THE PERILS OF POLYNUCLEOTIDES: THE EXPERIMENTAL GAP BETWEEN THE DESIGN AND ASSEMBLY OF UNUSUAL DNA STRUCTURES Nadrian C. Seeman, Hui Wang, Bing Liu, Jing Qi, Xiaojun Li, Xiaoping Yang, Furong Liu, Weigiong Sun, Zhiyong Shen, Ruojie Sha, Chengde Mao, Yinli Wang, Siwei Zhang, ‘Tsu-Ju Fu, Shouming Dut, John E. Muellerd, Yuwen Zhang? and Junghuei Chent Department of Chemistry ‘New York University New York, NY 10003 USA Abstract DNA computing relies on the successful implementation of physical chemistry techniques involving oligonucleotides of prescribed sequence. Our laboratory has been involved in the assembly and manipulation of designed oligonucleotides in order to pursue studies in genetic recombination and nanofabrication. We have constructed a large number of unusual DNA molecules, including branched DNA molecules, DNA polyhedra, DNA knots, DNA double crossover molecules, and DNA antijunctions and mesojunctions. Our experience with these systems has uncovered a large number of experimental pitfalls that may confront individuals working with DNA computing. We present our experience in this area with the hope that we can help investigators to anticipate the experimental problems that may affect their DNA. computing femes. Introduction DNA computing, as pioneered by Adleman (1), entails the operations of DNA hybridization, DNA ligation (2), and separation of target molecules. Lipton's elaboration of ‘Adleman's work also uses hybridization and separation. Other schemes for DNA computing involve the use of restriction enzymes (4) and more complex DNA structures, such as those containing branch points (5). The techniques used are primarily those of molecular biology, but the nature of the enterprise is chemistry. By this, we mean that the operations require relatively large-scale molecular operations that result in a substantial number of homogeneous target molecules that need to be amplified or purified in the course of the experiment. The answer must be contained in the products in sufficient quantity to be detected by a technique other than genetic selection. Although the correct solution might indeed be ‘selected’ by its sequence properties, it does not as a rule contain features known from the outset that can result in its selection from a large number of similar molecules by genetic means. Adleman's use of PCR at one step notwithstanding (1), successful chemistry with substantial yields appears to be an important component of DNA computing. This is in contrast to molecular biological techniques, in which small yields at successive steps do not prevent a successful cloning operation that selects for known genetic features in the target (e.g., 6). Our laboratory is not involved directly in DNA computing. However, since the early 1980's we have been performing experiments in the construction of DNA molecules with unusual structures and topologies, in the senses of helical branching and strand linkage. Our goals are twofold: [1] understanding the structural equilibria of branched DNA intermediates in the process of genetic recombination, and (2] the assembly of nano-scale objects and designed periodic matter as a way to solve the crystallization problem for biological macromolecules, and for use as potential components in nanofabrication (7,8). Both of these efforts center on branched DNA molecules called junctions. We have constructed stick-polyhedra, a cube and a truncated octahedron, from junctions that contain fixed branch points. These poiyhedra are complex catenanes of cyclic single-stranded DNA molecules. Besides conventional branched junctions, we general branched structures called antijunctions and mesojunctions (10,11). In addition, we have constructed DNA trefoil knots of both chiralities, as well as amphicheiral figure-8 knots (12,13).i Thus, we too have been involved in chemical-scale assembly of DNA molecules, through use of the molecular biological techniques of hybridization, ligation, restriction, electrophoresis and affinity separation. Several femtomoles of polyhedra have been assembled, the yields of knots ar= in the range of picomoles to nanomoles, and the individual small hydrogen bonded complexes arc ‘made in larger quantities. The constructions that we have undertaken appear to be relatively simple paper (8). Nevertheless, each of them has been completed successfully only by virtue of talists who did the laboratory work. We will review this work here, with emphasis on the (often unpublished) problems that have confronted and overcome in order to obtain the results that were sought. The purpose of this article is to summarize our experience in the experimental treatment of DNA, and to point out pitfalls that may be inobvious to investigators new to DNA chemistry. Branched DNA Molecules ‘There appear to be no stable DNA branches in nature. If one wishes to build them in the laboratory, and to use them as components for the construction of nanoscale objects, a method must be found to assign sequences to them. We have used sequence symmetry minimization techniques to order to achieve this goal (8). The fundamental assumption in all the DNA sequence assignment we have practiced is that the formation of Watson-Crick-paired double helical DNA is the most favorable state for DNA molecules. It is important to point out that this assumption appears to be true only at appropriate DNA concentrations in conventional neutral aqueous solutions, containing sufficient counterions and moderate salt concentrations, usually at ambient or low temperatures. We will stress the importance of experimental conditions throughout this paper. Where does the element of design enter into this picture? What problems can be encountered if the sequence is not carefully selected? The antithesis of control over chemical systems is symmetry, in its broadest sense, i.e., multiple outcomes that are equivalent, or nearly 0, from the standpoint of the free energy of the system. Thus, one must unambiguously —3) from all structures we have designed. Solution Protocols for Assembling Unusual DNA Molecules. It is not sufficient to design a molecule by means of sequence criteria if one wishes to make complex DNA molecules. It is also necessary to choose proper solution conditions. For example, short branched DNA structures, such as J1 (Figure 1) require Mg”* for stability (22): It is possible to visualize the junction as a single band on a non-denaturing gel only if multivalent cations are present in the clectrophoresis running buffer in sufficient quantities (ca. 1-10 mM). When Mg?* is omitted from the running buffer, the complex falls apart. The hybridization protocol is also important: The very first time J1 was formed, a solution containing strands 1 and 2 was mixed with a solution Containing strands 3 and 4 at 4 °C, to produce a solution ca. 10 uM in each strand, The branched junction failed to form until the solution was heated and then allowed to cool slowly. We believe the reason for this result is that the 3' half of strand 1 and the 5' half of strand 2 nucleated a stack that permitted weak associations between their respective 5' and 3° ends, while strands 3 and 4 behaved similarly. ‘This result was obtained, even though equimolar mixtures of strands I and 3 or 2 and 4 run as single strands on non-denaturing gels (23). Eventually, the junctions would have formed, but the barrier at 4 °C was too high to be overcome quickly just by mixing, followed immediately by spectroscopic examination. Structural Considerations in Forming DNA Objects. Structural features of DNA ‘must also be taken into account when a motif is designed. Pethaps the most important of these is the twist of the DNA doubie helix. The DNA backbone is often represented for com pair of parallel lines (e-g., Figure 1), but its helical nature cannot be ignored safely in the design of DNA structure have been worked out (25,26), but it is most often handled experimentally by using an integral number of helical half-turns between the vertices of objects. The cube and truncated