An RNAi Rogue's Gallery

High-throughput RNAi libraries for mammalian cells are nearing a new maturity with the first wave of virally transfected short hairpin (sh)RNA libraries finally making it into published work. Developers are still making major and frequent improvements, so comparisons are difficult. "Most of the RNAi field is unfortunately anecdotal," says David Sabatini of MIT's Whitehead Institute. "That's a very dangerous thing."

Commercially available short interfering (si)RNAs don't match the sophistication or long term knockdown of vector based systems that produce shRNAs, but they're more established and require less work. Several groups report that standardizations and comparisons across library types are coming. Until those reports are available, here's how available options measure up.

Commercial siRNA

For quick and dirty screening, commercially available siRNAs are a good place to start. They've been around the longest, and companies have poured vast resources into creating and validating sequences that give robust silencing. "If you have a cell line that you can transfect and an assay that's relatively short-term, siRNA is a perfectly good approach - it works great," says William Hahn, associate professor of medicine at Harvard University.

The effect from siRNAs lasts only from few days to a week. Some companies also manufacture synthetic short hairpins, but all synthetics share a common problem: While HELA cells or HEK 293 cells should take them up fine, primary neuronal cultures, hematopoietic cells, or T-cells may require viral vectors. Some companies are developing viral libraries, but versions designed in the academic sector are already available.

Pros
• Well-validated
• Good knock-down efficiency (75%)
• Most reliable
• Easy protocol

Cons
• Expensive
• Doesn't infect some cell types
• Short-term knock-down

Library availability
* Ambion
Silencer siRNA libraries for mouse and human
For human: 7,800 genes and 9,100 genes
For mouse: 6,400 genes
3 or 4 siRNAs per gene (depending on library)
Numerous sublibraries available
Price range: $2,000-$100,000+

* Dharmacon (Provided by Thermo Fisher)
siARRAY and siARRAY RTF
For human: 21,500 genes
For mouse: 16,500 genes
4 siRNAs per gene
28 pre-designed libraries

* Invitrogen
Stealth RNAi essential druggable collection
For human: 2,400 genes (17 gene family panels)
For mouse: 1,300 genes (4 gene family panels)
3 siRNAs per gene
Price range: $5,000-$49,000 per panel

* Sigma Aldrich
Mission siRNA Library, mouse and human
6,600 genes (17 gene family panels)
3 siRNAs per gene
Price range: $5,000-$50,000 per panel

* Qiagen
FlexiPlate siRNA and HP siRNA
For mouse and human: 17,000 genes
For rat: 6,000 druggable genes
4 siRNAs per gene
Price per siRNA on a FlexiPlate
1 nmol: $50
0.1 nmol: $22

esiRNA

Frank Buchholz at the Max Planck Institute for Molecular Cell Biology in Dresden and colleagues designed an algorithm to predict the region on each gene with the highest density of efficient silencing triggers. They then amplified those regions and tagged them with phage promoters to make dsRNA. By digesting the long strands with RNAse III or recombinant DICER, he ended up with a pool of endoribonuclease-prepared siRNAs, or esiRNAs. The approach reduced off-target effects 13-fold, compared to single siRNAs.

Pros
• Reportedly more efficient than synthetic siRNAs

Cons
• Doesn't infect some cell types
• Short-term knock-down

Library availability
• PCR products are distributed by the German Resource Center for Genome Research (RZPD) (www.rzpd.de/products/esiRNA/)
• Buchholz's lab is currently making plans for commercial distribution of esiRNAs.
• Probe sequences are published on the RiDDLE database. (http://cluster-12.mpi-cbg.de/bioinformatics_group/RiDDLE.html)

Viral libraries produce long-term silencing and are much cheaper - when you need more reagents, you can make them yourself. Aleister Saunders, whose RNAi Resource Center at Drexel University was one of the first to purchase a retroviral shRNA library two years ago, says, "The cost per gene was enormously low compared to si's." The downside, he says, is that they are not yet as validated as synthetics.

The newer genome libraries use lentivirus, which unlike retrovirus, can infect non-dividing cells. (Working with lentivirus does require Biosafety Level 2 or 2+ facilities.) Nevertheless, says Hahn, "no one's done a real head-to-head comparison of the libraries," he says.

Functionally, the difference between them lies in whether you plan to focus on arrayed or pooled screens. Arrays allow researchers to do image-based analysis (for example, changes in spindle morphology that accompany knockdown of certain genes), and detection of small changes in phenotype. Pooled screens require less reagent and cost thousands of dollars less. But, says Sabatini, negative screens are difficult to do in a pooled screen, and "you have to screen for things that preserve the viability of the cell."

Hannon-Elledge Library

Developed by Stephen Elledge at Harvard Medical School and Greg Hannon at Cold Spring Harbor Laboratory, this library can be purchased in either a retroviral (pSM2) or a lentiviral (pGIPZ) vector. Library cassettes are designed to be easily transposable into different vectors.

Hannon and Elledge modeled their shRNA library after early endogenous micro (mi)RNA sequences. A 30-mer sequence encoding an RNA and the use of the Pol II rather than the Pol III promoter is the key to their vector. According to studies by Ellege's lab, shRNAmir's give more potent knockdown. That potency is crucial in pooled experiments, says Elledge.

Pros
• Barcoded for easy use in pooled screens
• Both lentiviral and retroviral vectors available
• MicroRNA context for greater potency and single-copy use

Cons
• Shallow coverage - for some genes the library has just one hairpin

Library availability
• pSM2 retroviral human: 85,000 constructs targeting 31,000 unique accessions
• pSM2 retroviral mouse: 65,900 constructs targeting 28,000 unique accessions
• GIPZ lentiviral human: 38,287 construct targeting 28,000 unique accessions
• GIPZ lentiviral mouse: 33,600 constructs targeting 21,000 unique accessions
• Distributed by Open Biosystems

The RNAi Consortium (TRC) Library

The TRC library, developed by a public-private consortium based at the Broad Institute in Cambridge, exists on a lentiviral backbone. It was designed primarily for arrayed experiments. In theory, the hairpin itself can be used as the barcode, and according to Sabatini, director of the Consortium effort, the group is working on techniques to make its use easier in pooled experiments.

The TRC focused on depth per gene - each one is covered by at least five different shRNAs. "We don't think you can have enough," says Sabatini. A large institution with extensive resources, the TRC also plans to validate every hairpin in the library. "The ideal library not only covers the genome with several forms of redundancy, but also knows how well it works," says Hahn.

Pros
• Deep coverage - 5 hits per gene

Cons
• No barcodes - less readily applicable for pooled screens
• Lentiviral vector only

Library availability
• TRC lentiviral human: 75,883 constructs targeting 18,200 unique accessions
• TRC lentiviral mouse: 65,800 constructs targeting 15,000 unique accessions
• Distributed by Open Biosystems and Sigma-Aldrich

The Netherlands Cancer Institute (NKI) Library

The NKI library was the first library in use. "Most of the papers that identify interacting genes were done with our library," says Rene Bernards, who heads the division of molecular carcinogenisis at the NKI and developed it with colleagues there. Like the Hannon-Elledge library, NKI also uses barcodes for pooled screens. But its coverage for human genes is about half that of the TRC and Hannon-Elledge libraries, and it has not been updated since it was made. "I don't change a wining team," Bernards says. "I have an H1 RNA vector that works." Several researchers noted past problems with recombination.ﾠ Bernards says the library has been recloned in recombination resistant vector.

Pros
• Longest track record
• Barcoded for easy use in pooled screens

Cons
• Older generation
• Retroviral only
• Smaller breadth of coverage

Availability
• For human: 24,000 constructs targeting 8,000 genes
• For mouse: 30,000 constructs targeting 15,000 genes
• Available through collaboration with the Netherlands Cancer Institute
• According to Bernards, commercial distribution by Geneservice, in Cambridge, UK, is forthcoming.