Gene expression and morphology of the Drosophila embryo, calculated at cellular resolution for the three-dimensional surface of the blastoderm, can be projected onto a surrounding cylinder and unrolled to produce a planar map. (Anterior or head of the embryo is to the left, posterior to the right. Upper and lower edges of the map represent the dorsal or back midline of the embryo, with the ventral or front midline in the center.)

Meanwhile the blastoderm’s shape remains that of the egg, a smooth if somewhat irregular ellipsoid. But this simplicity is soon lost. At the next stage of development, called gastrulation, the embryo begins to crease, eventually folding into the fly’s intricately articulated body plan.

Because its geometry is uncomplicated and the nuclei and resulting cells all lie close to the surface, the blastoderm is a natural target of investigators who want to understand how genes turn on and off on their way to forming the whole animal, coordinating their expression in time and space.

Recently, for the first time, researchers at Berkeley Lab, UC Berkeley, and UC Davis, as members of the Berkeley Drosophila Transcription Network Project (BDTNP), have developed methods for looking at the Drosophila blastoderm in three dimensions, at the resolution of individual cells, while precisely measuring the expression of multiple genes as they interact to shape the fly.

“Visualizing gene expression and analyzing the morphology of an entire organism at cellular resolution has never been done before,” says David Knowles of the Berkeley Lab Life Sciences Division’s BioImaging Group.

3-D revelations at cellular resolution

In their first application, the new methods used by the BDTNP team detected morphological and gene-expression features never seen before, overturning some long-held assumptions in the process.

A striking feature of development during the blastoderm stage is the formation of “expression stripes” that encircle the embryo and move over its surface, the stripes moving from what would be the head of the organism toward the tail. The genes giving rise to these stripes control the later segmentation of the embryo. Prior to the work of the BDTNP, biologists studying this system were limited to two-dimensional images of the blastoderm taken at different stages from fixed embryos, no two of which were identical.

With this arrangement, gene expression could not be assigned to specific nuclei from one image of a fixed embryo to the next; the assumption arose, almost by default, that the spatial movement of stripe patterns was due solely to gene expression and not to the motion of the nuclei — a highly plausible assumption, considering that many of the expression products are transcription factors that regulate the expression of other genes.

Calculated from raw three-dimensional images of the embryo (left, with nuclei stained white, expression of a gene called snail stained red, and expression of a gene called eve stained green), PointClouds can be rendered as spheres or tiles (right), their combined colors indicating the degree of expression of the genes by each nucleus at a given stage. Arrows in the panels indicate the same three nuclei.

The BDTNP collaborators also began by imaging fixed embryos, but they imaged them whole in three dimensions. Then they applied a sophisticated set of data-analysis and modeling tools. “We set up a pipeline to make and compare thousands of high-resolution microscope images of different embryos at different stages of blastoderm development,” Knowles says. “For each image we stained the total DNA, which gives the location of each nucleus, plus we stained the expression patterns of two specific genes.”

By computer analysis of raw, three-dimensional embryo images, each some 500 megabytes in size, the images were converted into small, easy-to-manipulate text files called PointClouds. Each PointCloud records the center of mass of all nuclei and the expression of two genes in and around each nucleus in an embryo. By combining data for multiple PointClouds, the researchers could estimate the concentrations of multiple different gene products in each developing cell.

Biologist Mark Biggin of the Genomics Division says, “The developmental stage of each embryo that our PointClouds were derived from is known, but how can we say this cell in one embryo is the same as that cell in another? The answer is, we devised a model — an average, statistical model — based on our observations. And then we tested it.”

Biggin explains, “We were able to populate our model with points that we could use to predict any possible nuclear movement. And we could separately measure the expression flow by tracking the collection of expressed genes.” Unexpected features of the blastoderm immediately became apparent.

“With this new model of the embryo, we could see the previously reported shift in gene expression stripes,” says Biggin, “but we also saw that the nuclei do in fact move. The gene expression patterns move partly in sync with these changes in morphology but partly independently. There is an intricate interplay between which genes are expressed and how the nuclei move.”

Stained nuclei in a single living embryo were tracked through the blastoderm stage. Red arrows indicate nuclei that moved the farthest, blue arrows those that moved least. Movement of nuclei in live embryos corresponds to predictions of the model.

To substantiate the model’s prediction that the nuclei move, the researchers also followed the development of living embryos during the blastoderm stage. Although methods to examine living embryos provide only partial, less precise measurements, the model’s predictions were indeed confirmed.

The ability to model all three dimensions of the blastoderm revealed additional new information about the system. For one thing, says Knowles, “Expression of genes previously thought only to regulate development along the anterior-posterior axis” — the head to tail axis — “also changes along the dorsal-ventral axis” — the back to front axis. “As we go around the embryo in 3‑D, we see changes in expression along both axes.”

Future Directions

Having produced the first quantitative, three-dimensional description of gene expression and morphology of the whole embryo at cellular resolution, the researchers are eager to move to the next stage of discovery.

“We know as much about the biology of this organism as any other, but in fact we really don’t have enough data,” Biggin says. “A first-order approximation of what’s going on in the embryo is not good enough. When you have precision right down to the level of the cell, as we now do, all you have to do is ask a question and you’ll discover something. Our current study involves a limited number of genes acting in a small window of time. You just know you want this kind of information for the whole course of fly development, and for thousands of genes.”

Members of the Berkeley Drosophila Transcription Network Project include (from left) Hanchuan Peng, Angela DePace, Oliver Rübel, Jitendra Malik, Charless Fowlkes, Mark Biggin, Mike Eisen, Soile Keränen, Bernd Hamman, Cris Luengo, Damir Sudar, Gunther Weber, and David Knowles. (Image Courtesy of David Knowles, background visualization from BDTNP PointCloudXplore program.)

The Berkeley-based researchers have no intention of pursuing these questions by themselves. “The goal is to better understand the transcription network,” Knowles says. “We’ve released all of our current PointCloud data, unanalyzed, on the Berkeley Drosophila Transcription Network Project website, where anyone can download it and use it for their own research.”

The BDTNP’s remarkable discoveries to date are proof of the strength of “uniting different talents,” as Biggin puts it: “specialists in imaging, computer science, visualization, and biology.” It’s an approach that should lead to a better understanding of development not only in the fruit fly but in many other organisms as well.

Work conducted by the BDTNP is funded by grants from the National Institute of General Medical Sciences, the National Human Genome Research Institute, and the Department of Energy’s Office of Science.

Additional information

• More about the Berkeley Drosophila Transcription Network Project
• “3D morphology and gene expression in the Drosophila blastoderm at cellular level, part I: data acquisition pipeline,” by Cris L. Luengo Hendriks, Soile V. E. Keranen, Charless C. Fowlkes, Lisa Simirenko, Gunther H. Weber, Angela H. DePace, Clara Henriquez, David W. Kaszuba, Bernd Hamann, Michael B. Eisen, Jitendra Malik, Damir Sudar, Mark D. Biggin, and David W. Knowles, was published online by Genome Biology, 21 December 2006.
• “3D morphology and gene expression in the Drosophila blastoderm at cellular level, part II: dynamics,” by Soile V. E. Keranen, Charless C. Fowlkes, Cris L. Luengo Hendriks, Damir Sudar, David W. Knowles, Jitendra Malik, and Mark D. Biggin, was published online by Genome Biology, 21 December 2006.