|Home | About | Journals | Submit | Contact Us | Français|
This protocol describes imaging and computational tools to collect and analyze live imaging data of embryonic cell migration. Our five step protocol requires a few weeks to move through embryo preparation and four-dimensional (4D) live imaging using multiphoton microscopy, to 3D cell-tracking using image processing, registration of tracking data, and their quantitative analysis using computational tools. It uses commercially available equipment, and requires expertise in microscopy and programming that is appropriate for a biology laboratory. Custom-made scripts are provided, as well as sample datasets to permit readers without experimental data to perform the analysis. The protocol has offered new insights into the genetic control of cell migration during Drosophila gastrulation. With simple changes, this systematic analysis could be applied to any developing system to define cell positions in accordance with the body plan, to decompose complex 3D movements, and to quantify the collective nature of cell migration.
The combination of advanced imaging and image analysis techniques enables the investigation of large, dynamic cell populations within a developing embryo1,2. These imaging approaches provide a unique opportunity to study embryonic morphogenesis from the level of cellular processes to the scale of an entire tissue or organism. Gastrulation in the Drosophila melanogaster embryo is an excellent model system for the study of embryonic morphogenesis3. In less than two hours of development, ~6000 cells undergo stereotypical morphogenetic events, such as tissue invagination4, convergence-extension5,6, planar cell intercalation5,6, radial cell intercalation1, epithelial-to-mesenchymal transition7, synchronized waves of cell division1, and collective cell migration1. Although the geometry of the Drosophila embryo is relatively simple at early stages of development, the morphogenetic events involve highly dynamic processes and complex 3D movements of cells that prevent a complete investigation of most wild-type or mutant phenotypes based on the analysis of fixed embryos.
This protocol presents the quantitative imaging of complex cell migration in vivo, using mesoderm cell spreading during Drosophila gastrulation as a model system. The experimental strategy combines 4D in vivo imaging using 2-photon excited fluorescence (2PEF) microscopy, 3D-cell tracking using image processing, and automated analysis of cell trajectories using computational tools. This quantitative approach decomposes 3D cell movements, generating a precise description of morphogenetic events. Furthermore, this protocol describes the quantitative investigation of the collective nature of mesoderm cell migration. The reproducibility of morphogenetic events among wild-type embryos can be tested and mutant phenotypes can be dynamically analyzed. This approach provides a method to study complex or even subtle mutant phenotypes, such as the ability to distinguish cell populations that exhibit different behaviors1. We recently applied this approach to gain insights into the control of cell migration during mesoderm formation in Drosophila embryos1.
The experimental workflow is divided into five main parts (Fig. 1): the embryo preparation (steps 1–9), the 4D-imaging (steps 10–15), the 3D-cell tracking (steps 16–22), the tracking data registration (steps 23–27), and the tracking data analysis (step 28). Flies containing a fluorescent reporter are mated and embryos are collected. The chorion is removed and the embryos are mounted for live imaging and 4D image dataset acquisition using a 2PEF microscope. Typically ~2,000 mesoderm and ectoderm cells moving through the field of view are imaged during 2–3 hours of development. Each imaging dataset contains ~109 voxels and is processed using Imaris software to track the trajectories of the cell collection. Finally, a quantitative and automated analysis of the cell trajectories is performed using Matlab. Customized Matlab scripts required to perform steps 23–28 are provided in the supplemental section of this protocol (Supplementary Data 1). A sample dataset is also provided to allow readers to start the procedure at step 23 without having to collect experimental data (Supplementary Data 2). This protocol can be directly applied to study mesoderm spreading in gastrulating Drosophila embryos. However, the workflow is not specific to this particular stage or model system. In addition, each part described in Fig. 1 can be used independently and included into a different working strategy. In order to facilitate the adaptation of this protocol to other stages or model organisms, we discuss below each part of the workflow with general comments and advice that are summarized in Table 1. The specific experimental choices made to study Drosophila gastrulation are clearly indicated.
A critical component of this protocol is the choice of the fluorescent reporter, as this reporter must be suitable both for high quality imaging and cell movement quantification. To this end, fluorescent labeling of nuclei provides several advantages: (i) the nuclei are easier to segment and track from 4D image datasets than other cellular structures, such as membranes; (ii) the spatial position of the segmented nucleus can directly define the spatial position of a cell for cell movement analysis; (iii) nuclear fluorescent labeling provides a direct indicator of cell division; and (iv) transgenic lines of Drosophila with a strong, stable, and ubiquitous expression of fluorescent protein fused with histone or nuclear localization sequence are available (see Reagents section and Bloomington Stock Center, for instance). The lines expressing in-frame fusions of GFP to a nuclear localization sequence (NLS) have the disadvantage of producing a diffuse fluorescent signal each time the nuclear envelope breaks down during each cell division (Supplementary Movie 1). In this protocol, we used the transgenic line expressing GFP fused with Histone 2A available from Bloomington Stock Center (see Reagents section). The fluorescent Histone remains associated with the chromosomes even during nuclear envelope breakdown, giving an unambiguous signal for tracking1.
The scattering of light inside the biological sample is usually the factor limiting the depth of imaging. The scattering property of the embryonic tissue is developmental stage- and species-dependant (Box 1 and Fig. 2). During Drosophila gastrulation, a high density of sub-micrometer scale refractive vesicles, mostly lipid droplets, are observed in cells and at the surface of the yolk8. These lipid bodies are strong light scatterers, which results in the high scattering property of early embryos and prevents deep tissue imaging. The distribution of these lipid bodies is altered in klarsicht (klar) mutants: the lack of Klar in these embryos prevents the apical redistribution of lipid bodies at the end of cellularization, yet the homozygous mutants are viable9. As a result, klar cells appear more transparent than wild-type during gastrulation (Fig. 2a–b). We compared the optical properties of wild-type and klar embryos at stage 8 (stages defined by10) by measuring the scattering mean free path, lexs, of the near-infrared (NIR) light (Box 1 and Fig. 2c). lexs is ~56μm in wild-type embryos (blue in Fig. 2d) and ~76μm in klar embryos (red in Fig. 2d). From these measurements, the typical 2PEF signal decay depending on the depth of imaging can be plotted (Fig. 2e, see details in Box 1). It shows that the higher value of lexs in klar compared to wild-type embryos is sufficient to double the intensity of 2PEF signal recorded at 80 μm depth (compare blue and red curves in Fig. 2e). In this protocol, we used the klar background to improve the imaging depth and the level of signal - two criteria that significantly facilitate image processing. Of note is the fact that we did not observe any disruption of mesoderm migration in klar embryos1, therefore conducting experiments in a klar mutant background provides a good option to improve imaging capabilities.
To show the scattering properties of embryonic tissues and the subsequent limitation of imaging depth are stage- and species-dependant, we plotted the depth-dependent 2PEF signal from stage 5 Drosophila or early zebrafish mesoderm (gray curves in Fig. 2e) based on the previous experimental measurement of the scattering properties (gray in Fig. 2d). The signal decay demonstrates that stage 5 and stage 8 Drosophila embryos (dark gray and blue curves in Fig. 2e, respectively) exhibit significantly different properties, whereas these two stages are separated by only 1 hour of development. In addition, the 2PEF signal at 80 μm is expected to be 5 times weaker in Drosophila at gastrulation (blue curve in Fig. 2e) compared to early zebrafish embryos (light gray in Fig. 2e) for the same labeling and imaging conditions. Hence, the maximum depth of imaging and the choice of the microscopy technique depend on the stage and model system. For instance, as opposed to Drosophila embryos, the imaging of mesoderm structures at 80 μm in early zebrafish embryos is achievable with confocal microscopy and does not require 2PEF microscopy11.
The mounting procedure is a critical step of the embryo preparation for optimized imaging. The use of materials inducing optical aberrations on the optical path, such as agarose gel, should be avoided or limited. In order to enable a proper quantification of cell movements and avoid motion artifacts, the embryos must be precisely oriented and maintained in place during the image acquisition. Furthermore, the mounting of the embryos should not deform the embryo itself (for instance, by squeezing the embryo between coverslips), as this might alter the cell behaviors. In the case of Drosophila embryos, we found that mounting them in water and imaging without an additional coverslip between the specimen and the objective offered the best compromise between embryo health and image quality. This arrangement avoids the refractive index mismatch between embryo and immersion solution that would be present with an oil-immersion objective, prevents embryo hypoxia, and does not induce deformation. The embryos are oriented and maintained in place by gluing them on a coverslip. The orientation is first based on the shape of the embryo: the dorsal side has less curvature than the ventral side (Supplemental Movie 2). The well-oriented embryos are then selected at early stage 6 under the 2PEF microscope10 with the ventral side facing the objective. The onset of ventral furrow formation at stage 6 makes it easy to identify well-oriented embryos: the furrow should face the objective, in the middle of the field of view.
Choosing the appropriate microscopy technique to image living embryos depends on several criteria: the required spatial and temporal resolution, the size or shape of the embryo and volume to image, the sensitivity to phototoxicity, and the optical properties of the tissue. Imaging the early stages of Drosophila gastrulation is limited by two major factors: the light scattering properties of the tissue and the phototoxicity. These limitations are especially apparent when imaging mesoderm formation using confocal microscopy. When using confocal microscopy only half of the required depth is visualized and the required spatio-temporal sampling quickly induces strong phototoxicity (see below). 2PEF microscopy12 and other multiphoton microscopy techniques8, are better choices to support the 4D (3D in space and 1D in time), long-term, deep-tissue imaging of Drosophila embryos in a manner that does not compromise their viability.
In multiphoton microscopy, the sample is illuminated with NIR radiation and the spatial resolution is intrinsically three-dimensional, resulting in: (i) good penetration and low absorption of the excitation light, and (ii) efficient collection of the emitted light, including scattered photons, due to the absence of pinhole. We reported the imaging of internalized mesoderm cells up to a depth of 80μm within the embryo using 2PEF1. Another significant advantage of using NIR radiation, compared to the linear excitation at 488 nm used in standard fluorescence microscopy, is that the nonlinear excitation of GFP can be obtained using a wavelength (see below) inducing a lower background (i.e., auto-fluorescence).
The main limitation of 2PEF microscopy, as with any laser scanning microscopy, is the time of acquisition. Although Drosophila embryonic development is fast, the morphogenetic movements are slow enough to be captured with laser scanning microscopy. However, the acquisition speed becomes a limitation when imaging a large volume of cells while trying to maintain good spatial and temporal sampling. As a consequence and in order to avoid phototoxicity and obtain a signal level and spatio-temporal sampling suitable for proper image analysis, the 2PEF imaging of Drosophila mesoderm cells requires careful adjustment of the imaging parameters (i.e., objective, spatial and temporal sampling, field of view, resting time, laser power, wavelength).
The depth of imaging, the level of fluorescent signal, and the speed of acquisition required for this procedure can easily lead the investigator to use imaging conditions that induce phototoxic effects and prevent the normal development of the imaged embryo. For this reason, it is important to systematically check for any sign of photo-induced effects on movement. The imaging parameters must be carefully tuned in order to stay far away from phototoxic conditions while maintaining sufficient image quality to support the subsequent image processing steps. Though the molecular mechanisms resulting in phototoxicity in 2PEF microscopy are not fully understood, phototoxic processes usually appear to be highly nonlinear13, 14: meaning that the threshold is sharp and that small changes in imaging parameters are enough to switch from toxic to non-toxic conditions.
Several criteria can be used to identify phototoxic effects in Drosophila during gastrulation. The level of endogenous fluorescent signal (also called autofluorescence) is often a good indicator. If the endogenous signal from the yolk or the vitelline membrane begins to approach the level of the GFP fluorescent signal, it indicates that the imaging conditions will most likely induce phototoxicity. In this case, a different GFP labeling and/or a different excitation wavelength should be used. The cell movements can indicate phototoxicity: if these movements slow down independently of the temperature and specifically within the field of view, it is a clear effect of phototoxicity. Finally, it is possible to observe more subtle effects at low laser power level, including changes affecting cell division rates. Cell divisions occurring a few minutes earlier or later than normal induce a disruption of the cell division pattern that can be quantified1. We interpret this effect as a mild disruption of cytoskeleton dynamics. Lastly, it is important to note that phototoxic effects may result long before any photo-bleaching is induced. Hence, the mere absence of photobleaching is not a good indicator of non-invasiveness.
For the deep-tissue imaging of highly scattering tissue using 2PEF microscopy, the ideal objective must have a large working distance, a high numerical aperture (NA), a low magnification, and good transmission of NIR light. The large working distance prevents embryo hypoxia and allows deep-tissue imaging. The high NA improves the spatial resolution, the 2-photon excitation, and the light collection efficiency. The low magnification allows image acquisition from a large area, which significantly improves 2PEF signal collection efficiency15. For this procedure, we used a 40x water immersion objective with 1.1 NA and working distance of 600μm.
The choice of the excitation wavelength is critical to obtain an efficient fluorophore excitation, a low endogeneous signal (background), and low phototoxicity. Use of a tunable femtosecond laser allows the user to test different wavelengths and choose the best compromise. When imaging GFP, the optimal 2-photon excitation wavelength is 940–950nm. We observed that in gastrulating Drosophila embryos, the use of lower wavelengths leads to higher phototoxicity, lower GFP excitation efficiency, as well as higher levels of endogenous fluorescent signal. Consequently, in this case, the absorption of water in the 950nm wavelength range does not play a significant role in the phototoxicity.
In most techniques of fluorescence microscopy, such as confocal microscopy, only the ballistic photons that are not scattered from the emission spot en route to the detector contribute to the fluorescent signal. As the fluorescence excitation is restricted to the focal volume in 2PEF microscopy, every emitted photons can contribute to the signal, including scattered photons. In practice, it means that the signal collected from scattering tissue can be improved by collecting light in every spatial direction. For instance, the 2PEF signal can be collected in both the trans- and epi-direction if the microscope setup permits it. In our case, we added a silver mirror in the trans-direction, which reflects forward-directed photons and contributes to collection of some of them by the objective in the epi-direction. This straightforward procedure allowed us to collect up to 30% more 2PEF signal with the same illumination conditions, thus significantly improving the image quality and facilitating the image processing steps.
The spatial resolution has to be sufficient for the proper segmentation of nuclei. Even if the tracked objects are large (nuclei are of ~5–10 μm diameter), the gap between them can be small (< 2 μm). As a result, a high NA objective is required, especially for the segmentation of nuclei located deep within the embryo. A spatial sampling of 0.5 μm per pixel in x,y direction and 1 μm in z appears sufficient.
The time resolution is critical in order to ensure error-free cell tracking, and to avoid the incorrect assignment of cell identities due to temporal aliasing. Temporal aliasing occurs when 3D stacks of images are acquired with a time interval between two frames too large to permit faithful cell tracking. Indistinguishable nuclei travel with a velocity v and are separated by a distance d. When images are acquired with a time interval Δt between two stacks, for the nuclear trajectory to be extracted unambiguously the distance v.Δt travelled by the cell in between two stacks must be less than half the distance d (i.e., v.Δt < d/2)16. In our case, as v~5μm.min-1 and d~10μm, thus the requirement is that Δt < 1 min. We used Δt = 45–50 seconds.
Image processing techniques other than cell tracking have been successfully applied to quantify morphogenetic movements in embryos. For instance, image cross-correlation velocimetry12,17,18 isspecifically adapted to measure tissue deformation by direct differential analysis of the estimated velocity field12. However, the spatial resolution is limited by the size of the image interrogation window and this approach is usually limited to 2D. Cell tracking based on the segmentation and tracking of nuclei provides an opportunity to follow the behavior of individual cells in 3D with good spatial and temporal resolution (Fig. 3).
The quality of the image dataset is critical for the proper tracking of cell movements; any slight improvement of this dataset can drastically improve the image processing. For instance, as discussed above, the signal level as well as the spatial and temporal resolution is critical for proper nuclear segmentation and tracking (Table 1).
We choose to use Imaris software to perform 3D-cell tracking for several reasons. First, the user interface and the 3D visualization of the imaging dataset are extremely efficient. The cell tracks can be visualized, checked and manually corrected using the tracking editor (provided in version 5.7). The Imaris XT interface with Matlab improves the functionality of the software without extensive knowledge of computer programming: for instance, the data can be exported into Matlab for further analysis. Together, it appears to be a good compromise option, as it combines the user-friendly interface and standard analysis of commercial software with sufficient flexibility that the user can customize the tools for their applications without the need to write a completely custom software package. Because an improved background knowledge of Imaris software and its functionalities can drastically reduce the time spent performing 3D-cell tracking of a large dataset, users should consider obtaining experience with the software from Bitplane through user-training sessions (contact Bitplane customer service for details).
This protocol describes the tracking of two cell populations during Drosophila gastrulation: mesoderm and ectoderm cells. These two groups are defined by sorting the cell trajectories using Imaris functions. The mesoderm cells are those that have invaginated and the ectoderm cells stay at the surface of the embryo. A few midline cells (a sub-population of the ectoderm) are independently tracked and their trajectories are used for spatial registration (see below). The tracking of mesoderm and midline cells is carefully checked so that the trajectories span the entire time sequence.
The registration is an important step including any spatial or temporal transformation of the datasets that enables their comparison from one experiment to the other. This protocol describes three types of data registration: the correction of motion artifacts, the transformation of the adapted spatial coordinate system, and the synchronization of imaging sequences (Table 1 and Fig. 4–5).
In image analysis, different methods of registration exist. For instance, the distribution of specific markers in the sample can be used to correct its drift during time of acquisition (landmark-based spatial registration), or the voxel values of an image sequence can be used to synchronize several datasets (voxel-based temporal registration19). In this procedure, we used the segmented objects themselves to perform both spatial and temporal registration in a fully quantitative and automated manner. For this reason, the registration is performed after the 3D-cell tracking. Under some experimental conditions, spatial registration has to be done before 3D-cell tracking; for instance, strong motion artifacts during the image acquisition (embryo rolling, sample or stage drift, etc) can degrade the cell tracking process.
In this protocol, the spatial registration includes the definition of cell positions in accordance with the body plan. The choice of a spatial coordinate system adapted to the geometry of the tissue or embryo enable the user to investigate complex cell movements in 3D by decomposing their trajectories into components that have a biological meaning. The appropriate coordinate system depends on the biological model used: for instance, during early stages of development, a spherical coordinate system is adapted to the shape of zebrafish2 or Xenopus Laevis20, whereas a Cartesian coordinate system remains appropriate for avian embryos18. In the case of Drosophila gastrulation, the embryo has a cylindrical shape in the area where mesoderm spreading occurs (Supplemental Movie 2). The protocol shows first how a cylinder is fitted onto the spatial distribution of ectoderm cells (EctodermCylinderFit.m Matlab script, Supplemental Data 1 and Table 2) in order to identify the anterior-posterior axis of the embryo and to switch from Cartesian (x, y, z) to cylindrical (r, θ, z) coordinate system (Fig. 4). In this coordinate system, the movements in each direction (radial, angular or longitudinal) can be directly compared from one embryo to the other and correspond to specific morphogenetic events1.
The final step of spatial registration is the angular drift correction (Registration.m Matlab script, Supplemental Data 1 and Table 2). During the time of acquisition, the embryo can exhibit some rolling inside its vitelline membrane, corresponding to a solid rotation around the anterior-posterior axis (Supplemental Movies 2–3). This angular drift is corrected by tracking a few cells from the ectoderm midline and defining their angular position at each time point as θ=0 radian (Fig. 5a–c).
The temporal registration corresponds to the synchronization of image sequences based on the occurrence of a specific morphogenetic event (TimeSynchronization.m Matlab script, Supplemental Data 1 and Table 2). We choose the onset of germband extension (GBE)5, 6 as the time reference to synchronize the sequences and define t=0 min (Fig. 5d). At this time, both ectoderm and mesoderm cells start to move toward the posterior direction1.
It is important to notice that the references used for spatial and temporal registration are identical among embryos and are not disrupted in mutants. Hence, they depend on the model system studied. In this protocol, the estimation of the anterior-posterior axis using the shape of the ectoderm layer, the angular reference θ=0 rad using the ectoderm midline cells and the time synchronization based on the onset of GBE are independent of the mesoderm spreading process. In addition, we used these references for registration because they are not disrupted in the mutant we studied1.
Once the tracking data are registered, the cell trajectories can be analyzed directly and compared from one embryo to the other. We provide two examples of tracking data analysis useful for studying complex 3D movements of cell migration and quantifying the collective nature of this process: decomposition of cell trajectories along each cylindrical direction (Fig. 6) using MovementDecomp.m Matlab script (Supplemental Data 1 and Table 2) and mesoderm spreading analysis (Fig. 7) using SpreadingAnalysis.m Matlab script (Supplemental Data 1 and Table 2).
There are number of protocols available to investigate cell migration in tissue culture or in model organisms (see21 for instance). Here we discuss the advantages and specificity of this protocol for studying cell migration in vivo:
(i) The cells are imaged in challenging conditions: they move fast and deep inside a scattering and photo-sensitive embryo. Hence, we describe here an optimized imaging approach.
(ii) Most studies of cell migration are limited to 2D in space and to cells migrating on a fixed substrate; however, inside a living organism, it usually occurs in 3D, with the simultaneous combination of different movements. This protocol shows how to investigate such complex movements in 3D by choosing the appropriate spatial coordinate system and decomposing the cell trajectories into meaningful components. In this study, the mesoderm cells migrate on a moving cell layer (ectoderm): we recently demonstrated how the data generated by this protocol allowed us to investigate the mechanical coupling between two cell layers and to decouple their movements1.
(iii) During embryonic development, cells rarely migrate alone but more often as a collective. The method for tracking a large cell population described in this protocol allows for simultaneous observation of individual and collective behaviors, both of migrating and non-migrating cells. This approach allows the investigator to evaluate migration with a statistical analysis and to identify variability within the cell population1. By following a limited number of cells using techniques such as local photo-activation, one can focus on specific behaviors, but they may not necessarily be representative of the collective.
(iv) Whereas many studies analyze the cell tracking results using a qualitative or manual approach, we provide a quantitative and automated analysis of cell trajectories. In this protocol, the spatial and temporal registration of the data enables the investigator to quantitatively compare one experiment to the other, to test the reproducibility between embryos and to quantify mutant phenotypes1. In addition, the statistical analysis of cell trajectories presented here illustrates how to quantify the collective nature of a cell migration process.
(v) Sophisticated quantitative imaging of cell movements usually involves fully custom-designed approaches that are difficult to implement or modify by other laboratories without strong expertise2. This protocol uses commercially available equipment and software and provides customized Matlab scripts that are annotated and simple enough to be used and modified with minimal expertise. Imaris, the commercial software used to perform 3D-cell tracking is extremely user-friendly; its interface ImarisXT, can be used with classic programming languages and image processing software, such as Matlab or ImageJ, enabling a user with minimum skills in programming to improve the functionality of this software for specific scientific applications. Together, these aspects enable the user to implement, modify, or extend this protocol in a biology laboratory without extensive expertise in microscopy or computer science.
This protocol has two main limitations. First, cell migration is investigated by only tracking the cell nuclei. Although this approach can already generate a lot of biological insights, the analysis of other cell features, such as cell shape changes can be required for specific studies. In the case of mesoderm spreading in Drosophila, the challenging scattering conditions (see above) strongly limit the imaging of structures other than nuclei, such as cell membranes. The second limitation concerns the 3D-cell tracking: the fluorescent signal from the deepest nuclei is weak and their segmentation and tracking requires manual correction. This step, which is not fully automated, limits the number of cells segmented per experiment. For this reason, we limited our application of this protocol to ~100,000 segmented cell positions per embryo (including ectoderm and mesoderm cells)1. To increase this number, further improvement of imaging quality and/or of image segmentation/tracking strategy would be required. The subsequent computer analysis of cell trajectories provided here is automated and is not limited by the cell number.
Dissolve 22 g of sucrose in 350 ml of H2O and pour it into a 1 L bottle. Add 7 g of agarose into this bottle, mix by vigorous shaking. Microwave first for 2 min, and then 2 times for 1 min, mixing the solution in between.
▲ CRITICAL the agarose solution must boil in the microwave.
Put aside to cool to approximately 60 °C. Add 10 ml of ethanol and 5 ml of glacial acetic acid to the solution. Add 50 ml of apple juice and mix well. Pipette into 35 × 10 mm dishes (~60 plates/preparation) using 25 ml plastic pipette or syringe. The plates can be stored in a container at 4°C for weeks.
! Caution Acetic acid is corrosive. Handle with gloves.
! Caution Ethanol is flammable.
Dissolve 30 g of sucrose in 350 ml of H2O and pour it into a 1 L bottle. Add 10 g of agarose to the bottle, mix by vigorous shaking. Microwave first for 2 min, and then 2 times for 1 min, mixing the solution in between.
▲ CRITICAL the agarose solution must boil in the microwave.
Put aside to cool to ~60 °C. Pipette into 60 × 15 mm dishes (~20 plates/preparation) using a 25 ml plastic pipette or syringe. The plates can be stored in a container at 4°C for weeks.
Add short pieces (5–10 cm) of double-sided tape to a 200 ml glass bottle. Add heptane to cover the pieces of tape (typically 50 ml for 50 cm of tape). Gently shake the bottle at least overnight at room temperature (18–25 °C) to dissolve the glue. The heptane-glue bottle can be stored at room temperature for months. Prepare coverslips coated with glue at least 10 min before using them. Add a 60–100 μl droplet of heptane-glue to the middle of each coverslip and allow to dry for 10 min. The coated coverslips can be stored for a few days at room temperature in a box to protect them from dust.
Most of our imaging datasets have been acquired using a Zeiss LSM 510 microscope and a Chameleon Ultra femtosecond laser. However, the protocol can be accomplished with any similar 2PEF microscope. The embryos were imaged using C-Aprochromat 40×/1.1 N.A. W Corr UV-VIS-IR (Carl Zeiss Inc) objective at 940 nm. The non-descanned pathway is used with a single short pass filter (KP680nm) to cut out the laser light. 200×200×80μm3 3D-stacks with 0.5×0.5×1μm3 voxel size and 1.9μs pixel dwell time were acquired every 45–50 seconds for ~3 h.
The Matlab processing requires two freely available toolboxes: the Least Squares Geometric Elements (LSGE) library and the geom3d toolbox. The LSGE library was developed by the Centre for Mathematics and Scientific Computing (National Physical Laboratory, UK) and is available from the EUROMETROS website (http://www.eurometros.org/gen_report.php?category=distributions&pkey=14). The geom3d toolbox was developed by David Legland and is available from Matlab Central website (http://www.mathworks.com/matlabcentral/fileexchange/8002). Download the files, save the “lsge-matlab” and “geom3D” folders and their content on you computer and add both of them in the Matlab path (using “File/Set Path” from the Matlab menu).
Download the Matlab scripts from the supplemental section of this protocol (Supplementary Data 1). Unzip the corresponding file and place all contained files (Imaris2xyzt.m, EctodermCylinderFit.m, TimeSynchronization.m, Registration.m, MovementDecomp.m, SpreadingAnalysis.m, Browse.m, and cart2cyl0.m) in the same folder. The customized Matlab scripts included here are designed and annotated in order to allow the user to run and modify them with only basic knowledge of Matlab programming. However, to further manipulate the data, a working knowledge of Matlab is required. Table 2 provides a list of the scripts and their description.
In order to run the Matlab processing and start the procedure at step 23 without an imaging dataset, we provide sample tracking data files. Download the files from the supplemental section of this protocol (Supplementary Data 2). Unzip the corresponding file and place all of the files (Mesoderm.mat, Ectoderm.mat and Midline.mat) in the same folder as the Matlab scripts.
Steps 1–9, Embryo Preparation: 4 h per set of embryos for imaging.
Steps 10–15, 4D-Imaging: 3 h per imaging acquisition. Repeat steps 1–15 several times to obtain 3–4 correct imaging datasets: ~ 1 week.
Steps 16–22, 3D-Cell Tracking: several weeks per imaging dataset depending on the quality of the dataset and on the efficiency of the user to perform the tracking correction with Imaris.
Steps-23–27, Tracking Data Registration: 1 h maximum per dataset.
Step 28 Tracking Data Analysis: 1 h maximum per dataset.
The imaging and the 3D cell-tracking (steps 1–22) should result in the visualization of mesoderm and ectoderm cell distributions (Fig. 3b and Supplementary Movie 1) and the spreading movement of mesoderm cells during gastrulation (Fig. 3c). The visualization of a 4D imaging dataset is available within published work1.
The results of tracking data registration (steps 23–27) obtained with the sample tracking data provided in supplemental section of this protocol (Supplementary Data 2) are displayed in Fig. 4–5. First, the ectoderm cell positions are fitted onto a cylinder using EctodermCylinderFit.m script (step 25), which displays the distribution of a selected number of ectoderm cells on an estimated cylinder as in Fig. 4c. The time synchronization using TimeSynchronization.m script (step 26) shows the movement of the mesoderm cells toward the posterior direction and estimates the time point at which the onset of movement occurs (Fig. 5d). The Registration.m script (step 27) displays the angular movements of mesoderm cells before and after angular drift correction, as in Fig. 5a–c.
After the tracking data registration, the decomposition of mesoderm cell movements into their cylindrical components r(t), θ(t), and z(t) using MovementDecomp.m script (step 28A) should result in the three graphs of Fig. 6. Each of these graphs corresponds to a specific morphogenetic event: (i) r(t) shows the furrow collapse with the cells moving toward the periphery of the embryo (Fig. 6a); (ii) θ(t) shows the angular spreading of the mesoderm cells with movements toward the left and right directions (Fig. 6b); (iii) z(t) shows the movement of GBE with a concerted movement toward the posterior direction (Fig. 6c).
The analysis of mesoderm spreading using SpreadingAnalysis.m (step 28A) should display the two graphs of Fig. 7c and Fig. 7d. The first graph displays θ(t) for each cell with a color coding for the angular position at the onset of furrow collapse. It shows that the angular distribution of the mesoderm cells is maintained over the two hours of mesoderm spreading (Fig. 7c). The θend(θstart) graph (Fig. 7d), is used to investigate the collective migration of mesoderm cells during their spreading. The position of each cell in this graph corresponds to a specific movement behavior detailed in (Fig. 7a–b). When θend/θstart>1 (white areas in the graphs), the cells are spreading normally. If 0<θend/θstart<1 (light gray areas in the graphs), the cells are not spreading and move in the opposite direction, toward the midline. If θend/θstart<0 (dark gray areas in the graphs), the cells are not spreading, cross the midline and move to the opposite end of the embryo. In wild-type embryos, the cells position in the θend(θstart) graph are mainly distributed in the white area (Fig. 7d). In addition, they tend to be aligned along a specific line: a linear regression gives an estimation of the slope of this line (A) and of the correlation coefficient (R) (Fig. 7d). As previously reported1, A and R values should be close to 2 and 1, respectively. This statistical analysis provides a quantitative tool for investigating the collective behavior exhibited by mesoderm cells during their spreading. The behavior is quantitatively defined as the spreading strength A, which corresponds to the typical value of θend/θstart. The collective nature of the process is quantified by R: a value close to 1 demonstrates the spreading behavior A=θend/θstart is shared by the entire cell population, as in wild-type embryos; a lower value means the cell spreading is disrupted, as in mutants. This quantitative analysis has been used to demonstrate (i) the reproducibility of the collective behavior in wild-type embryos and (ii) the disruption of the process and the identification of different cell populations in a mutant embryo1.
The procedure described in this paper details every experimental step from the preparation of embryos for imaging to the quantitative analysis of mesoderm cell spreading. In addition to this analysis (step 28), the cell movements can be analyzed in whatever manner a user finds interesting by developing their own customized Matlab scripts to analyze the registered data (step 27).
In most biological tissues, light scattering is the main physical process limiting the depth of imaging. In 2PEF microscopy, it can be characterized experimentally by measuring , the scattering mean free path of the excitation light. This length provides an estimate of the maximum depth of imaging and allows for comparison of the imaging conditions between different biological samples. If light absorption and optical aberrations can be neglected, and assuming the fluorescence collection efficiency is constant within the depths of imaging15, the detected 2PEF signal F from a homogeneous fluorophore distribution is expected to scale as24:
where z and P0 are the imaging depth and the average incident laser power at the tissue surface, respectively. Hence, is experimentally estimated by acquiring a z-stack of images through the sample with a given incident power, by measuring the average 2PEF signal F(z) in a homogenous area at each z-position and the background signal Fbackground, and by plotting . A linear regression on G(z) provides an estimate of the slope as (Fig. 2c). We measured at 940 nm in the mesoderm and ectoderm tissues in wild-type and klar embryos at stage 8 as 56 μm and 76 μm, respectively (Fig. 2d). The estimation of displays the typical 2PEF signal decay based on equation  (Fig. 2e). This graph shows that at 80μm in depth, the signal in wild-type embryos at stage 8 is low (blue line) and twice as much signal can be expected in a klar mutant at the same stage (red line). As a comparison, we provide measurements and signal decay in stage 5 Drosophila embryos and in zebrafish embryos from previous reports11, 17 (Fig. 2d–e). It demonstrates that the optical properties of embryonic tissues and the subsequent limitation of imaging depth is highly stage- and species-dependant.
We thank M. Liebling for advice on Imaris and Matlab, and the Caltech Biological Imaging Center for sharing equipment. This work was supported by grants to A.S. from NIH(R01 GM078542), the Searle Scholars Program, and the March of Dimes (Basil O'Conner Starter Scholar Award, 5-FY06-145), grants to S.F. from the Caltech Beckman Institute and NIH (Center for Excellence in Genomic Science grant P50HG004071), and fellowship to W.S. from the Caltech Biology Division.
Competing Financial Interests The authors declare that they have no competing financial interests.