|Home | About | Journals | Submit | Contact Us | Français|
Drosophila ventral furrow formation has frequently been used as a model to study developmentally-regulated cell-shape changes. However, a technique to follow all cellular changes during this process within a single living embryo has been lacking. We describe a novel technique, called “end-on imaging”, to collect time-lapse images of transversely-mounted living embryos. End-on imaging revealed several new features of dorsoventral development. First, we observed a wave of syncytial nuclear divisions predicting the location of the ventral furrow. Second, we determined that there is a five-minute gap between the end of cellularization and the start of ventral furrow formation, suggesting that the two processes may share the same pool of cytoskeletal components. Lastly, we show that apical-membrane flattening, the first step in ventral furrow formation, is due to the ventral cells pushing against the vitelline membrane, rather than flattening the dome-shaped, apical surfaces of these cells by a pulling or constriction motion.
Gastrulation is composed of several morphogenetic events that require coordinated cell shape changes to generate the embryonic germ layers. The first morphogenetic event during Drosophila gastrulation is ventral furrow formation (VFF) where a swath of columnar epithelial cells become wedge-shaped and drive the invagination that produces the mesoderm (Leptin and Grunewald, 1990; Kam et al., 1991; Sweeton et al., 1991). Genetic analysis of VFF has revealed many components of the signaling pathway responsible for the initiation of the shape changes (Thisse et al., 1988; Moussian and Roth, 2005; Padash Barmchi et al., 2005). VFF cell-shape changes can be broken down into four morphogenetic components: (1) apical-membrane flattening, (2) apical-to-basal nuclear migration, (3) apical constriction and (4) apical-to-basal cell shortening (Leptin and Grunewald, 1990). These highly dynamic cell-shape changes take place primarily along the apicobasal cell axis.
Two of the most widely used methods to analyze cell-shape changes during VFF include fixed preparations of transversely sectioned embryos (Leptin and Grunewald, 1990; Dawes-Hoang et al., 2005), and three-dimensional time-lapse microscopy (Kam et al., 1991; Oda and Tsukita, 2001; Padash Barmchi et al., 2005). These approaches have been indispensable for characterizing the roles of the signaling molecules responsible for initiating VFF and describing the associated cell-shape changes. While transversely sectioned embryos provide adequate resolution to study these apicobasal cell-shape changes, the temporal sequence of events of this dynamic process must be inferred from numerous different embryos. Time-lapse imaging of ventrally-mounted embryos provides precise temporal information, but gathering sufficient apicobasal image data at high resolution by optical sectioning is limited by the axial resolution of the microscope and the imaging rate, which is slow relative to the rate of VFF.
To understand the connection between the signaling events and the mechanical mediators that drive VFF cell-shape changes, we developed a technique to capture time-lapse images of transversely mounted embryos. We refer to this new technique as “end-on imaging” because the embryos are imaged through their posterior ends. With end-on imaging, all apicobasal cell-shape changes along the dorsoventral axis can be imaged simultaneously. This technique provides the optimal perspective to obtain the information needed to fill the gaps in our understanding of how the signaling molecules that determine a cell's fate impinge on the cytoskeletal regulators that change the cell's shape.
In this report, we show that end-on imaging provides new insights into the events leading up to the invagination of the ventral furrow. First, a wave of syncytial nuclear divisions that predicts the location of the ventral furrow was observed using embryos that express nuclear-localized GFP (nGFP). Using GFP-tagged myosin II regulatory light chain (Spaghetti squash-GFP, Sqh-GFP), we measured the interval between the disappearance of myosin from the basal surfaces of cells completing cellularization and its reappearance at the apical surfaces of ventral cells. The duration of this interval indicates the same pool of myosin may be used for both processes. Lastly, using intervitelline injection of quantum dots, we show a decrease in the space between the apical surfaces of the ventral cells and the vitelline membrane during apical flattening, suggesting new models for the mechanism underlying this first step in VFF.
To visualize the apicobasal events that drive VFF, living embryos were transversely mounted by inserting them lengthwise into cylindrical wells molded into a thin polyacrylamide slab mounted on a glass coverslip. The embryos were covered with halocarbon oil and a clear Teflon membrane (Fig. 1). The oil prevents desiccation of the embryos and gel, while the Teflon membrane provides stability and allows ample aeration for development to proceed normally. The polyacrylamide slabs were prepared by pouring an acrylamide solution into a square, embedding mold and inserting a comb composed of twenty (two rows of 10) stainless steel pins. Once polymerized, the gel block was cut into slices such that each slab has a 2 × 10 array of cylindrical holes perpendicular to the surface of the slices (details provided in the Experimental Procedures section). The array of wells allows one to record the development of up to 20 embryos in a single session, which is essential for the analysis of zygotic mutants, where only one in four embryos will be mutant. This method allows one to visualize all apicobasal cell-shape changes around the dorsoventral circumference of embryos using a variety of fluorescent reporters.
To test the limits of end-on imaging, both single-photon and two-photon, laser-scanning, confocal microscopy were used to determine the range of focal planes at which individual cells expressing nGFP could be resolved. High resolution images were captured up to 65 μm using single-photon imaging from the posterior pole of the embryo, and up to 80 μm using two-photon imaging (Supplemental Fig. 1). At this focal plane, nuclei were seen even after they had entered the yolky interior of the embryo. At 50% egg length (approximately 250 μm from the posterior end), it was still possible to see a ventral furrow form, though the nuclei disappeared once they entered the interior of the embryo because of the light scattering properties of the yolk. Individual nuclei at the cortex were seen through the entire length of the embryo, though the resolution was dramatically decreased and there was shadowing over the part of the embryo closest to the gel as the focal plane was focused deeper into the embryo. It was possible to image nuclei all along the cortex without passing through the dense yolk, explaining why it was possible to image deeper into these preparations than in previous studies that used ventrally mounted embryos (Kam et al., 1991; Nikolaidou and Barrett, 2004).
End-on imaging can also be employed to study a host of other processes, such as mesoderm migration, pole cell formation and migration, head involution, brain development, and cell death and engulfment. Two-photon laser-scanning microscopy allows for the study of other processes that take place deeper in the embryo, such as the formation of the amnioserosa. It is also possible to use the prefabricated wells to study the development of other species such as Xenopus or zebrafish. This mounting method can be adapted to study any relatively small organism that needs to be immobilized in a specific orientation by using different size and shape molds for forming the mounting wells. These molds can be manufactured using pins of different dimensions as outlined above or by micro-machining intricately shaped molds.
The Toll signaling pathway, which directs VFF, is known to be activated more than an hour before VFF, during the syncytial nuclear divisions. Evidence of this early activation includes the nuclear localization of Dorsal, ventral expression of Twist and Snail, accelerated cellularization in ventral cells and the detection of proteomic differences between ventral and lateral cells prior to cellularization (Rickoll, 1976; Leptin and Grunewald, 1990; Gong et al., 2004; Moussian and Roth, 2005). We examined syncytial nuclear divisions to determine if they were also affected in ventral nuclei, compared to the other nuclei around the periphery. End-on imaging of nGFP-expressing embryos revealed a wave of nuclear divisions that propagated from the ventral nuclei around the periphery to the dorsal nuclei in nuclear cycles 11 to 13 (cycle 13 is shown in Fig. 2, Supplemental Movie 1). Nuclei entering mitosis lose their defined appearance as the nGFP diffuses into the cytoplasm during nuclear membrane breakdown. The wave of nuclear divisions was also seen in Sqh-GFP embryos by following the exclusion of the cytoplasmic myosin from the nuclear space during interphase (data not shown). The initiation point of this mitotic wave reliably predicted ±25° of where the ventral furrow would form in 90% (n=19) of wild-type embryos. This early VFF phenotype will be useful in identifying and analyzing mutations that affect VFF.
Ventralized embryos laid by Toll10B mutant females, where all cells express constitutively activated Toll receptor, cellularize uniformly. To test if the wave of syncytial nuclear divisions was also affected in ventralized embryos, Toll10B mutant embryos expressing nGFP were imaged (Fig. 3, Supplemental Movie 2). Three quarters of the Toll10B embryos analyzed by end-on imaging (17/22) had detectable syncytial mitotic waves. Seventy percent of the Toll10B embryos that had detectable mitotic waves (12/17) did not predict the location of the presumptive ventral furrow. The expected location of the ventral furrow was based on the location of the pole cells, which are typically seen directly opposite the ventral furrow in the amnioproctodeal invagination in about 90% of wild-type embryos (n=60). Of the Toll10B embryos where the wave did coincide with ventral location, the origin of the wave was much wider (about 1/3 of the embryo's circumference) than in wild-type embryos. The remaining Toll10B embryos analyzed (5/22) had either uniformly synchronous nuclear divisions or the amnioproctodeal invagination failed to form. These data show that the Toll signaling pathway affects many aspects of cellular behavior, including the nuclear syncytial divisions.
Given that ventral cell fate is established well before the start of cellularization, why don't some morphogenetic aspects of VFF occur during syncytial divisions and cellularization? One possible explanation is that the processes of syncytial division, cellularization, and VFF share the same cytoskeletal components, such as myosin II and actin. Thus, the completion of the syncytial divisions and cellularization is a prerequisite for VFF. It is known that the cellularization of ventral nuclei starts and finishes before that of other cells in the embryo, priming the ventral cells to begin morphogenetic movements as the other cells complete cellularization (Rickoll, 1976; Leptin and Grunewald, 1990; Hunter and Wieschaus, 2000). Additionally, during the last step of cellularization, myosin II and actin are required to form a contractile ring to pinch off the basal end of the nascent cells (Young et al., 1991; Royou et al., 2004). Likewise, myosin II is known to be apically activated during the initial stages of VFF (Young et al., 1991; Fox and Peifer, 2007). End-on imaging may help distinguish between two extreme hypotheses where (1) separate apical and basal pools of actin and myosin II are independently regulated, or (2) a single, shared pool of these proteins drives cellularization and VFF sequentially.
As an initial test of these hypotheses, we asked if there is a lag between the completion of cellularization and VFF, and if this lag is sufficient for the largest of these molecules, myosin II, to traverse from the basal end of the cell to the apical end. Embryos expressing Sqh-GFP were imaged using the same parameters discussed above (Fig. 4, Supplemental Movie 3). During cellularization, small puncta of Sqh-GFP were seen migrating toward and associating with the cellularization furrow, agreeing with a previous report (Royou et al., 2004). Rings of Sqh-GFP were seen during both the slow and fast phases of cellularization, but disappeared upon completion of cellularization (Fig. 4A-C). There was a five-minute gap between this basal disappearance and the apical reappearance of Sqh-GFP within ventral cells at 23°C (Fig. 4P, the time points between the arrows). Given the size of folded, inactive myosin II (546 kD and 10 S) and the viscosity of the cytoplasm (ηcyto/ηwater ≈ 3), it is estimated to take at least 80 seconds for folded myosin II to diffuse the length of the cell (~30 μm). This diffusion rate will surely be slower due to the sieving effect of filamentous and membranous structures in the cytoplasm and protein-protein interactions. These data suggest that cellularization and VFF may share cytoskeletal components, acting as a rate- or material-limiting step to regulate the timing of the two events. Further experimentation using end-on imaging coupled to photo-activation and photo-bleaching will be needed to test this diffusion hypothesis.
The first event of VFF is apical-membrane flattening. The current models suggest that the dome-shaped apical surfaces of the ventral cells are pulled away from the vitelline membrane as they flatten. If this is the case, there should be an increase in space between the cells and the vitelline membrane just prior to the invagination caused by apical constriction. Immediately after apical-membrane flattening, nuclei migrate basally and then the apical-membranes constrict.
To characterize the first two morphogenetic events of VFF, apical-membrane flattening and apical-to-basal nuclear migration, quantum dots (655 nm emission) were injected into the intervitelline space of embryos expressing nGFP. During cellularization, nuclei were seen around the periphery of the embryo, and the intervitelline space was visualized as a red-fluorescent signal between the cells and the vitelline membrane (Fig. 5A-F, Supplemental Movie 4). Over a five-minute period following the completion of cellularization, the ventral-most nuclei first migrated apically, and the apical membranes of ventral cells began to flatten (Fig. 5G-J and P, the time points between the arrows). This is the first time the apical migration of nuclei has been seen. Instead of observing a gain of fluorescence between the apical surfaces of ventral cells and the vitelline membrane as predicted if the apices of cells were pulling away from the vitelline membrane, a loss of intervitelline fluorescence was seen above the ventral cells, indicating a reduction in intervitelline space. The apical surface of ventral cells appeared to touch the vitelline membrane (Fig. 5G-J and P, the time points between the arrows). This did not occur in lateral cells, as fluorescence was consistently seen between their apical surfaces and the vitelline membrane. Kymographs comparing ventral cells and lateral cells clearly illustrate the difference between ventral and lateral intervitelline volume during VFF (Fig. 5P and Q).
There are several mechanisms by which the ventral cells could push against the vitelline membrane to flatten their apical surfaces. First, the ventral cells may be pushing each other or may be pushed by the lateral cells toward the vitelline membrane. These mechanisms take into account that the nuclei appear to be moving in the same direction as the apical surface, indicating whole cell migration. Second, as proposed by Wieschaus and colleagues (Dawes-Hoang et al., 2005), the apical adherens junctions may be drawn up to the apical surface. This may extend the apical surface to the vitelline membrane; however, it would require the addition of plasma membrane to extend to the vitelline membrane. Additionally, it could be that the ventral cells are swelling due to fluid uptake. Further experimentation will be needed to distinguish between these models.
Following apical-membrane flattening and the initial apical migration of ventral nuclei, the nuclei migrated basally (Fig. 5K). Next, the apical membranes of the ventral-most cells constricted, causing a large increase in intervitelline fluorescence between the apical surfaces of ventral cells and the vitelline membrane (Fig. 5L). As the apical membranes constricted further, the ventral furrow invaginated, which was seen as the ventral-most nuclei collapsed toward the center of the embryo and disappeared as they enter mitosis (Fig. 5M-N). At this focal plane it was also possible to see the amnioproctodeal invagination (Fig. 5G-O, arrowheads) and pole cells (Fig. 5L-O, within amnioproctodeal invagination). Individual images from these time series are very similar to those in published reports of transversely sectioned wild-type embryos (Leptin and Grunewald, 1990; Sweeton et al., 1991), however, these data provide much needed temporal resolution of these events and have revealed novel insights into how the first two cell-shape changes occur. The ventral nuclei initially migrate apically before their basal migration and the apical membranes of ventral cells appear to flatten by pushing into the vitelline membrane. These finer details should help us to understand how the shape changes are coordinated with one another.
To study the apicobasal cell-shape changes that occur during VFF in Drosophila embryos, we developed a technique that allows one to image along the dorsoventral axis and capture the morphological changes as they occur in vivo. Using this technique, we have shown that it is possible to view cell-shape changes and protein localization simultaneously in ventral, lateral and dorsal cells and to follow multiple processes: syncytial nuclear divisions, cellularization, VFF and amnioproctodeal invagination, within a single embryo. Quantifying the temporal sequence of events and the coordination of cell-shape changes has led to new insights into VFF. End-on imaging provides a unique perspective on dorsoventral development in Drosophila embryos and should serve to accelerate our understanding of complex morphogenetic events that occur along this axis.
The stocks of Drosophila melanogaster used in this study were yw,SqhAX3;Sqh-gfp42 for myosin II localization (Bill Sullivan, U. C. Santa Cruz) and yw;p(ubi-GFP.nls)ID-2;p(ubi-GFP.nls)ID-3 for nuclear migration (Bloomington Stock Center, 1691). The Tl10Bmwh,e/TM3,e,Sb,Ser/T(1:3)OR60 stock was used for constitutively active Toll receptor (Kathryn Anderson, Memorial Sloan-Kettering Cancer Center).
Embryos were collected on yeasted, apple-juice agar plates over periods of 2-3 hours. The embryos were dechorionated and viewed under a dissecting microscope (Wild) using transmitted light for staging according to Campos-Ortega and Hartenstein (1985). Embryos staged at the syncytial blastoderm (cycles 10-13) were selected for imaging. For end-on imaging, embryos were placed into holes within the polyacrylamide gel of the chamber and imaged as described under end-on technique.
Embryos to be injected were mounted on a coverslip as previously described (Gong et al., 2004). They were then dehydrated slightly, covered with halocarbon oil and injected with quantum dots (amino-PEG 655 nm emitting quantum dots, provided by Marcel Bruchez, Carnegie Mellon University). For end-on imaging, the injected embryos were removed from the coverslip by gently rolling the injection needle under them. They were then placed into holes within the polyacrylamide gel of the end-on chamber.
To obtain in vivo images of Drosophila embryo development along the anterior-posterior axis, we created a polyacrylamide gel mold that held embryos within cylindrical wells so that they were standing on their posterior ends. A gel block was created by pouring a 15% acrylamide solution (2.4 ml dH2O, 2.5 ml 0.5 M Tris, pH6.8, 30% Acrylamide/Bis solution, 100 μl 10% APS, 4 μl TEMED) into a square peel-away disposable embedding mold (Catalog #70182, Electron Microscopy Sciences). The cylindrical wells were created by inserting a comb composed of stainless steel pins (0.178 mm or 0.1905 mm diameter, Small Parts, Inc.) into the acrylamide solution. Once the acrylamide had cured, the mold was peeled away and the comb was removed while the gel was submerged in water to prevent tearing of the wells. The gel block was then rinsed and stored under water until sliced. The gel block was sliced to 400 μm thick with a vibratome (Leica, VT 1000S) at speed 7 and vibration frequency 9. The gel slabs were stored under water until use.
The comb was created by placing 2 cm long stainless steel pins (Small Parts, Inc. GWX-0070-60 or GWX-0075-60) onto a piece of double sided tape along the long edge of a glass slide so that the pins were perpendicular to the edge of the slide and 1.2 cm of the pin extended beyond the slide. The 2 cm pins were separated by four 0.5 cm long pins placed parallel to the 2 cm pins. Ten pins were placed on each side of the slide to create a 2 × 10 grid. A 2.5% agarose gel solution was poured into a square embedding mold to approximately 3 mm from the top of the mold. The comb was inserted into the agarose solution before the gel cured. Dental cement (A-M Systems, Catalog Numbers 525000, 526000) was poured over the solid gel block. The comb was removed from the gel after the cement solidified.
To setup the chamber for imaging, two pieces of double-sided tape were placed at either end of a cover glass (No. 1, 24 × 60 mm, Corning). Two dental wax strips were placed inside of each piece of tape. The gel slab was cut to fit onto the cover glass and placed between the two pieces of wax. The gel slab was covered with Halocarbon oil (700 series) to prevent desiccation of both the gel and the embryos. Embryos were inserted into the holes with their posterior ends down using blunted probes. The wax strips were then removed and a clear, oxygen permeable Teflon membrane (Yellow Springs Instrument Inc.) was firmly pulled over the gel to keep the embryos from floating during imaging (Fig. 1). The cover glass (Teflon membrane side) was placed onto a metal frame to provide stability to the cover glass. The cover glass (glass side) was placed onto a stage over a 63× glycerin (1.3 NA), 40× oil (1.3 NA), or 40× water (1.1 NA) objective of an inverted microscope. Time-lapse microscopy was performed using a Leica TCS-SP5 or a Zeiss LSM 510 Meta NLO confocal microscope system with an argon laser at 488 nm. Time-lapse recordings consisted of a single optical section focused at 50-75 μm in from the posterior end of the embryo taken at 30 second intervals over a period of 3-4 hours at 23°C. Studies to determine the full depth of imaging were done on a Zeiss LSM 510 Meta NLO confocal microscope using a Coherent Chameleon Ultra II femtosecond laser system operating at 860 nm.
Images were collected as either 8 bit or 16 bit data sets and normalized using ImageJ such that the mean for the highest pixel value was set to 255 (8 bit) and the lowest pixel value was kept at 0. This was done for each channel in RGB images. Kymographs were made by selecting an area that was 2 cells wide, spanning from the vitelline membrane to the center of the embryo. This area was then cropped and a single row montage of the stack was created using ImageJ. Two-photon images were despeckled using ImageJ software.
Supplemental Figure 1. VFF at different planes within an embryo using multi-photon imaging. A, D, G: Cellularizing blastoderm. B, E, H: Beginning of apical constriction. C, F, I: Sealing of the VF. Images were collected at 40 μm from the posterior end of the embryo for A-C, 80 μm from the posterior end of the embryo for D-F, and 120 μm from the posterior end of the embryo for G-I. Images were collected at 24°C. Dorsal is up, ventral is down. VF marks the ventral furrow. These images were collected using a 40× water objective. In optical sections deeper than 80 μm, a shadow was observed over part of the embryo closest to the gel. This shadow may be due to IR absorption by the polyacrylamide gel. Scale bar = 50 μm.
Supplemental Movie 1. A wave of syncytial nuclear divisions predicts VF location. Ventral nuclei (bottom of images) are the first to divide, and dorsal nuclei (top) are the last. Images were collected at 1 frame/30 seconds. Time-lapse recording shows a single optical section focused at 50-60 μm in from the posterior end of an embryo using a 63× glycerin objective.
Syncytial nuclear divisions do not predict VF location in ventralized, Toll10B embryos. Dorsal nuclei (top of images) are the first to divide, and ventral nuclei (bottom) are the last. The amnioproctodeal invagination, containing the pole cells, is seen at the top of this image. The ventral furrow would have formed at the bottom of the panel. Images were collected at 1 frame/40 seconds. Time-lapse recording shows a single optical section focused at 50-60 μm in from the posterior end of an embryo using a 63× glycerin objective.
Myosin II localization during cellularization and VFF. Sqh-GFP is seen along the cellularization front. Basal sqh-GFP disappears upon completion of cellularization. After about 5-6 minutes, myosin is seen apically in the ventral cells (bottom of images). Images were collected at 1 frame/30 seconds. Time-lapse recording shows a single optical section focused at 60-70 μm in from the posterior end of an embryo using a 40× oil objective.
Nuclear migration and apical flattening in wild-type VFF. Ventral nuclei (bottom of images) first migrate apically, then migrate basally as the VF forms. The space between the apical surfaces of the ventral cells and the vitelline membrane decreases just before VFF. (See also Fig. 2) Nuclei are green. Quantum dots (red) fill the intervitelline space. Images were collected at 1 frame/30 seconds. Time-lapse recording shows a single optical section focused at 60-70 μm in from the posterior end of an embryo using a 40× oil objective.
We would like to thank members of the Minden lab for critical reading of this manuscript. The microscopes used in this report are part of an imaging facility funded by grants from the NIH (S10-RR024716) and the Gordon Moore Foundation. We also thank Dr. Stefan Zappe for the use of his Leica TCS SP5 microscope. JAJF acknowledges support through the Pittsburgh Technology Center for Networks and Pathways (5U54-RR022241) funded under the NIH Roadmap for Biomedical Research.