The design, synthesis, and proof of new DNA motifs having novel geometries is a central endeavor in DNA nanotechnology. It typically involves much computer design, sometimes requires extensive DNA manipulation, and sometimes requires extensive characterization of the new motif. Two previously successful methods for designing DNA motifs are (1) Seeman's use of parallel DNA helices joined by antiparallel DNA crossovers to create a variety of small multi-stranded motifs and (2) Shih's use of paranemic crossover interactions to fold a long single strand of DNA into a large three-dimensional octahedron.

Here, we present a new method for folding single-stranded DNA into arbitrary two-dimensional shapes. The basic method draws on Seeman's technique of engineering DNA nanostructures using parallel double helices and antiparallel double crossovers. We have developed a computer program that allows the quick design of DNA motifs having arbitrary shapes. We experimentally demonstrate the folding of approximately seven kilobase single-stranded DNAs into several shapes --- a square, a triangle and a five-pointed star. The resulting DNA structures are approximately ten times larger (in linear dimension) than the double crossovers often used in DNA nanotechnology --- each with widths of approximately 100 nanometers. Proof of the structures is by correspondence of high resolution AFM imaging with the structural design. These structures represent the largest molecular structures ever constructed. Every atom occurs in a unique and specific position in a structure whose mass is 5 megadaltons, approximately twice the mass of a ribosome. The size of the structures generated appears limited only by the length of highly pure single-stranded DNA that may be practically used, currently limiting the size of the shapes created to approximately 10000 square nanometers. The resolution with which the shape is rendered is 6 nanometers in one direction and 3 nanometers in the other. It does not appear that this resolution may be easily improved.

Our work adds new understanding to the process of designing and constructing new motifs for DNA nanotechnology. First, it helps clarify when and how much computer design must be performed on DNA sequences for nanotechnology. Many of the sequences used in our DNA nanoconstruction fail sequence design constraints previously used (and held to be important) in the design of DNA nanostructures, and yet our structures fold correctly anyway. This does not mean that in general DNA design may be ignored, but in the specific instance or our designs, certain features of the design (i.e. intramolecular folding) weaken the need to apply some stringent constraints on the design. Second, we show that the space of sequences that may be assigned to antiparallel DNA crossovers is significantly larger than previously reported. Our designs employ hundreds of previously unexplored sequences for immobilized DNA crossovers. Third, a number of helper strands are employed to fold our designs. In a normal DNA nanoconstruction, these DNA strands would be purified and subjected to a combinatorially exhaustive analysis of the possible complexes that could be formed. We show that the purification of these strands and subsequent combinatorial characterization is unnecessary for this method of design.