Labeled DNA molecules line up like trains in a yard for genome mapping. Credit: BioNano Genomics

Optical DNA mapping may be the closest thing to taking out a ruler and measuring between the markers. DNA is typically pulled across a glass slide, then it is cut or hybridized to fluorescent probes at known sequences so that distances between spots or gaps can be calculated under the microscope. Although the method can be very effective, the difficulty of stretching the long and flexible DNA polymer uniformly has stymied its mainstream use.

Pui-Yan Kwok at the University of California at San Francisco, Ming Xiao at BioNano Genomics and their colleagues now show the potential for genome mapping to become widely accessible using a mass-produced chip that stretches DNA reliably for high-throughput automated measurements. The key to the chip is a set of 12,000 nanochannels that are each only 45 nanometers wide. “When DNA is in such a confined space, it cannot wiggle, so it's really stiff and straight, and the length is quite reproducibly set,” says Kwok. “It's stretched to a degree that you can measure [distances between sequences] much better.”

Maps derived from long single DNA molecules provide information that is difficult to obtain from sequencing and array-based genotyping. The authors genotyped molecules averaging >100 kilobases, which is substantially longer than reads generated by current sequencing methods. This capability is an advantage for repetitive DNA—regions which short sequencing reads are difficult to assign to unambiguously and for which physical maps provide effective scaffolds for genome assembly. The length also allows large structural variation to be reliably located. Hybridizing DNA to microarrays can find copy number changes due to insertions and deletions, but unlike the physical maps, it cannot always tell where in the genome multiple copies originate, and it misses changes that do not affect copy number.

A challenge for nanochannel devices is to thread matted coils of DNA through a narrow opening. To accomplish this, an electric field first sends DNA past progressively smaller pillars on the chip, causing it to unwind like a ball of yarn in a pinball machine before entering the nanochannels. The chip can be flushed and replenished several times, allowing 3 gigabases of DNA to be processed in one experiment. To visualize genotypes, the authors labeled DNA prior to loading with a sequence-specific nicking enzyme followed by a polymerase that adds fluorescent nucleotides. After automated imaging, they applied clustering software to find overlaps between molecules with identically ordered labels, merging signals to give distances with a standard deviation of ∼443 base pairs at 20× coverage. At this level of redundancy, essentially all missed or inappropriately labeled sites in single molecules are corrected.

As a proof of principle, the authors generated a map of the highly repetitive 4.7-megabase human major histocompatibility complex and used it as a scaffold to assemble the region from short-read sequence data. They detected structural and sequence differences between four parental DNA strands from two individuals, highlighting the fact that the method gives haplotype information. A haplotype is the unique string of sequences that originate from one parental chromosome. It is critical for many genetic analyses, but it cannot be derived directly from most sequencing and genotyping approaches.

Ultimately, it will be important to show that the method can tackle longer sequences. The groups have had success with human chromosome–length DNA and are working on high-density maps using multiple nicking enzymes and multiple fluorescent colors. “The mapping density [will] approach the sequencing level—not a single base but a few bases,” says Xiao. Resolution can also be improved to around 100 base pairs using super-resolution imaging. There will be a lot to see as DNA lines up for the camera.