NEW YORK (GenomeWeb) – A Broad Institute team has figured out a way to encode spatial information into DNA and perform microscopy using a next-generation sequencing instrument.
"It's imaging from the cell's point of view," said Jonathan Weinstein, a postdoc at the Broad Institute. This allows us to visualize biology as cells see it and not as the human eye does. The underlying idea is to treat the individual transcripts, or even proteins tagged with antibodies labeled with DNA as sensors in a network and to get this network to communicate in a way that allows for relative positions of molecules to be encoded into DNA."
Weinstein described the method, which he called DNA microscopy, in a paper published today in Cell, coauthored by Broad Institute Core Members Aviv Regev and Feng Zhang.
The authors claimed that the method doesn't require any specialized technology to generate cellular-resolution images that combine spatial and molecular information. "You can do this tomorrow, assuming you have a sequencer to run on," Weinstein said. The method does require reaction chambers that are not available off the shelf, though the description for how to make them is included in the paper.
In their paper the authors described the technology as "optics-free," though that is "kind of a misconception, because an Illumina sequencer is a fluorescence single-molecule microscope," said Long Cai, a researcher at CalTech who developed seqFISH+, a transcriptome imaging method, but was not involved in developing the new technology. "So imaging is done anyways, just packaged away in a black box in sequencing-based approaches to get spatial information." Coming up with ways to get spatial information through sequencing can come at a cost of resolution, sensitivity, and even throughput, he said. "Having said that, [DNA microscopy] is clearly an interesting and clever approach."
The technology is so nascent that its resolution limit and sensitivity have not yet been quantified, and will need to be done experimentally, Weinstein said. Nevertheless, he suggested the new technique could "take biological analysis a major step forward," especially in applications looking at "idiosyncratic" biological systems, like immune cell profiling or in the brain, where "you have to treat each biological sample as unique and not replicable."
The method is the latest entrant to the field of spatial genomics, which promises researchers a combination of spatial and molecular information. As reported by GenomeWeb, a number of companies are offering products to that end, including NanoString Technologies, 10x Genomics, Bruker, BioSpyder, and Cartana.
"I think we're just beginning to see people adopt this," said George Church, professor at Harvard Medical School, who was not involved in the research. "I'm not saying any particular technology will have a big impact or not, but it is a safe bet that one way or another we're going to have very high-quality data at whatever resolution we want, down to about 10 nanometers. And we'll have very fast methods as well." Church co-founded ReadCoor, a spinout which is commercializing FISSEQ, an in situ RNA sequencing technique developed in his lab.
The paper represents the end of a six-year process for Weinstein. While he credited his coauthors with support and direction, he admitted the execution of this idea has been somewhat of a "focused and determined solo effort."
"Josh has been absolutely committed to addressing this gap in molecular biology. The work has required a combination of imagination and skill, in both biology and math, that is almost inconceivable to find in just one person," Regev said in an email. "He devised the approach and spearheaded the experimental and computational development over the last several years, and it's been incredibly energizing to help shepherd this new technology for visualizing cells."
Weinstein, who has a background in physics and biophysics, said the key concept was to find a way to treat DNA the same way a light microscope treats photons. Super-resolution microscopy, which can achieve resolution of 10 nanometers, "exploits the fact that light is made of photons and treats them each as independent [statistical] trials in the interrogation of light-emitting molecules," he explained.
DNA microscopy's photons, so to speak, are sequences of DNA that are generated during an overlap extension PCR reaction that concatenates sequence tags of nearby target molecules. Starting with a fixed sample, the method labels every molecule of interest with a unique molecular identifier (UMI), in this case a 30-base-long, completely randomized oligo. "You are ensuring when one UMI labels one molecule, no other molecule has the same UMI," Weinstein said. The overlap extension PCR incorporates nucleotides into another reference sequence, dubbed a unique event identifier, or UEI. Every subsequent amplification step generates a new UEI and as molecules diffuse through the system, "the experiment records in detail the time course of events," he said.
Sequencing these UMIs and UEIs provides a matrix which can be analyzed to estimate every single molecule's position relative to every other molecule in the sample, thus producing the image.
"What's interesting about the UEI as an information unit is there's strong analogy between it and photons in light microscopy," Weinstein said. The resolution of both light and DNA microscopes is dependent on the standard error, which falls proportional to the square root of the total number of samples. This is partly why the Cell paper does not publish a resolution, Weinstein said. "Resolution is not a number, it's a number [related to the diffusion length scale of DNA] divided by an experimental parameter."
He estimated that the experiments disclosed in the paper allowed him to resolve individual molecules approximately 10 microns, or 10,000 nanometers, apart. "But you can alter the experimental parameters in productive ways that give you sub-cellular localization of the molecules," he said. "You can get significantly beyond that, provided you sequence that deeply."
For comparison, Cai said seqFISH+ offers 100-nanometer resolution with 50 percent detection efficiency, at a throughput of 2,000 cells, while Slide-seq, a droplet-based spatial genomics technique also developed at the Broad, offers 10-micron resolution, 0.1 percent detection efficiency, and a large field of view.
The authors disclosed in the paper that they have applied for patents related to the technique. Zhang and Weinstein applied for a US Patent for "Spatial and cellular mapping of biomolecules in situ by high-throughput sequencing" in 2014.
Weinstein said he's focused more on dissemination rather than commercialization, for the moment. "The main goal is to explore important areas of biology," he said. "There are too many potential avenues to narrow down where commercialization would be most appropriate." He suggested that making a reagent kit to sell through distributors "makes some sense." However, "it's not obvious yet whether that would be the right thing to focus on," he said.