June 15, 2006 (Vol. 26, No. 12)

Overexpression and RNAi Screens Are Complementary Ways to Uncover Gene Functions

The sequencing of the genome for human and several other mammalian species has been completed, and the focus of genomics has now shifted to developing methods to elucidate gene functions on a genome scale. Cell-based assays using human cell lines are a convenient and ethical way to screen for gene functions in a high-throughput manner.

Typically performed in microplate format, the cells in each well are perturbed by modulating the expression level of a different gene. When cells report a change in phenotype, it is inferred that the modulated gene plays some role in the biochemical pathway that produces the phenotype.

For genome-scale screening, the phenotypes of greatest utility are those that can be quantitatively scored in an automated system. For example, a promoter of interest may be fused to a chemiluminescent reporter, such as luciferase, and transfected into the cells to be screened. For genes that interact directly or indirectly with this promoter, modulation of their expression levels in a screen may increase or decrease the intensity of luminescence radiating from the well, generating a hit. High-content screening systems automate image acquisition and analysis in order to measure a broad range of structural and biochemical changes in cells and enable multiple endpoints to be scored.

Expression levels for a given gene can be modulated up or down, by means of cDNA overexpression or RNAi knockdown, respectively (Figure 1). In cDNA overexpression, the expression level of each gene is boosted by transfection of a plasmid containing the cDNA under control of a mammalian promoter. In RNAi knockdown, the expression level of each gene is reduced by introducing into cells an RNAi “trigger” that specifically interferes with the expression of that gene.

Fig.1: Cell-based overexpression and knockdown screens

RNAi-based Knockdown Screens

RNAi-based screening can be accomplished using siRNA or vector-based shRNA triggers. Although siRNAs are easy to use and often produce good knockdown, their effects are transient and restricted by the rate of cell division, delivery to hard-to-transfect cell lines, and limited functionality in vivo. shRNA and microRNA-adapted shRNA (shRNAmir) constructs, expressed from viral vectors, are more versatile, allowing transient transfection, stable integration, and in vivo RNAi applications. shRNA and shRNAmir cloned into lentiviral or retroviral vectors also have the advantage of infection-based delivery into most cell lines, including hard-to-transfect cells like primary and nondividing cells.

Whole genome shRNAmir lentiviral libraries have recently been created using vectors that incorporate GFP to track shRNAmir expression and molecular barcodes for deconvoluting data from complex genetic screens. Expression Arrest lentiviral and retroviral shRNAmir libraries targeting the entire human and mouse genomes with multiple constructs per gene are available from Open Biosystems (www.openbiosystems.com). Assay-ready DNA and viral formats of these shRNAmir libraries are also available to facilitate RNAi screening efforts.

Multiplexed Screening Options

Multiplexed (pooled) screening has both a cost and labor savings advantage over arrayed formats. The new generation of shRNAmir includes a molecular barcode in each construct, so that the genotype of each cell can be determined. After transduction and stable growth in culture, cells can be subjected to selective pressure and the identity of the knocked-down gene established by sequencing or hybridization to a barcode microarray.

A recently published positive selection screen used a pool of 28K shRNA clones targeting approximately 9,000 human genes. Infected cells were observed for anchorage-independent growth and approximately 100 anchorage-independent colonies were pooled and analyzed using a barcode microarray chip to identify each shRNA construct. The same set of clones was also analyzed by sequencing to identify the shRNA constructs. Both approaches yielded similar results and identified approximately 25 unique candidate tumor suppressor genes (Figure 2).

Negative selection screens that do not rely on survival of cells, but use the molecular barcodes to monitor relative abundance of specific shRNAmir constructs in complex populations, are also made possible by the Expression Arrest shRNAmir libraries.

Positive selection RNAi screen used to identify novel tumor suppressor genes

Overexpression is the New Alternative

While conceptually simpler than RNAi knockdown, cDNA overexpression is very much the newcomer to genome-scale screening. One obvious advantage is that a cDNA only overexpresses the one transcript that it encodes; the sequence-based off-target effects that afflict sub-optimal RNAi designs are not possible. Moreover, levels of an overexpressed protein can be many times higher than normal, which has the potential to alter the phenotype dramatically.

This may expand the dynamic range of the readout, increasing the sensitivity of the screen. (On the other hand, high concentrations of an overexpressed protein might be toxic to the cell or may increase the probability of low-affinity binding interactions, stimulating off-target biochemical pathways.) However, cDNA overexpression screens are limited by the availability of fully sequenced, full-ORF-containing cDNAs cloned into a mammalian expression vector.

The gold standard Mammalian Gene Collection (MGC) contains fully sequenced cDNAs verified to contain a complete coding sequence. A subset of this collection has been cloned into the mammalian expression vector pCMV-SPORT6 and contains 6,415 human cDNAs and 9,624 mouse cDNAs. This subset of MGC cDNAs (including human and mouse) has been employed in highly fruitful cDNA overexpression screens with human cell lines. It is currently the largest collection of full-length, expression-ready cDNAs and is available from Open Biosystems as lyophilized DNA in 384-well screening plates (Assay-Ready MGC cDNA), as normalized aqueous solutions of DNA, or as glycerol stocks.

Until now, cDNA overexpression has lagged behind RNAi in the range of vectors employed for delivery and the use of pooled screening formats. However, there is nothing intrinsic to the technology that prevents the use of these enhancements. The content available for genome-scale overexpression screening has also lagged behind RNAi, though with the continued growth of the MGC, it will gradually catch up.

One or the Other or Both?

These two different approaches are already complementary in practice. Overexpression is now used to validate hits from RNAi screens, while RNAi is used to validate hits from overexpression screens. Will the same researcher be interested in performing both types of screens on a genome scale? We think so, because the infrastructure required to perform both types of screen is essentially the same. The possibilities and limitations of each approach have not all been tested, however, we can be fairly sure that the application of both approaches in parallel will yield the most complete understanding of the complexities of the genome.

Previous articleLew Kontnik
Next articleRichard A. Gibbs, Ph.D.