DNA Nanotechnology from the Test Tube to the Cell Yuan-Jyue Chen1, Benjamin Groves1, Richard A

Home , DNA nanotechnology, Molecular machine, Nadrian Seeman, RNA origami, Spherical nucleic acid, Streptavidin

REVIEW ARTICLE PUBLISHED ONLINE: 3 SEPTEMBER 2015!|!DOI: 10.1038/NNANO.2015.195

DNA nanotechnology from the test tube to the cell Yuan-Jyue Chen1, Benjamin Groves1, Richard A. Muscat1 and Georg Seelig1,2*

The programmability of Watson–Crick base pairing, combined with a decrease in the cost of synthesis, has made DNA a widely used material for the assembly of molecular structures and dynamic molecular devices. Working in cell-free settings, researchers in DNA nanotechnology have been able to scale up system complexity and quantitatively characterize reaction mechanisms to an extent that is infeasible for engineered gene circuits or other cell-based technologies. However, the most intriguing applications of DNA nanotechnology — applications that best take advantage of the small size, biocompatibility and programmability of DNA-based systems — lie at the interface with biology. Here, we review recent progress in the transition of DNA nanotechnology from the test tube to the cell. We highlight key successes in the development of DNA-based imaging probes, prototypes of smart therapeutics and drug delivery systems, and explore the future challenges and opportunities for cellular DNA nanotechnology.

NA nanotechnology is a purist’s approach to biomolecular several recent results that show how DNA nanodevices can be pro- engineering. Te feld aims to create molecular structures grammed to interact with cell surface proteins, before turning to Dand devices through the exclusive use of DNA as an engi- work on the delivery of DNA devices and structures into cells. We neering material. Te well-characterized nature of DNA base- reach devices that operate inside live cells and review initial work pairing provides an easy means to control DNA interactions; this towards using DNA sensors and logic gates to detect, analyse and ‘sequence programmability’ has allowed the rational design of pre- regulate cellular RNA levels. We put this work into context by high- cisely defned structures ranging in size from nanometres to milli- lighting design principles identifed in the development of live-cell metres, and of molecular motors or circuits that can autonomously RNA imaging probes, small interfering RNAs (siRNAs) or anti- move or process information. Tere is currently no other molecular sense oligonucleotides (ASOs), which could be used to improve the engineering technology that enables the fully de novo design of a performance of DNA devices in cells. Finally, we make connections similarly complex and diverse set of biomolecular systems. to RNA nanotechnology and RNA synthetic biology, which have Te success of DNA nanotechnology comes from three key broadly similar aims to DNA nanotechnology but typically rely on ingredients: 1) our quantitative understanding of DNA thermo- the use of genetically encoded and transcribed RNA. dynamics, which makes it possible to predict reliably how single- stranded DNA molecules fold and interact with one another1,2; Cell-free DNA nanotechnology 2) the rapidly falling cost and increasing quality of DNA synthe- To operate reliably in complex, wet environments, living organ- sis3; and 3) the focus on cell-free settings, where designed reaction isms use molecular sensors to detect changes in that environment, pathways can proceed without interference from DNA and RNA motors and actuators to adapt to the environment, computational processing enzymes and other confounding factors that might be control circuits to convert sensor information into motor activity, encountered in cells. and structural elements that protect and organize these components. DNA nanotechnology has long been motivated by the goal of Intriguingly, cell-free DNA nanotechnology has made progress building ‘smart therapeutics’, drug delivery systems, tools for molec- towards the construction of most of the functional components — ular biology and other devices that could interact with or operate both structures and dynamic devices — required for creating molec- within living cells4–7 (Fig. 1). Such applications play to the obvious ular ‘robots’ that can emulate some of the behavioural complexity strengths of nucleic acid nanostructures and devices, particularly observed in biology. Here we review a few key results from cell-free their small size, biocompatibility and straightforward manner in DNA nanotechnology and point out potential applications in the which they could be programmed to interact with cellular nucleic cellular environment. acids through hybridization. However, to realize such applications using tools from DNA nanotechnology, it will be necessary Structural DNA nanotechnology. In the 1980s, Nadrian Seeman to bridge the gap between performing experiments in well-mixed developed the notion that DNA could be used as a structural reaction bufers and spatially structured, densely packed cellular engineering material8–10. In 1998, Winfree et al. provided the frst environments (Box 1). experimental demonstration of large-scale structure formation: In this Review, we summarize recent progress towards the goal they showed that micrometre-sized periodic DNA lattices could of bringing DNA nanotechnology into the cell. We focus on nucleic self-assemble from nanoscale DNA tiles that are themselves assem- acid nanodevices and nanostructures that are rationally designed, blies of multiple oligonucleotides11. Subsequently, tile assembly and chemically synthesized and then delivered to mammalian cells. We related techniques were successfully used to create a wide variety of begin with a brief overview of DNA nanotechnology in cell-free lattices and wireframe DNA structures11–19. settings, and then move to more cell-like environments, such as Rothemund further advanced structural DNA self-assembly cell lysates and fxed cells — settings that capture some, but not by developing DNA origami, a technique that is easy to use, fex- all, of the complexity of cellular environments. Next, we discuss ible enough to accommodate almost any two-dimensional (2D)

1Department of Electrical Engineering, University of Washington, Seattle, Washington 98195, USA. 2Department of Computer Science and Engineering, University of Washington, Seattle, Washington 98195, USA. *e-mail: [email protected]

748 NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology

© 2015 Macmillan Publishers Limited. All rights reserved NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2015.195 REVIEW ARTICLE structure of interest, and reliably results in a high yield of the target a Smart therapeutics b Drug delivery structure20. DNA origami relies on the use of a long single-stranded scafold strand that is folded into a target structure through hybridization with a large number of short staple strands. Tis technology was rapidly and broadly adopted, and was soon generalized to the AND self-assembly of three-dimensional (3D) structures21–24. DNA nanostructures are beginning to be investigated as tools for drug delivery and similar applications because they provide precisely programmable scafolds for the attachment of functional groups including drug and targeting moieties, and because 3D structures can be designed to act as protective enclosures for a cargo of interest.

Dynamic DNA nanotechnology. Dynamic DNA nanotechnology combines self-assembly through programmed hybridization with c Imaging d Cell biology DNAzyme catalysis or DNA strand displacement reactions — a form of competitive hybridization — to create devices with moving parts and time-varying behaviours. Dynamic DNA nano technology can Cell be traced to multiple sources, including Adleman’s work on DNA computation and research on the directed evolution and characteri- zation of functional nucleic acids25. However, Yurke and co-workers d truly launched the feld by demonstrating that a functional molecular ‘motor’ could be rationally designed and driven through its work cycle using only hybridization and strand displacement reactions26. Subsequently, the Winfree and Pierce groups demonstrated that multiple strand displacement reactions could be chained together to create complex reaction cascades27,28. Owing to their simplicity, Figure 1 | Applications of DNA nanotechnology at the interface with DNA strand displacement cascades have since been used widely and biology. a, Smart therapeutics could combine structural elements with efectively for molecular engineering and provide the mechanism molecular logic to target therapeutic actions to a specific cell or tissue that drives most dynamic DNA devices. type, thus minimizing side efects60. b, DNA nanostructures can serve as Dynamic DNA nanotechnology has resulted in molecular programmable scafolds for attaching drugs, targeting ligands and other motors29,30, including walking motors that autonomously move along modifications, such as lipid bilayers78. c, A novel class of sensitive and a track31–34, molecular circuits that can analyse information encoded specific imaging probes that takes advantage of DNA-based amplification in complex mixtures of molecules27,35–39, and catalytic amplifers that mechanisms can be programmed to sequence-specifically interact with can sense and amplify signals40–44. Many of these systems have obvi- cellular RNA52. d, DNA origami and other structures provide precise control ous potential for biotechnological applications: for example, Shapiro over the spatial organization of functional molecular groups, which makes and collaborators used DNA and a restriction enzyme to build a them intriguing tools for quantitative measurements in cell biology66. Figure molecular automaton that could diagnose the state of a disease by reproduced with permission from: a, ref. 60, AAAS; b, ref. 78, American detecting and analysing a set of molecular markers, thus realizing, in Chemical Society; c, ref. 52, American Chemical Society; d, ref. 66, Nature a test tube, a type of computation similar to those performed by gene Publishing Group. regulatory networks6,45. Conversely, the analysis and manipulation of molecular information in and on living cells is the one area of application in which molecular devices and structures can out perform afer incubation. Because detailed conditions for mixing the ori- their electromechanical counterparts. gami with cell lysate were not reported, it is difcult to evaluate how closely the reaction bufer approximated physiological conditions. DNA nanotechnology in lysates and fixed cells Furthermore, because DNA nanostructures are typically assembled Cellular conditions are signifcantly diferent from those used in in bufers with high Mg2+ concentrations (~10 mM), the addition cell-free experiments (Box 1): the presence of nucleic-acid-binding of large amounts of nanostructures could increase the Mg2+ level, proteins, including DNases and RNases, may interfere with device thus making the structures seem more structurally robust than what performance. Moreover, cellular environments are highly structured, might be expected in a cell. Still, such efects can be controlled, and which inhibits the free difusion of exogenously delivered nucleic lysates constitute a useful setting for exploring how nanodevices acids. Cell lysates, serum and fxed cells provide reaction environ- might fare in biological environments. ments that each capture some of the complexity of live cells and ena- Moving nanostructures into cell culture and animals will ble testing and optimization of nucleic acid devices in comparably require devices that are stable in the presence of serum and serum- well-controlled conditions. supplemented media. Like lysates, serum contains nucleases and lacks stabilizing salts such as magnesium. Conway et al. showed Stability of DNA nanostructures in cell lysates and serum. that small three-stranded nanostructures in the shape of a triangu- Lysates are mixtures of cellular components created from cells lar prism were more stable in serum than the individual compo- that have been homogenized. Because lysates lack any kind of cell nent strands47. A gel analysis showed that individual strands had a wall, nucleic acid devices can readily be placed into an environ- half-life of less than an hour in 10% fetal bovine serum, whereas ment imitating that found inside the cell, although the concentra- the half-life of intact structures was closer to two hours. Te use tions and activities of the cellular components encountered by the of chemically modifed strands resulted in structures with half-lives DNA nano structure are usually diferent. Yan, Meldrum and col- even longer than 24 hours. laborators tested the stability of DNA origami in cell lysate and In a comprehensive analysis, Perrault and colleagues tested found that origamis could be extracted from the lysate and char- three diferent 3D origamis in mammalian cell culture media sup- acterized following up to 12 hours of incubation46. In contrast, long plemented with serum, and showed that the structural integrity of single- and double-stranded nucleic acids could not be recovered origamis is strongly dependent on the origami design, the presence

NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology 749

Box 1 | Synthetic nucleic acids in the cellular environment.

Te diferent techniques, design considerations and limitations tend to have adverse efects on cell viability138. Terefore, when discovered by researchers working with ASOs, siRNAs, molecu- choosing modifcations for nucleic acid devices, it is important to lar beacons and related technologies help us highlight some of strike a balance between device stability and cell viability. the challenges of bringing nucleic acid nanotechnology to the cellular environment. Molecular crowding and cellular compartmentalization. Cells are densely packed with proteins and other macromolecules, Delivery. In test tubes, the concentration of all components can be which can adversely afect the performance of multi-stage, multi- precisely controlled and reaction kinetics can be monitored with input logic circuits and other systems with many interacting com- high time resolution. In contrast, to function in cells, nucleic acid ponents. Te difusion coefcient of synthetic DNA molecules in devices must frst cross the cell membrane. Diferent nucleic acid the cytoplasm is 5–100 times smaller than in water, depending on delivery methods can result in vast diferences in cellular uptake the size of the molecule139. Te rates of hybridization between com- timing, amount and subcellular distribution, and even cell viabil- plementary single-stranded nucleic acids are also diferent in the ity. For example, commonly used lipid-based transfection reagents cellular environment than in an aqueous bufer140. Furthermore, efciently deliver large numbers of probes to cells, but a signif- mammalian cells are organized in a variety of diferent compart- cant fraction are enclosed in endosomes and thus do not reach ments, and enclosures within such compartments could prevent the cytoplasm132,133. Conversely, microinjection can deliver nucleic distinct circuit components from encountering each other. acids directly to the cytoplasm or nucleus, but is limited to a relatively small number of cells. We refer the reader to Bao et al.134 for Immune activation. Exogenously delivered nucleic acids can trig- a more in-depth comparison of diferent methods for the delivery ger an innate immune response through the activation of Toll-like of synthetic nucleic acids to cultured cells, and to Davis et al.85 for receptors (TLR). TLR3 responds to double-stranded RNA; TLR7 a review of nanoparticle-based drug delivery. and TLR8 respond to single-stranded RNA; and TLR9 responds to unmethylated cytosine-guanine (CpG) motifs in DNA. TLR9 Stability and chemical modifcations. Te cellular half-lives serves to detect DNA of bacterial origin by exploiting the fact of short, unmodifed nucleic acids are of the order of minutes135. that in mammalian cell genomic DNA the dinucleotide ‘CpG’ However, a number of chemical modifcations to the sugar, base is generally methylated, whereas in bacteria it is not79. Double- and backbone of nucleic acids have been identifed that dramatically stranded RNA longer than 30 bp is bound by protein kinase R, enhance their stability. Te most commonly used modifcations which activates a cellular immune response that can result in cell include phosphorothioate inter-nucleotide linkages and 2ʹO-methyl death141. Activation of TLR9142- or PKR143-mediated responses are ribose modifcations136. Because chemical modifcations provide considered for therapeutic applications where immune stimula- efcient ways to protect nucleic acids against degradation by nucle- tion may be desirable. However, it is more common to avoid such ases, some of them (for example, phosphorothioate bonds137) also immune stimulation. of Mg2+ and the level of nuclease activity48. Using gel and transmis- nanostructures can retain their structural integrity over extended sion electron microscopy assays, they observed that two of their time periods in physiological salt conditions. Further research is three test constructs were partially denatured afer incubation in required to understand fully the interplay between the functionality cell culture media for a day, and that addition of 6 mM Mg2+ to of a given structure and its stability. the media was required to inhibit this efect. Intriguingly, a third structure, an origami nanotube, was structurally stable even with- DNA nanotechnology in fxed cells. Permeabilized cells and tis- out the added salt. Afer 24 hours, all three structures were partially sues also mimic some aspects of the cellular environment: fxed cells degraded by DNases when >5% fetal bovine serum was added to retain much of their structural organization, in particular the spatial the media. Importantly, nuclease degradation could be dramatically distribution of mRNA and proteins. Fixed cells provide a controlled reduced by addition of actin, a protein that binds competitively to setting for visualizing the subcellular distribution of mRNAs and nucleases; this modifcation was found to be compatible with cell proteins using immunostaining or fuorescence in situ hybridiza- culture conditions. tion (FISH); approaches from DNA nanotechnology have already To quantify the degree of nanostructure degradation by DNases, proved practically useful for increasing the sensitivity and specifc- Keum and Bermudez measured the half-life of wireframe tetra- ity of such imaging methods. For instance, molecular probes based hedral DNA nanostructures (TDNs) in the presence of DNase 1. on a hybridization chain reaction (HCR)28 have enabled the simulta- Tey found that TDNs were up to three times more stable than neous mapping of up to fve target mRNAs within intact vertebrate double- stranded DNA49. Likewise, DNA origami has also been embryos51,52. By hybridizing a set of adaptor strands to target mRNA shown to be more stable than duplex DNA in the presence of nucle- sequences, Choi et al. were able to controllably catalyse a polym- ases. Castro et al. incubated DNA origami with diferent nucleases erization reaction of two types of fuorescently labelled hairpin and assessed origami degradation afer 24 hours using transmission monomers; as a result of this catalytic hybridization reaction, the electron microscopy and gel assays50. Degradation was observed fuorescent signal associated with a given mRNA is amplifed and in the presence of DNase 1 and T7 endonuclease 1; however, at can be imaged readily using a fuorescence microscope (Fig. 2a). least for DNase 1, the measured rate of degradation was several By combining the ideas of strand displacement with single- hundred-fold slower than for duplex DNA. Comparing this result molecule FISH (smFISH), Raj and colleagues were able to detect with the TDN study suggests that DNA origami are even more sta- single-nucleotide variations within individual mRNA transcripts53. ble than smaller wireframe TDNs, perhaps because they are more When performing smFISH, a collection of singly labelled DNA highly interconnected. oligo nucleotides hybridize along the target RNA transcripts in fxed Together, these results suggest that DNA nanostructures can cells54. Co-localization of multiple probes on the same transcript withstand degradation by nucleases considerably better than sim- produces a discrete fuorescence ‘spot’ that is clearly discernable ple single- and double-stranded nucleic acids. Moreover, some using conventional fuorescence microscopy. Discrimination at the

750 NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology

a mRNA target

l1 50 μm

b Wild-type probe Mutant probe 1 2

Heterozygotic cell

Wild-type Mutant RNA RNA target RNA target 3 Mutant detection 4 SNV probe detection

Guide probe Guide probe Wild-type RNA 5 μm Mutant RNA Unclassiﬁed RNA

Figure 2 | In situ imaging of mRNA in fixed cells. a, HCR FISH52. Left: Initiator strands I1 and I2 hybridize to a target mRNA, which triggers a polymerization reaction between the two fluorescently labelled hairpin monomers H1 and H2. As a result, the target mRNA is connected to multiple fluorophores and can be visualized using fluorescence microscopy. Right: Confocal microscopy images at diferent z planes in a fixed zebrafish embryo. HCR probes are used to identify four diferent mRNAs (red: Tg(flk1:egfp); blue: tpm3; green: elevl3; yellow: ntla). b, Detection of a single-nucleotide variation using strand displacement probes53. Left: Reaction mechanism. Mutant and wild-type probes compete for binding to a target mRNA. Because binding kinetics strongly depend on toehold sequence, each probe type primarily binds to the cognate mRNA. Co-localization of single-nucleotide variation detection probes with multiple mRNA-targeting guide probes further shows that the signal is indeed triggered of the mRNA. Right: Fluorescence micrographs of BRAF mRNA detected using ‘guide probes’ (image 1), wild-type probes (image 2) and mutant probes (image 3). Image 4 shows mRNA classified as wild type or mutant. SNV, single-nucleotide variation. Figure reproduced with permission from: a, ref. 52, American Chemical Society; b, ref. 53, Nature Publishing Group. level of individual nucleotides was achieved using an additional means that the number of protein species that may be imaged is no strand displacement probe modifed with a distinct fuorophore/ longer constrained by the number of resolvable wavelengths avail- quencher pair53. Probe binding through toehold-mediated strand able to the microscope (generally around four), but is instead lim- displacement was dramatically slowed in the presence of a mis- ited only by the number of sequential hybridization/displacement match between the toehold and target. Co-localization of the single- cycles that may be performed. nucleotide variation probe with the transcript signal was used to DNA point accumulation for imaging in nanoscale topography verify the identity of the sequence (Fig. 2b). (DNA-PAINT) is an approach that similarly takes advantage of the Strand displacement in fxed cells has also been demonstrated for reversibility of DNA hybridization. Short, fuorescently labelled DNA-tagged proteins55. Te ability to hybridize and displace strands DNA imager strands are used to bind transiently to complementary

NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology 751

a Non-target cell DNA circuits Q F

Negative

Initiator Target cell F Q F Q Reporter Q Positive F Label EvaluateReport

b 0 1 0 1 AND 0 AND 0 AND 0 AND 1 0 0 1 1

Cell Cell Cell Cell

Figure 3 | Cell surface computation. a, In situ cell classification by evaluating specific surface markers59. Cells are first coated with DNA-modified antibodies (DNA circuits; antibodies are shown as rectangles or ellipses, DNA strands as coloured lines), and depending on the surface marker profiles of the cell type, either one or two gates can bind to cells. The subsequent introduction of an initiator strand (red) triggers a series of strand displacement reactions (fully complementary strands share the same colours). A soluble reporter complex can fluorescently tag only cells labelled with two surface-bound gates. b, Molecular robot for targeting a therapeutic action to specific cell types. The schematic shows how a barrel-shaped nanorobot responds to specific antigens (keys) expressed on cells surfaces60. The nanorobot is initially held in a closed configuration by two aptamer locks; only when it encounters a cell that displays two matching antigens can it be opened, thereby exposing a drug. Bottom: Transmission electron microscopy images of the closed and open states of the nanorobots (scale bars, 20 nm). Figure reproduced with permission from: a, ref. 59, Nature Publishing Group; b, ref. 60, AAAS. docking strands attached to a target56. Te spontaneous binding probes to protein antibodies59. Cells were labelled with one or two and unbinding causes the fuorescence at a given point to switch probes, depending on which proteins were displayed on the cell between the on and of state, thus allowing individual target sites to surface. Afer the binding stage, a trigger strand was added to acti- be imaged with sub-10-nm resolution using total internal refection vate a strand displacement cascade involving the attached probes. microscopy. As above, the reversible nature of DNA-PAINT means Te output of the cascade depended on whether one or both probes that it is not limited by the number of fuorophores, and sequen- were present, thus allowing the cell types to be distinguished tial labelling allows the reuse of fuorescent dyes. DNA-PAINT was (Fig. 3a). Although a similar outcome could be achieved by directly adapted to the in situ 3D imaging of fxed cells by targeting cellular labelling cells with two fuorescently tagged antibodies, this work proteins with antibodies conjugated to DNA docking strands57. demonstrates a more easily scalable approach for performing cell- state classifcation, potentially allowing many molecular markers to Interacting with cell surface markers be analysed in parallel and information to be summarized into an Mammalian cells are comprised of a number of compartments and easy-to-interpret actionable signal. structures, which all act as discrete vessels for biochemical reac- Douglas et al. created a DNA ‘nanorobot’ capable of delivering a tions. Te most accessible structure is the cell surface itself — a lipid molecular payload to particular cell types60. Te payload was enclosed bilayer incorporating many surface proteins that ofen diferenti- by a hinged origami container, which was initially held in a closed ate one cell type from another. Several recent papers demonstrated conformation to sequester the cargo. Aptamers — DNA or RNA that DNA nanosystems can be designed to interact with cell surface sequences selected to bind specifc proteins or even whole cells61,62 — markers; as with antibodies and aptamers58, the most mature exam- provided the means of targeting the nanorobot to specifc cells with- ples of potential DNA-based therapeutics target cell surface mark- out the need for covalent attachment of DNA strands to antibodies. ers and cells in the bloodstream — targets that do not require the Te same aptamers were also part of the locking mechanism; aptamer uptake of nanostructures into specifc cells and tissues. binding to the target protein triggered a conformational change, Stojanovic and colleagues directed the probes to particular cell thus allowing the origami lid to open and expose the cargo. AND surface proteins by covalently attaching DNA strand displacement logic — implemented by employing combinations of two diferent

752 NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology

© 2015 Macmillan Publishers Limited. All rights reserved NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2015.195 REVIEW ARTICLE aptamers — was used to target distinct fuorophores or a drug mol- Walsh et al. demonstrated that the uptake of TDNs into human ecule to a subset of cells (Fig. 3b). Intriguingly, a similar nanorobot embryonic kidney cells was similarly efcient with or without a was also shown to be active in the bloodstream of live cockroaches, lipid-based transfection reagent. A Förster resonance energy trans- with multiple robot species performing logic operations63. fer (FRET) assay was used to demonstrate that TDNs remain intact Te Tan group has also used aptamer-based logic gates to dis- for a long time afer cellular uptake74. Work by Schuller et al. showed tinguish cells in mixed populations64. In this implementation, a that DNA origami structures much larger than TDNs could also enter scafold is used to link a logic gate with multiple aptamers. Binding cells without the need for transfection reagents75. Internalization of the aptamers to surface proteins releases DNA strands that act as a fuorescently labelled strand attached to the origami made it pos- inputs to the logic gate. Tus, the logic gates will only be triggered sible to visualize origami uptake, although this assay could not be if the appropriate aptamer ligands are present on the surface of the used to determine whether the origami structures were still intact cell. Crosstalk between cells is minimal, which suggests that nearby in the cell. interactions are preferential owing to a higher local concentration Given the anionic nature of DNA, it is surprising that cells take of interacting species. Te same group has recently demonstrated up DNA nanostructures in the absence of transfection reagents. a more modular gate design that allows for a greater number of Liang et al. recently investigated the mechanism responsible for inputs and the potential for combining with the scafold approach TDN uptake and found that they enter mammalian cells through to improve of-target efects65. receptor-mediated endocytosis, specifcally the caveolin-dependent Not only can DNA devices interrogate combinations of markers pathway. Once inside cells, TDNs are actively transported along present on a cell surface, but the nanometre-precision addressability microtubules and eventually accumulate in lysosomes. To dem- of an origami can be used to control the density of ligands for cell onstrate that DNA nanostructures are capable of targeting difer- surface receptors66. Shaw et al. constructed ‘DNA nanocalipers’ by ent intracellular locations, Liang et al. coupled TDNs to nuclear arranging ephrin-A5 ligands on a DNA origami scafold. By taking localization signal peptides, thus successfully directing them to advantage of the positional addressability possible with DNA ori- the nucleus76. gami, they showed that cells are sensitive to the spatial arrangement Even though DNA nanostructures can enter cells with surprising of ligands. efciency, additional modifcations can be used to further enhance Te addressability of DNA systems can also be used to imbue uptake or increase stability. For example, Mikkila et al. demonstrated cells with a novel identity. Francis and colleagues developed a that rectangular DNA origami coated with viral capsid proteins method for afxing synthetic single-stranded DNA strands to liv- were taken up by human embryonic kidney cells at an efciency ten ing cells, thereby allowing an unlimited number of coding specifci- times that of the same origami delivered with Lipofectamine 200077. ties67. Cells were treated with a synthetic sugar that was metabolized Perrault and Shih constructed DNA nano-octahedrons encap- and integrated into the cell membrane where they could then sulated in a lipid bilayer. By incorporating lipid-coupled DNA serve as ‘chemical handles’. Phosphine-modifed single-stranded oligos, the octahedron served as a template for the formation of a DNA molecules were then attached to the ‘handles’ through surrounding lipid shell78. Te encapsulated octahedrons showed Staudinger ligation68. reduced immune activation and dramatically enhanced bioavail- Gartner and Bertozzi showed that this approach can be used to ability in circulation in mice, compared with non-encapsulated organize cells into 3D multicellular assemblies69. As a demonstra- controls (Fig. 1b). tion of structure guiding the function of a tissue, the authors gen- erated a ‘microtissue’ using assemblies of Chinese hamster ovary Drug delivery. CpG oligodeoxynucleotides (ODNs) are DNA cells expressing external growth factor interleukin-3 (IL-3) to sup- sequences with unmethylated cytosine-phosphate-guanine stretches port the growth of haematopoietic progenitor cells (FL5.12 cells); that can trigger a strong innate immune response by activating growth of the progenitor cells only occurred when they were associ- the Toll-like receptor TLR979. CpG ODNs are an attractive thera- ated with the IL-3 secreting cells. Te Gartner group went on to use peutic cargo because they can be integrated directly into any DNA this strategy to investigate the efect of cell-to-cell variation in Ras nanostructure through hybridization. Takakura and co- workers signalling on the morphogenesis of microtissues70. engineered Y-shaped DNA with CpG motifs to trigger immune In addition to mimicking cell–cell adhesion moieties, DNA can responses80,81. Tey found that Y-shaped DNA, compared with con- also be used to imitate other protein functions. Trough modifca- ventional single- or double-stranded DNA, are more efciently taken tion with hydrophobic groups, DNA structures can be inserted into up by macrophage cells, thus enhancing immune stimulation. Later lipid membranes71,72. Burns et al. demonstrated that a DNA nano- works have demonstrated efcient uptake and the activation of an pore channel inserted into the membrane of mammalian cells has a immune response with multi-arm82, TDN83 and origami structures75 number of cytotoxic efects73, possibly due to the free movement of functionalized with multiple CpG ODNs. critical ions, nutrients and other molecules across the membrane. Chang and co-workers created a synthetic vaccine complex by assembling TDNs that were modifed with both streptavidin and DNA nanostructures as drug-delivery vehicles CpG ODNs84. Streptavidin served as a model antigen, whereas the Te work discussed so far not only demonstrates the feasibility of CpG ODN was an adjuvant used to enhance the immune response. operating DNA nanodevices and structures in cell lysates and cell Te construct was frst tested in a mouse macrophage-like cell line culture (and in insects), but also shows that nanosystems can inter- and then injected into mice. Mice injected with the fully assembled act with cell surface proteins. We now move on to review the chal- complex developed higher levels of anti-streptavidin IgG anti bodies lenges associated with the delivery of nanodevices into mammalian than control mice injected with a simple mixture of streptavidin cells, and also discuss their use as vehicles for drug delivery. and CpG. DNA nanostructures have also been designed to serve as car- Cellular uptake of DNA nanostructures. By engineering folate- riers for doxorubicin, a cytotoxic drug that is used in a variety conjugated DNA nanotubes to target the folate receptors that are of cancer therapies. Previous work has shown that using nano- overexpressed on many cancer cells, Mao et al. successfully dem- particles to package doxorubicin could reduce side efects and onstrated that large DNA nanostructures can enter cells. Tey also dramatically increase circulation time in the body85. Moreover, further modifed the nanotubes with a fuorescent label to confrm because doxorubicin intercalates in DNA, it is a natural match for that the nanotubes, or at least nanotube fragments, were internalized delivery with a DNA vehicle — an idea frst introduced in the DNA upon receptor binding7. aptamer feld58.

NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology 753

a Short interfering DNA (siRNA) and ASO delivery have also been explored. By inserting a DNA loop containing an antisense sequence into one edge of a TDN, Keum et al. showed that the ASO H+ Salt reduced protein expression through RNaseH-mediated degradation SE TGN of a target mRNA in a sequence-specifc manner92. TDN-mediated Salt OH– siRNA delivery to tumours in a mouse model was demonstrated by Lee et al.93, who hybridized siRNA to TDNs that were also conju- RE gated to cancer-targeting ligands. Tey showed that in cell culture, both the number and relative orientation of ligands afected uptake LE and gene-silencing efciency. Te ligand-coupled TDNs exhibited H+ Salt high tissue specifcity in mice, accumulating mostly in the kidney and the tumour, but negligibly in other organs; it should be noted Salt OH– that the structures used in vivo difered in the number of attached Fu-IFu Tf-ITf TF receptor ligands from those characterized in vitro. Moreover, it remains difcult to judge how robust the structures were in such an environment b AND logic as no data was provided on the stability of the structures in vivo. Finally, Weizmann and collaborators created a periodic origami Substrate nanoribbon using a scafold strand obtained by rolling circle ampli- AND gate fcation. Nanoribbons entered cells through a clathrin-mediated pathway, and siRNA covalently attached to the structure resulted Input A in better knockdown than siRNA that were simply mixed in with miR21 Input B 94 miR125b the nanoribbons . Te work discussed in this section outlines an important proof-of- Figure 4 | DNA nanomachines and logic gates in mammalian cells. principle: DNA nanostructures can serve as drug-delivery vehicles. a,b, pH-sensitive DNA nanomachines for simultaneously probing the What sets DNA structures apart from approaches based on nano- furin (Fu) and transferrin (Tf) pathways97. Left: A transferrin-modified particles or polymers is primarily the programmability of a DNA DNA nanomachine Tf–ITf is confined to the transferrin pathway. The scafold, but also the very high degree of shape and size uniformity nanomachine enters a sorting endosome (SE), then a recycling endosome that can be achieved. (RE), and eventually returns to the membrane. The DNA nanomachine Fu–IFu targets the furin pathway: it enters the SE, then late endosome (LE), Dynamic DNA nanodevices inside living cells and eventually localizes in the trans-Golgi network (TGN). Nanomachine In this section, we review work on dynamic nucleic acid devices fluorescence is sensitive to pH, which varies between diferent endosomal that operate within cells and respond to specifc environmental compartments. Right: pH-sensitive elements of DNA nanomachines IFu cues; this includes devices that sense global environmental variables (green strands, top) and DNA nanomachines ITf (pink strands, bottom) such as pH, and recent progress towards detecting specifc molecu- form i-motif at low pH, which causes high FRET between the two lar information carriers such as cellular RNA. Finally, we discuss fluorophores. b, A DNAzyme-based AND logic gate operates inside living how the output of a nucleic acid device could allow modulation cells113. Left: Synthesized inputs with the sequences of miR-21 and mir- of gene expression levels and review frst steps towards the con- 125b are micro-injected together with the logic AND gate. Right: Reaction struction of logic circuits for the detection and analysis of multiple mechanism. Input B first binds to the hairpin (green segment), which is molecular markers. then available to interact with input A to join the two components of the AND gate. The joined DNAzyme complex can then cleave the substrate, Sensing the cellular environment. Functional nucleic acids such thus leading to high fluorescence by separating a fluorophore (red dot) as DNAzymes and aptamers have been used extensively within from a quencher (black dot). Figure reproduced with permission from: the biosensor feld to detect the levels of various molecular species a, ref. 97, Nature Publishing Group; b, ref. 113, Nature Publishing Group. within cells. Here we will highlight two results that combine such sensing moieties with DNA nanomotors or structures. Chang et al. were the frst to use DNA nanostructures — specif- Pei et al. constructed a set of TDNs with one or two reconfgur- cally wireframe icosahedra — to deliver doxorubicin to cancer cells86. able edges, which allowed the TDN to change its shape in response Jiang et al. built on this idea and created DNA origami triangles and to specifc molecular signals such as protons, ATP and mercury tubes to carry doxorubicin into human breast adenocarcinoma ions95. Using a FRET reporter strategy, they showed that a recon- cancer cells. Te much larger size of the origami structures further fgurable TDN changed conformation in response to intracellular enhanced the amount of doxorubicin that could be delivered, and ATP. Tis demonstrated the feasibility of combining a DNA nano- the drug-loaded complexes showed toxicity not only in regular can- structure with cellular sensors, which is an important property for cer cells, but also in doxorubicin-resistant cancer cells87. Kim et al. any potential ‘smart drug’. similarly found that the TDN-based delivery of doxorubicin resulted Along a similar line, Modi et al. used a DNA-based sensor to in drug activity in an otherwise drug-resistant cell line88. When map the pH of endosomal pathways in living cells96. Te design was employing DNA origami nanotubes with varying degrees of global based on Yurke’s DNA tweezers — essentially two double helices twist, Hogberg and collaborators found that doxorubicin loading, connected with a fexible hinge — but incorporated an i-quadruplex delivery and release could be tuned by changing the amount of twist structure that acts as a pH-sensitive switch to open and close the in the structure89. Zhu et al. constructed a DNA polymer based on tweezers (Fig. 4a). T e sensor was taken up by fy haemocytes an HCR-like mechanism to deliver doxorubicin90. Te polymers through endocytosis and trafcked from early endosomes (pH ~6) could specifcally target cancer cells via the recognition ability of an to late endosomes (pH ~5.5), and fnally to lysosomes (pH ~5). aptamer, inhibiting tumour growth in mice. Finally, afer injecting Te increasingly acidic environment resulted in quantifable fuo- DNA origami triangles loaded with doxorubicin into the tail veins of rescence changes and thus an indirect measurement of the pH. tumour-bearing mice, Zhang et al. found that origami-based deliv- Coupling the sensor to the protein transferrin allowed the pH ery resulted in a faster reduction of tumour mass than delivery of changes to be mapped along a specifc receptor-mediated endo- equal amounts of doxorubicin without a DNA origami carrier91. cytic pathway. In a follow-up study, the same group showed that

754 NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology

a mRNA 3´ UTR target b

123123

β-actin mRNA Scrambled probe Merge

c Control Target mRNA

mRNA 20 µm

Figure 5 | mRNA imaging in living cells. a, Ratiometric bimolecular beacons (RBMBs)101. Top: Binding to a target mRNA separates the reporter dye (red dot) from the quencher (black dot), which results in high fluorescence. Multiple RBMBs can bind to the tandem repeat targets in the 3’UTR of a heterologous mRNA, thereby enabling visualization of a single transcript in living cells. A reference dye (pink dot) is used to control cell-to-cell variation in molecular beacon delivery. Bottom: Fluorescence images of HT1080 cells using RBMB and FISH probes for the same mRNA. Image 1: Fish probes; Image 2: RBMB reporter dye; Image 3: A merged image that also includes nuclear DAPI stain (blue). b, Multiply labelled tetravalent RNA imaging probes (MTRIPs)103. Top: MTRIPs consist of multiple fluorophore labelled oligonucleotides attached to streptavidin (purple). Multiple MTRIPs can be designed to hybridize to a target mRNA, thus making single mRNA visible in living cells. Bottom: Deconvoluted confocal microscopy images of individual β-actin mRNA in an A549 cell. Image 1: MTRIPs; Image 2: Scrambled probes; Image 3: A merged image that includes nuclear DAPI stain. c, Nanoflares104–107. Left: A nanoflare contains long ‘capture strands’ and fluorophore-labelled ‘flare strands’, which are initially quenched by the gold nanoparticle. Target mRNAs can bind to ‘capture strands’, displace the ‘flare strand’ and trigger an increase in fluorescence. Right: Confocal fluorescence microscopy images of HeLa cells treated with either control nanoflares (left) or Survivin (target mRNA) nanoflares (right). Figure reproduced with permission from: a, ref. 101, © Oxford Univ. Press; b, ref. 103, Nature Publishing Group; c, ref. 106, American Chemical Society. pH-sensitive nanomachines could simultaneously track multiple a transfer RNA (tRNA) sequence resulted in active export of the pathways in the same cell97. probe from the nucleus to the cytoplasm, thus facilitating interactions with mRNA99–101 (Fig. 5a). Chemical modifcations, in par- Sensing cellular RNA. Te identity and health of a cell can ofen be ticular 2ʹOMe RNA, have been used to enhance probe stability by inferred by its RNA repertoire. However, the detection of specifc protecting against degradation by cellular nucleases102. As in the cellular nucleic acids can be challenging because it requires nanode- single-cell FISH techniques described above, co-localization of vices to access the cytoplasm, where most mature mRNA or miRNA multiple probes on the same transcript can improve the signal-to- are located. Furthermore, the low copy number of many RNA spe- background ratio101. Multivalent probe designs such as MTRIPs, in cies may make signal amplifcation necessary, and the secondary which several linear oligonucleotides are attached to a streptavidin structure or RNA-binding proteins can reduce the accessibility of core, can also result in a stronger signal103 (Fig. 5b). certain sequences. Nanofares, developed by Mirkin and co-workers, provided the frst Te live cell imaging feld has addressed many of these issues example of a DNA strand displacement reaction with an RNA input in and has developed a number of nucleic acid probes for detecting live cells (Fig. 5c). Nanofares consist of gold nanoparticles function- specifc mRNA sequences in live cells. Chemical and structural alized with DNA oligonucleotides complementary to an mRNA or modifcations used to improve the performance of live-cell imaging miRNA target. Shorter fuorescently labelled oligos are hybridized to probes could also be used to enhance the intracellular performance the nanofares and quenched by proximity to the gold nanoparticle104. of DNA nanodevices. Molecular beacons — stem-loop probes with Binding to the target displaces the fuorescently labelled strand, which a fuorophore and quencher attached to the stem — are probably results in increased fuorescence. A modifed version of the nanofare the best-studied class of probes for detecting mRNA in living cells98. technology used an LNA-modifed oligonucleotide to increase bind- Fluorescence is quenched when the probe is delivered but becomes ing strength with RNA targets while simultaneously targeting them unquenched when the probe hybridizes to the target mRNA. for degradation105,106. Halo et al. further demonstrated that nanofares, Variations on this basic design have resulted in improved perfor- in combination with fow cytometry, can be used to distinguish live mance: the addition of a longer double-stranded RNA domain or circulating tumour cells in the context of whole blood107.

NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology 755

Modulating cellular RNA. Existing non-coding nucleic acid tech- transcription requires extensive adjustments to the experimental nologies for gene-regulation, such as ASOs, ribozymes and siRNAs, approach and molecular design. Although this may seem like a sig- not only inform the design of cellular DNA nanosystems, but could nifcant challenge, considerable progress has already been made. also be integrated with DNA nanodevices as a means of controlling Here we highlight a few intriguing results using transcribed RNA the cellular environment. systems that straddle the increasingly blurred line between nucleic Afonin et al. demonstrated the assembly of a functional siRNA acid nanotechnology and synthetic biology. from two DNA:RNA complexes that individually did not enter the Cell-free RNA nanotechnology has resulted in a variety of 2D RNAi pathway in cells108. Te reaction was initiated by hybridiza- and 3D structures, and the size of structures that can be realized tion of complementary single-stranded RNA overhangs present in with RNA is rapidly increasing115–120. For example, Afonin et al. the two inactive complexes and is likely to have proceeded through demonstrated co-transcriptional, isothermal assembly of RNA four-way strand exchange109. siRNA activity was observed in cell cubes made from six or ten ~40 bp strands120. Even more recently, culture and tumour xenograf mouse models. Geary, Rothemund and Andersen demonstrated the feasibility of It would be even more intriguing if the activation of a regula- the computational design and experimental implementation of co- tory response could be conditional on the detection of a specifc transcriptional folding of RNA origami tiles up to 660 nucleotides in molecular marker. Benenson and colleagues took a frst step in size121. Tese tiles also self-assembled into larger lattice structures, this direction by designing nucleic acid displacement circuits that with dimensions reaching hundreds of nanometres. interact with components of a cell lysate110: detection of an RNA Delebecque et al. created repeating RNA scafolds that self- sequence added to the lysate triggered a strand displacement reac- assemble in bacteria. Transcribed from plasmids in Escherichia coli, tion, leading to the creation of a functioning siRNA. Pierce and these RNA scafolds were used to facilitate fux through a metabolic collaborators demonstrated a more general mechanism for the pathway. Te enzymes [FeFe] hydrogenase and ferrodoxin cata- conditional formation of a Dicer substrate RNA in a cell-free bio- lyse the reduction of protons to hydrogen. Fusing these proteins to chemical assay111. Yokobayashi and co-workers built a genetically RNA-binding domains allowed them to interact with RNA aptam- encoded RNA hairpin system that formed a substrate for the RNAi ers expressed in the same cells. Scafolding was achieved by chain- pathway upon activation by a synthetic, exogenously delivered ing the aptamers into repeating 2D units using a double-crossover input oligonucleotide112. motif; the scafold-associated enzymes improved hydrogen output almost 50-fold122. Tis approach has recently been further refned Molecular computation. Te realization of multi-input, multi-layer and extended to increase the efciency of pentadecane synthesis in molecular circuits is one of the major accomplishments of dynamic E. coli 123. DNA nanotechnology. But what unique advantages would DNA Bhadra and Ellington used products from in vitro transcrip- nanotechnology bring to engineering cellular ‘biocomputers’, com- tion reactions to demonstrate dynamic RNA strand displace- pared with alternative technologies based on synthetic gene regu- ment cascades124. Tey employed transcribed RNA hairpins to latory networks? First, DNA circuits rely on components that are construct circuits capable of cascading, amplifcation and logic. mechanistically simple and rationally designed at the molecular Although developed in vitro, this demonstration — the feasibility level, which provides a high degree of control over the reaction path- of using products of transcription for the construction of dynamic way. Second, new, orthogonal components can be designed simply devices — suggests that a similar approach may be used in the cell. by changing sequence, which makes it easy to increase system size Crossing into the realm of synthetic biology, Isaacs et al. engineered in a modular fashion. Tird, most dynamic DNA devices have a a class of riboregulators that rely on a strand-displacement mecha- relatively small DNA footprint compared with systems assembled nism for activation. Riboregulators inhibit bacterial mRNA transla- from genetically encoded proteins. tion by hiding the ribosome binding site inside the stem of a hairpin Te Shapiro group microinjected a DNAzyme AND gate along loop; the repression can be relieved by expressing a short RNA that with miRNA-derived inputs into MCF7 breast carcinoma cells113 hybridizes to the loop domain and unfolds the stem structure125. (Fig. 4b). Te gate was protected from nucleases by the addition of Tese riboregulators have been further modifed by Green et al. to inverted thymidine groups to the 3ʹ ends. Gate activation was quan- relieve many of the sequence constraints126. tifed using fuorescence microscopy and, consistent with AND Exogenously produced DNA devices have the advantage of logic, fuorescence increased only in the presence of both inputs. closely mimicking those that have been developed for use in vitro. Strand-displacement DNA logic gates have also recently been On the other hand, by moving to systems that are transcribed within used to detect combinations of miRNA in living cells. Based on the cell, problems of delivery and expression become soluble using designs frst demonstrated in vitro27, Hemphill and Deiters used the more familiar tools of genetic engineering. As transcribed RNA an AND gate constructed from DNA to detect the endogenous nanotechnology develops, tools from more mature disciplines of miRNAs miR-122 and miR-21 in Huh7 hepatocellular carcinoma genetic engineering will become increasingly valuable. cells114. Gates were delivered using standard transfection reagents Recently, the possibility to reconcile these two approaches has and gate activation was observed only in cells that produced both emerged, thereby potentially allowing DNA systems to be expressed input miRNA. directly in cells rather than having to reinterpret them in a new Te results reviewed in this section suggest that the feld is mak- nucleic acid substrate. Te Lu group used a retron — a bacteria- ing rapid progress towards the design of dynamic DNA devices derived reverse transcriptase — to express single-stranded DNA in that can sense information in cells, analyse that information using bacteria. Tey used these DNA species to incrementally modify the embedded molecular control circuits and then respond by efecting specifc regions of the bacterial chromosome, thus creating a popula- changes in the cell. tion-level analog timer127. It is not much of a stretch to imagine that the same system could be further expanded to create components for Genetically encoded structures and devices DNA-based structures and dynamic systems. Applications from gene therapy to metabolic engineering require long-term embedded control of gene expression. RNA scafolds Outlook and regulatory elements that can be genetically encoded and DNA nanotechnology has made remarkable strides towards practi- transcribed in living cells are likely to be a better match for such cal applications in cellular settings. Nucleic acid structures in par- applications than transiently delivered synthetic DNA systems. ticular have already quite successfully made the transition from the Modifying existing DNA nanotechnology to be compatible with in vitro to the in vivo environment. Structures from tetrahedra to

756 NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology

70 as programmable, multifunctional, therapeutic systems that could In vitro circuits eventually rival viruses in sophistication. 60 In vivo circuits 37 DNA-based therapeutics and diagnostics are set apart from 39 more established approaches because of their capacity to respond 50 to the surrounding environment. Molecular logic and conditional (un)hiding of drug moieties could decrease side efects and increase 40 ations specifcity. Even the relatively simple one- or two-input systems built so far have resulted in increased specifcity and performance, and Oper 30 could be further improved with more complex multi-input logic. 34 35 Diagnostic and therapeutic decisions are routinely based on the 20 27 59 analysis of panels of multiple molecular markers, be they proteins, 97 38 ? RNA, DNA, lipids, sugars or metabolites. For example, immunolo- 10 6 41 96 113 106 60 114 107 gists must ofen consider large numbers of cell surface proteins to 0 delineate all of the various cell types in a blood sample. Gene expres- 2003 2006 2009 2012 2015 sion classifers that reliably distinguish diferent tissues and disease Year of publication states are typically built on measurements of tens or hundreds of diferent RNA species. Given the success of dynamic DNA nano- Figure 6 | Complexity break for cellular DNA nanodevices? The complexity technology in scaling up the size and reliability of molecular circuits of cell-free DNA logic circuits and similar dynamic devices has increased by in cell-free settings, it is intriguing to think that DNA ‘biocomput- almost two orders of magnitude over the past decade. In cellular settings, ers’ could eventually perform complex diagnostic tasks based on the dynamic devices with only two or three independent operations have so analysis of tens of molecular markers directly in living organisms. far been demonstrated. This suggests that design principles for adapting Beyond diagnostic and therapeutic devices, we could imagine dynamic DNA nanodevices to cells are yet to be uncovered. Each coloured synthetic DNA ecosystems that integrate motors, logic, structural dot and number represent a specific reaction network and associated elements and more to control and interrogate cellular behaviour in publication (reference number); trend lines are included to guide the eye. time and space. To realize such a vision and go beyond the delivery An operation is defined as a unique (sequence-specific) connection, such of mostly static structures, we still need to identify broadly appli- as a strand displacement reaction or DNAzyme cleavage event within a cable design principles that make it easy to translate any device network. A circuit with n gates arranged in a cascade is considered to be that works reliably in cell-free settings to the cellular environment equally complex as a circuit with n independent gates operating in parallel, (Fig. 6). New design strategies might include the delivery method, even though the latter is probably easier to realize experimentally. Moreover, nucleic acid chemistry and sequence design, or even diferent reac- multi-turnover catalytic reactions are weighed equally against single-step tion mechanisms. However, given the progress that has already been reactions, which potentially underestimates the complexity of the former. made, it is quite likely that DNA nanotechnology will become a useful complement to more traditional approaches for manipulating origami have been shown to be stable in cells and can be readily and controlling biological information. modifed into molecular transportation devices for siRNA, antibodies or small-molecule drugs. Te most immediate application Received 28 January 2015; accepted 29 July 2015; might be to use DNA nanostructures as programmable tools for published online 3 September 2015 interrogating cellular processes66. Even the use of nanostructures as multifunctional carriers for drug delivery seems to be within References reach, with several groups already demonstrating functionality of 1. Bloomfeld, V. A., Crothers, D. M. & Ignacio Tinoco, J. Nucleic Acids: nanostructure-based therapeutics in mouse models. Structures, Properties and Functions (University Science Books, 2000). What challenges need to be met before DNA nanostructures can 2. SantaLucia, J. & Hicks, D. Te thermodynamics of DNA structural motifs. compete with alternative delivery technologies based on liposomes, Annu. Rev. Biophys. Biomol. Struct. 33, 415–440 (2004). polymers, aptamers and others58,85? First, we must develop a bet- 3. Carlson, R. Te changing economics of DNA synthesis. Nature Biotechnol. ter understanding of the pharmacokinetics and biodistribution 27, 1091–1094 (2009). 4. Dittmer, W. U., Reuter, A. & Simmel, F. C. A. DNA-based machine of DNA nanostructures. For example, one challenge is to design that can cyclically bind and release thrombin. Angew. Chem. Int. Ed. approaches that allow the selective uptake — while inhibiting the 43, 3550–3553 (2004). non-selective uptake — of DNA nanostructures by target cells; this 5. Yurke, B., Mills, A. P. Jr & Cheng, S. L. DNA implementation of addition in is a common hurdle for any drug carrier, but is something the DNA which the input strands are separate from the operator strands. BioSystems nanotechnology feld will need to work out. Additionally, potential 52, 165–174 (1999). immunostimulatory properties of DNA nanostructures must be 6. Benenson, Y., Gil, B., Ben-Dor, U., Adar, R. & Shapiro, E. An autonomous molecular computer for logical control of gene expression. Nature investigated. Finally, production cost could be a concern, especially 429, 423–429 (2004). for large nanostructures such as DNA origami, which will prob- 7. Ko, S., Liu, H., Chen, Y. & Mao, C. DNA nanotubes as combinatorial vehicles ably be at least as expensive to produce as antibodies ($300 g–1 or for cellular delivery. Biomacromolecules 9, 3039–3043 (2008). less128) or aptamers ($50 g–1, ref. 58). Moreover, if the DNA nano- Cellular uptake of large DNA nanostructures was frst demonstrated in structures primarily serve as drug carriers, they will need to com- this work. pete with polymer materials that can cost less than $1 g–1. However, 8. Seeman, N. C. Nucleic acid junctions and lattices. J. Teor. Biol. 99, 237–247 (1982). alternative methods for synthesizing large amounts of high-quality 9. Kallenbach, N. R., Ma, R.-I. & Seeman, N. C. An immobile nucleic acid DNA are being explored. For example, recently developed high-cell- junction constructed from oligonucleotides. Nature 305, 829–831 (1983). density bioreactors can efciently generate large quantities of M13 10. Chen, J. & Seeman, N. C. Synthesis from DNA of a molecule with the phage genome129. Combined with other approaches for producing connectivity of a cube. Nature 350, 631–633 (1991). high-quality short oligonucleotides in cells or enzymatically130,131, 11. Winfree, E., Liu, F., Wenzler, L. A. & Seeman, N. C. Design and self-assembly this technology provides a promising start towards developing low- of two-dimensional DNA crystals. Nature 394, 539–544 (1998). 12. Goodman, R. P. et al. Rapid chiral assembly of rigid DNA building blocks for cost production methods for DNA nanostructures in the future. molecular nanofabrication. Science 310, 1661–1665 (2005). Tus, it seems very likely that these challenges will be overcome, 13. Zheng, J. et al. From molecular to macroscopic via the rational design of a self- given the truly unique potential of DNA nanostructures to serve assembled 3D DNA crystal. Nature 461, 74–77 (2009).

NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology 757

14. Shih, W. M., Quispe, J. D. & Joyce, G. F. A 1.7-kilobase single-stranded DNA 50. Castro, C. E. et al. A primer to scafolded DNA origami. Nature Methods that folds into a nanoscale octahedron. Nature 427, 618–621 (2004). 8, 221–229 (2011). 15. Yan, H., LaBean, T. H., Feng, L. & Reif, J. H. Directed nucleation assembly of 51. Choi, H. M. T. et al. Programmable in situ amplifcation for multiplexed DNA tile complexes for barcode-patterned lattices. Proc. Natl Acad. Sci. USA imaging of mRNA expression. Nature Biotechnol. 28, 1208–1212 (2010). 100, 8103–8108 (2003). 52. Choi, H. M. T., Beck, V. A. & Pierce, N. A. Next-generation in situ 16. Schulman, R. & Winfree, E. Synthesis of crystals with a programmable kinetic hybridization chain reaction: higher gain, lower cost, greater durability. barrier to nucleation. Proc. Natl Acad. Sci. USA 104, 15236–15241 (2007). ACS Nano 8, 4284–4294 (2014). 17. Rothemund, P. W. K., Papadakis, N. & Winfree, E. Algorithmic self-assembly 53. Levesque, M. J., Ginart, P., Wei, Y. & Raj, A. Visualizing SNVs to quantify of DNA Sierpinski triangles. PLoS Biol. 2, e424 (2004). allele-specifc expression in single cells. Nature Methods 10, 865–867 (2013). 18. He, Y. et al. Hierarchical self-assembly of DNA into symmetric supramolecular Taking advantage of the specifcity of toehold-mediated strand polyhedra. Nature 452, 198–201 (2008). displacement reactions, this work demonstrated that single-nucleotide 19. Yan, H., Park, S. H., Finkelstein, G., Reif, J. H. & LaBean, T. H. DNA-templated variants on single RNA transcripts can be detected using smFISH-based self-assembly of protein arrays and highly conductive nanowires. Science imaging probes. 301, 1882–1884 (2003). 54. Raj, A., van den Bogaard, P., Rifin, S. A., van Oudenaarden, A. & Tyagi, S. 20. Rothemund, P. W. K. Folding DNA to create nanoscale shapes and patterns. Imaging individual mRNA molecules using multiple singly labeled probes. Nature 440, 297–302 (2006). Nature Methods 5, 877–879 (2008). 21. Douglas, S. M. et al. Self-assembly of DNA into nanoscale three-dimensional 55. Duose, D. Y. et al. Confguring robust DNA strand displacement reactions for shapes. Nature 459, 414–418 (2009). in situ molecular analyses. Nucleic Acids Res. 40, 3289–3298 (2012). 22. Ke, Y. et al. Scafolded DNA origami of a DNA tetrahedron molecular 56. Jungmann, R. et al. Single-molecule kinetics and super-resolution container. Nano Lett. 9, 2445–2447 (2009). microscopy by fuorescence imaging of transient binding on DNA origami. 23. Andersen, E. S. et al. Self-assembly of a nanoscale DNA box with a Nano Lett. 10, 4756–4761 (2010). controllable lid. Nature 459, 73–76 (2009). 57. Jungmann, R. et al. Multiplexed 3D cellular super-resolution imaging with 24. Dietz, H., Douglas, S. M. & Shih, W. M. Folding DNA into twisted and curved DNA-PAINT and Exchange-PAINT. Nature Methods 11, 313–318 (2014). nanoscale shapes. Science 325, 725–730 (2009). 58. Keefe, A., Pai, S. & Ellington, A. Aptamers as therapeutics. 25. Adleman, L. M. Molecular computation of solutions to combinatorial Nature Rev. Drug Discov. 9, 537–550 (2010). problems. Science 266, 1021–1024 (1994). 59. Rudchenko, M. et al. Autonomous molecular cascades for evaluation of cell 26. Yurke, B., Turberfeld, A. J., Mills, A. P. Jr, Simmel, F. C. & Neumann, J. L. surfaces. Nature Nanotech. 8, 580–586 (2013). A DNA-fuelled molecular machine made of DNA. Nature 406, 605–608 (2000). Tis work successfully used strand displacement cascades to classify 27. Seelig, G., Soloveichik, D., Zhang, D. Y. & Winfree, E. Enzyme-free nucleic diferent cell types, thereby demonstrating a scalable approach for the acid logic circuits. Science 314, 1585–1588 (2006). analysis of cellular information. 28. Dirks, R. M. & Pierce, N. A. Triggered amplifcation by hybridization chain 60. Douglas, S. M., Bachelet, I. & Church, G. M. A logic-gated nanorobot for reaction. Proc. Natl Acad. Sci. USA 101, 15275–15278 (2004). targeted transport of molecular payloads. Science 335, 831–834 (2012). 29. Kay, E. R., Leigh, D. A & Zerbetto, F. Synthetic molecular motors and Proof-of-principle demonstration of a novel class of conditional mechanical machines. Angew. Chem. Int. Ed. 46, 72–191 (2007). therapeutics that combine protective DNA origami structures with 30. Bath, J. & Turberfeld, A. J. DNA nanomachines. Nature Nanotech. molecular logic. 2, 274–284 (2007). 61. Tuerk, C. & Gold, L. Systematic evolution of ligands by exponential 31. Omabegho, T., Sha, R. & Seeman, N. C. A bipedal DNA Brownian motor with enrichment: RNA ligands to bacteriophage T4 DNA polymerase. Science coordinated legs. Science 324, 67–71 (2009). 249, 505–510 (1990). 32. Lund, K. et al. Molecular robots guided by prescriptive landscapes. Nature 62. Ellington, A. D. & Szostak, J. In vitro selection of RNA molecules that bind 465, 206–210 (2010). specifc ligands. Nature 346, 818–822 (1990). 33. Muscat, R. A., Bath, J. & Turberfeld, A. J. A programmable molecular robot. 63. Amir, Y. et al. Universal computing by DNA origami robots in a living Nano Lett. 11, 982–987 (2011). animal. Nature Nanotech. 9, 353–357 (2014). 34. Wickham, S. F. J. et al. A DNA-based molecular motor that can navigate a 64. You, M. et al. DNA ‘nano-claw’: logic-based autonomous cancer targeting network of tracks. Nature Nanotech. 7, 169–173 (2012). and therapy. J. Am. Chem. Soc. 136, 1256–1259 (2014). 35. Chen, Y.-J. et al. Programmable chemical controllers made from DNA. 65. You, M., Zhu, G., Chen, T., Donovan, M. J. & Tan, W. Programmable and Nature Nanotech. 8, 755–762 (2013). multiparameter DNA-based logic platform for cancer recognition and 36. Qian, L., Winfree, E. & Bruck, J. Neural network computation with DNA targeted therapy. J. Am. Chem. Soc. 137, 667–674 (2015). strand displacement cascades. Nature 475, 368–372 (2011). 66. Shaw, A. et al. Spatial control of membrane receptor function using ligand 37. Qian, L. & Winfree, E. Scaling up digital circuit computation with DNA strand nanocalipers. Nature Methods 11, 841–846 (2014). displacement cascades. Science 332, 1196–1201 (2011). By showing that cells are sensitive to the spatial organization of protein 38. Elbaz, J. et al. DNA computing circuits using libraries of DNAzyme subunits. ligands arranged on a DNA origami, the authors provide an intriguing Nature Nanotech. 5, 417–422 (2010). example of the use of nanostructures as tools for cell biology. 39. Pei, R., Matamoros, E., Liu, M., Stefanovic, D. & Stojanovic, M. N. Training a 67. Chandra, R. A., Douglas, E. S., Mathies, R. A., Bertozzi, C. R. & molecular automaton to play a game. Nature Nanotech. 5, 773–777 (2010). Francis, M. B. Programmable cell adhesion encoded by DNA hybridization. 40. Seelig, G., Yurke, B. & Winfree, E. Catalyzed relaxation of a metastable DNA Angew. Chem. Int. Ed. 45, 896–901 (2006). fuel. J. Am. Chem. Soc. 128, 12211–12220 (2006). 68. Saxon, E. & Bertozzi, C. R. Cell surface engineering by a modifed Staudinger 41. Zhang, D. Y., Turberfeld, A. J., Yurke, B. & Winfree, E. Engineering reaction. Science 287, 2007–2010 (2000). entropy-driven reactions and networks catalyzed by DNA. Science 69. Gartner, Z. J. & Bertozzi, C. R. Programmed assembly of 3-dimensional 318, 1121–1125 (2007). microtissues with defned cellular connectivity. Proc. Natl Acad. Sci. USA 42. Zhang, D. Y. & Winfree, E. Dynamic allosteric control of noncovalent DNA 106, 4606–4610 (2009). catalysis reactions. J. Am. Chem. Soc. 130, 13921–13926 (2008). Tis work demonstrated a novel strategy for the bottom-up construction 43. Turberfeld, A. J. et al. DNA fuel for free-running nanomachines. of ‘microtissues’ using DNA sequence-programmed connectivity. Phys. Rev. Lett. 90, 118102 (2003). 70. Liu, J. S., Farlow, J. T., Paulson, A. K., Labarge, M. A. & Gartner, Z. J. 44. Bois, J. S. et al. Topological constraints in nucleic acid hybridization kinetics. Programmed cell-to-cell variability in Ras activity triggers emergent Nucleic Acids Res. 33, 4090–4095 (2005). behaviors during mammary epithelial morphogenesis. Cell Rep. 45. Benenson, Y. et al. Programmable and autonomous computing machine made 2, 1461–1470 (2012). of biomolecules. Nature 414, 430–434 (2001). 71. Langecker, M. et al. Synthetic lipid membrane channels formed by designed 46. Mei, Q. et al. Stability of DNA origami nanoarrays in cell lysate. Nano Lett. DNA nanostructures. Science 338, 932–936 (2012). 11, 1477–1482 (2011). 72. Burns, J. R. et al. Lipid-bilayer-spanning DNA nanopores with a bifunctional 47. Conway, J. W., McLaughlin, C. K., Castor, K. J. & Sleiman, H. DNA porphyrin anchor. Angew. Chem. Int. Ed. 52, 12069–12072 (2013). nanostructure serum stability: greater than the sum of its parts. 73. Burns, J. R., Al-Jufali, N., Janes, S. M. & Howorka, S. Membrane- Chem. Commun. 49, 1172–1174 (2013). spanning DNA nanopores with cytotoxic efect. Angew. Chem. Int. Ed. 48. Hahn, J., Wickham, S. F. J., Shih, W. M. & Perrault, S. D. Addressing 53, 12466–12470 (2014). the instability of DNA nanostructures in tissue culture. ACS Nano 74. Walsh, A. S. et al. DNA cage delivery to mammalian cells. ACS Nano 8, 8765–8775 (2014). 5, 5427–5432 (2011). 49. Keum, J.-W. & Bermudez, H. Enhanced resistance of DNA nanostructures to 75. Schüller, V. J. et al. Cellular immunostimulation by CpG-sequence-coated enzymatic digestion. Chem. Commun. 7036–7038 (2009). DNA origami structures. ACS Nano 5, 9696–9702 (2011).

758 NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology

76. Liang, L. et al. Single-particle tracking and modulation of cell entry pathways 103. Santangelo, P. J. et al. Single molecule–sensitive probes for imaging RNA in live of a tetrahedral DNA nanostructure in live cells. Angew. Chem. Int. Ed. cells. Nature Methods 6, 347–349 (2009). 53, 7745–7750 (2014). 104. Rosi, N. L. et al. Oligonucleotide-modifed gold nanoparticles for intracellular 77. Mikkilä, J. et al. Virus-encapsulated DNA origami nanostructures for cellular gene regulation. Science 312, 1027–1030 (2006). delivery. Nano Lett. 14, 2196–2200 (2014). Nanofares provided the frst example of strand displacement reactions with 78. Perrault, S. D. & Shih, W. M. Virus-inspired membrane encapsulation of DNA an endogenous RNA input inside living cells. nanostructures to achieve in vivo stability. ACS Nano 8, 5132–5140 (2014). 105. Alhasan, A. H., Patel, P. C., Choi, C. H. J. & Mirkin, C. A. Exosome encased Te authors showed that lipid encapsulation of DNA octahedrons results in a spherical nucleic acid gold nanoparticle conjugates as potent microRNA reduced immune response and greatly enhanced bioavailability in circulation regulation agents. Small 10, 186–192 (2014). in mouse models. 106. Prigodich, A. E. et al. Nano-fares for mRNA regulation and detection. 79. Hemmi, H. et al. A Toll-like receptor recognizes bacterial DNA. Nature ACS Nano 3, 2147–2152 (2009). 408, 740–745 (2000). 107. Halo, T. L. et al. NanoFlares for the detection, isolation, and culture 80. Nishikawa, M., Matono, M., Rattanakiat, S., Matsuoka, N. & Takakura, Y. of live tumor cells from human blood. Proc. Natl Acad. Sci. USA Enhanced immunostimulatory activity of oligodeoxynucleotides by Y-shape 111, 17104–17109 (2014). formation. Immunology 124, 247–255 (2008). 108. Afonin, K. A. et al. Activation of diferent split functionalities on re-association Te frst demonstration of drug delivery using DNA nanostructures; of RNA–DNA hybrids. Nature Nanotech. 8, 296–304 (2013). Y-shaped DNA nanostructures decorated with CpG motifs were used to 109. Chen, S. X., Zhang, D. Y. & Seelig, G. Conditionally fuorescent molecular trigger immune responses in living cells. probes for detecting single base changes in double-stranded DNA. 81. Rattanakiat, S., Nishikawa, M., Funabashi, H., Luo, D. & Takakura, Y. Te Nature Chem. 5, 782–789 (2013). assembly of a short linear natural cytosine-phosphate-guanine DNA into 110. Xie, Z., Liu, S. J., Bleris, L. & Benenson, Y. Logic integration of mRNA signals by dendritic structures and its efect on immunostimulatory activity. Biomaterials an RNAi-based molecular computer. Nucleic Acids Res. 38, 2692–2701 (2010). 30, 5701–5706 (2009). 111. Hochrein, L. M., Schwarzkopf, M., Shahgholi, M., Yin, P. & Pierce, N. A. 82. Mohri, K. et al. Design and development of nanosized DNA assemblies in Conditional dicer substrate formation via shape and sequence transduction polypod-like structures as efcient vehicles for immunostimulatory cpg motifs to with small conditional RNAs. J. Am. Chem. Soc. 135, 17322–17330 (2013). immune cells. ACS Nano 6, 5931–5940 (2012). 112. Kumar, D., Kim, S. H. & Yokobayashi, Y. Combinatorially inducible 83. Li, J. et al. Self-assembled multivalent DNA nanostructures for noninvasive RNA interference triggered by chemically modifed oligonucleotides. intracellular delivery of immunostimulatory CpG oligonucleotides. ACS Nano J. Am. Chem. Soc. 133, 2783–2788 (2011). 5, 8783–8789 (2011). 113. Kahan-Hanum, M., Douek, Y., Adar, R. & Shapiro, E. A library of 84. Liu, X. et al. A DNA nanostructure platform for directed assembly of synthetic programmable DNAzymes that operate in a cellular environment. Sci. Rep. vaccines. Nano Lett. 12, 4254–4259 (2012). 3, 1535 (2013). 85. Davis, M. E., Chen, Z. (Georgia) & Shin, D. M. Nanoparticle therapeutics: 114. Hemphill, J. & Deiters, A. DNA computation in mammalian cells: microRNA an emerging treatment modality for cancer. Nature Rev. Drug Discov. logic operations. J. Am. Chem. Soc. 135, 10512–10518 (2013). 7, 771–782 (2008). 115. Yu, J., Liu, Z., Jiang, W., Wang, G. & Mao, C. De novo design of an RNA tile 86. Chang, M., Yang, C.-S. & Huang, D.-M. Aptamer-conjugated DNA icosahedral that self-assembles into a homo-octameric nanoprism. Nature Commun. nanoparticles as a carrier of doxorubicin for cancer therapy. ACS Nano 6, 1–6 (2015). 5, 6156–6163 (2011). 116. Lee, J. B., Hong, J., Bonner, D. K., Poon, Z. & Hammond, P. T. Self-assembled 87. Jiang, Q. et al. DNA origami as a carrier for circumvention of drug resistance. RNA interference microsponges for efcient siRNA delivery. Nature Mater. J. Am. Chem. Soc. 134, 13396–13403 (2012). 11, 316–322 (2012). 88. Kim, K.-R. et al. Drug delivery by a self-assembled DNA tetrahedron 117. Severcan, I. et al. A polyhedron made of tRNAs. Nature Chem. for overcoming drug resistance in breast cancer cells. Chem. Commun. 2, 772–779 (2010). 49, 2010–2012 (2013). 118. Ohno, H. et al. Synthetic RNA-protein complex shaped like an equilateral 89. Zhao, Y.-X. et al. DNA origami delivery system for cancer therapy with tunable triangle. Nature Nanotech. 6, 116–120 (2011). release properties. ACS Nano 6, 8684–8691 (2012). 119. Chworos, A. et al. Building programmable jigsaw puzzles with RNA. Science 90. Zhu, G. et al. Self-assembled, aptamer-tethered DNA nanotrains for targeted 306, 2068–2073 (2004). transport of molecular drugs in cancer theranostics. Proc. Natl Acad. Sci. USA 120. Afonin, K. A. et al. In vitro assembly of cubic RNA-based scafolds designed in 110, 7998–8003 (2013). silico. Nature Nanotechnol. 5, 676–682 (2010). 91. Zhang, Q. et al. DNA origami as an in vivo drug delivery vehicle for cancer 121. Geary, C., Rothemund, P. W. K. & Andersen, E. S. A single-stranded therapy. ACS Nano 8, 6633–6643 (2014). architecture for cotranscriptional folding of RNA nanostructures. Science 92. Keum, J. W., Ahn, J. H. & Bermudez, H. Design, assembly, and activity of 345, 799–804 (2014). antisense DNA nanostructures. Small 7, 3529–3535 (2011). 122. Delebecque, C. J., Lindner, A. B., Silver, P. A. & Aldaye, F. A. Organization 93. Lee, H. et al. Molecularly self-assembled nucleic acid nanoparticles for targeted of intracellular reactions with rationally designed RNA assemblies. Science in vivo siRNA delivery. Nature Nanotech. 7, 389–393 (2012). 333, 470–474 (2011). 94. Chen, G. et al. Enzymatic synthesis of periodic DNA nanoribbons Tis work used self-assembled RNA scafolds to increase the efciency of for intracellular pH sensing and gene silencing. J. Am. Chem. Soc. hydrogen production in bacteria, thus demonstrating the functional use of 137, 3844–3851 (2015). RNA architectures in vivo. 95. Pei, H. et al. Reconfgurable three-dimensional DNA nanostructures 123. Sachdeva, G., Garg, A., Godding, D., Way, J. C. & Silver, P. A. In vivo co- for the construction of intracellular logic sensors. Angew. Chem. Int. Ed. localization of enzymes on RNA scafolds increases metabolic production in a 51, 9020–9024 (2012). geometrically dependent manner. Nucleic Acids Res. 42, 9493–9503 (2014). 96. Modi, S. et al. A DNA nanomachine that maps spatial and temporal pH changes 124. Bhadra, S. & Ellington, A. D. Design and application of cotranscriptional inside living cells. Nature Nanotech. 4, 325–330 (2009). non-enzymatic RNA circuits and signal transducers. Nucleic Acids Res. 97. Modi, S., Nizak, C., Surana, S., Halder, S. & Krishnan, Y. Two DNA 42, e58 (2014). nanomachines map pH changes along intersecting endocytic pathways inside the 125. Isaacs, F. J. et al. Engineered riboregulators enable post-transcriptional control same cell. Nature Nanotech. 8, 459–467 (2013). of gene expression. Nature Biotechnol. 22, 841–847 (2004). 98. Tyagi, S. & Kramer, F. R. Molecular beacons: probes that fuoresce upon 126. Green, A. A., Silver, P. A., Collins, J. J. & Yin, P. Toehold switches: hybridization. Nature Biotechnol. 14, 303–308 (1996). de-novo-designed regulators of gene expression. Cell 159, 925–939 (2014). 99. Chen, A. K., Davydenko, O., Behlke, M. A. & Tsourkas, A. Ratiometric 127. Farzadfard, F. & Lu, T. K. Genomically encoded analog memory with precise bimolecular beacons for the sensitive detection of RNA in single living cells. in vivo DNA writing in living cell populations. Science 346, 6211 (2014). Nucleic Acids Res. 38, e148 (2010). 128. Kelley, B. Industrialization of mAb production technology: the bioprocessing 100. Mhlanga, M. M., Vargas, D. Y., Fung, C. W., Kramer, F. R. & Tyagi, S. tRNA- industry at a crossroads. MAbs 1, 440–449 (2009). linked molecular beacons for imaging mRNAs in the cytoplasm of living cells. 129. Kick, B., Praetorius, F., Dietz, H. & Weuster-Botz, D. Efcient production Nucleic Acids Res. 33, 1902–1912 (2005). of single-stranded phage DNA as scafolds for DNA origami. Nano Lett. 101. Zhang, X., Song, Y., Shah, A. & Lekova, V. Quantitative assessment of ratiometric 15, 4672–4676 (2015). bimolecular beacons as a tool for imaging single engineered RNA transcripts and 130. Ducani, C., Kaul, C., Moche, M., Shih, W. M. & Högberg, B. Enzymatic measuring gene expression in living cells. Nucleic Acids Res. 41, e152 (2013). production of ‘monoclonal stoichiometric’ single-stranded DNA 102. Bratu, D. P., Cha, B.-J., Mhlanga, M. M., Kramer, F. R. & Tyagi, S. Visualizing the oligonucleotides. Nature Methods 10, 647–652 (2013). distribution and transport of mRNAs in living cells. Proc. Natl Acad. Sci. USA 131. Gu, H. & Breaker, R. R. Production of single-stranded DNAs by self-cleavage 100, 13308–13313 (2003). of rolling-circle amplifcation products. Biotechniques 54, 337–343 (2013).

NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology 759

132. Gilleron, J. et al. Image-based analysis of lipid nanoparticle–mediated siRNA 141. Manche, L., Green, S. R., Schmedt, C. & Mathews, M. B. Interactions between delivery, intracellular trafcking and endosomal escape. Nature Biotechnol. double-stranded RNA regulators and the protein kinase DAI. Mol. Cell. Biol. 31, 638–646 (2013). 12, 5238–5248 (1992). 133. Sahay, G. et al. Efciency of siRNA delivery by lipid nanoparticles is limited 142. Krieg, A. M. Terapeutic potential of Toll-like receptor 9 activation. by endocytic recycling. Nature Biotechnol. 31, 653–658 (2013). Nature Rev. Drug Discov. 5, 471–484 (2006). 134. Bao, G., Rhee, W. J. & Tsourkas, A. Fluorescent probes for live-cell RNA 143. Shir, A. & Levitzki, A. Inhibition of glioma growth by tumor-specifc detection. Annu. Rev. Biomed. Eng. 11, 25–47 (2009). activation of double-stranded RNA-dependent protein kinase PKR. 135. Fisher, T. L., Terhorst, T., Cao, X. & Wagner, R. W. Intracellular disposition Nature Biotechnol. 20, 895–900 (2002). and metabolism of fuorescently-labeled unmodifed and modifed oligonucleotides microinjected into mammalian cells. Nucleic Acids Res. Acknowledgements 21, 3857–3865 (1993). We would like to thank S. Douglas, N. Pierce, M. Schwarzkopf, S. Pun and D. Soloveichik 136. Watts, J. K., Deleavey, G. F. & Damha, M. J. Chemically modifed siRNA: for insightful comments and helpful feedback on the manuscript. Tis work was tools and applications. Drug Discov. Today 13, 842–855 (2008). supported by NSF grants CCF-1317653 and CAREER CBET-0954566. 137. Amarzguioui, M., Holen, T., Babaie, E. & Prydz, H. Tolerance for mutations and chemical modifcations in a siRNA. Nucleic Acids Res. Author contributions 31, 589–595 (2003). All authors contributed equally to this work. 138. Bramsen, J. B. et al. A large-scale chemical modifcation screen identifes design rules to generate siRNAs with high activity, high stability and low Additional information toxicity. Nucleic Acids Res. 37, 2867–2881 (2009). Reprints and permissions information is available online at www.nature.com/reprints. 139. Lukacs, G. L. et al. Size-dependent DNA mobility in cytoplasm and nucleus. Correspondence should be addressed to G.S. J. Biol. Chem. 275, 1625–1629 (2000). 140. Schoen, I., Krammer, H. & Braun, D. Hybridization kinetics is diferent Competing ﬁnancial interests inside cells. Proc. Natl Acad. Sci. USA 106, 21649–21654 (2009). Te authors declare no competing fnancial interests.

760 NATURE NANOTECHNOLOGY | VOL 10 | SEPTEMBER 2015 | www.nature.com/naturenanotechnology