DNA Droplets: Intelligent, Dynamic Fluid

Breathtaking advances in DNA nanotechnology have established DNA as a promising biomaterial for the fabrication of programmable higher‐order nano/microstructures. In the context of developing artificial cells and tissues, DNA droplets have emerged as a powerful platform for creating intelligent, dynamic cell‐like machinery. DNA droplets are a microscale membrane‐free coacervate of DNA formed through phase separation. This new type of DNA system couples dynamic fluid‐like property with long‐established DNA programmability. This hybrid nature offers an advantageous route to facile and robust control over the structures, functions, and behaviors of DNA droplets. This review begins by describing programmable DNA condensation, commenting on the physical properties and fabrication strategies of DNA hydrogels and droplets. By presenting an overview of the development pathways leading to DNA droplets, it is shown that DNA technology has evolved from static, rigid systems to soft, dynamic systems. Next, the basic characteristics of DNA droplets are described as intelligent, dynamic fluid by showcasing the latest examples highlighting their distinctive features related to sequence‐specific interactions and programmable mechanical properties. Finally, this review discusses the potential and challenges of numerical modeling able to connect a robust link between individual sequences and macroscopic mechanical properties of DNA droplets.


Introduction
Biological cells are highly intricate chemical reactors capable of various functions. Reproducing biological cells from scratch using a minimal set of biomolecules has attracted increasing attention as a rich source of inspiration for engineering and science. Creating cell-like systems may provide fundamental insights into the origin of life and its evolutionary pathways [1] by decoupling specific but fundamental aspects of complex biological phenomena. In addition, a newly gained insight into biological phenomena might revolutionize biotechnology engineering and sciences. Several projects are ongoing in the global research community. [2,3] Artificial cell studies have now expanded across interdisciplinary research lines, ranging from biology and chemistry to physics, information science, [2] and engineering. [4] Traditionally, most efforts to build artificial cell architectures have been concentrated on semipermeable membranes that delimit the interior regions. [5] The membranes preferentially formed at the oilwater interface are composed of phospholipids, [6] synthetic protein-polymer conjugates, [7] and diblock copolymers. [8] Recently, membrane-free systems have been increasingly explored because of their dynamic, open nature. The membrane-free architectures are formed via liquid-liquid phase separation (LLPS). [9,10] In the context of artificial cell studies, DNA droplets have emerged as a promising biomaterial for the construction of intelligent artificial cells ( Figure 1A). [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27] DNA droplets are micrometer-scale liquid-like condensate composed purely, or at least mostly, of DNA, a biomolecule in which genetic information is encoded. Since the first report published in 2018, [14] several groups have explored DNA droplets to demonstrate their fundamental functions, properties, and advantages, which are not parallel to other biomolecules. [11,24,28] One of the most notable features of DNA droplets is "programmable intelligence" in fluidic behavior ( Figure 1A-i). DNA droplets preferentially coalesce with peer droplets with the same sequences encoded in the SEs, but do not coalesce with nonpeer droplets with orthogonal, i.e., incompatible, sequences encoded. [11] Such sequence specifically engineered fluid dynamics reflects the fact that DNA droplets are a unique hybrid of two lines of research: One is DNA nanotechnology. [35,36] The exploitation of DNA as design material has become a powerful, Breathtaking advances in DNA nanotechnology have established DNA as a promising biomaterial for the fabrication of programmable higher-order nano/microstructures. In the context of developing artificial cells and tissues, DNA droplets have emerged as a powerful platform for creating intelligent, dynamic cell-like machinery. DNA droplets are a microscale membrane-free coacervate of DNA formed through phase separation. This new type of DNA system couples dynamic fluid-like property with long-established DNA programmability. This hybrid nature offers an advantageous route to facile and robust control over the structures, functions, and behaviors of DNA droplets. This review begins by describing programmable DNA condensation, commenting on the physical properties and fabrication strategies of DNA hydrogels and droplets. By presenting an overview of the development pathways leading to DNA droplets, it is shown that DNA technology has evolved from static, rigid systems to soft, dynamic systems. Next, the basic characteristics of DNA droplets are described as intelligent, dynamic fluid by showcasing the latest examples highlighting their distinctive features related to sequencespecific interactions and programmable mechanical properties. Finally, this review discusses the potential and challenges of numerical modeling able to connect a robust link between individual sequences and macroscopic mechanical properties of DNA droplets. reliable molecular platform for the creation of stable functional nano/microstructures. [35,37] The most appealing feature of DNA is its sequence-specific binding capability, known as Watson-Crick base-pairing. Sequence specificity allows for predictable molecular interactions, offering a solid basis for highly programmable biomaterial engineering, normally illustrated as "bottom-up" construction. [35] The molecular selectivity and programmability of DNA have revolutionized nanotechnology and science. To date, these advantages have been exploited for the precise fabrication of well-defined higher-order structures, including DNA origami, [31,[38][39][40] DNA tiles, [32,41,42,43] DNA 2D arrays, [44] DNA supramolecular polyhedra, [45] and DNA nanostructures with dynamic mechanical motion. [46][47][48] Beyond nanometer scales, the sequence specificity of DNA has been applied to biomolecular polymerization for the fabrication of DNA hydrogels, space-spanning 3D networked structures with elasticity and water retention. Since the pioneering work, [30] DNA hydrogels have attracted significant attention as smart biomaterial with stimuli-responsiveness, [49][50][51][52][53][54][55] biocompatibility, [56][57][58][59] controllable physical [60][61][62][63] and chemical [64,65] properties, and ease of functionalization. [66,67] Indeed, DNA droplets and DNA hydrogels possess these features in common. As explained later, DNA droplets show gel-liquid phase transitions with changes in applied temperature. Thus, DNA droplets are the latest derivative of DNA nanotechnology.
Another line of research related to DNA droplets is the LLPS of molecules in biological cells. [68,69] Increasingly recognized is that LLPS leads to the spontaneous formation of biomolecular condensates with fluid-like properties, [70,71] such as organelles and stress granules. Living cells compartmentalize and spatiotemporally coordinate their interlaced chemical reaction pathways within such phase-separated subsystems. [34] In contrast with membrane-based compartmentalization strategies such as vesicles and liposomes, [37,72] phase-separated droplets are not bound by membranes. This membrane-free nature allows phase-separated droplets to behave in a highly dynamic manner, [69,73] such as droplet coalescence, [33] active molecular exchange with the surrounding environment, [71,74,75] high stimuli-responsiveness, [76] selective molecular uptake, [77] intracellular molecular transport, [78] and genome DNA packaging. [79,80] Hence, reproducing dynamic subsystems in a confined microenvironment is a crucial step toward creating artificial cells. [73] Another distinct feature of DNA droplets, with which DNA hydrogels have in common, is high programmability in the mechanical properties ( Figure 1A-ii). As detailed below, DNA sequence programmability in the building block of DNA hydrogels/droplets allows for fine-tuning of larger-scale phase behavior and rheological properties. This trans-scale tuning depends on, e.g., stiffness, binding strength, and number of SEs of each building block. Because the link between the encoded sequences and the resulting mechanical properties is not as clear as that of the sequence-directed droplet interactions, high-resolution prediction of the macroscopic states requires numerical approaches as extensively explored for DNA hydrogels.
DNA droplets, micrometer-sized condensates of DNA nanostructures resulting from phase separation, [11,14] possess unique features inherited from both lines of research. DNA droplets are unparalleled design material for the creation of dynamic and intelligent artificial cells. This Review focuses on the basic features and functions of DNA droplets by presenting recent demonstrations of their unique features potentially applicable to various purposes from biomedical engineering to information science. We begin by providing a concise yet comprehensive introduction of DNA hydrogels, a precursory form of DNA droplets, both of which share many features in terms of fabrication strategies and physical properties. Next, we describe recent progress toward DNA droplets by showcasing increasing observations of liquid-like states in DNA hydrogel studies. The Review then puts DNA droplets in the context of the evolutionary path of DNA nanotechnology from rigid and static nanostructures to soft and dynamic microstructures. Next, we describe the basic features of DNA droplets as intelligent fluid with high programmability in droplet interactions and macroscopic mechanical properties. As a distinct example of the intelligent dynamic systems, we highlight sequence-specifically directed restructuring dynamics of DNA droplets and DNAbased information processing combining stimuli-responsive nature with sequence recognition capability. Additionally, we discuss the great potential and existing limitations of numerical modeling of DNA droplets as a means of high-resolution prediction of their mechanical properties, including the full phase behavior and elastic properties. Because numerical modeling of DNA droplets remains largely unexplored, we instead present numerical studies of DNA hydrogels readily available in the literature. Finally, this Review concludes with a brief description of the challenges and perspectives for further sophistication and widespread use of DNA droplets as a platform of creating artificial cells.

DNA Hydrogel
Nascent DNA droplets and DNA hydrogels, well-established DNA self-assembled soft matter, [49] have much in common in terms of physical properties (soft DNA-based material) and fabrication methods. In this review, both constructs shall be referred to as "DNA condensates." We begin this paper with a brief description of DNA hydrogels (Figures 2 and 3) to provide a basic understanding of the fabrication strategies and physical properties of DNA droplets. DNA hydrogels, with sizes ranging from nanometer to centimeter scales, have a mesh-like 3D network structure with properties such as biodegradability, high permeability, biocompatibility, low elasticity, and stimuli-responsiveness. [30,49,[81][82][83][84][85] DNA hydrogels have witnessed various applications, such as cell-free protein production ( Figure 2B), [86][87][88] pH sensing ( Figure 2C), [89] DNA-based laser printing ( Figure 2D), [90] photolithographic gel shape control, [91] multi-florescence emission ( Figure 3B), [67] and dynamic soft material with flow-regulatable locomotion ( Figure 3E). For a detailed description, refer to more comprehensive reviews of DNA hydrogels in the literature, [30,49,[81][82][83][84]92,93] together with some reviews for hydrogel in general. [94,95] DNA hydrogels can be broadly classified into two groups based on their fabrication process: 1) motif-based methods and 2) amplification methods using circular DNA. In 1) motif-based www.advanced-bio.com methods, DNA hydrogel is formed through the self-assembly of DNA branched nanostructures ( Figures 1A and 2A), [30,[96][97][98] typically with X-shaped and Y-shaped structures. In a single motif, multiple single-stranded DNAs (ssDNAs) are hybridized to form stem segments. The arms of the DNA nanostructure possess a sticky end (SE), a single-stranded overhang. Because each SE is designed as a palindromic sequence, the base-paring of SEs leads to larger-scale DNA hydrogels. Motif-based fabrication methods can be further categorized by the use of enzymatic ligation of branched DNA motifs. For the construction of stable bulk-scale material, the self-assembly of DNA motifs is coupled with ligase-catalyzed reactions, leading to covalent cross-linking of DNA motifs. [30,86] Additionally, in the case of more dynamic uses of DNA hydrogels, such as biosensing and bioprinting, variable motif interaction is key to controllable assembly and disassembly of the hydrogel structure. Hence, ligase-free methods, which do not depend on the ligase-catalyzed reactions, are preferred for achieving dynamic changes in physical properties.
In 2) another fabrication method, DNA hydrogel is synthesized using circular template DNA in conjunction with DNA polymerase (Φ29) and a primer ( Figure 3A). This well-established DNA amplification technology [99] is known as "rolling circle amplification" (RCA). [61] In RCA, DNA polymerase travels along the circular template DNA to elongate the primer. A notable feature of polymerase is its strong capacity for strand displacement. This ability enables the continuous production of long ssDNA with repeated sequences subjected to amplification and weaving of DNA. For more efficient amplification, chain reactions are initiated by adding multiple primers complementary to the primary and secondary ssDNA products. [61] As a result of the physical entanglement of the amplified products, a pleated structure of the DNA hydrogel was constructed ( Figure 3B). [54,61,67] The distinct advantage of the RCA-based method over motif-based fabrication is its ability to directly encode genes and functional sequences within the template DNA. [61,92,100,101] Significantly, the basic concept of RCA can be extended to RNA gel formation by replacing DNA polymerase with RNA polymerase ( Figure 3D-i), as demonstrated for RNA microsponge [102] and free-standing RNA membranes ( Figure 3D-iii). [101] The RCA-based methods coupling template programmability with hydrogel fabrication have witnessed various applications, from cell-free protein expression ( Figure 3C) [87] and biosensing [100] to in vitro evolution within microfabricated droplets. [103] An interesting example of the RCA-based method is the execution of the physical entanglement step in a microfluidic device with a periodic arrangement of pillars ( Figure 3E). [104] Upon injection of a generation solution containing RCA-amplified precursor DNA into a microfluidic device ( Figure 3E-i), vortex flow induced by the array of pillars enabled the spontaneous formation of DNA hydrogel ( Figure 3E-ii). The distinctive feature of this study is a realization of the biological metabolic process by coupling the entanglement-induced DNA gel formation with microfluidic regulation: There were three inlets in the microchannel-one was for the generation solution containing DNA polymerase, and the other two side inlets were for a degeneration solution containing DNA-hydrolyzing enzyme. By carefully engineering the coflow of the three solution flows, alternative gel disassembly and assembly continued, giving the appearance of dynamic locomotive behavior of DNA hydrogel migrating toward the upstream region ( Figure 3E-iii). [104]

From Gel to Liquid
Increasingly, fluid-like properties of DNA hydrogels have been observed by changing the applied temperature, [60] pH, [105] and hydration condition in the buffer. [61] In DNA hydrogels, sequence design is a powerful platform for their controlled phase behavior, which transitions between gel and liquid states. Specifically, DNA motifs provide a programmable template for engineering the valency of nanoparticles (Figure 4A-i). At a specific valency, a solution of DNA motifs phase-separated into DNA-poor and DNA-rich phases ( Figure 4A-ii), the latter of which was characterized by fluid-like properties. DNA motifs with varied valencies f adopted distinct phase behaviors. [106] Notably, by increasing the valency from f = 3 to f = 4, the accessibility of phase-separated state increased ( Figure 4A-iii). In a later study, the fluid-like property of the DNA-rich phase was explicitly demonstrated by the dynamic coalescence of two droplets in contact and microrheological measurements. [14] The choice of DNA motifs as a template provides a powerful, flexible means of designing DNA droplet structures, functions, and behaviors [11,18,19] for the following two considerations. 1) First, the motif template is readily amenable to various software suites for numerical analysis and design of nucleic acid systems. [107][108][109][110] The software outputs a numerical estimate of the stability, or technically "melting temperature" (T m ), of input sequences (the motif and its SEs, separately), generating a rough but easy-to-make prediction of the phase behavior of the DNA droplets of the input sequences. Such a user-friendly computer-aided design (CAD) approach, frequently used in DNA nanotechnology, is accessible to DNA droplets. For more precise but more time-consuming simulation of the full phase behavior, numerical modeling of DNA gels and liquids has been the subject of increasing interest, as discussed below. 2) Second, the motif-based strategy offers a limited set of design parameters to take into account for designing their interactions and mechanical properties (e.g., encoded sequence information, valency, flexibility, etc.), as will be explained in detail below. Thus, the multi-branch motifs provide scientists and engineers with a robust and facile route toward well-engineered hierarchical droplet architectures [11,16,19,20] and flexibility-controlled phase behavior. [18,25] In addition to the motif-based fabrication, other methods are available for the fabrication of DNA droplets. As illustrated in Figure 4D-i, entangled RCA-amplified ssDNAs form DNA droplets as a result of a well-optimized thermal cycling process. [21,22,28] Repeated sequences in RCA-amplified ssDNAs determine the temperature-dependent phase behavior. The use of multiple sequences with different T m values allowed different phase behaviors to be coupled within a single droplet ( Figure 4D-ii). Upon heating and cooling thermal cycling, the faster gel formation of one sequence formed the shell before encapsulating the slowly growing liquid phase of another sequence, leading to the formation of a caged droplet ( Figure 4D-ii). Another fabrication strategy of DNA droplets uses double-stranded DNA (dsDNA) as an ingredient. [111,112] Generally, in the presence of cationic polypeptides, an increased fraction of base-paired sequences preferentially leads to gel-like structures with less fluidity due to a decreased internal flexibility, as demonstrated for DNA [113,114] and RNA. [113] However, when coupled with poly-l-lysine (PLL), a positively charged cationic peptide, plus a well-optimized salt concentration, negatively charged short dsDNAs favored liquid crystal coacervate via end-to-end stacking. [115][116][117] At low salt concentrations, strong electrostatic interactions caused rigid precipitation of dsDNAs; at very high salt concentrations, strong screening effects on electrostatic interactions resulted in the disassembly of the DNA-peptide complex. At moderate salt concentrations, the DNA-peptide complex favored a well-ordered but fluid-like state (liquid crystal) ( Figure 4E). In another study using dsDNA, [111] fluid-like DNA coacervates were generated by exploiting a photoswitchable surfactant capable of hydrophilic-to-hydrophobic switching ( Figure 5). Last, ssDNA-based droplets have been reported; the hybridization-dehybridization switching of ssDNAs caused the assembly-disassembly of DNA droplets ( Figure 4F). [118,119] While RNA shares many features with DNA, RNA molecules favor 3D-structured intrastrand double helices and elaborate tertiary and quaternary structures, in striking contrast to DNA. [120,121] RNA transcripts possessing multiple repeat sequences were shown to form condensates: they exhibited solid-like behavior when transcribed in vitro but liquid-like behavior when transcribed within biological cells. [113] www.advanced-bio.com Phase behaviors of f = 4 nanostar confined within a capillary tube, which equilibrates in (left) dispersed phase at a higher temperature and in (right) liquid state at a lower temperature. Each state is marked by a symbol of the corresponding color in a (iii) phase diagram. Reproduced with permission. [106] Copyright 2013, National Academy of Sciences. B) DNA liquid assembled using (i) DNA nanostars. Unpaired "A" (in red) separates the stem and SE as a spacer for flexibility. (ii) (Left) Confocal microscopy images of phase-separated DNA liquid, with (right) fluorescence visualization. Scale bar, 20 µm. (iii) (Left) Time-evolution of coalescence of DNA droplets. Scale bar, 5 µm.
(Right) Time-dependent aspect ratio of the coalescing DNA liquid of width W and length L. Reproduced with permission. [14] Copyright 2018, Royal Society of Chemistry. C) Enzymatically induced sub-compartmentalization of a DNA droplet. (i) Two orthogonal motifs are coupled by a cross-linker, a six-branched DNA nanostructure. (ii) Equilibrium state of the sub-compartmentalized droplets with Janus shape. Scale bar, 20 µm. Reproduced with permission [11] under the terms of the Creative Commons CC BY-NC 4.0 license. Copyright 2020, American Association for the Advancement of Science. D) RCA-fabricated DNA droplets of ssDNAs with repeat block sequences. (i) RCA-generated ssDNAs favor a demixed state; upon heating and cooling, the entangled strands are hybridized in the repeat block sequences to form a mesostructure. (ii) By coupling gel-liquid phase behaviors with different temperature dependencies, core-shell DNA structures are formed. Reproduced with permission. [28] Copyright 2018, Springer Nature. Scale bar, 10 µm. E) DNA liquid crystal coacervates of negatively charged double-stranded DNAs (dsDNAs) and cationic peptide poly-l-Lysine (PLL). Reproduced with permission. [112] Copyright 2020, American Chemical Society. F) Gel-liquid phase control of DNA-peptide polymer complexes based on the hybridization state. Reproduced with permission. [118] Copyright 2018, American Chemical Society. www.advanced-bio.com

Crystalline or Amorphous?
It is worthwhile to discuss the design factors in motif sequences leading either to amorphous, soft DNA condensates or to highly crystallized structures of DNA tiles, both of which use scalable nano-scale motifs as a module. Prior to the explosion of DNA hydrogel studies, [30] Seeman suggested, in his early work, [122] the possibility of crystalline assembly of multibranched motifs into scalable 3D lattice-like structures through SE ligation. Following this suggestion, a few efforts to create crystalline 3D DNA structures were reported using three-branched [123] and four-branched motifs. [124] More recently, rigid macroscopic 3D DNA crystals were successfully constructed, with well-defined pores [125] and multi-layered structures. [126] The common strategy of the motif-based amorphous DNA condensates is to encode the identical self-complementarity in the SE sequences. The motifs bind with each other via the SE hybridization, leading to nondirectional growth to fully bonded network-like 3D microstructures. In contrast, the basic strategy to form the crystallized structures of DNA tiles is to introduce highly directional coordination between tiling units by assigning differentiated complementarities to their SEs, [32,41,42,[127][128][129] as well demonstrated in algorithmic assembly of DNA tiles into Sierpiński triangle (as seen in Figure 6), a well-known fractal pattern. [41] For example, if a unit tile possesses four SEs, "A" and "B" on one end and "a" and "b" on the other end (the upper case is complementary to the lower case), 1D, periodic growth is preferentially chosen.
The stiffness of modular motifs is another critical parameter for the crystalline assembly, as suggested by Seeman's early work. [122] Since the persistence length of dsDNA is 50 nm under physiological conditions, the tiling units, usually with the length of ≈10 nm, [32,41] can be considered as sufficiently stiff 2D plates. In practice, the key to the formation of the tiling units is to design a double-crossover (DX, Figure 7A) [130][131][132][133] or a triple-crossover [134] unit to avoid the formation of floppy, "marshmallow-like" components. [135] In a DX unit, the crossover strands are encoded to be immobile as opposed to mobile Holliday junctions, which are naturally encountered for gene swapping in living systems. As a result, the crossover strands function as a clamp in the tiling unit. The DX design disallows the deformation such as twisting and bending, and consequently, reinforces the directional growth. In contrast, the multibranched motifs, although normally possessing the double-stranded stem arms with less than 50 nm in length, are designed to have nonbinding nucleotides at specific sites, e.g., in the central junction ( Figure 7A) [11] and between the stems and the SEs ( Figure 4B-i), [14,15] to introduce the internal flexibility. This flexibility permits a higher degree of freedom in bending, twisting, and orientation, and thus the nondirectional growth. In general, colloidal particles with the encoded flexibility and limited valency (known as patchy particles) thermodynamically prefer liquid state rather than a crystalline state. [136] An interesting but seemingly irrelevant exception for this explanation was given by demonstrating the self-assembly of multi-branched motifs into 3D-structured crystals ( Figure 7C). [137][138][139] The key to this crystallization is to replace the monovalent SEs with non-specific, cholesterol-tagged strands to allow for multivalent interactions at the end of the stem arms. Furthermore, numerical simulations, complemented with X-ray scattering data, showed that the crystallization of amphiphilic DNA motifs needed a sufficient degree of internal flexibility. [139] Figure 5. Photoswitchable phase behavior of DNA droplets using photo-responsive surfactant (azoTAB). A) UV encourages hydrophilicity of the surfactant, leading to the dispersed state. Blue light, by contrast, enhances hydrophobicity, leading to phase separation. B) Microscopy images of photoswitchable phase behavior of DNA droplets as indicated by the turbidity of the samples observed (inset). Scale bars, 10 µm. Reproduced with permission. [111] Copyright 2019, John Wiley and Sons. www.advanced-bio.com

A Broad Perspective of the Evolutionary Path toward DNA Droplets
The aforementioned brief description of the preceding DNA hydrogel and the later emergence of DNA droplets presents a perspective of DNA droplets in the context of long-established DNA nanotechnology ( Figure 6). DNA nanotechnology has evolved from static and rigid nanostructures such as DNA origami, DNA arrays, DNA polyhedral, and DNA tiling. Later, the dynamic rearrangement of rigid DNA nanostructures with the introduction of stimuli-responsive molecules into DNA nanostructures has been reported for DNA origami [46,140] and DNA tiles. [32,127] Beyond nanofeatures, the application of DNA for polymerization has created DNA hydrogels, the subject of intense research for fundamental and industrial purposes as smart, soft biomaterial. DNA hydrogels as a cell-free protein factory are designed to avoid dynamic behaviors, whereas DNA hydrogel with capabilities such as stimuli-responsiveness, photoprinting, and flow-regulated locomotion are designed to possess dynamic nature. The brief sketch featuring the evolution from DNA hydrogel to DNA droplets highlights DNA droplets as an inherently dynamic soft biomaterial.

Dynamic Fluid
Broadly, the fabrication strategy of DNA hydrogels can be applied to DNA droplets. [14,106] The only factor distinguishing DNA droplets from DNA hydrogels is the higher structural flexibility. [14] A typical fabrication process of DNA hydrogels is In "DNA droplets," scale bar, 10 µm. Reproduced with permission [11] under the terms of the Creative Commons CC BY-NC 4.0 license. Copyright 2020, American Association for the Advancement of Science. In "DNA hydrogels," (P-gel pad) Scale bar, 500 µm. Reproduced with permission. [86] Copyright 2009, Springer Nature. (Swollen DNA hydrogel) Scale bar, 1 cm. Adapted with permission. [30] Copyright 2006, Springer Nature. (DNA NF) Reproduced with permission. [67] Copyright 2014, John Wiley and Sons. (pH-responsive DNA gel). Adapted with permission. [89] Copyright 2009, John Wiley and Sons. (Locomotive DNA gel). Reproduced with permission. [104] Copyright 2019, American Association for the Advancement of Science. (Laser printing/erasing) Scale bar, 100 µm. Reproduced under the terms of the Creative Commons CC-BY license. [90] Copyright 2022, The Authors, published by John Wiley and Sons. In "DNA nanostructures," (DNA origami) Scale bar, 100 nm. Adapted with permission. [31] Copyright 2006, Springer Nature. (DNA 2D array) Reproduced with permission. [44] Copyright 2005, American Chemical Society. (Algorithmic self-assembly of DNA tiles into Sierpiński triangle) Reproduced under the terms of the Creative Commons Attribution License. [41] Copyright 2004, The Authors, published by Public Library of Science. (Cycle of disassembly and reassembly of DNA nanotubes) Adapted with permission. [32] Copyright 2019, Springer Nature. (DNA origami-based rotary nanomotor) Reproduced under the terms of the Creative Commons CC BY license. [48] Copyright 2022, The Authors, published by Springer Nature.
www.advanced-bio.com a ligase-catalyzed self-assembly process in combination with a relatively high DNA concentration (≈200 × 10 −6 m [30] ). This strategy aims to achieve enhanced structural stability sufficient to support a large-scale 3D mesh-like bulk structure. By contrast, DNA droplets do not rely on either catalytic treatment or high DNA concentrations (≈5 × 10 −6 m [11] ). Hence, a lowered DNA concentration favors phase separation into DNA-rich and DNA-poor phases rather than bulk network formation. Additionally, the lowered structural strength allows for continuous reshuffling of the SEs, providing DNA droplets with fluidlike properties. [11,14] Thus, DNA droplets possess high diffusivity in the internal liquid phase. Normally, experimental support for the diffusive properties of phase-separated droplets is obtained using fluorescence recovery after photobleaching (known as "FRAP"). [11,14] The membrane-free surface of DNA droplets enables dynamic behavior, as is commonly encountered in other biomolecule condensates. For example, DNA droplets favor coalescence upon immediate contact with neighboring droplets. [11,14] High enzymatic sensitivity is another characteristic feature of the dynamic nature, as demonstrated using a restriction enzyme. [17] An enzymatically induced fast breakdown in DNA droplets resulted in not only time-dependent droplet shrinkage but also distinct vacuole formation due to increased osmotic pressure. Other dynamic behaviors seen in DNA droplets are selective molecular uptake based on nucleotide length [15] and diffusion-mediated molecular transport between DNA droplets encapsulated within a water-in-oil (W/O) droplet [24] as mentioned later. Very recently, reaction-diffusion waves were generated within a spherical geometry of DNA droplets. [20] In the presence of free ssDNAs of varied length outside the droplets, the strands diffused into the droplets at different paces, determined by the length-dependent diffusion coefficients. As a result, waves of strand displacement reactions propagated in a radial direction in the droplets, generating the reaction-diffusion patterns. Thus, DNA droplets, as membrane-free compartments, are a promising platform for the creation of dynamic synthetic cell-like systems. For a general description of phaseseparated condensates in the context of creating artificial cells, see some comprehensive reviews in the literature. [10,73] [135] Copyright 2003, Springer Nature. B) A representative example of a spacer encoded in Y-motif as "TT" in the middle of each strand. Reproduced under the terms of the Creative Commons CC BY-NC 4.0 license. [11] Copyright 2020, The Authors, published by American Association for the Advancement of Science. C) Amphiphilic DNA nanostar. In the place of a monovalent SE, a cholesterol-tagged anchor strand is used to provide non-specific multi-valent interactions of the hydrophobic tags. Reproduced under the terms of the Creative Commons CC-BY license. [137] Copyright 2017, The Authors, published by American Chemical Society. www.advanced-bio.com

Intelligent Fluid
Another distinguishing feature of DNA droplets is their sequence-specific selectivity in inter-droplet interactions. Because of the continuous reshuffling of the SEs that serve as sequence recognition sites, droplet coalescence dynamics can be regulated as prescribed by the SE sequence design (Figure 8): [11] Two DNA droplets in contact whose constituent motifs possess identical SEs (e.g., 5'-GCTCGAGC-3') favor coalescence due to high affinity in the SE interaction, as do the other biomolecular condensates. By contrast, two droplets with orthogonal sequences in the SE, where, e.g., one SE is designed as aforementioned, and the other SE is 5'-CTCGAGAG-3'), disfavor coalescence even in immediate contact due to little affinity. Rigorously speaking, the SE-encoded complementary interactions enable droplets to overcome the electrostatic repulsion of negatively charged DNA strands toward complete coalescence. Thus, salt concentration significantly affects not only the fluidity of DNA droplets [14] but also their interactions, because of salt screening effects that counteract the electrostatic repulsion.
Membrane-free DNA droplets with the sequence recognition capability offer an advantageous platform for the reconstitution of an intracellular network of molecular transport pathways. A recent experiment showed one-way and two-way transport of ssDNAs between two neighboring DNA droplets confined within a W/O droplet. [24] Cargo (ssDNA) transport was regulated by the azobenzene-based photoswitching in two neighboring DNA droplets designated as a sender and receiver. Inter-droplet transport was mediated by Brownian motion and directed by programmed sequences of DNA droplets.
The sequence recognition capability of DNA facilitates the sequence-directed construction of higher-order structures. An example of a well-programmed structure was demonstrated for hierarchical DNA droplets. [19] In the "core-shell" droplets, the core region was composed of DNA motifs that possessed cholesterol-functionalized branches and thus amphiphilic properties; the outer region was composed of unfunctionalized normal DNA motifs and sealed the amphiphilic nature of the inner region. Upon a molecular cue input and triggered dissociation of outer motifs via toehold displacement, their amphiphilic nature was exposed. When crowded in contact with a membrane-bound vesicle, the core-shell DNA droplets exposed their inner amphiphilic region and disrupted the vesicle membrane, owing to increased permeability.
In living systems, the sub-compartmentalization of phaseseparated condensates is crucial for the spatiotemporal coordination of highly intertwined biochemical reactions occurring within biological cells. The immiscibility of two DNA droplets with orthogonal sequences (Figure 8) offers a facile route for the formation of such sub-compartmentalized structures. The basic strategy is to bridge orthogonal-SE droplets with a cross-linker DNA that possesses binding affinities for both DNA droplets.
A mixture of two orthogonal DNA motifs and a DNA crosslinker with binding affinities for both motifs favored the coexistence of two immiscible phases within a single droplet upon annealing (Figure 9A-i). [11,12,16] Notably, over a spectrum of concentration ranges of the cross-linker, the sub-compartmentalized conformation varied significantly ( Figure 9A-ii): at very low concentrations, bulb-shaped compartments adhered to each other; at moderate concentrations, distinctly separated subcompartments constituted a single spherical droplet without forming bulb-shaped subunits; and at very high concentrations, the sub-compartmentalized structures disappeared and equilibrated in a completely mixed state. [16] This conformational control using the cross-linker concentration can be explained by the interfacial free energy between the immiscible DNA phases: increasing the concentration of the cross-linker leads to a decrease in the interfacial energy, which allows for a larger area of the interface; in contrast, decreasing the linker concentration and thus increasing the interfacial energy forces the immiscible phases to minimize the interfacial area. Thus, the cross-linker serves as a neutralizer for immiscible DNA droplets with orthogonal SEs.  [11] Copyright 2020, The Authors, published by American Association for the Advancement of Science.

www.advanced-bio.com
Cross-linker-based sub-compartmentalization offers an effective means of the dynamic remodeling of DNA droplets. [11,27] As shown in Figure 9B-i, enzyme-responsive cross-linkers with six branches were considered in Sato et al. [11] Hereafter,  [11] Copyright 2020, The Authors, published by American Association for the Advancement of Science.

www.advanced-bio.com
we refer to the six-branched cross-linker as the "S-motif." The S-motif, which cross-links the two orthogonal motifs, neutralize the immiscibility between them, creating fully mixed droplets ( Figure 9B-ii). As illustrated in Figure 9C-i, the key to dynamic remodeling is to cut a S-motif into two fragments, each of which is for the corresponding DNA motif. To introduce nicks in the motif, specific DNA sequences were replaced with RNA of the equivalent sequences. Hereafter, we refer to this chimeric design containing DNA and RNA as the "CS-motif." In the presence of ribonuclease A (RNase A), the RNA segment underwent rapid degradation upon enzymatic reaction, and the CS-motif was subsequently cleaved into two fragments. Depending on the mixing ratio of the S-and CS-motifs, an initially mixed droplet dynamically rearranged to different subcompartmentalized states: when the cross-linkers were composed exclusively of the CS-motif (0:1 mixing ratio of S-motif and CS-motif), a DNA droplet initially in the mixed state showed drastic fission upon enzymatic reaction with RNase A ( Figure 9C-ii). For a 1:9 mixing ratio of the S-motif and CSmotif ( Figure 9D-i), a completely mixed DNA droplet dynamically transitioned to a Janus-shaped droplet, where two bulbshaped immiscible hemispheres were in contact ( Figure 9D-ii). For an equimolar mixture of S-and CS-motifs, the equilibrium state was a patchy-like droplet. The splitting behavior for the 0:1 mixing ratio may provide a straightforward route toward in vitro emulation of biological cellular division and hence provide a highly programmable model system in the study of the origin of life. [141][142][143] Other subcompartmentalization methods using W/O droplets as a template for artificial cells [144] have been achieved with phase transfer [145,146] and microfluidic devices; [147,148] each of these methods is limited by low programmability and time-consuming device fabrication steps, respectively. The high programmability and controllability of multi-compartmentalization is an incomparable advantage of DNA droplets.
The sequence-based programmability of DNA droplets in their interaction and structure design is an as yet unmatched feature in other biomolecular condensates. The intelligent aspect of DNA droplets lays the groundwork for far-reaching applications in various fields, from biomedical engineering and science to computer science, as presented below.

DNA-Based Information Processing Systems
Biological systems rely on complex networks of their internal chemical reactions to regulate their behaviors in response to the external environment. An assembly of DNA-based constructs processes sequence recognition of component strands in a computer-like manner for sequence-directed ordering of higher-order structures. This intelligent capability makes DNA a unique biomolecular platform for modeling bio-inspired information processing systems. Since the pioneering work of DNA computation, [149] a myriad of DNA-based information processing systems has been applied to many diagnostic technologies, e.g., early detection of biomarkers and intracellular pH sensing. [150] These functional DNA systems are broadly classified into two categories: One category is DNA-based computation that executes, e.g., various algorithms [149,151] and Boolean logic operations (AND, OR, etc.). The DNA-based logic operating systems use receive biomolecules (such as DNA, [36,151,152] RNA, [153][154][155] protein, [156,157] proton, and ATP [158] ) as binary inputs (0 and 1) and generate outputs interpreted as the binary digits. The other category is a biosensor [159][160][161][162][163] that upon receiving targeted stimuli, signals a specific form of outputs, usually fluorescence. In these functional DNA systems, the information processing often carries out the strand displacement reactions, where a free strand displaces a member strand of the hybridized strands with itself and forms a new duplex. [164] The architectures of these implementations range from a mixture of multiple strands [149,151,155] to nanoscale (origami and tiles) [156][157][158][159] and microscale structures (hydrogels). [52,58,163,165] The underlying mechanisms and basic strategies employed in these implementations have been detailed in the literature and overviewed in other review papers. [58,[165][166][167][168] In biological cells, phase-separated condensates utilize their membrane-less nature for timely response to environmental stimuli and subsequent dynamic restructuring for spatiotemporal regulation of chemical reactions. Thus, the construction of intelligent microscale robots eagerly awaits such stimuli-sensitive mechanisms. A recent study has employed DNA droplets to detect cancer biomarkers by combining molecular sensing and DNA computing. [13] This computational DNA droplet can recognize specific combinations of tumor biomarker micro-RNAs (miRNAs) as molecular inputs and output the results of logical operations through physical DNA droplet phase separation (Figure 10). The computing is logically controlled by a linker encoding two receptor sequences that recognize two inputs and fulfill the characteristics of an AND gate. Two linker insertions were applied to four miRNA detections and miRNA pattern recognition for breast cancer diagnosis. [169]

Phase Behavior
The fluidity of DNA droplets is another programmable feature. Generally, the fluidity of phase-separated droplets significantly affects their dynamic nature such as diffusivity, enzymatic sensitivity, and inter-droplet coalescence. In biological cells, decreased fluidity, or liquid-to-solid phase transition, of phase-separated condensates can lead to decreased molecular exchange with the surrounding environment. It has been increasingly suggested that for phase-separated biomolecular condensates, the liquid-to-solid phase transition causes various diseases, such as amyotrophic lateral sclerosis (known as "ALS") [170] and Alzheimer's disease. [171] Thus, fluidity regulation via phase transition is a crucial mechanism underlying the dynamic nature of artificial cells.
The phase behavior of the DNA droplets can be reversibly controlled by the applied temperature field (Figure 11). [11,28] The sequence design of SEs determines their binding affinity, expressed as − ΔH (ΔH: binding-unbinding enthalpy difference). For a fixed value of − ΔH, the phase behavior shows a monotonic dependence on temperature ( Figure 11). An increase in the valency of motifs leads to an increase in the critical temperature, [172], i.e., higher preference to bonded states. Thermodynamically, this is a consequence of a decreased total of ΔH within a system. ΔH can be well engineered through the www.advanced-bio.com Figure 11. Temperature-controlled phase transition of DNA droplets. Phase transition is reversible. Figure 10. Molecular computing based on DNA droplet phase separation. [13] A) Design of molecular computing reaction. B) Schematic of a computational DNA droplet operating ((miRNA-1 ∧ miRNA-2) ∧ (miRNA-3∧ ¬miRNA-4)). C) Confocal microscopy images of ABC-mixed-droplets and phase separation corresponding to miRNA inputs. Scale bars, 10 µm.
www.advanced-bio.com design of the SEs of DNA motifs, usually with the assistance of CAD software. Thus, temperature-dependent phase behavior is another programmable property of DNA droplets.
As a regulation mechanism of phase behavior, isothermal control is more suitable than temperature control for minimizing non-specific effects on other targets. Because of its high locality and minimal invasiveness, photocontrollability of the phase behavior is an effective approach to isothermal phase control. As shown in Figure 5, the isothermal phase transition of the DNA droplets was demonstrated using a photoswitchable surfactant (azobenzenetrimethylammonium bromide, known as "azoTAB"). [111] Upon UV and visible (vis) light irradiation, cis/trans isomerization of the surfactant yields hydrophilic/ hydrophobic interactions with dsDNAs. This provides a photoswitching mechanism between the dispersed and liquid states of the DNA droplets.
Photoswitching of the phase behavior of DNA droplets has also been achieved by introducing photoisomerizable chemical compounds such as azobenzene [173] and arylazopyrazoles [174] in DNA oligonucleotides. [119,175] The photoregulatory mechanism is based on cis/trans isomerization of chemical compounds inserted into DNA sequences. [173,174] UV and vis irradiation induce cis and trans isomers of the compounds, respectively: the former state favors enhanced base-stacking of DNA strands, and the latter favors unstacking due to steric effects. [173] This molecular-level photoswitching translates into a macroscopic phase transition. Generally, for the creation of stimuli-responsive biomaterial, inserting photoisomerizable compounds provides a powerful photoswitching mechanism. [176]

Complex Relationship between Individual Sequences and Mechanical Properties of DNA Condensates
As mentioned earlier, CAD software for designing 3D DNA nanostructures gives engineers and scientists a rough estimate of the phase behavior of micro-scale DNA condensates. However, a robust link between the thermodynamic parameters of individual DNA sequences and the mechanical properties (microrheology and full-phase behavior) of the resulting DNA condensates is still missing. DNA motifs offer multiple design parameters that govern the mechanical properties of DNA condensates: [177] i) the number of SEs in a single motif, or val ency. [11,29,60,106,178] A decrease in the motif valency led to a limited access to the phase-separated state ( Figure 4A). [106] ii) the SE binding strength. [11,60,85] iii) the concentration of DNA. [85] iv) the presence of competition between two different binding mechanisms toward specific motifs. For example, consider gelforming Motif A and free strands B and C, both of which can bind with the SEs of Motif A; [179,180] At very high temperatures, A, B, and C are dispersed. At moderate temperatures, A forms solid structures, while B and C still remain dispersed; At very low temperatures, B and C jointly replace AA bonding with AB and AC bonding, resulting in fluid-like states again. In general, this nonmonotonic temperature dependence of phase behavior is known as re-entrant melting. [181,182] v) the presence of another DNA motif with orthogonal sequences within a twocomponent suspension. [183] Dynamic light scattering measurements showed that the two-component suspension exhibited faster restructuring dynamics than single-component suspensions. vi) the buffer conditions, e.g., ionic strength [172,178,184] and cationic polypeptides, [113,114] which affect the electrostatic interactions. Finally, vii) the structural flexibility imparted by a spacer of nonbinding nucleotides inserted at different sites (e.g., between the stems and SEs, [14,15,25,185] in the central segment of the motif, [25,139] and in the middle of the stems [18] ). The flexibility of a spacer depends not only on the length but also on the encoded base sequences, [18] because poly-dA sequences are more rigid than poly-dT sequences. [186] Apart from the sequentially designed internal flexibility, salt concentration can be an external control parameter to the overall motif shape due to screening effects against electrostatic repulsions of negatively charged DNA. [14,172,178] The flexibility in the building blocks significantly affects the rheological properties of hydrogel in a manner, which is largely dependent on the design of the building blocks (Figure 12). [18,177,187,188] For systems of Y-motifs cross-linked with linkers with different spacer sequences, flexible linkers showed lower values of the storage modulus than rigid linkers at moderate temperatures ( Figure 12A). Counterintuitively, for systems of chain-like structures cross-linked with each other, a greater fraction of flexible linkers enhanced the storage modulus ( Figure 12B). [187] Similar counterintuitive phase behavior was also seen for systems of Y-motifs possessing a single-stranded spacer in the middle of each stem. More flexible motifs (i.e., larger spacer length) exhibited a significant shrinkage of the condensed-phase region in the phase map ( Figure 12C). [18] In agreement with this experimental observation, simulations showed that a topologically assembled structure of more flexible motifs was more vulnerable to an irreversible breakup due to enhanced thermal fluctuation ( Figure 12C-iii). In contrast, when spacers were inserted outside the stem segments (i.e., in the junction center and between the stem and SE), higher flexibility led to stronger preference of liquid state rather than solid gel state. [25] Due to the complicated dependency of the mechanical properties on many parameters, high-resolution predictions are substantially required for rational design of DNA condensates. To bridge the gap between individual sequences and resulting mechanical properties of DNA condensates, numerical approaches able to generate experimentally unachievable insights are desperately needed. For an informative guide on the rheology of DNA hydrogels and related numerical/experimental methods, see some reviews in the literature. [177,189,190]

Coarse-Grained Modeling
In numerical modeling of DNA condensates, the most popular method uses coarse-grained models, because molecular-level resolutions are far beyond the limits of existing computational powers. The most well-known coarse-grained model is oxDNA, [191,192] which has been extensively applied to calculation of the thermodynamics of DNA constructs. [18,139,172,177,178,188,193,194] Besides Monte Carlo calculation, oxDNA executes molecular www.advanced-bio.com Simulations demonstrate that with increasing the flexibility, the hexagonal topology is more prone to a breakup due to enhanced thermal fluctuation conferred by increased structural flexibility. (iv) Simulations show that higher rigidity leads to a higher probability in the SE bonding. Reproduced with permission. [18] Copyright 2021, Royal Society of Chemistry.
www.advanced-bio.com dynamics (MD) algorithms, which allow individual molecules to move according to Newton's law of motion. oxDNA represents nucleotides as connected rigid bodies possessing interactive sites, which model various interactions, such as the base stacking, hydrogen bonding of opposing bases, and backbone interactions with the neighboring rigid bodies.
Nevertheless, oxDNA-based simulation is still computationally demanding, limiting its realistic use to the case spanning a time duration of only a few µs. For example, in a computational box containing a total of 19 600 rigid particles interacting with each other, the time evolution over 10 µs required 10 9 MD steps for each state point even with GPU-boosted computational power (×40 faster than CPU-intensive calculations). [195] A practical alternative to oxDNA may be LAMMPS (large-scale atomic/molecular massively parallel simulator), [196,197] another platform of MD simulation. LAMMPS, equipped with parallel computing capabilities in MD simulation, runs on multi-core CPU architectures. The all versions of oxDNA have been implemented into LAMMPS. [198] As in Figure 13A, LAMMPS was applied to Y-shaped units of connected beads as a model of Y-motifs to evaluate the elastic properties of hydrogels: [199] In the proposed model, besides the excluded volume as a repulsive potential between the beads, two types of attractive patchy sites, patch A and patch B, were defined on the surface of the end beads as a single-stranded SE ( Figure 13A-i,ii). The attractive interaction was limited to a pair of patch A and patch B, and inhibited for identical patches (A-A and B-B). In a numerical domain, there were two groups of motifs containing different combinations of patches, A-A-B and A-B-B, to secure homogeneously connected states. LAMMPS-based simulations demonstrated experimentally unachievable high-resolution data on time evolution of the pairwise bonding ( Figure 13A-iv). Besides, by applying an oscillatory shear on the computational box ( Figure 13A-v), the simulations further extracted the elastic properties of the fully bonded systems ( Figure 13A-vi).
The parallelized computing of LAMMPS seems to reduce an elapsed computation time in oxDNA simulation and allow for an access to much larger systems. [199] In a recent report of benchmark tests to compare the performances of both platforms, [200] for large system sizes, LAMMPS deployment of multiple MPI tasks showed much better performance than oxDNA simulations using a single CPU by a few orders of magnitudes. For small system sizes, however, the performance of LAMMPS on multiple CPUs was marginally faster than the single-CPU oxDNA simulations. A later benchmark report using a single Intel Core i7 processor showed a higher performance of singlecore-based oxDNA than LAMMPS by a factor of 1.6. [198] This less-than-expected performance of LAMMPS may have resulted from a larger number of degrees of freedom per nucleotide to take into account (13 and 18 degrees of freedom for oxDNA and LAMMPS, respectively), which can lead to increased communication overheads. Thus, the efficiency of LAMMPS largely depends on the computational architectures employed and the attributes and scales of the systems to simulate.
Another efficient solution to reduce the computational demand is to combine theory and oxDNA simulations. [172,178] The most common theory employed is Wertheim thermodynamic perturbation theory [201,202] (WTPT), developed to predict the thermodynamic properties of associating fluids. Associating fluids, characterized by strong, directional, short-range interactions, prefer to form fully connected network structures due to the limited number of connected bonds, as opposed to simple fluids possessing symmetric short-range repulsion and weak long-range attraction. [203,204] Prior to the emergence of the hydrogel-forming DNA motifs, patchy colloids modeled as repulsive spherical particles carrying a fixed number of attractive sites on the surface were the ideal target to which WTPT can be applicable. [205,206] Thus, the complementation of the numerical modeling with WTPT appears to be a natural choice, since DNA motifs and patchy colloids have many features in common. In WTPT, the free energy of the system is a sum of two contributions from 1) a reference part, where the interactions are modeled as a spherical, repulsive potential to model the reference fluids and 2) an associating part, which arises when taking into account the bond formation for the targeted associating fluids. [172,178,203,204] The basic strategy employed to combine the theory and numerical modeling is to calculate the effective potential in two-body interactions and the related coefficients as inputs for the theoretical framework. [172,178] By scrambling the encoded sequences in the associative SEs, nonbonding DNA motifs were considered to model the reference fluid. As an associative contribution, the bonding probability between the SEs was evaluated using the well-established Santa Lucia nearest neighbor model. [207,208] Figure 13B-ii shows the coexistence regions in the phase map spanning the temperature-density space for four-valency motifs ( Figure 13B-i) for different-order virial approximations. [172] Figure 13B-iii demonstrates the dependence of the critical temperature on salt concentrations for systems of different valencies.

Beyond Coarse-Grained Modeling
The numerical approaches introduced so far have been applied to equilibrium DNA hydrogel systems, which are less dynamic than DNA droplets. Apart from the equilibrium phase states, more dynamic aspects of DNA droplets should emerge as a significant rich source of insights into their mechanical properties. Examples include coalescence/break-up, [209] sub-compartmentalization, and diffusion-mediated molecular exchange between droplets, where some droplets grow in size at the expense of other smaller ones as a result of chemical potential difference. [210][211][212][213] These dynamic behaviors take place, usually on a time span of more than tens of seconds [11] in a microscale field, which are much larger spatiotemporal scales than the coarsegrain model can reasonably access on the existing computational power. Besides theoretical frameworks, [211,212] this consideration indicates the desperate need for continuum modeling that accounts for thermodynamic parameters while calculating hydrodynamics to bridge the gap between the individual sequences and macroscopic dynamic behaviors of DNA droplets. [214,215] Such numerical modeling could extract experimentally unachievable quantitative data as well as high-resolution visualization of dynamically restructuring DNA droplets in response to stimuli.

Conclusions: Challenges and Prospects
For widespread use and further sophistication of DNA droplets, there are two as yet poorly explored research directions: www.advanced-bio.com Figure 13. Numerical simulations using coarse-grained models. A) Numerical studies using LAMMPS, parallelized molecular dynamics simulator. [196,197] (i) Schematics of coarse-grained models of a Y-motif based on a bead-spring representation to model the double-stranded stems and excluded volumes around the beads to introduce repulsive interactions. At the surface of the end beads, attractive patches are imposed to model complementary SEs binding. (ii) Profiles and schematics of the interactions between the beads. (iii) A representative initial configuration of Y-motifs in a simulation box. (iv) Time-dependent number of connected pairs toward equilibrium states. (v) Snapshots of fully associated systems when (left) unsheared and (right) sheared. (vi) Simulated storage modulus G' for different number densities of Y-motifs ρ. Reproduced with permission. [199] Copyright 2019, American Chemical Society. B) A hybrid approach combining a theoretical model of associating fluids [201,202] with oxDNA, [191,192] a coarse-grained model evaluating the thermodynamics of DNA constructs. (i) Snapshots of DNA motifs with different valencies. (ii) Simulated plots of the coexistence regions in the T − ρ map (T: temperature) for varied salt concentrations S. Solid lines and symbols refer to second-and third-order virial approximation. (iii) Critical temperatures T C as a function of S for different valencies. B 2 and B 3 are the second and third virial coefficients in the virial expansion of the reference free energy. Reproduced under the terms of the Creative Commons CC-BY license. [172] Copyright 2018, The Authors, published by Multidisciplinary Digital Publishing Institute.
www.advanced-bio.com 1) One is the experimental evaluation of their physical properties, as discussed earlier. The physical properties of DNA droplets provide useful guidelines for the fabrication and manipulation strategies of DNA droplets in various applications. The micro-rheology (e.g., viscosity, diffusion coefficient, and interfacial tension) of DNA droplets has been evaluated experimentally. [14,114] The devices and methods used therein include conventional rheometers, microbead tracking within the droplets, and observation of coalescing droplets. Despite its significance, the microrheology of DNA droplets remains less studied. 2) The other research direction worth exploring is well-controlled organization of multiple DNA droplets to form networked structures. As demonstrated for W/O droplets, wellcontrolled organization of networked droplets in highly ordered formations are an essential step toward collective intelligence and behavior, analogous to those of biological tissues. [216][217][218] For this purpose, 3D printing technology is available for controlled packing of W/O droplets assembled in a layer-by-layer manner. [219] Recently, DNA-functionalized lipid-layered particles (solid particles [220] and oil-in-water (O/W) droplets [221] ) were demonstrated to achieve spontaneous organization as encoded in DNA linkers freely mobilized on the droplet surfaces. Specifically, the DNA linkers freely mobilized on the O/W droplet surfaces underwent a sequence of strand displacement reactions as programmed, leading to propagative organization. [221] For these purposes, direct accessibility of DNA droplets should be an eagerly awaited feature in various situations from generation to manipulation. For example, a series of droplet coalescence events yields a good estimate of the interfacial tension of liquid phase, based on the empirical relation between the characteristic time of coalescence and the interfacial tension. [222,223] The direct manipulation capability would significantly reduce the working time in the case of large sampling sizes. Moreover, the generation and manipulation of DNA droplets with well-controlled size will benefit the construction of functionalized DNA droplets, such as DNA-based computing, artificial cell synthesis, and controlled organization of tissue-like structures. However, self-assembly processes of DNA droplets, currently limited within reaction tubes, suffer from lack of direct accessibility and uncontrollable polydispersity, as seen in vinaigrette.
Generally, the prevailing method for the generation and manipulation of droplets relies on microchannels, [224][225][226] a powerful platform for the artificial cell synthesis and manipulation. [227][228][229] The well-established droplet-generating microfluidic devices are flow-junction systems that exploit Rayleigh-Plateau instability as an underlying mechanism. [230,231] However, ultralow interfacial tension of DNA droplets as water-in-water (W/W) systems, [14] much lower than that of highly immiscible W/O systems by orders of magnitudes, notoriously prohibits microfluidics-based construction and manipulation of W/W droplets, with the exception of a few successful reports. [232][233][234] This is because W/W systems cannot readily achieve a welloptimized balance between the surface tension and the viscous effect within reasonably sized geometries. [230,235] DNA droplets, as intelligent and dynamic fluid systems, are at the forefront of increasingly intense research activities to create artificial cells. DNA droplets are a unique hybrid inherited from DNA nanotechnology and phase-separated condensates. The most notable features of DNA droplets include, but are not limited to, programmable interactions, high enzymatic sensitivity, programmable sub-compartmentalization, programmable phase behavior, biosensing and logic operations, and well-controlled molecular transport within DNA droplets. DNA droplets would serve various purposes, including the construction of artificial organelles, controlled droplet division, and cell-free protein production. Further, combining the DNA-based bottomup methods with top-down additive manufacturing may create a breakthrough toward the synthetic cell construction. [4,236,237] Biological cells are highly complex chemical reactors; thus, further research is required to overcome the current limitations in the artificial cell studies. We envision that DNA droplets will provide smart, reliable design/fabrication tools to overcome these challenges.
www.advanced-bio.com Yusuke Sato is an associate professor at the Kyushu Institute of Technology. He received his Ph.D. degree in Engineering at Tohoku University in 2018. He was a superlative postdoctoral research fellow of Japan Society for the Promotion of Science and worked at Tokyo Institute of Technology (2018-2020). He then moved to Frontier Research Institute of Interdisciplinary Sciences as an assistant professor at Tohoku University (2020-2022). His research interests focus on constructing artificial molecular devices and systems based on DNA nanotechnology.
Masahiro Takinoue is a full professor at the Department of Computer Science, Tokyo Institute of Technology (Tokyo Tech), Japan. He received his B.Sc., M.Sc., and Ph.D. in Physics at The University of Tokyo, Japan, in 2007. After a postdoctoral fellow at Kyoto University, an assistant professor at The University of Tokyo, and an associate professor at Tokyo Tech, he has held a biophysics research group studying artificial cell science and molecular robotics using living soft matter and DNA nanotechnology at Tokyo Tech.