-
[
International C. elegans Meeting,
2001]
We have developed a computerized system that automatically acquires a cell lineage of early C. elegans embryo. At first, the system prepared a 4D Nomarski DIC microscope image of a target embryo that consists of 3D images taken every minute for 2 hours. A 3D image consists of more than 50 focal plane images at 0.5-micrometer intervals. The system then detected the regions of cell nucleus in each focal plane image. We found that the image entropy efficiently distinguishes nucleus regions from cytoplasm regions because it measures the roughness of images. All the focal plane images, whose size was about 600x600 pixels, were scanned by a small square window of 10x10 pixels measuring the image entropy of the inside regions, in order to convert the original images into entropy-value images. The regions of lower entropy are the nucleus regions. Finally, the system created a cell lineage based on the nucleus regions. Nucleus regions that represent the same nucleus at the same time point were grouped into a 3D nucleus region, then each pair of them were connected if they represent the same nucleus at two consecutive time points. The connections generated the lineage of 3D nucleus regions, which is the cell lineage. The system output three-dimensional positions of nuclei at each time point and their lineage. Currently, our system can generate the cell lineage from the 1-cell stage to about the 20-cell stage.
-
[
International Worm Meeting,
2003]
We have been developing a computerized system to measure and visualize early embryonic cell lineage of C. elegans, which aims to obtain precise positions of cells, directions of cell cleavages and timings of cell divisions. Time-lapse and a multi focal plane images of embryos are recorded by 4D DIC microscope system. In order to distinguish nucleus regions from cytoplasm regions, the image entropy, which reflects the roughness of the image texture, is utilized. A small window, which calculates the entropy value of a local part of an image, scans the entire image. The regions of lower entropy, which are equivalent to the regions of smoother image texture, are detected as nucleus regions. When nucleus regions in different focal planes at the same time point are overlapped each other, the system regards them as representing the same nucleus and groups them into a 3D nucleus region. Then 3D nucleus regions between adjacent time points are connected when they are overlapped each other and regarded as representing the same nucleus. These processes are operated through series of images, which outputs a cell lineage. However, because of the nature of the DIC microscope and the entropy filter, some nuclei were falsely detected as one merged 3D nucleus region when those nuclei were closely positioned. In order to solve this problem, we developed an algorithm to separate a merged 3D nucleus region into several 3D nucleus regions which properly represent individual nuclei. We assumed no nuclei merge together during development. Due to this assumption, merged 3D nucleus regions were automatically identified when several 3D nucleus regions were connected to one 3D nucleus region in the course of development. The system searched the smallest cross section of such a merged 3D nucleus region in between the regions corresponding to the two parental 3D nucleus regions. The merged 3D nucleus region was cut by that cross section so that each resulted 3D nucleus region represents an individual nucleus. The algorithm properly constructed 11 lineages from 1-cell stage up to 24-cell stage out of 22 image sets of wild-type embryo that contained merged 3D nucleus regions. Meanwhile, GUI based software which permits manual removal of falsely-detected non-nucleus regions and helps finding the optimal cutting section for merged 3D nucleus regions was also implemented. Currently we are considering other image processing filters to minimize the number of non-nucleus regions involved in the output of nucleus detection processes.
-
[
International Worm Meeting,
2003]
Aiming to understand the roles and the mechanisms of temporal and spatial dynamics of cellular structure during embryogenesis, we are analyzing early embryonic cell lineages of RNAi-treated embryos.L3-L4 stage worms were placed onto plates containing seeded bacteria expressing dsRNA (provided from J. Ahringer through UK MRC HGMP Resource Centre) and were incubated for 40-48 h at 22C. Then, the worm was bisected and an embryo was obtained just after the sperm entry. Development of the embryo was recorded at 22C by 4D DIC microscope system with 56 different focal planes at a distance of 0.5 m between two focal planes, at intervals of 40 sec for 2 h. The recorded images were processed by our automatic cell lineage acquisition system (Onami et al. 2001; Hamahashi et al. this meeting), which detects nucleus regions in each image, groups regions that represent the same nucleus at each time point, connects those groups that represent the same nucleus in adjacent time points and outputs cell lineage that consists of 3D positions of nucleus at each time point and their lineage from 1 cell stage to 24 cell stage.First, cell lineage was measured for 20 individual wild-type worms. The shape of the resulted lineage was similar to that of Sulston et al.'s. Variation of the lineage among different embryos was rather small, e.g. the standard deviation for cell-cycle length of each cell was less than 6 min (less than 2.5 min in most cases). Thus, we established a standard wild-type cell lineage that consists of the mean and the SD for each cell-cycle length and direction of each cell division.Then, cell lineage of RNAi-treated embryos was measured. Genes on the chromosome I whose RNAi phenotypes were reported as 100% embryonic lethal (Fraser et al. 2000) were chosen as targets. All 18 genes with 3-letter gene name (
dad-1,
dhc-1,
eft-2,
gsk-1,
hlh-2,
hmp-1,
lam-1,
mei-1,
mei-2,
mel-26,
mex-1,
rba-1,
par-6,
pfn-1,
pop-1,
rba-2,
tba-1 and
tba-2) and 15 out of 229 hypothetical genes have been examined. Unexpectedly, more than 25% (5/18) genes did not show 100% embryonic lethality in our experiments. Cell lineage measurement for the remaining genes is in progress. Detailed analysis of measured cell lineages will be presented.
-
[
International C. elegans Meeting,
2001]
We have developed a system that automatically acquires cell lineages of C. elegans from the 1-cell stage up to approximately the 25-cell stage. The system utilizes a set of 4D Nomarski DIC microscope images of C. elegans embryo consisting of more than 50 focal plane images at each minute for about 2 hours. An image-processing algorithm, utilizing the image entropy, detects the region of cell nucleus in each image, and 3D nucleus regions, each of which is a complete set of nucleus regions that represent the same nucleus at the same time point, are made. Each pair of 3D nucleus regions is then connected, if they represent the same nucleus and their time points are consecutive, and the cell lineage is created based on these connections. The resulting cell lineage consists of the three-dimensional positions of nuclei at each time point and their lineage. The system utilizes our Beowulf PC cluster, made up of 32 PC, to execute all the above processes and can deduce the cell lineage within 9 hours. We also developed a software package that three-dimensionally visualizes the resulting lineage data, which may help three-dimensional understanding of nucleus movement and division. Moreover, with this package, lineages of two different individuals - e.g. wild-type and mutant - can be visualized on the same screen. The cell lineages of 10 individual N2 worms were deduced, which were quite similar to each other and to the Sulston's cell lineage. We are establishing a standard wild-type cell lineage, which describes the mean value of nucleus position at each time point together with some statistical data, such as the variance, error distribution, etc. The lineages of
par-1 (
b274) mutants were analyzed and the difference from wild-type was recognized as reported previously. Encouraged by the performance of our system, we have started systematic cell lineage analysis of knock-out animals. Studies of nucleus movement will also be presented.
-
[
Japanese Worm Meeting,
2002]
We developed a system that automatically acquires cell lineage from the 1-cell stage up to the 24-cell stage. The system utilizes a set of 4D Nomarski DIC microscope images of embryo consisting of 55 focal plane images at each 45 seconds. An image-processing algorithm detects regions of cell nucleus in each image, and tracking these regions creates the cell lineage. The cell lineage consists of the three-dimensional positions of nuclei at each time point and their lineage. In order to start systematic cell lineage analysis, we accelerated the system by transplanting to a newly assembled PC cluster machine (48 x Pentium4 - 2.0GHz), and automating the file reformatting processes, which needed complicated human intervention in the previous system. The total execution time was shortened from 9.5 hours to 5.5 hours. As the first step of systematic cell lineage analysis, 4D microscope image sets of 30 wildtypes and 15
par-1 mutants were recorded. Cell lineage studies of these individuals and other strains will be presented.
-
[
East Asia Worm Meeting,
2004]
Aiming to understand roles and mechanisms of temporal and spatial dynamics of cellular structure during embryogenesis, we are analyzing four-dimensional (three-dimensions plus time) cell division pattern of RNAi embryos using image processing-based measurement system. L3-L4 stage worms were placed onto plates containing seeded bacteria expressing dsRNA (provided from J. Ahringer through UK MRC HGMP Resource Centre) and were incubated for 40-48 h at 22C. The worm was bisected and an embryo was obtained just after the sperm entry. Development of the embryo was recorded at 22C by 4D DIC microscope system with 66 different focal planes at a distance of 0.5 (mu, Greek)m between two focal planes, at intervals of 40 sec for 2 h. The recorded images were processed by our measurement system (Onami et al. 2001), which detects regions corresponding to nucleus in each image using image processing-based algorithm, groups regions that represent the same nucleus at each time point and connects those groups that represent the same nucleus in adjacent time points. The system outputs cell division pattern that includes 3D positions of nucleus at each time point and their lineage from 1 cell stage to 24 cell stage. All 129 genes on chromosome I whose RNAi phenotypes were reported as 100% embryonic lethal (Fraser et al. 2000) were analyzed. Four features of division pattern, i.e., life time of cell, symmetry of cell cycle, distance between cells and angle made up of 3 cells, were mathematically defined and evaluated for each RNAi embryo based on the measured cell division pattern. Clustering analysis was applied to those features and several genes whose RNAi exerts an interesting cell division pattern were identified; the interesting patterns include normal division pattern up to 24 cell but resulting in embryonic lethal, normal cell division timing but abnormal cell arrangement and slow cell division timing but normal cell arrangement. We are currently analyzing 100% embryonic lethal genes on chromosome III. Detailed analysis of cell division pattern will be presented.
-
[
International Worm Meeting,
2005]
To understand the mechanisms of spatiotemporal cell positioning during embryogenesis, we are measuring and analyzing quantitative cell division patterns of wild-type and RNAi-treated embryos. Early development for 85 wild-type embryos and 604 RNAi-treated embryos (245 genes on the chromosome I and III) whose phenotypes were reported as 100% embryonic lethal (Fraser et al. 2000; Kamath et al. 2003) was recorded by 4D DIC microscope system with 66 focal planes at a distance of 0.5m between adjacent planes, at intervals of 40 seconds for 2 hours. The recorded images were processed by our cell lineage acquisition system (Onami et al. 2001), which detects nuclear regions in each image by image-processing, and then groups regions that correspond to the same nucleus at each time point and connects those groups that correspond to the same nucleus in adjacent time points by using tracking algorithm. Produced cell division pattern consists of 3D positions of nucleus at each time point and their connections from 1-cell stage up to 24-cell stage. To explore function of genes during early embryogenesis, we are developing bioinformatics methods for analyzing these quantitative cell division pattern data. To facilitate the analysis, we developed systematic method to quantify various features of cell division pattern, such as timing of cell division, direction of cell division, relative position between nuclei and cell movement. These quantified features were analyzed by using multivariate analysis methods, such as clustering and pair-wise correlations methods. We also developed methods for finding RNAi phenotypes on various quantified features and those for investigating properties of each quantified feature, such as normality of cell cycle duration. Quantitative analyses of cell division patterns using these methods are in progress.
-
[
East Asia C. elegans Meeting,
2006]
One of the challenges for analyses of developmental process is that manipulation of gene function can affect not only fate of cells but also their location, size, movement and developmental timings. Therefore, to understand the genes function comprehensively, 4-dimensional (= 3-dimension + time) description of cellular information is crucial. The C. elegans embryo is one of the best model systems for the 4D analysis of developmental processes, because of its transparency and invariant cell lineage, as well as amenability to genetic and cellular manipulations. Recent advancement of 4D-microscopy and computer technology has led to the development of several approaches to 4D data analysis (1-5), mainly focusing on nuclear divisions and lineage analysis. Here, to study cellular dynamics of early embryogenesis in C. elegans at a higher spatiotemporal resolution, we aim to establish a new 4D image analysis algorithm by incorporating GFP markers that highlight specific subcellular components. We have been testing GFP markers available in the community (e.g., histone, β-tubulin, and γ-tubulin), and also constructing potentially useful new markers. We will present our progress on the optimization of fluorescent 4D image capturing with a high-speed laser scanning confocal microscope, and the algorithms for quantitative analysis of the 4D datasets. 1. Bao, Z., et al. (2006) PNAS 103, 2707-2712. 2. Hamahashi, S., Onami, S. and Kitano, H. (2005) BMC Bioinformatics 6, 125-140. 3. Heid P., Voss E. and Soll D. (2002) Developmental Biology 245, 329-347. 4. Schnabel R., et al. (2006) Developmental Biology 294, 438-431. 5. Bischoff M. and Schnabel R. (2006) Developmental Biology 294, 432-444.
-
[
International Worm Meeting,
2007]
A 4-D nuclear tracking system, which automatically tracks 3-dimensional (3-D) regions of nuclei of early C. elegans embryo from Nomarski DIC microscopy images, has been developed in our laboratory. The system works well on images that are conditioned so as to have a fixed image quality (e.g. brightness, contrast). It is, however, difficult for the system to process unconditioned images of different image qualities. In order to overcome this limitation of our system, we are developing software that can track nuclear regions from movies of various image qualities. Our target is tracking 2-D regions of nuclei from the fertilization to the 4-cell stage because many basic processes in development are expected to be involved in this period. In our software, local image entropy is used for detection of nuclei as in our existing system [1]. For each image the software automatically varies thresholds of local image entropy to distinguish nuclear regions from cytoplasm, whereas our existing system used a fixed threshold optimized for the conditioned images. The software first calculates a threshold for the entire image by Watsons approach [2] and determines nuclear regions using the threshold. Next, the software defines an oval area that contains each nuclear region and that is two times as large as the region, recalculates a threshold locally within the oval area by thrno method [3], and redetermines the nuclear region. This redetermination of nuclear regions is repeated 20 times for each image. This algorithm can extract regions of nuclei in the images of various image qualities. In addition to this improved nuclear detection algorithm, the software includes a forward tracking algorithm that connects a nuclear region overlapped with nuclear regions at adjacent time point, and a backward tracking algorithm that selects longer tracks of nuclear regions to exclude falsely-detected regions. As a pilot experiment, we applied the software to 27 movies of RNAi embryos in the PhenoBank database. For 26 movies, the software extracted nuclear regions well corresponding to contours of actual nuclei. For 12 in the 26 movies, the software tracked the nuclei until the 4-cell stage although sometimes there are falsely detected regions. In the remaining 14 movies, nuclei were out of focus for a long time or images leaped too much, resulting in the termination of tracking. The software will be useful not only for our 4-D nuclear tracking system but also for quantitative analysis of movies available from public databases or recorded in individual laboratories. [1] Hamahashi et al. (2005) BMC Bioinformatics. 6, 125. [2] Watson et al. (1987) Cytometry 8, 1-8. [3] Otsu. (1979) IEEE Trans. Syst. Man & Cybern. SMC-9, 62-66.