Professor of biology
Name: Daniel Branton
Position: Professor of biology, Harvard University, since 1973
Experience and Education:
— Associate and full professor of botany, University of California, Berkeley, 1963-1973
— PhD in plant physiology, University of California, Berkeley, 1961
— MS in pomology, University of California, Davis, 1957
— AB in mathematics, Cornell University, 1954
Researchers in the Harvard Nanopore Group led by Gene Golovchenko and Daniel Branton are working to develop a nanopore-based DNA sequencer that can analyze a mammalian genome for under $1,000 with very long reads, measuring the DNA’s nucleotides directly on the basis of their physical and electrical properties.
In Sequence recently spoke to Branton to find out where the idea for nanopore sequencing came from, what the major remaining challenges are, and how far he and his colleagues have proceeded in making nanopore sequencing a reality.
How did the idea of nanopore sequencing develop?
In a sense, it was started many years ago, in the 1920s and 1930s. [Wallace] Coulter had the idea of detecting small particles in a Coulter Counter, [which he patented in 1953], and that’s essentially a small hole. He never thought of going down to the atomic or molecular scale.
Then Hagan Bayley, [now at the University of Oxford], started using the alpha-hemolysin [protein nanopore], not to detect DNA, but for a variety of other reasons. And then Dave Deamer and I were thinking about ways of sequencing DNA, and we thought, because both of us are membrane biologists, that maybe you could do some interesting detection by getting DNA to move through a biological pore in a membrane. At the time, we didn’t know what pore to use, but then we learned about Hagan’s work, and we looked up the pore he was using, and it seemed to have just the right diameter for a single strand of DNA, and so we put DNA through it [see Kasianowicz et al., PNAS, 1996]. And that was sort of the beginning of work with nanopores and DNA. But it all arose as a result of Hagan having been using alpha hemolysin to detect other small molecules.
Since then, the field has grown enormously, and now we have a lot of collaborators and competitors.
How is the Harvard Nanopore group planning to use nanopores for DNA sequencing?
Most of our early work was done using a biopore, the alpha-hemolysin pore, which a lot of people have been using, and there will probably, in the future, be other biologically motivated pores, probably native or slightly modified native proteins, which serve as nanopores.
All of our work today is really now focused on solid-state pores, because we are interested in developing a new detector. The initial use of the nanopore, and still today even with the solid-state pores, has been to detect the presence of DNA, and the activities of the DNA in the pore, by means of monitoring the current through the pore itself. In other words, just having the pore, without any other construct being involved, gives you a monitor, in that there is a current that flows through the pore when the pore is open, and when the pore is blocked or partially blocked, the current changes with a given voltage. When DNA is going through the pore, it blocks the current.
It also is the case that when DNA is folded upon itself, so that you have two strands in the pore, if the pore is large enough for two strands at once, then you see that fold, so that you can detect the way in which the DNA is folded. All of the initial work with the biopores has been done, and probably for the future will be done, using just the blocking of current as the primary monitor.
We are interested in developing a monitor [that] we will expect will be much more sensitive to the local atomic-scale passage of individual nucleotides. The pore as a whole, that is to say, the blocking of the pore, is unable to detect individual nucleobases, because inevitably, a pore has a finite length. You can dream of an imaginary pore, which is a hole through an infinitely thin sheet, but there is no such thing as an infinitely thin sheet. In reality, the pores, for example the biological pore, has a length of about 5 nanometers. The solid-state pores that we use can vary somewhere between 5 [nanometers] up to whatever length we want. And the reason we are interested in solid-state pores is that the kind of detector we are constructing depends upon electronics, and you can’t fabricate electrical wires and so forth on a biological pore.
The type of probing that we expect to be doing involves tunneling microscopy. It depends upon atomic scale properties of the material under the probe. One usually has something called an emitter, which is a very sharp, pointy object, usually a piece of tungsten or other kind of metal that you can fabricate into a very sharp point, and a collector. The collector is usually the object on which the specimen sits. In a scanning tunneling microscope, the emitter is what actually moves and scans across the specimen and shoots electrons into the specimen, which then move to the collector on which the specimen is sitting. In other words, the specimen is sitting on ground potential.
And indeed, early electron microscopy, or early sequencing, thoughts were that one could use a scanning tunneling microscope to actually sequence DNA, and there was a lot of activity in this direction in the 1950s, 60s, [but] nothing actually transpired. Part of the problem is, of course, [that] it’s extremely difficult to scan across this very curly object, because DNA is not a straight line. If you put the molecule down, it’s all over the place; it balls up into twisted material.
So one of the initial reasons, indeed, for using a nanopore, is that, if the pore is small enough, then only a single strand can go through at a time. It can’t go through in a twisted configuration. It may wind up like a ball of string or knotted on one side, after it’s gone through, but while it’s going through, it has to be untangled.
In effect, what we are thinking about is tunneling into a molecule [that] is moving by the probe, [or] the emitter, rather than having the emitter scan across a specimen. The emitter, in our case, is going to be a [carbon] nanotube. One of the reasons for using carbon nanotubes is that they come in very small sizes. One can easily generate a 1-nanometer diameter carbon nanotube, and single-wall carbon nanotubes are usually between 0.8 to 1.5 or 2 nanometers in diameter.
In our case, the collector will also be a nanotube. And our reason for being interested in having the collector be a nanotube also is, it turns out that the atomic structure of the nanotube, essentially carbon in hexagonal-like arrays, has the property that it binds the individual bases of DNA. And as you may know, the bases in DNA stack on each other. That is called pi-stacking, which connotes a particular electronic configuration around the atoms. It turns out that individual nucleobases pi-stack on the surfaces of nanotubes. And it is this pi-stacking which can hold each nucleobase in the same orientation and configuration, with respect to the collector, each time. In using tunneling microcopy or in doing the kind of tunneling we are talking about, because it is so sensitive to the atomic-scale orientation of the specimen itself, it’s necessary to have some means of orienting the specimen in a reproducible manner. And that’s why we are using carbon nanotubes.
What are the main remaining challenges to what you are trying to do?
One of the key challenges is assuring a reproducible orientation of the nucleobases with respect to the emitter and collector, and we are doing this by providing a collector [that] tends to orient the nucleobases.
The other major challenge has to do with the speed with which the DNA is sliding or moving across the nanotube. Obviously, the individual nucleobase has to be in the detector long enough for detection and identification to take place. That has to do with the speed with which the DNA is being transported through the nanopore, which is functionalized with a nanotube.
We haven’t yet shown this, and that’s why it remains the furthest out yet to be done, is to actually demonstrate that we can control the motion of the DNA on the nanotube. We have learned a great deal about the fact that nanotubes orient the nucleobases, and we have demonstrated what the binding energies between DNA and the nanotubes are, and those are within the range that we believe will afford us the control we need on the sliding, because we can apply a voltage bias to the nanotube.
As you know, the DNA itself is negatively charged because of the phosphates along the backbone. So if one applies a negative bias to the nanotube on which the DNA is resting or sitting or sliding, it will tend to separate or move the DNA a little bit away from the nanotube. And by controlling that, we should be able to reduce the atomic scale friction between the DNA and the nanotube, so that it will slide more easily, or with greater difficulty. That’s going to be the mechanism that we hope will control the rate at which the DNA moves through the nanopore. DNA has to move at a rate [that] is commensurate with, essentially, the bandwidth that we have in our collector.
What is that bandwidth?
In theory, given current day amplifiers and bandwidth, it should be possible to do it at about a base per microsecond. But we anticipate that there are going to be problems going that fast, and it will be around 104 bases per second, rather than 106 bases per second.
On what timescale do you think you will be able to make an actual sequencing device?
Our current program is oriented towards about five years. A lot of work will be done in the meantime using sequencing by synthesis, well ahead of the time when nanopores come online to do sequencing. But the virtue of using nanopores is, in the first place, extremely long read lengths, which of course saves a huge amount of time and expense doing computational work to align all the fragments, which are otherwise sequenced in shorter bits by other methods.
What do you think the error rate will be?
In principle, we have to be as error-free as any other method. Of course whether that gets achieved is the big question. But the field of scanning tunneling microscopy, and in general the field of tunneling electrons, has really been driven experimentally, rather than theoretically. I am not saying theory is unimportant, it’s essential to understand the theory and to have a good working knowledge of what’s happening, but the whole notion of using scanning tunneling microscopy came because somebody built the instrument and showed that it could do something and work. And there were a large number of experimental surprises that could not have been predicted. Subsequently, they were understood and the theoretical framework built.
But the same is in a sense true for what we are trying to do. Without being able to fabricate the nanopore, articulated with nanotubes, it’s very difficult to get going. So, one of the reasons things have been slow, and not a great deal of progress has been made, is that no one has had a pore with the tunneling probes in it, in other words, with the nanotube collector and nanotube emitter. We have that now. That’s a major step forward, I would say, that’s gone on in the last year or so. It’s the first time that such a device has been built, and the basic electronic tests have been done that show that it is working as expected. But now we are at the stage where we can start putting DNA through it and making measurements, and that will reveal a great deal.
What is the next major goal?
First, I think, to demonstrate that we have control over DNA sliding on a nanotube outside of our device. Subsequently, there are some obvious improvements to our device that it is already evident we are going to have to also address. Making those improvements will take probably another year or so.
One company that wants to commercialize protein nanopore-based sequencing is Oxford NanoLabs in the UK (see In Sequence, 4/8/2008). Is anyone interested in commercializing your approach?
We are working with Oxford NanoLabs. [The company] is using only the biological pores, and that’s their focus and is going to be their focus for the next few years. That’s why they, in a sense, have partnered with several different groups, us included, to develop the other portion of the nanopore, that is to say, work with solid-state pores.
Are other single-molecule technologies that promise long reads, like Pacific Biosciences [or VisiGen Biotechnologies], going to be competing with nanopore sequencing?
[Pacific Biosciences] has a very powerful technology that looks, at least from the distance, as though it seems to be the major competitor for nanopore sequencing. I do think, however, that in the long term, if nanopore sequencing works, it’s the only procedure [that] doesn’t require any chemicals. It doesn’t require expensive enzymes or means of attaching those enzymes to surfaces, or anything like that, all of which is part of what, for example, Pacific Biosciences requires.
And it’s very difficult to approach the cheapness of a chip with chemicals. As you know, the price of computers keeps going down, the speed goes up. We anticipate that once the first chips with nanopores are available and begin to be used, their price will go down, so that rather than having a chip costing $1,000 for the first sequence that’s been done, it will keep going down. And if it keeps going down at the same rate as other things in the computer world, it should be possible to sequence a human genome for $10, or maybe $100.