A University of Wisconsin at Madison-led team has used single-molecule optical mapping to unearth structural changes in two human tumors, en route to more extensive analyses of cancer samples using a related single-molecule approach currently under development.
As they reported online in BMC Genomics late last month, the University of Wisconsin, Madison's David Schwartz and colleagues applied single-molecule optical mapping to DNA from two oligodendroglioma tumors, uncovering thousands of suspected structural variants that spanned a few thousand to hundreds of thousands of bases of sequence in each tumor.
"What this paper reports, I believe for the first time, is the use of a purely single-molecule approach to do novel discovery of mutational structural variants in a solid tumor," Schwartz told In Sequence. "We have revealed many, many novel mutational events that are quite complex, which you wouldn't be able to see with sequencing."
Though his own group is transitioning its cancer analysis studies over to a nicking enzyme-based approach called nanocoding — which he and collaborators are working to commercialize through a startup called Industry 3200 — Schwartz said the current proof-of-principle study highlights the wide range of structural information, somatic mutation data, and clinical clues that can be gleaned from optical mapping.
In some instances, the method is able to pick up changes to individual bases, particularly when these swaps wipe out or introduce sites that are recognized by the restriction enzyme used to cut DNA during the optical mapping process.
But the approach is most adept at picking up larger structural variations — including balanced rearrangements — that involve anywhere from a few kilobases to around a million bases of sequence, noted Schwartz, who came up with the restriction enzyme-based optical mapping methods that the Maryland-based company OpGen licensed from the University of Wisconsin several years ago.
Trevor Wagner, senior manager of applications with OpGen, said he was not surprised by the extent of the structural variation that could be identified in the tumor genomes with an optical mapping approach, since several publications have highlighted the potential of whole-genome optical mapping for finding large structural variants that other technologies miss.
"It's really able to detect things that sequencing can't and that microarrays can't," Wagner told IS. "It's able to discover new and novel things that might be part of — in this case — human disease."
Wagner said he expects to see optical mapping applied to a range of basic research problems and clinical applications, just as next-generation sequencing has been.
Last month, for instance, OpGen opened a CLIA-certified lab that provides surveillance and diagnostics services centered on organisms involved in hospital-acquired infections.
OpGen CEO Douglas White noted that that the company's CLIA lab does not currently offer clinical tests on human chromosomes, though its more research-focused service lab has used the Argus whole-genome optical mapping system to assess a wide range of human tissue and cell types.
OpGen's service lab is also applying optical mapping to epidemiological studies of healthcare-acquired bugs, foodborne pathogens, and other infectious microbes (see IS 2/19/2013).
In general, the optical mapping method used by Schwartz's research team and OpGen involves mapping the DNA fragments formed when many individual DNA molecules from a given sample are cut with a restriction enzyme.
For instance, cutting DNA from a typical human genome with a restriction enzyme such as BamHI produces somewhere in the neighborhood of 360,000 restriction fragments, Schwartz explained. By tiling fragments across the length of the genome, it's possible to pick up changes in fragment size and orientation relative to the pattern produced when the same enzyme dices up the human reference genome.
Such optical maps can be used to detect everything from insertions, deletions, and copy number changes to balanced rearrangements and inversions — in some cases uncovering complex sequence patterns surrounding such rearrangements. The technique has also been gaining favor as a tool for helping with the de novo assembly of plant, animal, and pathogen genomes.
For the current single-molecule optical mapping study, Schwartz and his team decided to use the approach for yet another application: tackling a human cancer.
Despite the advances that have been made in characterizing various cancer genomes, there are still mutations that are tricky to see by sequencing, he and his co-authors noted, particularly structural variants involving a few thousand to several hundred thousand bases of sequence.
Meanwhile, techniques such as karyotyping, fluorescence in situ hybridization, SNP array analysis, or array comparative genomic hybridization tend to offer either low resolution across the whole genome or high resolution within a targeted portion of it, they noted. And many methods are prone to miss rearrangements where no DNA is added or subtracted, such as inversions.
On the other hand, the study's authors reasoned that "[o]ptical mapping offers several unique advantages toward assembling the complex structure of a cancer genome."
"Genomic DNA isolated directly from cells is analyzed, thereby obviating any bias introduced by amplification or cloning steps," they wrote. "Moreover, because the DNA is of high molecular weight … segmental duplications and other repeat-rich regions of the genome are revealed, and additionally, the structure and long-range context of any aberration can be determined."
Using oligodendroglioma samples provided by MD Anderson Cancer Center's Oliver Bogler, a co-author on the study, the researchers produced optical maps using SwaI — a restriction enzyme selected for the relative uniformity of the fragment sizes produced when it chops up a typical human genome.
Starting from slices of tissue from two oligodendroglioma tumors, the team mashed up cells from each sample and broke them open to retrieve relatively large DNA molecules, ranging in size from around 500,000 bases to roughly 1 million bases apiece.
From that collection of DNA in the test tube, which Schwartz likened to a "floppy ball of yarn," the investigators then stretched out individual molecules of DNA using microfluidic devices.
From there, they chopped up the immobilized DNA with SwaI, visualized each fluorescently labeled molecule with a light microscope, and used an existing pipeline to analyze fragment patterns for structural variant detection.
With the help of automated scanners, the team looked at thousands of molecules in parallel on each microfluidic chip, ultimately teasing apart restriction maps, or Rmaps, for hundreds of thousands of molecules spanning the two tumor genomes.
"Our datasets for human and cancer genomes consist of hundreds of thousands, if not millions, of these single-molecule restriction maps," Schwartz explained, noting that each Rmap is akin to an individual sequence read.
In the case of the oligodendrogliomas tested for the current study, the group assembled an Rmap consensus sequence that covered nearly 97 percent of one tumor genome at a depth of nearly 37-fold, on average. The Rmap consensus for the other tumor spanned almost 94 percent of the genome at an average depth of more than 25-fold.
Since they did not have access to matched normal samples for the oligodendrogliomas on hand, the researchers compared the tumor maps to optical maps representing the human reference genome, uncovering almost 1,100 apparent structural variations in each of the tumors.
To narrow in on those most likely to be authentic mutations, they ultimately weeded out structural variants described in available databases and publications, an analytical strategy that Schwartz called "very conservative."
"Any structural variant that was scored in the tumor that has also been discovered or is present in a database of genomic variants, we count as a polymorphic event," he said.
All told, he and his colleagues tracked down 94 apparent somatic mutations affecting two-dozen genes — a set that included alterations verified with the help of alternative approaches such as PCR, SNP microarray, and array comparative genomic hybridization. They also detected suspicious structural changes falling in non-protein-coding portions of the tumor genomes.
Among the structural changes detected in one of the tumor genomes was a chromosome 7 inversion similar to that described in Williams-Beuren syndrome, a neurodevelopmental condition characterized by learning and other disorders, particular personality traits and unusual facial features.
The inversion detected in one of the oligodendroglioma tumors is "not quite the canonical" Williams-Beuren syndrome inversion, Schwartz said, though he and his co-authors suspect that the newly detected inversion may be a variant contributing to oligodendroglioma, which affects support cells in the brain's frontal lobe.
While the optical mapping approach used to test the tumor samples does not provide restriction mapping profiles for individual cells, investigators can get a good idea about the various genotypes present within a given tumor sample by looking at many long DNA molecules from cells in that sample.
"Because molecules are very long, you can actually phase them — in other words, determine that this mutation exists in the same proximity as this mutation," Schwartz said.
The approach also appears to have potential for distinguishing between different sub-clonal populations present in the same tumor. In one of the two oligodendrogliomas tested, for example, optical mapping unearthed different loss of heterozygosity patterns in two nearby samples from the same tumor.
On its own, the single-molecule optical mapping method provides information that resembles what might be possible if karyotyping could be done at much higher resolution, Schwartz said. And in combination with sequencing, such optical maps offer what he called a "very complete view of the genome."
In late 2011, OpGen introduced software called Genome-Builder for bringing together optical mapping information with DNA sequence data (IS 10/18/2011).
While approaches such as optical mapping "will not give you a lot of nucleotide-level information," Schwartz said, "they can complement what you do with sequencing or maybe even obviate it, depending on what you want to know."
For their part, Schwartz and his team are also pursuing another single-molecule approach known as "nanocoding," which centers on the use of so-called nicking enzymes. In that system, each DNA molecule is assessed in tiny channels, or "nanoslits," as opposed to sticking to the surface of a chip.
Schwartz is involved with the startup Industry 3200 that is developing a commercial nanocoding device, planned as the company's first commercial offering. The anticipated device is based on the latest version of the team's nanocoding technology, first described in a 2007 paper in the Proceedings of the National Academy of Sciences.
Since that proof-of-principle study was published, the team has gone on to automate the nanocoding system and has worked out issues related to the physics of the process, Schwartz noted.
He and his collaborators are now gearing up to look at many more tumor samples and tumor types using nanocoding devices, while at once working on ways to develop a nanocoding device that can perform both structural variant detection and DNA sequencing.
"What we're in the midst of doing is trying to build a cluster of [nanocoding] scanning stations and so on to have very high throughput in terms of tumors," Schwartz said.
"We're also building a pipeline that combines nanocoded data with sequence," he noted. "So in essence what's going to happen is a whole discovery engine."