NEW YORK – Researchers based in Spain and Germany have created CloneTracer, a computational method that infers single-nucleotide variants, mitochondrial SNVs, and copy number variants from single-cell RNA-seq data to provide information on cell lineage in acute myeloid leukemia (AML) samples.
The scientists, led by Lars Velten from the Center for Genomic Regulation in Barcelona and Carsten Müller-Tidow of the Molecular Medicine Partnership Unit of the European Molecular Biology Laboratory (EMBL) and the University of Heidelberg, posted a preprint describing CloneTracer to BioRxiv last month.
Their study used 10x Genomics single-cell gene expression data to analyze differentiation in samples from 19 AML patients. Velten and his team believe they've been able to identify rare cancer stem cells, which are hard to pick out because they often resemble healthy stem cells. Moreover, they've identified some important differences between those two kinds of cells and were able to predict response to first-line chemotherapy.
Overall, the study looked at more than 88,000 cells from 25 samples, across two cohorts, for a cost of less than $.20 per cell.
The new method is one of several attempts in the field of AML research to employ so-called "natural barcoding" systems to trace the origins of cancerous cells. "I don't think they've provided something revolutionary," said Asiri Ediriwickrema, a fellow in hematology at Stanford Medicine and an AML researcher. "But I think they've provided something that is a substantial step forward in this overall goal."
CloneTracer is able to integrate different types of mutations through a single protocol. "The 10x kit usually gives you very sparse coverage of the gene body," Velten said, instead offering more coverage toward either the 3' or 5' end. "If you're interested in any particular nuclear mutation, such as a point mutation, you would not have it covered very well."
The researchers modified the 10x library preparation protocol to amplify sites of interest based on targeted perturb-seq (TAP-seq), a targeted single-cell sample prep method Velten published in 2020. To optimize the sample prep, they take an aliquot of the cDNA library created by the 10x library preparation and amplify particular targets using PCR, including nuclear and mitochondrial genes of interest.
"The real advance is that we have the software and the statistical model to properly analyze the data," Velten said. "Imagine you have a gene that is decently expressed and a mutation in that gene." Drop-offs, lack of allelic coverage, and lowly expressed genes can lead to a high false negative rate. "This makes any type of even qualitative analysis very difficult. To quantitatively do it, just taking these data at face value is not possible. We needed a way to account for the statistical uncertainty and make use of any marker mutation we can get," he said. To solve this problem, they used a Bayesian statistical model to infer relationships between mutations and used that to assign cells to a lineage.
A charm of CloneTracer is that it can also be applied to existing datasets. "But the confidence you get, or the clonal resolution you get, is lower," Velten said.
The method isn't without its limitations. Gene dropout remains high. "We're limited by how much these droplet-based assays capture," Velten said.
Ediriwickrema noted that he wished the authors would have included comparisons of their findings to similar, but orthogonal, observations made with other clonal lineage tracing pipelines. Similar work from Harvard University Stem Cell Institute researcher Peter van Galen, which also uses mtDNA, is an "obvious comparator," he said.
He would also ultimately like to see functional validation studies of their findings. "It would be nice if they went back to their samples and sorted them using various genes or surface markers, and studied them in vitro or in vivo," he said.
Velten said comparisons with other methods is "something that we already have first data on and that we will most likely include in a revised manuscript," adding that functional studies will also likely be addressed.
Study size is another area in which Velten wants to improve. He'd like to use the method on larger AML cohorts.
Another advantage is that CloneTracer could theoretically be used on any cancer lineage. It's simply more illuminating in AML, he said, because those cancers often have lower numbers of mutations. For future studies, though, Velten would like to stick with AML.
"But in parallel, I think we should start to look into developing 'simpler' assays," such as fluorescence-activated cell sorting, he said.