86
Genome Informatics 16(2): 86–93 (2005)
Objective Measurement of Spindle Orientation
in Early Caenorhabditis elegans Embryo
1
2
Shugo Hamahashi1,2
Shuichi Onami1,2
[email protected]
[email protected]
Graduate School of Science and Technology, Keio University, 3-14-1 Hiyoshi, Kohoku, Yokohama
223-8522, Japan
Institute for Bioinformatics Research and Development (BIRD), Japan Science and Technology
Agency, 5-3 Yonbancho, Chiyoda, Tokyo 102-0081, Japan
Abstract
The spindle orientation is a crucial piece of information to understand the development of
embryo. The spindle forms during cell division and the cell divides along the spindle axis. Spindle
orientation was measured in many different mutant embryos of Caenrohabditis elegans. However,
the objectivity and the productivity of these measurements were low because these measurements
were made manually. Here we present a system that automates the measurement of the spindle
orientation in C. elegans embryo. Automation increases the objectivity and productivity of the
measurement. We confirmed the applicability of the system by applying it to spindles during the
second divisions in wild-type and mutant C. elegans embryos.
Keywords: cell division, embryogenesis, objective measurement, automation
1
Introduction
The spindle orientation during the cell division is a crucial piece of information to understand the
development of embryos. The spindle forms during cell division and the cell divides along the spindle
axis. In Caenorhabditis elegans embryo, the germ-line cells divide along the anterior-posterior axis,
whereas the somatic cells divide orthogonally to the previous division [12]. The pattern of cell divisions
from one-cell stage to the adult is roughly the same among the wild-type embryos [15]. Many mutants
were identified in which the pattern of cell divisions is remarkably different from that in wild-type.
Spindle orientation was measured in many different mutant embryos [2, 7, 17].
The spindle orientation has been measured manually at a single time point [2, 13, 15, 17]. Two
methods have been used to measure the spindle orientation, namely that using positions of nuclei
and that using positions of centrosomes. The orientation of cell division, which is parallel to the
spindle orientation, was measured using a line connecting the positions of daughter nuclei. In this
method, nuclei were visualized in a living embryo through a Nomarski differential interference contrast
(DIC) microscope [10] and the positions of the nuclei were determined manually at the birth of
these nuclei [13, 15]. The spindle orientation was measured using a line connecting the positions of
centrosomes. In this method, centrosomes were visualized either by using a DIC microscope [3, 17]
or by immunostaining [2, 17], and the positions of centrosomes were determined manually at a single
time point. Because the positions of nuclei or centrosomes are determined manually, the objectivity
and productivity of the measurement are low in these methods. In addition, because the spindle
orientation has been measured in a single time point in these methods, the time-transition of the
spindle orientation remains to be measured.
We previously developed a system [4] that detects nuclei in DIC images of C. elegans embryos
(we call this system “nucleus detection system” (NDS) in this paper). This NDS uses the local image
entropy to discriminate the image texture of nuclei from that of cytoplasm. Because the image texture
Objective Measurement of Spindle Orientation
DIC images
87
y
nuclear region
regression line
Local image
entropy
pixel
Image regions
of spindles
Figure 1: Detection of image regions of spindles
in the Caenorhabditis elegans embryo.
θ
x
(anterior-posterior)
Figure 2: Orientation of image region of spindle.
of spindle is similar to that of nucleus, the NDS detects not only nuclei but also spindles in DIC images
of C. elegans embryos. Because the NDS almost automatically detects spindles, the objectivity and
the productivity of the detection of spindles are high.
This paper reports a system that automates the measurement of the spindle orientation in early
embryos of C. elegans. This system uses NDS to objectively produce the image regions of spindle and
uses the regression line [9] to objectively calculate the spindle orientation. This system enables highly
objective and productive measurement of the spindle orientation in early C. elegans embryos.
2
Production of Image Regions of Spindle from DIC Images
In this section, we explain NDS [4]. NDS uses the local image entropy [5] to quantify the smoothness
of image texture in various regions of a DIC image of a C. elegans embryo and produces image regions
of spindles.
First, we explain how the local image entropy quantifies the smoothness of image texture in various
regions of an image. The local image entropy calculates the image entropy of a local image. The
definition of the local image entropy is as follows: Let [xij ] be the matrix representing a digital input
image, where (i, j) is {(i, j)|x ≤ i < x+w, y ≤ j < y +h}. The local image entropy E(x, y) is described
P −1
by the following equation: E(x, y) = − N
k=0 P (k) log2 P (k), where N is the number of gray levels in
the image and P (k) is the probability of occurrence of gray level k in the local image whose size is
w × h. If both the size of w × h and the number of gray levels are fixed, E(x, y) becomes smaller when
larger number of pixels in the local image have the same gray level, i.e. the image texture in the local
image is smoother. On the contrary, E(x, y) becomes larger when the smaller number of pixels have
the same gray level, i.e. the image texture is rougher. Therefore, the local image entropy quantifies
the smoothness of image texture in various regions of an image.
Next, we explain how NDS produces image regions of spindles from a DIC image of a C. elegans
embryo. NDS uses difference in the smoothness of image texture between the spindle and the cytoplasm
in DIC images, i.e. image texture is smoother in the spindle than in the cytoplasm (“DIC images”
in Figure 1). Because local image entropy quantifies the smoothness of image texture, local image
entropy is smaller in the spindle than in the cytoplasm (“Local image entropy” in Figure 1). NDS
produces image regions of spindles by selecting pixels whose local image entropy is lower than a
threshold (“Image regions of spindles” in Figure 1).
88
3
3.1
Hamahashi and Onami
Methods
Preparation of Sets of DIC Images of C. elegans Embryos
Embryos immediately after fertilization (before the meeting of female and male pronuclei) was dissected
from a gonad, placed on a slide coated with 0.01 % poly-L-lysine (Sigma) in M9 solution, covered with
a coverslip, and sealed with petroleum jelly. Differential interference contrast images were obtained
using a Leica DMRE microscope equipped with an HCX PL APO 100×/1.40 NA objective, whose
illumination intensity and objective-side Wollaston prism were adjusted to obtain images of the same
quality. Digital images of 600 × 600 pixels with 256 gray levels (0.1 µm/pixel) were recorded with an
ORCA CCD Camera (Hamamatsu Photonics). The recording system was controlled by IP Lab 3.5
software (Scanalytics). The set of images of each embryo was recorded at 22˚C. The time interval
between time points of images was 40 s.
3.2
Production of Image Regions of Spindles in NDS
DIC images of a C. elegans embryo were applied to NDS [4], and image regions consisting of pixels
whose local image entropy is lower than the threshold were produced. The window size for calculating
the local image entropy was 10 × 10 pixels. The value of threshold to binaries the local image entropy
was 95 for DIC images of wild-type embryos and 110 for those of par-2 (RNAi) and par-3 (RNAi)
embryos. The image regions of nuclei were automatically selected from the produced regions using
an object-tracking algorithm of NDS [4]. If image regions of cytoplasm were falsely selected by the
object-tracking algorithm, we manually removed these regions.
3.3
Measurement of Spindle Orientation Using the Regression Line
The spindle starts to form at prometaphase of cell cycle, when the centrosomes move at the spindle
poles. The time point 0 of each set of images was set to the earliest time point after the centrosomes’
movement. The spindle orientation was defined using the line connecting two positions of centrosomes
at the spindle poles. To measure the spindle orientation using image regions, we assumed that the
orientation of image region of spindle should correlate with the line connecting two positions of centrosomes. With this assumption, the spindle orientation was calculated using the regression line of
the image region of spindle (Figure 2).
The calculation of the angle of the regression line of an image region is as follows: All pixels in
an image region of a spindle were used to calculate the angle of the regression line. When Sxx is the
sum of squared errors in x and Sxy is the sum of cross-products of x and y, Sxx and Sxy are described
P
P
P
P
as follows: Sxx = (x − x)2 and Sxy = (x − x)(y − y), where x = x/N , y = y/N , N is the
number of data (number of pixels in the image region). The angle of the regression line is calculated
as Sxy /Sxx and the angle θ [degree] is described as follows: θ = arctan(Sxy /Sxx ).
4
4.1
Results
Spindle Orientations in Wild-Type C. elegans Embryo
We measured the spindle orientation of wild-type C. elegans embryos using our system. In this paper,
we focused on the second cell divisions of embryo because the entire shape of the spindles during
the second cell divisions, in which the embryo grows from 2-cell to 4-cell embryo, can be observed
in a single focal plane of the DIC microscope. During the second cell divisions, the spindle in the
anterior cell (called “AB”) forms transversely along the anterior-posterior (AP) axis, whereas the
spindle orientation in the posterior cell (called “P1 ”) is parallel to the AP axis [12]. We recorded DIC
images of wild-type embryos and examined whether the entire shape of spindles were observed. In the
Objective Measurement of Spindle Orientation
89
DIC images
Local image
entropy
Image regions
of spindles
AB
AB
69
Spindles'
orientations
AB
86
46
66
AB
P1
44
-15
0s
P1
-20
P1
40 s
-16
80 s
AB
P1
12
11
P1
120 s
160 s
Figure 3: Image regions of spindles and spindle orientations in a wild-type
Caenorhabditis elegans embryo. Panels in the top row show time-series DIC images
of the same wild-type Caenorhabditis elegans embryo. Each panel in the second row
shows local image entropy corresponding to the DIC image at the top panel in each
column. The brighter color represents the higher local image entropy. White regions
in the third row shows image regions of spindles. Measured orientations of spindles
are shown with angles (degree) and lines in the bottom row in each column. The
axes of reference orientation to calculate the spindle orientation are shown as white
arrows in the third top panel at time 0 s. The anterior-posterior axis of the reference
orientation is parallel to a line connecting centroids of image regions of AB and P1
at the last time point before spindle formations.
most of recorded sets (8/10) of DIC images, both AB spindle and P1 spindle were observed in a single
focal plane. In the other sets (2/10), one of the poles of AB spindle went to a different focal level so
that the entire spindle was not observed in a single focal plane.
We selected DIC images in which the entire shape of spindles were observed, applied these images
to the NDS, and obtained image regions of spindles (“DIC images” and “Image regions of spindles”
in Figure 3). Our system calculated the angle of the regression line of these image regions (“Spindles’
orientations” in Figure 3). We obtained the angles of the spindle orientations during the second
divisions of embryo’s development (“Spindles’ orientations” in Figure 3). For example, we obtained
86˚(40 s in Figure 3) for the AB spindle which is known to be transverse along the AP axis, and –20˚
(40 s in Figure 3) for the P1 spindle which is known to be parallel to the AP axis. From the viewpoint
of the measurement system, we examined the reproducibility of the measured values of our system.
We confirmed that the measured values were completely the same when we measured the spindle
orientation several times at the same time point using the same set of images (data not shown).
4.2
Spindle Orientations in RNAi Embryos
To examine if this system can measure the spindle orientations in mutant embryos, we measured spindle
orientations of par-2 (RNAi) and par-3 (RNAi) embryos using this system. The spindle orientations
in these embryos are different from that in wild-type embryo [3, 7]. In the par-2 (RNAi) embryo, both
AB and P1 spindles are transverse along the AP axis [7]. In the par-3 (RNAi) embryo, both AB and
P1 spindles are parallel to the AP axis [7].
We recorded DIC images of par-2 (RNAi) and par-3 (RNAi) embryos and applied these images
to our system. We obtained the angles of the spindle orientations of par-2 (RNAi) and par-3 (RNAi)
embryos. Figure 4 shows results of a par-3 (RNAi) embryo. In this example, we obtained 23˚(0 s in
Figure 4) for the orientation of AB spindle and 14˚(0 s in Figure 4) for the orientation of P1 spindle.
90
Hamahashi and Onami
DIC images
Local image
entropy
Image regions
of spindles
23
Spindles'
orientations
AB
P1
22
14
0s
AB
P1
2
40 s
80 s
13
AB
23
AB
-3
P1
-7
P1
120 s
39
AB
-6
P1
160 s
Figure 4: Image regions of spindles and spindle orientations in a par-3 (RNAi) embryo. Panels in the top row show time-series DIC images of the same par-3 (RNAi)
embryo. Each panel in the second row shows local image entropy corresponding to
the DIC image at the top panel in each column. The brighter color represents the
higher value of the local image entropy. White regions in the third row show image
regions of spindles. Measured orientations of spindles are shown with angles (degree) and lines in the bottom row in each column. The axes of reference orientation
to calculate the spindle orientation are shown as white arrows in the third panel at
time 0 s. The anterior-posterior axis of the reference orientation is parallel to a line
connecting centroids of image regions of AB and P1 at the last time point before
spindle formations.
4.3
Time-Transition of Spindle Orientation
Figure 5 shows the time-transitions of the spindle orientations during the second divisions of embryo’s
development in wild-type, par-2 (RNAi) and par-3 (RNAi) embryos. For each condition of embryos,
the spindle orientations were measured in five embryos and the results were averaged in each time
point. We found clear differences in time-transitions among these embryos. In wild-type embryos, the
orientation of AB spindle was about 85˚(transverse to the AP axis) and that of P1 spindle was about
–5˚ (parallel to the AP axis) at the beginning of cell division. As the cell division proceeded, the
orientation of AB spindle decreased and the P1 orientation increased so that orientations gradually
became nearly parallel. In par-2 (RNAi) embryos, orientations of both AB and P1 spindles were about
89˚and 91˚respectively (almost transverse to the AP axis) at the beginning of cell division, and did
not change a lot within 120 s. In par-3 (RNAi) embryos, orientations of both AB and P1 spindle were
about 2˚and 9˚respectively (almost parallel to the AP axis) at the beginning of cell division, and
did not change a lot within 120 s.
5
Discussion
In this paper, we reported a system that automates the measurement of the spindle orientation during
the second cell division in C. elegans embryos. We consider that the system has three main advantages. The first advantage is the high objectivity of the measurement of spindle orientation. The
spindle orientation has been manually measured by discriminating the difference of image texture
between the spindles and the cytoplasm in DIC images [3, 17] or detecting the spindles stained using
immunostaining technique [2, 17]. Thus, objectivity of the measurement has been low. To address this
problem, we used NDS [4] and the regression line [9]. NDS can objectively produce the image region
Objective Measurement of Spindle Orientation
100
91
100
100
80
80
60
40
20
0
0
40
80
-20
AB spindle
60
40
AB spindle
20
P1 spindle
0
120
0
40
80
Spindle orientation [deg]
P1 spindle
80
Spindle orientation [deg]
Spindle orientation [deg]
AB spindle
P1 spindle
60
40
20
0
0
120
40
80
120
-20
-20
Time [s]
Time [s]
Time [s]
wild-type
par-2 (RNAi)
par-3 (RNAi)
Figure 5: Time-transitions of spindle orientations in wild-type, par-2 (RNAi), and
par-3 (RNAi) embryos. Horizontal value represents the averaged orientation of spindles (N = 5), and the error bar represents the standard deviation. Time point 0
was defined as the time point when all centrosomes in AB and P1 cell positioned on
the cell-division axis.
of the spindles [4]. The regression line can objectively calculate the angle of image region. Because
all tasks in the measurement are carried out objectively in our system, we consider that our system
measures the spindle orientation highly objectively.
The second advantage is the high productivity of the measurement of spindle orientations. The
spindle orientation has been measured by manually looking at images one by one to discriminate
the image texture of spindle from that of cytoplasm or to determine the positions of centrosomes [2,
3, 17]. Because the measurement of spindle orientation has been laborious, the productivity of the
measurement has been low. To address this problem, we automated most of tasks in the measurement
of spindle orientation in our system. We used NDS, which automatically produces image regions
of spindles so that the productivity of producing image regions of spindles is high. We used the
regression line, which automatically calculates the spindle orientation from an image region of spindle.
We measured spindle orientations of five wild-type, par-2 (RNAi) and par-3 (RNAi) embryos using
our system (Figure 5), suggesting that our system can measure the spindle orientation with high
productivity. This high productivity will allow a large-scale analysis of spindle orientation. Because
of a benefit of 4-dimensional DIC microscope system [6, 16], numbers of DIC images of C. elegans
embryos can be recorded automatically. In addition, we can obtain time-lapse DIC images of RNAi
embryos of C. elegans for many genes from online databases [14]. Appling these data to our system,
we have started a systematic analysis of spindle orientation.
The third advantage of our system is that the system enables the analysis of time-transitions of
spindle orientations. Although the spindle orientation changes during cell division (Figure 5), the
spindle orientation has been usually measured at a single time point during cell division [2, 3, 17].
Because our system can measure the spindle orientation highly objectively and highly productively,
time-transitions of spindle orientations can be measured at very short intervals. Analysis of these data
will provide a new insight into the mechanisms of development.
The major disadvantage of our system is that the system is only applicable to 2-dimensional images.
Because of this limitation, in this paper, we focused on the second cell division of C. elegans, in which
the entire shape of the spindle can be viewed usually in a single focal plane. However, sometimes
the shape of the spindle cannot be viewed in a single focal plane even in the second cell division.
In our experiments, one of the poles of AB spindle went to a different focal level and could not be
viewed with a single focal plane in 2/10 image sets. After the second cell divisions, the entire shape
of the spindle cannot be viewed in a single focal plane. To overcome this limitation, a combination of
the 4-dimensional (4D) DIC microscope system [6, 16] and NDS [4] might be effective. The 4D DIC
microscope system automatically shift its focus position, records images, and repeats the recording
92
Hamahashi and Onami
with a specified time interval. Using this microscope system, sets of 4D DIC images can be obtained.
NDS can produce image regions of spindle from the set of 4D DIC images. Thus, the combination
of these systems will be able to measure the spindle orientation after the second cell divisions in C.
elegans embryo.
The objectivity and the productivity of the measurement are expected to contribute to understanding the development of embryos. We have been focused on the highly objective and highly productive
measurement of nuclear behavior in C. elegans embryos [4, 8, 11, 18]. NDS [4] automatically detects
nuclei in a set of DIC images of C. elegans embryos. The high objectivity of the detection of nuclei
greatly helped to reveal a mechanism of male pronuclear migration in the very early C. elegans embryo [8]. The highly objective and highly productive measurements of the nuclei by NDS are used
in our research project of a systematic analysis of patterns of cell division in C. elegans embryos
(Onami et al., in preparation). The highly objective and highly productive measurement of spindle
orientations by our system will provide crucial clues to understanding the development of embryos.
6
Conclusion
We reported a system that measures the spindle orientation in C. elegans embryos. Using the nucleus
detection system and the regression line, our system objectively measured the spindle orientation
during second cell division in wild-type, par-2 (RNAi) and par-3 (RNAi) embryos. We consider that
our system contributes to understanding the development of C. elegans.
Acknowledgments
We are grateful to Dr. K. Kyoda who helped a task of image processing and provided useful discussions
and comments. We would like to thank Dr. H. Amano, Dr. K. Oka, Dr. A. Kimura, Mr. M. Urai
and all members in the Onami laboratory for their supports. This work was supported in part by a
grant from the Special Coordination Funds for Promoting Science and Technology, the Ministry of
Education, Culture, Sports, Science and Technology of Japan.
References
[1] Alberts, B., Johnson, A., Lewis, J., Raff, M., Roberts, K., and Walter, P., Molecular Biology of
the Cell, Garland Science, 2001.
[2] Cheng, N.N., Kirby, C.M., and Kemphues, K.J., Control of cleavage spindle orientation in
Caenorhabditis elegans: the role of the genes par-2 and par-3, Genetics, 139:549–559, 1995.
[3] Gomes, J.E., Encalada, S.E., Swan, K.A., Shelton, C.A., Carter, J.C., and Bowerman, B., The
maternal gene spn-4 encodes a predicted RRM protein required for mitotic spindle orientation
and cell fate patterning in early C. elegans embryos, Development, 128:4301–4314, 2001.
[4] Hamahashi, S., Onami, S., and Kitano, H., Detection of nuclei in 4D Nomarski DIC microscope
images of early Caenorhabditis elegans embryos using local image entropy and object tracking,
BMC Bioinformatics, 6:125, 2005.
[5] Handmann, U., Kalinke, T., Tzomakas, C., Werner, M., and W. Seelen, An image processing
system for driver assistance, Image Vision Comput., 18:367–376, 2000.
[6] Hird, S.N. and White, J.G., Cortical and cytoplasmic flow polarity in early embryonic cells of
Caenorhabditis elegans, J. Cell. Biol., 121:1343–1355, 1993.
Objective Measurement of Spindle Orientation
93
[7] Kemphues, K.J., Priess, J.R., Morton, D.G., and Cheng, N.S., Identification of genes required for
cytoplasmic localization in early C. elegans embryos, Cell, 52:311–320, 1988.
[8] Kimura, A. and Onami, S., Computer simulations and image-processing reveal length-dependent
pulling force as the primary mechanism for C. elegans male pronuclear migration, Dev. Cell,
8:765–775, 2005.
[9] Montgomery, D.C. and Peck, E.A., Introduction to Linear Regression Analysis, John Wiley &
Sons, 1982.
[10] Nomarski, G. and Weill, A., Application à la métallographie des méthodes interférentielles à deux
ondes polarisées, Rev. Metall., 2:121–128, 1955.
[11] Onami, S., Hamahashi, S., Nagasaki, M., Miyano, S., and Kitano, H., Automatic acquisition of
cell lineage through 4D microscope and analysis of early C. elegans embryogenesis, In Foundations
of Sysmtes Biology, Kitano, H. Eds., MIT Press, 39–55, 2001.
[12] Riddle, D.L., Blumenthal, T., Meyer, B.J., and Priess, J.R., C. Elegans II, Cold Spring Harbor
Laboratory Press, 1997.
[13] Schnabel, R., Hutter, H., Moerman, D., and Schnabel, H., Assessing normal embryogenesis in
Caenorhabditis elegans using a 4D microscope: variability of development and regional specification, Dev. Biol., 184:234–265, 1997.
[14] Sönnichsen, B., Koski, L.B., Walsh, A., Marschall, P., Neumann, B., Brehm, M., Alleaume, A.M.,
Artelt, J., Bettencourt, P., Cassin, E., Hewitson, M., Holz, C., Khan, M., Lazik, S., Martin, C.,
Nitzsche, B., Ruer, M., Stamford, J., Winzi, M., Heinkel, R., Roder, M., Finell, J., Häntsch, H.,
Jones, S.J., Jones, M., Piano, F., Gunsalus, K.C., Oegema, K., Gonczy, P., Coulson, A., Hyman,
A.A., and Echeverri, C.J., Full-genome RNAi profiling of early embryogenesis in Caenorhabditis
elegans, Nature, 434:462–469, 2005.
[15] Sulston, J.E., Schierenberg, E., White, J.G., and Thomson, J.N., The embryonic cell lineage of
the nematode Caenorhabditis elegans, Dev. Biol., 100:64–119, 1983.
[16] Thomas, C., DeVries, P., Hardin, J., and White, J., Four-dimensional imaging: computer visualization of 3D movements in living specimens, Science, 273:603–607, 1996.
[17] Watts, J.L., Etemad-Moghadam, B., Guo, S., Boyd, L., Draper, BW., Mello, C.C., Priess, J.R.,
and Kemphues, K.J., par-6, a gene involved in the establishment of asymmetry in early C. elegans
embryos, mediates the asymmetric localization of PAR-3, Development, 122:3133–3140, 1996.
[18] Yasuda, T., Bannai, H., Onami, S., Miyano, S., and Kitano, H., Towards automatic construction
of cell-lineage of C. elegans from Nomarski DIC microscope images, Genome Inform., 10:144–154,
1999.