Dr. Tinkertoy

DNA is more than the storehouse of life's secrets, it's also a marvelous construction toy.

By Robert Pool|Saturday, February 01, 1997

Hanging from the ceiling of Ned Seeman’s office, suspended by fishing wire, is a model of . . . um, something. Its half dozen coils of plastic tubing are paired up in a way that resembles the double helix of a DNA molecule, but DNA was never like this. Instead of forming one simple, linear strand, the plastic coils of this thing snake around, separating and coming together again to trace out a complex, cube-shaped framework. Across the room, on a table between two windows, sits an intricate assembly of sticks and balls. It’s the sort of thing a young boy might make with a set of Tinkertoys--assuming the set was large enough and the boy patient enough--but in the office of a chemistry professor at New York University, it is somewhat disconcertingly playful.

After the office, Seeman’s laboratory is a bit of a disappointment--it holds nothing you can’t see in thousands of molecular biology labs around the world. Here is a DNA synthesizer, there a gel electrophoresis unit, and in the corner sits a series of heating blocks for keeping reactions at set temperatures--all standard tools for exploring DNA and the genes that it holds. Of course, there’s nothing wrong with that-- gene mapping and genetic engineering are very hot fields--but Seeman’s office seemed to hint that something more unusual might be going on here.

And, actually, it is. As his equipment indicates, Seeman does indeed work with DNA, but it’s not to study or to manipulate genes. Instead he uses DNA as a raw material for microscopic construction projects. By using chemical reactions to assemble short bits of DNA, he builds frameworks so small that 10 million of them lined up end to end wouldn’t stretch across your palm. The cube made of coiled plastic tubing is a representation of one of his assemblies, but a better way to visualize them is to ignore for a moment the helical structure of the DNA and think of each length of DNA as just a short stick. Then Seeman’s research amounts to building with molecular Tinkertoys.

His program has immediate appeal for anyone who has ever dumped the colored sticks of a Tinkertoy set onto the floor and wondered what to make. A bridge? A rocket ship? A molecular cage smaller than a virus? And though the idea of using DNA--the molecule of life--as one’s construction material lends a whimsical charm to the whole endeavor, it also marks it as visionary.

For decades, scientists have dreamed of designing and building objects not much larger than individual atoms. And when that day comes, futurists have promised, the list of ensuing wonders will be endless: microscopic computers that store information in a single molecule, tiny machines that reproduce themselves by the billions, miniature medical robots that roam a patient’s body to destroy viruses. Seeman’s simple DNA assemblies are certainly far from any of these futuristic applications, but they are perhaps the most important step to date toward this distant goal.

Seeman admits that his choice of building material seems strange. In life DNA looks like a mess of spaghetti, the individual strands twisting this way and that, often curving back on themselves to form loops and figure eights. Not to worry, Seeman says. This pasta is al dente, not the mushy, overcooked stuff, and short pieces of it are relatively stiff. Since the bits he uses to build with are typically just three and a half times as long as they are wide--about the same proportions as a quarter- inch snippet of spaghetti--he avoids the floppiness problem almost completely.

At the same time, DNA has a property that makes it a natural choice for construction projects, particularly those in which the pieces are too small to handle or even see: it assembles itself. Combine the right sorts of DNA strands in a test tube, under the right conditions, and they will automatically latch onto one another. Indeed, millions or billions of these assemblies will occur simultaneously, depending on how much DNA is placed in a test tube. This facility for self-assembly is at the heart of DNA’s ability to carry the genetic code, but no one before Seeman came along had thought to put it to work in construction.

As every high school biology student knows, DNA occurs naturally as a double helix--two long molecules that have zipped together and twisted into a spiral staircase. Under certain conditions--such as when a cell is dividing and needs to copy its DNA--the double helix will unzip into two strands, but these single strands are unfulfilled and will, given the chance, spontaneously link with each other again. The explanation for this mutual attraction can be found in the makeup of the strands. Each single strand of DNA consists of a series of units called bases, of which there are four types, labeled A, T, G, and C (for adenine, thymine, guanine, and cytosine). A and T are chemically attracted to each other, as are G and C.

Because of this attraction, a strand of DNA consisting of, say, agga, will grab onto its complementary strand, tcct. On the other hand, two strands that are not complementary will not bind to each other. None of this is news, of course--the basics have been known since the Nobel Prize- winning work of James Watson and Francis Crick in the early 1950s--but the way that Seeman applies it is new. Traditionally, scientists have thought of the attraction between complementary strands of DNA purely as a way of creating long double strands of the sort that appear naturally in the cell. Seeman realized that it could be used to make much more interesting objects.

It was the sort of realization that demanded a little distance, that needed someone who could look at DNA and see more than the genetic code. Seeman fit the bill. When he first thought of transforming DNA into molecular Tinkertoys, he was in an area of science far removed from genes. He was looking for a way to make crystals.

In the late 1970s, Seeman was working as a crystallographer at the State University of New York in Albany. He would take small biological molecules, such as short pieces of DNA, and try to coax them into forming crystals. The idea was to aim an X-ray beam at the crystallized molecules and, by analyzing the pattern of radiation that came out the other side, to study their structure. Such X-ray crystallography determines the structure of a molecule--exactly where each atom sits in relation to other atoms-- which in turn is crucial to understanding how a molecule works. Drug companies, for instance, spend a small fortune on such techniques to learn the structures of proteins and other molecules in the body so that they can either mimic their functions or interfere with them.

Seeman specialized in rna, a molecule closely related to DNA, and had succeeded in pinning down the structure of several rna fragments-- short stretches of the much longer rna molecule. But then he hit a dry spell. I couldn’t grow any of the crystals I was interested in, he recalls. It’s one of the trickiest, most difficult jobs in science to persuade biological molecules--which tend to be irregularly shaped and floppy--to line up neatly and stiffly in a crystalline array, and it’s not unusual for a research team to spend years crystallizing a single complex molecule. Sometimes a researcher will choose a particularly difficult molecule and never get it into crystalline form. Such was the case with Seeman, and it left him in an uncomfortable position. I had to do something, he recalls. A crystallographer without crystals is in trouble.

So when a colleague, Bruce Robinson, asked Seeman for help on a problem that had nothing to do with crystallography, Seeman said yes. Robinson was interested in Holliday junctions, four-way intersections that can form when two strands of DNA come together, and he wanted Seeman to build a physical model of a Holliday junction. That investigation, in early 1979, set Seeman on the path toward building with DNA.

To picture a holliday junction, start with four single strands of DNA, each of them, say, 20 bases in length. Now zip two of them up, but only halfway, so that there is a pair of single strands, each ten bases long, hanging free. Repeat with the remaining two strands. You now have two T-shaped assemblies, the base of each T being a double strand and the arms single strands. Take one arm from the first T and one from the second, and zip them together; repeat with the remaining two arms. You are left with a single molecule of DNA that has four branches, each of them a double strand ten bases long. This joining of four strands is the Holliday junction.

It is well known among those who study DNA that the molecule of life takes on such atypical shapes during various processes in the cells. It may look like pieces of spaghetti most of the time, but occasionally DNA will rearrange itself into the Holliday junction or some more exotic configuration. But before Seeman, no one saw the opportunity that these shapes held.

As he played around with his models of Holliday junctions, Seeman found he could make them behave quite differently by changing the sequence of bases that made up the strands of DNA. A typical Holliday junction, made from two identical pairs of strands, is mobile--the intersections tend to move around because two arms will zip up while the other two unzip. If an intersection moves far enough, the Holliday junction separates into two double strands. Seeman realized, however, that he could design the pairing scheme of the four single strands so that they would assemble themselves into a Holliday junction that was completely locked in place.

Seeman soon recognized that there was no reason to limit himself to junctions with four arms. With a little cleverness, he thought, he could assemble intersections with as many as eight branches. Nature might never have designed a three-armed or a six-armed DNA junction, but he could.

But . . . so what? Figuring out how to build these junctions was intellectually challenging--and fun--but would they be good for anything? Seeman didn’t know, but he thought he ought to figure it out.

Then, Seeman recalls, he did what he often did when he needed to ponder. I went to the campus pub and I had a beer. And as he was sitting and sipping, the proverbial lightbulb went on. I was thinking about six- arm junctions, he says, and Escher’s ‘Depth’ came to mind. The famous Escher drawing shows a three-dimensional array of stylized flying fish, each with a head and a tail and four prominent fins pointing up, down, right, and left. The fish reminded Seeman of the six-arm junctions, with one arm corresponding to the head, one to the tail, and the other four to the fins, so he imagined lining up his junctions the same way Escher had lined up the fish. Each junction would have six others neighboring it, above and below, in front and behind, and to the right and left. Unlike Escher, who left space between the fish, he would link the junctions arm to arm, and--voilà--he would have a large framework resembling the steel skeleton of a skyscraper, but much smaller and made out of DNA.

That idea was quite exciting to a crystallographer frustrated with the difficulty of getting molecules to line up in crystalline order. If he could build such a DNA framework, Seeman thought, he could use it as a scaffolding to hold molecules in place and so crystallize almost any molecule he wished.

Seeman has been pursuing that heady vision for 16 years now. I think that ultimately, he says, we’re going to get where we’re headed.

More important than the specific goal Seeman set for himself that day in the campus pub was the shift in thinking that it represented. Previously he had conceived the various junctions--four-armed, six-armed, or whatever--merely as isolated objects, interesting to design or study but having no larger significance. Suddenly, in a brilliant insight, he saw these bits of DNA as building blocks for making bigger, more complicated structures. The bad news was that his insight proved to be several years ahead of its time.

In 1980 tools for working with DNA were primitive. Take, for example, the synthesis of DNA: To build a single six-arm junction, Seeman would need six different strands, each at least 16 bases long; 20 bases would be better. But in 1981, the year after his pub-based eureka moment, even a topnotch chemist using state-of-the-art techniques would need about three months to create only one 12-base strand of DNA. Six strands with 16 bases each, just to build a DNA Tinkertoy--that was out of the question.

Fortunately for Seeman, the necessary technological breakthroughs were just around the corner. Beginning in the early 1980s chemists developed--and then automated--methods for quickly producing longer strands of DNA, and today commercially available machines can turn out custom-made strands more than 100 bases long. When Seeman needs, say, a 100-base strand for building a DNA cube, he punches in the specifications on a computer attached to his synthesis machine, and within 18 to 20 hours he has a test tube full of the stuff.

More generally, Seeman notes, his program has benefited from a number of dramatic advances in molecular biology over the past decade and a half. Researchers chasing information about DNA’s role in the cell have developed techniques not just for synthesizing strands of DNA but for cutting, bonding, purifying, and analyzing them. Seeman has adapted those techniques for building with DNA.

For much of the 1980s, Seeman says, he was simply laying the groundwork to pursue his ambitious program. It was five years before we got a DNA synthesis machine, he recalls. And he had to train himself in a whole new area of science. He began testing his ideas by building simple, two-dimensional structures, such as four-armed junctions and rectangles. Even those were not easy. I didn’t know how to do any of the molecular biology techniques, he says. There are a million little things that you have to learn not to do wrong. Performing the reactions at a temperature that’s too high or too low, adding too much or too little of a particular enzyme, having the wrong concentration in a solution--any of these things could cause a trial to fail.

Eventually, however, Seeman got it right, and by 1988 he decided he had learned enough to make the jump into three dimensions. He would build a cube. This step was the defining work of his DNA construction program. It was the first time Seeman had pieced together something significantly more complicated than is built by nature. Learning how to do it led to the development of a number of new DNA construction techniques that differentiate Seeman’s lab from the typical DNA laboratory, where most work involves little more than attaching long double strands. For instance, the computer software available for modeling DNA is not appropriate for Seeman’s purposes, so he doesn’t use any. What’s good enough for molecular biology is often not good enough for the synthetic chemistry that we’re doing, Seeman says. Instead, he first works out all of his structures on a DNA Erector set made of tubes and colored plugs.

The way seeman went about building his cube was quite different from how an eight-year-old would construct one with Tinkertoys. There’s no easy way to take 12 equal-length pieces of DNA and join them to form a cube. The problem is the corners--the world of DNA has nothing like the junctions in a Tinkertoy set, those round pieces with holes. Theoretically, it would have been possible to build the eight corners of the cube, each one a three-armed junction, and then join them together, but Seeman never considered this option. It would demand eight separate joining--or ligation--steps, and those steps, Seeman says, are relatively inefficient. Instead he chose a method completely foreign to the Tinkertoy mind-set but quite natural to DNA.

It begins with two 80-base strands of DNA, each of which is formed into a loop by sealing the ends together. These two pieces will eventually become the left and the right sides of the cube. To each loop Seeman adds four more DNA strands, each about 40 bases long, that zip up partially with the larger piece, creating a pair of half-cubes--squares with a leg sticking out from each corner. Finally, he joins the two half cubes by linking the legs end to end. The resulting cube has 12 edges, each a double strand of DNA 20 bases long, and eight corners that are three- armed junctions.

Three years after building the cube, Seeman used a more complicated procedure to construct a truncated octahedron--an object whose six apexes have been cut off so that it has six square sides and eight hexagonal sides. This shape was so complicated and involved so many steps that Seeman had to anchor parts of the structure to a Teflon support to keep them from drifting around in solution and attaching where he didn’t want them to. This success gave him the confidence that he could make still more complicated objects if he wished. We believe we know how to make almost any topological object out of DNA, he says. But he has one final big problem to solve before his DNA Tinkertoys are ready for prime time.

Although the short stretches of DNA he uses in his constructions are stiff, the junctions where they come together are not. They’re floppy. It’s as if his Tinkertoys have marshmallows in place of the usual wooden junction pieces, Seeman says, so the corners of his cube, for instance, do not stay at perfect right angles. This allows the edges to move around and the cube to collapse, as though somebody let the air out.

Seeman had actually learned that his junctions were floppy before he built the cube, but he decided to forge ahead anyway just to prove that he could build in three dimensions. Then he began to look for a way to make his structures more rigid. Now he thinks he’s got it.

The solution comes in two parts. First, Seeman has replaced rectangles with triangles as his basic building blocks. If the corners on a rectangle aren’t rigid, the sides can move relative to one another. But if the corners on a triangle aren’t rigid, it doesn’t matter--as long as the sides of the triangle are rigid, nothing will move. Of course, when two triangles are joined, they can still move relative to each other unless their intersection is fixed somehow, and Seeman thinks he knows how to do this too. He has developed a rigid type of junction called a double- crossover molecule, consisting of two Holliday junctions linked very closely to each other. These double-crossover molecules have allowed Seeman to connect more than a dozen short lengths of DNA in a row and keep them from drooping back on themselves and forming circles. Combining triangular construction with straight lines made from double-crossover molecules should allow him to build two-dimensional structures that are not floppy, Seeman believes.

In three dimensions, Seeman has sketched out a number of rigid frameworks that could be assembled out of triangles. It’s pure Buckminster Fuller, he says, referring to the engineer whose famous geodesic dome depended on triangles for its structural strength. Like the geodesic dome, three-dimensional frameworks made from double-crossover molecules could be remarkably strong and rigid.

For now, Seeman is working on assembling the double crossover molecules into a triangular two-dimensional framework, something he hopes to have done within two years, although he cautions, You can’t put a time frame on anything in this business. After that, he’ll start working in three dimensions and eventually, he hopes, develop a way to use his constructions to crystallize biological molecules. With luck, he could be finished in time for the thirtieth anniversary of that day in the campus pub. Which would be good--30 years is a long time to be playing with Tinkertoys.

Comment on this article