Four-Dimensional Imaging and Quantification of Gene Expression in Early Developing Zebrafish (Danio rerio) Embryos

Toxicological Sciences, Apr 2006

Four-dimensional (4D) imaging is a powerful tool for studying three-dimensional (3D) changes in an organism through time. Different imaging systems for obtaining 3D data from in vivo specimens have been developed but usually involved large and expensive machines. We successfully used a simple inverted compound microscope and a commercially available program to study and quantify in vivo changes in sonic hedgehog (shh) expression during early development in a green fluorescence protein (GFP) transgenic zebrafish (Danio rerio) line. We applied the 4D system to study the effect of 100 μM cadmium exposure on shh expression. In control zebrafish embryos, shh:GFP expression was detected at about 9 h post-fertilization (hpf) and increased steadily in the next 7 h, peaking at about 17 hpf and decreasing in the following 4 h. In the same time period, different shh expression volumes were observed in cadmium-treated and control embryos. Embryos affected by cadmium-exposure demonstrated a down-regulation in shh expression. The number of GFP-expressing cells measured by flow cytometry decreased, and expression of neurogenin-1, a downstream target of the shh signaling pathway, was down-regulated, providing additional supporting data on the effects of cadmium on shh. In summary, we demonstrated the setup of a 4D imaging system and its application to the quantification of gene expression.

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

Four-Dimensional Imaging and Quantification of Gene Expression in Early Developing Zebrafish (Danio rerio) Embryos

Richard M. K. Yu 3 4 Chun Chi Lin 3 4 P. K. Chan 3 4 Elly Suk Hen Chow 1 3 4 Margaret B. Murphy 3 4 Barbara P. Chan 0 3 Ferenc Mu ller 2 3 Uwe Strahle 2 3 S. H. Cheng 3 4 0 Medical Engineering Program, Department of Mechanical Engineering, The University of Hong Kong , Pokfulam Road , Hong Kong 1 Current address: Division of Biology, California Institute of Technology , Pasadena, CA 91125. ogy and Chemistry , City University of Hong Kong , 83 Tat Chee Avenue, Hong Kong. Fax: (852) 2788 7406 2 Institute of Toxicology and Genetics, Forschungszentrum Karlsruhe , 76021 Karlsruhe , Germany 3 The Author 2006. Published by Oxford University Press on behalf of the Society of Toxicology. All rights reserved 4 Department of Biology and Chemistry, City University of Hong Kong , 83 Tat Chee Avenue , Hong Kong Four-dimensional (4D) imaging is a powerful tool for studying three-dimensional (3D) changes in an organism through time. Different imaging systems for obtaining 3D data from in vivo specimens have been developed but usually involved large and expensive machines. We successfully used a simple inverted compound microscope and a commercially available program to study and quantify in vivo changes in sonic hedgehog (shh) expression during early development in a green fluorescence protein (GFP) transgenic zebrafish (Danio rerio) line. We applied the 4D system to study the effect of 100 mM cadmium exposure on shh expression. In control zebrafish embryos, shh:GFP expression was detected at about 9 h post-fertilization (hpf) and increased steadily in the next 7 h, peaking at about 17 hpf and decreasing in the following 4 h. In the same time period, different shh expression volumes were observed in cadmium-treated and control embryos. Embryos affected by cadmium-exposure demonstrated a downregulation in shh expression. The number of GFP-expressing cells measured by flow cytometry decreased, and expression of neurogenin-1, a downstream target of the shh signaling pathway, was down-regulated, providing additional supporting data on the effects of cadmium on shh. In summary, we demonstrated the setup of a 4D imaging system and its application to the quantification of gene expression. - These methods have some limitations in terms of sample analysis, as destruction of sample structure is inevitable, and the data are only representative of a particular developmental stage. No more data can be extracted from the same specimen in subsequent developmental stages, and only relative differences, such as presence/absence or increase/decrease, are reported. Although semiquantitative reverse transcription polymerase chain reaction (RT-PCR) can provide data on changes in gene expression (Chabicovsky et al., 2003; Ren et al., 2003), the results are still relative data. Four-dimensional (4D) is a term describing the change of a three-dimensional specimen through time. In order to keep the specimen alive throughout the experiment, optical sectioning is the only way to study three-dimensional structure. Magnetic resonance imaging (MRI), computerized tomography (CT), multiphoton microscopy, confocal laser scanning microscopy, and wide field deconvoluted microscopy are imaging techniques that fulfill this requirement of 4D imaging. Moreover, distinguishing or in vivo labeling the target cells or tissues, which provides the ability to trace any changes, is another key point in 4D study. One of the common methods used is fluorescent labeling and fluorescence microscopy. A fundamental problem with any form of 3D fluorescence microscopy is out-of-focus signals at each plane of the resulting image stacks. These signals are light from throughout the specimen that results in imprecise fluorescent images and affects any data obtained from the images. Deconvolution is a computational method used to determine how much out-offocus light is expected for the optics in use and then redistribute this light to its point of origin in the specimen (McNally et al., 1999). There are two families of deconvolution algorithms, deblurring algorithms and restoration algorithms (McNally et al., 1999; Swedlow and Platani, 2002; Wallance et al., 2001). Several deconvolution methods have been developed based on these two algorithms, including the no neighbors method, nearest neighbors method, linear methods, constrained iterative methods, and blind deconvolution (McNally et al., 1999). After deconvolution, fluorescence signals are isolated from background by thresholding. This region-oriented segmentation technique is used to classify image points as objects and to distinguish between in-focus pixels or background, such as noise signals, according to their intensity values (Gerlich et al., 2003), resulting in binarized images for quantification. A recent application of fluorescence microscopy has been the use of green fluorescence protein (GFP) transgenic fish models to study gene expression and tissue or organ development (Gong et al., 2001). Some features, including external development, optical clarity of embryos to visualize internal structure, and short generation time, make zebrafish (Danio rerio) a good GFP transgenic model. There are many GFP transgenic zebrafish lines, each labeling different cell types or tissues. Using GFP transgenic zebrafish, Blechinger et al. (2002) demonstrated the effect of cadmium on development, and Hill et al. (2003) reported quantitative analysis of gene expression after TCDD exposure. Likewise, in toto imaging has enabled the tracking of every cell in a GFP transgenic embryo through space and time during development (Megason and Fraser, 2003). Sonic hedgehog (shh) is a signaling molecule that has a wide range of biological functions in vertebrate and invertebrate development. It is involved in a number of early patterning processes, including left-right asymmetry, dorso-ventral patterning of the central nervous system and somites, patterning of limbs, as well as some aspects of organogenesis such as eye development (reviewed by Hammerschmidt, 1997; Marti and Bovolenta, 2002). In zebrafish embryos, shh is involved in the development of the trigeminal neurons (Ota and Ito, 2003). Exposure to methoprene photolytic compounds induced shhrelated developmental defects in zebrafish spinal motor and optic nerve axons (Smith et al., 2003), and defects in prechordal plate migration due to ethanol exposure (Blader and Strahle, 1998) have also been reported. Cadmium, a well-known heavy metal and environmental pollutant, has been reported to be a teratogen and developmental toxicant to embryos of various vertebrate species (Domingo, 1994; Webster, 1990). Our team reported previously that cadmium decreased shh expression in the brain region in zebrafish embryos (Cheng et al., 2000). In chicks, mice, and rats, exposure to cadmium resulted in effects including reduction in embryo body size, microphthalmia, growth retardation, and morular degeneration (Chen and Hales, 1994; De et al., 1993; Gilani and Alibhai, 1990). Furthermore, cadmium was also suspected of decreasing birthweight in humans (Frery et al., 1993). Here we report, using an automatic 4D acquisition system and the imaging processes of deconvolution and thresholding, on the effects of cadmium on shh expression in early developmental stages measured quantitatively in the GFP transgenic zebrafish line developed by Albert et al. (2003). In order to provide further evidence of the reliability of the results from 4D acquisitions, flow cytometry cell sorting and in situ hybridization of one shh-modulated gene, neurogenin-1, were performed. MATERIALS AND METHODS Zebrafish maintenance and embryo collection. In our experiment, the shh:GFP transgenic zebrafish (Danio rerio) line developed by Albert et al. (2003) was used. Zebrafish were maintained, and embryos were collected as described previously (Cheng et al., 2000). Briefly, embryos were collected during the first hour of the light period in the 14:10-h light:dark cycle. All the collected embryos were incubated in 30% Danieaus solution (19.3 nM NaCl, 0.23 mM KCl, 0.13 mM MgSO4 7H2O, 0.2 mM Ca(NO3)2, 1.67 mM HEPES [pH 7.2]) at 28.5 C until chemical treatments. Chemical treatments. Cadmium treatment took place during the gastrulation and segmentation period. Cadmium chloride (CdCl2 2H2O) (Sigma Chemical Company, St. Louis, MO) was dissolved in 30% Danieaus solution at 100 lM according to a previous report (Cheng et al., 2000). Embryos were selected at 4 h post-fertilization (hpf), and 20 embryos were kept in a 60-mm Petri dish with 6 ml of prepared cadmium chloride. For the control group, embryos were selected at 4 hpf and kept continuously in 30% Danieaus solution. Each treatment was replicated two times. Embryos were incubated at 28.5 C until further processing for imaging. Two different morphologies were observed in the cadmium-treated group. We termed those with abnormal mid-/ hindbrain morphology as cadmium-affected (Cd-af) embryos, while embryos with similar appearance to the control embryos were termed normal-looking (Cd-nl) embryos. Four-Dimensional Imaging In our experiment, up to 18 specimens bright-field and fluorescent 4D images were captured in a single time point. The system included an inverted microscope (ECLIPES TE200, Nikon, Kanagawa, Japan), a bright field (VMMD1, Uniblitz, Rochester, NY), and a fluorescence-controlling shutter device (Lambda 10-C, Sutter Instrument Company, Novato, CA), a mercury burner (OSRAM, Munich, Germany) as the light source of fluorescence, ignited by a power supply unit (HB-10104AF, Nikon) and a heat plate (MATS-U505R30, Tokai Hit, Shizuoka-ken, Japan), a motorized stage (ES100S3, Prior Scientific, Cambridge, UK), and a cooled color digital (CCD) camera (SPOT RT, Diagnostic Instruments Inc., Sterling Heights, MI). A commercially available computer program, MetaMorph version 5.0r2 (Universal Imaging Corp., Downington, PA), was used on a Windows 2000 platform in the experiment. The general concepts are summarized below; the whole imaging system setup and the detail of Metamorph processes are described in the Supplementary Material. Journal setup for imaging. A journal is a record of commands that allows a series of steps to be run automatically. In order to meet the need for both a fluorescent stack and a bright field stack for each specimen at every single time point, two sets of journals were set for the 4D imaging. One set included a series of time point (TP) journals, controlling all the imaging steps at each time point, including controlling the shutters, loading settings for imaging, and saving of image stacks as sequential names. The other set had only one journal, which controlled when the TP journals would function. Preparation of the embryos for imaging. Treated embryos at about 8 hpf were transferred to a 1% agarose (Invitrogen, Carlsbad, CA) coated 90-mm Petri dish to prevent any damage of the embryos during dechorionation. Chorions were removed using a pair of fine forceps under a dissecting stereomicroscope (SZX-12, Olympus, Tokyo, Japan). During the process, embryos were incubated in culture medium, 30% Danieaus solution with 0.016% Tricaine (Sigma) and 0.003% 1-phyenyl-2-thiourea (PTU) (Sigma). 100 lM cadmium chloride was added to Petri dishes containing embryos from the cadmium-treated group. Chorion-free embryos were transferred to wells made using a gel comb in 1% agarose in a 60-mm Petri dish with the help of a glass transfer pipette. A small hole was created in each well for keeping embryos in position during imaging. Embryos were positioned on their sides, with the yolk plug pointing downward and the dorsal side (thicker blastoderm) of the embryo on the right hand side. The Petri dish was transferred carefully to the stage of the inverted microscope and covered by a prewarmed transparent heat plate to keep the temperature at 28.5 C during the experiment. 4D imaging. In order to capture the whole embryo, the 43 lens (Plane Fluor, Nikon) was chosen. Before starting the 4D acquisition, the stage speed and the position (XYZ coordinates) of each specimen were inputted. Each embryo was focused until a sharp outline was observed, indicating a z coordinate that was approximately in the middle of the embryo, and allowing the setting of a range of z-series values covering the whole embryo for 4D imaging. The stage was moved slowly to prevent any change in embryo orientation. Image Data Processing Images of each embryo at each time point were saved in stack format. Before data extraction, images were deconvoluted and scaled down to 8-bit. All planes in each stack were saved separately in .tiff format. The next step was computerized thresholding of the images, a segmentation process. Finally, all images were quantitatively measured, and the data were logged to a Microsoft Excel file. All processes were conducted using MetaMorph except thresholding. Deconvolution and thresholding. Each plane from the image stack was binarized by thresholding after removing out-of-focus signals by deconvolution (Fig. 1). MetaMorph version 5.0r2 provided two types of deconvolution function: no neighbors and nearest neighbors. The main difference between these two methods was that the no neighbors method only used information from a single plane, while the nearest neighbors method only used adjacent focal planes, the one above and the one below the plane of interest (McNally et al., 1999). The nearest neighbors method was applied in our experiment, and 8-bit images were thresholded by a computer program developed by our laboratory using the adaptive thresholding method. Images from the same stack were read into the program and thresholded automatically. At the same time, the integrated intensity, the sum total of the pixel intensity of each plane, and total intensity, the summation of integrated intensity of whole embryo, were recorded. Quantitative measurement and data extraction. In our experiment, because an image stack (3D data) was separated into a series of independent images (2D data) for image processing, the reconstruction of images to a 3D stack was required for volume measurement. Both reconstruction and measurement were carried out by MetaMorph. To measure the volume in metric units (e.g., cm3, mm3, or lm3) rather than voxels, a calibrated distance was required. The volume was automatically calculated in lm3, and data from each image stack was logged to a Microsoft Excel file. The intensity per volume for each embryo was calculated by dividing the total intensity by the total volume (unit: lm 3). Flow cytometry cell sorting. Control and cadmium-treated embryos from shh:GFP transgenic fish were dechorionated at 24 hpf. The embryos were then added to 5 ml trypsin lysis solution (0.5 mg/ml trypsin in a solution of 0.14 M NaCl, 0.05 M KCl, 0.005 M glucose, 0.007 M NaH CO3, and 0.7 mM EDTA) and triturated through a narrow-bore Pasteur pipette until they were dissociated by continued checking under a dissection microscope. The resulting cell suspension was then centrifuged at 1000 3 g for 9 min at 4 C. The supernatant was discarded, and cells were resuspended in 5 ml phosphate buffered saline (PBS) to wash away the trypsin. They were again centrifuged at 1000 3 g for 9 min at 4 C, followed by resuspension in 5 ml PBS. Cells were analyzed by flow cytometry at a low rate, not faster than 2000 events/second. To detect GFP expression, we used wild-type zebrafish embryos as a negative control. In situ hybridization of neurogenin-1 (ngn-1). Embryos at 24 hpf were dechorionated and fixed overnight in PBS with 4% paraformaldehyde at 4 C for further investigation. Whole mount in situ hybridization was carried out as described by Westerfield (1994) with modifications (Cheng et al., 2000; Hen Chow and Cheng, 2003). Briefly, antisense RNA was synthesized by linearizing the plasmids containing cDNA probes for ngn-1 (Blader et al., 1997) and transcribing them with T7 polymerase and Digoxigenin-11-UTP (Roche, Basel, Switzerland). Embryos were transferred to methanol and stored at 20 C to increase permeability. They were then washed and digested with 10 lg/ml proteinase K in PBS with 0.1% Tween-20 before being incubated with the antisense probes at 55 C overnight. Following hybridization, probes were removed with high-stringency washes, 23 saline sodium citrate (SSC) and 0.23 SSC each twice for 30 min at 62 C. Embryos were subsequently incubated with preabsorbed sheep anti-digoxigenin-alkaline phosphatase Fab fragments (Roche) for 2 h at room temperature on a nutator. After washing, 5bromo-4-chloroindolyl phosphate was added as the substrate, and nitro blue tetrazolium was added as the coupler (Roche). The time required for color development was the same in both control and cadmium-exposed embryos. Statistical analysis. Repeated measures analysis of variance (repeated measures ANOVA) was carried out to determine significant differences in the magnitude of GFP volume and the intensity per volume for two factors, time and group, and their interaction. For statistical purposes, undetectable signals were set to one-tenth of the lowest value measured at the same time point. Levenes test was used to verify data normality. Significant differences between the control treatment and the two cadmium-exposed morphologies were determined using post-hoc Bonferronis tests. In order to determine any significant peak volume difference and whether delays in GFP expression occurred, the peak GFP volumes and the time point at which they were identified were analyzed by ANOVA with the use of post-hoc Bonferronis tests. A Students t-test was used to determine significant differences in the number of shh:GFP events measured by flow cytometry between control and cadmium-treated embryos. The level accepted for statistical significance in all cases was p < 0.05. The program used for statistical analysis was SPSS (SPSS Inc., Chicago, IL, USA). Four-Dimensional Acquisition of shh in Early-Developing Zebrafish Embryos Figure 2 shows a series of lateral view bright-field and fluorescence images of a shh:GFP transgenic zebrafish embryo, from 10 hpf to 23 hpf. Each bright-field image was the central plane of a stack of embryo images at a particular time point. The fluorescent images were maximum projections of all deconvoluted planes of the image stack at the corresponding time points. Expression of shh occurred along the floor plate and extended into the brain region in the same manner reported by Krauss et al. (1993). Quantitative Measurement of shh During Cadmium Exposure In order to demonstrate the utility of our system for toxicological studies, we performed an experiment with cadmium chloride. Bright field and maximum-projected fluorescent images of control (n 7), Cd-nl (n 4), and Cd-af embryos (n 3) at three different stages are shown in Figure 3. Bright field images showed that some cadmium-treated embryos exhibited abnormal mid-/hindbrain structure. However, significant differences in shh expression between the control and cadmium-treated embryos were difficult to determine visually. We therefore quantified the volume of fluorescence in these embryos to determine any alteration in expression. The change of volume of shh expression revealed an interesting pattern when mean SEM volume was plotted against hpf, from 10 hpf up to 23 hpf (Fig. 4a). The variation in volume of shh expression during the segmentation period is clearly demonstrated. Expression of shh started at about 9 to 10 hpf in normal zebrafish embryos at a relatively small volume of 1000 2000 lm3. Expression of shh increased steadily in the next 7 h until it reached a peak of approximately 5.4 3 103 lm3 at approximately 16 hpf. Afterwards, the volume dropped during the next 4 h, to about 3.3 3 103 lm3, and remained at this volume until the end of segmentation. Cd-nl and Cd-af embryos produced dome-shaped patterns similar to those of the control embryos (Fig. 4a). However, the magnitude of shh expression in Cd-nl embryos was quite similar to the control treatment, while there was a clear drop in expression volume in Cd-af embryos. Both groups of cadmiumtreated embryos started to express shh at about 9 to 10 hpf in small volumes (approximately 102 lm3). In Cd-nl embryo, shh expression followed a similar pattern as in the control group, peaking at a volume of approximately 5.2 3 103 lm3 and having a final volume of approximately 2.6 3 103 lm . 3 However, in Cd-af embryos, a decrease in shh expression was observed when compared with the control. This group had peak and final volumes of approximately 3.2 3 103 lm3 and 1.7 3 103 lm3, respectively. Statistical analysis indicated that the expression volume pattern were significantly different between control and Cd-af, and also Cd-af and Cd-nl embryos (control and Cd-af p 0.001, Cd-nl and Cd-af p 0.049; repeated measures ANOVA). Figure 4b shows the mean SEM intensity per volume during the same time period. However, there were no significant differences in intensity per volume between those groups (control and Cd-nl p 0.706; control and Cd-af p 0.889; Cd-nl and Cd-af p 0.200, repeated measures ANOVA). Apart from the different patterns, the effects of cadmium on gene expression were also evident in the volume and time of peak expression. Peak volume within three groups were significantly different from each other (p 0.002, ANOVA). Bonferronis FIG. 2. A series of a bright-field and fluorescent images of a control shh:GFP zebrafish embryo in lateral view from 10 hpf to 23 hpf. Bright-field series images (a1a14) are central planes of the embryo at each time point. Fluorescent series images (b1b14) are maximum projections of deconvoluted images of a stack from different time points. post-hoc test showed that peak volume in Cd-af embryos was significantly different from both normal (p 0.001) and Cd-nl embryos (p 0.048), and peak expression of shh:GFP was delayed 1 to 2 h visually from the graph in both cadmium-treated groups (Fig. 4a). However, there was no statistically significant difference in peak expression time between the groups within the experiment (p 0.072, ANOVA). Flow Cytometry of shh:GFP Transgenic Zebrafish Embryos Single suspended cells from wild-type and shh:GFP (n 3; 100 embryos per experiment) transgenic zebrafish embryos were analyzed by flow cytometry (Fig. 5). In reference to the wild-type negative control, we defined particles with fluorescent intensity higher than 101 to be shh:GFP positive. Of the total number of sorted events (250,000 events), 13.18 2.38% were shh:GFP-positive in control embryos, while only 11.57 2.87% of shh:GFP-positive events sorted in the cadmiumtreated group. Statistical analysis indicated that shh expression was significantly reduced in zebrafish embryos in the cadmium-exposed exposure regimen (p 0.047, Students t-test). Gene Expression Downstream of the shh Signaling Pathway Once shh expression is altered, the expression of downstream genes should also be affected. Upon completion of segmentation (24 hpf), ngn-1 transcripts were detected in the forebrain, midbrain and hindbrain in control embryos. In particular, ngn-1 was expressed in the midbrain tectum and in the hindbrain rhombomeres (Fig. 6a). In contrast, 18.33 3.76% (n 3; 60 embryos per experiment) of the cadmiumtreated embryos demonstrated prominent reduction of ngn-1expressing proneural clusters in these regions (Fig. 6b). By using sonic hedgehog:green fluorescent protein (shh:GFP) transgenic zebrafish, we report the availability and reliability of applying automatic four-dimensional (4D) imaging in vivo to measure gene expression quantitatively in whole embryos. The quantified down-regulation of shh expression in cadmium-exposed malformed embryos is a novel finding which appears to indicate that there is a range of cadmium effects on zebrafish embryonic development. We provide further confirmation of the decrease of shh expression caused by cadmium exposure using flow cytometry cell sorting to support the reliability of our system. In addition, we show that neurogenin-1 (ngn-1), a proneural gene found downstream of the shh signaling pathway, is also down-regulated in cadmiumexposed embryos. Studying changes in the expression of critical genes is one of the ways to understand any morphological malformation in developmental toxicology, especially when changes can be continuously recorded and quantified. In vivo 4D image acquisition, together with the use of a fluorescent transgenic model, provides a direction in the development of methods for the quantitative measurement of gene expression. In the past decade, 4D imaging has been widely applied in studying developmental biology, from the cellular level to the observation of specific structures such as chromosome movement FIG. 4. Changes in shh:GFP volume (mean SEM) in early development of the zebrafish embryo (a). Significant differences between the control (n 7) and Cd-af (n 4) individuals, as well as Cd-nl (n 3) and Cd-af were found (control and Cd-af, p 0.001; Cd-nl and Cd-af, p 0.049; repeated measures ANOVA). Intensity per volume (mean SEM) in early development of the zebrafish embryo (b). There were no significant differences between groups (control and Cd-af, p 0.889; control and Cd-nl, p 0.706; Cd-nl and Cd-af, p 0.200; repeated measures ANOVA). (Gunawardena and Rykowski, 2000), remodeling of cardiac fiber structure (Chen et al., 2003), development of optic tectal interneurons (Wu and Cline, 2003), and cell divisions during neurogenesis in the fish retina (Das et al., 2003). Additionally, it is also widely used in tumor studies (Herborn et al., 2003; Khoshyomn et al., 1998; Yang et al., 2000). 4D imaging has the potential to quantify changes through time in specific targets, such as organs, tissue, or gene expression, if labeling can be accomplished without any damage to the model. For example, 4D imaging has been used to determine the volume change of live chondrocytes under osmotic stress (Errington et al., 1997). The majority of in vivo studies to this point have used cell or tissue cultures because it is difficult to study whole-body gene expression quantitatively. With the use of GFP transgenic zebrafish, our system allows fluorescence data from the whole embryo to be captured. Automatic 4D Imaging System An automatic imaging system is more economic and flexible than time-lapse imaging that requires waiting periods of 1 to 2 h, FIG. 5. Percentage of the shh:GFP expressing cells in control and cadmium-exposed embryos measured using flow cytometry. Cadmium-exposed embryos showed a significant reduction in GFP expressing cells when compared with control embryos (p 0.047, Student t-test). Three experimental replicates of 100 embryos each for both the control and cadmium treatments were used. especially when many specimens are being processed. It also helps to minimize the error generated during moving and focusing the specimens. Our setup, which is novel in 4D imaging due to the use of a simple compound microscope rather than more expensive imaging equipment, provides automated control of the imaging process. A commercially available program, MetaMorph, was chosen to operate imaging devices and settings. Shutters controlling light sources for bright-field and fluorescence were shifted alternatively during imaging, providing a pair of similar time-image stacks for all embryos at each time point. These similar time-image stacks allow for comparisons between the bright-field and fluorescence conditions, confirming the location of the expression domains and presence of any morphological alterations. The stage and the focal plane are moved according to the settings to bring each embryo to the same coordinate and to acquire and enclose the whole embryo at the same coordinate on the z-axis at each time point respectively. This well-organized and ordered imaging not only allows huge amounts of data to be stored in an easy and systematic way, but also minimizes the error generated by a hand-operated system. If two image stacks of the same embryo at different stages are captured at different height ranges along the z-axis, the comparison between them may not be accurate, especially if some signal falls outside of the imaging z-range. Such an error can seriously affect the quality of the quantitative results. Apart from the minimization of error, the time required to complete the imaging process is relatively short. Since the journal that drives the imaging is optimized before the experiment, the main time-limiting factor is the manipulation of the embryos in each experiment. About 1.5 h are needed for a skilled person to dechorionate about 60 embryos at 8 hpf and settle them in the prepared agarose wells. The only input to the computer needed for the automatic imaging processes is the coordinates of each embryo, which requires about 5 min. More detailed information on method optimization and procedure is included in the Supplementary Material. FIG. 6. Expression pattern of ngn-1 transcripts in proneural clusters in control (a) and cadmium-treated (b) zebrafish embryos at 24 hpf. Cadmium treatment caused a reduction of ngn-1 transcripts in the embryonic brain, especially in the tectum (tct) and rhombomeres (Rbs). The images are dorsal views with the anterior to the left. Three experimental replicates of 60 embryos each for both the control and cadmium treatments were used. Scale bar equals 500 lm. Image Processing Captured images have the problem of being out-of-focus, whereby signal released from one plane will also be included on other planes, especially in fluorescent images. Deconvolution is a calculation to determine how much out-of-focus light is expected for the optic in use and then redistribute this light to its points of origin in the specimen. Many parameters are included in this calculation, including the refractive index, numerical aperture of the lens, and the z interval. There is currently no perfect deconvolution method. Incomplete deconvolution produces different intensity levels that result in different volume data. For example, a specimen is imaged as a stack with 30 planes and the central midline of the specimen lies on plane 15. After deconvolution, signals from the specimen are found between planes 10 to 20. However, if the specimen intensity doubles under the same deconvolution settings, signals can be found outside planes 10 to 20. The final result is that different quantitative data are generated. Before volume measurement, segmentation is required for deconvoluted images to distinguish signals being measured. Thresholding, used in our system, is a simple segmentation method provided by most image processing programs. For MetaMorph, the limitation of this method is that judgment is visual rather than computerized. Error is easily generated, since it is quite arbitrary and based on personal experience. In order to solve this problem, a program developed in our laboratory was used to threshold images automatically. Effect of Cadmium on shh Expression In order to apply our system to the study of developmental toxicology, we performed 4D imaging on cadmium-exposed early-developing shh:GFP transgenic zebrafish embryos. Due to the importance of shh in development, it is likely that any alteration in the signaling pathway would result in dramatic defects, both morphologically and functionally. In a previous study, our team observed six major categories of developmental abnormalities in zebrafish, including head and eye hypoplasia, hypopigmentation, cardiac edema, yolk sac deformities, altered axial curvature, and tail malformations. We also observed a loss of shh expression in the brain and anterior floor plate after cadmium treatment (Cheng et al., 2000). The results of 4D imaging help us to quantitatively determine the degree of effect of cadmium on shh expression in early-developing zebrafish embryos. When compared with the control, we found that there was no delay in shh expression, but the peak volume of the Cd-af group was reduced by nearly onehalf. At the same time, there was no change in intensity per volume between groups, showing that there was no change in the shh:GFP expression level in each cell, and the difference was due to the decrease in shh:GFP volume. To provide further evidence for the significance of our 4D imaging system, we performed flow cytometry cell sorting to count the number of events of shh:GFP-positive single suspended cells in both control and cadmium-exposed embryos at 24 hpf in order to detect the effects of cadmium on shh expression. Since all cadmium-treated embryos were analyzed together in the flow cytometry experiment, we did not distinguish embryos by their severity of malformation, and hence the difference was smaller than what we found in the 4D study. However, the results still showed a significant decrease in the number of shh:GFPexpressing cells in cadmium-treated zebrafish embryos. Apoptosis is one of the possible explanations for this decrease in expression, since cadmium can induce apoptosis in malformed tissue in zebrafish embryos (Chan and Cheng, 2003). Once the expression of shh is affected, the signaling pathway downstream of the gene will also be altered. In zebrafish, ngn-1 is a proneural gene regulated by shh signals (Blader et al., 1997). It is required for commitment of neural progenitor cells in early development. Ngn-1 is first detected during late gastrulation and later is expressed dynamically in the zebrafish embryonic brain (Blader et al., 1997; Khoshyomn et al., 1998). A positive relationship between shh and ngn-1 has been reported (Ota and Ito, 2003). We examined mRNA expression of ngn-1 by in situ hybridization and observed a reduction in expression in the midbrain tegmentum and in the hindbrain rhombomeres in cadmium-exposed embryos, indicating that the decrease in ngn-1 expression was due to the effects of cadmium on shh. Both the results of the flow cytometry and ngn-1 expression experiments provide strong support for the reliability of our system. Application and Improvement This simple 4D imaging system provides an easy way to study gene expression quantitatively in a developmental toxicology context. We can easily study the changes in expression of a particular gene within a specific period. Since toxicant exposure usually results in a range of dose-dependent adverse effects, the relationship between the expression volume of the target gene and a toxicant can be determined using bright-field image data to detect and classify morphological alterations at a particular toxicant concentration. In addition, the system can also be applied to study specific organs if the organ tissue can be labeled by fluorescent proteins. It is also possible to calculate the proportional volume change of the whole embryo by using a vital counterstain such as BODIPY TR methyl ester dye (Cooper et al., 2005). However, this calculation will increase the image-capturing time, and also requires a microscope with automatic changeable fluorescence filters. Although our system can provide relevant and informative data, improvements can be made. The half-life of the fluorescence reporter protein is a point for improvement. Since the target of our study was to demonstrate the applicability of our system, the half-life of the GFP was not our main concern, and a GFP transgenic zebrafish line with a long half-life of up to one day was used. The expression of a fluorescent protein with a long half-life can last for a number of time points during 4D imaging, especially when the experimental time interval is short, which may result in a false signal when gene expression is not actually present. Destabilized enhanced fluorescent protein (dEFP) variants have rapid turnover rates and a shorter half-life (Li et al., 1998), providing fewer false signals and improved accuracy. Additionally, time spent on data analysis is greatly affected by the efficiency of the computer used. Since both deconvulation and thresholding are mathematical calculations, the CPU speed and RAM of the computer play a critical role. Usually a computer should have at least three times as much RAM as the size of the image to be deconvoluted (Wallance et al., 2001). As discussed previously, there is no perfect deconvolution method. Users should compare different methods to find the most suitable one. In summary, with the use of an inverted compound microscope and a commercially available computer program, MetaMorph, we successfully developed a 4D imaging system and quantified shh expression in shh:GFP transgenic zebrafish embryos during early stages of development. We also applied the system to measure significant down-regulation of shh expression due to cadmium exposure in the same transgenic line. SUPPLEMENTARY MATERIAL A detailed instruction on journal setup and equipment operation during the experiment and a figure (supplementary figure 1) illustrating our setup, including the connection of all equipment and the optical pathways for imaging are provided in the Supplementary Material available online at www. ACKNOWLEDGMENTS The work described in this paper was substantially supported by a grant from City University (Project # 7001208) to S.H.C. It was also supported by the EC Integrated Project ZF-Models grant to U.S.

This is a preview of a remote PDF:

Richard M. K. Yu, Chun Chi Lin, P. K. Chan, Elly Suk Hen Chow, Margaret B. Murphy, Barbara P. Chan, Ferenc Müller, Uwe Strähle, S. H. Cheng. Four-Dimensional Imaging and Quantification of Gene Expression in Early Developing Zebrafish (Danio rerio) Embryos, Toxicological Sciences, 2006, 529-538, DOI: 10.1093/toxsci/kfj115