High-content screening gives researchers lots of cellular pathways to watch. But watching all of this content without proper data management tools can make the average viewer a little bleary-eyed.
The coming of the 21st century has revealed that we are truly in an information age. Internet and email access through our television sets and cell phones bombards us with high-volume information 24/7.
The pharmaceutical industry has also embraced the information age, adopting the highly information-dense “omics” technologies with their accompanying bioinformatic tools and other kinds of high-throughput technologies. One technology more recently adopted by is high-content screening (HCS). This technology is a high-throughput method of screening compound libraries against multiple signal transduction pathways in parallel and in live cells.
“If you think about years ago, people looked down microscope lenses at cells on slides and they made qualitative judgments,” says Ger Brophy, PhD, general manager of advanced systems, GE Healthcare Life Sciences in Uppsala, Sweden. “The advantage of HCS and HCA is that you can now do that with hugely improved throughput. You can put those cells in microtiter plates. You can analyze populations of those cells. You can digitize those images and turn them into quantitative data, which can be analyzed by analysis software.”
Hardware, software
GE Healthcare develops hardware and software tools for HCS. In terms of hardware, they have developed the IN Cell Analyzer, an automated, epifluorescent system for the analysis of cells on 96- and 384-well plates. The analyzer performs cell-based assays using fluorescent microscopy and bright-field microscopy coupled with the power of genetically-engineered or chemical fluors. According to Brophy, the fluors are subjected to lamp-based excitation, which gives the assay flexibility in terms of the type of fluor used.
In addition to hardware, GE Healthcare has also developed a number of software products to aid in HCS. One product, Investigator, visualizes and analyzes cellular and sub-cellular activities. Another, called Miner, allows for the data-basing and data-mining of large image data and is scheduled to roll out in March, 2008.
Too much data
Because it analyzes the responses of several cellular pathways to compound libraries, a single run of HCS produces an enormous amount of data. And much of these data might not be needed to interpret the experiment from which it was generated. But what happens to all of the remaining data? The answer: it is stored. And that presents a huge bioinformatics problem for HCS users and developers.
“We can certainly store the data. But in addition to storing the raw data, in order to get value from the technology we need to be able to retrieve and analyze that data. This requires processes and a robust infrastructure,” says Andrew Hill, PhD, principal research scientist I, Biological Technologies, Wyeth Research, Cambridge, Mass.
The approach Hill and his colleagues have used to overcome this problem has been to adapt core components used for some of their genomics technologies to their HCS operation. “By scripting SQL-level access to our relational databases that hold HCS data, and using things like our compute cluster to put more compute power behind our analysis for these bigger data sets, we have been able to pull out these relatively large data sets and do effective analysis,” says Hill.
Two classes of data are generated from their HCS experiments: images and statistics representing the biological activity in those images. On average, a typical experiment might generate between 1 million and 10 million data points. To crunch the numbers in experiments comparing control data to that of data from a compound-treated group, they might generate a Kolmogorov-Smirnov (KS) statistic or other multivariate statistics.
Assay development
Of course, before one gets to analyze HCS data, they must first develop an assay to generate it. One such assay is the phosphoflow platform developed by Stanford University scientists Gary Nolan and Peter Krutzik, which detects phosphorylation events in single cells by flow cytometry. This platform uses fluorescently-tagged antibodies that bind specifically to phosphorylated forms of proteins, in particular, kinases. In addition, antibodies against cell surface markers are used to simultaneously analyze signaling in multiple cellular subsets. The platform involves multidimensional flow cytometry, which can detect 10 to 15 different colors of fluorescence, and thus can potentially analyze 10 to 15 different pathways and/or cell types simultaneously.
Although this platform works on flow cytometry principles, it does not have the common limitation of flow cytometry because it not only detects molecules on the cell surface, but also within cells. “What always has been the case is that immunologists—as much as they might want to understand the molecular mechanisms of the disease—were pretty much stopped at the cell surface in so far as flow cytometry was concerned and had to take upon themselves standard biochemical tools to go inside cells, which usually meant bulk lysis approaches,” says Garry P. Nolan, PhD, associate professor, Baxter Laboratory in Genetic Pharmacology, Department of Microbiology and Immunology, Stanford University School of Medicine, Stanford, Calif. Such biochemical tools included Western blots, kinase assays, and pull-down assays.
A typical experiment starts out with live primary murine or human cells, usually ex vivo, though in vivo experiments are possible in mice. Signaling, induced by a stimulus such as a cytokine, results in phosphorylation events. The induced signaling pathways must be frozen in time so that they can be analyzed later, therefore, the cells are fixed with a fixative such as formaldehyde. Cells are then permeablized and antibodies are added to react with the phosphorylated proteins and cell surface markers of interest.
Nolan and his associate researcher Peter Krutzik, PhD, have applied the phosphoflow platform all the way from initial compound library screening to pre-clinical work in mice. The platform offered them the opportunity to analyze protein phosphorylation events in a number of signaling pathways, but also across a number of different cell types assayed in a highly parallel and high-throughput manner. “That really let us see the drug activity in a whole different light, it allowed us to see not only the effect on signaling pathways, but also the specificity that a compound might have for a particular pathway and the specificity or selectivity that a compound has for a particular cell type,” says Krutzik.
“But you don’t have to do just kinase inhibitors. You can do any kind of inhibitors. You’re acting against the pathway, but you are reading out the output at the phospho-signaling event,” says Nolan. “For us, [the main goal of the project] is to develop a tool to develop inhibitors.”
According to Krutzik, Amgen is one of the main users of the PhosphoFlow technology. Other pharma companies also use it. “There are people using it not necessarily at the drug screening stage but at the lead validation and hit-to-lead time point,” says Krutzik.
Another HCS assay used to study cell signaling is Invitrogen’s in-house reporter gene technology, CellSensor. The reporter used here, beta-lactamase, is under the transcriptional control of pathway-specific DNA promoter elements. Although this assay can be used for single pathways, Invitrogen’s Qtracker has been used to multiplex the assay for HCS.
The reporter assay is quantitative and is described as follows. Expression of the beta-lactamase enzyme is induced using different stimuli. Its proprietary substrate is a fluorescein molecule linked to a cumarin molecule by a beta-lactam ring. Intact substrate undergoes FRET yielding green fluorescence, whereas cleaved substrate will, upon excitation, produce blue fluorescence. “The blue-to-green ratio gives you a measure of to what degree your pathway is activated,” says Thomas Machleidt, PhD, group leader, Emerging Targets, Discovery Sciences, Invitrogen, Madison, Wis.
Machleidt has successfully used Qtracker to optically tag individual CellSensor cell lines, which allowed him to assay up to three independent cell signaling pathways simultaneously in an image-based format. And “there is really is no reason why you should not be able to extend that to use more complex mixtures. So that means instead of having to run three plates, you can run the entire experiment in one plate.” Some of the pathways investigated using this assay include NF-kB and AP-1 (both of which are good general readouts for growth factor-activated pathways) and Jak-Stat signaling.
In summary, HCS is a powerful tool for screening investigative drug compounds against cell signaling pathways in live cells in a high-throughput and highly-parallel manner. But as with all high-throughput technologies, the enormous amount of data produced challenges many of its users to search for data-management solutions. Although some companies have managed to develop such solutions and others have utilized solutions developed by vendors, the data-management challenge of HCS remains unresolved. And finding a solution to this problem is on the HCS users’ wish-list.
This article was published in Drug Discovery & Development magazine: Vol. 11, No. 3, March, 2008, pp. 18-22.
Filed Under: Drug Discovery
