DNA and RNA 2015; 2: 13–22

Research Article Open Access

Seth G. Abels and Emil F. Khisamutdinov* Nucleic Acid Computing and its Potential to Transform Silicon-Based Technology

DOI 10.1515/rnan-2015-0003 occurring molecular is a living cell. Cells Received July 21, 2015; accepted September 11, 2015 continuously process an enormous amount of signals Abstract: Molecular have existed on our planet (inputs) generated by a broad range of environmental for more than 3.5 billion years. Molecular computing factors such as temperature, pH, pressure, nutrients, devices, composed of biological substances such as nucleic signaling chemicals, macromolecules, etc. [1-5]. Once the acids, are responsible for the logical processing of a variety signal is processed, the appropriate response (output) is of inputs, creating viable outputs that are key components effected. Examples of such a response can include gene of the cellular machinery of all living . We have silencing, enzymatic activity, cell proliferation, migration, begun to adopt some of the structural and functional and apoptosis. In the field of , there is knowledge of the cellular apparatus in order to fabricate tremendous interest in the fabrication of artificial “smart” nucleic-acid-based molecular computers in vitro and in nano-devices that can autonomously perform functions vivo. Nucleic acid computing is directly dependent on similar to the physiological behavior of living cells. These advances in DNA and RNA nanotechnology. The field functions, which utilize macromolecules such as DNA, is still emerging and a number of challenges persist. RNA, and , include the storage, retrieval, and Perhaps the most salient among these is how to translate processing of inputs. a variety of nucleic-acid-based logic gates, developed Molecular computers use inputs of materials and by numerous research laboratories, into the realm of energy to achieve a specific purpose by repeatedly silicon-based computing. This mini-review provides some cycling them through certain states. They differ from basic information on the advances in nucleic-acid-based conventional computers in fundamental ways. Most computing and its potential to serve as an alternative that molecular computers are nanoscale objects, thus are can revolutionize silicon-based technology. subject to manifestations of universal physical laws which differ greatly from the macroscopic world of our senses Keywords: Molecular computer, RNA nanotechnology, [6-8]. For example, molecular computers have a small DNA nanotechnology, Logic Gates, Biocomputing, Silicon mass-to-volume ratio and move in viscous media rather based computing than in a vacuum or air. As a result, they cannot store momentum or kinetic energy as usefully as macroscopic devices do. In addition, they cannot store thermal energy for a significant period of time due to their extremely small 1 Introduction size, and must operate isothermally. Conformational changes in the moving parts of Molecular computers are natural and/or artificial molecular machines, driven by thermal agitation, devices in which macromolecules, including proteins create, modify and disrupt binding and catalytic sites and nucleic acids, mediate necessary functions. These for substrates, and after the free-energy landscapes that functions usually include three basic operations: govern their motions. This characteristic gives molecular sensing inputs, processing the inputs, and generating computers a distinct advantage, although harnessing specific outputs. The best-known example of a naturally this capability of nanoscale soft matter is proving to be one of the most challenging aspects of molecular design. Molecular devices can, however, store elastic potential *Corresponding authors: Emil F. Khisamutdinov, Department of energy with spring-like conformational distortions of their Chemistry, Ball State University, 3401 N Tillotson Ave, Muncie, Indiana 47306, USA, E-mail: [email protected] macromolecular parts or substrates [8]. They consist of Seth G. Abels, Department of Chemistry, Ball State University, soft, conformationally-flexible matter that can function Muncie, IN 47304, USA

© 2015 Seth G. Abels, Emil F. Khisamutdinov, licensee De Gruyter Open. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License. 14 S.G. Abels, E.F. Khisamutdinov as ratchet-and-pawl devices to bias molecular motions made of semiconductor materials like silicon are tasked of their substrates or flexible subsystems in desired with the primary role of amplifying and switching directions [6]. electrical power. The semiconductor has positively- In bionics, many engineering principles have been and negatively-charged areas. Electricity will not flow adopted from the study of biological machineries (6-8). between these areas unless the conducting channel is A few historical examples of bionics include fixed-wing open, as shown in Figure 1B. The channel opens when a aircraft, whose design was both inspired and informed by conductor, such as an insulated metal plate, is electrified. the study of birds in flight, and Velcro, which was based For electrons to flow though the conducting channel, on the hook-and-loop mechanism by which burs cling to there must be a source (input) and drain (output) made clothing and fur. As the bionic approach to engineering of metal. Even if the input is charged, it cannot flow to ventures more into the micro- and nanoscale, many fields, the output unless the gate (metal plate in the middle) from computer science to medicine to robotics, could be is also charged. When the gate is charged, it opens a transformed. For instance, consider ribosomal RNA. The conducting channel that allows the electrons to flow from study, incorporation, and even flat-out mimicry of the the negative to positive areas, or from the source to the features of this naturally-occurring molecular computer drain (Figure 1b, right panel). The most fascinating quality may help us clear hurdles and uncover answers (as well about this device is that there are no moving parts, in the as more questions) related to the design and behavior of mechanical sense. Electricity alone is used to perform nanocomputers. Current research in the fields of DNA [9-13] the computer’s functions, toggling between ON and OFF. and RNA [14-18] , as well as in molecular This turns out to be ideal for the assembly of logic gates engineering [19-21], support the model of nucleic acid - digital circuits which alternatively permit or deny the molecules as promising candidates to fabricate molecular passage of an electrical signal. The signal can pass only if devices or bio-computers. DNA and RNA molecules certain LOGICAL conditions are satisfied. As exemplified possess the programmable folding properties necessary to in Figure 2 A, an OR gate can be constructed by connecting carry out various applications ranging from a drug carrier two to different power supplies (inputs) and [22, 23] to nano-chips [24]. allowing a current (output) to flow from either or both of This mini-review focuses on the recent advances these supplies directly to a light bulb. The ON and OFF in nucleic acid computing. It is organized as follows: states, in the case of the switches, can be represented as an introduction to the simple operations carried out by “1” and “0,” respectively. If at least one of the switches computer microchips, including their limitations; a review is ON, the light bulb will be powered and in its ON (or of DNA and RNA computing as an alternative to silicon- “1”) state. An AND gate can be assembled using similar based computing technology; and a summary of, and an principles. For this, the output wire of the first outlook for, nucleic-acid based computing. needs to be connected to the input wire of the second transistor, as shown in the Figure 2B . In this case, both switches need to be ON for the bulb to be lit. The basic, 2 Silicon-Based Technology and its binary language of modern computers lies in the realm of Limitations Boolean logic, which is a mathematical system featuring just two variables, “1” and “0”. The common In computing devices, the input information is functions (AND and OR, as touched upon, but also XOR, mathematically processed into a digital signal. For NAND, NOR, and XNOR) employed in Boolean algebra, as example, in the case of a binary code, the basic unit of well as the truth tables which define each function, are information is written as a series of “0’s” and “1’s,” summarized in Figure 3. Note that some gates, such as indicating the two states of the logic circuit. To gain better NAND, NOR, and XNOR, are in their ON state by default, insight into molecular computer operations, we need to unlike the AND, OR, and XOR gates. In fact, NAND, NOR, look at the core of a computer – its microprocessor. The and XNOR gates can be viewed as inverses of AND, OR, microprocessor consists of different modules, each one and XOR gates, respectively. performing different operations, including adding and The overall computer performance (e.g. system storing numbers. The modules are made of numerous availability, response time, latency, throughput etc.) transistors, which are widely used in electronics, from roughly depends on the numbers of transistors embedded calculators to spacecrafts. A transistor is an electrically within the microchip or integrated circuit (IC). The driven switch that permits or denies the passage of capabilities of digital electronic devices have increased in electrons, as demonstrated in Figure 1. In brief, transistors lock-step with Moore’s Law, which posits that the number Nucleic Acid Nanotechnology: Promising Path to Altering Silicon Technology 15

Figure 1. Representation of a transistor - the core of any compu- Figure 2. Exemplified view of (A) an OR logic gate and (B) an AND ting device. (A) Example of a microprocessor chip showing variety logic gate, both emloying transistors. or colored modules. Each module or area of the chip contains multiple transistors, as demonstrated on the right. The transistor is composed of a semiconductor material (base), a negatively charged area (source), a positively charged area (drain), and a gate build of conducting material like metal as shown in 3D view. (B) Different transistors states, ON and OFF. The electricity will not pass from source to drain until the gate is open (ON), even though there is a high potential on the source. of transistors in a microchip will double approximately every two years [25]. This trend is poised to reach a limit, though, as the unending quest to miniaturize transistors is expected to come to a halt due to the quantum tunneling effect[26]. When the distance of a gate is scaled down about 10 nm, its electrons will jump spontaneously from source to drain, and the control over the flow of electricity will be lost. To overcome this problem, it is possible to increase the size of microchips or fabricate them layer by layer (forming a stack of microchips). Nevertheless, this will not solve the problem of keeping up with Moore’s Law; it will only delay the stalemate. In anticipation of this stalemate, new technologies offering alternatives to silicon-based computing are in high demand.

Figure 3. Summary of the common Boolean logic gates with symbols 3 Molecular Technology and truth tables.

One of the most promising alternatives to silicon-based variability, versatility, etc.) are just as important as, and computing is molecular computing; in particular, perhaps more important than, the quantity of molecules computing based on nucleic acids. Molecular computing used. This may obviate the problem of trying to fit an has an enhanced ability to provide parallel computation. increasing number of transistors into a decreasing volume Harnessing this ability may be the most effective means of of space. keeping up with Moore’s Law. The nature of the molecules There are many reports available in which researchers used to perform the computations (their complexity, have directly or indirectly used nucleic acids, both DNA 16 S.G. Abels, E.F. Khisamutdinov and RNA, to construct artificial architectures for a variety probed the possibility of developing a new generation of applications. These include “smart” devices capable of molecular logic gates and molecular computers based of doing simple and complex molecular computations, on the advantages of DNA molecules [20, 30-40]. DNA similar to those of transistors. Below, we summarize molecules are attractive from two main reasons - they are some of the papers that demonstrate the successful amenable to well-regulated, programmable folding, and employment of DNA and RNA to generate they have the unique ability to store genetic information. logic gates, which are the elementary building blocks of As such, DNA-based computing has been intensively used a digital circuit. for solving a variety of computational problems [41-44], e.g. Computing using nucleic acids is an integration of satisfiability problems, in which the computing time can biochemistry and molecular biology with the goal of grow exponentially with the problem size. The principle designing and creating algorithmic processes. To date, behind the use of a DNA duplex as computing material a variety of nucleic-acid-based logic gates have been is based on Watson-Crick (WC) base pair formations, built as defined counterparts of a biocomputer system, as shown in Figure 4A. Here, a set of DNA sequences (see references [16, 27, 28] for review articles). We briefly can be used as inputs, and the output is a duplex describe them here, starting with DNA biomolecules: formation which can be monitored using fluorescence, e.g. fluorescence resonance energy transfer (FRET) [45-49], or molecular beacons [35, 50-52] (Figure 4B). The 4 Computing with DNA molecules formation of a DNA duplex requires that the inputs be perfectly matched to the set of solution molecules. The The concept of DNA molecular computation was proven so-called strands hybridization approach can serve as in 1994 when Leonard Adleman demonstrated the ability an in vitro option for engineering a feasible logic-based of synthetic DNA oligonucleotides to solve a seven-point network [49]. Two-input AND, OR and NOT logic gates Hamiltonian path problem by performing a sequence of were previously constructed using a branch-migration logical operations [29]. Since then, many studies have

Figure 4. Demonstration the basic principles behind the use of a DNA molecules in logic gate operations. (A) Simplified view of DNA two strand hybridization, in which single-stranded DNA can be used as inputs and the duplex as an output. (B) An example of an AND gate demonstrating the molecular-beacon approach for monitoring DNA hybridization based on fluorophores and quencher pairs, see Ref # 35 for more details. (C) AND logic operation in vivo by DNA origami “nanorobot”. The construct opens only when Key A and Key B are both present in a cell, Ref 63. (D) Schematic design principle of AND logic gate utilizing reengineered Deoxyribozyme, Ref #68. Nucleic Acid Nanotechnology: Promising Path to Altering Silicon Technology 17 scheme with a mechanism built on strand recognition and Other examples of in vitro DNA-Logic-Gate strand replacement [37, 53]. Single-stranded nucleic acids engineering utilize the properties of deoxyribozymes were used as inputs and outputs in this work. The gate [64-67] (Figure 4D). To engineer, for example, an AND function was created by sequential base pairing, triggered gate, two different oligonucleotide inputs were hybridized by toehold-toehold binding between single strands and with corresponding controlling elements [66, 68]. This subsequent breaks. The combination of such molecular caused the cleavage of the substrate in the presence of events is similar to parallel computing, in which processes both inputs, leading to the subsequent conformational are carried out simultaneously rather than sequentially. change of controlling elements. NOT and XOR gates were The massive parallel computation power and colossal constructed using a similar idea [68]. memory capacity [54, 55] are among the most attractive Many studies have focused on engineering a logic qualities of DNA computation. Much progress has been gate using DNA in vivo. To this end, a core machinery is made in the performance and reliability of DNA computing usually selected based on gene expression regulation [66, [56, 57], though improvement in specificity of interaction 69-73]. One of the many examples use two inputs - beta- and chemical stability are likely to be necessary to make D-thiogalactopyranoside and anhydrotetracycline (aTC) these devices robust enough. - with a green fluorescent (GFP) as the output DNA self-assembly properties through WC base [74]. To construct this complex logic system, a DNA interactions can lead to the formation of a variety of plasmid was designed consisting of three transcription- structures, including an arrangement of tiles [58-60]. In factor-encoding genes (LacI, TetR, and lambda cI) with principle, the macroscopic self -assembly of different their corresponding promoters. The binding states of DNA-based tiles can be used to perform DNA-based LacI and TetR were modulated with the input molecules. computations. This was demonstrated by the one- This system was composed of five additional promoters dimensional algorithmic self-assembly of DNA triple- that were regulated by the three transcription factors. crossover molecules for executing four steps of a logical Two of the promoters were repressed by LacI, one was XOR operation [61]. The value of a tile, 0 or 1, could be repressed by TetR, and the remaining two were regulated modulated by the presence of a restriction site (e.g. Pvu II (positively or negatively) by lambda cI, resulting in 125 represents 0, and EcoR V represents 1). To relay the answer possible networks. Various GFP-expressing systems were after self–assembly, each molecular tile was embedded formed using a combination of different promoters, input with a reporter strand. The answer produced a barcode molecules and host strains, e.g. E.coli. A set of functional display on a PAGE gel. networks was fabricated with the logic operations NOR, DNA aptamers [35, 62, 63] can be used to build logic NOT, and NAND. A novel in vivo system, called the operations in vitro and in vivo. As demonstrated by , has been used to build permanent amplifying Yoshida et al., AND gates can be fabricated by fusing an AND, NAND, OR, XOR, NOR, and XNOR gates to control adenosine-binding DNA aptamer to a thrombin-binding transcription rates [75]. DNA aptamer in one construct [35]. The individual aptamer sequence binds to partially complementary fluorescence- quencher-modified , QDNA1 and 2 respectively. 5 Computing with RNA molecules When the two inputs, adenosine and thrombin, are bound The progression in the field of RNA nanotechnology makes simultaneously, the QDNAs release, causing an increase RNA molecules perhaps the most promising candidate in in fluorescence. Other input combinations, e.g. 0 + 0, 0 + the construction of bio-computers [76-78]. The reason for 1, and 1 + 0, lead to the presence of “0” or “1” QDNA and this is twofold: RNA (i) possesses the structural properties a weaker fluorescence. Similarly, other logic gates can be of DNA and (ii) mimics the functional properties of created, if the positions of the fluorophore and QDNA are proteins [14-17, 79]. At the beginning of its exploration, modified. RNA molecules were viewed as little more than an DNA nanorobots that can exist in open and closed intermediate (albeit a crucial one) with the main role of conformations are another example of the successful use passing information from genome to proteome in all living of DNA aptamers [63]. The principle of a DNA aptamer creatures. However, the discovery of non-coding RNAs lock mechanism is based on the binding of DNA strands revealed that RNA performs more versatile functions, to their antigen keys (Figure 4C). This lock functions as an including gene expression and regulation. This broadened AND gate, in which the aptamer-antigen activation state the view of RNA’s traditional role as a genetic intermediate. is the input and the conformation of the DNA nanorobot Indeed, the famous Adleman molecular computing (open or closed) is the output. 18 S.G. Abels, E.F. Khisamutdinov approach has been expanded to RNA molecules [80]. A between a protein-coding exon and an alternatively spliced molecular algorithm was developed and applied to solve exon (Ex) containing a stop codon. These sequences a chess problem, using specific ribonuclease digestion were followed by another intron, another protein-coding to manipulate strands of a 10-bit binary RNA library. exon, and the herpes simplex virus-thymidine kinase There are many examples involving RNA molecules that (HSV-TK) gene. The product of this gene was an activator can be viewed as computational algorithmic processes, of ganciclovir (GCV). Binding of β-catenin with the RNA including RNA editing (RNA sequence alternation) [29, aptamer led to mature mRNA which lacked Ex, leading to 81] and RNA-based regulatory networks [82, 83]. The latter the expression of HSV-TK. Early termination have been described as normal forms of logic function in occurred only when Ex was included in the mature mRNA, the form of input, logic gate and output [84, 85]. In most resulting in the synthesis of a nonfunctional peptide. For RNA computational systems, the inputs are small RNA the induction of apoptosis as output, both the expression elements or motifs, and the output is mRNA [82, 86, 87]. of HSV-TK and the presence of GCV were required. Later, There are various classes of RNA functional molecules, the same group produced AND, NOR, NAND, or OR gates such as ribozymes, RNA aptamers, riboswitches, miRNA based on RNA aptamers by constructing a high-ordered and siRNA, and orthogonal [88-90] that enable RNA device which included an RNA aptamer sensor, the fabrication of RNA-based nanoparticles to advance ribozyme and a connecting component [99]. modern technology [91-96]. Riboswitches are RNA elements that act by binding Ribozymes are catalytic RNA molecules that have a small molecule, which switches gene expression on or been shown to perform different functions in computing off [107-110]. Ligand-binding that can cause suppression processes [33, 97] For example, the hammerhead ribozyme, or activation of transcription and/or translation has been which site-specifically cleaves RNA strands [98], can demonstrated by Sudarsan et al. using a tandem riboswitch function as an actuator in an RNA computing device. In core machinery in vivo [111]. This machinery facilitates this case, the input-binding is translated to a change in sophisticated control of gene expression discovered in the activity of the actuator structure and the functional the 5’ untranslated region of Bacillus clausii metE. Two (active) state results in self-cleavage of the ribozyme [99]. naturally occurring RNA riboswitches bind independently The RNA device is placed at the 3′-untranslated region to two different metabolites, one to S-adenosylmethionine (3’-UTR) of the target gene, where ribozyme self-cleavage (SAM) and the other to coenzyme B12 (AdoCbl). This results in the inactivation of the transcript and therefore binding induces the transcriptional termination of a gene lowers gene expression [99]. Different signal integration of interest through cis-acting corresponding riboswitches. schemes act as various logic gates. Logic gates have been This serves as a NOR gate, as the absence of both inputs engineered with ribozymes in vivo [100, 101] and in vitro (SAM and AdoCbl) is required for the complete length of [102, 103]. Hybridization of two oligonucleotide inputs transcript produced as output. with the ribozyme leads to its activation, demonstrating Recently, particular interest in computing with an example of AND gate [102]. Using the ribozyme system, RNA molecules has been sparked by RNA interference Chen et al. were able to design a YES gate (an input of “0” (RNAi). RNAi is natural process that stimulates post- produces an output of “0” and an input of “1” produces transcriptional gene silencing in higher eukaryotes and an output of “1”) in vivo. The system was inserted into utilizes a class of short RNA duplexes and RNA hairpins the 3′-UTR of a target transgene where the ribozyme was [small interfering RNAs (siRNA) and micro RNA (miRNA) inactivated in the presence of theophylline, resulting in respectively [112, 113]]. With the introduction of synthetic expression of transgene [100]. siRNAs, the RNAi machinery can be harnessed [114, 115]. Another important class of functional RNA candidates In recent studies, it is been shown that it is possible for computing are aptamers, artificial RNA sequences that to design a synthetic gene network that implements have been selected in vitro through a SELEX (systematic general Boolean logic to make certain decisions based evolution of ligands by exponential enrichment) process on endogenous molecular inputs [116, 117]. Using inputs to tightly bind their ligands, such as DNA, proteins, other in the form of two groups of miRNAs and expression of RNA and small organic molecules [104]. This aptamer the hBAX protein as output, it is possible to build an property has been widely used to engineer input sensors AND-type logic gate, as demonstrated by Xei, et. al. [116]. in devices [99, 105]. For example, Binding of miRNA, along with the consequent repression RNA-aptamer-binding β-catenin has been used in the of translation, makes miRNA suitable to serve as sensory engineering of an AND gate in vivo [106]. To accomplish modules for RNA- and DNA-based digital logic circuits [53, this, the aptamer sequence was embedded into an intron 87, 116, 118]. Nucleic Acid Nanotechnology: Promising Path to Altering Silicon Technology 19

Collectively, advances in RNA biochemistry, computing? An increase in the number of elements in specifically in RNA nanotechnology, offer precise control RNA- and DNA-based logic operations is accompanied over the programmability of the size and shape of RNA by a drastic increase in noise [124, 125]. The common nanoparticles, as well as over the composition and approaches for noise reduction may include following: stoichiometry of the delivered inputs [23, 79, 119-121]. This (i) better optimization of individual logic gates, (ii) field is emerging and has contributed novel engineering utilization of network topology [125, 126], and (iii) for concepts for biological systems, as it has been utilized to even larger networks, the introduction of new network generate RNA-based algorithmic operations within a cell, elements to suppress the redundancy of the elements, within a test tube, or on a surface. thereby improving the signal-to-noise ratio. We still have a long way to go before we can claim to understand nano-molecular computers in sufficient detail to be able 6 Conclusion to “reverse engineer” an existing molecular computer or design an entirely new one. Biological computers possess several distinct advantages over silicon computers [122, 123] as they follow very different Aknowledgements: We thank Ball State University for principles than their macroscopic analogs. Biomolecules, providing Start-Up fund to establish Dr. Khisamutdinov’s including DNA, RNA and proteins are major elements in laboratory as well as BSU Chemistry Department’s CRISP logic gates operations. Their enzymatic selectivity (they program for supporting undergraduate researchers. often process a specific chemical function) gives them an advantage over silicon-based computing in both Conflict of interest: Authors state no conflict of interest specificity and usability in an intracellular environment. As most of the biological reactions controlled by specific enzymes are interconnected with other functional inputs, References it is possible to fabricate DNase- and/or RNase-based informational processing units. This processing unit can [1] Acker H. Mechanisms and meaning of cellular oxygen sensing be scaled-up to fabricate artificial biocomputing networks in the . Respir Physiol. 1994;95:1-10. [2] Daly K, Darby AC, Hall N, Wilkinson MC, Pongchaikul P, Bravo D, possessing variety of logic functions. et al. Bacterial sensing underlies artificial sweetener-induced As touched upon earlier, molecular computers cannot growth of gut Lactobacillus. Environ Microbiol. 2015. store kinetic energy or momentum. Consequently, they [3] Green J, Paget MS. Bacterial redox sensors. Nat Rev Microbiol. rely mostly on thermal energy from the environment for 2004;2:954-66. the delivery and removal of substrates and other large- [4] Vogel V, Sheetz M. Local force and geometry sensing regulate scale motions, including significant motions of their own cell functions. Nat Rev Mol Cell Biol. 2006;7:265-75. [5] Yoney A, Salman H. Precision and variability in bacterial moveable parts. These features of the nano-world pose temperature sensing. Biophys J. 2015;108:2427-36. new and unique challenges to nano-molecular engineers. [6] Feynman RP. The Feynman Lectures on Physics. Massachusetts, Nature, however, has provided us with examples of USA: Addison-Wesley. 1963;1. remarkably sophisticated and highly evolved algorithmic [7] Smoluchowski Mv. Experimentell nachweisbare, der Ublichen processes from which we can learn and implement for Thermodynamik Widersprechende Molekularphenomene. Phys Zeitshur. 1912:1069. our own advantage. Cells can be engineered to sense [8] Finkelstein ASSaAV. „The as a Brownian Ratchet and respond to environmental signals or inputs such as Machine“ chapter 9. in Molecular Machines in Biology edited temperature, pressure, radioactivity, or toxic chemicals. by Joachim Frank. 2011:158-90. Biological systems have the ability to adapt to new [9] Seeman NC. DNA nanotechnology: novel DNA constructions. information from an altered environment. They can Annu Rev Biophys Biomol Struct. 1998;27:225-48. self-assemble and self-reproduce, which might provide [10] Seeman NC. DNA engineering and its application to nanotechnology. Trends Biotechnol. 1999;17:437-43. some economic advantages. Thus, the ultimate goals of [11] Sun L, Yu L, Shen W. DNA nanotechnology and its biocomputing are the monitoring and control of biological applications in biomedical research. J Biomed Nanotechnol. systems. Although some solutions have been discussed 2014;10:2350-70. above, many natural limitations to the engineering of [12] Turberfield AJ. DNA nanotechnology: geometrical biological computers remain. For example, how do self-assembly. Nat Chem. 2011;3:580-1. [13] Zakeri B, Lu TK. DNA nanotechnology: new adventures for an we eliminate or reduce noise in nucleic-acid based old warhorse. Curr Opin Chem Biol. 2015;28:9-14. 20 S.G. Abels, E.F. Khisamutdinov

[14] Guo P. The emerging field of RNA nanotechnology. Nat [36] Zhang C, Yang J, Xu J. Circular DNA logic gates with strand Nanotechnol. 2010;5:833-42. displacement. Langmuir. 2010;26:1416-9. [15] Guo P. RNA Nanotechnology: methods for synthesis, [37] Qian L, Winfree E. Scaling up digital circuit computation conjugation, assembly and application of RNA nanoparticles. with DNA strand displacement cascades. Science. Methods. 2011;54:201-3. 2011;332:1196-201. [16] Qiu M, Khisamutdinov E, Zhao Z, Pan C, Choi JW, Leontis [38] Weitz M, Kim J, Kapsner K, Winfree E, Franco E, Simmel FC. NB, et al. RNA nanotechnology for computer design and Diversity in the dynamical behaviour of a compartmen- in vivo computation. Philos Trans A Math Phys Eng Sci. talized programmable biochemical oscillator. Nat Chem. 2013;371:20120310. 2014;6:295-302. [17] Grabow WW, Jaeger L. RNA self-assembly and RNA [39] Winfree E. Algorithmic Self-Assembly of DNA: Theoretical nanotechnology. Acc Chem Res. 2014;47:1871-80. Motivations and 2D Assembly Experiments. J Biomol Struct [18] Westhof E, Masquida B, Jaeger L. RNA tectonics: towards RNA Dyn. 2000;17 Suppl 1:263-70. design. Fold Des. 1996;1:R78-88. [40] Barish RD, Rothemund PW, Winfree E. Two computational [19] Bai X, Edwards J, Ju J. Molecular engineering approaches primitives for algorithmic self-assembly: copying and counting. for DNA sequencing and analysis. Expert Rev Mol Diagn. Nano Lett. 2005;5:2586-92. 2005;5:797-808. [41] Braich RS, Chelyapov N, Johnson C, Rothemund PW, Adleman [20] Wang K, Tang Z, Yang CJ, Kim Y, Fang X, Li W, et al. Molecular L. Solution of a 20-variable 3-SAT problem on a DNA computer. engineering of DNA: molecular beacons. Angew Chem Int Ed Science. 2002;296:499-502. Engl. 2009;48:856-70. [42] Lipton RJ. DNA solution of hard computational problems. [21] Wu Y, Kim Y, Tan W. Molecular engineering of DNA bases: Science. 1995;268:542-5. building block for functional molecular probes in biomedicine. [43] Liu Q, Wang L, Frutos AG, Condon AE, Corn RM, Smith LM. DNA Nucleic Acids Symp Ser (Oxf). 2008:65-6. computing on surfaces. Nature. 2000;403:175-9. [22] Afonin KA, Kasprzak WK, Bindewald E, Kireeva M, Viard M, [44] Xu J, Qiang X, Yang Y, Wang B, Yang D, Luo L, et al. An Kashlev M, et al. In silico design and enzymatic synthesis of unenumerative DNA computing model for vertex coloring functional RNA nanoparticles. Acc Chem Res. 2014;47:1731-41. problem. IEEE Trans Nanobioscience. 2011;10:94-8. [23] Khisamutdinov EF, Li H, Jasinski DL, Chen J, Fu J, Guo P. [45] Buckhout-White S, Spillmann CM, Algar WR, Khachatrian A, Enhancing immunomodulation on innate immunity by Melinger JS, Goldman ER, et al. Assembling programmable shape transition among RNA triangle, square and pentagon FRET-based photonic networks using designer DNA scaffolds. nanovehicles. Nucleic Acids Res. 2014;42:9996-10004. Nat Commun. 2014;5:5615. [24] Endo M, Sugiyama H. DNA nanotechnology: Measuring chloride [46] Chakraborty S, Mizusaki H, Kenney LJ. A FRET-based DNA in live cells. Nat Nanotechnol. 2015;10:569-70. biosensor tracks OmpR-dependent acidification of Salmonella [25] Livstone MS, van Noort D, Landweber LF. Molecular computing during macrophage infection. PLoS Biol. 2015;13:e1002116. revisited: a Moore‘s Law? Trends Biotechnol. 2003;21:98-101. [47] Kim J, Lee JS, Lee JB. Investigation of Forster resonance energy [26] Keyes RW. Physical limits of silicon transistors and circuits. transfer (FRET) and competition of fluorescent dyes on DNA Reports on Progress in Physics. 2005;68:2701-46. microparticles. Int J Mol Sci. 2015;16:7738-47. [27] Moe-Behrens GH. The biological microprocessor, or how [48] Mohd Bakhori N, Yusof NA, Abdullah AH, Hussein MZ. to build a computer with biological parts. Comput Struct Development of a Fluorescence Resonance Energy Transfer Biotechnol J. 2013;7:e201304003. (FRET)-Based DNA Biosensor for Detection of Synthetic [28] Miyamoto T, Razavi S, DeRose R, Inoue T. Synthesizing Oligonucleotide of Ganoderma boninense. Biosensors (Basel). -based Boolean logic gates. ACS Synth Biol. 2013;3:419-28. 2013;2:72-82. [49] Zadegan RM, Jepsen MD, Hildebrandt LL, Birkedal V, Kjems J. [29] Adleman LM. Molecular computation of solutions to Construction of a fuzzy and Boolean logic gates based on DNA. combinatorial problems. Science. 1994;266:1021-4. Small. 2015;11:1811-7. [30] Cornett EM, Campbell EA, Gulenay G, Peterson E, Bhaskar N, [50] Deng H, Yang X, Gao Z. MoS2 nanosheets as an effective Kolpashchikov DM. Molecular logic gates for DNA analysis: fluorescence quencher for DNA methyltransferase activity detection of rifampin resistance in M. tuberculosis DNA. Angew detection. Analyst. 2015;140:3210-5. Chem Int Ed Engl. 2012;51:9075-7. [51] Sato Y, Tian J, Ichihashi T, Chinda Y, Xu Z, Pang Y, et al. [31] Ogawa A, Susaki Y. Multiple-input and visible-output logic Enhancement in fluorescence response by a quencher for gates using signal-converting DNA machines and gold amiloride upon binding to thymine opposite an abasic site in a nanoparticle aggregation. Org Biomol Chem. 2013;11:3272-6. DNA duplex. Anal Chim Acta. 2010;675:49-52. [32] Okamoto A, Tanaka K, Saito I. DNA logic gates. J Am Chem Soc. [52] Zhang J, Tao M, Jin Y. An enzyme-aided amplification strategy 2004;126:9458-63. for sensitive detection of DNA utilizing graphene oxide (GO) as [33] Penchovsky R. Engineering integrated digital circuits with a fluorescence quencher. Analyst. 2014;139:3455-9. allosteric ribozymes for scaling up molecular computation and [53] Seelig G, Soloveichik D, Zhang DY, Winfree E. Enzyme-free diagnostics. ACS Synth Biol. 2012;1:471-82. nucleic acid logic circuits. Science. 2006;314:1585-8. [34] Pu F, Liu Z, Yang X, Ren J, Qu X. DNA-based logic gates [54] Ajo-Franklin CM, Drubin DA, Eskin JA, Gee EP, Landgraf D, operating as a biomolecular security device. Chem Commun Phillips I, et al. Rational design of memory in eukaryotic cells. (Camb). 2011;47:6024-6. Genes Dev. 2007;21:2271-6. [35] Yoshida W, Yokobayashi Y. Photonic Boolean logic gates based [55] Goldman N, Bertone P, Chen S, Dessimoz C, LeProust EM, Sipos on DNA aptamers. Chem Commun (Camb). 2007:195-7. B, et al. Towards practical, high-capacity, low-maintenance Nucleic Acid Nanotechnology: Promising Path to Altering Silicon Technology 21

information storage in synthesized DNA. Nature. [76] Benenson Y, Gil B, Ben-Dor U, Adar R, Shapiro E. An 2013;494:77-80. autonomous molecular computer for logical control of gene [56] Lin CH, Cheng HP, Yang CB, Yang CN. Solving satisfiability expression. Nature. 2004;429:423-9. problems using a novel microarray-based DNA computer. [77] Gil B, Kahan-Hanum M, Skirtenko N, Adar R, Shapiro E. Biosystems. 2007;90:242-52. Detection of multiple disease indicators by an autonomous [57] Wu H. An improved surface-based method for DNA biomolecular computer. Nano Lett. 2011;11:2989-96. computation. Biosystems. 2001;59:1-5. [78] Shapiro E, Gil B. Cell biology. RNA computing in a living cell. [58] Aghebat Rafat A, Pirzer T, Scheible MB, Kostina A, Simmel FC. Science. 2008;322:387-8. Surface-assisted large-scale ordering of DNA origami tiles. [79] Afonin KA, Grabow WW, Walker FM, Bindewald E, Dobrovolskaia Angew Chem Int Ed Engl. 2014;53:7665-8. MA, Shapiro BA, et al. Design and self-assembly of siRNA- [59] Wei B, Dai M, Yin P. Complex shapes self-assembled from functionalized RNA nanoparticles for use in automated single-stranded DNA tiles. Nature. 2012;485:623-6. nanomedicine. Nat Protoc. 2011;6:2022-34. [60] Wei B, Vhudzijena MK, Robaszewski J, Yin P. Self-assembly of [80] Faulhammer D, Cukras AR, Lipton RJ, Landweber LF. Molecular Complex Two-dimensional Shapes from Single-stranded DNA computation: RNA solutions to chess problems. Proc Natl Acad Tiles. J Vis Exp. 2015. Sci U S A. 2000;97:1385-9. [61] Mao C, LaBean TH, Relf JH, Seeman NC. Logical computation [81] Daniel C, Lagergren J, Ohman M. RNA editing of non-coding using algorithmic self-assembly of DNA triple-crossover RNA and its role in gene regulation. Biochimie. 2015. molecules. Nature. 2000;407:493-6. [82] Benenson Y. RNA-based computation in live cells. Curr Opin [62] Jiang Y, Liu N, Guo W, Xia F, Jiang L. Highly-efficient gating of Biotechnol. 2009;20:471-8. solid-state nanochannels by DNA supersandwich structure [83] Cukras AR, Faulhammer D, Lipton RJ, Landweber LF. Chess containing ATP aptamers: a nanofluidic IMPLICATION logic games: a model for RNA based computation. Biosystems. device. J Am Chem Soc. 2012;134:15395-401. 1999;52:35-45. [63] Douglas SM, Bachelet I, Church GM. A logic-gated nanorobot [84] Davidson EA, Ellington AD. Synthetic RNA circuits. Nat Chem for targeted transport of molecular payloads. Science. Biol. 2007;3:23-8. 2012;335:831-4. [85] Rodrigo G, Landrain TE, Jaramillo A. De novo automated design [64] Lederman H, Macdonald J, Stefanovic D, Stojanovic MN. of small RNA circuits for engineering synthetic riboregulation in Deoxyribozyme-based three-input logic gates and construction living cells. Proc Natl Acad Sci U S A. 2012;109:15271-6. of a molecular full adder. Biochemistry. 2006;45:1194-9. [86] Lucks JB, Qi L, Mutalik VK, Wang D, Arkin AP. Versatile [65] Pei R, Macdonald J, Stojanovic MN. Development of trainable RNA-sensing transcriptional regulators for engineering genetic deoxyribozyme-based game playing automaton. Methods Mol networks. Proc Natl Acad Sci U S A. 2011;108:8617-22. Biol. 2012;848:419-37. [87] Leisner M, Bleris L, Lohmueller J, Xie Z, Benenson Y. Rationally [66] Stojanovic MN, Semova S, Kolpashchikov D, Macdonald designed logic integration of regulatory signals in mammalian J, Morgan C, Stefanovic D. Deoxyribozyme-based ligase cells. Nat Nanotechnol. 2010;5:666-70. logic gates and their initial circuits. J Am Chem Soc. [88] Chappell J, Watters KE, Takahashi MK, Lucks JB. A renaissance 2005;127:6914-5. in RNA synthetic biology: new mechanisms, applications and [67] Zhang HR, Wang YZ, Wu MS, Feng QM, Shi HW, Chen HY, et al. tools for the future. Curr Opin Chem Biol. 2015;28:47-56. Visual electrochemiluminescence detection of telomerase [89] Gallivan JP. RNA synthetic biology: from the test tube to cells activity based on multifunctional Au nanoparticles modified and back again. ACS Synth Biol. 2015;4:493-4. with G-quadruplex deoxyribozyme and luminol. Chem Commun [90] Isaacs FJ, Dwyer DJ, Collins JJ. RNA synthetic biology. Nat (Camb). 2014;50:12575-7. Biotechnol. 2006;24:545-54. [68] Stojanovic MN, Mitchell TE, Stefanovic D. Deoxyribozyme-based [91] Afonin KA, Bindewald E, Kireeva M, Shapiro BA. Computational logic gates. J Am Chem Soc. 2002;124:3555-61. and experimental studies of reassociating RNA/DNA [69] Goni-Moreno A, Amos M. A reconfigurable NAND/NOR genetic hybrids containing split functionalities. Methods Enzymol. logic gate. BMC Syst Biol. 2012;6:126. 2015;553:313-34. [70] Ji W, Shi H, Zhang H, Sun R, Xi J, Wen D, et al. A formalized [92] Afonin KA, Bindewald E, Yaghoubian AJ, Voss N, Jacovetty design process for bacterial consortia that perform logic E, Shapiro BA, et al. In vitro assembly of cubic RNA-based computing. PLoS One. 2013;8:e57482. scaffolds designed in silico. Nat Nanotechnol. 2010;5:676-82. [71] Ran T, Douek Y, Milo L, Shapiro E. A programmable NOR-based [93] Khisamutdinov EF, Bui MN, Jasinski D, Zhao Z, Cui Z, Guo device for transcription profile analysis. Sci Rep. 2012;2:641. P. Simple Method for Constructing RNA Triangle, Square, [72] Wang B, Kitney RI, Joly N, Buck M. Engineering modular and Pentagon by Tuning Interior RNA 3WJ Angle from 60 orthogonal genetic logic gates for robust digital-like synthetic degrees to 90 degrees or 108 degrees. Methods Mol Biol. biology. Nat Commun. 2011;2:508. 2015;1316:181-93. [73] Moon TS, Lou C, Tamsir A, Stanton BC, Voigt CA. Genetic [94] Khisamutdinov EF, Jasinski DL, Guo P. RNA as a boiling-resistant programs constructed from layered logic gates in single cells. anionic polymer material to build robust structures with Nature. 2012;491:249-53. defined shape and stoichiometry. ACS Nano. 2014;8:4771-81. [74] Guet CC, Elowitz MB, Hsing W, Leibler S. Combinatorial [95] Jasinski DL, Khisamutdinov EF, Lyubchenko YL, Guo P. Physico- synthesis of genetic networks. Science. 2002;296:1466-70. chemically tunable polyfunctionalized RNA square architecture [75] Bonnet J, Yin P, Ortiz ME, Subsoontorn P, Endy D. Amplifying with fluorogenic and ribozymatic properties. ACS Nano. genetic logic gates. Science. 2013;340:599-603. 2014;8:7620-9. 22 S.G. Abels, E.F. Khisamutdinov

[96] Shu Y, Haque F, Shu D, Li W, Zhu Z, Kotb M, et al. Fabrication [113] Fire A, Xu S, Montgomery MK, Kostas SA, Driver SE, Mello CC. of 14 different RNA nanoparticles for specific tumor targeting Potent and specific genetic interference by double-stranded without accumulation in normal organs. RNA. 2013;19:767-77. RNA in Caenorhabditis elegans. Nature. 1998;391:806-11. [97] Lambowitz AM, Zimmerly S. Group II introns: mobile [114] Timmons L, Fire A. Specific interference by ingested dsRNA. ribozymes that invade DNA. Cold Spring Harb Perspect Biol. Nature. 1998;395:854. 2011;3:a003616. [115] Bich Ngoc Dao MV, Angelica N. Martins, Wojciech K. Kasprzak, [98] Khvorova A, Lescoute A, Westhof E, Jayasena SD. Sequence Bruce A. Shapiro, Kirill A. Afonin. Triggering RNAi with elements outside the hammerhead ribozyme catalytic core multifunctional RNA nanoparticles and their delivery. DNA and enable intracellular activity. Nat Struct Biol. 2003;10:708-12. RNA nanotechnology. 2015;2:12. [99] Win MN, Smolke CD. Higher-order cellular information [116] Xie Z, Wroblewska L, Prochazka L, Weiss R, Benenson Y. processing with synthetic RNA devices. Science. Multi-input RNAi-based logic circuit for identification of specific 2008;322:456-60. cancer cells. Science. 2011;333:1307-11. [100] Chen YY, Jensen MC, Smolke CD. Genetic control of [117] Rinaudo K, Bleris L, Maddamsetti R, Subramanian S, Weiss mammalian T-cell proliferation with synthetic RNA regulatory R, Benenson Y. A universal RNAi-based logic evaluator systems. Proc Natl Acad Sci U S A. 2010;107:8531-6. that operates in mammalian cells. Nat Biotechnol. [101] Lee SK, Calin GA. Genetic control of mammalian T-cell 2007;25:795-801. proliferation with a synthetic RNA regulatory system - illusion [118] Leisner M, Bleris L, Lohmueller J, Xie Z, Benenson Y. or reality? Genome Med. 2010;2:77. MicroRNA circuits for transcriptional logic. Methods Mol Biol. [102] Penchovsky R, Breaker RR. Computational design and 2012;813:169-86. experimental validation of oligonucleotide-sensing allosteric [119] Afonin KA, Kasprzak W, Bindewald E, Puppala PS, Diehl AR, Hall ribozymes. Nat Biotechnol. 2005;23:1424-33. KT, et al. Computational and experimental characterization of [103] Penchovsky R. Computational design of allosteric ribozymes as RNA cubic nanoscaffolds. Methods. 2014;67:256-65. molecular biosensors. Biotechnol Adv. 2014;32:1015-27. [120] Guo P. RNA nanotechnology: engineering, assembly and [104] Ellington AD, Szostak JW. In vitro selection of RNA applications in detection, gene delivery and therapy. J Nanosci molecules that bind specific ligands. Nature. 1990;346:818-22. Nanotechnol. 2005;5:1964-82. [105] Hermann T, Patel DJ. Adaptive recognition by nucleic acid [121] Afonin KA, Viard M, Kagiampakis I, Case CL, Dobrovolskaia aptamers. Science. 2000;287:820-5. MA, Hofmann J, et al. Triggering of RNA interference with [106] Sudarsan N, Hammond MC, Block KF, Welz R, Barrick JE, RNA-RNA, RNA-DNA, and DNA-RNA nanoparticles. ACS Nano. Roth A, et al. Tandem riboswitch architectures exhibit complex 2015;9:251-9. gene control functions. Science. 2006;314:300-4. [122] Tan C, Song H, Niemi J, You L. A synthetic biology challenge: [107] Garst AD, Edwards AL, Batey RT. Riboswitches: structures and making cells compute. Mol Biosyst. 2007;3:343-53. mechanisms. Cold Spring Harb Perspect Biol. 2011;3. [123] Benenson Y. Biomolecular computing systems: principles, [108] Serganov A, Nudler E. A decade of riboswitches. Cell. progress and potential. Nat Rev Genet. 2012;13:455-68. 2013;152:17-24. [124] Fedichkin L, Katz E, Privman V. Error correction and digita- [109] Topp S, Gallivan JP. Riboswitches in unexpected lization concepts in biochemical computing. Journal of places--a synthetic riboswitch in a protein coding region. RNA. Computational and Theoretical Nanoscience. 2008;5:36-43. 2008;14:2498-503. [125] Privman V, Arugula MA, Halamek J, Pita M, Katz E. Network [110] Breaker RR. Riboswitches and the RNA world. Cold Spring Harb analysis of biochemical logic for noise reduction and stability: Perspect Biol. 2012;4. a system of three coupled enzymatic and gates. J Phys Chem B. [111] Sudarsan N, Hammond MC, Block KF, Welz R, Barrick JE, Roth 2009;113:5301-10. A, et al. Tandem riboswitch architectures exhibit complex gene [126] Privman V, Strack G, Solenov D, Pita M, Katz E. Optimization of control functions. Science. 2006;314:300-4. enzymatic biochemical logic for noise reduction and scalability: [112] Tabara H, Grishok A, Mello CC. RNAi in C. elegans: soaking in how many biocomputing gates can be interconnected in a the genome sequence. Science. 1998;282:430-1. circuit? J Phys Chem B. 2008;112:11777-84.