NEW YORK (GenomeWeb) – A team led by researchers at the Broad Institute has described a new molecular barcoding method that it hopes can help accelerate drug discovery by enabling screening of drug candidates against hundreds of cell lines pooled togetherin vitro or potentially using in vivo mouse xenografts.
The group described the method, called PRISM, in a letter in Nature Biotechnology today, sharing data from a variety of experiments assessing its feasibility and validity compared to traditional single cell line results.
Todd Golub, the study's senior author and director of the Broad Institute's cancer program, told GenomeWeb that molecular barcoding as a tool to deconvolute mixed targets is not new, but this type of technology hasn't previously been used to support highly multiplexed testing of drugs.
Golub noted that a number of researchers have shown that using cancer cell lines to study relationships between drug sensitivity and molecular biomarkers is a "very powerful" approach.
"The challenging thing is that it's now clear that to have real confidence in cell line studies you have to profile a large number of lines otherwise you can get confused by the genetic diversity of cancers," Golub said.
"Relying on your favorite one or two or three cell line models to understand drug sensitivity or resistance doesn't suffice," he added. You really need to look at hundreds to develop real confidence in knowing where a drug may or may not work."
According to Golub, while some large pharmaceutical companies have been able to automate and implement factory-style efforts that test compounds against many different cell lines in parallel, the expense and time requirements for this make it impossible for groups with fewer resources.
"It's arduous work, to take out vials of hundreds of cells and grow them up uniformly every time you have a new drug to test," he said, "and that has limited the practical utility of the large panels of cancer cell lines that have been developed."
In developing PRISM, Golub and colleagues hoped to create a way to be able to test drugs on a mixture of different cell lines in a single go. Before settling on a molecular barcoding strategy, the group considered several other ways to enable this, one of which was using the unique genomic features of each cell line to distinguish different lines from each other.
"That turns out to be pretty tricky in practice, to have a highly multiplexed amplification of different loci across the genome that are informative and uniquely identify a particular cell line," he said.
Instead, the team decided to use a molecular barcoding system supported by Luminex's FlexMap detector.
Briefly, the PRISM approach involved integrating 24-base-pair DNA barcodes into individual tumor cell lines. These cell lines are then individually frozen and later thawed to generate mixtures of equal numbers, frozen again, thawed, and arrayed into tissue culture assay plates.
These plated mixtures can then be treated with test compounds or controls. At the conclusion of treatment, a PRISM user harvests DNA from the mixture of remaining viable cells, amplifies the barcode sequences using universal PCR primers, and hybridizes the amplified sequences to microbeads for fluorescence-based quantification using the Luminex FlexMap detector.
According to the group's protocol the signal for each barcoded cell line can be scaled to that of a control to adjust for different barcoding efficiencies and cell doubling.
Golub told GenomeWeb that he and his coauthors were careful to perform extensive technical validation of the approach. "Any new method has a series of tradeoffs, in terms of performance and cost and speed and accuracy, [so] we wanted to really have a deep understanding of the method."
In the group's validation experiments PRISM was able to sensitively detect differences in cell number in mixtures with varying levels of a single cell line against an invariant background of four others. According to the authors, the assay could detect as few as 10 cells in a mixture of 4,000.
When the group pooled 25 lung adenocarcinoma cell lines with different known molecular features using PRISM, and tested compounds known to have genotype-specific cell killing patterns, they saw the expected patterns borne out in the mixed-cell PRISM pool. For example, treatment with puromycin resulted in uniform cell death across the pool, while treatment with the EGFR inhibitor erlotinib resulted in dose-dependent killing of the four EGFR-mutant cell lines in the pool.
Similarly, treatment with the ALK kinase inhibitor NVP-TAE-684 resulted in death of one cell line harboring an EML4-ALK translocation, but not another with a different translocation where the group expected intrinsic resistance.
Testing the method using a panel of 100 barcoded cell lines divided into four pools of 25 lines each, the group was able to compare results using PRISM to sensitivity data measured in conventional unpooled assays for 23 different compounds. Results from the pooled PRISM testing matched closely the expected patterns of cell killing seen in the conventional unpooled assays, the authors wrote.
According to Golub, the team's early results suggest a number of promising applications for PRISM. One area the method could be useful is in supporting more comprehensive characterization of the activity of existing drugs.
"You could have an advanced small molecule of interest, maybe in clinical or preclinical development, or even an FDA-approved drug for another indication," Golub said, "And you want to characterize it across the breadth of cancer."
Alternately, PRISM could be useful as a part of the process of optimizing a candidate drug, as part of improving half life, potency, or other pharmacologic properties, without "wandering off target," Golub said.
"This is a particular challenge in oncology because when the readout is killing cells, there are a lot of ways to do that … so you can easily get off target," he explained. "With PRISM, you can easily look across a panel of hundreds of cell lines where you have an expected pattern of killing and non-killing [that is] correlated with a particular mutation of interest …. So as you continually test each analogue of the compound … if you see the same patterns, you [know] that you are still on target."
In addition to evaluating the method's feasibility and its validity compared to testing of individual cell lines, the group collected data that hint at some other ways PRISM may enhance or support the development and testing of drugs in wider practice.
For example, the researchers performed a screen of 102 cell lines across 8,400 compounds, which led to the identification of one, BRD-7880, as a potent and highly specific inhibitor of aurora kinases.
This type of broad screening is an area where Golub said he and colleagues at the Broad are interested in applying PRISM themselves. Another application the investigators demoed in their study is the use of PRISM to pool cell lines not only in vitro, but also in vivo, by implanting the pooled mixes of cell lines into mice.
In their experiments with mouse xenografts, the Broad team injected a pool of 24 barcoded lung adenocarcinoma cell lines subcutaneously into ten recipient mice. Euthanizing these mice several weeks later, the researchers then resected the implanted mixed-cell tumors and applied PRISM, which was able to detect three out of the 24 initial cell lines.
While the 23 lines appeared to grow at different rates from mouse to mouse, their relative abundances within the tumors were nearly identical across the ten xenografts, the authors reported.
"We've done far less characterization of the method in this setting," Golub said, "But in the preliminary work, the data are encouraging."
Golub said that PRISM will need to be used in larger studies to bear out its ultimate utility. While the investigators developed it primarily for their own use, he said, they hope publishing it will catalyze interest in both academia and industry.
The team is already engaged in some industry collaborations using the method, according to Golub.
He and his colleague Channing Yu, the study's first author, have also applied for a patent concerning the PRISM approach. Right now, though, they have no plans to commercialize PRISM as a routine service.