David Bell was game to give it a try. About six years ago, William Glover, president of ZS Genetics, a DNA sequencing company based in North Reading, MA, asked Bell, an electron microscopist at Harvard University, to develop a DNA sequencing method using transmission electron microscopy. Bell said OK, but he wasn't sure how it was going to work.
When scientists talk about reaching this goal, the conversation usually revolves around the usual suspects: next-generation or third-generation sequencing technologies such as Pacific Biosciences’ single-molecule sequencing, Ion Torrent’s semiconductor sequencing, and Oxford Nanopore’s nanopore sequencing techniques. Electron microscopy is rarely mentioned.
But Bell’s hoping to change the conversation. In a proof-of-principle paper published in Microscopy and Microanalysis (1), Bell and colleagues described a new electron microscopy approach for DNA sequencing that may win that gold prize. In that study, the group was the first to successfully prepare a sample of DNA, take a picture of it with a transmission electron microscope (EM), and read the sequence directly from the picture.
So, why hasn't electron microscopy been successfully developed for DNA sequencing? Well, for starters, the four different bases that make up DNA differ by only a few lightweight atoms. As a result, they’re not dense enough to reflect an electron beam, and consequentially don't produce enough contrast to be accurately differentiated from one another by an EM. They look almost transparent.
To solve this, Glover sought help from geneticist Kelley Thomas at the University of New Hampshire who had extensive experience working with DNA molecules. To provide the needed contrast, Thomas and colleagues developed a technique to tag DNA bases with large, heavy metal atoms. In that process, they used PCR to copy the strand of DNA, replacing a specific base with the heavy metal-tagged version.
"They started with iodine," says Bell. "That worked pretty well. We could see variations along the strand as to where iodine was." Then, they moved to mercury, a much larger atom. "Mercury scatters really well. It shows up really, really clearly. As we show in the paper, we don't have to do image processing. We can just read the bases right off the picture."
The technique tags the bases with high specificity: you can specify that mercury attaches only to cytosine, for example, and iodine attaches only to guanine. Getting the heavy metal atoms to stick correctly to the right bases was a key component of the technology. Because the bases that make up DNA are complementary, one can read an entire sequence using just two different heavy metals.
Others have tried—and failed—to develop a DNA sequencing method that uses transmission electron microscopy. For example, California-based startup Halcyon Molecular was developing a similar electron-microscopy sequencing technology. Halcyon's approach stained the entire DNA structure with heavy metals, rather than just the bases. But the company quietly shuttered earlier this year.
"It's nothing magic," says Thomas of their specific tagging approach, "just template-directed enzymatic incorporation of modified nucleotides. We haven't tried to do anything weird." In the end, Thomas' team only had to work out some details empirically, as some enzymes are more robust, and some heavy metals incorporate into the strand better than others.
To be read by the microscope, the DNA needed to be imaged in a single, long strand on an almost transparent substrate. The substrate has to be as transparent as possible to reduce the signal-to-noise ratio; right now Bell is using a carbon monolayer substrate.
But stretching out the labeled strands of DNA was a larger challenge. And that stretching issue can turn into spreading one, a problem that plagued researchers who tried sequencing by EM in the 1960s and 1970s. Spreading happens when the base-to-base spacing becomes irregular, preventing a reliable read. ZS Genetics has overcome this by using a combing system that uniformly pulls the strand across the plate using a constant force. "It works with very good repeatability," said Bell, "and even if it's stretched a bit, that's okay, as long as it's uniform along the entire strand."
In their proof-of-principle paper (1), Bell and colleagues reported that they were able to identify one type of nucleotide in one of the two strands of a DNA molecule using electron microscopy. In this experiment, the team started out by adding a mercury atom to a deoxyuridine triphosphate (dUTP) base. Then they incorporated this mercury-labeled dUTP into double-stranded DNA by DNA synthesis on a primer-bound, single-stranded DNA template. To image these mercury-labeled DNA molecules, the group used annular dark-field scanning transmission electron microscopy, which provided higher contrast than standard EM because the technique collects more of the electrons scattered by the sample.
A Third-Generation Competitor?
Bell is optimistic about his technique’s ability to compete. One reason in particular stands out: this technique can produce long contiguous reads—thousands or even tens of thousands of bases long, compared to hundreds with other approaches.
"You don't have to throw anything away," says Bell. "You can look at everything in one go." Those lengths enable de novo sequencing of a variety of types of samples that are currently hard to sequence with short reads. For instance, the repeat elements in eukaryote genomes can be longer than the read length, making accurate assembly close to impossible. With a longer stretch of intact, fairly accurate sequence, researchers can look at genome structure in a much more powerful way.
So, this would make Bell’s technique particularly well-suited for environmental samples and metagenomic samples. It's also perfect for virus identification, because it requires only small amounts of DNA, unlike the Sanger method. And with viruses you want to image the whole strand in one shot, and the EM can do that.
"It's no longer a science problem. It's really a matter of engineering. It's a matter of scaling it up," says Bell. The next step toward commercialization would be to make a custom, automated transmission EM that would do the reads. "Currently, we can only image a certain section of the strand at a time. We'd like to make a system that would move along the entire length of the strand without having to splice the images together." An EM designed specifically to do reads of DNA would be cheaper to make than a general-purpose EM, and Bell believes the cost will be reasonable.
They've calculated that in theory, at current technology speeds, one could prepare a sample and sequence it in about 24 hours—the magic number. That's an order of magnitude better than anything else available. Currently, it takes a full day to prepare the sample with the heavy metal bases, but Bell says that part of the process may be able to be done more quickly once it's engineered on a production scale.
While the technique has the potential to be affordable and quick, those aren't enough, of course, to make a DNA sequencing technology successful. It also has to be accurate, and so far, this approach is passing the test. "We think we've got read errors in the 1-2% range," says Bell. "And with this technique, you don't get mislabels, where the label is on a base you don't want. The read errors are from the PCR labeling missing a base every now and then."
"The technique they are describing sounds completely feasible," says Derek Stein, a molecular biophysicist at Brown University. "The data looks really good. But is it going to be competitive? That's where I'm not as certain.”
The next major problem for Bell’s group will be to develop the technique to be cheaper and quicker without losing the ability to accurately distinguish the base pairs from one another. This has been true for the other so-called next-generation or third-generation sequencing approaches. For example, the technology to identify the bases with fluorescent molecular groups and then read them under a fluorescent microscope has been out there for quite a while, but engineering the process to bring down cost and speed and increase efficiency continues.
Even if the scaling-up proves more problematic than anticipated, Bell says that the DNA labeling technique may have a future in other sequencing methods. "For example, you don't get a lot of signals from the bases as they go through the nanopores. If you put heavy metal labels on the bases, you can amplify the signal."
But for now, the scale-up is in a holding pattern as ZS Genetics looks for investors to finance the project. When asked how long until the technology is commercialized, Bell says: "It depends on the venture capital situation, which is not so great for anyone right now. If we had the money, we could get it built tomorrow."
- Bell, D. C., W. K. Thomas, K. M. Murtagh, C. A. Dionne, A. C. Graham, J. E. Anderson, and W. R. Glover. 2012. DNA base identification by electron microscopy. Microscopy and Microanalysis 18(05):1049-1053.