Scientists from the European Molecular Biology Laboratory have used automated time-lapse imaging technology to create a platform for high-throughput RNAi screening that can overcome the limitations of existing high-content RNAi screens, according to research published this week.
The Heidelberg, Germany-based institution has also begun collaborating with researchers from other European nations to make the technology widely available, according to Jan Ellenberg, an EMBL researcher and co-author of the paper, which appears in the May issue of Nature Methods.
"After the completion of the human genome sequencing project, the task of functional genomics is to discover protein function genome-wide," the researchers wrote in the paper. "Currently, RNAi is the method of choice to study loss-of-function phenotypes in human cells by specifically suppressing the expression of virtually any desired protein-coding gene."
While several RNAi screens of human cells have been reported, these have "typically been based on endpoint assays of cells transfected in microtiter plates," the researchers added. "This allowed reasonable throughput but limited the information content of the phenotypic readout."
In doing a large-scale genome-wide experiment, "most people just chose a certain time point after RNAi knockdown to look at the phenotypes they are interested in … because of experimental limitations," Ellenberg told CBA News sister publication RNAi News this week. However, "phenotypes arise at different points in time and they are usually not static."
"Phenotypes arise at different points in time and they are usually not static."
As a result, if an assay is run shortly after gene knockdown, it is usually very specific, but "you only get the phenotypes of proteins that are relatively labile and are run down quickly — you only get part of the answer you are looking for," he said.
Recent notable examples of such screens include experiments conducted by academic members of the public-private The RNAi Consortium (see CBA News, 3/24/2006); researchers from the Max Planck Institute of Molecular Cell Biology and Genetics in Dresden (7/25/2005); and scientists from the Whitehead Institute for Biomedical Research (6/8/2004).
Conversely, assays run at late time points are more comprehensive but lead to specificity problems since it becomes difficult to differentiate between primary consequences of knockdown and secondary effects for rapidly turned-over proteins, the paper's authors wrote.
To tackle the problem, Ellenberg and his colleagues developed their own automated time-lapse microscope and image-analysis software to capture phenotypes of cells over time in large-scale RNAi screens.
"Genome-wide screening means you do on the order of 20,000 separate experiments, and phenotypic scoring is a lot of work even if you do it in microtiter formats or [on] microarrays," he said. "If you want to do this in a time-resolved fashion over several days you have many more data points you need to collect … [and] you deal with issues like cell synchronization and reproducibility at different time points.
"Being microscopists, we said, 'In cell biology, what you do if you look at things that change over time is you just take a video of the cells over the duration that it takes to carry out the process and you use that for screening,'" Ellenberg said.
To conduct their screens, the researchers took "advantage of the miniaturized RNAi delivery offered by transfected cell microarrays," developed in the lab of the Whitehead Institute's David Sabatini, "in which individually spotted siRNA transfection mixes are directly taken up from the solid phase by cells seeded on top of the array."
By spotting siRNA microarrays in live-cell imaging chambers, "we were able to perform time-lapse microscopy of HeLa cells on the arrays," the researchers wrote. "By massively increasing the throughput of fluorescence imaging and developing computerized analysis of the phenotypes by digital image processing, we established a fully automated high-throughput and high-content workflow of RNAi screening by time-lapse imaging."
The researchers imaged cells every half-hour for two days. Their homemade imaging platform consisted of an automated Olympus epifluorescence microscope with some modifications to make it more conducive to the long-term live-cell imaging being conducted. More importantly, they used EMBL-developed image-analysis software to extract the specific phenotypes from the time-lapse images.
High-content screening systems, such as those offered by BD Biosciences, Cellomics, Evotec Technologies, GE Healthcare, and Molecular Devices are capable of performing similar experiments. To wit, all of the aforementioned earlier screening experiments used one of these types of instruments. However, Ellenberg told CBA News that when his group started this work, that none of the platforms had the long-term live-cell modifications needed for this type of experiment.
"Since then, many systems have been developed for live-cell imaging, with proper incubation and other features," Ellenberg said. "Even so, those are not necessarily optimized to this specific assay, for the type of imaging we wanted to do, in terms of the illumination needed, for instance. They are more general platforms."
Ellenberg pointed out that the success of the screening system — which was validated in a pilot screen assaying cell division — is based not only on advances in microscopy but advances in "dealing with the kind of data you get from video imaging if you do it on such a large scale. It's a huge amount of pictures that you need to analyze, and [the Nature Methods paper described the] method of how the computerized imaging process is absolutely essential to extract the phenotypic data from the videos."
Although the RNAi screening method involves technologies not readily available to the average researcher, Ellenberg anticipates it is within reach of the major life science research institutions, which often open their facilities to visiting scientists.
For example, "at EMBL … visitors come to use [the] central infrastructure," he said. "EMBL each year has 2,000 scientists visiting, not only for screening but all kinds of things. This screening technology will also be available for other European scientists at EMBL, and we're already collaborating with people from different countries who want to take advantage of the technology."
Ellenberg also pointed out that the technology — especially the image-analysis algorithms — are very specific to his group's work, and would not be conducive to a straight open-source software type of situation.
"We are more than willing to make the codes available, but it's not a general software platform," he said. "It is very specific to this assay In terms of extracting phenotypes from the time-lapse images."
A version of this story appeared in the April 20 issue of RNAi News, a CBA News sister publication.