A group led by scientists at the Weizmann Institute of Science has used time-lapse fluorescence microscopy and advanced image-analysis techniques to examine changes in protein expression levels and localization in individual living cells.
The research is the latest in a recent trend of using cellular imaging techniques to conduct dynamic proteomic studies in order to better understand protein regulation in cells. It may also help spawn a niche market for high-content analysis technology — if big-ticket imaging systems become more flexible and less costly, a research group member told CBA News this week.
Researchers looking to determine protein expression levels in cells typically turn to techniques such as using DNA microarrays to measure mRNA expression, or mass spectrometry. However, cellular imaging — in particular live-cell microscopy — is quickly gaining popularity for this application and could even complement these other inherently in vitro techniques.
"There are quite a few differences between imaging and [other techniques]," said co-investigator Ron Milo, currently a research fellow in the department of systems biology at Harvard Medical School. "One is the ability to analyze single cells. Also, if you want to have temporal resolution, this approach is more conducive."
Milo participated in the study, which appears in the July issue of Nature Methods, while still a researcher in the department of physics of complex systems at the Weizmann Institute, based in Rehovot, Israel, along with colleague Uri Alon. Leading the study was Alex Sigal, who works in Weizmann's department of molecular cell biology. Other contributors included scientists from the Whitehead Institute for Biomedical Research, Harvard Medical School, and Stanford University.
Cellular imaging — in particular live-cell microscopy — is quickly gaining popularity for this application and could even complement time-tested techniques such as DNA microarray analysis and mass spectrometry.
According to Milo, imaging approaches are superior to microarray analyses when researchers are interested in actual protein levels rather than just mRNA levels.
"If you have a protein-reporter fusion, you can do that, but you can't with a microarray," he said. "Obviously microarrays have had an immense impact and are great for answering different questions, but for these types of questions, it seems that microscopy is the way to go. I think more and more people will begin using these types of approaches once the tools become more widely available."
As described in the Nature Methods paper, the researchers generated 20 clones of human lung cancer cells, with each clone expressing a yellow fluorescent protein molecule tagged to a different nuclear protein. To label the proteins, they used a previously described method known as the "central dogma tagging approach," in which a retrovirus is used to insert the YFP coding region into the appropriate regions of the cell's genome, resulting in endogenous expression patterns.
The scientists subsequently acquired images of individual cells every 10 minutes for 48 to 60 hours using a homemade automated imaging platform, which comprised a Leica fluorescence microscope, YFP filter set, motorized stage, CCA camera, incubator, and Media Cybernetics control software.
To analyze the images, the researchers used a custom-written image-analysis algorithm developed by Milo using Matlab. In addition, they used a version of an algorithm in CellProfiler — the open-source image-analysis software being developed by Whitehead Institute scientists (see CBA News, 7/4/2005) — to segment the nuclei in the cellular images.
They used the image-analysis algorithms to examine how the localization and expression levels of the nuclear proteins changed during different stages of the cell cycle. Because each cell was in a different stage of the cell cycle during the experiments, the scientists used an in silico method to synchronize the cells without perturbing them, and established a normalized time base for all the examined cells.
The researchers found that of the proteins whose cellular localization had been previously characterized, 74 percent showed the same localization patterns as revealed by the YFP-tagging method. This discrepancy could be a result of several factors, Milo said — in particular, different cell lines, buffers, and methods used in previous experiments.
Furthermore, the scientists discovered that approximately 40 percent of the nuclear proteins examined demonstrated a dependence in expression level — and in some cases, localization — on the cell cycle.
Although the researchers tackled only 20 cellular proteins in the Nature Methods study, they have actually developed a library of about 300 clones, each of which has a different cellular protein labeled with YFP. However, for the purposes of this study, they focused on proteins contained to the cells' nuclei.
"Although the entire clone library is an order of magnitude larger, the whole process [of imaging] is quite time-consuming, so we had to choose a feasible number to work with," Milo said. "Also, when you are trying to do image analysis, choosing to work with a subset that has nuclear localization makes the whole approach must simpler," he said, because reliable image-analysis programs exist for segmenting the nucleus from the rest of the cell.
This is not the first time scientists have used an imaging approach for proteomics studies. In fact, the technique is gaining popularity as researchers have begun to realize the importance of looking at proteins in intact cells to paint a more accurate picture of their activity.
One notable example of this kind of research is the "location proteomics" approach developed in the lab of Robert Murphy at Carnegie Mellon University, where researchers have developed methods for high-content imaging of sub-cellular protein localization patterns (see CBA News, 2/15/2005).
More recently, Edward Marcotte's lab at the University of Texas at Austin used cell chips and automated imaging to examine phenotypes and protein localization patterns in yeast cells in response to various stimuli (see CBA News, 5/12/2006); while Jonathan Weissman and colleagues at the California Institute for Quantitative Biomedical Research also quantified protein expression levels in yeast cells using GFP tagging and flow cytometry (see CBA News, 6/23/2006).
These examples represent only a smattering of the work that has recently been done in this area. Still, the research done by the Weizmann scientists is one of the first published examples in which scientists were able to characterize protein activity in living human cells in real time, and the researchers said this may be more physiologically relevant than previous methods due to the endogenous expression technique and the minimally disturbing in silico synchronization method.
One result of all the work as a whole is that an important niche market may be evolving for high-content automated imaging platforms such as those sold by Cellomics, GE Healthcare, Molecular Devices, Evotec, and BD Biosciences — not to mention flow cytometers and imaging cytometers.
"We are presently extending this dynamic proteomics approach to cover a large fraction of the proteins expressed in this cell line," Milo and colleagues wrote in Nature Methods. "The present [central dogma]-tagging protocol can generate [about] 1,000 tagged and identified clones in a few months.
"High-speed and high-throughput microscopy systems can generate time-lapse movies on hundreds of clones per week per microscopy station," they wrote. "We therefore believe that with the present approach, it will become feasible to study proteome dynamics of human cells under diverse conditions and stimuli."
Before any of this happens, however, big-ticket imaging systems may need to become more flexible and less costly. Milo told CBA News that his team decided to put together its own automated imaging platform after examining a few turnkey options.
"One problem, I think, is that these are quite expensive — typically around $500,000 each — and although they had very nice features, it seemed like they weren't always that flexible," he said. "Some things they can do much better than we can do with any homemade system, but for other things where we wanted to be able to configure it differently and play around, it didn't have those options."
The image-analysis landscape, Milo said, is progressing more quickly. The group initially tested programs such as MetaMorph and ImagePro, but "it seemed like for the time-lapse movies with the specific problems we had, it was very difficult to make these software packages work for us in a productive way. Therefore we decided to write [one] ourselves."
Milo said if they started the project today, however, they might explore other options — including the CellProfiler software, which has since been further developed into a more complete image-analysis package.
"I believe that a vision of the field is that there will be tools good enough to do that without the need for people to write their own code, which is very demanding, and makes everybody have to invent the wheel again and again," he said.
— Ben Butkus ([email protected])