethanol at 4°C. The agarose cubes were incubated in increasing concentrations (40% then 70%) of Spurr's low viscosity embedding medium (EMS, Kit 14300) in ethanol for 2-hr intervals and then incubated in 100% Spurr overnight. The agarose plugs were oriented in molds filled with 100% Spurr. The molds were cured at 60°C overnight. The molds were sectioned using a Reichert ultramicrotome and placed on copper grids (EMS). Some sections on grids were poststained for 15 min with 2% uranyl acetate in 50% methanol:50% ethanol and subsequently stained with 2% lead citrate for 15 min. Grids were examined using a Philips 410 transmission electron microscope equipped with an Advantage HR CCD camera (AMT). Images were acquired and analyzed with AMT's imaging software.

Acknowledgements

We thank Andy Crawford for his assistance with data analysis. We also thank Jurgen Schulze and Andrew Forsberg for their assistance in developing the software.

Ancillary