Researchers at Novartis recently published a study using a lentiviral based library containing many millions of unique sequences (barcodes) to label and track erlotinib-resistant non-small lung cancer cells. The approach provided a way to differentiate whether these resistant cells were already present in the initial population or arose anew during the drug treatment.
The researchers developed a library—called ClonTracer—that is very similar to Cellecta's CellTracker Barcode Library, also mentioned in a previous blog post. Non-small lung cancer cells were labeled with ClonTracer sequences starting with inoculating several million cells with approximately 1 million viral particles (ensuring a low MOI) so that, by and large, only one barcode was picked up by each of the cells. Selection of transduced cells, then, left about 1 million cells—each with a different barcode integrated in its genomic DNA. This large population of labeled cancer cells enabled the authors to test whether resistance to cancer drugs arises in a population during treatment, or whether a small number of resistant cells are already present in a population before treatment.
To tease out which of these two possibilities drives the development of resistance, the initial population of uniquely labeled cells were carefully expanded to make several starting populations each having similar representation of about 20 cells with each barcode. Eight of these similar starting populations were then treated with the drug erlotinib. The idea was that, if the population of resistant cells was already present, then cells with the same barcodes should survive the drug selection in each culture. On the other hand, if the resistance arises as a result of mutations occurring during exposure to the drug, then any of the originally plated cells could become resistant, so the barcodes of the resistant cells would be randomized between replicates.
After isolating genomic DNA and running NGS, it turned out that 462 barcodes were enriched in the selected populations. Further, 40% of the barcodes were shared across all the replicates, and 90% were shared in two or more. Clearly, the selection was not random. The original population contained a small number of cells already resistant to erlotinib—in fact about 0.05%. Subsequent work showed that most of these cells could be eliminated with crizotinib, so they are cMet dependent.
The above study demonstrates the utility of using highly complex barcode libraries to identify and understand how a few cells or rare sub-populations can influence and affect larger heterogeneous populations. Recent publications substantiate the application of clonal cell tracking with barcode sequences to assess cell culture heterogeneity and identify and track select clonal populations to better investigate proliferation, growth dynamics, and cell differentiation.
25 million barcodes were sequenced in a NGS run 45 million reads deep. 40% of these reads (10 million) were unique, indicating the library contains more than 50 million unique barcodes. Further, only 20 barcodes had more than 30 reads and the most prevalent barcode appeared 46 times. With 45 million reads, this is equivalent to 1-in-a-million representation. Thus, almost every clone will be unique in a random selection of 1 million clones from the library, so it can be used to uniquely label about 1 million cells.