NEW YORK (GenomeWeb) – An international team has developed a sequencing pipeline for finding transposon mutagenesis insertion sites in an unbiased manner in individual cells or bulk tumor tissue — capabilities intended to help in uncovering early-stage tumor drivers and in addressing tumor heterogeneity.
Researchers from the US, Singapore, New Zealand, and the UK combined random genomic DNA shearing with liquid phase capture to pull down insertion sites for subsequent sequencing and bioinformatic analyses in a sequencing pipeline called "Sleeping Beauty capture hybridization sequencing," or SBCapSeq, which has reduced reliance on PCR-based amplification.
The SBCapSeq approach makes it possible to "drill down to understand the heterogeneity of tumor clones and transposon insertions in a given tumor," co-lead author Karen Mann, a cancer researcher affiliated with the Houston Methodist Research Institute and Singapore's Institute of Molecular and Cell Biology, told GenomeWeb.
"We can sequence more of the events because we don't have any limitations based on the PCR," she explained, "and we can use sequencing read depth as a proxy for clonality, which again is due to the unbiased nature [of SBCapSeq]."
In a study published last week in Nature Biotechnology, for example, she and her colleagues demonstrated that they could use SBCapSeq with the Ion Torrent Proton sequencing platform on both bulk tissue samples or individual cells from a mouse model of myeloid leukemia. In the process, they identified insertions in genes already implicated in acute myeloid leukemia (AML) in humans, while getting clues to early mutation events that have been selected for individual tumor sub-clones.
"We basically could show that we have very complex tumors that are driven by a transposon and that we could look at different tumor cell populations, or sub-clones," Mann said. "We could even go at the single-cell level, which is unprecedented with [Sleeping Beauty] transposon analysis."
Transposon insertion sequencing has traditionally relied on DNA fragmentation and amplification methods that can skew insertion sites detected, Mann explained. In DNA chopped up with restriction enzymes, for example, some transposon insertion sites might be missed if they are too far from restriction sites. Likewise, extensive PCR-based amplification can swamp out rare or low frequency insertions and may be biased towards certain sequence contexts.
Such methods "basically give you a very good global view of insertions," she said, but do not provide complete profiles of the insertion-mediated mutations present across a given tumor genome.
"Most simply put, we were missing [transposon insertion] events that were occurring," Mann said. "We found by comparing our new method to our old method that we were missing major events — not just minor, one-off insertions."
In contrast, SBCapSeq relies on enzymatic or sonication-based shearing to cut genomic DNA more randomly before grabbing transposon insertions by liquid hybridization capture, regardless of the placement of these insertions.
"We're basically pulling down any piece of DNA, using our probes, that have a transposon repeat at the end," Mann explained. "So, there are no assumptions made of where those should happen and how many we should get."
To avoid capturing transposons that haven't been inserted into the genome, the team uses capture probes targeting inverted repeat sequences in the transposon, as well as blocking probes designed to catch non-inserted transposons.
Though some PCR amplification is still required in the SBCapSeq pipeline, the DNA fragmentation and liquid capture steps enrich for the Sleeping Beauty insertion sites in a non-biased manner and decrease the number of PCR rounds needed, making it possible to get away with just eight to 10 PCR cycles instead of multiple PCR rounds with 20 or more cycles apiece.
Reads containing sequences from the Sleeping Beauty transposon and the mouse sequences from the insertion site are then mapped to the mouse genome with clues from the quality scores associated with the reads, the team explained. The capture step also helped in dialing down the amount of input DNA needed so that insertion sequencing could be applied to smaller tissue samples or even to individual cells.
The researchers demonstrated that SBSeqCap yields could identify more Sleeping Beauty insertion sites than a transposon insertion sequencing method based on a method called SB splinkerette PCR, apparently providing a less-skewed representation of these insertions.
For their proof-of-principle study, they also applied SBCapSeq to a mouse model of myeloid leukemia, produced using an inducible form of the Sleeping Beauty transposases and a beta-actin linked Cre transgene.
In bulk spleen tissue from 168 mice with myeloid leukemia, the team uncovered nearly 28,000 distinct transposon insertion sites affecting 466 genes suspected to contribute to cancer development.
Of those, three-dozen had human homologs in the Cancer Gene Census database in CSMIC, the researchers reported, along with more than 100 genes mutated in sequencing studies of AML in humans.
They subsequently tapped read-depth data to narrow in on mutations in 35 genes that appear to represent potential cancer drivers. Eighty percent of those genes were related to MAP kinase or JAK-STAT signaling.
Based on the insertion patterns present in 26 individual cells from mouse tumors, meanwhile, the team tracked down more than 3,800 insertions and nearly 1,600 genes impacted by at least one of these insertion events — a set that included new candidate cancer contributors not detected in the bulk tumor samples.
By folding in RNA sequencing data from a myeloid leukemia-affected mouse spleen sample, the researchers got transcriptional hints that helped to tease apart the driver and passenger mutations spurred on by Sleeping Beauty mutagenesis. Again, the expression analysis showed overlap with AML, Mann said, suggesting the mouse mutagenesis model is relevant to human disease.
While SBSeqCap should be compatible with a variety of sequencing platforms, she explained, the Ion Torrent platform offered an advantage in that it requires very little input DNA and produces relatively clean SBSeqCap profiles.
Mann noted that it is important to account for background events by sequencing normal samples alongside the tumor — particularly when looking across different tumor models — since some Sleeping Beauty transposons produce more extensive genome insertions than others.
The general SBCapSeq is expected to be applicable to other transposon mutagenesis experiments, including those done with PiggyBac transposons, by tweaking the capture and blocking probes used.
The team is looking at ways of doing more extensive multiplexing in the SBCapSeq pipeline, which would bring down the cost of the approach, and is making improvements related to transposon sequence capture efficiency.
The researchers are also using the SBCapSeq approach to assess bulk tumor samples and individual cells from squamous cell carcinoma, melanoma, and pancreatic cancer models. The latter, single-cell analyses may provide insights into mutations that work together to drive cancer development cooperatively, Mann explained, which is relevant to future targeted treatment design.
"Oftentimes a gene may be very important for the disease," she said, "but there's no targeted therapy available or it's a difficult [mutation] to target, but there may be a cooperating driver that's always found that might be very amenable to targetable therapy."