To pursue his interest in understanding the biological basis of genomic instability and DNA damage repair systems, Alex Bishop, D.Phil., assistant professor, cellular and structural biology, at the University of Texas Health Science Center, is using RNAi screening to probe cell survival and response mechanisms following exposure to DNA-damaging agents.
A cell’s DNA-damage repair system coordinates a multifaceted, pleiotropic response that involves multiple pathways affecting, for example, cell-cycle regulation, cell growth, DNA repair, transcription and translation control, and proteasome degradation. The result may be a controlled shut-down of cell function leading to apoptosis, progression down a path of DNA repair and the return to a growth state, cell senescence, or some intermediate response.
Because so many cellular processes are involved, “this is a perfect situation to look at the entire genome,” says Dr. Bishop. And RNAi screening in Drosophila offered a means to do just that. By exposing Drosophila cells to DNA-damaging agents and introducing RNAi reagents to achieve targeted gene silencing across the genome, Dr. Bishop is hoping to identify genes—and the pathways in which they function—that are activated when cells carry out DNA repair.
This work has a direct translational component, as many existing chemotherapeutic drugs used to treat cancer exert their effects by causing DNA damage in dividing cells, impeding their ability to replicate. How these drugs work and how cells develop resistance to them is not well understood, and Dr. Bishop believes that these types of RNAi screening studies will help shed light on the critical mechanisms involved in these processes.
Dr. Bishop has had to overcome several challenges in applying RNAi assay technology for whole-genome analysis in Drosophila. The first is the technology’s intrinsic propensity for off-target effects. About 40% of the data is “real” and readily validated, he notes, whereas the remaining 60% may be junk or at least of questionable utility.
This same issue plagues mammalian RNAi screening. Another stumbling block has involved translating the information obtained in Drosophila to humans, as most of the bioinformatic tools available for genomic analysis were developed for human data and most existing orthology mapping tools were inadequate.
Dr. Bishop has adapted the RNAi assay system he is using to a robotics system to allow for high-throughput screening. To enable better comparison of data between microtiter plates and to help normalize the data and reduce plate-to-plate variation as well as plate location effects, Dr. Bishop’s group loads samples onto the plates in a distributed pattern, interspersing siRNA across the plate and utilizing 10% of the plate’s real estate for controls. This strategy has increased the stringency and sensitivity of the screen.
In addition to mining the screening data to look for individual genes linked to DNA repair and cell survival, another component of Dr. Bishop’s data-validation strategy is to do pathway analysis. The ability to link multiple genes in a pathway to a cellular activity of interest increases one’s confidence in the strength of that association, Dr. Bishop explains.
He is also probing the protein interactome to support data validation and minimize the risk of false-negative results. By combining three databases that cover five species, he is able to compare interactions across species and to configure protein interactions into pathways. This has helped to simplify the interactome, making it a more useful tool for data mining and intuitive gene discovery.
Thermo Fisher Scientific conducts in-house RNAi-based screens as a service and to assess best practices, evaluating screening-design strategies and hit-stratification methods.
James Goldmeyer, Ph.D., a field application scientist at Thermo Scientific Genomics, a Thermo company, describes a study designed to evaluate different screening strategies, incorporate siRNA reagents that minimize off-target and secondary knockdown effects, avoid the selection of promiscuous molecules, and optimize data interpretation and validation. “I don’t think we will ever completely eliminate the off-target events, but there are ways to mitigate them and follow-up strategies that allow for easier interpretation of the data,” says Dr. Goldmeyer.
In one screen, Thermo scientists utilized a ubiquitin-EGFP assay designed to identify druggable genes required for proteasome function. They selected a proteasome screen because it is well studied and well understood, relevant gene ontology tools are available for performing whole-genome searches, and the assay design is amenable to multiplexing.
The question they wished to examine is the following: is it better to “start out with a broad assay design and then qualify candidate hits with a more specific assay (such as cell death followed by pathway analysis), or can you start out with a specific assay and then drill down even further (for example, a cell-based assay followed by target-specific assays)?” They tested two different assay strategies and two siRNA design approaches.
The assay strategies included a common cell-viability assay and a pathway-specific GFP fusion assay that directly targets proteasome degradation. The results indicated that “the more specific the assay, the more specific the data,” Dr. Goldmeyer says. Both approaches yielded “high-confidence candidate hits”; the screeners’ design choice, he concludes, will depend on factors such as time, reagent and consumable costs, and how many false positives they are willing to accept to minimize the risk of missing a false-negative result.
One common practice is to develop an assay using a pooled reagent for the primary screen followed by deconvolution of the pool and rescreening of candidate hits using individual constructs. The screen described by Dr. Goldmeyer incorporated two genomic siRNA libraries, one of which was chemically modified for enhanced specificity. The screens were run in parallel as gene-specific pools and the individual constructs that made up these pools.
The first approach—screening with pools and then with individual siRNAs—yields more data points per gene, and the number of individual construct hits can be rank-ordered based on increasing confidence. The alternative approach offers higher throughput and generates a smaller number of yes/no data points; multiple positive results indicate higher confidence.
The outcomes for various combinations of reagents yielded a hit list with overlapping high-confidence candidates. Cost/benefit considerations and whether users are more comfortable applying bioinformatics tools to a large dataset or working with less data in which they have greater confidence regarding their validity will drive screening design decisions. The specificity of the assay and how much noise is in the system will also be contributing factors.