In the latest sign that toxicology is a hot area for gene expression analysis tools, CuraGen and Bayer announced last week that they have developed a microarray-based assay to screen compounds for liver toxicity in vitro. The assay is the first fruit of an 18-month research collaboration that goes beyond the toxicology arena. “What we intend to do is identify toxic compounds much earlier, so we spend less money on optimizing compounds,” said Werner Kroll, vice president of technology collaborations at Bayer. This assay, when fully validated, promises to lead to “substantial savings” in cost, he said.
The assay measures gene expression in primary rat hepatocytes after treating them with a pharmaceutical compound, and is able to distinguish between nine different kinds of liver toxicity, among them cholestasis, hypertrophy, and necrosis, according to CuraGen.
Calling Genes with Curagen
To develop the assay, scientists determined gene expression profiles in response to about 150 well-known toxic compounds, both in rats and in rat cell lines, using CuraGen’s GeneCalling technology, an “open” system for measuring differential gene expression. Unlike a microarray, it does not rely on pre-selected probes.
While Bayer performed the animal experiments and histopathology, CuraGen was responsible for the expression analysis. The two companies then correlated the in vivo and in vitro gene expression results with the pathology data and built a reference library, to which expression profiles of new compounds can now be compared. These compounds are then ranked according to their predicted liver toxicity, one of the main reasons for failure of lead candidates.
In order to perform the assay on a focused microarray, the scientist selected a number of marker genes. The array is a glass slide spotted with 1,000-2,000 features of 30mer oligos representing liver-specific rat genes. But it is the patent-pending expression profiles that are most important, not the array technology, which was chosen both for reasons of cost and amount of RNA required. “We are not married to [the array platform],” said Pascal Bouffard, the project leader of the toxicogenomics screening group at CuraGen. “If we find a better way to do the readout, we will use it.”
The next step will be to introduce the assay slowly into routine research. Ultimately, “we want to look into the toxic potential of these compounds directly after high throughput screening, when we still have the possibility to modify their structure during lead optimization, either to eliminate their toxic potential or to choose other compounds with similar efficacy but much less toxcicity,” said Kroll.
The assay will not replace any in vivo studies later on, but will help eliminate toxic compounds at an early stage. One of its advantages is that only a small amount of compound is required. “We are talking about less than a milligram, whereas in classical tox studies, you need hundreds of milligrams to put into a rat over time,” said Kroll.
This toxicity screening assay is only part of a larger collaboration between CuraGen and Bayer, the aim of which is to find and develop targets for diabetes and obesity, said Kroll, who has been working at CuraGen’s site in New Haven for the last 18 months. CuraGen said it had already started developing the assay for other organs such as heart, kidney, brain, and muscle. It is planning to offer the screen to its pharmaceutical partners, adding it to its portfolio of technologies as another negotiating tool, said Mark Vincent, CuraGen’s director of corporate communications and investor relations.
But will pharmas be interested? “Some of our pharma customers are building their own databases to help them do predictive tox screening,” said David Craford, vice president for marketing at Affymetrix. “This is nothing that new. Gene Logic has been doing this for over a year, and many of our customers have been doing this for well over a year or two.”
Pharmas certainly acknowledge that using arrays and cell lines is cheaper than doing classical toxicology in rats. But their use of arrays is not limited to focused arrays like the ones Curagen and Bayer have developed. An alternative strategy is to rank compounds by their selectivity — a highly selective compound will likely have fewer toxic side effects — while “also keeping an eye on genes thought to be known to be problematic with toxicity,” said Paul Kayne, a senior research investigator in the applied genomics department at Bristol-Myers Squibb. His company uses both their own and commercial arrays for toxicity studies, he said. Besides looking for selectivity, it also searches for markers of toxicity in various organs, he added, but generally in human systems.
Array makers agree that there is no single mode of doing array-based tox screens. John Jaskowiak, marketing manager for bioresearch solutions at Agilent Technologies, said that some of his customers use large-scale rat arrays for toxicology studies, so they don’t miss any effects, while others have requested custom arrays with a content of 1,000 genes or so for tox screens. On the whole, the percentage of arrays sold for tox studies is still small, he said, and Agilent does not offer a catalogue tox array at the moment. However, “I wouldn’t be surprised if within the next 12 to 18 months there would be a specific array,” said Jaskowiak. “The key thing is really knowing ’What exactly is the proper gene content to put down.’ I think that is the part that the industry as a whole is still trying to understand.”
Affymetrix agrees on this — it once sold a catalogue tox chip, but “what we found out by selling that product was that different people wanted different things on the array,” said Craford. Some of their customers use whole genome arrays for toxicology screens, he said, while others make use of Affy’s custom arrays, usually with a content of 100 to 1,000 toxicity-specific genes. “My suspicion is there is going to be a market for both,” said Jaskowiak. “Probably for the earlier stage of toxicology, there will be more broad arrays, but as you get ready for preclinical screens, you might use smaller subsets of genes.”