This articles aims to explain the basic concepts, perspectives and the main ethical considerations regarding the concept of DNA nano-robots. This outreach paper has been written with equal contributions from all the DNA-Robotics Early stage researches. Authors are listed in random order:
Quentin Vincentini, Lorena Baranda Pellejero, Aitor Patiño Díaz, Alba Monferrer i Sureda, Michael Pinner, Yash Bogawat, Minke Nijenhuis, Angel Santorelli, Nestor Sampedro, Marco Llocaico, Igor Baars, Mihir Dass, Karol Kolataj, Joakim Bohlin, Rafael Carrascosa Marzo.
Before explaining what ‘DNA Nano-robotics’ is, we first have to answer a much more generic question: “What is a robot?” Robots are generally thought of as humanoid devices with artificial intelligence, but is that really all they are? This definition of robots is of course mainly ingrained in public awareness thanks to pop-culture juggernauts such as “Star Wars”, “iRobot”, or “RoboCop”. The term ‘robot’, however, refers to a much wider spectrum of devices. In essence, robotic systems are composed of sensors and actuators connected to and coordinated by an information-processing unit. Sensors provide information about the environment, which is evaluated by a computer (the information-processing unit) and used to decide on necessary actions – for example mechanical motion (enabled by the actuator). As a result, robots are capable of automatically carrying out pre-designed series of actions. Robotic systems currently transform the way we work and live – they greatly speed up and enhance manufacturing processes, they provide assistance in diverse areas such as health care or environmental remediation, they perform in environments inaccessible, too harsh, or too dangerous for humans – not least, our autonomous cars of the future are likely to be robots on four wheels.
1.2 Robots at the nanoscale
Now we know what robots are and how they are used to automate many different tasks in our everyday life. However, all of these macroscopic robots are in the range of meters or millimeters, which makes it trivial to manipulate and control them. To fully appreciate nano-robots, you must first remember that they are much, much smaller than classical robots. To be more precise, nano-robots are measured in the nanometer scale where one nanometer is one billionth of a meter. When you work with such small structures, you have to take into account that they are subject to a completely different environment than macroscopic structures. For example, nanoscale structures are in constant motion, known as Brownian motion, where molecules in solution constantly collide at high speed. You must also take into consideration that some physical interactions such as electrostatic repulsion and intermolecular interactions are stronger at the nanoscale. Taking into account the nanoscale environment, you can start thinking about the designing nano-robots with specific features and applications. In the design of nanoscopic structures, you have to forget about classical approaches that we use to fabricate macroscopic robots. Mostly because we do not have the tools to specifically and precisely manipulate matter at the nanoscale. In other words, we do not have a nano-hammer, a nano-drill or nano-nails that we can use for building our nano-robot.
1.3 DNA as a building material
DNA is a great building material for your nano-robots. “Why is that so?” the curious reader may ask and the answer to this question lies in the structure and properties of DNA itself. Think of double-stranded DNA as a ladder, which twists upon itself along its length, forming the immediately recognizable double helical shape that is ubiquitous today (Figure 1). The chemical make-up of DNA includes a backbone – the legs of the ladder – made of sugar (a Deoxyribose, which is a ribose sugar that is missing an oxygen atom) and phosphate (a phosphorus atom connected to 4 oxygen atoms) molecules. Sticking out of this backbone (like the rungs/steps of a ladder) are the molecules called Nucleic Acids, of which there are 4 types. They have longer names, but for this purpose, we will just call them A, C, G and T; and now things get interesting. A programmable system is called so because on providing certain inputs, consistent outputs can be obtained based on the programming of the system. How is it related to DNA? From the 4 bases A, C, G and T; A will only bind to a T or vice versa (and not to G or C) and the same relation holds true for G and C. Therefore, a sequence of DNA like ‘AAGGTTCC’ will most stably bind to its complementary sequence i.e. TTCCAAGG.
This property, combined with the ability to synthesize DNA of any user-defined sequence, gives rise to the possibility of rationally designing structures from DNA. Let’s take the simplest example. Imagine 2 single strands of DNA; let’s call them Scaffold and Staples. As indicated by name, Scaffold has more bases (let’s say 60) and therefore is a larger DNA sequence than the staple, which has only 20 bases. Now the Staple can only bind to 20 complementary bases on the Scaffold. However the trick here is, knowing the sequence of bases in the Scaffold in advance, we synthesize the staple such that half of the staple’s bases are complementary to 10 bases at the Scaffold’s one end, and the other half are complementary to Scaffold’s other end. In this way, in the process of stably binding together, the scaffold will bend so as to accommodate the much shorter staple. This will give rise to a U-turn, a structural feature missing in the original scaffold strand. Similarly, by using 2 staples which attach at different locations on the scaffold, you could create a U-turn in the opposite direction. Now the scaffold will look like an ‘S’. Using short DNA strands (usually called staples) to bend a large DNA strand (called scaffold) into desired shapes is essentially the technique now called DNA Origami, due to its similarity to the eponymous paper folding technique (Figure 2).
Going back to the problem of accurately placing objects in fixed positions at the nanoscale, the DNA origami technique allows us to do exactly that with nanometre precision. Consider an example. We can synthesize these staple strands such that a portion of them is complementary to the scaffold and a portion just consists of multiple ‘A’ bases. At the same time, we can attach short DNA strands that are just multiple ‘T’ bases on the objects that we want to arrange on the DNA origami. The T-strands will bind with the complementary A-strands. Since we can synthesize any of the staple strands with these extra ‘A’ bases, we can choose where the objects with the T-strands will attach.
2. DESIGN AND EXPERIMENTS
Just like a piece of paper can be folded into complex and interesting origami structures, a long “scaffold” strand of DNA can be bent and folded into a desired shape, held together by short “staple” DNA strands. Because of the similarity, this is called DNA origami. DNA origami can be designed in elaborate ways and functionalities can be incorporated into the folded DNA allowing for basic locomotion, actuation, sensing or computing forming the building blocks for a DNA-robot. The sequences of the staple strands determine which part of the scaffold they bind to and bring together. It would be very time consuming to calculate the correct sequences by hand, but luckily there are software tools, such as caDNAno, available to help.
The way that caDNAno operates is by starting first with the structural design of the desired figure. Choosing between two possible layouts (honeycomb or square), you distribute the DNA helices that will be used for building your origami in a cross section. Following this, in an adjacent window, you determine the distribution and length of the scaffold strand in these “empty” helices. Once the scaffold is in place, the program adds the necessary amount of short staple strands to make the scaffold bend in the shape determined by the initial structural design (Figure 3).
Up until this point, no sequences have been considered in the design; so far the effort has been purely from a structural perspective. There is a set of commonly used scaffolds that have varying lengths and they are included in the program’s database, but you can use your own custom sequence if you so desire.
By introducing a “break” in the continuous strand of the scaffold, you create a starting point for its sequence. Because the base complementarity is an immutable law of DNA, the software can determine and list the sequence of the short complementary staple strands as soon as it knows the exact scaffold sequence. The list of staple strands will consist of a variable number of short DNA sequences, depending on the size of the origami you are making. The next step is to purchase the sequences, but it is always a good idea to run some preliminary simulations first as detailed in the next section.
With the stable strand sequences provided by caDNAno, you could simply order them and try them out in the lab directly. But a good way to save time (and money) would be to first run a simulation, verifying that your design works as intended. There are many simulation tools that can read your caDNAno design file as input, some more detailed than others. A less detailed (more coarse-grained) simulation model will run much faster, but it also tends to be less accurate, so you have to pick a model that suits your purpose. One of the fastest and easiest simulation options is CanDo. It will give you a plausible 3D structure and a nice movie, but the model lacks excluded volume, meaning that DNA strands can pass through each other without colliding, which will never happen in the lab. A relatively new option is mrDNA web tool, which simulates your structure at increasingly less coarse-grained levels, transitioning from fast to more accurate. Although it has excluded volume, it still ignores hybridisation, meaning that helices cannot form or unzip themselves during the simulation. If you care about the thermodynamics of the origami assembly, oxDNA simulations are the way to go. Less coarse-grained than the methods above, because it models individual nucleotides, oxDNA can tell you how your DNA will react to being melted, pulled apart1, or (more nicely) how it goes about walking2.
To get even more accuracy you could, of course, try to simulate every single atom in your DNA robot, but for that you would need a supercomputer and a lot of time on your hands.
When you are satisfied with your simulations, you are finally ready to order your strands. There are different companies online from which you can buy the customized sequences by sending the list to them, along with other technical specifications. They can produce the short staple strand sequences with perfect fidelity out of individual, chemically synthesized nucleotides and after about a week of waiting, you will receive your collection of synthetic DNA strands with the exact sequences required for your design. Now you are ready to go ahead and produce your DNA nano-robot in the laboratory.
3.1 Nano-robot production
So, you have finally finished the design of your nano-robot with the aid of the really handy software tools described above and have just received the ordered robotic “parts”, the staples. It is very simple to produce the robot from this point forward, but you need a set of starting materials and tools to do so.
You may remember that, apart from the staples, a scaffold is needed. These are commercially available, but you can also make them yourself from plasmids grown in bacterial cultures. Salts will also be needed, magnesium is usually the go-to salt, but sodium can be used as well. Regarding the tools, a machine like a PCR machine that can produce and control certain temperatures with great precision and then reduce them in a gradual fashion is also needed. The first step in our recipe is to mix the scaffold with the staples. Usually the staples are added in excess, to ensure that they all occupy their determined positions. Along with the DNA, the salts are also added. Their role is to allow the sequences to get together close enough so that the design can be materialized.
The driving force of the self-assembly of our nano-robot is heat. More specifically, a decreasing gradient of heat. Therefore, once you have the mix ready, you introduce it in the machine that will produce these gradients or “ramps” of heat. Starting off at the highest temperature and lowering it over time, you allow for the staples to find their complementary sequences in the scaffold and bend it into the designed shape. This process takes some hours to complete, after which some purification will have to be done as the excess of the staples can interfere with later steps. Bellow you can see how the process plays out (Figure 5).
3.2 Nano-robot characterization
What you have in your hands right now is a large number of your nano-robots floating in solution. But, as they are right now, it is impossible to tell if the procedure has gone according to the design or if some unexpected interactions has caused deformations. There are different methods of checking, or characterizing, the nano-robot. Each of them has their advantages and caveats and which technique is best depends on your structure. Here we present you a small list with some of them.
3.3 Atomic Force Microscopy (AFM)
One of most widely used techniques for producing images at the nanoscale is the Atomic Force Microscope (AFM). In this technique, a probe tip (which is a few nanometres wide) hovers over the surface of the sample to be studied. When the tip approaches a structure in the sample surface, attractive and repulsive forces are exerted on the probe. These atomic forces are measured by a laser that is constantly pointing at the end of the cantilever bearing the probe tip. When the tip, and thus the cantilever, scans over the sample surface, its up- and downward movements are detected by a laser detector. In this way, the detection of the reflected laser beam allows the microscope to construct a precise topography of the nanorobot at a nanometer scale by detecting the changes in laser reflection from the cantilever, which are then used to generate an image of the sample (Figure 6).
There are both some advantages as well as some drawbacks to this technique. The biggest advantage of AFM imaging is the relatively simple generation of 3D images that does not require special sample treatment, which might damage the sample itself. However, the obtained 3D information is limited compared to alternative methods such as TEM and cryo-EM because it only shows height differences without additional information such as the internal folding. Compared to other methods, the scanning speed is slower and the scanned area is smaller. Another disadvantage is that the scanned molecules are adsorbed on a flat mica surface and might display different properties compared to those they have when they are freely floating in solution.
3.4 Transmission Electron Microscopy (TEM)
Another type of microscope that can be used for the characterization of our nano-robot is the Transmission Electron Microscopy (TEM). Much like the previously explained AFM, it does not use visible light to visualize the sample. Since our nano-robots have sizes measured in nanometres, a million times smaller than a millimetre, visible light is insufficient when trying to study these structures. Instead TEM uses electrons for visualisation, which, due to their physical characteristics (wavelength), have greater resolution and are able to image the nano-robots.
Before using the TEM, the nano-robots must be pre-treated so that they can properly interact with the electrons that will be shot at them. The electrons will mostly only go through the DNA origami sample and not the area around it and the electrons passing through can be analysed to obtain a projection of the sample. Having a projection means that, instead of getting a direct image of the sample, you get the shadow against the wall of it. Because these electrons go through the nano-robot, some internal features can usually be seen.
Another useful way to characterize your structure is to make it fluorescent, to emit light, to signal different events. You can attach certain molecules to your robot that absorb and emit light of different wavelengths and a common setup, called FRET, requires two different molecules: The first one absorbs the light you shine on it and, if it is close enough, transfers the absorbed energy to the second one, which will then light up in a specific colour. This way, you can tell if the two molecules in your nano-robot are near each other or not. There are also variants where the second molecule quenches the light emitted from the first when they are close, but the principle is similar.
You could, for example, put the two molecules at the ends of two single DNA strands and know that they have hybridised (formed a double helix) when you see your sample glow with the intended colour. If you have arms on your robot that open and close (perhaps it is doing yoga?), you will know that the arms are close, because the sample glows when the fingers entwine.
So far, we have covered how to design, assemble and visualize our DNA-robot, but that alone would not make a complete robot. As discussed in previous chapters, a robot must be able to perform complex tasks. One way to achieve this is to functionalize our DNA nanostructures. In this context, functionalization refers to the process in which new features, properties and capabilities are added to our initial structure using chemical, biological and physical means. Modern technologies in these fields allow us to attach different specialized molecules with specific functions to DNA.
Even though synthetic DNA is traditionally produced using bio-molecular methods (like PCR), it is now also possible to make short DNA-strands (oligomers) by a process that relies completely on organic chemistry synthesis. This method brings with it a greater freedom to customize the DNA because different functional groups can be added to specific sites of the oligomer. One way to do so is through bio-conjugation where the addition of a functional group enable us to permanently link our DNA-oligomer to a biomolecule with specialized biological functions such as antibodies, enzymes and peptides. A DNA-Biomolecule conjugate can then be attached to our DNA-nanorobot through complementary base pairing, thus adding new functions to our nanostructure. Just like an electric drill can change attachments to be functionalized for different jobs, the use of bio-conjugation allows us to couple different “attachments” to our robot to ensure that more complex and specific tasks can be carried out.
4.2 Nanoparticles as DNA functionalization
New and interesting applications arise when inorganic chemistry and material science is combined with DNA technology and one example is DNA-functionalized nanoparticles . Nanoparticles are inorganic nanoscale particles (usually metallic crystals) that exhibit unexpected physical and optical properties. The combination of DNA’s programmability and ability to self-assembly with the special properties of nanoparticles enables us to assemble more complex structures.
The complex combinations of inorganic particles with DNA have potential applications in many fields including sensing, gene delivery, medicine and optics and some of these have already been realized. DNA- functionalized quantum dots are for instance used in in vivo and in vitro imaging due to their high fluorescence and excellent photo-stability. These nanoparticles are part of the arsenal we can use to functionalize our robot for the different tasks to be performed.
4.3 Small molecule actuators
So far, we have covered some tools that we can attach to our nanorobot to increase its functionalities. However, it is also possible to modify the main scaffold with specialized small molecules able to perform different task and using organic synthesis, different moieties can be inserted that respond to different stimuli. For instance, some molecules can stretch or expand upon specific light irradiation, others can change conformation upon the addition of a consumable reagent (just like a fuel), while others exhibit changes depending on the pH, temperature or salt concentration. Introducing these kinds of moieties in the design, allow us to make our robot more programmable and capable of carrying out specific actions.
5. FUNCTIONS AND APPLICATIONS
5.1 Giving our robot a good start
A nano-robot without any functions is not really a robot. In this chapter, we will guide you through some of the functions that our nano-robot potentially could have. When reading this, you should imagine that each of these functions can be built into a single module (much like LEGO bricks) and then different module can be combined to create a modular nano-robot with very specific functions according to the modules it is built from.
Now imagine that we have created a cool nano-robot, which is able to do many of the things we describe below. If we want to send our nano-robot to work inside a patient we must protect it against all the dangers of a living organism, because otherwise it will be shredded into pieces although it is ever so cool. The DNA that makes up our nano-robot is the target of DNA degrading nucleases, enzymes found everywhere that chop in into small pieces of DNA, and the DNA itself may even induce an inflammatory response. Of course, we want to avoid both of these scenarios.
Trying to avoid that our nano-robot becomes a nuclease victim, we can try to mimic strategies found in nature. A family of viruses called “enveloped viruses” are remarkably stable inside the organisms due to their “envelope”, a special lipid membrane that masks them from the host’s defence mechanisms.
5.2 Protecting DNA nanostructures: Examples from the lab
Inspired by the enveloped viruses, a research team from Harvard mimiced the lipid membrane and coated DNA to improve its stability3. The DNA itself formed an octahedron to better mimic the virus’ spherical shape. On the surface of the octahedron, short DNA strands are exposed, and they can bind complementary DNA strands, which are functionalized to assemble a lipid coat on top of the DNA (Figure 9). If you can picture the exposed DNA strands as the loops of a Velcro system and the complementary DNA strands as the hooks, the hooks have small lipids on the assembled structure.
The researchers first tested the stability of the lipid-coated DNA octahedron, before they labelled it with fluorescent molecules and introduced it in a mouse model to see how it behaved compared to a non-protected octahedron. The results showed that octahedrons without a lipid coat accumulated in the bladder, which indicates that the particle is quickly eliminated. In contrast to this, the coated octahedrons were distributed throughout the mouse (Figure 10).
6. Sensing and Signalling
If we want our nano-robot to perform specific complex tasks based on a given input, it must be able to sense and react to its environment. Sensors in robots, just like your senses, respond to something in the environment, transmits the information, before they provide a given output. Take for instance your sense of sight: Your eyes focus light rays, determines the information in neuronal cells inside it, transmits the information to the brain, where it is processed and interpreted to form the images we see. From this example it is clear that sensing and signalling are very tightly connected – there is no use in sense something, if the signal cannot be transmitted. In macroscopic robots that we know from our everyday life, engineers have tried to mimic natural sensory systems. Depending on the task to be performed, sensing modules are integrated in the robot design and they generally consist of a three basic parts: A transductor, which amplifies a given signal, and it is connected to an electric circuit, and together they transform a physical stimulus into a quantifiable electric signal communicated to a last part, a central processing unit. In modular nano-robots, biosensors are often used as the sensing module an through an integrated biological component attached to a physicochemical detector, they allow the detection (or perhaps even the measurement) of minuscule amounts of analytes. Because our nanorobot is built from DNA, it is clear-cut that we will use DNA-based biosensors. The current technology allows us to exploit DNA’s versatility to build modules responsive to different molecular cues, and so far researchers have successfully detected metallic ions, pH changes, small molecules, proteins, and antibodies.
6.1 Sensing and signalling: Example from the research lab
We are still in the early process of making a modular robot. However, examples already exist of relatively simple DNA-based robots that are able to sense specific signals and provide an output. To understand our example, please recall from previous in the text, that DNA is both stable, programmable, and it may undergo specific changes when it binds to complementary DNA. Biosensors can take advantage of these characteristics, for instance in aptameric sensors. Aptamers, which are short nucleic acid oligomers with well-defined 3D structures, change their shape when they bind a specific molecular target (Figure 11).
Taking advantage of the aptamer switch technology described above, researchers have designed a simple barrel-shaped nano-robot, which is able to carry a small cargo inside (Figure 12). This hexagonal cage robot consist of two halves permanently attached via a hinge only in one side to ensure that the halves can be opened and closed depending on the presence of a specific signal4. The other side of the two halves have protruding DNA strands that comprise a lock mechanism based on the DNA aptamer switch technology and it has been designed to respond to signals present in its environment. If a certain signal (consider it a key for the lock) is not present, the aptamers are closely entwined in a helix. When the signal is present, on the other hand, they change conformation and the helix opens, so does the cage, and the cargo placed inside is released. If you want to be very specific about when the cage can be opened, it is possible to design two different locks. Now you will need two different signals to be present before the cage is opened. The signals required for opening can vary. As an example, specific activating cells can either carry the required signals on their surface or release the signal molecules into the environment. In this way, the nano-robot can go searching fort he signals required before it releases its cargo, which can also be designed to have different functions. For instance, the cargo may bind to specific molecules (such as cancer cell markers) on the cell surface and emit fluorescent light upon binding, thus selectively labelling the cells that present the suitable keys. If the nano-robot carries drugs or biologically active cargo, it will be able to scavenge for a specific delivery site before releasing its cargo.
Knowing that we can create modules for sensing specific signals, it would be handy to also have module that makes the robot able to move because it can then actually “go” searching for signals. To achieve this, the nano-robot would need an actuator, which is commonly known as the part of a machine responsible for controlling its movements. With some source of energy as input, the output of the actuator is either a rotary or a linear movements. In simple terms, the actuator is the “muscles” of a robot. Rotary actuator movements from everyday life, includes electric motors, which use electrical energy to rotate a wheel. The rotary motion of a helicopter blade providing the upthrust and the rotary motion of an old watch gear that helps you keep track of time are other examples. Linear actuators, on the other hand, are used to control modern industrial robots.
Although you will probably recognize the man-made examples from above, actuation modules are also found in Nature’s own nano-machines. Take for instance the famous ATP synthase, which is found in all cells and that generates molecular energy in the form of ATP. Similar to the macroscopic actuators mentioned above, the ATP synthase requires energy to generate a movement. Because it is always located in a membrane that separates two compartments with different levels of electrochemical energy, it can harvest energy from this gradient to power a rotary movement within itself. This can in turns power the recharging of the ATP energy storage molecules vital for cell survival.
7.1 Actuation: Example from the research lab
Inspired by Nature’s ATP synthase and other nano-machines, researchers are creating actuation modules that will become important parts of nano-robots in the future. One example is a nanometer-scale rotor designed as a 3D DNA origami.5 The module self-assembles into a structure that resembles a helicopter blade and it consists of different subunits: a rotor component, a stator, and an axle (blue, red and white, respectively in Figure 13). The rotor assembles with a stator (platform) and it can rotate freely on an axle, thanks to Brownian motion. This early, but its structurally complex, attempt of mimicking the ATP synthase’s rotational movement is an important first step for the making of actuation modules for a nano-robot5.
A nano-robot with both sensing/signalling and actuation module can turn to reach a signal, which it senses. However, if the signal is too far away it will not be able to neither sense nor reach it, if it does not have a translocation module. Translocation, or the ability to move from one place to another, is therefore incredibly important because it supports and improves the other modules. When you take a look around, you will see animals (including yourself) moving around using their limbs, fins, and wing and even in the microscopic world, some bacteria grow external appendages (somewhat similar to the structure shown in Figure 13), which allow them to move. If you look even further, inside our cells, you will find an example of a natural nano-robot: Kinesin. This protein is present in all of your cells and using what you may think look likes two legs, it ‘walks’ along the cell’s internal scaffold. Fuelled by ATP, kinesin walks from place to place delivering its cargo.
The eventual applications of nanorobots, which are able to move, are multifold, but they are believed to be very promising in disease diagnosis and treatment. Just imagine a nano-robot that both moves around, senses disease biomarkers, and elicit a specific immune response by delivering the necessary drug. This would have an immense impact on global health.
8.1 Translocation: Examples from the lab
Because kinesis is a natural nano-robot, many scientists have been inspired by it to explore whether nanorobots can move in a similar way. Researchers have designed a nano-robot that walks on a flat DNA Origami surface, where it picks up multiple cargos and accurately deliver them at predefined sites. Its movements can be switched on and off using DNA-DNA interactions, but it moves 100,000 times slower than the kinesin. Researchers are currently testing different strategies to improve the speed of the robot. Another example is a DNA arm controlled by an electric field where the researchers used a flat DNA origami structure as the actuator unit. The arm was a six-helix bundle attached to the surface via a small loop in the bundle.6
To move the arm, they took advantage of DNA’s inherent negative charge. If you apply a positive electric field to the system, you can move the arm around the surface of the actuator within milliseconds, just like you can pull a magnet around on a tabletop if you use the opposite pole of a second magnet. Because the structures are too small to see with your naked eye, small fluorescent signals are attached to one end of the arm. The researcher can see this light in a specialized microscope to find out how the arm behaves, when you move it with the electric field (Figure 14). Compared to simpler set-ups, where scientists have achieved translocation at nanoscale, the nano-robot arm was at least five order of magnitude faster, and they aim to expand and improve the system to transport a cargo between different arms.
If we are able to mix’n’match between all the functions described above, we can create rather amazing nanorobots. But we still need to make sure that the nano-robot know what to do and when to do it and this is where computing comes into the picture.
In general terms, computing refers to any activity that creates, uses, or benefits from computers and as such, computing is already an integral part of any robot. DNA computing, on the other hand, is a discipline where researchers use DNA as computer hardware based on DNA’s inherent advantages. First, DNA strands have a high density of information, which allows a DNA computer to perform a very high number of parallel operations. Second, the Watson-Crick base pairing described in the beginning are very helpful from a computational point of view because we always know which bases are opposite to each other in double stranded DNA. The computing component of our nano-robot should allow it to make decisions. For instance, IF the robot senses “A”, THEN it must act by doing “B”.
So far, there are a a range of different strategies for creating small DNA computers. One is to use nucleases that chop up the DNA strands, where the information is encoded. You can imagine that the nucleases are the hardware of a traditional computer and the DNA is the programmable software. Another option is to use sequences of DNA that change shape when they sense the right input and the altered shapes of these DNAzymes can then be measured directly. A third strategy is to use toehold mediated strand displacement, where two DNA strands compete about binding to a third strand. One of the first two DNA strands binds to a single stranded end of the third strand, the toehold, but the second strand is able to displace the first strand by a mechanism, which does not require any input from outside the system. The third strategy with the toehold-mediated strand displacement has been used to design logic components. In macroscopic system, the computer interested reader may know these as logic gates, which conduct electricity based on a given rule. Different kinds of gates, with each their own rule, are normally present in electronic circuits but the simplest is the AND gate. In AND gates, an output is only given if you have two inputs present at the same time. This gate can be built using only DNA strands and is composed of three strands. The output strand is called E(out) and is partially complementary to the long F strand. The other part of the F strand is complementary to the G strand. The input strands are called F(in) and G(in). The F(in) is complementary to the F strand and the G(in) is complementary to the G strand. When the G(in) is added to the mix, it can recognize and bind the G strand, but there is no output. If only the F(in) strand is present, there is no output as well. So, the gate is on only if the G(in) and the F(in) strands are present together; the G(in) strand can bind to the G strand and free the toehold region for the F strand. At this point, the E(out) strand is free and can be detected (Figure 15) .
9.1 DNA computing: Examples from the lab
Sorting is all around us: Children sort LEGO bricks by colour, ants sort their larvae by size, and proteins are delivered to different places in the cell after being synthesized in the same place. Thus, arranging objects by an arbitrary recognizable feature is an integral part of life. Our first example of DNA computing from the research lab is therefore a sorting robot. To design the robot, researchers had to brake down the sorting problem into distinct functionalities. First, the robot has to be able to move between different locations. Second, the robot must be able to pick up cargo of different types. Third, it must be able to recognize and drop off the cargo at a goal of matching type. During the design process, the researchers prioritized reliability over speed and smartness to limit the complexity of information processing, which also resulted in a quite small nano-robot compared our earlier examples.
The robot moves around as a random walker by wandering around aimlessly, displacing itself from one partially complementary track position to another on a DNA origami platform with short single stranded DNA strands as tracks.7 Some positions on the platform contain a cargo to be picked up or goal strand, which indicates a drop-off site for a given cargo. Each cargo has a type-specific domain and a general cargo-domain. If the robot happens to bump into a cargo, it will pick it up through strand hybridization with the general domain, it does not differentiate between different types of cargo. Once it has picked up a cargo, it will again walk randomly until it bumps into a goal location. If the type-specific domain of the cargo matches that of the goal, the cargo is dropped off permanently using strand displacement. The robot continues its random walk, repeating this process until all cargo is sorted.
It is at this last step where the genius of the robot’s design shines through: Processing a match between cargo and goal is not done by the robot, but solely by the goal strand, which takes away the cargo from the robot if, and only if, it has the correct type-specific domain. You can compare this to the kid sorting LEGO bricks by colour: The kid has to recognize many different colours (types) and has to decide which pile of LEGOs match to the block (cargo). It would be a much easier task if the kid only had to walk by the piles and the blocks automatically dropped from the hand once they have reached the correct pile. The latter is what happens to the sorting nano-robot. As of now, the robot only sorts arbitrary DNA sequences but it would be interesting to expand the nano-robot’s cargo repertoire to functionalized DNA, aptamers, and eventually active cargo such as proteins.
Our next example of a nano-robot is composed of both a sensor and an actuator connected to and controlled by an information processing unit. This nano-robot is more abstract than the previous examples because the three different functions are not found in one integrated structure but are distributed over separate DNA-strands in solution. In fact, to call it a robot might be a bit of a stretch, so “multiplier” is a better name. The multiplier can preform multiplications with numbers between 0 and 9. As numbers it uses DNA-strands containing an identifier-sequence representing a number ranging from 0 to 9, but the result are displayed as an Arabic numeral. They provide multiple display scales, both macroscopic and nanoscopic. We will focus on the latter, as this employs the already familiar DNA-origami template. Interestingly, rather than processing the input signals through a network of logic gates to calculate a solution (used for instance in Winfree and coworker’s DNA-based square root calculator), the multiplier operates as a lookup table where the correct result is selected from a library of potential outputs. It is comparable to small children memorizing multiplication of single-digit numbers from a table, rather than doing the actual calculation in their head. This is called a combinatorial approach, rather than a computational one.
There are three functionalities involved in the multiplier: the selection of the right result if and only if the X and Y are both present, selecting the right translator strands which are needed for the downstream display, and finally the readout generation. As mentioned before, these functionalities are distributed. Sensing is performed by hybridizing X and Y directly (the approach shown in the figure), or in a more modular fashion with a linker strand. The linker binds X and Y in a stable complex that exposes their unique identifier domains (red and green). Only the combination of two identifier domains can bind the correct result strand, resulting in a bigger complex. This complex is then added to a library of translator strands, and takes only those that have a domain complementary to a domain on the result (cyan). The entirety of complexes is incubated with a flat DNA-origami sheet. The translator strands contain domains (purple) complementary to specific locations on the origami. The entire complex binds to only those positions on the DNA-origami, thus revealing the Arabic number when analyzed with AFM (Figure 16).
Gothelf et al. showed the proof-of-concept by making a mathematical multiplier, but it should be noted that their work can cover far more application.8 In general terms it is a method to retrieve a relation between two inputs that has been encoded in a result library, and to display that relation in an immediately readable format. A possible application would be to make the multiplier sense two disease-marker oligonucleotide inputs of biological origin. The multiplier could then select an appropriate downstream response to that specific combination of biomarkers from a preprogrammed library.
Now you should have a fair idea about what we are working on and as you probably have figured out by know, a modular robot can be designed to carry out many different tasks. But before we can put a nano-robot into use, it is very important to take some ethical issues into consideration. It may sound boring… But it is not. Therefore, in this last chapter, we will answer some of the ethical questions that we are often asked about DNA nano-robots when we engage with the general public.
10.1 Are DNA robots GMOs?
Nope. Although there is no clear border between living organisms and other biological entities such as viruses, current definitions of life normally state that a living being has a metabolism and the ability to reproduce and grow. These vital functions are extremely complex and we are by no means able to create life from scratch by simply mixing components together. The current state of DNA nanotechnology is more comparable to sculpting than to life. We are getting better at creating various shapes on the nanoscale but, just like sculptures, a DNA origami structure by itself is nothing more than a material in a particular shape. Moreover, the biological function, which DNA is famous for, relies on the interplay of many hundreds and thousands of other molecules. By itself, DNA is not “alive” or able to bring dead matter to life.
The DNA sequences used in DNA origami are often entirely synthetic or genetically modified versions of natural sequences derived from phages, which are viruses that infect bacteria only. To produce these sequences, we use genetically modified bacteria and a safe routine method that has been used by scientists for decades. The sequences from these bacteria are modified and some encode phage proteins, but others do not encode for a functional protein at all. The genetically modified bacteria used in the experiments will be killed either by the phages they themselves produce, or by heat sterilisation at the end of the experiment.
10.2 Is it self-replicating?
No. The structure of DNA allows for it to be replicated by cells, but this process requires the interplay of many other proteins and molecules. In a standard experimental setting, or even if injected into the blood stream, the required components will not all be present and replication of the DNA is thus prevent. Furthermore, the way DNA itself is folded into complex shapes is another reason why replication outside controlled experimental settings is extremely unlikely. To replicate a DNA, short single-stranded snippets of DNA must anneal to the usually much longer single-stranded DNA. The short strand binds to the long single-stranded template before the replication machinery travels along the longer strand while replicating it. In a DNA origami, the folded structure is locked into place and the replication machinery is simply not able to travel very far. At most, one might receive short snippets of harmless single-stranded DNA, but no fully replicated DNA robot.
10.3 Isn’t DNA a dangerous thing to use?
DNA is everywhere. Every animal and every plant contains DNA in every single one of its cells. This means that nearly everything we eat contains DNA. Yet food does not make us sick, because DNA alone is a harmless chemical molecule. It is only when it encodes for dangerous proteins and finds a way into our own genome that it can be a safety concern. Naturally, none of the sequences we use for the construction of DNA origami objects encode for anything dangerous. Even if we did, the folding of a DNA origami ensures that proteins, that may or may not be encoded in a folded DNA origami structure, cannot easily be produced. At the very most, one can expect to find truncated proteins that are unlikely to have any function at all.
Because of its ubiquitous nature and important biological role, our bodies have developed powerful mechanisms to destroy potentially harmful DNA. This is also the reason why everybody working with DNA is wearing gloves: It is (often) not that the DNA is dangerous to us, but rather that we are dangerous to the DNA. The proteins found on our skin rapidly destroy DNA. For potential applications in medicine, DNA nanostructures must be stabilised to prevent their rapid destruction, and the same stabilisation may further prevent the DNA from being read by a cell’s replication or protein production machinery. Our DNA nano-robots are quite fragile in nature and require controlled conditions, which means that unless stabilised, they would fall apart even in pure water.
10.4 What are the worst-case scenarios?
When considering DNA as a material for treatment or function, several things can theoretically go wrong with technologies from the near future.
One problem that comes to mind is that it is possible for the DNA nano-robot to gain unfavourable functions or accidental side effects. For example, imagine a structure was made to deliver drugs to the lungs, but instead deposits them in the intestines. In this case, the problem likely originated from the design of the structure. Luckily, we will be able to see these effects early on, during the initial experiments, like with any other drug. Furthermore, it is very unlikely that a structure “develops” a new function. This is because DNA structures are not suitable for long term use in the body, they get degraded and excreted relatively quickly. This makes it close to impossible for them to actually mutate sufficiently to gain new functions.
Another potential problem is clogging of the structure. This is due to the nature of DNA folding. When making DNA structures, some structures stick together and aggregate into “clumps” of structures. Now, this is a step long before actual application, meaning the correct structures can be purified before ever injected into a living being. It is possible that under some conditions, DNA structures start to aggregate after injection and “clog” areas in the body. Although a possibility, these effects can be relatively easily counteracted. Namely, by adding enzymes that degrade DNA, the structures and subsequent aggregations can be undone. Furthermore, the immune response will also remove these structures using similar methods.
This brings us to the next possible problem, the immune response itself. It is possible that DNA structures can promote an immune response or allergic reaction. As stated before, the structures can be treated with DNA degrading enzymes to remove them. Also, immune suppressors will prevent any major harm from the immune system itself during an allergic reaction. Note that the actual application of DNA structures in living beings is still very much in the research stage, so many of these potential problems will be ironed out eventually.
Finally, one problem that people can fear is that the DNA starts becoming functional as information-holding DNA, rather than just a 3D structure. We already discussed above why DNA structures cannot replicate. For very similar reasons, the DNA also cannot become functional, as it would require the exact same conditions. Furthermore, all DNA sequences used are screened against a database of known “dangerous” genes (e.g. ones from viruses or harmful bacteria). This way the chance of accidentally using a dangerous DNA strand is also highly reduced.
Ironically, the most likely “worst case scenario” is the loss of function, rather than gain of function. That is, the DNA nano-robot will not actually perform any function and just be there, hanging around, and doing nothing. Now this case is completely harmless, as it will be removed from the body relatively quickly and will not do anything at all.
10.5 Is it immoral to use DNA structures?
The answer to this question will obviously depend on your personal moral values, but in most cases the answer will simply be “no”. This is because the generation of these structures and their intended purpose is at least equal to, and in many cases better than, regular drug development. The intended purpose of many DNA structures, at least in the biological context, is usually to replace or enhance drugs and to limit their side effects. Whereas the non-biological ones are often diagnostic or simply technical in nature. None of these intentions are in any way more harmful or ethically debatable than traditional medicine or mainstream technology. Furthermore, the generation of these structures is completely animal-free, and the function is tightly controlled. To then say that it is “immoral” to use DNA in any of these contexts, would mean that it is also equally immoral to use proteins and other drugs or pursue any progress in the medical field.
10.6 Aren’t you playing God when working with DNA?
Not really. To play God would imply that you would have the power to either create life, or substantially affect it. Now to address the first part, we simply cannot create life from simple DNA strands, especially when they are as interlocked in the DNA origami structures. As for the second part of the ”God-playing” definition, this technology can impact one’s life significantly, but only in the same sense as traditional medicine can. So, this answer will again depend on your personal interpretation of life. If making medicine to cure diseases or performing surgery to save a life is playing God, then using DNA nanotechnology will likely have the same answer.
10.7 Is this a precursor to a more dangerous technology?
This is a rather difficult question to answer, because literally any technology can be transformed into something dangerous. But in DNA nanotechnology, the chances of this are extremely low for several reasons: Firstly, while robotics and computation will become more important in future iterations of DNA nanotechnology, the basis, i.e. the DNA, will remain the same. This leads us back to the question “Isn’t DNA a dangerous thing to use?”, to which the answer was “no”. Other, more long-term applications of DNA nanotechnology will not have biological applications and will therefore not pose a threat to us. Based on this, it is extremely unlikely for this type of technology to give rise to anything threatening in the context of weaponry. Even more so when considering the robustness and relative ease of other potential agents such as bacteria, viruses and toxic chemicals to achieve the same thing. Furthermore, the relative difficulty of “spontaneously” gaining new (and dangerous) functions is also much more limited than in its bacterial or viral alternatives, making it much less likely for it to “accidentally” become a dangerous technology.
10.8 Are there other ethical questions that might play a role in the future of DNA nanotechnology?
Currently and in the near future, DNA nanotechnology will be somewhat limited to simple mechanical and biological functions such as opening and closing or extending a part of its structure, making the structure itself relatively harmless. However, when considering the more distant future, ethical questions regarding robotics may start to apply, especially in the field of computation. This field is growing and will continue to do so in the foreseeable future. One day, DNA circuits and DNA computers may become a fully functioning and more widely applied method, due to its smaller size and potential complexity. The more this technology develops, the more relevant the question of the laws of robotics becomes. Again, because this technology is still very new, reaching the question of how to handle an Artificial Intelligence for example, is still very far ahead in the future and unlike more mainstream forms of programming and computing.
The writing of this work was supported by the EU MSCA-ITN grant number 765703
1. Engel, M. C.; Smith, D. M.; Jobst, M. A.; Sajfutdinow, M.; Liedl, T.; Romano, F.; Rovigatti, L.; Louis, A. A.; Doye, J. P. K., Force-Induced Unravelling of DNA Origami. ACS Nano 2018, 12 (7), 6734-6747.
2. Sulc, P.; Ouldridge, T. E.; Romano, F.; Doye, J. P. K.; Louis, A. A., Simulating a burnt-bridges DNA motor with a coarse-grained DNA model. Nat Comput 2014, 13 (4), 535-547.
3. Perrault, S. D.; Shih, W. M., Virus-inspired membrane encapsulation of DNA nanostructures to achieve in vivo stability. ACS Nano 2014, 8 (5), 5132-40.
4. Douglas, S. M.; Bachelet, I.; Church, G. M., A logic-gated nanorobot for targeted transport of molecular payloads. Science 2012, 335 (6070), 831-4.
5. Ketterer, P.; Willner, E. M.; Dietz, H., Nanoscale rotary apparatus formed from tight-fitting 3D DNA components. Sci Adv 2016, 2 (2), e1501209.
6. Kopperger, E.; List, J.; Madhira, S.; Rothfischer, F.; Lamb, D. C.; Simmel, F. C., A self-assembled nanoscale robotic arm controlled by electric fields. Science 2018, 359 (6373), 296-301.
7. Thubagere, A. J.; Li, W.; Johnson, R. F.; Chen, Z.; Doroudi, S.; Lee, Y. L.; Izatt, G.; Wittman, S.; Srinivas, N.; Woods, D.; Winfree, E.; Qian, L., A cargo-sorting DNA robot. Science 2017, 357 (6356).
8. Liu, H. J.; Wang, J. B.; Song, S. P.; Fan, C. H.; Gothelf, K. V., A DNA-based system for selecting and displaying the combined result of two input variables. Nat Commun 2015, 6.