16.3 N&v 283 Mh New

283 As civilization has developed over the past 10,000 years, humankind has learned how to build larger and larger structures; over the past two decades, we have begun to learn how to build smaller and smaller structures. On page 297 of this issue, Paul Rothemund presents a material step forward in this second arena: he describes a stunningly simple and versatile approach to the fabrication, by self-assembly, of two-dimensional DNA nanostructures of arbitrary shape. DNA has emerged over recent years as the molecule of choice for nanodesigners. There are two reasons for this. First, in the 50 years since the discovery of the DNA double helix, a detailed understanding of the energetics of its formation has developed. This allows one to predict, with reasonable success, the shapes into which a DNA molecule of a given sequence will fold in solution. Second, the advent of automated chemistry for the rapid synthesis of DNA molecules has made it possible to easily obtain DNA molecules of any desired sequence, of lengths up to 100 nucleotides or so. (A nucleotide is the monomer unit of DNA.) Longer molecules may also be obtained using biological methods such as the polymerase chain reaction or now-standard molecular cloning approaches. The molecular engineer is thus armed with two of the basic elements needed to build structures of interest: the materials for building and an understanding of their properties. The missing ingredient has been a versatile design strategy. The best-developed model for DNA design is the ‘tile model’ developed in Ned Seeman’s laboratory. This uses as its basic buildingblock DNA in two-dimensional, rectangular shapes. These tiles are designed to include crossover points between DNA strands, so imparting stiffness to the structure. Free single-stranded regions (‘sticky ends’ in the parlance of the molecular biologist) extrude from each corner of the tile and permit tiles to selfassemble into larger, two-dimensional sheets. This basic concept has been adapted and extended many times, enabling demonstrations of, for instance, the templated selfassembly of protein arrays, and the fabrication of the fractal patterns known as Sierpinski triangles. A second design theme was introduced by William Shih and colleagues in 2004. They showed that a single strand of DNA, 1,669 nucleotides long, could be driven to selfassemble into a nanoscale octahedron by the addition of five short DNA strands complementary to selected regions of the original strand. Again, structural rigidity was obtained by means of crossover points. This work was notable in two respects. First, it permitted assembly of a three-dimensional structure, rather than the two-dimensional sheets that had been the primary focus of most previous work. Second, it introduced the concept of using short DNA strands to direct the folding of a longer DNA strand. Rothemund builds upon both these models, and expands their generality to permit the fabrication in DNA of any two-dimensional shape. Just as in the approach described by Shih et al., he directs the folding of a long single-stranded DNA molecule, in this case the genome of the widely used cloning vector M13 mp18. M13 is a bacteria-destroying virus with a single-stranded DNA genome about 7,000 nucleotides in length; its known sequence and ready availability make it convenient for this application. The design process has five steps (see Fig. 1 on page 298). First, the desired shape is chosen and is filled from top to bottom by an even number of parallel double helices. Second, a single, long scaffold strand is folded back and forth along the double helices, so introducing periodic crossovers (again for rigidity) between parallel helices. Third, a computer program generates the sequences of many short ‘staple strands’. These bind to the DNA scaffold strand — making it double-stranded — and create crossovers between strands. In the two final steps, the design is examined and refined by computer to relieve strain and to strengthen the structure at the nicks and seams produced in the initial design process. The results that emerge are stunning. Rothemund shows the generality of the approach with six different structures (Fig. 2 on page 299), notably a five-pointed star and a smiley face — myriads of which are a disconcerting sight in an atomic force microscopy image (Fig. 1, above). He further demonstrates the assembly of the individual structures into rather beautiful, higher-order patterns, reminiscent of the designs found in Persian carpets, and shows the absence of any symmetry requirement in the designs by fabricating a map of the world (Fig. 3g on page 300). Rothemund’s basic method is fairly straightforward, and ample experimental and design details are provided in the many pages of supplementary material. He notes that there is a plethora of widely available chemical modifications to DNA strands; this should make it possible to incorporate, for example, dyes or binding elements into these structures at any desired position. The extension of this model from two to three dimensions should not prove too difficult either, given the threedimensional precedents, opening up further possibilities in the design and construction of functional materials at the nanoscale. Thus equipped not only with DNA building NANOSTRUCTURES