1 DNA Nanotechnology: A Personal Account

I stumbled into DNA nanotechnology at the turn of the millenium and since then actively participated in about half of its 40 year history which is celebrated here. With a background in solid-state nanophysics—having spent time in clean rooms for the fabrication of semiconductor nanodevices—I was initially fascinated by the promise of “self-assembled electronic devices” that would become possible with the help of DNA’s magic programmable base-pairing rules in the future. In fact, I was completely unaware of DNA nanotechnology until I saw a talk by Uri Sivan from Technion in 1998 who, together with Erez Braun and co-workers, had just realized the first DNA-templated metal nanowires [1]. I actually first considered doing a postdoc with Uri Sivan, but during a visit at Bell Laboratories in 1999, Bernard Yurke introduced me into his work on “DNA tweezers” that he had started with Allen P. Mills and Andrew Turberfield (who did a Sabbatical at Bell Labs in this year).Footnote 1 The vision of making artificial molecular machines from DNA molecules seemed so unconventional that I felt immediately attracted to it. Fortunately, Bernie Yurke offered me a position, but due to a delay in funding I could only start at Bell Labs in early 2000. Bernie and his co-workers were so nice to let me do a series of control experiments for the DNA tweezers paper [2], which was already under review at that time, and made me a co-author on this extremely important paper. During our work on strand displacement-based DNA machines I finally became aware of Nadrian (Ned) Seeman’s pioneering work in DNA nanotechnology—in fact, together with Chengde Mao he had already published a different type of DNA nanodevice in 1999, which was based on the B-Z transition of CpG rich DNA sequences [3].

A major initiating event for me in this period was the seventh International Workshop on DNA computing in Tampa, FL, where for the first time I was exposed to the deep connections between theoretical computer science and self-assembly processes—I still remember a wonderful tutorial introduction by Erik Winfree—and also various ideas on computing in biological systems—topics that have immensely widened my view on the subject and have fascinated me ever since.

2 Designing and Programming with DNA

DNA nanotechnology is an extremely interdisciplinary field of research. Compared to many other fields, it also has a relatively low entry barrier for researchers with diverse scientific backgrounds. There are various reasons for this, but in a deep sense really the unique properties of DNA molecules are responsible for it.

2.1 DNA—A Programmable Molecule

DNA’s biological role is intimately connected with its molecular and structural properties, which are heavily utilized in DNA nanotechnology. Of course, there is base-pairing, adenine (A) pairs with thymine (T), guanine (G) with cytosine (C), but in such a manner that A-T and G-C base-pairs have the same size in the context of the double helix. This makes a DNA duplex a structurally very uniform heteropolymer, which is biologically important because this allows DNA and RNA polymerase to run smoothly over it regardless of its sequence. Of course, this also means that information can be encoded in the sequence of base-pairs. As a double-stranded helix, DNA is mechanically relatively rigid. Binding between complementary strands is highly specific and depends on the base sequence. The thermodynamic properties can be determined in an approximately additive manner by just summing up the contributions of nearest neighbors [4]—there are no long-range interactions between distinct sequences along a double helix that complicate matters. Put in the nanotechnology context: Double-stranded DNA is a rigid, information-encoding polymer with uniform geometric properties. The interactions between DNA single strands can be programmed by choosing the appropriate sequences.

This is almost everything you need to know to get started with DNA nanotechnology—which is quite astonishing, because really a lot is known about DNA—you can ignore the details in the beginning, but they become important later. Because of its central role to life, DNA has attracted scientists from very different disciplines and next to biologists, of course, chemists who developed schemes to synthesize and chemically modify DNA at will and who also came up with synthetic DNA and RNA analogues. Physicists like to study the structure of DNA, its mechanics and dynamics as a polymer, its thermodynamics, and its charge interactions [5, 6]. A whole sub-discipline of computer science—bioinformatics—is devoted to the study and comparison of DNA sequences. DNA nanotechnology greatly benefits from all these achievements—automated synthesis, structural and thermodynamic information, and computational tools are all available. This makes DNA nanotechnology to be the most advanced molecular technology so far, and because it is sequence based, it is also programmable in a relatively straightforward manner.

2.2 Learning by Building

Researchers in synthetic biology often talk about the design-build-test-learn (DBTL) cycle and about understanding by building. In DNA nanotechnology, these principles are in fact realized! The availability of computer tools combined with automated synthesis and a wide range of established characterization tools enable even inexperienced researchers to quickly create novel DNA nanostructures, study their behavior, and use them in applications. When mistakes are made, one can relatively easily go through another DBTL cycle to improve the results. Importantly, this makes it possible to find a good balance between rational design (taking into account the wealth of nucleic acids knowledge) and a learning-by-doing approach.

Among the many fascinating achievements of DNA nanotechnology [7], DNA origami is certainly the best known outside of the community [8, 9]. DNA utilizes a long single-stranded DNA molecule termed the “scaffold” and a large number of shorter staple strands that sequence specifically bind to designed positions along the scaffold. This creates links between these positions, which fold the scaffold into a three-dimensional molecular shape. The same scaffold strand can be “programmed” to adopt completely different shapes simply by the choice of the staple sequences. Using a DNA origami design program allows researchers to design such origami structures on the computer and choose staple sequences for synthesis that will later physically assemble the desired shapes in the test tube [10]. An interesting development sets in here: Designing DNA nanostructures using computer programs and the corresponding DNA nanotech methodology represents a new skill set that draws from, but is distinct from that of traditional biophysics or biochemistry. As stated already before, you do not need to know every chemical detail about DNA to build such structures. At the same time, origami researchers develop a “feeling” for what can be built, and what not.

Think of modern computer programmers—it is not required to have detailed knowledge about computer hardware (or electronics and solid-state physics, for that matter) to be able to write code in a high-level programming language and solve complex tasks with it. While it may be a little sad that programmers do not know how certain things work in detail—it is also a big achievement that they do not need to know! Modularization and abstraction help to make much bigger leaps much faster. And, more or less, this is what happens in DNA nanotechnology—this is the power behind the idea of molecular programming [11, 12].

2.3 Challenges and Limitations

Already now the ability to create DNA-based nanostructures of almost arbitrary shape combined with the availability of a wide range of chemical modifications for these structures has found use in nanomedicine, biosensing, nanoscale science, and biophysics. In such applications, DNA nanostructures are typically used to spatially organize molecules and nanoparticles with nanometer precision.

More or less obvious questions (and challenges) for the future of static DNA nanostructures regard the scale of the structures: Can we make bigger and bigger functional structures, what quantities and at what cost [13]? Can we achieve greater precision in the arrangement of molecules and increase the chemical diversity and functionality of DNA nanostructures? Using modified DNA and nucleic acid analogues, e.g., it would be great if one could rationally design DNA nanostructure-based catalysts with similar catalytic versatility and power as enzymes. These challenges are rather chemical in nature and will hopefully be addressed by the gifted chemists in the field (if we come back to the comparison with computer programming: chemists provide the hardware for DNA nanotechnology and thus define and extend the capabilities of the hardware-agnostic molecular programmers).

In my opinion, a greater challenge lies in the realization of dynamic functions and the control and utilization of non-equilibrium processes. Over the past two decades, DNA nanotechnology has been concerned with the realization of molecular switches (such as the tweezers), machine-like devices (molecular walkers [14, 15], rotors [16], and the like), DNA-based chemical reaction networks (CRNs), and DNA computers [17, 18]. There are different motivations for creating such systems: On a fundamental level, one would simply like to learn how to build artificial molecular machines or generally realize dynamic molecular functions. Then, such systems should be of great use in nanotechnology—examples for applications of dynamic DNA processes already exist: enzyme-free DNA-based sensors (hybridization chain reaction [19], catalytic hairpin assembly [20]) and DNA-based super-resolution microscopy (DNA-PAINT) [21].

In several aspects, however, dynamic DNA systems are still very limited. Next to their low chemical functionality (as for static structures, see above), their main restrictions are the difficulty of operating them continuously and autonomously and their relatively low speed. If one considers DNA nanotechnology as an approach to emulate and thus understand biological self-organization, these will be major hurdles.

3 From Self-Assembly to Non-equilibrium Dynamics and Self-Organization

Realizing dynamic, biology-inspired self-organizing systems, and building molecular machinery in particular, involves the challenge of controlling the flow of chemical energy through a non-equilibrium chemical system to generate interesting dynamics and structure (cf. Fig. 1).

Fig. 1
An illustration of nanostructures. It features the following, equilibrium and non-equilibrium self-assembly, transient, dissipative assemblies, non-autonomous devices, autonomous devices and machines, energy flow dynamic reconfiguration, and operation in cells.

DNA nanotechnology has very successfully generated nanostructures, which self-assemble as the thermodynamically most stable structure of a mixture of components. Various molecular machines and devices have been realized by periodically driving switchable molecular structures out of equilibrium in a non-autonomous manner. In the future, DNA nanotechnology is anticipated to increasingly operate further away from equilibrium, which will be of interest in various contexts: accessing far-from-equilibrium self-assembled and dissipative structures; driving molecular machines continuously and autonomously; realizing active matter and cell-like soft robotic systems that can respond to their environment and quickly switch between different states of organization. A specific challenge is the production and operation of nucleic acid devices inside living cells. In the figure, \(\lambda \) denotes some parameter that characterizes the deviation of the system from equilibrium

Full size image

3.1 Molecular Machines

Molecular machines and devices made from DNA have usually been based on DNA conformational changes (formation of duplexes, hairpins, triplexes, i-motifs, etc.) that take place in the presence of a DNA input or in response to a change in environmental conditions (salt, pH, light). Due to the nature of these stimuli, most such systems are operated in a clocked manner—they are kicked out of equilibrium by a change in conditions (addition of fuel), followed by relaxation to a new equilibrium state.

Continuous and autonomous operation out of equilibrium is challenging. An ingenious idea to achieve this was developed by Turberfield and Yurke and further developed by many others [22, 23]. DNA fuel can be forced into an unreactive secondary structure (like a hairpin loop), which can be activated via strand invasion by a DNA trigger strand. The activated sequence domains can then invade another inert fuel hairpin (or secondary structure). This principle can be used to store chemical (hybridization) energy in hairpin loops, which can be released in a dynamical DNA system in catalytic cycles or cascades. In essence, this is an example of kinetic pathway engineering: The inert fuels are present in a metastable state, which can only relax through reactions that drive a process of interest [23]. This principle has been used to drive autonomous molecular walkers and also lies at the heart of a variety of nucleic acid amplification schemes [24].

Even though these achievements are extremely impressive, the field of DNA machines has not yet “taken off,” and the capabilities of DNA nanomachines are still far from those of biological machinery. DNA-based molecular machines are slow, they have not been used to exert appreciable forces to carry out tasks or move objects over larger distances. Why is that?

Using DNA as a fuel is one of the major conceptual developments in DNA nanotechnology—DNA is an information-bearing fuel and can simultaneously act as a molecular address code that switches only specific molecular processes. As can be seen from the applications that have emerged, this is particularly useful for biosensing applications and for the realization of chemical reaction networks. DNA fuels have not been very successfully used, however, to generate fast movements and quick conformational changes as found in biological molecular machines.

Biology does not use an information-bearing fuel molecule—it uses a couple of small molecule fuels (ATP, GTP) for almost everything. These molecules are present at high concentrations (in the mM range), are constantly (re)generated by cellular metabolism, and shared by a large diversity of different processes that run in parallel.

Using small molecule fuels at high concentrations results in much faster kinetics. Millimolar concentrations are much higher than the typical concentrations used for DNA fuels, with an accordingly higher on-rate for the fuel.Footnote 2 Furthermore, binding of a small molecule to its binding site is faster than binding between large molecules with multiple interactions or complex interaction surfaces—DNA hybridization requires nucleation of a few base-pairs “at the right position,” for which the participating DNA molecules must be in the appropriate orientation to each other. Of course, also the off-rate is affected: While a small molecule with a comparatively small \(\Delta G\) for binding will quickly dissociate, we have to use branch migration or other means to wrest off a DNA fuel molecule from a DNA machine to start another machine cycle.

Other aspects that play a role in biological machines and may guide the further improvement of artificial machines relate to the coupling of fuel consumption to conformational changes. In biology, sometimes only a small change—binding of ATP into a tight pocket or changing a single charge by phosphorylation—is transduced mechanically to a large change in geometry. DNA nanotechnology has not achieved this level of molecular control yet. It remains to be seen whether DNA—even in chemically modified form—can be used for such mechanisms at all.

If we want to stick with chemical driving mechanisms rather than physical manipulation of DNA nanodevices with magnetic [26] or electrical fields [27], or light [28,29,30], we will also need to figure out a mechanism to provide high-energy fuels for longer periods of time. If we do not want to use manual or microfluidic addition of fuels and removal of waste [31], we will need to embed the DNA devices within a reaction cycle that generates the fuels—in other words, some kind of metabolism is required. This probably cannot be achieved with DNA alone.

Potential tasks for building future molecular machines based on DNA might thus be the following: (i) find the “ATP equivalent” for DNA nanotechnology. It would be great to have a universal fuel that is used by all kinds of DNA machinery (which, of course, would mean that we have to give up sequence-addressability through the fuel); (ii) find a way to couple fuel consumption to a quick conformational change; (iii) realize some kind of metabolism that generates fuel at high concentrations or find some other clever way to replenish it.

What will be the role of DNA programmability in this context?

3.2 Non-equilibrium Chemical Dynamics and Self-Assembly

Similar considerations as for molecular machines apply to dynamic and growing biological structures and materials in general. Structures in biology are rarely built to last. There is a constant turnover of molecules within supramolecular assemblies like the cell membrane or the cytoskeleton, which also gives biological systems the ability to respond to their environment, to grow, change shape, move, etc.

For instance, treadmilling of cytoskeletal filaments [32] involves addition of new monomers at one end, consuming ATP or GTP, and dissociation at the other end, which results in an apparent movement of the filaments. Several groups have succeeded in generating polymers from DNA tiles or helix bundles [33,34,35], but the polymerization process was usually based on DNA hybridization and was not coupled to consumption of a high-energy fuel. In biology, the growth of filaments can actually exert forces on membranes (in addition to the Brownian ratchet force generated by growth of the filaments, molecular motors are involved [36]). It is currently unclear, how the slow growth of DNA filaments could be applied for something similar. At this point, DNA nanotechnology can provide the building blocks, but DNA-fuelled processes do not yet compete with the ATP/GTP-driven processes involved in the formation of non-equilibrium structures.

There are opportunities, obviously, in fusing ATP consuming proteins (or other, potentially synthetic catalytic units) to DNA structures and thus generate molecular systems with active dynamics. Here the DNA components will provide the spatial organization, while the dynamics are driven by some other chemical process, which can be supplied with fuel more easily. In principle, one could also use cellular metabolism itself to drive DNA or RNA-based systems and thus solve (or circumvent) the problem of energy supply and creation of non-equilibrium conditions. Several examples of toehold-mediated strand invasion processes have been demonstrated in vivo, where all of the components were generated by transcription reactions. This so far mainly refers to switchable regulatory RNA molecules (toehold switches [37] or conditional guide RNAs [38, 39]), but recent developments (still in vitro) also suggest a route to realize more complex RNA strand displacement circuitry from transcription products [40, 41].

Somewhat intermediate in this context is the utilization of DNA or RNA polymerases, ligases, and nucleases that can be adopted to power dynamic nucleic acid-based systems in vitro  [42,43,44]. These enzymes accomplish the production and degradation of RNA or DNA fuels consuming nucleotide triphosphates and can therefore keep a system out of equilibrium as long as the NTPs are supplied and waste products are removed. Enzyme-driven systems retain much of the sequence programmability of pure DNA systems and have been used to create bistable systems and oscillators [42, 43, 45, 46], pattern forming systems [47, 48], transient dynamics [44, 49, 50], or to control assembly/disassembly reactions [35]. In contrast to dynamic DNA systems based on hybridization interactions alone [51], however, enzyme-driven systems have to cope with the biochemical idiosyncrasies of the enzymes used and are intrinsically less programmable.

In spite of the limitations mentioned, it is still conceivable to extend dynamic DNA nanotechnology to certain far-from-equilibrium processes and assemblies based exclusively on DNA and thus to create assemblies that are not the thermodynamically most stable ones. For instance, control of kinetics has previously been shown to be important in the context of origami folding and allowed to direct the folding process toward one of several alternative possible structures[52, 53]. Kinetics of strand displacement reactions can further be tuned via the length of the toehold [54], or by using tricks such as remote toeholds [55], which allow the engineering of kinetic pathways. “Handhold-mediated strand displacement” has just recently been demonstrated to enable the formation of templated far-from-equilibrium DNA assemblies [56].

3.3 Robots

Of course, speed is not everything. Many applications will not require fast movement or response. For instance, if we just want to arrange molecules into specific geometries, static DNA nanotechnology is sufficient. Also growth and pattern formation processes do not strictly have to be fast—think of emulating plant growth or programming the development of an organism. Where speed probably matters, is in robotics [27, 57,58,59,60,61,62,63]. At least in my interpretation, robotic systems should be able to interact with their environment—sense, make decisions, respond, and act on their surroundings. In order to be able to do that sensing, computation, and action have to take place on a relevant timescale. In this context, it may be useful to adopt an engineering approach toward molecular robotic systems. We need to identify tasks for the robots and define desired performance characteristics and benchmarks. Very likely, for many applications it will not be possible to reach the aims set for the robots using DNA alone—any physical or chemical trick to speed up the systems will here be welcome.

Another challenge for the realization of DNA-based molecular robotics is the integration of different robotic components into a consistent system. DNA nanodevices have already been shown to be capable of sensing, computation, and movement, but these have been combined into a molecular robotic system only in a few cases. Rather than programming DNA robots at the sequence level, programming at a modular level could be interesting. Taking inspiration from macroscopic modular robots, one could strive for generating optimized DNA robotic components with standardized interfaces that allow the reuse of known functional DNA modules.

Already now, many researchers like to apply the same type of DNA origami structures (think of Paul Rothemund’s rectangle [8]) to many different tasks—simply because they are useful and have been proven to work—generating ever more differently shaped structures is not necessary for many applications. Further, using localization and spatial organization of DNA components not only allows speeding up slow hybridization reactions by generating high local concentrations [62, 64,65,66,67]—it also allows reuse of the same components by spatial separation and thus avoids cross-talk. Thus, strategies for defined modular interfaces and spatial arrangements of known components might become a new branch of molecular programming.

4 What Lies Ahead?

DNA nanotechnology 40 years after its inception by Ned Seeman is still developing rapidly. Hundreds of laboratories worldwide harness the power of DNA self-assembly to arrange molecules and particles into specific geometries to address scientific questions in a wide range of research areas. Due to all the favorable properties of DNA already mentioned above, DNA nanotechnology is here to stay—it has (or will) simply become a standard approach for molecular engineering, similar as lithography or the production of nanoparticles in nanoscience. Further progress in the field will require improving the chemical versatility of DNA and DNA analogues without sacrificing the unique capability for sequence-programmable self-assembly.

The greatest challenges for the field lie in the realization of dynamic functions. Can one generate DNA-based molecular machines that are really useful? Can one generate complex chemical dynamics based only on DNA? Can one emulate biological behaviors with programmed DNA nanosystems? Will it be possible to speed up the systems to make them useful for applications? If not, what are the general principles, the insights we can derive from DNA-based model systems?

It is likely that we will need to use some kind of dissipative process—chemical catalytic cycles, enzymes, cellular metabolism—or physical driving to generate interesting dynamic behaviors. The challenge, then, is to use the power of DNA nanotechnology to arrange the respective active components in space and to direct the non-equilibrium energy flow to generate and control self-organization processes. A major challenge will be to develop schemes to abstract these behaviors and ultimately make them as programmable as static DNA structures.

Let me end with a personal note. I am glad and grateful that I can be part of this great interdisciplinary adventure called DNA nanotechnology—DNA nanotech is a wonderful scientific community, with so many highly inspiring personalities from all fields of research, starting, of course, with Ned Seeman who in his talks always emphasized the importance of knowing everything about chemistry, physics, biology, math, ..., and arts. Even though mentioned several times in my text, one also should not overemphasize usefulness and applicability. Using DNA as a generic, programmable molecular substrate to explore interesting concepts and ideas “in the test tube” is a worth on its own. It allows you to stand back and approach problems in self-assembly and chemical dynamics at a more abstract level—and thus gain insights into their general governing principles.