Sure, single-cell sorting devices are cool and useful and all, but Harold Craighead’s lab at the Cornell University Department of Biomedical Engineering is developing a microfluidic device that can separate individual methylated DNA fragments from a single cell’s total genetic content. In a lab test reported in their recent open-access paper (pdf) in Proceedings of the National Academy of Sciences, the team used the device to separate methylated plasmids from among 11 femtograms of mixed DNA, hitting a 5.6 percent false-positive rate and 3.5-fold enrichment. That level of enrichment is typical of immunoprecipitation methods that need about 1,000 times as much input DNA.
I spoke to Craighead last week to ask him a little more about the device and its design, as well as when we might expect a commercial version fit for regular labs.
But first, a quick explanation. In general, the device uses electrical current to pull fluorescently labeled single DNA molecules through a nanofluidic channel, much like capillary electrophoresis. This current pulls the molecules to a Y-shaped fork, where a scanning laser excites any attached fluorophores. If the device doesn’t detect fluor-emitted light, the same current continues to pull everything through the same branch of the fork. But if it detects the right wavelength—or wavelengths—the device suddenly switches its current to pull DNA through the Y’s alternate branch.
To detect methylated DNA, the researchers labeled a collection of methylated and unmethylated plasmids with a fluorophore that emits red light. They mixed that with methyl binding domain protein-1—MBD1—labeled with a green light-emitting fluor, and let the protein do its thing. Using two lasers to excite molecules at the fork, and requiring two emission wavelengths to switch electrical current, the Cornell team was able to sort only those bits of DNA that had bound to MBD1.
That’s the experiment with a 5.6 false-percent positive rate and 3.5-fold enrichment. Unfortunately, it also had a 81 percent false-negative rate—meaning the device had missed about four of every five molecules of methylated DNA.
But Craighead says his team can almost certainly improve those odds with a few chemical and design tweaks. It’s the concept they’re interested in.
Why did you decide to focus on epigenetic analysis for this paper?
We’ve been working on different aspects of single-molecule nucleic acid analysis for a while, and developing approaches for detecting various features. But with epigenetics, the motivation was really to try to deal with the issue of sample inhomogeneity that’s inherent in mixing together samples from inhomogeneous sources, and also to develop techniques that could be used to track small samples over time, as sort of a research vehicle.
Also, we wanted to expand the depth of information that we could get from tissue and other samples that have some of those inhomogeneity problems.
It’s applicable beyond epigenetics, though?
Correct. We’re developing a generic molecular sorting approach and demonstrating it for sorting DNA with methylation differences. Any optical signature that you can identify in a molecule, we use as a key to sort it out and analyze it.
In what situations might come in handy?
There are two general areas that I can imagine. One is the work we do with cancer motivation. So, rare circulating tumor cells that you might want to catch and find out what stage of cancer development they represent, or what treatments might be appropriate—and you may only have a few cells. Phenotypic variations may not be obvious, and this gives us the ability to look at the underlying epigenetic state to determine their characteristics.
So that would be a few rare cells that you want to look at, and so you only have a small amount of material, so you want to look at the material that comes from a few selected cells.
Another case would be an organ whose condition you want to sample, but the types of cells available are very inhomogeneous, so you want to sort out the cells that represent exactly the sample that you want, so you can analyze those in detail. So, sorting out a few cells that you care about, and analyzing those with the assumption that you won’t be able to amplify those samples, so you have to be able to develop the techniques that are sensitive enough to look at the material from just a few, and that motivates the single-cell approach—to use every molecule that we extract from these cells.
This technique is very similar to single-cell sorting—why wasn’t this developed sooner?
It’s basically the signal-to-noise [ratio]. So with an individual molecules, you can’t put a whole lot of label molecules on each, typically. Your single fluorphore, for example, only has a few photons to deliver, so the background may overwhelm that [signal].
Did the high level of noise interfere with the device’s performance?
Yes. The major technical issue is how you increase the signal-to-noise to allow you to rapidly identify these optical signatures.
If you want to run molecules very fast through the system, you won’t collect enough signal to get a reliable sorting signature.
The whole system is designed to increase the signal-to-noise, including optical collection in a very small volume to eliminate the background fluorescence. [Our signal] filtering technique is one way to improve the recognition of the signature that we want, by matching the shape of what you see with what you believe it should be. So, some knowledge of what the signal should be allows you to pick out with greater efficiency the correct one.
How might you improve this device’s ability to pick out epigenetic marks?
There are two components to the overall efficiency. There’s the optics and how well you can identify something, and then how well you can actuate an electrical signal to sort that. The other side is how well the chemistry works in highlighting the feature that you want.
Both of those aspects have to work. The chemistry has to reliably bind and keep the fluorescent label where it’s supposed to be, and then you have to efficiently collect that. This paper is predominantly on the hardware aspect of that, because the chemistry can be done independent of the sorting hardware.
You’re planning to commercialize this through your company Odyssey Molecular—what are you working on?
Yes—we’re just starting out, trying to push this toward a higher-throughput version. So what we’re doing now is basically demonstrating the concept and the capabilities, but we’re imagining we’ll use more modern fabrication techniques to make the system more compact and to have more channels operating in parallel.
When do you hope to have a product?
That’s an interesting question! This type of technology takes a few years, at least. I’d say that the first step is to get it to work in the lab and understand the basic issues, and then worry about scaling it up and making it cost-effective.
[Reproduced above with express permission of Harold Craighead is part of Figure 1 from "Real-time analysis and selection of methylated DNA by fluorescence-activated single molecule sorting in a nanofluidic channel," by Benjamin R. Cipriany, Harold G. Craighead, et al., Proceedings of the National Academy of Scientists, DOI 10.1073.]
Cipriany, B., Murphy, P., Hagarman, J., Cerf, A., Latulippe, D., Levy, S., Benitez, J., Tan, C., Topolancik, J., Soloway, P., & Craighead, H. (2012). Real-time analysis and selection of methylated DNA by fluorescence-activated single molecule sorting in a nanofluidic channel Proceedings of the National Academy of Sciences DOI: 10.1073/pnas.1117549109