NEW YORK (GenomeWeb) – Researchers from New York University have developed a method for detecting very rare mutations in a population of cells using error-corrected sequencing.
The approach, called maximum-depth sequencing (MDS), adds barcode and adaptor sequences directly onto a short sequence of interest so it can be amplified and sequenced many times in parallel to produce a consensus sequence.
In so doing, explained co-senior author Evgeny Nudler, a professor of biochemistry and molecular pharmacology at New York University School of Medicine, the approach applies "all the power of a high-throughput sequencing machine on a small region."
In a study published earlier this month in Nature, the team used MDS to scrutinize mechanisms behind bacterial mutagenesis and the frequency of such events, demonstrating that it could pick up rare mutations in an unbiased manner without long-term growth assays or selection. In an Escherichia coli strain examined for the analysis, the researchers saw mutation rates that were on par with those reported previously.
Still, the MDS approach made it possible to see site-specific mutation rate differences and get clues to the mechanisms of mutation that might otherwise be missed, Nudler said.
"Almost every aspect of bacterial mutagenesis is sort of debatable, so there are opposite views on whether highly expressed genes are [more or less mutation-prone], whether antibiotics cause mutations directly or indirectly, and what mechanism could lead to that," he explained. "Using this method, we get pretty solid data … addressing those questions."
The team has filed for a patent on the MDS approach and permutations of it that use CRISPR-Cas9 — rather than restriction enzymes — to more precisely cut the DNA adjacent to targeted sequences of interest.
"You can imagine that, in the future, when we work with precious DNA samples isolated from human tissues, the starting amount of DNA potentially could be limiting and there could be advantages to using that same sample for [testing] multiple different genes," explained co-author Aviram Rasouly, a postdoc in Nudler's lab.
"We've already adapted [MDS] so that the cleavage is with Cas9 instead of with restriction enzymes and we also modified our adaptors now in a way that reduces background [for] even higher sensitivity and lower error rate," Rasouly said.
Though MDS development is still in progress, Nudler said he would consider future commercialization of the method, particularly in clinical areas such as cancer diagnostics, tumor mutation profiling for treatment selection, or treatment response tracking with cell-free DNA.
"We're actually using … a slightly improved version of our MDS [method] to detect rare mutations, not just in bacteria but also in populations of cells in blood, for example, where you're looking for rare cancer mutations," he added.
The approach differs in several ways from traditional barcode-based sequencing methods, he explained. Rather than simply using a primer to introduce a barcode and adapter sequence to a given sequence, for example, the team starts by using a restriction digest to chop the DNA next to the region of interest.
Then, after one round of PCR with a barcoded primer that anneals to the 3'-end of the sequence of interest, the researchers remove barcode sequences that were not used in the reaction. Because the targeted sequence has been cut with the restriction enzyme, the PCR reaction produces new sequence that's not only complementary to the sequence of interest but also sequence directly attached to the target that complements the adapter and barcode sequences.
"Because of the exposed 3' site on the genomic DNA molecule left by the restriction enzyme, the genomic DNA molecule acts as a 'primer,' causing the barcode and an adapter to be synthesized onto the end of the [region of interest]," the researchers explained.
Through subsequent linear amplification of the sequence containing the adaptor, barcode, and the sequence being targeted, the investigators can produce many copies of the original sequence and compare them with the help of the barcode to weed out sequence errors introduced by PCR and identify authentic variants or new mutations.
"By synthesizing unique barcodes directly onto the [region of interest] of an original genomic DNA molecule and then copying that molecule using linear amplification, we increase yield and substantially reduce both polymerase and sequencing errors," the MDS developers wrote.
"By increasing the number of reads used to call a consensus sequence, MDS can lower error rate indefinitely," they suggested, "given sufficient coverage."
The team noted that the approach may get a further boost in accuracy by introducing a second barcode sequence following the linear amplification step, though it did not take that tack for their E. coli experiments.
For their Nature study, Nudler and his colleagues demonstrated the utility of MDS for tallying up mutation rates at specific loci in E. coli, using cultures grown from the MG1655 strain for no more than 120 generations.
"The motivation was to determine the frequency of mutation in bacteria directly," Nudler said, noting that it's been "a challenge for decades" and usually requires some sort of selective pressure on bacteria.
"We looked for an unbiased method that could directly detect rare mutational events," he noted. "Next-generation sequencing obviously provides this opportunity, but the existing methods have their own shortcomings."
The team focused on six 100-base regions — parts of the genome that were selected for their potential biological or clinical significance. These included an RNA polymerase enzyme-coding sequence implicated in resistance to the antibiotic rifampicin, for example, along with the coding sequence for a protein related to penicillin binding.
After applying the MDS protocol, the researchers performed multiplex sequencing on the Illumina HiSeq platform, aiming for around 15 million reads per targeted site, Rasouly said.
Although the general MDS approach is platform agnostic, he noted that the relatively low error rate and throughput of the Illumina technology were advantageous when trying to pick up rare mutations, even when many reads from the same locus were available.
The team developed a consensus sequence for each group of reads with the same barcode. Compared with results from past mutation rate studies of E. coli — many of them done with long-term evolution experiments that rely on whole-genome sequencing of cultures grown over long periods of time — the locus-specific mutation rates identified by MDS generally agreed, explained Rasouly.
"The power of MDS," he explained, "is the ability to measure mutation rate in the absence of selection. One of the things that we show in the paper is the ability to measure, very accurately, the mutation rate at any given site."
Along with small changes such as substitutions or small insertions and deletions, the approach appeared to pick up transient substitution mutations that clustered in certain parts of the genome and affected just one strand of bacterial DNA.
"In vivo, these misincorporations must be reversed before genome replication," the authors wrote. "However, our observations represent a snapshot of this dynamic process before this repair can occur."
To distinguish such temporary mutations from those maintained in the cells, the team did further analyses of cultures grown over 20 generations or less, a time frame that is expected to be too short for the accumulation of many authentic mutations.
On the other hand, experiments using E. coli treated with relatively low doses of antibiotics, such as ampicillin and norfloxacin, provided a peek at mutations that begin to arise in response to such drugs.
The team is currently testing out MDS for finding mutations in cell-free tumor DNA in the blood of patients with non-small cell lung cancer, focusing on mutations in genes with potential prognostic or treatment relevance, such as EGFR or KRAS.
Preliminary unpublished results suggest that the same mutations present in biopsy samples from a patient's tumor can also be detected very sensitively with MDS, Nudler noted.
The team is also looking at how the accuracy, mutation detection sensitivity, yield, and other features of MDS stack up against duplex barcoding and circular consensus sequencing methods to attain the type of data needed to pursue a commercial version of the test.