Skip to content

Shedding Light on the 3-D Structures of Cells

SDSC RESEARCH |Contents| Next
Jon Genetti
David R. Nadeau
Bernard Pailthorpe
Lubor Borsig
Eric Elenko
Jeff Esko
Marilyn Farquhar
James R. Feramisco,
Maria Pinhal
Brian Smith
Hiro Tsukada
Ajit Varki,
Jean J. Y. Wang
UC San Diego

P rofessor James Feramisco and his associates at the UC San Diego School of Medicine's Cancer Center are working with David Nadeau, a principal scientist in SDSC's Scientific Visualization group, to create 3-D images of the interiors of cells. "SDSC's 3-D rendering software gives us valuable perspective views into the structures of cells that enable us to perceive spatial relationships between regions that contain active proteins," Feramisco said. The collaboration will provide valuable feedback to SDSC's visualization toolkit developers on how researchers can take best advantage of computer graphics in their scientific applications.




"Scientific visualization is a cornerstone of high-performance computing," said Bernard Pailthorpe, SDSC's associate director for Scientific Visualization. "As we develop visualization toolkits for scientific applications, we work with researchers to ensure that we create needed, useful tools for the community. Our latest collaboration is with cellular biologists at the UCSD Cancer Center."
"As far as I'm concerned, without 3-D imaging there is not biology," said Jean J. Y. Wang, professor of Biology and associate director of Basic Research for the Cancer Center. "The era of grinding up cells and studying them in a test tube taught us a lot, but now we need to understand how cellular functions are coordinated in space and time."


The Cancer Center's Digital Imaging Shared Resource (DISR) assists research into the detailed examination of cellular structure, especially in such areas as cell-to-cell contact, cell differentiation, apoptosis (cell death), intra-cellular transport of substances, and metastasis that are relevant to studies of cancer. DISR resources include a deconvolution microscope that can generate multi-spectral digital images at wavelengths from 340 to 700 nanometers, and a microinjection system that can insert molecules and dyes into living cells. The UCSD Cancer Center is supported by the National Cancer Institute as an NCI-Designated Cancer Center.

"In a technique called immunofluorescence microscopy, we utilize purified antibodies attached to fluorescent dyes, applying this dye-antibody complex to a cell or tissue section where it binds to its corresponding antigens," said DISR Resource Leader James Feramisco, a professor at the Cancer Center and a member of the departments of Medicine and Pharmacology. "Illuminating the specimen with light of appropriate wavelengths causes the labeled antibodies to fluoresce. Since each chosen dye-antibody complex binds only to a specific protein in the cell, the fluorescence acts as a structural marker for regions of interest that contain the protein. By treating a specimen with several differently-colored labeled antibodies, multiple proteins can be localized within the same cell. When we view the cell with a high-resolution microscope, we capture multiple sets of images, one per color." The researchers also microinject fluorescently labeled proteins into living cells, so they can follow the intracellular dynamics of specific proteins.

In each image, the portions of the cell at the plane of focus are sharp and well-defined, while sections of the cell that are closer or further away are out of focus. Using a commercially available wide-field fluorescence microscope and deconvolution software, researchers capture stacks of images of each specimen while moving the plane of focus from near to far. The out-of-focus portions of each image are removed by applying a deconvolution algorithm to the image stack, resulting in sequences of cross-section images that correspond to slices through the cell specimens.

Top| Contents | Next


"SDSC's visualization group is creating a suite of versatile, cross-platform tools for rendering, visualizing, and interacting with very large data sets from a variety of scientific disciplines," said software architect Nadeau. "The tools are designed to work with data sets up to hundreds of gigabytes in size--too large to fit in a computer's main memory--and support simulations and analyses on teraflops computers."

To visualize the cells in 3-D, Nadeau imported the image stacks into the Volume Scene Graph Toolkit. The toolkit filtered the data, mapping low-intensity backgrounds to transparency and leaving cell structure opaque. Both single frames and movies were rendered in perspective using the VISTA volume renderer.

"VISTA is a toolkit that will enable researchers to render very large data sets in 3-D on parallel supercomputers, such as NPACI's IBM SP, Cray MTA, and Sun HPC machines," said Jon Genetti, VISTA's software architect, "but it will scale from single-CPU to multi-CPU architectures and integrate with desktop computer visualization software and applications."

VISTA is an application programming interface that can be linked into existing code packages. VISTA precursors already had been used in biomedical applications for visualizations of brain tomography and of the National Library of Medicine's "Visible Human" virtual cadavers. The Scientific Visualization group's collaborations with astronomers at the American Museum of Natural History's Hayden Planetarium to create accurate views of astronomical objects from unearthly vantage points were of particular interest to the Cancer Center researchers.

Top| Contents | Next


"We need the ability to render and store larger data sets," Feramisco said. "Currently, a typical volume derived from fluorescence microscopy might consist of 500 images, a stack of 100 image planes in each of five colors. We intend to capture larger volume views of living cells in time sequence--ten or more in succession--and a typical day's run could generate many gigabytes of data. SDSC's computational and storage resources are ideal for our needs."

"We started working with the Cancer Center because our toolkit software could do two things that other packages for rendering large volume data sets could not," Nadeau explained. "They knew our 3-D renderings of the Orion Nebula for the Hayden Planetarium in New York were based on volume elements that each had separately settable parameters for transparency versus opacity and for luminosity versus non-luminance. This is what's needed to accurately render images of glowing gases in interstellar space, and it's also what's needed to depict fluorescing dyes inside cell structures and enhance image understanding by facilitating the modulation of transparency and opacity."

"We also are interested in true perspective rendering," Feramisco said, "especially in video sequences in which the viewpoint moves. Perspective rendering indicates depth and allows us to better view relationships among components in cells within the resolution limits of optical microscopy."

"An important aspect of these visualizations is the ability to see quantitative spatial relationships between structures," said Marilyn Farquhar, professor and chair of the new Department of Cellular and Molecular Medicine and professor of Pathology at UC San Diego. "Determining the proximity and amount of overlap of the cell components that contain various proteins is important to our understanding of cell processes. Our lab studies protein targeting during membrane trafficking and investigates the functions of proteins involved in signaling. Another interest is the cellular and molecular basis of kidney diseases, especially autoimmune diseases, toward which we are mapping protein binding sites for antibodies and ligands."

The initial results from the collaboration's 3-D fluorescence microscopy reconstructions of human and mouse tumor cells have been quite successful (Figures 1 and 2). Two papers currently are submitted for journal publication, with more to follow. Professor of Medicine Ajit Varki and postdoctoral researcher Lubor Borsig have discovered differences in the amount of platelets and macrophages associated with metastatic tumor cells dependent on the protein P-selectin; Professor of Cellular and Molecular Medicine Jeff Esko, postdoctoral researcher Maria Pinhal, and researcher Brian Smith of the Cancer Center have identified protein domains responsible for Golgi localization.

Cancer Center researchers are continuing to acquire cell images, and Nadeau and Genetti are creating many more volume-rendered visualizations. In addition to the ongoing work with fluorescence microscopy images of several types of normal and abnormal cells, the collaboration is investigating whether the visualization toolkits can be used with bright-field microscopy and with Nomarski interference microscopy, a technique that uses refraction of polarized light to highlight regions in the cell of differing thicknesses and refractive indices. The Nomarski technique can produce sequences of images at different focal planes that yield a volume representation of the specimen.

"This is just the beginning," Feramisco said. "We have a wonderful new interaction, and we can't wait to make full use of it."

"SDSC's work with the Cancer Center is a high-payoff project," Pailthorpe said. "We view this effort as an important 'reality check' that demonstrates the usefulness of the SciVis toolkits. A 'plug-and-play' visualization environment, with modular tools that scale from desktop machines to supercomputers, will enable researchers to tackle not just larger problems, but new kinds of problems as well." --MG *

Top| Contents | Next
blue_angled-cmyk Figure 1. Golgi Bodies in 3-D
Golgi bodies are the "packaging plants" of cells, where materials manufactured by the endoplasmic reticulum are modified, stored, and shipped out. The Golgi apparatus is extensive in cells that are specialized for secretion. For this 3-D visualization, two different protein components of the Golgi bodies (green, red) and nuclei (blue) were imaged using the deconvolution microscope. The data were obtained by Maria Pinhal in the lab of Jeff Esko, Department of Cellular and Molecular Medicine at UCSD School of Medicine, with the assistance of Brian Smith of the DISR.
Figure 2. Abetting Leukemia Cells
"Nurse-like" cells, which spontaneously differentiate from the blood of leukemia patients, participate in a strange symbiotic relationship--leukemia cells promote their formation, and nurse-like cells protect leukemia cells from dying. Around the cell nucleus (blue) are the protein chemokine (red), which attracts leukemic cells and the protein vimentin (green). These frames from a motion sequence created by David Nadeau that shows the value of 3-D perspective and moveable viewpoint in examining cell structures. The microscope images were acquired by Hiro Tsukada, a member of the laboratory of Thomas Kipps, deputy director for Research Operations of the UCSD Cancer Center, with the assistance of Brian Smith of the DISR.