By combining highly parallel processing with long-term gene knockdown, researchers at Harvard Medical School and Cold Spring Harbor Laboratory have managed to design an RNA knockdown screening method that allows them to simultaneously achieve two conflicting goals: screen very large numbers of genes for their effects on cell growth, and detect genes that have fairly subtle effects.
Oh, and it's cheap, too.
The methods, are published in two papers - one focusing on breast cancer cell lines, the other on both breast and colon cancer - in the Feb. 1, 2008, issue of Science.
In gene knockout studies, cells lacking different genes usually need to be kept physically separate from each other to be able to ascribe effects on growth to the proper gene. And when a project investigates many genes, such separation gets complicated fast, necessitating robotics to handle the cell cultures: "It's not trivial in terms of cost or infrastructure," Michael Schlabach told BioWorld Today.
Schlabach is the first listed author on one of the papers, though he shares equal first author credit with colleagues Ji Luo, Nicole Solimini and Guang Hu. He also is a co-author on the second paper. The research was led by senior scientists Stephen Elledge and Gregory Hannon, respectively.
The new method combines stable shRNA transduction with so-called "barcoding" of the short hairpin RNAs used for gene knockdown. The barcoding, or the addition of an identification sequence to the shRNA, allows the researchers to transduce different cells with many different shRNAs without needing to keep them physically separate.
In their paper, titled "Cancer Proliferation Gene Discovery Through Functional Genomics," the researchers transduced two colon cancer cell lines, a breast cancer cell line and a normal breast tissue cell line with a total of more than 8,000 shRNAs targeting roughly 3,000 genes.
Cells were grown for several weeks; over that time, cells transduced with shRNAs that affect proliferation grew more slowly than others. And fewer cells produce less of the barcoded shRNAs.
By using microarrays with the barcodes, the scientists were able to quantify how much of each shRNA was present in the mixture, which in turn allowed them to infer which shRNAs were affecting growth.
In the cancer cell lines, 3 percent to 5 percent of genes affected cell growth when they were knocked down, while 25 percent of genes affected the growth of the normal breast cell line when they were knocked down. The genes that affected growth in the different cell lines overlapped, including 19 genes whose knockdown affected cell growth in all four. But the researchers also identified a number of genes whose knockdown specifically slowed growth in one or more cancer cell lines. They wrote in their paper that those cell-line specific genes "are particularly interesting because they . . . represent potential cancer-selective drug targets."
Nicole Solimini told BioWorld Today that "we do think that there are interesting targets in the papers to pursue."
Still, the researchers pin higher hopes on further scaled-up experiments. "When we expand this screen to the genome scale, we might stumble across proteins that have inhibitors," but have not been implicated in cancer cell proliferation, Luo said.
As researchers have developed the technical means to screen large numbers of genes, one new problem has been that such large-scale screens bring with them an increased risk of false positives. In genomewide association studies, the answer to that problem has been to set the statistical criteria extremely high. But that in turn means that subtle effects currently have next to no chance of being detected in genomewide association studies.
In the experiments described in Science, the research team counteracted the risk of false positives in two ways: by transducing very large numbers of cells, and then culturing them through many generations.
Luo likened the effect of culturing the cells for several weeks to that of compound interest. "With sustained knockdown, more subtle differences can be magnified over time," he said.
In addition, the sheer number of cells the researchers are able to transduce with their multiplexing method makes it much easier to protect the findings against false positives than is the case in genomewide association studies. Each shRNA is delivered into roughly a thousand cells, and each experiment is repeated three times, giving the researchers "a reasonable degree of statistical certainty" that the effects they see are not false positives, Schlabach said.
The team hopes that its work will be complementary to the Cancer Genome Atlas, another project whose goal it is to take a comprehensive view of how cancer cells differ from normal ones. (See BioWorld Today, Dec 22, 2005.)
The Cancer Genome Atlas "takes more of a physical approach," mapping mutations to see where cancer cell DNA differs from that of normal cells. "We take more of a functional approach . . . [that] takes an unbiased view," Luo said. "Regardless of what mutations there are, we are interested in what we can inhibit in the existing genetic context."