Research Projects

Environmentally Adaptive Self-Assembly

Lithography has given us the ability to take an exact spatial arrangement of circuits, etc., and produce a result that exactly fits that specification. Likewise, scaffolded DNA-origami has allowed the production of beautiful nanoscale shapes to order: smiling faces, dogs, maps of the Western hemisphere and even giant "buckeyballs". But what do you do if what you know what thing you want to make should do -- maybe act as a wire between two devices -- but not exactly the size and shape it needs to be to do that? In biology, the specification for complex objects is often imprecise. Structures can either be somewhat random, or more interestingly, details of the environment where the organism grows determine exact feature shapes. Trees, for example, grow branches towards light.

I am interested in understanding how we can program environmentally adaptive self-assembly processes. One well-known example of this type of assembly in biology is the mitotic spindle (Image from Anne Knowlton, Stuken/Burke labs, University of Virginia):


Here, green fluorescing microtubules connect two chromosome centers, ensuring that during cell division each end of a cell ends up with exactly one of each of the organism's chromosomes. The spindle structure must connect these two centers, but its exactly spatial shape is otherwise different from cell to cell.

With Jan Liphardt (UC Berkeley) and with assistance from the biological nanostructures group at Lawrence Berkeley labs, I am working on using DNA tile nanotubes to connect two chemically marked locations in 3-dimensional solution, a start and a finish. The design of the system is such that tubes should span the start and finish even when the orientation of these points varies in distance or orientation.


Above, two views of the schematic arrangement of DNA within a nanotube, the arrangement of the tile monomers (each of which is a two-helix structure given its own color in the schematic) is somewhat like the arrangement of tubulin monomers in microtubules.


Above, an optical microscopy image of fluorescently labelled DNA nanotubes in solution. The field of view in this image is about 100 microns.

Directing Crystal Growth Pathways

The energy of attachment of a monomer to a growing crystal is the difference between the enthalpy gain due to the formation of new bonds at the interface of the monomer and the crystal and the entropy change of the system due to the attachment of the monomer.

DAO-E DNA tiles crystallize via hybridization of their "sticky" ends. Monomer attachments in which more than one sticky end bond is formed simultaneously are more energetically favorable than those that form only one attachment.

Above, a sample "tile set" consisting 6 monomer types (tiles). Each tile is shown as a ribbon diagram on the left and a rectangle and claw diagram to the right. The colors of the rectangles represent the identity of the monomer type which the claw colors show the affinity of the sticky ends --- interlocking ends of like colors hybridize, while other ends do not have significant interaction energies.

Because we can design these monomers and their attachment energies to other monomers (much like designing a set of jigsaw puzzle pieces), we can equivalently design crystal growth pathways. In our work we have shown that we can design sets of monomers for which we control the nucleation rate as well as control the rate of facet nucleation.

Above, a designed growth pathway. While many reactions are possible, the illustrated pathway is particularly likely given the possible reactions between monomers. We've engineered, here, the critical size of the crystals that are made from the set of monomers.

Additionally, we can design tile sets that reduce the rate of defects during growth. Further, we can increase this robustness (manuscript in preparation) by adding more and more types of monomers to a reaction.

We've also shown that we can design seeds for crystals such crystals grow only off seeds, and the DNA composition of the seed determines the arrangement of monomers in the crystal that does grow.

Above are 2 examples where we use a DNA origami seed and a set of tiles that can propagate any arrangement of stripes. Each experiment uses the same set of monomers but the presence of differently programmed seeds produces the differences in the ribbons in the two cases. These images are atomic force microscopy images taken under fluid, scale bar is 50 nanometers, and some monomers have additional DNA to produce height contrast (brighter = higher off the surface) that can be seen in the pictures. I am currently working with Constantine Evans and Erik Winfree to investigate new ways to produce crystals with complex patterns with high yield, and to produce crystals of defined size: the set of monomers defines a process of growth which eventually terminates at a fixed size.

Self-Replicating Materials

What is a minimal life form like? In the 1990's a panel gathered by NASA concluded that the essential attribute of life was the ability to undergo self-sustained replication and evolution. Erik Winfree and I have postulated that DNA tile crystals could be used to make a very simple experimental system satisfying this criterion.

The basic ideas come from the work of A. Graham Cairns-Smith, who thought that crystals could carry information in their arrangements of monomer types and/or defects in such a way that the information was propagated during crystal growth. If physical forces in the environment (such as those in non-uniform flows) broke a crystal into pieces, each piece could then grow and propagate the same information, thereby replicating it.

I am currently working with Bernard Yurke and Erik Winfree on experimentally demonstrating the replication of DNA tile crystals bearing particular sequences. We use the system of templated crystal growth to produce elongational fluid flows to stimulate their breakage into pieces.

The Computational Complexity of Crystal Evolution

If we could replicate DNA tile crystals bearing particular patterns of monomers, and occasional mistakes occurred during the propagation of these patterns, we would see Darwinian evolution. That is, crystals with arrangements that could enable crystals to grow and replicate faster would be selected for.

But so what? What does a set of arrangements of DNA tiles do? Chemically, we might make functional sequences by attaching tiles to Is any selection process ever going to produce? From a computer scientist's perspective, the surprising answer is that the evolution processes possible with DNA tile crystals are as algorithmically complex as any we may ever hope to see! There are sets of monomers for which the arrangement propagated also runs a program of arbitrary complexity. Additionally, evolution that produces complexity is a pretty common phenomenon among all possible sets of DNA tile monomer, and these sets can be very small, small enough to make in the laboratory.

I am currently working on characterizing this frequency and investigating whether it is easy enough that natural crystals, if they were to evolve, might ever also evolve complex particle arrangements as a result of repeated growth and breakage.