Researchers from the Genomics Institute of the Novartis Research Foundation have published one of the most comprehensive papers to date demonstrating how robotics and automation can be combined with cell-based assay technology to conduct rapid, large-scale compound library profiling early in the industrial drug-discovery continuum.
Specifically, the GNF scientists have developed a system they call the automated compound profiler, or ACP, which comprises automated cell culture, cell dispensing, incubation, reagent addition, imaging, and detection components — and used it to characterize a set of 1,400 kinase inhibitors in a panel of 35 tyrosine-kinase-dependent cell-based assays in a single experiment.
With much of the proof-of-principle work done, the researchers will now begin to adapt the system to screen other molecular libraries, such as antibodies and siRNAs; investigate pharmacological properties of candidate and existing drugs; and incorporate image-based high-content screening to study specific cellular pathways, GNF officials told CBA News this week.
"We've kind of turned the paradigm around and we're doing many assays — hundreds of assays — against a collection of molecules."
The development is significant because GNF straddles the border between industry and academia, and its work may provide a blueprint for the growing number of pharmaceutical and academic scientists wishing to scale up the throughput, automation, and multiplexing capabilities of cell-based drug screening.
"Recently, and especially at GNF over maybe the past five years, [researchers] have gone back to cell-based screening, doing one cell-based assay against millions of compounds in one run," Jeremy Caldwell, director of lead discovery at GNF and corresponding author on the paper, told CBA News. "That's where the state of the art is right now. You can do high-throughput cell-based screens one at a time, and maybe one or two a week, against millions of compounds.
"But now, we've kind of turned the paradigm around and we're doing many assays — hundreds of assays — against a collection of molecules," he added. "In terms of a precedent for this type of profiling, a lot of companies and CROs — like MDS Pharma, Upstate and Chemicon, and Invitrogen — have biochemical assay profiling capabilities, but those are all done using workstations, and they've come up with clever ways to reach economies of scale. They can do tens or thousands of compounds, but it's certainly not automated."
The ACP research, published in the Feb. 28 issue of Proceedings of the National Academy of Sciences, was done in collaboration with scientists from GNF neighbor Scripps Research Institute, though the bulk of the work was done at GNF, officials said.
The current ACP system contains three 486-plate or -flask incubators; a 1,536-well cell and reagent dispenser; control software running custom scheduling software; PerkinElmer ViewLux plate reader; Olympus inverted microscope with an automated stage and MetaMorph software from Molecular Devices; compound transfer station; tissue culture station; BD Biosciences fluorescence-activated cell sorter; and the "workhorse," a custom-built gripping robot arm.
The platform is reminiscent of the robotic high-throughput screening system marketed by San Diego-based drug-discovery firm Kalypsys (see CBA News, 7/13/2004), and for good reason: Kalypsys was spun out of GNF in 2001 and, according to GNF director of business development and licensing Avi Spier, the branded platform that Kalypsys sells is essentially the one developed and designed by GNF.
However, Spier stressed that the new ACP system is independent of the Kalypsys technology and has been developed by GNF in the past few years. In particular, he said, GNF re-worked the process of incorporating cell culture into a large-scale screen.
"In this system we pretty much reinvented how automated cell culture is done," Spier said. "We have actually created a new vessel that is harmonious with robotic manipulation and cell culture." That particular component of the ACP is currently available from Greiner Bio-One, Spier said.
"This is a very nice illustration of the power of a robotic platform enabling scientists to conduct industrial-scale screening."
"The process needed to be reinvented," he added. "Rather than trying to invent automation around how humans do tissue culture, we basically took an engineering approach to how tissue culture should be done in a robotic system."
As described in the PNAS paper, Caldwell and colleagues established a panel of 35 tyrosine-kinase-dependent cellular assays by creating stably expressed Tel-tyrosine kinase fusions representing each branch of the tyrosine kinase phylogeny.
Using the ACP system, the researchers then assayed a chemical library of 1,400 unique small molecules targeting tyrosine kinases against the cell lines in 1,536-well microtiter plates in a dose-dependent fashion. The cell-compound mixtures were then incubated overnight, and then assayed for cell viability using Promega's Cell Titer Glo luminescent cell viability reagent and the PerkinElmer ViewLux.
The ACP system enabled all of the experimental steps to essentially be carried out in complete walk-away fashion.
"This is a very nice illustration of the power of a robotic platform enabling scientists to conduct industrial-scale screening," Jim Inglese, director of biomolecular screening and profiling at the National Institutes of Health's Chemical Genomics Center, told CBA News. Typically, "these kinds of experiments are tedious and usually carried out around the world on bench tops, often by graduate students."
Inglese said that the NCGC, which has a GNF-Kalypsys automated screening system, has developed its own "novel screening paradigms" on it, and would likely be publishing these methods in the near future.
"This is the kind of work that we at the NCGC believe is going to propel biological research forward in the future, but we believe that it doesn't necessarily need to be in an industrial setting."
Although the research was intended to be proof of principle, the GNF scientists feel they have scratched the surface of how this type of automated compound profiling can further drug discovery in the near future.
In the particular case of kinases, the researchers wrote in their study that "automated kinase profiling expands [structure-activity relationships] from one target against a set of compounds to many targets against many compounds, thus providing a more comprehensive dataset." One of the most powerful consequences of this, according to Caldwell, is the ability to prioritize "hit" compounds much earlier in the drug-discovery process, immediately following high-throughput biochemical screens.
"Using the standard paradigm, what's often done is to take hits, and then test each one for its potency against the target, select the potent ones, and move those forward," Caldwell said. "But what's completely missing is biological specificity. You might have a compound that's incredibly potent on its target … but you have no idea if that compound might hit tens of other targets and be incredibly non-specific.
"So using this ACP, you can add a new dimension of information to make better choices early on as to which compounds to pursue," he added. "We like to call it biological intuition, as opposed to chemical intuition, which is so often superimposed during prioritization."
In addition, the GNF researchers believe this type of profiling can lead to benefits such as finding previously unknown effects of established drugs. Indeed, the kinase profiling experiment identified two new side activities for the Novartis leukemia drug Gleevec that are currently being investigated for asthma and gastrointestinal stromal disorder. Several other tested kinase inhibitors also showed previously uncharacterized activities.
From this point, the GNF researchers will begin to explore the many options beyond kinase screening for the ACP system.
"Now that we've done proof-of-concept, we can radiate outwards to all the things that we mentioned in the paper, from GPCR discovery to taking patient-derived tumor cells and finding out which compounds are most efficacious in vitro for those tumors, moving towards personalized medicine," Caldwell said. "We can also take libraries of natural products and begin to find out which herbal medicines can do what, and confirm all the anecdotal information."
In addition, GNF is currently evaluating high-content imaging systems that can be incorporated into the ACP for conducting pathway-based screens, protein translocation, and their ilk.
The combination of an inverted microscope and MetaMorph software that the team is currently using is strictly for measuring cell population density. "This can be adapted to doing [HCS], but probably not at the throughput that we require," Caldwell said. "This experiment was 1.5 million data points, and that's just the tip of the iceberg. We consider this to be a small-scale experiment.
"So we're really going to need to go with a more souped-up imaging system," he said. "We're evaluating a number of systems, and we have a few in house, such as the EIDAQ 100 from Q3DM, the [CompuCyte] iCyte, and the Evotec Opera, especially, which can keep pace with profiling capacity."
— Ben Butkus ([email protected])