ADVERTISEMENT
ADVERTISEMENT

Advances in Cellular Image Processing

EMBRYOGENESIS UNFOLDING IN 3-D:Left and Right Image: Courtesy of Wen Bin Tsai & W. Kinsey Center image: Courtesy of H. Matsumoto & S. K. DeyThree-dimensional projections created from Z-stacks of a zebrafish embryo at the four-cell stage (left), a blastocyst (center), and a more fully developed zebrafish embryo (right). DAPI-stained nuclei are colored blue, while various specific proteins are labeled green (FITC/FITX) and red (rhodamine).Like much of science, imaging has become almost ent

Mike May
<p>EMBRYOGENESIS UNFOLDING IN 3-D:</p>

Left and Right Image: Courtesy of Wen Bin Tsai & W. Kinsey Center image: Courtesy of H. Matsumoto & S. K. Dey

Three-dimensional projections created from Z-stacks of a zebrafish embryo at the four-cell stage (left), a blastocyst (center), and a more fully developed zebrafish embryo (right). DAPI-stained nuclei are colored blue, while various specific proteins are labeled green (FITC/FITX) and red (rhodamine).

Like much of science, imaging has become almost entirely computerized, with digital capture devices replacing more traditional film. Capturing the data in digital form simplifies work, for instance by cutting out lengthy film processing steps, and it aids in data archiving. More important, however, digital imaging enables a new variety of experimental approaches.

When asked to pick the most important recent advance in imaging cells, John Kirn, associate professor of biology at Wesleyan University in Middletown, Conn., points to the two-photon, laser-scanning confocal...

GOING LIVE

<p>WHAT NERVE!</p>

Courtesy of Yongjun Sui and O. Narayan

An optical section of cultured neurons. Shown are Hoechst-stained nuclei (blue), a FITC-labeled cytoskeletal marker (green), and TRITC-labeled cytokine label (red).

The digital side of cell imaging comes from a charge-coupled device (CCD). When an image is focused on a CCD sensor, the photons are absorbed where they strike the silicon, producing electron-hole pairs. The charges can be measured separately in each pixel on the silicon surface, thereby creating an image.

According to Pawley, one new high point is the electronic multiplier CCD camera, which he calls the "first significant improvement in CCD since its inception." In essence, the electron multiplier is a series of low-gain amplifiers that, working together, boost the electron signal before it reaches the normal CCD amplifier. Thus, the device can detect the signal from a single photoelectron that hits the sensor, and it can do so quickly. "In the past, you could have high sensitivity or fast readout," says Pawley. "With the electronic multiplier CCD you can have both, and this will be particularly useful when using disk-scanning CCDs on living cells."

Pawley works on making new imaging systems for living samples.1 In general, dead cells can take a beating and still be observed, but in living cells, the combination of fluorescent dyes and light can create toxic substances. Consequently, a live-sample detector needs to grab up as much fluorescence as possible so that less light can be used. Along those lines, Pawley is developing a new diode that is 10 times more efficient than the ones used in conventional confocal microscopes.

To get a good picture of living cells, though, scientists would also like a more realistic view, replacing the flatness of a 2-D image with a more lifelike 3-D one. "Prior to the last four to five years, even the last two to three years, it wasn't very easy to do 3-D analysis," says Christopher K. Rodesch, director of the Fluorescence Microscopy Core Facility at the University of Utah School of Medicine in Salt Lake City.

In the last few years, though, such analysis has become routine. "A number of commercial outfits put out decent, user-friendly 3-D analysis packages," says Rodesch. Image processing and analysis software is also available for free by using the National Institutes of Health's Java-based ImageJ http://rsb.info.nih.gov/ij. These packages generally process a collection of optical slices to produce a 3-D rendering of the original sample.

Though such software can generate spectacular images, 3-D also delivers new information. "3-D imaging has helped in the exploration of structure-function relationships at the levels of single cells and whole brains," says Kirn, a neurobiologist who studies the birth of new neurons in adult brains. In single cells, confocal imaging can reveal fine anatomical processes, even in thick slices of brain tissue. "Optical sections are stored on computer," Kirn explains, "and there are programs that allow for the rotation of cells in three dimensions." For whole brains, Kirn uses positron emission tomography and nuclear magnetic resonance, which record the activity of large cell populations. "All we need now," Kirn says, "are the algorithms to merge the information from the two methods so we can see the 'trees' and the 'forest' at the same time."

ADVANTAGES OF AUTOMATED COUNTING

<p>COMING INTO FOCUS:</p>

© 2003 Nature Publishing Group

A fluorescently labeled microtubular network in a human embryonic kidney cell imaged under standard confocal conditions (left) and with Stefan Hell's STED-4Pi microscope (right). At bottom is an intensity diagram of the image section under the dotted lines above. Note the sharpness of the right-hand image compared to the left. (from Nat Biotechnol, 21:1303–4, 2003)

Like many areas of modern biology, imaging creates enormous amounts of data. Elke Weiler at Ruhr University in Bochum, Germany, searches through thousands of images of brain slices.2 "We are interested in brain development, function, and how it can be restored when lesions occur," Weiler says. "We look at how cells around a region take over a function or express another genetic factor," by labeling specific cell populations to reveal expression changes. But, she adds, "We don't want to count the cells by hand."

Manual counts take time, produce errors, and require experienced counters, according to Weiler. She says commercial programs cost too much and still require supervision from an experienced worker. To tackle those issues, Weiler and colleagues created a software tool called the automatic cell-counting method (ACCM; http://www.neurop.ruhr-uni-bochum.de/~alia/accm.html).3

To test this system, Weiler's team made sections taken from rat cortex and labeled the cells with various stains, including one for the neurotransmitter gamma-aminobutyric acid (GABA). They imaged the cells using a CCD-equipped microscope. The ACCM software scanned the images for brightness (darker areas contain more stain), grouped pixels into clusters that represent cells, and then counted them.

Weiler wanted to know how the software would perform compared to manual counting and a commercial counting system. The experienced counter and the ACCM gave equivalent counts. But, the program was faster: ACCM counted more than 10,000 images of GABA-stained cells (which would have taken a pathologist four months to complete manually) in just two days.

Then, Weiler and her colleagues compared the software with a commercial product called MetaMorph from Universal Imaging of Downingtown, Pa. The ACCM and MetaMorph performed about the same, as long as the cell density exceeded 1,000 cells per image. Below that threshold, Weiler writes in her paper, "MetaMorph values were often above those of ACCM and outside the variance of the manual counting, independent of the magnification (50× to 400×) and the marker." She concludes, "In the presented case, the correlation between manual counts and ACCM counts was better than the correlation between manual counts and the commercially available program that we used." In other words, Weiler says in an interview, "Our program was more accurate than the program that we bought."

<p>I CAN SEE CLEARLY NOW:</p>

Courtesy of Michael Taylor and D. Wright

Three-color separation of a 3-D projection created from a z-stack (A,B,C). Two sensory nerves are shown (C); the one at bottom left is degenerating, while the one at top is normal, and looks white (D) where it wraps around the muscle bag fiber (red). Image (D) is an overlay.

Even the best counts from a static image, however, may not answer every research question. The imaging might also need a dynamic side. "Adding the time element," says Jeff Lichtman, professor of anatomy and neurobiology at the Washington University School of Medicine in St. Louis, "one can directly observe many phenomena that are totally impossible to even infer from single time points." For example, in developmental neurobiology, an organism starts out with more neurons than it ends up with as an adult. In many places, neurons compete to make connections to other neurons or muscles, adding and removing branches as needed. "No single-time-point data can tell you that," Lichtman says. "You need to watch over time."

Lichtman and graduate student Mark Walsh watched places where two neurons innervated a single muscle.4 As their camera rolled, the neurons competed for control until only one remained. "We watched one neuron get bigger over time," Licht-man says, "and it took over the other neuron's territory." Most surprising, a neuron that is smaller at the start can outgrow the other. "You could never know that from single-time-point data," Lichtman says. "It altered our way of thinking of this competition." Watching movies like this could answer many questions, often with unexpected answers.

The temporal dimension also helps some scientists to study mechanisms within cells. Gary Sieck, professor and chair of the Department of Physiology and Biomedical Engineering at the Mayo Medical School in Rochester, Minn., says, "Using high-speed (about 500 frames per second) and high-resolution (less than one millimeter) confocal imaging, we and others have been able to determine dynamic changes in intracellular calcium signaling."

BREAKING THE LAW

<p>FREE, MODULAR IMAGE ANALYSIS:</p>

Courtesy of Lary Reinkling

The NIH's free ImageJ software is a modular image-processing tool. A variety of third-party add-ons have been written, and are available at http://rsb.info.nih.gov/ij/plugins/index.html. These two photos are from a collection of echinoderm embryos used for particle counting and analysis.

In many cases, the interesting action takes place at such a small scale that ordinary light can't image it. Since the 19th century, scientists have known that wave diffraction limits the resolution of light microscopy to about 200 nanometers. More than a decade ago, however, Stefan Hell, director of the Department of Nanobiophotonics at the Max Planck Institute of Biophysical Chemistry in Göttingen, Germany, was working on ways to circumvent this restriction.5

Hell hoped to push the resolution down to a handful of nanome-ters, and thought fluorescent labels might provide the key. His basic idea sounds simple enough: Sharpen a fluorescent spot by damping it at its periphery, which leads to scanning with a smaller spot, providing increased spatial resolution. To do that, Hell hypothesized that he could use stimulated emission deletion (STED).

Hell focuses laser light to a 200-nanometer spot, which excites molecules for fluorescence. A second, donut-shaped pulse of laser light interacts with the already-excited molecules, bringing them back to the ground state so that they no longer fluoresce. "The second pulse somehow wiggles the molecules, pushes them back to the ground state before they emit a fluorescent photon," he says. Using this strategy, Hell was able to get beneath the diffraction limit by more than fivefold. For even better resolution, Hell combines STED with his improved confocal microscope, called 4Pi, which uses two objective lenses focused on the same spot. With his STED-4Pi microscope, Hell is able to zoom in on bacterial membranes that are only 35 nanometers wide.

Hell speculates that his technique could also reveal substructures such as dendritic spines along a neuronal membrane, the intracellular Golgi apparatus, or the nucleus in a living cell. He adds, however, "This has not been done. I'm a pure physicist, and my job is only to work out the problems of physics." Nonetheless, some scientists expect great results from Hell's work. Douglas Taatjes, professor of pathology and director of the Microscopy Imaging Center at the University of Vermont in Burlington, says that breaking the diffraction limit "is truly an exciting breakthrough but is in the developmental stage."

As physicists advance the imaging technology field, it will expand its appeal to biologists and clinicians alike. Kenneth D. Tew, director of the Cell Imaging Facility at the Fox Chase Cancer Center in Philadelphia, says, "I think that the capacity to do FRET [fluorescent resonance energy transfer] imaging has been a major step forward." He and his colleagues used FRET to watch the enzyme glutathione S-transferase-π interact with JNK kinase, which is often associated with environmental stress.6

As Tew watches such protein-protein interactions, other scientists will explore their favorite new territories. From 3-D images and movies of neurons in action to live structures in the nano-meter range and pathways in process, today's imaging will unveil new discoveries in frame after microscopic frame.

Mike May mikemay@mindspring.com is a freelance writer in Madison, Ind.

Interested in reading more?

Magaizne Cover

Become a Member of

Receive full access to digital editions of The Scientist, as well as TS Digest, feature stories, more than 35 years of archives, and much more!
Already a member?
ADVERTISEMENT