The Proceedings of the 1989 NanoCon Northwest regional nanotechnology conference, with K. Eric Drexler as Guest of Honor.

Return to NanoCon Top
Return to NanoCon Proceedings Table of Contents
Return to Jim's Molecular Nanotechology Home Page


B. Ned Seeman: Nucleic Acid Structural Engineering.


"NUCLEIC ACID STRUCTURAL ENGINEERING AS THE BASIS FOR NANO-SCALE DEVICES AND LATTICES" Nadrian C. Seeman, Department of Chemistry, New York University, New York, NY 10003 and Bruce H. Robinson, Department of Chemistry, University of Washington, Seattle, WA 98195.
There are two fundamental approaches to nano-scale construction: either honing or manipulating large objects to make small ones, or direct bottom-up fabrication. We have adopted the latter approach in attempting to construct specifically shaped materials. The molecule that we have used so far is synthetic DNA, whose sequence is selected in such a fashion that it can form branched structures, rather than linear duplex. These branched structures, termed junctions, can be ligated together in routine biotechnological applications involving molecular cloning. We have formed macrocycles that incorporate between 3 and 8 individual monomers when we oligomerize individual junctions with a pair of complementary cohesive ends. We have formed a particular cyclic tetramer when the ends were specifically designed to form only that macrocycle. Attempts are currently underway to synthesize a closed, 3-connected object. One of our major goals is the formation of predetermined periodic lattices from these materials. Modification of the DNA to attach specific groups is planned, as is the neutralization of the charged properties of the polyanion. Applications envisioned include nanomanipulators for the joint determination of structural and thermodynamic properties, and biochips that can function as components of computers.
J. LEWIS: Our next speaker is Ned Seeman of the Chemistry Dept. of New York University. Ned has a background in both computers and chemistry. He started doing molecular biology studying how DNA molecules recombine, but about 10 year ago had the insight that DNA had some other interesting properties that led him to the idea of using DNA to design nanostructures.
Note: In November of 1995, at the Fourth Foresight Conference on Nanotechnology, Nadrian Seeman was awarded the 1995 Feynman Prize in Nanotechnology, the second researcher to be awarded a Feynman Prize in Nanotechnology. See also the story in Foresight Update #23.

SEEMAN: I will discuss the work that we have been doing in the laboratory using DNA as a possible medium with which to construct nano-scale devices. We haven't gotten very far as yet, but we are moving in that direction. We really are not very far behind where we had expected to be at this time. However the path of our progress was different from what we had expected. Some of the technologies moved more slowly than we had expected, but other things moved more rapidly.

1. DNA as a Polymer

Let me introduce the nucleic acids that we work with. These are chemicals that we work with on the bulk scale - microliters of material. This slide shows the structure of an RNA molecule. There is nothing that prevents what I will say about DNA from applying also to RNA; it is just that RNA is chemically a good deal less stable because of its autocatalytic properties, and it is more difficult to make. Let me point out some of the features of the molecule.

DNA is a polymer - a copolymer of an ionized phosphate (i.e. it carries a negative charge) and a sugar (the nature of the sugar makes the difference between DNA and RNA) that connects every phosphate to its neighbor. The real business end of DNA is a series of flat molecules called bases. There are four different kinds: two big ones (purines) and two little ones (pyrimidines). The big ones are abbreviated A (for adenine) and G (for guanine) and the little ones are C (for cytosine) and T for Thymidine. In RNA, U (uracil) substitutes for T. The interesting property of DNA is that DNA strands pair such that a T on one strand is always opposite an A on the other, and a C is always opposite a G. The two strands form a helix, with the bases pairing inside the double helix; the polymer backbone of phosphate and sugar loop around the outside. This (Watson-Crick) base-pairing (A with T and G with C) is how DNA replicates itself. A further feature is that the two chains are antiparallel. This means that the two ends of each chain are not equivalent. One end is the 5` (pronounced 5-prime) end and the other the 3` end. The two strands pair such that the 5` end of one is opposite the 3` end of the other strand (rather than having the two 5` ends paired and the two 3` ends paired).

2. Branches in DNA

A major feature of DNA in nature is that, with the exception of some very transient events, the helix axis is linear in a topological sense: the axis is unbranched. You could follow a chromosome from one end to another, regardless of all of the coiling of the DNA molecule, without meeting a branch.

One thing that happens to DNA during the life cycle of the cell is DNA's own version of molecular sex, called recombination. Most of our work has been on recombination. When DNA recombines, it loses its linear character and becomes a branched molecule. Two double helices combine and cross-over such that the two double helices reciprocally exchange one strand at the point(s) of recombination. Thus each double helix will retain one of its original two strands, while the other strand will derive from the original strand for part of its length and from the recombinant partner for the other part. This crossing over occurs at branched structures that are called cruciforms.

I originally got interested in these structures after Bruce Robinson wondered into my office ten years ago and wanted to discuss the structures of these recombinant "cruciform" structures after he had done some molecular dynamic studies on them. The salient experimental fact about this kind of branched DNA molecule that makes it highly unstable is that these structures form with a two-fold symmetry; i.e, the sequence of the double helices in the two halves of the cruciform is the same. This allows the molecule to isomerize. It can isomerize in one of two ways: the helices on each side of the cruciform have a choice of which strand on the other side that they will pair with. The consequence of this is that over time the cruciform will resolve itself into two linear duplexes again, making it very difficult to study these structures.

Ten years ago we established an algorithm that allows the minimization of the symmetry of oligonucleotides (small segments of DNA chains) such that a molecule would form a branch that would be stable. In general, control over a system is inversely proportional to the amount of symmetry that is present. Thus we destroy symmetry to gain control over a chemical reaction.

3. Making and Analyzing DNA

Let me tell you about the two experimental devices that we use to analyze everything that we do. The first is a commercially available DNA synthesizer [Editor's Note: The slide showed a picture of a box a yard long and a couple feet deep and high sitting on a lab bench]. This machine uses solid phase synthesis (a variation on the Merrifield method for synthesizing proteins) to make DNA molecules of any sequence, simply by punching in the sequence desired. Three sequences at a time can be made, up to 125 to 150 nucleotides (one nucleotide is a single A,C,G, or T) long, adding one nucleotide to each sequence every 10 minutes. The scale of each synthesis is about 100 nanomoles. The advance in this technology has drastically speeded up our work compared to where we expected to be a decade ago. When we started, making a 16-mer was the state of the art. This machine is easy to use and is a standard feature of every molecular biology laboratory.

This other device is a gel electrophoresis apparatus, a cheap way to analyze the DNA that we have made. A long, thin gel of polyacrylamide between two glass plates, with a potential field of 500-1000 volts applied so that the negatively charged DNA molecules migrate toward the positive electrode, is used to fractionate the DNA molecules according to size and shape. The electrophoresis can be done either under "native" conditions, in which the strands of the double helix remain associated, or under "denaturing" conditions to dissociate the strands, which then form random coils.

We synthesized the four individual strands of the cruciform that we had designed and these associated by Watson-Crick base pairing. The next part of our work uses the same restriction enzymes that biotechnologists use. These enzymes recognize specific short sequences of DNA (typically palindromes of 6 base pairs) and cut it, usually leaving short "sticky" end sequences. These sticky ends can be mixed with the sticky ends of the same sequence generated from other DNA molecules, and will adhere. An additional enzyme is used to covalently join the adhering sticky ends. Thus one molecule is formed from pieces of two molecules. Typically one of these pieces is a portion of DNA that contains the signals necessary for the DNA to grow and express itself well in bacteria, and the other codes for a valuable protein produced in a higher organism, but in very small quantities. The interest of the biotechnologist is to place the DNA into a bacterium, have it make a lot of this valuable protein, and thus "excrete money." This process always yields an unbranched DNA molecule, either a long linear piece or a circle.

4. Building in 3D with DNA

Our interest is to make a branched DNA structure, which we call a junction. With the proper sticky ends, these branches can be put together to form networks. When I came up with this concept, I was very excited because I grew up as an X-ray crystallographer, which meant that I studied the 3-dimensional shapes of molecules in crystals. I found that I was not good at growing crystals, which was a real difficulty in the way of a career in X-ray crystallography! What excited me was that such a network is a periodic array of molecules. A crystal is a parallel array of molecules in 3 dimensions. This computer graphics representation is of a 3-arm junction, instead of the 4-arm junctions found in biological recombination, and shows how it could form a 2-dimensional network where the individual junctions are separated by an even number of half-turns of the DNA helix. Because DNA is a helix, we are not confined to a plane, but can make 3-dimensional objects. I expect that in 1991 or 1992 we will be synthesizing junctions based on a truncated octahedron.

We have chosen to build with DNA because it is easy to synthesize and work with. To make a square, for example, we first have to decide what the building blocks will be. In this case, we chose 3-arm junctions (4, 5 or 6 are other possibilities) so that there will be 4 sets of cohesive ends (designated A, B, C, and D). Choosing different identities for these ends will give us very different results. If we let all 4 be equal, we will not be able to control how they assemble because high symmetry means little control.

Once we have chosen the connectivity, we have to choose the separation from junction to junction. Since DNA is a helix we want to have either even or odd numbers of half turns of the helix between junctions. We have written a user-friendly program to help predict what structures a particular choice of sequence will yield.

Let me tell you now about some of our experimental results. Here we have taken three strands to make a junction. These cost ~$5000 per strand in 1983, and they had to be specially ordered from a company that synthesized them. Today we could make them in our own laboratory for ~ $120. We control whether or not anything happens at any particular end of a strand by whether or not we place a phosphate group on that end. We can thus assure that no interaction happens at the blunt end, but we have complementary (cohesive) ends on the other end that have a minimal symmetry to give us a little more control. When we mix these strands together "in a pot" the cohesive ends stick to one another. We use the enzyme ligase to close the gaps at the ends and make a covalent bond. We get a large number of linear molecules and a certain number of cyclic molecules.
AUDIENCE: (from John Cramer): Is it not possible to have a closure of just two?

ANSWER: No. That's a good point. DNA is a stiff molecule, which is one of the good things about DNA for use in molecular construction. Its persistence length is ~600Å. The distances we're discussing here are ~60-70Å, one tenth of the persistence length. What persistence length means is that two base pairs would have to be separated by at least 600Å before the orientation of one would have no correlation with the orientation of the other.

AUDIENCE: In other words, that's about the length that it would take to bend this one molecule into a complete circle without stress?

ANSWER: No, that's closer to making a right angle. Actually, you can make circles of that size without too much stress.

J. CRAMER: Do these junctions bend?

ANSWER: We have done this experiment with three different 3-arm junctions, and each time we get a whole series of structures, so that the junction is obviously not highly rigid. We have done the same experiment with a 4-arm junction and get the same answer - we get a lot of linears and also a lot of different closed structures, indicating again that there is sloppiness. The angles formed by the junctions are probably different in different structures. Furthermore, we don't know the 3-dimensional structures of any of these.
Here are some pictures of what some of these structures should look like in a gravitational field. The first closed cycle product of the 4-armed junction ligation is a triangle of ~70Å (7.0 nm). The second product is a square.

We did a second experiment in which the junctions were separated by an odd number of half-turns instead of an even number, so that it is a turn and a half from junction to junction. The product there starts with the square instead of the triangle. The only way that I could model this is by putting a differently-shaped junction structure at the different corners. All of this put together gives us a paradigm for the junction, a marshmallow impaled by three pieces of spiral pasta!

The differences between the two junctions can be seen most clearly by looking at the Y-helix axis: in one case they meet at the corner; in the other case they do not. The way in which the corner is turned is very different in the two cases.
AUDIENCE: Have you tried making a rectangular structure in which you have one and a half turns on one side and two turns on the other?

N. SEEMAN: No. That would be very difficult because the twist constrains what we can make.

When we have a reasonably stiff rod-like element forming the edge of a polyhedron, and we have a reasonably floppy vertex, nature and Buckminster Fuller decided the best solution was to build primarily using triangles. For example, a large number of viruses use icosahedral structural motifs.

To make periodic networks, another element that Fuller discussed called the octahedral strut is useful. It is commonly used in scaffolding, for example, to hold up the roof in airports. Two edge-sharing octahedra can be combined in a strut to eliminate the wobble about the edge. A materials scientist would notice that in this structure, the relative positions of all the vertices form a face-centered cubic lattice. The good news is that these things are probably tough enough to build tanks out of; the bad news is that each of these individual vertices is 12-connected as opposed to the 3- or 4-connected sort of things that we have been talking about.

Right now we have trouble building things much over 4-connected. We built 5 and 6 arm junctions, but they are not stable with 8 base pairs per arm as are 3 and 4 arm junctions. We have done theoretical studies that indicate why that is so. If we made the arms longer, which makes more flop, we might be able to make as high as 12 arms.

Another bit of good news about this system is that we really do have control over what we synthesize by means of particular cohesive ends. I will now describe an experiment in which we made the extra effort and made A, B, C and D all different, rather than making them all equal, as in the experiments that I just finished describing. We tried to make a square. In this case, we made a 3-arm junction in which two of the strands were joined by a little loop at the end. We separated them by a turn and a half of DNA. The product should give two circles of DNA interlocked to each other 6 times. Each of these different 4 ligation sites is different. We incorporated two restriction sites on an exocyclic arm so that the synthesized product can be analyzed by cutting with a restriction enzyme and then sequencing the DNA to determine what has been made. We succeeded in making this particular square, isolating it from failure products by treating with enzymes that digest structures that have not closed (i.e., have free ends).

A current project is to synthesize a cube, using the same technology as I have discussed. A few points to notice about this cube: (1) It is ~7 nm on a side. (2) If we made a lattice of this, it would have a big hole in the middle of each cube so that I could solve my problem of crystallizing proteins by putting proteins in there. Talking with Eric last night, he's had similar ideas. (3) We could also put different proteins in different cubes to form assemblages of proteins to collaborate on some task. Proteins in cells are often found as clusters (multi-enzyme complexes) to achieve various purposes. Instead of waiting for the right enzyme needed for the next step of a complex synthesis to diffuse to where it is needed, you put all of them together so that the first product is immediately transferred to the second enzyme, etc. With the technologies that we know about today, we could start to try these experiments, which are sort of like making DNA zeolites.

Someone said last night that a lot of the human element might go out of the world if nanotechnology comes in. I'd just like to point out some of the humans involved in doing this work. Jung-Huei Chen made the DNA square, and Neville Kallenbach has been a long-time collaborator.

5. A DNA Computer

Let me now talk about the work we published several years ago ["The design of a biochip: a self-assembling molecular-scale memory device" BH Robinson, NC Seeman. 1987. Protein Engineering 1:295-300] on a rough design for a molecular computing device that we hoped would operate at electronic speeds using the DNA junction as scaffold. We not concerned here with any properties of DNA except its mechanical properties and our ability to control the structures it forms. Bruce will talk about the details of the conducting polymers that we want to assemble on the scaffolding.
This particular lattice happens to have 6-arm junctions, which are not yet as stable as we would like them to be. A structure with a volume of 104 cubic nanometers constitutes one bit.

6. A DNA Nano-Manipulator

Besides making immobile structures, we would also like some components that actually move. I would thus like to discuss some aspects of nanomanipulation, specifically having controllable isomerizations of molecules that we can deal with today.

Going back to junctions of DNA formed during recombination, there is a preferred way in which the DNA junction straightens itself out. The strands in the junction cross over in a particular way [Editor's Note: two strands (of the two double helices) are unperturbed helices while two strands cross over]. The two strands that cross over can assume either of two conformations, [Editor's Note: i.e., strands 1 and 2 can cross over while 3 and 4 are unperturbed, or vice versa] as a function of the particular base sequence that flanks a given junction. We have recently found ways that we can control the isomerization between the two conformations. One conformation is slightly favored except in the presence of stress, although the equilibrium is not far from an equal mixture. Another aspect of this isomerization has to do with branch migration, which involves the instability of the junction so that the cross-over point can move down the helix. Others have shown (in systems called supercoiled DNA that involve large increments of torque) that a small amount of torque on one pair of these arms can control that isomerization.

Another transformation of DNA turns DNA into a funny structure almost the mirror image of normal DNA. We have a very crude concept of a nanomanipulator, in which two objects are initially in contact. A DNA molecule is attached to one object. This DNA segment has an appropriate sequence that would favor a structural transformation so that instead of being the normal double helix geometry, it assumes the other conformation. Thus if we make this DNA segment of the particular sequence that favors this transformation, and then put it in a different chemical environment, it will change conformation, moving the one object out of contact with the other.

7. Engineering with Weak Bonds

I'd like to emphasize something about the interactions of these segments of DNA and of the slide that Eric showed last night of a protein bound to a stretch of DNA. This DNA-protein complex involved several thousand atoms and several thousand bonds. Moving the big protein off the DNA would take about 20 kcal/mol. I think that it is important to emphasize that that is a weak interaction. To break any one of the thousands of covalent bonds in the DNA or protein would take much more energy than to move the whole protein "blob" off the DNA. It is important when thinking about what can be made and what can't be made to realize that there are two stages of what Eric is talking about as nanotechnology. In one case, he is talking about moving molecules away from or toward other molecules where you are making weak intermolecular contacts. Whereas making particular molecules (like the crankshaft) involves an order of magnitude more energy. I don't see this second phase as coming during the remainder of my scientific career (which I hope will go on another third of a century), while the first phase should be feasible soon because we are talking about lower energies.

E. DREXLER: I agree entirely.

N. SEEMAN: What direction am I stumbling in now? A Dutch biologist not that long (1872) after Wöhler had synthesized urea and founded organic chemistry decided that he was going to try some synthetic morphology but didn't get very far. To show how far back these ideas go, the floor of a temple in a museum in East Berlin dating from the first century AD shows helical bars, making lattices... so these traditions go back a long way.

In addition to this talk, Dr. Seeman has also contributed a paper to this conference

Next page of the NanoCon Proceedings
Previous page of the NanoCon Proceedings
Return to NanoCon Proceedings Table of Contents
Return to Jim's Molecular Nanotechology Home Page
Return to Web Site Design and Authoring Services

Comments and correspondence? Send email to Jim Lewis:
NanoCon Proceedings © Copyright 1989, by NANOCON
This page is part of Jim's Molecular Nanotechnology Web, copyright ©1996 James B. Lewis Enterprises. All rights reserved.
Last updated 24June96.
The URL of this document is: