By Adrienne J. Burke
Some call it the most revolutionary biological tool since polymerase chain reaction. Nobel laureate and Biogen founder Phil Sharp says it will “fundamentally change how we do cell biology.” And technology licensing officers at the Carnegie Institution and MIT, which sell commercial rights to two separate pools of patents for RNA-mediated-interference technologies, say that these could well be the most significant inventions, at least in biology, ever to cross their desks.
However you measure its importance, there’s no question that RNAi has electrified the functional genomics field. In a nutshell, RNA interference is a phenomenon by which a double-stranded RNA inserted into a cell deactivates the gene bearing its complementary sequence by triggering the destruction of its messenger RNA. A follow-on discovery revealed that a small, 21-nucleotide duplex, called short interfering RNA, performs the same trick in a mammalian cell when it is incorporated into a protein complex that targets and cleaves mRNA. Experimentally demonstrated in a model organism only five years ago and for the first time in a human cell in 2001, this knockout method is now touted as the single best way to study how genes work.
While investors go wild about the therapeutic possibilities for siRNAs (VCs have already poured more than $75 million into companies trying to figure out how to administer them directly to patients), a whole new business sector is emerging around the use of RNA interference for gene-function analysis and drug target validation.
Pharmaceutical companies including Abbott, AstraZeneca, BMS, GSK, J&J, Merck, Novartis, Pfizer, Pharmacia, and Procter & Gamble have launched RNAi target validation studies, some of them in dedicated new gene silencing or RNAi departments. Among companies that have turned the technology into a new brand of high-throughput validation service for pharma are Exelixis and Sequitur as well as newcomers Atugen, Cenix, DevGen, GeneExpression Systems, Genetica, Intradigm, and Ribopharma. And tool and reagent providers are surfacing like worms in a rainstorm to sell synthetic siRNA oligos, vectors and transfection reagents, and siRNA selection and design software. A survey of the sector turns up at least 35 such companies.
Most remarkably, some pioneering scientists have already applied the technology to entire genomes. RNAi studies of all C. elegans and Drosophila genes are soon to be published. One company and at least three public-sector groups have announced plans to systematically screen the whole human genome with siRNAs. And Harvard Medical School awaits funding to open what would be the first RNAi core facility to provide Drosophila genome screening services.
Surprisingly, in a field that’s been burned time and again by over-hyped technologies, this one seems to be without detractors. Explaining that because “the technology isn’t 100 percent robust yet” and that “others are already doing” high-throughput genome scans, Christopher Austin, senior advisor to NHGRI’s director for translational research, says in an e-mail to GT that the institute “decided not to do anything in siRNAs for now.” But of more than 30 people interviewed for this story, none expressed doubts about the usefulness of the technology for gene function analysis, and several consider NHGRI’s decision a mistake.
As Tony Hyman, a group leader and director of the Max Planck Institute of Molecular Cell Biology and Genetics in Dresden, Germany, points out, this is a use for the human genome sequence that five years ago no one could have envisioned. Now, picking siRNA sequences seems to be exactly what GenBank was made for.
Taking the early winnings from the RNAi revolution are IP owners and their attorneys. Carnegie Institution of Washington is lead licensor of what are widely considered to be the fundamental patents in the area. Andy Fire of Carnegie and Craig Mello of UMass Medical School coined the term RNAi in Nature in 1998 when they showed that long, double-stranded RNA effectively silenced targeted nematode genes.
While Boston biotech attorney Louis Myers says he’s following the progress of more than 30 related patent applications, in February Fire and Mello became the first in the US to be awarded an RNAi patent.
For an up-front fee of $35,000, annual payments of the same amount, an additional $50,000 upon each of three milestones — one being the issuance of the first patent — and a “reasonable” royalty for “sales of products embodying the invention or produced using the same,” anyone can use, make, or sell products incorporating Fire and Mello’s discoveries. As of last month, about two dozen companies had taken licenses, adding up to nearly $2 million so far for Carnegie, UMass, and the inventors.
MIT, meanwhile, is licensing a pool of discoveries that enable RNA interference specifically in mammals. An interferon effect induced by longer strands prevented the Fire-Mello approach from working in higher organisms. But in 2001 Tom Tuschl, then of the Max Planck Institute for Biophysical Chemistry in Göttingen, Germany, used 21-mer RNA duplexes that included two-nucleotide 3’ overhangs to silence genes in human embryonic kidney and HeLa cells. Tuschl’s work was based on a series of earlier experiments he carried out with MIT professor Phillip Sharp, Phillip Zamore of UMass Med School, and Whitehead’s Dave Bartel. MIT handles the bundled intellectual property for the respective institutes.
Many use the word “murky” to describe the landscape of RNAi patents, and some are particularly skeptical that the Sharp group will be granted one. Some speculate that Fire and Mello’s work does indeed extend to mammals; others say that a pair of German researchers, Roland Kreutzer and Stefan Limmer, whose company Ribopharma already has secured a European patent, had prior art; another bystander says Gary Ruvkun’s genetics lab at Harvard published the earliest evidence that small RNAs were involved in gene regulation. Regardless, MIT sells rights to biotech and pharmaceutical companies that wish to use Tuschl’s recipe to make, or have made, siRNAs for internal target discovery and validation research, and it granted exclusive therapeutic uses to Alnylam, a company founded by Bartel, Sharp, Tuschl, and Zamore.
The university also sold “co-exclusive” rights to synthesize and sell siRNAs to four vendors, selected for their ability to control quality and meet demand. When they buy 21-mer RNA oligos from these vendors, labs get MIT’s go-ahead to use them for functional genomics research.
Until patents are issued, of course, others may, and do, sell synthesized siRNAs. Some, such as Molecular Research Labs and Eurogentec, advise in fine print that their customers contact MIT for licensing details; others like Australian company Benitec maintain that their own IP overrides MIT’s. But it appears that most buyers are playing it safe by sticking with the companies with MIT’s seal of approval. A recent study by market research firm Bioinformatics LLC estimates that three of the MIT licensees — Ambion, Dharmacon, and Qiagen — are responsible for as much as 80 percent of siRNA sales worldwide. Bruno Poddevin, VP of Proligo, the fourth licensee, notes that his siRNA business has more than doubled in the past year and says the report underestimates his market share.
Some observers say these companies were duped into paying for methods that aren’t yet patented. MIT won’t tell how much in annual subscriptions or royalties it collects from them, but considering the way the sector is exploding, even if the patents never issue, the cachet the licensees get from MIT’s endorsement could be well worth it.
Ambion says 1,000 people visit its online RNAi Resource Page per day. And Qiagen’s product manager for gene silencing, Michael Sturges, says he draws throngs to his siRNA seminars: “I had one seminar, an hour talk, and 110 people showed up. That’s not typical for a vendor seminar.” Asked how business is going, Sturges gushes, “It’s fantastic! This technology is on fire.”
Front Line Strategic Consulting estimates the current market for siRNAs at $38 million and predicts it will more than quadruple within five years. Robin Rothrock, who polled 800 researchers for the Bioinformatics study, says one of its most striking findings was that three-fourths of scientists using siRNAs began doing so during the last year. “Based on the responses that we got, the market could triple within the next 12 months,” she says. “It’s just amazing.”
Adoption and Adaptation
Seldom has a technology been adopted so quickly by so many biologists. A PubMed search for “‘siRNA’ OR ‘RNA interference’” brings up more than 750 papers published in the five years since Fire and Mello ignited the industry. According to Dmitry Samarsky, director of technology development for Sequitur, which sells custom siRNAs for target validation that it says are manufactured under the MIT license, “The technique has become so popular that people are trying to adjust their [research] to the technology … asking, ‘What should I knock down to use this?’ It’s fun to use and could help you to get good papers.”
Phil Zamore, who has three siRNA patent applications under his belt, cites two precedents for this sort of craze: “the use of bacterial enzymes to clone things in plasmid vectors and PCR.” The reason RNAi has taken off? “’Cause it’s so unbelievably cool,” Zamore gasps.
What renders him and others so breathless is the first reliable, reproducible way to silence a gene and connect genotype to phenotype. Where microarrays allow a user to correlate an overexpressed gene and a tumor, siRNAs provide a causal link. They’re far easier and quicker to use than that high-maintenance gene-silencing tool, the knockout mouse. There’s no new instrument or system required to do RNAi, and the reagents are readily available to researchers on tight budgets. (Chemically synthesized siRNAs cost upward of $250 per pair and a single gene experiment typically requires three to five pairs; a less costly option is to generate hairpin RNAs in vitro with plasmid- or virus-based DNA molecules.)
Tom Tuschl, the 36-year-old siRNA star who moved his lab from Germany to New York’s Rockefeller University in January, has a special arrangement with Dharmacon whereby he synthesizes his own siRNAs with the company’s reagents. “If I want to knock out a gene it takes one day to synthesize the RNA, the second day I transfect, and two to three days after transfection I look for my phenotype in my cells. Within four days I know whether the gene is essential or not. If I know what to look for I can do the downstream analysis and I can work on essential genes that no mouse geneticist can work on.”
Tuschl notes, “There are effective antisense technologies that use DNA oligonucleotides, but these reagents were never made publicly available. They were inside biotech companies.” Academics couldn’t get access. Not so with RNAi. Anybody who wants to do it can, and just about everyone is.
Sequence and Snafus
That’s not to say that the art of RNA interference has been perfected. Far from it. It’s simple enough to copy and paste a few 21-base sequences from GenBank into a vendor’s e-mail order form, but consider some of the things that can go wrong in an siRNA experiment: Transfection — just getting the siRNA into cells — doesn’t take; you induce toxicity by adding too much siRNA to a cell; because of SNPs or database errors, the GenBank gene sequence that you’ve designed your siRNA against is not the same one that exists in the sample in your dish (an obvious mistake, Zamore notes, “but even smart people forget it”); the 21-base segment you’ve selected is homologous to more than one gene and ends up silencing the wrong one; darn, you should have confirmed before buying that the sequence in your mouse tumor study has homology with a human gene; or, oops, you ordered your sequence from the vendor in a 3’ to 5’ format rather than the recommended 5’ to 3’ direction.
Other snafus remain to be explained, such as why high GC content can, but won’t always, trip things up; why the rates of suppression vary among siRNAs; or why off-target effects, as one pharma team that will soon publish a paper on the topic claims, seem to be an inherent feature of short oligos.
Several companies see opportunity in helping researchers select siRNAs that work. Dharmacon, for instance, hawks its Smart Selection tool. The company says the siRNAs it chooses have a 99.99 percent chance of knocking down the activity in your chosen gene by at least 75 percent. Cenix BioScience, a startup spun out of EMBL in 1999 to apply RNAi techniques to genome-wide screening and target validation, says its tool for picking potent siRNAs could give its new partner Ambion an edge over Dharmacon. Baltimore informatics firm ReceptorBase says its new MapRNAi product can identify every potential RNAi and associated networks for a given genome. And Compugen, which argues that you need to consider splice variants when you select your siRNAs, struck a deal last September to use its transcriptome database to do just that for 27,000 genes in Novartis’s database.
On the public-sector side, Fran Lewitter’s bioinformatics team at Whitehead, with input from Tuschl, unveiled an siRNA design program in February. The online tool, visited by more than 300 people worldwide in one month, helps users avoid known polymorphic sites, exclude regions with exon boundaries, choose sequences specific to the targeted gene or gene family, and account for alternative splice variants when selecting siRNA sequences. Lewitter’s group is also at work on a public database called sirBank that will catalog sequences known to suppress gene activity.
But Zamore says the ability to select an siRNA is overrated. “I don’t know anybody who can’t find one that works. It’s a very stochastic thing. It’s not that hard to find a good siRNA.”
Tuschl agrees, and shakes his head at companies hawking “black box” siRNA design programs, and is particularly irritated with vendors that don’t reveal the exact sequence they’ve sold to a customer. “There are no stringent rules for design except they should be 21 nucleotides in length and have a two-nucleotide overhang. Essentially any sequence would qualify,” he says. If you make one siRNA for every gene you want to look at, the general probability of success, he says, is 70 percent. “So you may miss 30 percent of genes that could be important. … [But] you file IP on those 70 percent.” Public labs without a lot of money to throw around on reagents are one thing, but pharmas that get mired in siRNA selection rules, Tuschl says, are just “delaying the process and they’re going to miss the race.”
Cagey and Competitive
The endgame of this race, of course, is siRNA-validated drug targets. And as pharmas hone their techniques, competition is heating up. An attendee to an siRNA conference in San Diego in late March says the rivalry is evident: “A lot of companies were very careful about what they were disclosing. You got the feeling they had patents or trade secrets.”
Several pharmas and biotechs, presumably interested in maintaining confidentiality about the targets they’re looking at, have purchased the MIT license to make their own siRNAs. But vendors of the compounds, who say pharmas are still their biggest customers, offer some clues about how drug companies are applying the technology.
The small-batch orders that Dharmacon gets indicate that pharmas are revisiting known targets in specific disease categories, not screening the entire human genome with siRNAs. “We’ve done orders for 500 to 1,000 genes … and then you talk to them and they say, ‘We’re still putting out data, we’re buried.’ These are places with high-throughput platforms,” CEO Steven Scaringe says. “To our surprise, there’s much more interest in buying five, 20, or 30 siRNAs … and we’re dealing with large pharmas.”
But Proligo’s Bruno Poddevin says some pharma customers are indeed building genome-wide siRNA libraries and “biotechs are doing what they can afford — some in the range of a few hundred or a few thousand oligos.”
Pharmas are also striking deals with biotechs such as Cenix BioScience and Ribopharma, both in Germany, that provide RNAi target discovery and validation services. Ribopharma’s Roland Kreutzer says dozens of requests came in when he announced his intention to begin such a service. And Christophe Echeverri, CEO of Cenix, which has already conducted RNAi screens of the entire Drosophila and C. elegans genomes and soon will begin gene family-focused and genome-wide RNAi screens in human cells, says his company is negotiating several target-discovery and -validation deals with pharma customers.
The Wall Street Journal reported last December that Exelixis had provided Pharmacia new Alzheimer’s drug targets by conducting RNAi experiments on 10,000 C. elegans genes, and in March Sequitur expanded its target validation service contract with Bristol-Myers Squibb to include RNAi studies.
Mark Cockett, BMS’s executive director of applied genomics, says his staff “is investigating the potential of RNAi and systematically applying it to target and biomarker validation as well as target identification.” In January he told GT he would like to build a library of reagents for performing RNAi experiments on every target-class gene.
A Merck scientist who asked that his name not be used says his team has screened hundreds of genes with siRNAs, and he assumes “other people are doing similar things.” He says he doubts, however, that any pharma is ready to conduct siRNA screening on a genome-wide scale — “in practice, a pharma would probably just revisit the genes it knows.” Instead, whole genome screens are more likely to be the pursuit of academia, he says.
Quest for Funding
Make no mistake, academic labs are raring to go. Max Planck’s Tony Hyman, a Cenix cofounder who also sits on a selection committee for the EU’s 6th Framework program, says the EU fielded numerous funding requests this year from labs proposing to do genome-wide RNAi screens in cancer.
Some say that, given current costs, such projects are out of reach. At $250 per duplex, even if a lab succeeded in picking just the right one for each gene, it would need to spend nearly $8 million to screen an entire human genome. “And I’m going to go out on a limb and say [you’re] stupid if [you’re] thinking of doing one per gene,” says the Merck scientist, acknowledging the less-than-100-percent chance that a given siRNA will suppress a gene.
Norbert Perrimon, whose lab at Harvard took nine months to complete a genome-wide RNAi Drosophila screen, agrees that pulling that off in a mammalian genome will be expensive and time consuming. Even once the obvious unknown is solved — “I can’t find anyone who can tell me exactly how many human genes there are anyway,” Perrimon says — he predicts it will take a few years to complete a full human-genome siRNA screen.
But earlier this year, René Bernards at the Netherlands Cancer Institute and Julian Downward of Cancer Research UK announced plans for a joint human-genome-wide RNAi project to determine function of any genes with a role in cancer. They’re starting out with 8,000 genes, but told the Lancet in February that, if all goes well, they could get through 35,000 by mid-2004.
Hyman’s own lab has teamed up with Vienna’s Research Institute of Molecular Pathology, EMBL, and the Sanger Institute to raise money for an siRNA screening network. “This is a project where one group makes the library, one group does the screening, according to our expertise,” he says. If the money comes, Hyman predicts that an entire human genome screen could be complete in 2004. The cost? Anywhere between $300,000 and $3 million, he estimates. “There are cheap and easy ways,” Hyman says. “If you do it with chemical 21-mers it’s expensive, but if you make RNAi yourself from cDNAs and chop it up yourself, it’s cheaper.”
Greg Hannon has proven that by making hairpin siRNAs from standard DNA oligos, a technique he is credited with perfecting, the job can be done on the cheap. Hannon’s Cold Spring Harbor lab took about eight weeks to screen some 5,000 human genes with three homemade hairpin constructs apiece. In all, he intends to screen 9,300 genes chosen for cancer relevance or pharmaceutical accessibility. With $1 million from NCI, another $4 million from DOD, and some corporate backing, Hannon says he has enough cash to finish the project and then do it over again.
Ultimately, Hannon plans to make his siRNA libraries available to the public through multiple vendors. Perrimon, meanwhile, awaits NIH funding to build the world’s first centralized RNAi screening facility for Drosophila at Harvard Medical School. He envisions a shop equipped with robotics and a staff of competent technicians who could run screens on cell-based assays submitted by outside scientists.
Such ambitious initiatives point up the prominent role that the public sector will play in propagating RNAi technologies and data. While IP issues get sorted out and pharmas race each other to validate targets, the public availability of siRNA screens and libraries will ramp up functional genomics research and generate innumerable new commercial opportunities — RNAi chips, databases, customized high-throughput screening platforms and robotics are but a few.
NHGRI, however, is looking for some answers before it gets behind any sort of siRNA project. Institute advisor Chris Austin suggests that the agency needs to do more investigating. He wonders how completely genome-wide projects such as Hannon’s will cover the genome and in what time frame, about the best type of siRNA to use, how much that would cost, and how they’ll manage quality control and variances in gene activity suppression.
To be sure, the field is nascent, with new research being published every week and patent applications being filed almost as frequently. But the pioneers admonish those who are waiting for improvements before jumping in. Holding off on launching RNAi studies because the technology is not robust enough “would be a self-fulfilling prophecy,” Hannon warns. “If you don’t take the step down this road, you create inertia.”
Tom Tuschl echoes those sentiments. Labs waiting for better selection tools or cheaper oligos are just wasting precious time, he says. “Why not just go for it?”
Hyman says the RNAi revolution offers a lesson to people who couldn’t see how the human genome sequence would be useful: “You don’t know until you do it” how other scientists are going to make use of something. As for whole-genome siRNA screening, he says, “The technology is ready to do it. There’s no question about it; the time to do it is now.”
THE PATENT PIONEERS
More than 30 patents have been applied for in the area of RNA interference. Here are the inventors and assignees of the early wave of methods and technologies. Only the first two patents listed here have issued so far.
Andy Fire & Craig Mello — Carnegie Institution & University of Massachusetts
US Patent 6,506,559 B1, issued January 2003. Demonstrates how specially designed double-stranded RNAs can silence targeted genes.
Roland Kreutzer & Stefan Limmer — Ribopharma
European patent 1144623 B1, issued August 2002. Method for inhibiting expression of gene that impedes or prevents apoptosis of tumor cell by insertion of dsRNA of less than 25 nt. World intellectual property organization application WO 02055693 A2, filed January 2002. A method of inhibiting expression of a target gene with dsRNA that is a maximum of 49 nt in length with 1-4 nt overhang.
Tom Tuschl — Max Planck Institute
WIPO application WO 0244321 A2, filed November 2001. Short interfering RNA discovery, shows how 19-23 nt RNA fragments are the sequence-specific mediators of RNA interference.
Phillip Zamore — University of Massachusetts
WIPO application WO 03006477, filed January 2003.
Invention of in vivo production of siRNAs.
Tom Tuschl, Phillip Zamore, Phillip Sharp, David Bartel — Whitehead Institute
US patent application 20020086356 A1, filed March 2001. Demonstrates that 21-23 nt fragments are the sequence-specific mediators of RNA degradation.
US patent application 20020162126 A1, filed May 2001.
Describes methods for attenuating gene expression in a cell using gene-targeted dsRNA. The dsRNA contains a nucleotide sequence that hybridizes under physiologic conditions of the cell to the nucleotide sequence of at least a portion of the gene to be inhibited.