This article has been updated from a previous version, which mistakenly reported that Klaus Melchers presented information on Altana's research instead of Michael Byrne.
SAN FRANCISCO – Pharmaceutical firms are finding cell-based RNAi screens to be a valuable drug-discovery tool, but are finding that the approach still has a number of limitations that need to be taken into account, according to speakers at Cambridge Healthtech Institute’s Molecular Medicine Tri-Conference, held here this week.
Several speakers at,the conference noted a number of advantages of cell-based RNAi screens over small-molecule screening for target identification. The reasons include the fact that most cell types are amenable to RNAi, it’s relatively easy to knock down any gene of interest, and the resulting data is extremely informative.
The approach still has some limitations, however, most notably RNAi’s well-known tendency to produce off-target effects. In addition, several speakers discussed challenges related to the cells themselves.
For example, John Reidhaar-Olson, research leader in the department of research informatics for genetics and genomics at F. Hoffman-La Roche, noted that cell-based RNAi assays are particularly prone to edge effects because the cells in the outer wells of the plates grow at a different rate than the cells in the inner wells. As a result, his group usually ignores the outer wells in most studies, he said.
Steven Haney, group leader in oncology genomics in the department of biological technologies at Wyeth Research, said that his lab has had problems with the “penetrance” of some RNAi screens, in which the level of GFP in the cells is heterogeneous, making it difficult to interpret.
After further study, Haney said his team found that the expression levels of several proteins varied significantly within cells grown in culture, meaning that the problem is not heterogeneity of the siRNA knockdown, but heterogeneity of protein expression.
The issue is an “artifact of the cell culture itself,” Haney said, perhaps due to the fact that cells proliferate faster in culture than they would under normal biological conditions. While low penetrance can ultimately “reduce the impact of the data” from an assay, Haney said that his group has found they can improve their analysis of the data by looking at single cells rather than entire wells.
Michael Byrne, director of biochemistry at Altana Pharma, discussed another challenge related to cell lines. In this case, Altana was using the LAD 2 cell line in an RNAi screen to identify kinases related to mast cell degranulation. The problem, he said, is that LAD 2 is “a difficult cell line to work with” and “notoriously difficult to transfect.”
While the Altana team was able to account for these challenges in its assay, siRNA transfection caused a very strong interferon stress response in the cells – an unexpected side effect, Byrne said.
While the researchers first feared that this effect might be a “showstopper,” Melchers said that after further study, they found that the effects of degranulation could be “uncoupled” from the stress response, permitting them to proceed with the assay. Ultimately, he said, they were able to identify 10 novel kinase targets “that we wouldn’t have found any other way.”
Growing Awareness of QA/QC
Several speakers noted that they have begun to perform statistical analysis on their raw screening data prior to further analysis in order to account for experimental bias or flaws as soon as possible.
This appears to be an area of growing awareness in the field, but Dmitry Samarsky, director of technology development at Thermo Fisher subsidiary Dharmacon, discussed preliminary results from a project that may focus a bit more attention on the issue.
The issue of low penetrance is an “artifact of the cell culture itself,” Haney said, perhaps due to the fact that cells proliferate faster in culture than they would under normal biological conditions.
Samarsky outlined the initial findings of a project coordinated by the RNAi Global Initiative ─ an alliance of Dharmacon and several international non-profit biomedical research centers ─ to establish RNAi screening standards. Under the auspices of the initiative, 10 labs used Dharmacon’s siArray human genome siRNA library to conduct the same experiments on the same cell lines with the same protocols in a multi-site comparative RNAi screen.
Samarsky said that there was a surprising amount of variability between labs, between screens, and even within screens ─ variability that must be carefully accounted for when running such experiments, he said.
After running a battery of statistical analyses on all the data from the 10 different labs, Samarsky said that the primary sources of variability were poor pipette delivery, miscalibrated robotics, procedural deviations (such as wash stringency and incubation times), differences in reagent concentrations, and differences in cell growth patterns due to clumping.
After performing meta-analysis across all 10 screens, Samarsky said the researchers subjected a short list of hits to secondary screening, and were only able to confirm five of the top 10 hits ─ a much lower than expected success rate. Samarsky said that his team is still trying to determine the source of this problem, but noted that the results only confirm the importance of standards for the field.
One outcome of the study is the proposed MIARE (Minimal Information About RNAi Experiments) standard [CBA News 01-05-07], which Samarsky said should help account for variability across experiments.
Samarsky said that the RNAi Global Initiative currently has a paper in press at Nature Biotechnology outlining the study, as well as the proposed MIARE standard.
Samarsky told CBA News after his talk that the variability seen in these experiments is not a reflection on the quality of individual labs, but an inherent feature of RNAi-based screens. Therefore, he said, better metrics are required to help researchers ensure that they are getting accurate results.