Challenges and Perspectives of DNA Nanostructures in Biomedicine

Abstract DNA nanotechnology holds substantial promise for future biomedical engineering and the development of novel therapies and diagnostic assays. The subnanometer‐level addressability of DNA nanostructures allows for their precise and tailored modification with numerous chemical and biological entities, which makes them fit to serve as accurate diagnostic tools and multifunctional carriers for targeted drug delivery. The absolute control over shape, size, and function enables the fabrication of tailored and dynamic devices, such as DNA nanorobots that can execute programmed tasks and react to various external stimuli. Even though several studies have demonstrated the successful operation of various biomedical DNA nanostructures both in vitro and in vivo, major obstacles remain on the path to real‐world applications of DNA‐based nanomedicine. Here, we summarize the current status of the field and the main implementations of biomedical DNA nanostructures. In particular, we focus on open challenges and untackled issues and discuss possible solutions.


Introduction
DNAi st he carrier of hereditary information and therefore serves as ac entral component of all life on Earth. That said, due to its unique chemical and structural properties, DNAc an also be used as ap rogrammable material [1] for the controlled synthesis of molecularly defined artificial nanostructures and even switchable nanodevices. [2] Long considered an exotic niche topic, the field of DNAnanotechnology has immensely grown over the past decade,b oth on the experimental and computational side. [3] As an indication of this,D NA nanostructures (DN) are currently applied in research areas as diverse as nanoelectronics,chemical sensing, molecular computing,a nd biomedicine. [4] Thel atter field in particular has recently made impressive progress toward the utilization of DN in various therapeutic and diagnostic applications.
Compared to other more conventional nanomaterials,DN have some significant advantages when it comes to biomedical applications.F irst, whereas many nanoparticle systems have raised concerns regarding possible adverse effects, [5] DN are essentially biocompatible,b iodegradable,a nd non-cytotoxic. Second, DN and especially DNAorigami (DO) [6] can be assembled in well-defined yet almost arbitrary sizes and shapes and thereby provide ameans to tuning their biological availability and activity.Third, their surfaces can be modified in ap recisely controlled manner with molecular accuracy. This is particularly true for DO,w hich are based on the folding of along, single-stranded DNAscaffold into adesired nanoscale shape by hybridization with as et of short staple strands.Each of these staples has aunique sequence and can thus be unambiguously addressed and modified to carry different entities such as dye molecules,p roteins,n anoparticles,and drugs.Inthis way,dozens of different functional species can be arranged with nanometer accuracyo nt he interior and exterior surfaces of DO.This approach thus holds great promise for various biomedical applications,a si td oes not only enable the defined loading of the DN with various therapeutic cargos,b ut may also be used to facilitate cell targeting,c ellular uptake, target binding,a nd their conformational switching in response to various external stimuli. In general, DN may bridge biochemically relevant length scales and sub-nanometer precision to macroscopic dimensions.
However,D Na re intrinsically less stable than inorganic nanomaterials,which may result in serious limitations regarding their applicability in physiological environments that are equipped with sophisticated machinery for identifying and degrading foreign DNA. Stabilizing DN under such adverse conditions without impairing their desired biomedical function thus represents the most prominent challenge that we currently face on the road to real-world applications.I nt his Review,w es hall first introduce the various therapeutic and diagnostic applications of DN before summarizing the challenges imposed on DN stability and functionality by the physiological environment. Finally,w ew ill discuss strategies and solutions to these challenges and, in particular, address DNA nanotechnology holds substantial promise for future biomedical engineering and the development of novel therapies and diagnostic assays.The subnanometer-level addressability of DNAnanostructures allows for their precise and tailored modification with numerous chemical and biological entities,w hichmakes them fit to serve as accurate diagnostic tools and multifunctional carriers for targeted drug delivery.The absolute control over shape,size,and function enables the fabrication of tailored and dynamic devices,s uch as DNAn anorobots that can executeprogrammed tasks and react to various external stimuli. Even though several studies have demonstrated the successful operation of various biomedical DNAn anostructures both in vitro and in vivo,m ajor obstacles remain on the path to real-world applications of DNA-based nanomedicine.Here,wesummarize the current status of the field and the main implementations of biomedical DNA nanostructures.Inparticular,wef ocus on open challenges and untackled issues and discuss possible solutions. shortcomings,u nsolved issues,a nd potential conflicts with regard to DN functionality.

Biomedical Applications of DNA Nanostructures
Theb iomedical applications of DN are just as numerous as their shapes.With some exceptions such as the field of drug discovery,w hich is seeing more and more DNA-nanotechnology-related works, [7] most of these applications require the exposure of the DN to biological media, either in vivo or ex vivo.T he latter automatically provides ad istinction between the two major application areas:t herapeutic applications typically aim at employing the DN inside the human body,w hereas diagnostic applications often (although not always,see Section 2.2) only require exposure to (sometimes diluted or purified) blood, serum, or tissue samples.I nt his section, we will provide an overview of these two major application areas of DN in the biomedical field. Forf urther in-depth discussions of the specific applications,the reader is referred to the large number of recent Reviews that focus specifically on these topics. [8,9] 2.1. Therapeutic Applications Doxorubicin (DOX) is widely and commercially used in cancer therapy-especially for solid tumors-and it is awellknown DNAi ntercalator by its nature.T herefore,a tl east in principle,i ts hould be one of the most promising candidates for DN-based delivery,a si tc an be loaded into customized nanostructures that may have ap lethora of other functions. There are multiple examples of DN that have been employed as DOX carriers such as DNAt etrahedra (DT), [10,11] twisted 3D DO, [12] DO triangles, [13][14][15] rectangles, [11] and helix bundles, [13] as well as tubular DO loaded into liposomes. [16] In addition to DOX, its close molecular relative daunorubicin has also been used as ad rug for a( Tr ojan) "DNAh orse". [17] Although the efficacyofDOX-loaded DNAnanocarriers was verified in several in-vivo models, [10,11,13,14,16] each approach has its own and different loading and purification strategy, environment, pH, as well as DOX and ion concentrations, thus making the results extremely hard to compare to each other. Not only are the spectroscopic [18] properties of DOX strongly ion-and pH-dependent, [19] but DOX is also commonly employed in substantial excess to DN in the loading process,a lthough it is known to self-aggregate at high concentrations. [20] Moreover,D OX can also bind to partially hybridized or self-hybridized staples that are used in excess to DO-scaffold strand during folding.A sthe binding affinity of DOX is only slightly DNA-sequence-dependent, [21] the effects of staples should not be ignored. That being the case,studies relying purely on the spectroscopic properties of DOXa nd not taking into account all the above-mentioned factors may produce ambiguous results and leave plenty of room for speculation.
Besides broadly employed DOX, there are other potential drugs that can be loaded into DN.A gain, the nature of the interaction between the chosen drug and DN depends on the prevalent conditions,b ut the loading efficiencym ay also depend on the DN superstructure.T his has been demonstrated for intercalating YOYO-1 and acridine orange molecules [22] as well as for groove-binding methylene blue. [23] In contrast to the supposedly simple loading of the DNA nanocarriers with intercalators or groove-binders,D Nc an also be employed for the spatially controlled presentation of functional molecules.Mçser et al. recently conjugated ephrinmimicking peptides that bind to EphrinA2 receptors to the tips of aD NA three-arm junction. [24] Ephrin-signaling pathways are involved in tumor development and may thus be utilized in cancer therapy.T he authors observed that the oligovalent presentation of three ephrin-mimicking peptides on one DN resulted in significantly increased EphA2 phosphorylation in PC-3 cells compared to monomeric peptides.DN-templated oligovalence thus represents apromising concept for various therapeutic applications.F urthermore,i tw as found that even monomeric peptides showed higher potencyw hen coupled to the DN,w hich may provide arather simple route to tuning drug stability,distribution, and activity. [24] In cells,c ompartmentalizationa nd precise organization of active compoundss ucha se nzymes is vitalf or specificity, control, andenhancement of reactions. Thereare many routes to achievea rtificialc ompartmentsf or enzymes, butt he unprecedenteda ddressabilityo fD Nm akes them highly attractive candidates fort hisp urpose.P rotein encapsulation is beneficial noto nlyf or multipurpose delivery applications rangingf romi nfectioust og enetic-disease treatments, [25]  also form odulatingp rotein properties such as stabilitya nd function. [26] It hasb eens hown that thec ellulard eliveryo f luciferase-loadedD Oc an be achieved andt hatt he enzymes retain theiractivityinthe process. [27] Moreover,the activityof theseenzymes canbefurther modulatedwithcationicpolymer coatings of theh ollowD Oc ontainer (see also Section4.2). [28] Therei sawide varietyo fD NA-based vesselsf or enzymatic cargos such as cascaden anoreactors, [29] tubularh osts, [30] reconfigurable vaults andc apsules ( Figure 1a,top panel), [31,32] andvarious DNAcages (Figure1a, bottom panel) [33,34] as well as nanosheets forn ucleased elivery. [35] Theset ypes of (multi-)enzymes ystems andv ehiclesw ithe nzymatic payloadsh ave been reviewed in Refs. [26,36].Itisnoteworthy that harnessing DNAtemplates forprotein assembly mayhaveratherintriguinga nd unconventional implementationsi nt ailoredp rotein design,a sd emonstratedb yR osiere tal. [37] They assembled af unctionala poptosomeb yc o-localization of multiple caspase-9 monomers with theh elpo faDO platform. This approach may pave thew ay fort he engineeringo fa rtificial enzymest hata re involved in processess ucha si nflammation, innate immunity,and necrosis.
In addition to drug and enzyme delivery,D Na re investigated as delivery vehicles for therapeutic nucleic acids such as small interfering RNA( siRNA), antisense RNAs (asRNA), and genes. [38][39][40] In as eminal study by Lee et al.,the authors employed DT to deliver siRNAsequences into cells. [39] They observed the suppressed expression of the targeted genes only when the amount of cancer-targeting ligands (folates) and their orientation was appropriate,t hus underlining the importance of the spatial addressability of the DN.M oreover,t hese particles exhibited longer blood circulation half-lives than the parental siRNA. Similar DNA shapes have also been used in an anogel-based siRNAdelivery system. [40] Very recently,Lietal. showed that aptamer-equipped DN could be used in the repair of cerebral ischemia-reperfusion injury (IRI) in rats. [42] Oxidative stress combined with inflammation is the main contributor to brain IRI, which is intensified by the complement component 5a (C5a). Therefore,t he authors utilized DNAf rameworks conjugated with anti-C5a aptamers to selectively reduce C5a-mediated neurotoxicity and effectively relieve oxidative stress in the brain.
Immunostimulatory CpG oligonucleotides are as pecial class of therapeutic nucleic acids.T hese CpG sequences stimulate intracellular Toll-like receptors in macrophages and dendritic cells,w hich results in T-cell activation. [9] They can thus be used in cancer immunotherapy and as vaccine adjuvants.N umerous studies used DN to display CpG oligonucleotides and deliver them into macrophage-like cells without the need for transfecting agents. [43,44] Enhanced immunostimulation due to DN-mediated CpG delivery was also validated in vivo. [45] In contrast to chemo-, immuno-, or gene therapy,p hotodynamic [46] and photothermal therapy [47] employ inert compounds and materials that become active only upon interaction with light. These can be either photosensitizers that generate reactive oxygen species upon illumination or plasmonic nanomaterials such as gold nanoparticles that heat up due to the resonant absorption of photons of acertain wavelength. Many of these photosensitizer and nanoparticle systems,h owever, suffer from low solubility and low cell/ tissue uptake.A ts everal instances,t he application of DNbased carriers was shown to overcome these drawbacks and result in improved anticancer activity both in vitro and in vivo. [48][49][50]  On top of the above-mentioned examples,D Na ddressability,activity,and emerging multifunctionality may find uses in targeted therapy,w here the designed vehicle can perform multiple tasks.One of the earliest works manifesting this kind of utility was ad ynamic logic-gated nanorobot designed by Douglas et al. (Figure 1b,t op left panel). [41] Thea uthors equipped ah ollow-shell-like and spring-loaded DO with antibodies and closed the device using an aptamer "lock" system (dsDNA) that can only be opened through binding of as pecific antigen "key". They demonstrated the logic gating with at wo-input system by assembling al ogic AND gate through aptamer encoding,t hat is,ag ate where both locks needed to be opened simultaneously to activate the robot (Figure 1b,b ottom panel). Furthermore,t hey presented the versatility of the system by building ah andful of distinct versions of these aptamer-encoded devices and testing the response with multiple different cell lines.L ater on, similar nanorobots were employed in performing universal computing in living cockroaches. [51] This approach by Amir et al. was based on dynamic interactions between the robots.T hese interactions served as logical outputs that were further relayed to switch molecular payloads on or off,thus opening new avenues in the computational control of therapeutics.
Along these lines,L ie tal. created their own version of ad ynamic in-vivo nanorobot (Figure 1b,t op right panel). [52] They used at hrombin-loaded rectangular DO that was further wrapped into at ubular shape with DNA" fastener" strands and functionalized with targeting aptamers.T he fastener strands were designed in such aw ay they could open through the interaction with nucleolin proteins that are expressed at the surface of actively proliferating tumor vascular endothelial cells.W hen the encapsulated cargo was displayed and exposed through the nucleolin-induced reconfiguration of the robot, thrombin activated blood coagulation at the tumor site.T he authors used mice to demonstrate the specific delivery of robots to tumor-associated blood vessels and the resulting intravascular thrombosis.F inally,this led to tumor necrosis and inhibition of tumor growth.
Related to above examples,Liu et al. have also shown that drug-molecule-loaded delivery devices with multifunctional properties can be assembled using the DO technique. [14] The authors integrated gene delivery with cancer therapy by loading aDOtriangle with DOX molecules and equipping it with two linear tumor-therapeutic genes (p53). In as imilar manner,t he same group used ac ombination of RNA interference (RNAi)a nd chemotherapy by incorporating siRNAa nd DOX into as ingle DO. [53] These multifunctional devices could enter multidrug-resistant tumors (MCF-7R) and inhibit their growth both in vitro and in vivo.F urthermore,t he research group of Ding has also shown that DOX, gold nanorods,and the tumor-specific aptamer MUC-1 can all be incorporated into one DO vehicle for effective circumvention of drug resistance. [54] Jiang et al. recently demonstrated that even non-modified DO can have therapeutic potential ( Figure 2). [55] They observed that the DO preferentially ended up in the kidneys (Figure 2b), whereas partially folded nanostructures and unfolded scaffold strands were sequestered by the liver or experienced rapid renal clearance.M ost astonishingly,t he authors also found that at least one of the DO shapes exhibited renal-protective properties.Rectangular DO could efficiently alleviate acute kidney injury (AKI) in mice via the scavenging of reactive oxygen species (ROS,F igure 2a). The therapeutic response was rapid, which indicates that DO may be promising candidates for the treatment of various kidneyrelated diseases.Importantly,the authors also showed that the selected rectangular DO shape was not toxic to organ functions and it did not elicit an immune response in vivo. However,i tshould be pointed out that DO may still require additional protection mechanisms for efficient targeting and enhanced pharmacokinetic bioavailability,w hich might change their biodistribution from what has been shown here.N evertheless,t he possibility of using DN themselves as drugs by exploiting their ROS-scavenging abilities is very appealing and deserves further in-depth investigation.

Diagnostic Applications
Being fully composed of DNA, it is rather straightforward to decorate DN with aw ell-defined arrangement of capture probes for the specific binding of preselected, medically relevant nucleic-acid sequences such as cancer-related micro-RNAs (miRNA) [56] or disease-or pathogen-specific genes. Target binding can be detected using various techniques,with the DN often being used for signal enhancement, transduction, or the implementation of logic operations.One of the earlier demonstrations by Ke et al. used barcoded DO substrates to arrange three different capture sequences complementary to regions of three different genes and detected the site-specific binding of target RNAs using atomic force microscopy (AFM, Figure 3a,r ight panel). [57] Wang et al. further advanced this concept by implementing strand-displacement-based logic operations for the simultaneous detection of two different input miRNAs. [58] Kuzuya  [59] Target binding resulted in achange in DO shape that could be detected either by AFM or by fluorescence readout using dye-modified DO constructs. [60] As of now,f luorescence-based signal readout is the most widely used technique for the detection of target DNA/RNA binding to DN.T his is mostly because the use of DN substrates for fluorophore presentation provides several advantages over free nucleic-acid probes.F or instance,D N enable the construction of multicolor-fluorescence systems that can enter living cells for the detection of intracellular RNAs. [61] Furthermore,D Np rovide several means for enhancing fluorescence intensities.D ecoration of beaconlike DO with multi-fluorophore arrays resulted in enhanced FRET signals for target DNAd etection down to 100 pm concentrations. [62] Zhu et al. recently demonstrated the pHcontrolled intracellular release of hairpin probes from triplehelix-functionalized DT,which initiated ahybridization chain reaction upon interaction with target mRNAfor the amplification of fluorescence signals. [63] By modifying DO with gold or silver nanoparticles,p lasmonic nanoantennae can be constructed for the plasmonic enhancement of the fluorescence signal obtained from single fluorophores by several orders of magnitude. [64] Ochmann et al. combined such nanoantennae with fluorescence-quenching hairpins to detect Zika-virus DNAa nd RNAs equences. [65] This approach not only enabled target detection at 1nm concentrations in human serum but was also extended toward multiplexing by combining multiple antenna designs.
In addition to AFM and fluorimetry,v arious electrochemistry-based sensing concepts for the detection of nucleicacid binding to electrode-immobilized DN have also been evaluated (Figure 3b). [66][67][68] Here,i mmobilizing target-binding probe sequences on electrode-supported DN typically resulted in ah igher detection sensitivity and sequence specificity compared to direct immobilization at the electrode surfaces.
Employing dynamic DN allows for the implementation of novel sensing concepts.F unck et al. recently demonstrated the detection of ah epatitis Cv irus RNAs equence using ac ross-shaped DO comprised of two gold nanorod-carrying arms connected with af lexible pivot point. [69] Target RNA binding triggered astrand-displacement reaction that resulted in the switching of the DO device from am ostly achiral to ar ight-handed chiral geometry,w hich could be detected by circular-dichroism spectroscopy.T he sensitivity of the device was determined as 100 pm and successful detection of 1nm target RNAwas accomplished in 10 %serum.
In ar ather different approach, DO have been used as shape IDs not for DNAdetection but for genotyping.T othis end, Zhang et al. developed as et of differently shaped and modified DO that were used for the site-specific labeling of genomic DNAextracted from human blood samples. [72] Using AFM for shape ID visualization, the authors could detect and distinguish various single nucleotide polymorphisms (SNPs) with alateral solution of about 10 nm. This approach was later extended to identify the genotype of hepatitis Bv iruses. [73] Thegeneral detection concepts introduced above can also be adapted for the detection of proteins and other biomarkers.T his usually requires the introduction of target-specific aptamers into the DN. [74] Detection of the bound analyte can then be achieved using AFM, [70,75] electrochemistry, [76] fluorimetry, [60] and CD spectroscopy, [77,78] and successful target detection was recently demonstrated even in whole blood. [76] Nevertheless,most of these works can be considered proof-ofprinciple studies that employed well-characterized aptamers with ah igh affinity toward some model targets such as thrombin, [75,78] ATP, [60,76] or adenosine. [77] Successful aptamerbased detection of an infection-related target was demonstrated by Godonoga et al.,w ho incorporated aptamers against the malaria-protein biomarker PfLDH in 2D DO substrates (Figure 3c). [70] Using AFM, target-protein binding could be detected at aconcentration as low as 500 nm and also in the presence of human plasma. Kwon et al. decorated starshaped DN with ar egular pattern of aptamers that matched the expression pattern of their target proteins in the denguevirus envelope,resulting in strong oligovalent virus binding. [79] By introduction of fluorophore-quencher pairs,D N-virus binding could be detected in human serum and plasma at as uperior sensitivity compared to PCR-based detection. Furthermore,a rranging the aptamers on the DN also enhanced their inhibition efficacy.
As an alternative to aptamers,D Nc an also be directly modified to carry antibodies or antigens.For instance,Kuzuya et al. used fluorescein-modified DO pliers for anti-fluorescein IgG detection. [59] Pei et al.,o nt he contrary,d eveloped as andwich-type assay in which an antibody against TNFa was attached to DT immobilized on ag old electrode. [80] After capturing TNF-a from solution, as econd antibody carrying aHRP enzyme was bound to the captured TNF-a to translate target binding into ad etectable electrochemical signal. This sandwich-type assay was subsequently further advanced toward the detection of antibodies [81] and pathogens. [82,83] Fori nstance,W ang et al. recently reported the highly sensitive detection of pneumococcal surface protein A from Streptococcus pneumoniae lysate. [83] This assay not only had an extremely low limit of detection of 0.093 CFU mL À1 equivalent of S. pneumoniae lysate,b ut was also able to quantify S. pneumoniae in different swab samples from ahuman subject.
DN can also be employed as carriers of functional species for in-vivo-imaging applications. [84] Many studies have employed fluorophore-and quantum-dot-labeled DN to visualize their biodistribution by fluorescence microscopy. [13,39,85,86] While most of these works focused on the efficacy of the DN as vehicles for the transport of therapeutic cargo, this approach may also find its way into purely diagnostic applications.F or instance,K im et al. used af luorophorelabeled DT for sentinel-lymph-node imaging. [87] Here,the use of aD Nr esulted in enhanced lymph-node translocation and aprolonged retention time at the node.
In addition to fluorescence imaging,D Nh ave recently also been utilized in other in-vivo-imaging techniques.J iang et al. employed ad ual-modified DT carrying an ear-infrared (NIR) emitter and ar adioactive Tc isotope for combined NIR-fluorescence imaging and single-photon-emission-computed tomography of targeted tumors in mice (Figure 3d, right panel). [71] Similarly,t he in-vivo biodistribution of 64 Culabeled DO was evaluated using positron-emission tomography. [55] Finally,g old-nanorod-modified DO have been employed for two-photon luminescence (Figure 3d,l eft panel) [49] and optoacoustic imaging. [50] 3. Challenges for Applying DNA Nanostructures in the Physiological Environment Thea pplication of DN in physiological environments faces two major challenges:l imited stability in biological media and the induction of an adverse immune response.On top of that, the structures usually suffer from poor cell uptake. Naturally,t hese challenges are more severe for in-vivo applications than for ex-vivo diagnostics.H owever,e ven though many ex-vivo assays may use biological media that have been purified, inactivated, supplemented with stabilizing salts,orsimply diluted, the ultimate goal is to directly analyze patient samples with as little pre-processing as possible,i n particular in the field of point-of-care diagnostics. [88] 3.1. Limited Stability Early on, the stability of DN in biological media has attracted considerable attention and initial results appeared very promising.Keum and Bermudez have shown that DT are more stable in the presence of diluted serum than their linear counterparts. [89] Similarly,M ei et al. demonstrated the stability of various 2D and 3D DO in cell lysates. [90] Nevertheless, Castro et al. observed the complete degradation of 3D DO in the presence of nucleases in less than one hour. [91] Finally, Hahn et al. have identified two major factors that limit DO stability in cell-culture media, namely low Mg 2+ concentrations and the presence of nucleases. [92] Thev ast majority of protocols for DN assembly employ Mg 2+ concentrations in the mm range in order to compensate the electrostatic repulsion between neighboring DNAh elices. [93] Relevant biological media such as blood or serum, however, have much lower Mg 2+ concentrations. [92] Even though physiologically more abundant monovalent cations such as Na + or K + are,ingeneral, also able to stabilize the DN, they are less efficient and may again require concentrations that exceed those of physiological environments. [94,95] This is due to ion-specific differences in the type of interaction, binding site,a nd binding affinity between DNAa nd the individual ion species. [94,96] Recently,R oodhuizen et al. performed atomistic molecular-dynamics simulations of DO stabilization by various cations. [96] They found that Mg 2+ ions bind strongly to minor-groove atoms and the backbone phosphates,whereas Na + binding is much weaker and barely involves the phosphates.T his strong binding of the Mg 2+ ions is responsible for the somewhat surprising observation that DO assembled in aM g 2+ -containing buffer can be gently transferred into Mg 2+ -free buffers and even pure water without al oss of integrity. [94,97] Under such conditions, however, other buffer components such as EDTAo rp hosphate ions also become important as they may interfere with the DNA-bound Mg 2+ ions and thus promote DO denaturation. [94] Furthermore,D Os hape and superstructure also appear to play as ignificant role in DO stability and denaturation under low-ionic-strength conditions. [92,94] While these ionic-strength-related destabilizing effects are typically more pronounced in large DO that feature ad ense arrangement of many negatively charged double helices than in smaller DN, [98] DT [99] and single-strandedtile (SST)-based DNAn anotubes [100] may also turn out sensitive toward low mono-and divalent salt concentrations. In particular,K ocabey et al. observed that SST-based DNA nanotubes may undergo complete denaturation in phosphatebuffered saline with ap hysiological Na + concentration of 135 mm when supplemented with less than 2mm Mg 2+ . [101] Even more intriguing, they found that low-salt denaturation is more pronounced when siRNAs equences are hybridized to single-stranded overhangs on the nanotube surface.
DNA-degrading nucleases are found in virtually all types of tissues and bodily fluids, [102,103] and thus represent aserious threat to the integrity of therapeutic or diagnostic DN. Consequently,an umber of studies have evaluated DN stability in the presence of selected nucleases or serumcontaining media. Unfortunately,the obtained results remain somewhat ambiguous so far.W hile Keum and Bermudez observed that DT are more stable in 10 %fetal bovine serum (FBS) and in the presence of DNase It han simple doublestranded DNA, [89] Goltry et al. have shown that the situation is more complex. [104] In particular,they observed not only that the stability of as imple DNAn anomachine in 70 %h uman serum depends on its local topology,b ut also that it was in general less stable than corresponding duplex DNA. Furthermore,t he introduction of fluorophore and quencher labels resulted in areduced stability of both the nanomachine and the duplex DNA. This observation is in line with the results by Kocabey et al.,w ho found that siRNA-decorated DNAn anotubes were degraded in 10 %F BS within 1h, whereas unmodified nanotubes survived for at least 8hunder identical conditions. [101] As was recently shown by Lacroix et al.,t his rapid enzymatic degradation of fluorescently labeled DN in cell-culture medium may cause severe artifacts in in-vitro cellular-uptake studies. [105] Forlarger DO,exposure to 10 %FBS resulted in notable degradation already after two hours. [92] This fast degradation is caused by the very high nuclease activity of FBS.I nh uman serum, about sixfold longer lifetimes have been observed. [104] Using FBS,the lower nuclease activity of human serum can be mimicked by heat inactivation, [104] which resulted in completely intact DO after 24 hincubation. [92] Castro et al. investigated the susceptibility of 3D DOhelix bundles toward degradation by various nucleases and observed significant digestion only for T7 endonuclease Iand, most importantly,D Nase I, [91] which is the most abundant nuclease in blood and serum. [102] Nevertheless,D Nase-Iinduced degradation was much slower for the DO than for duplex plasmid DNA. [91] Absolute time scales of DNase-Iinduced DO degradation, however, strongly depend on the experimental conditions.F or instance,C astro et al. observed the complete digestion of aDO24-helix bundle (2 ng in 20 ml) by one unit DNase Iwithin 1hat 37 8 8C. [91] Auvinen et al.,on the contrary,o bserved no degradation at all for a6 0-helix bundle (310 ng in 20 ml) exposed to one unit DNase Iw ithin 1hat room temperature. [106] These differences may result not only from the different temperatures and DNAc oncentrations employed in these experiments,b ut also from the different DO shapes.I ndeed, as was recently shown by our labs,D Os hape and superstructure play an important role in modulating global as well as local DNase-I activity (see Section 4.1). [107] 3.2. Adverse Immune Response Albeit DNAm olecules are inherently biocompatible polymers and act as key players in many biological and cellular processes,t hey may nonetheless elicit severe inflammatory responses. [108] In the case of DN,Perrault and Shih [85] as well as Auvinen et al. [106] observed remarkable immune activation in mice splenocytes incubated with plain 3D DO just by monitoring their cytokine production, namely interleukin 6( IL-6) and/or interleukin 12 (IL-12) levels.N evertheless,w hen these primary spleen cells were treated with lipid-or protein-coated DO (see also Section 4.2), both groups demonstrated ap roper camouflage of DNA, in other words,anegligible immune response.Schüller et al. (see also Section 2.1) also probed the immune response of mice splenocytes,b ut in this case,t hey used tubular DO with and without immunostimulatory CpG oligonucleotides and followed the secretion of three different immune markers,IL-6, IL-12, and transmembrane C-type lectin (CD69). [44] Cell incubation with CpG-modified tubes resulted in elevated levels of all markers,while unmodified DO induced only IL-6 and IL-12 responses.I nc ontrast, as tudy by Xia et al. employing dendritic RAW264.7 cells (mouse-macrophageprecursor immune cells) and small DT did not show any IL-6 or IL-12 activation. [109] Nevertheless,the direct comparison of these results is rather challenging,a sthe amount and size of DN vary from study to study. [110]

Design Factors
DN can be designed in avariety of shapes,and for along time,i tw as speculated that their shape and size should have as ignificant influence on the efficacyo ft heir cellular transport, similar to nanocarriers made of other materials. Bastings et al. thus systematically studied the living-cell uptake of DO having asize range of 2to5MDa and multiple structures such as different-diameter bundles,circles,barrels, and other nanoshapes (Figure 4a,l eft panel). [111] They observed that large and compact structures are preferentially internalized (Figure 4a,right panel), indicating that both the aspect ratio and mass play an important role in the process. Nevertheless,t hey reported that the transfection rates are actually even more dependent on the used cell type than the DO themselves.
In any case,t hese structures will eventually face the adaptive immune system (see above) and endonucleases, which may digest and eventually destroy them. Keum and Bermudez have reported that simple DT with different shapes and sizes may be resistant to specifically and non-specifically digesting nucleases. [89] Kim et al. used aDNA cage library of 16 different shapes for the in-vivo screening of tumor targeting in mice and found that the tumor specificity was closely related not only to the cellular uptake of the cages but also to their nuclease resistance. [112] By selecting the most potent cages,the group demonstrated tumor-specific damage by delivering macromolecular apoptotic proteins solely into the tumor tissue.
Ramakrishnan et al. studied how DNase Iw ould enzymatically cleave multiple distinctly designed DO by monitoring the structures in real time on mica substrates using highspeed AFM. [107] They observed that these 2D shapes exhibit structure-specific degradation profiles,t hus confirming that digestion is structure-dependent (Figure 4b). In particular, DO degradation by DNase Is eems to be most efficient at mechanically flexible sites that can accommodate the structural duplex distortions associated with DNase-I binding.
Similar observations were also made by Stopar et al. who investigated the enzymatic cleavage of the DO scaffold by several restriction endonucleases at their respective recognition sites. [114] They found that these sites can be either active or strongly resistant toward enzymatic cleavage,d epending on the local mechanical and topological properties in the vicinity of the individual site. [114,115] Theb ulk of these data suggests that DN can be rationally designed for increased serum stability and cell uptake,f or instance,b ym aking them mechanically very rigid to suppress nuclease binding. However,t he desired performance of ag iven DN depends not only on its stability and uptake but also on other aspects such as drug loading efficiency and release kinetics,w hich are affected by the same design factors.I ndividual design-related properties will thus have to be weighed against each other, favoring,f or instance,e fficient drug intercalation over serum stability.
As an interesting addition to the different designs discussed above,K im et al. employed ab io-orthogonal base-pairing system, that is, l-DNAi nstead of natural d-DNA, for their DT vehicle to avoid undesirable interaction between the carrier and the attached proliferative aptamer cargo (Figure 4c). [113] This modification also led to strongly enhanced serum stability and increased intracellular delivery rates.S imilarly, Liu et al. introduced unnatural base pairs into DNAjunctions and nanotubes,r esulting in increased melting temperatures and exonuclease resistance. [116] It remains to be seen, however,h ow such oligonucleotide modifications affect the loading with intercalating or groove-binding drugs.

External Modifications
Themost intuitive way to tackle possible immunogenicity, poor transfection rates,and low stability is to employ coating or self-healing strategies.Perrault and Shih demonstrated that the DNAstrand-mediated lipid-bilayer coating of aspherical DO not only attenuated the immune response (see Section 3.2), but also improved the stability against nucleases and increased the in-vivo pharmacokinetic bioavailability (Figure 5a,t op left panel). [85] Thes tability and resistance to nucleases can also be increased by decorating the exterior of DN with dendritic oligonucleotides as shown by Kim and Yin (Figure 5a,t op right panel). [117] Lacroix et al. used ah umanserum-albumin (HSA) coating based on the attachment of  [111].C opyright2 018 American Chemical Society.b )Reproducedw ith permission from ref. [107].C opyright2 019 John Wiley and Sons. c) Reproduced with permission from ref. [113].

Angewandte Chemie
Reviews dendritic alkyl-conjugated DNAs trands for achieving serum stability and protection against nucleases (Figure 5a,bottom right panel). [120] It is noteworthy that the employed HSA coating did not hinder the activity of structure-bound genesilencing oligonucleotides inside cells.Inaddition to the study by Lacroix et al.,ithas also been shown that discrete protein modifications can enhance the cellular delivery of DN. Schaffert et al. incorporated multiple DNA-modified transferrin proteins into aplanar DO and managed to increase the transport rates into cancer cells up to 22-fold compared to naked DO. [118] Yeta nother direct DNA-linking-based approach to increase the bioavailability of DN is self-healing (Figure 5a,b ottom left panel). Li and Schulman demon-strated that the degradation process of long poly(ethylene glycol) (PEG)-coated DNAnanotubes in 10 %FBS could be reversed by an excess amount of small PEG-conjugated DNA tiles to seal the broken parts and repair defects. [119] With this strategy,t he serum lifetime of DNAn anotubes could be extended to several days.
All the coating approaches mentioned above rely on modified DNAstrands that can be anchored via hybridization to the designed positions in the larger structure,w here they may equally serve as binding sites for further functionalization. While the addressability is arguably an advantage,these methods may require expensive modifications of dozens of individual strands.T herefore,i nstead of DNAc onjugation, simple electrostatic coating and shielding of the negatively charged DN may be more feasible in many occasions.There is ap lethora of techniques that take advantage of cationic polymers,s uch as poly(2-dimethylaminoethyl methacrylate) (PDMAEMA)-PEG copolymers (Figure 5b,t op right panel), [28] chitosan and linear polyethyleneimine (PEI), [124] oligolysine-PEG (Figure 5b,b ottom right panel), [86] and polylysine-PEG copolymers (Figure 5b,l eft panel). [121] Recently,i tw as demonstrated that the nuclease resistance of DO coated with PEGylated oligolysine can be further enhanced by ac hemical crosslinking of the polymer coating. [125] Besides synthetic polymers,itis equally possible to apply protein-based coatings using electrostatic interactions.M ikkilä et al. disassembled chlorotic cowpea mottle virus (CCMV) to obtain single capsid proteins (CPs) with positively charged residues at the N-terminus. [122] These CPs were then complexed with rectangular DO,r esulting in wrappedup morphologies (Figure 5c,l eft panels) or fully CP-encapsulated DO.T he HEK 293 cell-transfection efficiencyofCP-  [117].Bottom left panel reproduced with permission from ref. [119].C opyright 2019 American Chemical Society.Bottom right panel reproduced with permission from ref. [120].C opyright 2017 American Chemical Society.b )Left panel reproduced with permission from ref. [121].C opyright2 017 John Wiley and Sons. Topright panel reproduced with permission from ref. [28].B ottom right panel reproduced with permission from ref. [86].c)Left panel reproduced with permission from ref. [122].C opyright 2014 AmericanC hemical Society. Topright panel reproduced with permission from ref. [123].C opyright 2017 AmericanC hemical Society.B ottom right panel reproduced with permission from ref. [106]. coated DO could be increased up to 13-fold compared to bare DO.K opatz et al. used as imilar strategy as above,b ut their components were CPs of simian virus 40 (SV40) from the polyomavirus family and an early spherical 3D DO with aw ell-defined diameter. [126] In this case,C Ps were fully encapsulating the DO,and the complex resembled the native SV40 in shape,symmetry,and size.Inaddition to viral capsids, serum albumin has also been used for the protection of DO. Auvinen et al. demonstrated the bovine serum albumin (BSA) coating of ab rick-like DO by covalently coupling positively charged and branched dendron molecules to BSA (Figure 5c,bottom right panel). [106] It was not only shown that the BSA corona attenuated the immune response as explained earlier, but also that the cell-transfection rates and stability against DNase Iw ere clearly improved. Recently,c ationic HSA (cHSA), an albumin derivative,w as utilized in as imilar way to complex rectangular DO. [127] Yet another way of making use of protein-based shielding is to attach protein polymers [128] or diblock polypeptides [123] to DN to protect them from enzymatic degradation (Figure 5c,t op right panel).
Several issues need to be considered when employing such coatings in biomedical DN.W hile it has been demonstrated that oligolysine-PEG coating of DO did not compromise the functionality of single-stranded DNAh andles protruding from the DO surface, [86] this may not be the case for all coatings and functional surface modifications reported in literature.Inparticular,the binding affinity of surface-bound aptamers may be drastically reduced by the application of such coatings.F urthermore,m any of the coatings discussed above are not compatible with the conformational switching of dynamic DN.F inally,c oating of DN will restrict access to any encapsulated cargo [28] and thus (positively or negatively) affect drug-loading and release properties.

Enzymatic and Chemical Modifications
Ageneral problem faced by virtually all DN results from the fact that they consist of short oligonucleotides that are hybridized with each other via domains comprising only asmall number of base pairs.T herefore,the melting temperatures of these oligonucleotides are often rather low.F urthermore,e ven am oderate nuclease attack resulting in only af ew cleaved oligonucleotides may already lead to the spontaneous dehybridization of the even shorter fragments and thus the complete collapse of the DN.C onsequently, various studies have attempted to increase DN stability by introducing covalent bonds between neighboring oligonucleotides.
Ar ather obvious strategy for the covalent linking of (selected) oligonucleotides is their phosphorylation and postassembly enzymatic ligation. ONeill et al.,f or instance, demonstrated that ligation of tile-based DNAn anotubes not only increases their melting temperature but also renders them stable in pure water. [129] Even though the dense duplex arrangement in DO most likely makes asignificant fraction of nicks non-accessible for the ligase, [107,114] the enhanced stability of ligated DO under denaturing conditions was recently demonstrated (Figure 6a). [130] Hamblin et al. further demonstrated that decreasing the number of backbone nicks renders DNAn anotubes resistant against nuclease degrada- Bottom panel:A fter the UV-treatment the DO pointer structure retains its shape when incubated for 48 hinlow ionic strength phosphatebuffered saline (PBS) at 40 8 8C. a) Reproducedwith permission from ref. [130].b)Reproducedwith permission from ref. [133].C opyright 2014 John Wiley and Sons. c) Reproducedw ith permission from ref. [135].d)Reproducedw ith permission from ref. [136].C opyright 2019 AmericanC hemical Society. tion in 10 %F BS. [131] In their case,t his was achieved not by enzymatic ligation but through the use of continuous backbone strands produced by rolling-circle amplification.
Similar covalent crosslinks can also be achieved via chemical modifications of the employed oligonucleotides. Casinelli et al. assembled SST-based six-helix tubes from (3'alkyne,5'-azide)-modified oligonucleotides. [132] Each of these oligonucleotides then was cyclized by covalently connecting its ends in ap ost-assembly click reaction, which resulted in the topological interlocking of the SSTs.T his interlocking resulted in as ignificantly enhanced stability of the DNA nanotubes under low-salt conditions and enhanced resistance against exonuclease digestion. Kalinowski et al. followed ad ifferent approach and chemically ligated the oligonucleotides in aD Ov ia phosphoramidate linkages between 3'amino-modified staples and their 5'-phosphorylated neighbors (Figure 6b). [133] Recently,R aniolo et al. evaluated the stability of various crosslinked and non-crosslinked DN both in 10 %F BS and inside cells. [134] They found that enzymatic ligation of DNAc ages as well as cyclization of SST-based DNAnanotubes resulted in asignificantly enhanced stability compared to both their non-crosslinked counterparts and non-ligated DO.
In addition to these purely chemical crosslinking strategies,v arious routes for the photo-induced crosslinking of DO have been reported. Rajendran et al. incubated preassembled DO with the drug 8-methoxypsoralen (8-MOP). [137] Subsequent exposure to UV light resulted in the formation of covalent adducts between 8-MOP and pyrimidine bases.T he UV-crosslinked DO exhibited drastically enhanced melting temperatures.G erling et al. employed am ore direct photo-crosslinking. [135] They modified the staple strands with single-stranded thymidines at predefined positions and employed UV irradiation to crosslink thymidines in close proximity via the formation of cyclobutane pyrimidine dimers (CPDs). In this way,t hey introduced various interhelical as well as some intrahelical crosslinks (Figure 6c). TheU V-crosslinked DO displayed superior stability under high-temperature and low-salt conditions as well as in 10 %FBS and in the presence of various nucleases. Engelhardt et al. further advanced this approach by utilizing aspecifically designed scaffold with AA motifs at all possible staple crossover positions,w hich therefore leads to adjacent thymidines in staple strands that can be readily UV-crosslinked without requiring any additional staple modifications (Figure 6d). [136] Finally,p hoto-crosslinking can also be reversible,a sr ecently demonstrated by Gerling and Dietz, who employed 3-cyanovinylcarbazole-modified staple strands in DO. [138] Upon irradiation with UV light at 365 nm, this modification can form acovalent bond with athymine base in its close vicinity.T his bond can be reversibly cleaved upon irradiation at a3 10 nm wavelength, which enables the transient stabilization of switchable DO devices. Covalent crosslinking strategies as described above have proven very successful in stabilizing DN under physiological and denaturing conditions.H owever,there are certain issues associated with the crosslinking concept in general that may somewhat limit the application of these strategies in biomedical settings.For instance,inmany cases,covalent crosslinking may lock dynamic DNAdevices in afixed conformation and thus inhibit any switching action in response to ad etected stimulus.Furthermore,crosslinking may alter the mechanical properties of individual duplexes and thereby affect drugloading and release properties.I nt his regard, the impact of covalent crosslinking on the performance of ag iven DN is rather difficult to predict and needs to be assessed individually in each application.
As an alternative to covalent crosslinking,n on-crosslinking-based strategies may also be able to enhance DN stability. Conway et al.,for instance,demonstrated an improved serum stability of prism-like DNAc ages assembled from oligonucleotides carrying terminal hexaethylene glycol and hexanediol modifications. [139] In particular,the introduction of hexaethylene glycol modifications increased the lifetime of the DNAc ages in 10 %F BS from about 2hto about 15 h. This strategy has the great advantage that interference with structural properties and cargo accessibility is kept at aminimum. Whether these modifications are also able to stabilize larger DO,however, remains to be seen.

Summary and Outlook
Folding DNAi nto customized shapes with desired functions is no more just ab lack-box system, since we continuously gain more knowledge of its inner workings.This progress has enabled better designs,faster production, higher fabrication yields,a nd in some cases,m ore stable structures. As ar esult, DN have found their way into numerous application fields.I np articular,t he field of biomedical DNAn anotechnology has seen immense advancements in the last decade,w ith DN being employed in numerous applications,r anging from pathogen detection to genotyping to drug delivery and targeted therapy.S everal successful tumor treatments in animal models have been demonstrated using DN delivery vehicles,a nd DN have even shown the potential to be used as active therapeutics themselves. Nevertheless,t he limited stability under physiological conditions and possible immunogenicity of DN have raised concerns regarding impaired functionality and insufficient biodistribution and circulation time on top of undesired side effects.
Consequently,m ore and more research efforts focus on elucidating and controlling the molecular mechanisms that govern DN stability and degradation under physiological conditions.Several routes toward controlling DN stability and immunogenicity have already been explored, including rational (re)design strategies,e nzymatic ligation, chemical and photo-crosslinking,p rotein and lipid encapsulation, and polyelectrolyte coatings.W hile many of those approaches have indeed resulted in significant improvements in DN stability,b iodistribution, and cellular uptake,t hey may also carry the risk of interfering with the therapeutic performance of the DN,p rimarily with the loading and release of therapeutic cargo and the binding of diagnostic biomarkers. Furthermore,many of the stabilization strategies investigated so far are incompatible with adynamic switching of the DN,as used in several stimuli-responsive DNAnanorobots.T ailoring DN stability,i mmunogenicity,d ynamic switching,c argo loading and release,and analyte binding both simultaneously and independently thus represents the most eminent challenge that biomedical DNAn anotechnology currently faces.
Another challenge that we did not discuss so far,e ven though it will become ah ighly pertinent question at some point in the future,r egards the clinical translation of therapeutic DN.I mportant subjects in this respect include long-term storage and shelf-life,scalability,CMC-and GMPcompliant production, and cost. An umber of studies have already addressed the issues of long-term storage of DN and their pre-assembled components by lyophilization and cryostorage,and further identified appropriate conditions to keep them intact for up to several years. [99,140] With regard to scalability and cost, previous in-vivo studies using intravenous administration of DOX-loaded DN typically applied doses in the range of about 100-600 mgD Np er kg animal. [10,13,14] For ahuman with ab ody weight of 75 kg, this would translate to aa bout 10-50 mg DN per dose.E mploying biotechnological mass production, DO can be produced at an estimated E0.18 per mg, [141] resulting in total DNAc osts of less than E10 per dose.This number,however, does not include additional costs arising from CMC-and GMP-compliant production, which may be rather challenging for biologics concerning issues of sterilization, purity,a nd batch-to-batch consistency, [142] and thus increase the production costs significantly.H owever, there may also be additional regulatory matters.While several nucleic-acid-based drugs have been approved by the FDAand are already in clinical use, [143] these comprise only synthetic oligonucleotides without any genomic material. Therefore, we expect that fully synthetic DN such as DT and SST-based DN will have to overcome fewer hurdles on their way toward FDAapproval than DO that are based on agenomic scaffold. Such DN also present the advantage that they can be assembled from other DNA-like materials that are not potentially genetically active. [113] While GMP-compliant oligonucleotides are significantly more expensive than standard ones,l arge-scale synthesis in the multi-kg range will reduce the price to af ew euros per mg, which is comparable to monoclonal antibodies.T his would raise above estimate to af ew E10 per dose,w hich is fairly moderate for ab iopharmaceutical. Forinstance,under the above considerations,one dose of Trastuzumab in the AC-TH regimen for the treatment of HER2 + breast cancer costs around $900, [144] while the antisense oligonucleotide drug Nusinersen for the treatment of spinal muscular atrophy is prized at $125,000 per injection. [143] Therefore,for therapeutic DN to enter clinical trials, significant investments will be required due to the comparatively large costs of small-scale GMP-compliant oligonucleotide synthesis.H owever,o nce aD N-based drug formulation receives regulatory approval and enters the market, we expect production costs as well as the final product prices to drop to reasonable levels,t hus rendering DN promising therapeutics for the treatment of numerous diseases.