Three-dimensional culture of human breast cancer cells, with DNA stained blue and a protein in the cell surface membrane stained green. The cancer in these cells is driven by the ERBB2 gene. Image: NIH
Three-dimensional culture of human breast cancer cells, with DNA stained blue and a protein in the cell surface membrane stained green. The cancer in these cells is driven by the ERBB2 gene. Image: NIH
Harvard Medical School scientists have developed an improved method for quantifying how sensitive cells are to cancer drugs. The approach works by zeroing in on an important characteristic that current methods do not take into account: the varying rates at which cells divide.
The research team, led by Peter Sorger, the Otto Krayer Professor of Systems Pharmacology at HMS and head of the Harvard Program in Therapeutic Science, published its findings May 2 in Nature Methods.
Multiple factors can influence cellular growth rate in the lab, ranging from the genetics of the cells to the number of cells in the experimental system to the type of medium in which the cells are grown. The new quantification method developed by the HMS team corrects for the influence of growth rate during experiments on the quantification of drug sensitivity and resistance.
“We believe that using growth rate metrics instead of traditional measurements will improve our ability to identify genes and biological processes responsible for drug sensitivity and resistance.” Peter Sorger
The effects of variation in growth rate, the authors argue, may explain why large-scale drug response data vary from one study to another, making results difficult to reproduce.
“We believe that using growth rate metrics instead of traditional measurements will improve our ability to identify genes and biological processes responsible for drug sensitivity and resistance,” Sorger said.
For precision medicine to fulfill its promise of matching treatments to patients, Sorger and his co-authors said, we must gather accurate information about drug sensitivity and resistance. Their research follows earlier work from the HMS Laboratory of Systems Pharmacology exploring how well drug response is captured in research experiments.
Typically, assays measure the concentration of drug that inhibits growth to the point that the number of cells in a treated sample grown over a period of time is only half the number of cells in an untreated sample grown in parallel; that concentration is known as IC50. In the simplest tests, cells are grown in a lab dish. More sophisticated approaches include growing cells in three-dimensional culture, introducing the cells into a mouse as a xenograft, or taking cells from a patient and introducing them into a mouse, a method known as a patient-derived xenograft.
“We wanted a better understanding of sensitivity or resistance at the molecular level, and to use that understanding to better design treatments, especially combination treatments. To do that, we had to quantify the growth rate.”—Marc Hafner
It was their interest in taking a closer look at predictors of drug response that led the scientists to study the growth rate of dividing cells, said Marc Hafner, HMS research fellow in systems biology and co-first author of the paper.
“We wanted a better understanding of sensitivity or resistance at the molecular level, and to use that understanding to better design treatments, especially combination treatments,” Hafner said. “To do that, we had to quantify the growth rate.”
IC50, which measures cell viability only at the end of an experiment, does not provide a reliable measure of a drug’s overall effectiveness if the number of cell divisions during an assay varies, as often is the case when comparing different cancer cell lines. For example, oncogenic mutations that promote cancer progression sometimes also slow down the rate of cell division, so in a typical drug-response assay a drug could appear to be less effective at inhibiting cell growth than it really is when compared to cells without this oncogenic mutation.
“There’s a fundamental problem with the way the IC50 is assessed: You cannot accurately compare cell lines if they grow at different rates,” said Mario Niepel, HMS instructor in systems biology and co-first author of the paper. “It is this intrinsic property of the cell lines, irrespective of variants in the experimental setup, that really plagues large data sets measuring drug sensitivity in hundreds of cell lines.”
The scientists tested their theory in computer simulation models and in lab experiments with existing cancer drugs. Their results evaluated drug response using growth rate metrics, derived by comparing growth rates in the presence and absence of drugs. Their analysis quantified the potency of a drug on a per-division basis, making sure that fast- or slow-growing cells with the same biochemical responses to drugs were scored the same way.
“That’s the beauty of it—there is hardly any experimental or computational change required for somebody to use these new metrics, and there’s even the possibility that groups could retroactively correct their data sets.”—Mario Niepel
The computational tool they used—available to the research community at www.grcalculator.org—can easily be incorporated into any users’ analytical pipeline, the authors said.
“That’s the beauty of it—there is hardly any experimental or computational change required for somebody to use these new metrics, and there’s even the possibility that groups could retroactively correct their data sets,” Niepel said.
More work needs to be done before their findings can be brought to the clinic, Sorger said.
“It’s a case where data drives a new theory and hopefully a new theory drives the collection of new and potentially more useful data,” he said.
This work was funded by grants U54-HL127365 and P50-GM107618, by the Giovanni Armenise-Harvard Foundation, and by a fellowship from the Swiss National Science Foundation (P300P3_147876).
