associate professor, Department of Mechanical Engineering, Northwestern University
Name: Sandip Ghosal
Title: Associate professor, Department of Mechanical Engineering, Northwestern University, since 2000
Experience and Education:
Scientific research staff member, Combustion Research Facility, Sandia National Laboratories, 1998-99
Postdoctoral Fellow, Center for Nonlinear Studies, Los Alamos National Laboratory, 1995-1998
Research Fellow, Center for Turbulence Research, Stanford University, 1992-95
MPhil and PhD in physics, Columbia University, 1992
BSc in physics, Presidency College, Calcutta, India, 1985
A number of research groups are developing nanopore sequencing approaches in an effort to bring the cost of DNA sequencing within the range of the $1,000 genome. However, the goal of single-base resolution has proven difficult because DNA molecules move so quickly through the pore — at around 10-8 seconds per base pair in a typical experiment — that the bases can’t be read one at a time.
In an effort to better understand the resistive forces that are in play in this process, Sandip Ghosal, a fluid-dynamics researcher at Northwestern University, conducted a theoretical hydrodynamics study of the interaction between the DNA molecule and a nanopore.
The results of his study, published in the June 8 issue of Physical Review Letters, indicate that the fluid surrounding the DNA within the nanopore creates some resistance, although not enough to counter the electrical force that pulls the molecule though the pore.
However, these findings indicate that developers of nanopore sequencing methods might only have to tweak the surface chemistry of the pores in order to increase the resistive force of the fluid and slow down the molecule so that it can be read, according to Ghosal.
Last week, In Sequence spoke to Ghosal about these findings and their implications for nanopore sequencing.
Can you provide some background on your work and the motivation for this study?
First of all, in contrast to most work in this field, this study is entirely theoretical. Most papers in the nanopore area are written by experimentalists who are fabricating nanopores that they can test. But this is not the only theoretical paper. There are about three or four other theoretical papers in this area.
One of the difficulties in this nanopore sequencing area is single-base resolution. That’s the Holy Grail, so to speak. People have already demonstrated that you can detect homopolymers where different bases can be distinguished, and even a change in the base sequence can also be detected, but the question is, ‘If I give you a random sequence of bases, can you read off the bases from the signal?’
One of the difficulties in doing that is that both for the solid-state nanopores [which use a material such as silicon oxide or silicon nitride] and alpha-hemolysin pores [which are natural protein nanopores in a lipid bilayer], in single and double-stranded DNA, the DNA goes through the pore at a rate that is way too fast. No known method can detect [single bases] at that high frequency. The big problem is how can you revise the method so that DNA doesn’t go through so fast so you can read one base at a time?
So the theoretical question that I asked is, ‘What determines the speed of the DNA in the first place?’ If you don’t know what determines the speed, you can’t slow it down. Well, you can, but in a rather random way. If you want to do it in a systematic way, you better understand what is going on.
That was the intent of the paper, and the situation in this area is, we know very roughly what determines the DNA speed. We know that there is an electrical force that pulls the DNA through. We know that there is some kind of resisting force — there has to be, otherwise it would go at infinite speed. But as to the nature of the resistive force, there are various rough ideas and no clear consensus as to what is resisting the electrical force and what determines the speed of the DNA.
So I took an idea that was around that fluid viscosity in the pore could cause it, and instead of speculating that this could be a possible mechanism, I actually calculated using an analysis that if you assume that fluid viscosity is the only or the main resisting force, what would be the velocity? And then I compared it with the experimental [data], and showed that it agrees — to within a factor of two or three, we can’t be too accurate in this area — but within reasonable errors, it does agree with the experimental data, which would make it very plausible that it is the hydrodynamic friction, or the viscous force, that is the main resistive force.
And that gives a theoretical foundation to understand these DNA translocation velocities.
What are the implications of this for nanopore sequencing?
In terms of practical implications, once you understand what is making it go as fast as it is going, you can explore parameters that you can change to make it go slower. One of the parameters that people had thought of changing before is the molarity of the solution. This happens in a potassium chloride solution, and people have changed the potassium chloride concentration for the solid-state nanopores — in the alpha-hemolysin [nanopores] you can’t do that — in the hope that it would change the translocation velocity, and the fact is, it does not.
My theory explains why it does not, and the theory also suggests what might change it, and one of the things I propose in this paper is that if the substrate in the solid-state [nanopore], the surface of the pore, can be treated in a way that the surface charge is changed, then the translocation speed can be slowed down greatly. In fact, theoretically it can be made zero.
Once you do the analysis it’s easy to understand it. The point is, when you have a silicon nitride or a silicon oxide surface, the surface is actually charged. And because the surface is charged, if you apply an electric field, even without the DNA, there’s a strong flow in the direction opposite to the direction in which the DNA would normally travel. This is electro-osmotic flow.
So what is happening is the DNA tries to go through the pore, and there is fluid rushing in the opposite direction because of the electro-osmotic flow. It’s like trying to swim down a hydrant with the flow going against you. And by controlling the amount of flow in this hydrant, you can reduce your swim velocity to zero essentially. And that can be done by adjusting the surface charge. There are various chemical ways of doing that.
That’s something I suggested in this paper. It remains to be seen if an experimentalist will pick it up and actually show that you can have control over the translocation speed by changing the charge. That’s still up in the air.
Are you working with any experimentalists or any developers of nanopore sequencing technologies to determine whether that approach might work?
Not in terms of a direct collaboration, but I talk to various people. I send them my papers and we communicate via e-mail, but we don’t have a formal collaboration. This is a small part of my research. I’m not primarily a sequencing type of person. I do other problems in fluid mechanics and electrical effects in fluids and so on.
Do you have any particular ideas regarding how the surface charge could be modified, and what a realistic expectation would be for slowing down the DNA?
The surface charge can be modified in various ways. A simple way is to apply a coating. There is a substance called PMMA — polymethyl methacrylate — but if you coat a silicon surface with this substance, it changes the charge. The reason I mention this coating is that it’s something that people use all the time, and using this coating has the correct amount of charge that [would cause] the DNA to slow down. You can actually slow down the velocity to a very small value if you use this coating.
This is something I’ve suggested, but it has not been done. Not being an experimentalist, I don’t quite know if this coating can be made to go inside the nanopore. That might be a non-trivial problem. It’s one thing to coat a flat surface, but it’s another thing to coat the interior walls of a nanopore — it’s not a hole big enough to put your hand in. But if chemists can do the trick of getting this coating into the pore, then that is a possibility.
The other possibility could be to use a different material altogether. So silicon oxide and silicon nitride are typically what people use. There have been some attempts to do these nanopores in plastics, but that hasn’t been very successful. I don’t know what other tricks the chemists have.
There is also a way to treat the surface with ion beams to change the surface charge. So there are various tricks that they can do, but I’m not an expert on that subject.
The paper mentions that you predicted the resistive viscosity in a solid-state nanopore, but not in the alpha-hemolysin nanopore. Can you discuss why that was the case and what the implications of that might be for people working in this field?
The alpha-hemolysin pore is a biological pore. It’s a bunch of protein molecules that can adsorb onto a lipid membrane and it self-assembles. The thing about the alpha-hemolysin pore is that because it’s made of a particular set of protein molecules, every alpha-hemolysin pore is exactly identical to every other. So that’s the great advantage that they have — [they’re] very reproducible.
On the other hand, the solid-state nanopores — and you have to keep in mind that this is a really new and emerging technology — they cannot yet make them in a way that is reproducible. Going down to one nanometer, you can’t always control the shape and the size, so there is some scatter in the data.
But the advantage with the solid-state nanopores is that you have a lot more flexibility. The alpha-hemolysin pores, once they form, you can think of it like a soap bubble: It floats around for a while and then it pops, and when it pops it’s gone. And also, it floats around on the lipid membrane. It’s never quite still. You do experiments as long as it’s there, and then when it’s gone, you make a new one.
And another difficulty is that you can’t do these experiments at any salt concentration, because if you use other salt concentrations then the pore is not stable. It will disappear. But with a solid-state nanopore, you can use any salt solution, so it gives you a larger parameter space, so to speak, to do the experiments.
There is also a difference in the size. In the alpha-hemolysin pores the smallest diameter is about 1.8 nanometers. That’s so small that even a single DNA molecule — the normal double-stranded DNA molecule — can’t go through it. Only single strands can go through.
In the solid state pores, the good pores are typically somewhat larger, around the 5-nanometer range, and the experiments are usually done with double-stranded DNA. Now you can make solid-state pores around one nanometer, but those are not as good. So most of the experiments — at least the ones I have looked at — were done with nanopores in the 5- to 10-nanometer range, perhaps because we can’t get good reproducibility if they go down to 1 nanometer.
The difference in the translocation speed is about one to two orders of magnitude — a factor of about 10 to 100. That could be because the pores are larger.
The reason I used the solid-state pores to do the calculations is the following: If you go down to distances that are that small — 1.8 nanometers — you come to a regime where the distances become comparable to the size of the water molecules. That is a difficulty because the theoretical framework that I’m applying is called continuum hydrodynamics. So here, you treat the water as if it was a continuum — like Jello — but not as discrete particles. This works very well as long as you’re not looking at pores comparable to the size of the water molecule.
So that’s the difference, and I felt that since this is the first attempt to make a comparison, I should pick a case where I can be more confident that classical fluid dynamics would be at least roughly applicable. So that was the reason I picked this.
One could try to push it and see what you get for alpha-hemolysin, but I haven’t done those calculations yet. Those have to be done in a somewhat different way and would certainly be less reliable than these solid-state pore calculations.
Why has no one else applied theoretical classical hydrodynamics to study the interaction of DNA with a nanopore?
It’s the breadth of the field and the expectations of people. There is a large fluid-mechanics community, but very few of them look at problems involving these nanopores and phenomena on that scale. There was a paper from [David] Lubensky and [David] Nelson from Harvard [Driven polymer translocation through a narrow pore. Biophys J, October 1999, p. 1824-1838, Vol. 77, No. 4], which is probably the first theoretical analysis ever published in this nanopore area. They do take hydrodynamics into account, but they didn’t do a detailed calculation like I had done. They had made some estimates and those estimates were not quite right.
In my first paper [Electrophoresis of a polyelectrolyte through a nanopore. Phys Rev E Stat Nonlin Soft Matter Phys. 2006 Oct ;74 (4-1):041901 17155090], I explained some reasons why it wouldn’t be so. They considered hydrodynamics, and they said that the viscous force would be too small by a factor of 100. But they looked at various other things in that paper. There’s a small section where they consider hydrodynamics, but they didn’t think it would be the main effect.
I was the first who did a detailed calculation, and in my first Physical Review E paper I sort of explain why that initial estimate wasn’t quite right.
What’s next? You mentioned that nanopore sequencing isn’t the focus area of your research, so is this your last look at this particular problem?
No, I’m continuing to look at various problems related to nanopores. One thing, for example, that I would like to look at in more detail is the following: I know that the fluid resistance in the nanopore is the main effect, but I also know that it’s not the only effect because if it was the only effect, then the speed of the translocation of the DNA would be independent of DNA size. But that’s not quite the case. There is experimental evidence that indicates that velocity does depend on DNA length, but that dependence is rather weak.
There has been work on this other effect — why is it dependent on DNA length — and fluid viscosity is involved in that, but I think a precise enough analysis hasn’t been done, so that’s an area to look at. Once part of the DNA goes through the pore, there is another part that is still in solution and that is sort of being kicked around by Brownian motion, so there is some sort of viscous drag associated with this part of the DNA that remains outside.
The analogy would be: Imagine you have a swimming pool with a hole at the bottom and you’re trying to drag a long rope through it. So let’s say 50 feet of the rope is in the swimming pool, and it’s being moved around by the fluid while you pull from the other side. There is resistance within the pore, but there is also resistance from the fluid outside the pore, through which the DNA has to go. And these effects are more complicated because statistical effects are involved because the rope is not standing still — it’s been coiled in a complicated way, it’s being uncoiled and it’s moving around, and there are interesting effects there.
Have you had any feedback on this work from the nanopore sequencing community?
I’m in contact with David Deamer at [the University of California] Santa Cruz. He gives me encouragement and good feedback. He’s very supportive of theoretical work in this area. In fact, he’s the person who got me into it.
I presented this in the March meeting of the [American Physical Society], and I had contact there with a broad section of this community. A lot of these people were there. I’m new in the field, but I’m getting good feedback. They think it’s useful.
This must be a very interesting problem for somebody coming from a fluid dynamics background.
It is interesting. Fluid mechanics has been applied to so many things — from airplanes to submarines to whatnot — but never before has it been applied to such small-scale systems, and never before has there been the opportunity to actually experimentally verify this. The ability to control and measure things relating to one DNA molecule — this is a capability that has existed only over the last 10, 20 years. It wasn’t possible before this time. And the fact that you can do fluid mechanics calculations and compare to experiments — that’s just unprecedented and it’s really fascinating.