Investigating Morphogenesis in Xenopus Embryos: Imaging Strategies, Processing, and Analysis

Abstract

Methods have been developed for visualizing cell movement and protein dynamics during morphogenesis within live multicellular tissues isolated from Xenopus laevis embryos. These include the preparation and use of reporter constructs in Xenopus embryos, microsurgical techniques for isolating embryonic tissues, and methods for culturing live tissues for extended periods. In this article, we present strategies for successful imaging of large thick embryonic tissues by improving the signal and minimizing damage to cells and tissues from overexposure. We also describe strategies for image analysis, including construction of kymographs, use of time- and z-projected confocal stacks, and approaches to segment images using regions of interest. With these imaging tools, the “cut-and paste” embryology of the Xenopus model system allows remarkable access to both the mechanics of cells and tissues as well as the complex cell biology of adhesion and cytoskeleton during morphogenesis.

Introduction

Light microscopy techniques used for live-cell or live-embryo imaging require considerable optimization. Damage to cells and tissues from over-exposure to high intensity lasers during laser-scanning confocal microscopy or high intensity Hg or Xe arc-lamps during spinning disk confocal or conventional epifluorescence microscopy can lead to dead or dying cells, erroneous results, and inaccurate interpretations. In some cases steps to reduce photodamage are obvious, such as the reduction of incident excitation, whereas other efforts, such as the use of lower magnification objectives are not. In this section we suggest several strategies that may improve the viability of cells subjected to intensive live-imaging.

In addition to reducing the power or intensity of the excitation illumination, increasing the interval between image acquisition in a time-lapse sequence, or opening the confocal pinhole the user can reduce pixel dwell-times or spread the energy deposited by excitation over a longer time. Laser-scanning confocal microscopes typically provide control over the speed at which an image or single confocal section is collected or the frequency of collection. The speed of collecting a single image is often provided in Hz (or Hertz), or how many scan-lines are sampled per second. A 1024 by 1024 pixel image (1024 lines with 1024 pixels sampled per line) collected at 200 Hz requires about 5.1 seconds (1024 lines / 200 lines per second) to acquire each section. Increasing the speed to 400 Hz reduces the time to 2.6 seconds. Speeding acquisition reduces the amount of laser excitation that falls on each pixel. Depending on the confocal imaging system, the degree of “zoom” magnification can also increase the amount of laser excitation that falls on the sample, increasing the total laser energy on a smaller region of the cell or tissue. In contrast to changing the zoom, reducing the dimensional size of the image should not change the amount of time the laser spends on each sample.

It has been demonstrated in some systems that the time-course of exposure can also influence the severity of photodamage (Pawley, 2006). Continuous exposure may be more damaging to a cell than iterative exposures spread over a longer time even when the total integrated exposure is the same. One hypothesis is that cellular mechanisms that repair photodamage can be saturated allowing cellular injury to spread; however, if the damage is spread over a longer time, cellular mechanisms may be able to repair the damage. Depending on the rate of change of the cell process one can collect a short series of images at the same focal plane at reduced intensities instead of collecting a single image or confocal sections with a high laser intensity. Such a series can be summed or averaged to obtain the same quality as the high intensity image.

Choice of fluorescent protein or fluorescent molecule for tagging target proteins can strongly influence the degree of photodamage. It is thought that fluorophores that are resistant to photobleaching may also produce lower levels of photodamage. We have typically used fluorescein, rhodamine, cy3, EGFP, and mRFP to tag proteins (Fig. 1). Among these, the most phototoxic has been fast bleaching fluorophores such as fluorescein. Another consideration is the level of expression or amount of tagged protein injected. Photodamage may be greater from high levels of expressed protein than from lower levels. Photodamage from an introduced fluorophore can be discerned from endogenous sources of photodamage such as pigment granules or NADPH if photodamage increases with expression level or with the degree of accompanying photobleaching.

Figure 1.

Live-cell confocal microscopy during morphogenesis. (A) Mediolaterally intercalating mesoderm cells labeled with both rhodamine-dextran to highlight the cytoplasm and nucleus and negatively mark yolk granules and intracellular vesicles and mem-GFP to highlight the plasma membrane at cell boundaries. Mem-GFP appears twice as intense when two membranes fall within the confocal depth of focus or in a single set of pixels. (B) GFP expressing live mesoderm assembles fibronectin fibrils along the interface where mesoderm abuts agarose. Synthesized fibronectin fibrils are labeled with a Cy3-conjugated fibronectin mAb (4H2) in real time. (C) Paxillin-GFP expressing animal cap edge cells cultured on a fibronectin coated glass substrate. (D) Frames from a confocal time lapse sequences of utrophin-mCherry expressing mesendoderm cells cultured on a fibronectin coated substrate.

Choice of objective is another key element in enhancing image quality and reducing photodamage. Depending of the resolution needed, either a high numerical aperture, long working distance 20X or a high n.a. oil immersion 40X or 63X objective can be used. High n.a. oil immersion lenses combined with scanner zoom magnification is required when attempting to resolve cytoskeletal dynamics. Increased magnification does not always capture a better image of the cytoskeleton. For instance, use of the 100X oil immersion objective produces a dimmer image than the 63X objective and results in greater levels of photodamage. In addition, high n.a. oil immersion objectives do a poor job of deep tissue imaging, being optimized for use on thin samples within 25 μm of the coverslip and have very limited working distances. Several microscope companies offer water immersion lenses or so called “dipping” lenses. These have relatively high n.a. and offer working distances up to 2 mm. However, they were designed for use with upright compound microscopes and due to the rapid evaporation of water are not well suited for long term time-lapse imaging. In the future, microscope manufacturers and objective designers will need to address this gap in their product lines. Depending on the structures being studied, a high n.a. oil immersion lenses is not always required; for instance, when cell membranes or nuclei are being tracked a confocal-optimized 20X objective may be sufficient. Even though the 20X objective is not as efficient as the oil immersion lenses in capturing emitted light it can resolve more cells and can spread the negative effects of photodamage over a larger area, often reducing the degree of photodamage per cell.

The key to testing the viability of cells after exposure is to check cell and tissue health after the experiment. Typically we collect both pre- and post-confocal-session still images of cells and tissues. Viable cells retain their participation in the process of morphogenesis. Cells that have been overexposed or have suffered photodamage typically round up, and are excluded or dissociated from the explant. Since we can control the field of view we can also carry out control experiments where cells that have been intentionally exposed to high levels of illumination can be compared, often in the same field of view, with those exposed to minimal levels of illumination.

One concern with live-imaging is that the behavior of the fluorescently tagged protein is different from endogenous proteins. The first step in assessing the veracity of tagged proteins is whether their distribution or localization patterns differ from of endogenous proteins that have been fixed and immunofluorescently imaged. Since fixation techniques are also subject to artifacts one needs to first optimize protocols for immunofluorescent labeling. Once convinced that fluorescently tagged proteins reflect the localization of endogenous proteins you must consider whether the dynamics of the tagged protein represents dynamics of the endogenous protein.

Validating the dynamic behavior of fluorescently tagged proteins can be challenging and somewhat subjective but we suggest several strategies. First, test different tagged proteins that are known from biochemical studies to assemble into the same complex. These should co-localize to similar locations. Additionally, it must be remembered that the fluorescently tagged proteins may have different biochemical rate constants than endogenous proteins. The addition of a bulky, 27 kD molecular weight fluorophore like GFP may alter the affinity of the tagged protein for its normal target binding sites. Similarly, over-expression of a tagged protein may saturate or sterically-block binding sites of endogenous proteins altering normal behavior of large multi-protein molecular complexes. One strategy for establishing a reliable level of expression is to lower the expression level to the lowest possible degree and assess localization patterns and determine whether macroscopic cell behaviors are unchanged from those that do not express tagged proteins.

Image processing and analysis

Unlike single cell culture, development progresses over longer time scales may take hours to accomplish each step. Various image analysis techniques are needed to detect and interpret the dynamic nature of multicellular tissues and intracellular molecules from long time-lapse movies of developing embryo. Flexible application of these approaches allows quantitative analysis of cell movements and correlation of patterns of protein localization with cellular functions that are thought to drive morphogenesis. In this section we will introduce some approaches to extract quantitative data from time-lapse sequences and analyze cellular dynamics from live confocal images of isolated explants.

a. Kymographs

The kymograph is one of the most simple and effective ways to show spatial changes of moving structures over time. To construct a kymograph one must first pick a line as a “region of interest” (ROI). Pixel intensities along the line in subsequent frames can be reassembled as a 2D image with one axis representing intensity along the line and the second axis representing intensity changes along the line over the time-course of a movie. A kymograph can reduce complex patterns of movement in a movie to reveal trends in a readable way. For instance, the velocity of moving features in the kymograph can be measured as a slope. A limitation of this method, since intensities are collected along a fixed line, is that kymographs may not useful in analyzing movements over large distances, movements that are not very persistent, or structures that overlap, join, or separate during their movement.

b. Projecting images

Multiple images in 3D confocal stacks, or z-series, and time-lapse sequences can be projected to a single image to yield information on the connectivity of structures in space or time, respectively.

Z-series

Although collecting live 3D confocal images of frog tissue is challenging, limited 3D stacks over the thickness of a single cell can be collected at high resolution. Confocal sections from two z-positions collected 5 μm apart, e.g. at z=0 the basal surface and 5 μm from the basal surface of cell, over time can be color encoded with two primary colors to reveal temporal dynamics at the basal and mid-cell surfaces in a single, color image plane (fig. 2A). Confocal stacks at 1 μm intervals may be collected over a range of 10 μm at each time-point. Small z-step intervals, e.g. 0.1 μm, collected into a stack can allow post-collection reconstruction to reveal a “lateral” view of cell boundaries that can reveal cell-cell interactions within the tissue. Interesting targets for this type of imaging and analysis include proteins that move from the cytoplasm to the cell cortex or those that dynamically change their location in the cell cortex. Collecting high resolution 3D confocal stacks over time can lead to photobleaching and photodamage and we advise caution when interpreting dynamics within these data sets (see discussion in Section IV above).

Figure 2.

Processing and Analyzing Images. (A) Two confocal z-sections of mem-GFP are collected from a marginal zone explant, color encoded, and combined into a single merged image. Membrane protrusions are visible at z =0 and cell boundary is clear at the 5 μm level. Merged z-levels encoded by different colors can show how mesoderm cell protrusions on the substrate are coordinated with shape changes in the cell body. (B) A single frame from a confocal time lapse of tau-GFP collected for 10 minutes at a 10 second interval. Average or maximum time projections of tau-GFP present the dynamics of microtubule within elongated mesoderm cells over the time course of the movie. (C) Frames from a time-lapse of mem-GFP expressing cells can be used to extract cell shapes (C′), designate ROIs, and follow the shape changes over time for individual cells (marked in red and blue in C′).

Time series

Projections of confocal time-lapse sequences can be useful to explore dynamics of labeled proteins (fig. 2B). Different approaches can be used to project images depends on phenomena; for instance, summation or averaging of time lapse images can represent the accumulated intensity values of each pixel over the field of image which can indicate the frequency of a structure’s appearances or localization of the tagged proteins over time. Maximum intensity projection of short time series can often show the movement individual structures during the time series. Unlike conventional 3D volume projection, projection of time domain allows the visualization of movement with a simple way and can help formulate more sophisticated methods for analyzing time lapse movies. Color can be used to distinguish structures appearing early in the time-lapse from those that appear later.

c. Managing your ROIs

To identify or quantify any interesting feature from your collected images, you first need to separate that feature from the background through thresholding or background subtraction (Russ, 2006), then extract features for further quantification (fig.2C).

Image J

Among several image processing and analysis programs available, we routinely use the free software ImageJ (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, http://rsb.info.nih.gov/ij/, 1997–2009). ImageJ is an open-source image processing program that runs on Mac, PC, and Linux operating systems that can display, edit, analyze and process many types of images. Users can freely write their own or download additional useful plug-ins to carry out more complex or sophisticated image analysis.

ROI manager

Identifying structures or fluorescent features in images is a key step in assessing their function. In the jargon of image analysis these features are called regions of interest, or ROIs. ROIs may be manually selected, hand-drawn, or segmented by an automated process. An important feature in ImageJ, the ROI Manager, allows the automated handling and manipulations of these ROIs. Once created, ROIs can be packaged and stored by the ROI Manager into a file containing the location, size, and shape of the listed ROIs. The ROI Manager is getting a powerful tool when dealing with hundreds or even millions of ROIs identified from time lapse images. Listed ROIs can be named, colored, measured for several factors, in addition, these ROI databases can be used as a static data sets to test different image analysis algorithms. Thus we have found the ROI manager an essential tool in analyzing cell behaviors and protein function during morphogenesis.

Conclusion

Recent studies of morphogenesis have come to rely on sophisticated molecular engineering, live fluorescent and confocal imaging, and computationally intensive image analysis tools to illuminate the cell and mechanical basis of morphogenesis. This chapter has presented standard methods to carry out these types of studies, from expressing molecularly engineered probes, to chambering and imaging the cells and proteins in morphogenetically-active tissues, and in quantitative analysis of these images. Future efforts are needed to improve both the fluorescent reporters and the automated image analysis tools used to assess their function. Improved automation will enable high throughput experiments and extend our mechanistic understanding of the molecular basis of morphogenesis to the interaction between cellular mechanics and molecular machines that drive them.