Beyond Watson-Crick: The Next 40 Years of Semantomorphic Science

Abstract

It should come as no surprise that the world of DNA nanotechnology is still learning how to fully master the different steps of the self-assembly process. Semantomorphic science, as the late Ned Seeman would describe DNA nanotechnology, relies on the programmability of nucleic acids (semanto-) to encourage short oligomers to put themselves together (-morphic) into designed architectures (science?). In the same way that Gibson assembly frustrates the molecular biologist, semantomorphic self-assembly has for decades, and continues to, defy the scientist in question. In a brief analogy, Gibson assembly can be thought of as enzymatically directed self-assembly [1] that follows the same general rules as Seeman assembly: (1) guess conditions; (2) set up reaction; (3) pray to entity of choice; (4) check result; and (5) repeat as needed. In other words, when it works, it works well; when it doesn’t, troubleshooting the sticky-ended cohesion between too-large or too-small building blocks with imperfect assays can take months. Returning to semantomorphic science, it is still mesmerizing that any of this works at all, and for that, we owe our deepest gratitude to Ned and his generations-spanning vision.

1 A Brief Retrospective

If we look back at the last forty years of work done in Ned’s Lab, we can roughly break this period into decades, anchored to the founding ideology. In 1982, Ned proposed the idea of using self-assembling DNA oligomers as crystalline scaffolds for the structure determination of biomolecules [2]. By 1991, with Junghuei Chen’s development of the first complex topological object, a DNA cube, it was apparent that Ned’s vision could be possible [3]. In 1998, Ned worked with Erik Winfree to establish that periodic 2D DNA crystals were attainable through self-assembly [4]. This study and the ensuing cluster of papers on 2D lattices [5−9] may be described as a time when it was clear that Ned’s vision should be possible. The following decade brought nanomechanical devices (Chengde Mao’s B-Z device [10], Bernard Yurke’s nano-tweezers [11], and Hao Yan’s PX-JX₂ actuator, [12] among others), DNA origami [13] and LEGO-ology [14], and a vast proliferation in the complexity of self-assembled shapes and devices. In 2004, Chengde Mao published the tensegrity triangle motif [15], which Jianping Zheng modified in 2009 to self-assemble into the first 3D macroscopic DNA crystal [16], demonstrating that rigid, simple, and programmable self-assembly could be carried out in 3D: Ned’s vision would be possible after all.

The start of the 2010s brought about an expansion of complexity: devices involving crystalline logic gates [17], self-replicating DNA machines with Paul Chaikin [18], and natural DNA computing with Natasha Jonoska [19, 20]; and for the first time, these devices were no longer able to be exchanged between laboratories in the same way—the outputs and techniques had now become so complex that composable parts were no longer as simple or modular. Rather than a single motif or reaction, DNA nanodevices had begun to involve a vast number of moving parts, atypical constituents, and specialized conditions. Describing Hongzhou Gu’s nanoparticle assembly line [21] for grant applications remains difficult to this day, let alone the modifications necessary to change its function and the critical role that this device plays as a use case for DNA nanotechnology. The current state of semantomorphic science in the Seeman Lab, with the anticipated publication of our first biomolecular structures attained through 3D DNA crystals (manuscript in preparation at the time of writing), is that Ned’s vision is possible. Not only is it possible, but the feedback gained through modification of simple parts and predetermined solutions to the diffraction phase problem can allow rapid screening of a target library in a way that is impractical using state-of-the-art crystallography. While the advent of cryogenic electron microscopy has already shaken traditional crystallography with promises of imminent obsolescence, there are use cases for Ned’s 3D diffraction lenses that we believe are critical for the next 40 years—particularly in the realm of non-canonical, non-Watson-Crick architectures, which are chronically overlooked by dogmatic crystallographic science. Watson-Crick interactions can be, in broad strokes, defined as double helical interactions involving G:C and A:T base pairs. From this point on, we believe that eclipsing Watson-Crick semantics will allow for technologically relevant morphologies, i.e., semantomorphic science as Ned once envisioned it.

2 A Science Allegory

Ned’s last publication in 2021 was Brandon Lu’s hexagon [22], a structure that can in many ways be thought of as a representation of DNA nanotechnology more broadly. In the outward-facing story of the hexagon, it was discovered that use of non-Watson-Crick sticky-ends in a tensegrity triangle caused non-canonical cohesive interactions involving A:C and G:T. Rather than inhibiting self-assembly, this non-Watson-Crick “mismatch” drives the growth of much larger and, apparently, more thermodynamically-favored crystals with bent Holliday junctions, or branch points. The crystals possess a pseudo-infinite hexagonal channel along the crystallographic P6₃ screw axis, presenting a technological opportunity for nanoscopic organization of composable parts in a metachiral system (see Sect. 4 for elaboration).

Preceding this publication came years of head-scratching and some tunnel vision by way of optimism. Ned once said that the use of a microscope—any kind of microscope—was a heuristic trap: the observer would, more often than not, see whatever they were looking for. In the world of self-assembly, this led to almost fifteen years of curating rhombohedral—and only rhombohedral—crystals. Lurking in the corners of thesis figures and oral histories since the first days of Ned’s 3D assemblies are needle-like, chunky crystals that appeared at first glance to be failure products. As the designs in our laboratory became more complex, the chemistry further from Watson-Crick, and the geometries more distant from B-form double helices, we began seeing crystals with jagged, improper shapes and space group “failures” in the X-ray diffraction patterns (see Fig. 1). Hao Yan published a similar study in collaboration in 2016, in which a “tensegrity square” became an unwound, flower-like design with P3₂ symmetry [23]. In the summer of 2020, the conundrum of the hexagon began a rapid convergence. Updated processing software and the simultaneous result of three separate projects with P6₃ symmetry made it clear that our “accidents” and “failures” were telling us something. In abject disregard for Ned’s adage “garbage in, garbage out,” Brandon Lu and Karol Woloszyn—two extremely talented researchers in the laboratory—on the same day obtained molecular replacement results that showed tensegrity triangles arranged in a chair-hexagon conformation.

Fig. 1

Self-assembly of tensegrity triangles [24]. a–c Various crystallization conditions generate rhombohedra (orange), hexagons (green), and spherulites (blue) in the same drop, despite general optimism for rhombohedral assembly. d Two identical hanging drops on the same glass slide, with hexagonal crystals (left) and rhombohedral crystals (right) attained simultaneously. It is clear that macroscopic effects can result from minute differences in the crystal microenvironment

It was at that moment starkly clear that self-assembly had produced something different: packing effects had imposed curvature on the double helix to form a corkscrew-like architecture from a linear monomer. Nature finds a way, especially when trying to impose Watson-Crick symmetry on strained oligomers engaging in the dubious process of self-assembly. But it is precisely this departure from Watson-Crick that uncovered a new mode of self-assembly—a change to DNA primary sequence had profound implications on the quaternary structure of those oligomers. To this end, semantomorphic science may find its next relevant morphologies by altering its core semantics and by attending to the unexpected results.

3 A Roadmap

We envision the departure from Watson-Crick architectures through (1) modification of DNA semantics by way of new base pairing rules; (2) the augmentation of DNA syntax leading to diverse secondary and tertiary structures; and (3) the expansion of the DNA operating system through modifications to the sugar-phosphate backbones. We elaborate upon these categories below.

3.1 DNA Semantics: Schrödinger Crystals Versus Seeman Crystals

In meetings and conferences, Ned would often repeat that “symmetry is the opposite of control!” As it is well known, the very first step toward designing and programming synthetic branched junctions involved the development of SEQUIN, a FORTRAN algorithm explicitly implemented for symmetry minimization [25]. In a computer science sense, this can be thought as the greatest common substring optimization algorithm, and in a biochemical sense the free energy minimization of desired secondary structures. Schrödinger posited in his 1951 treatise, “Was ist Leben,” that the fundamental nature of life required an aperiodic crystal as the carrier of genetic information [26]. Only by minimizing repeating segments of information in a geometrically isomorphic crystal would living systems be able to propagate the stored information between generations. Schrödinger’s postulation preceded SEQUIN by four decades, but proposed a similar principle of symmetry minimization contained in topological units unperturbed by the information carried within.

By contrast, Seeman crystals can be thought of as mostly periodic crystals, whose sequences repeat (tesselate) in the semantomorphic bundles that we call motifs. These motifs are subject to crystallographic symmetries to generate theoretically infinite nanomaterials. Within this system, local symmetry minimization is critical in imposing long-range order: Schrödinger’s aperiodic crystal must be tightly constrained. The tensegrity triangle, Ned’s crowning joy, contained a periodic unit with only 42 nucleotides (nt). Geometric asymmetry, in the form of 3:3:1 annealing ratios of three interwoven oligomers, counteracts the rotational symmetry of the triangle arms to allow for sequence symmetry down to 42 nt (Fig. 2a vs. Fig. 2c). A similar motif with asymmetric sequences across the triangle and no rotational symmetry between arms contains 126 nt in each periodic unit for the same unit cell, this time with symmetric strand ratios between seven unique oligomers (1:1:1:1:1:1:1) (see Fig. 2d). There is a clear exchange between geometric and sequence symmetries—penalties in sequence minimization require more complex geometry, which in turn makes experiments more difficult and the analysis more complex. It is clear that Schrödinger’s prescription holds for DNA nanotechnology and can be amended for Seeman crystals: the more aperiodic a crystal’s sequence, the more symmetric the geometry can be. With this in mind, expanding the genetic code will have a clear effect on the types of structures that nanotechnologists can engineer.

Fig. 2

Tensegrity triangle motifs. a–b The symmetric two-turn triangle developed by Jianping Zheng employs a repeating 42 nt unit that attains higher-order complexity by asymmetric strand ratios (strands S1–S3) and sequence weaving. c–d The asymmetric version of this motif, developed by Rachel Chernet, only repeats over 126 nt, but involves easier annealing with symmetric and logical strand ratios (strands S1–S7)

There are many existing strategies for expanding “vanilla DNA”—as Ned would describe Watson-Crick chemistry. The most obvious lies in nucleobase analogs, which typically employ rearranged hydrogen bonding groups: inosine (I) [24], isocytosine (S), isoguanine (B) [27], 2,6-diamino-purine, 2,4 diamino-pyrimidine [28], and “hachimoji” nucleotides dZ and dP [29]. Indeed, very extensive applications of these nucleotides have been carried out by generations of Ned’s students, and it is clear that they satisfy the requirements for Seeman crystals [30]. The full implementation of an expanded lexicon has, to our knowledge, not been carried out in a self-assembly context.

A further development lies in metal-mediated DNA base pairing (mmDNA), which has emerged over the last two decades as a viable means to add symmetry (reducing the aperiodicity of DNA crystals) by allowing non-Watson-Crick pyrimidine:pyrimidine stabilization through the coordination of diverse metal ions (most commonly Ag⁺ and Hg²⁺) [31, 32]. The payoff here lies in reducing the regularity of the overall behavior—metal base pairs are known to be pseudo-covalent, stabilize the helix, and add novel electronic and magnetic properties [33,34,35], all of which represent a general departure from hydrogen bonding as the sole means of implementing Schrödinger and Seeman crystals. By adding bioinorganic diversity to the properties and behaviors of nucleic acids, there lies an opportunity to increase the complexity of DNA structural motifs—more than compensating for the increase in information entropy. A main focus of the work in Ned’s group lies in these types of architectures, and we look forward to sharing Ned’s latest works with the community in the coming months and years.

We envision that the expansion of the coding language of DNA will enable more precise, geometrically minimized, and behaviorally complex semantomorphism, and we have only begun to scratch the surface of these techniques.

3.2 DNA Syntax—Information Bundles and Secondary Structures

With the advent of the first functional poly-crossover (PX) motifs by Zhiyong Shen [36], Ned found a way to augment the complexity of DNA branches (Fig. 3b). Beyond topologically-closed knots and catenates [37], PX motifs represent a wonderful conundrum of the double helix that has driven the creation of new dynamic devices. For years, Ned tried to convince biologists to utilize PX, to search for it in biological systems that Xing Wang demonstrated [38]. The utility of PX DNA was made clear with the PX-JX₂ device, first designed by Hao Yan in 2002, in which geometric control was attainable through the organization of DNA within differentially shaped information packets [12]. Through the addition of small changes attained through strand displacement, it was possible to reorganize the whole semantomorphic structure—this represents a fundamental change in the grouping and behavior of information, and in the same way syntax dictates the order and function of words in a sentence (Fig. 3). To this end, expanding the syntax of DNA nanotechnology will enable more complex materials in the coming decades. A road to syntactic expansion follows Ned’s ideas, involving hairpins, triplexes, quadruplexes, and yet-to-be -discovered or developed secondary structures.

Fig. 3

Archival chalk drawings from Ned’s office of various DNA secondary structures. a Duplex DNA; b PX DNA; c JX₂ structure; d PX DNA with hairpins

The hairpin was one of the earliest tools in the semantomorphic book of style—Hao Yan used this technique to give a 2D crystal the power of nano-actuation: by interfacing with strand displacement, hairpins can act as geometric parentheses, putting topological information on hold until it is needed [39]. Hao’s motif changed shape like an omni-directional accordion, performing nanoscale work stored in hairpin parentheticals. Hairpin technology has been extensively explored, but likely has new tricks and interfaces to tell us, especially when implemented with other syntactical elements.

Triplex-forming oligomers (TFOs) exploit the major groove of pyrimidine-enriched DNA sequences to bind a third strand. The overall architecture of the DNA is likely altered to some degree, as Seeman crystals do not form with TFOs during self-assembly—there is, however, no difficulty in decorating these crystals with TFOs after the fact. In this way, the effect of a TFO is not unlike an exclamation mark—placed properly it can manifest a new behavior in an existing sentence. This technology, developed in collaboration with David Rusling, has been interfaced with various attachment chemistries [40] and shows strong promise for adding reporter and sensor behavior to semantomorphic structures. A complete structural description of TFO-bound DNA nanostructures has not been attained, but it is clear that these motifs will enhance and emphasize DNA technologies in the coming years.

It is known that repeats of guanine bases can interrupt the double helix by forming a weaving tetraplex structure (G4), and Ruojie Sha and Nongjian Tao built on this idea to show that G4 molecules are excellent electronic conductors [41, 42]. Unlike Watson-Crick DNA, which could be generously described as a weak resistor (or a fantastic molecular heater), G4 motifs are now known to be wide-bandgap semiconductors. In essence, these structures bind alkali cations between poly-guanine stacking planes and impart electronic functionality to an otherwise inert DNA. This structural organization changes both the orientation and the underlying electron delocalization of the material—in this way, G4 acts as a semicolon. To be useful in a semantomorphic sense, it must come in the middle of a DNA sentence, but it changes the focus and structure of the surrounding information. One does not use a semicolon lightly; and conversely, G4 must be used sparingly for its symmetry penalties and strong departure from Watson-Crick helical parameters. The implementation of G4 within structural DNA motifs has been sparing, but as designs become more complex, we can expect G4 to become increasingly common in the future.

There exist other secondary structures that rely on semantomorphism [43], such as aptamers [44], i-motifs [45], and of course, paranemic cohesion (PX motifs). The future of DNA-based technologies will strongly rely on sequence context and behavior, and the ability to code for novel heterostructures is a powerful and underexplored aspect of semantomorphic science.

3.3 Nucleic Acid Operating Systems: XNA and Beyond

If polymerases can be considered the read and write heads of the DNA hard drive, the metaphor can be extended to consider the sugar-phosphate backbone as the operating system in which nucleobase information is stored. Entire classes of enzymes are employed to extract DNA information into the RNA operating system and furthermore translate that information into peptides. Alterations to the nucleic acid operating system generate what are known as xenobiotic nucleic acids (XNAs) [46], and they offer what any orthogonal digital platform can: data encryption, novel behaviors and features, and the need to rewrite basic programs.

Encryption is a result of enzymatic orthogonality—with a different form of aperiodic crystal, information can neither be read nor copied and therefore cannot escape from a synthetic biological system encoded using XNA [47]. This concept has powerful implications for synthetic polymers and synthetic life. The novel features result from the chemical modifications present in XNA via the addition of hydrophobic groups, linkers, and various anchors and substituted sugar moieties [46], leading to different interactions with the surrounding solvent and ultimately presenting the opportunity to program interfaces with chemistry beyond neutral (physiologically adjacent) aqueous systems. Finally, a departure from Watson-Crick geometry changes the shape of the ensuing molecule, including the helical period, the radial dimensions, and the overall pitch of the nucleobases. In order to be successful in achieving nanotechnological programming in an XNA system, the connection between semantics and morphology must be painstakingly re-established for each polymer. The benefits of this expansion have been clear for some time, and work toward structural DNA(+) nanotechnology has been carried out robustly for at least a decade.

Cody Geary and colleagues demonstrated this concept by folding RNA nanostructures while they were transcribed from a DNA template [48]. This represented a fundamental shift toward “living nanotechnology” or “in vivo nucleic acid nanotechnology” by showing that the RNA operating system could be (occasionally) relied upon to generate semantomorphic structures. Ned implemented true XNA backbones within double-crossover motifs, measuring the double helical periodicity of peptide nucleic acids (PNA) [49] and 2′-fluoro-deoxyribonucleic acid (F-DNA) [50]. PNA has subsequently been shown to operate well in organic solvents [51], which has long considered a barrier to self-assembled architectures predicated upon hydrogen bonding.

Ned and Jim Canary at NYU have been collaborating since 1997 to attach organic polymers to the sugars of 2′-O-propargyl-modified ribonucleotides. This approach employs a sterically-unhindered linking site to “plug in” a variety of molecules to the double helix. This was most elegantly demonstrated by Xiao Wang through polyanaline/emeraldine functionalization of a DNA cage that imparted optoelectronic functionality into Seeman crystals [52]. This has been further extended into nylon-DNA [53] and tertiary structure-like crosslinking [54].

Further orthogonal backbones of interest might overcome charge polarity in the phosphates, impart an integer number of nucleotides per helical turn (to avoid the “curse” of 10.5 bp/turn), increase thermostability, frustrate endonucleases, provide a sequence-agnostic attachment strategy for nanomaterials, or even impart electrical or magnetic functionality to the polymer of life [46, 55]. It is clear that backbone modifications are an underexplored, technically difficult, but rewarding addition to nucleic acid nanotechnology. With any operating system change, the new environment must be first explored, structurally characterized, and tested. As with PNA and RNA nanotechnologies, future XNAs can be expected to mature and allow a fully-elucidated and adaptable semantomorphic coding environment.

4 Beyond Watson-Crick: A Call to Action

The benefits of escaping Watson-Crick geometry are hopefully made clear: semantomorphic science stands to gain (1) a much greater geometric diversity through a more versatile genetic code; (2) expanded diversity through syntactic groupings of bases employed in new secondary structures (see Fig. 4); (3) orthogonality to enzymatic and aqueous systems through altered backbone operating systems; and (4) an integration of these various approaches to impart novel mechanical, material, electronic, or catalytic behaviors into semantomorphic polymers. As with DNA nanotechnology in its early days, the changes required to carry out this paradigm shift will require an immense amount of experimental exploration and community action; but, unlike the first forty years, the next forty years can rely on the incredible foundation laid out by the pioneering work of Ned Seeman. It is with the deepest gratitude that we remember Ned and his mission and the painstaking groundwork that he and his early team performed to create the field from scratch. With the guidance left in his writings and with his students, colleagues and collaborators across the globe, there exists a full and flourishing roadmap forward. In his work, “DNA Nanotechnology at 40,” Ned concluded “Every day I open a journal and I’m surprised by another unit of progress … I like being surprised that way.” [56] Let us continue to surprise him.

Fig. 4

Self-assembly beyond Watson-Crick: cross-section of 3D hexagonal DNA assembly, PDB 7R96 [22]. The periodic unit forms a tensegrity triangle (orange center, see Fig. 2) which packs into a star-like lattice. The space group P6₃ indicates a sixfold symmetry in each channel (see the hexagon-like nature of the triangle arrangement) and that three layers of the material are required in the z-direction for the meta-helix to make a full turn around the cavity (the crystallographic “screw axis”). This discovery was a joyful surprise to Ned