By Mike May

This Roche scientist is using a liquid-handling robot for automated RNAi, which mixes siRNA and transfection reagents for use in cellular assays. (Image courtesy of Roche.)
In 2001, the publication of the human-genome sequence triggered excitement and trepidation. "At that time," says William Marshall of Dharmacon (Lafayette, CO), which is owned by Fisher Scientific (Hampton, NH), "we suspected that we had decades of work to understand the functions and interrelationships of all of the genes." Unraveling what the thousands of human genes do, though, and even putting them to work might go much faster, all because of automating a powerful gene switch - RNA interference, or RNAi.

Rachele Kardon of Ambion (Austin, TX) - acquired by Applied Biosystems (Foster City, CA) early in 2006 - says, "RNA interference is a cellular mechanism that provides a tool to silence specific genes in cultured cells or animal models." It knocks down genes for a wide range of organisms, from C. elegans to humans. To automate that process, researchers look at three general steps: getting the short-interfering RNA (siRNA) that drives the mechanism, putting the siRNA into cells, and, finally, analyzing the results.

Laying out the libraries
"In the early days of RNAi, which was only about five years ago," says John Reidhaar-Olson of Roche (Nutley, NJ), "we designed our own siRNAs. Everybody did." Today, scientists at Roche, and most other academic and industrial labs, simply buy siRNAs. "They are quite good," says Reidhaar-Olson.

In addition, Eric Lader of Qiagen (Germantown, MD) says, "To automate RNAi, you need libraries." Lader's company made the first siRNA library in 2002, consisting of only a few hundred siRNAs. "At that time," says Lader, "it was a shock that anyone could screen several hundred siRNAs at once." One year later, Qiagen made a genome-wide library for Novartis.

In the past, researchers found functional siRNAs by making lots of them. Now, companies use algorithms that design the siRNA against a specific gene. In the March 2004 issue of Nature Biotechnology, Dharmacon scientists described a technique for rationally designing siRNAs. Marshall says, "The concept of rational design is now validated by labs around the world, but we did it first." In addition, Marshall notes that pooling - making mixtures of siRNA that all target the same gene - can increase specificity. "The mixture is more specific than the individual siRNAs," says Marshall, "because of dilution and a competition phenomena."

At Ambion, Kardon says that 80 percent of the siRNAs generated by their algorithm can knock down a gene by at least 70 percent. In addition, scientists can buy siRNAs - usually a few versions - for almost any human gene, not to mention rat and mouse. Kardon adds, "We can customize a library to make it easier for a researcher to use." Moreover, a library today can have as few as 50 up to thousands of siRNAs. Libraries can also include positive and negative controls, which look for desired and undesired results, respectively.

When RNAi came along in 1995, scientists used it on one gene at a time. "As you move up to a thousand genes," says Reidhaar-Olson, "it becomes more difficult to do that manually."

Dharmacon’s bioinformatics algorithms and siRNA chemical modifications produce functional and specific siRNA molecules. (The company provides plated siRNAs.) Design and production steps include:
1. Advanced rational siRNA design algorithms select highly functional siRNA sequences.
2. Sequences with high identity to other genes, microRNA (miRNA), seed regions, and toxic motifs are eliminated.
3. Chemical modifications are applied to the siRNA sense strand.
4. Additional chemical modifications are applied to the antisense strand further increasing specificity.
5. As an optional step, siRNAs can also be pooled to further reduce off-target effects. (Photo courtesy of Dharmacon.) Click here to enlarge.
Getting inside
With siRNA on hand, getting it in the cells comes up as the second challenge. Transfection makes up one of the most popular delivery methods, and this can be sped up in various ways.

Dharmacon supplies products for reverse transfection. Instead of growing cells in 96- or 384-well plates and then adding the siRNA to cells, Dharmacon delivers the plates with the siRNA already in the wells. Other companies do this too. "If you are going from manual to automated," says Marshall, "that innovation can knock off a couple days of preparation."

Part of the key to automating RNAi, though, revolves around a complete system. "Transfection and chemical agents must be tested for each model system," says Kardon. "You want to optimize transfection, but you also need positive and negative controls to gauge the variability between plates and between runs on different days."

At Roche, the use of liquid-handling robots speeds up the process. Reidhaar-Olson says, "We've automated some of the steps, such as transfecting cells in a 96-well plate." He adds, "We keep human intervention all the way through to provide flexibility in design."

Following the data flow
The blessing and the curse of automated RNAi comes from data - there is tons of it. Lader says, "A genome-wide library can make up 80,000 data points, with a couple siRNAs per gene. Duplicating the transfections ups that to 160,000 data points. If you are looking microscopically at individual cells, that could lead to 200 cells per well." Add it all up, and you get 32 million data points from one genome-wide run.

The results from RNAi, though, can be analyzed in different ways. Maybe scientists just want to know what genes got knocked down, so they just measure the mRNA with quantitative PCR. Robotics can keep such processes moving fast. For example, Qiagen developed robotics platforms for RNAi, which are being used at the National Cancer Institute and the Max Plank Institute in Berlin.

An even faster future
Despite the options that are available for automating RNAi research, Lader still sees some scientists and companies lagging behind. He says, "Lots of researchers still think gene by gene. Instead, they could ask bigger questions, like: What happens if I knock down the entire kinase pathway instead of just one gene related to it?" Moreover, some big labs continue to steer clear of some forms of RNAi. "Even at large labs with incredible expertise in automation and assays," says Lader, "some do no RNAi, because it is different from using small-molecule inhibitors."

Also, just because something is possible does not mean everyone will believe in it. For example, Reidhaar-Olson says, "We have not decided to do genome-wide screening with RNAi. We are not convinced that it is beneficial to look across all of the genes." Nonetheless, Reidhaar-Olson does see great value in screening specific groups of genes, especially the so-called druggable genome.

Plates can be purchased with siRNA already in the wells to speed processing; scientists just need to add cells. (Photo courtesy of Dharmacon.)
For the most part, most anyone involved in RNAi would probably agree with Reidhaar-Olson's conclusion, "It is amazing what is happening in this field!"