DNA cytoskeleton for stabilizing artificial cellsChikako Kurokawaa, Kei Fujiwarab, Masamune Moritac, Ibuki Kawamatad, Yui Kawagishid, Atsushi Sakaia,Yoshihiro Murayamaa, Shin-ichiro M. Nomurad, Satoshi Muratad, Masahiro Takinouec,1, and Miho Yanagisawaa,1

aDepartment of Applied Physics, Tokyo University of Agriculture and Technology, Tokyo 184-8588 Japan; bDepartment of Biosciences and Informatics, KeioUniversity, Kanagawa 223-8522, Japan; cDepartment of Computer Science, Tokyo Institute of Technology, Kanagawa 226-8502, Japan; and dDepartment ofRobotics, Tohoku University, Sendai 980-8579, Japan

Edited by Dan Luo, Cornell University, Ithaca, NY, and accepted by Editorial Board Member Herbert Levine May 30, 2017 (received for review February 9, 2017)

Cell-sized liposomes and droplets coated with lipid layers havebeen used as platforms for understanding live cells, constructingartificial cells, and implementing functional biomedical tools suchas biosensing platforms and drug delivery systems. However,these systems are very fragile, which results from the absence ofcytoskeletons in these systems. Here, we construct an artificialcytoskeleton using DNA nanostructures. The designed DNA olig-omers form a Y-shaped nanostructure and connect to each otherwith their complementary sticky ends to form networks. Toundercoat lipid membranes with this DNA network, we usedcationic lipids that attract negatively charged DNA. By encapsu-lating the DNA into the droplets, we successfully created a DNAshell underneath the membrane. The DNA shells increased in-terfacial tension, elastic modulus, and shear modulus of thedroplet surface, consequently stabilizing the lipid droplets. Suchdrastic changes in stability were detected only when the DNA shellwas in the gel phase. Furthermore, we demonstrate that lipo-somes with the DNA gel shell are substantially tolerant againstouter osmotic shock. These results clearly show the DNA gel shellis a stabilizer of the lipid membrane akin to the cytoskeleton inlive cells.

cytoskeleton | self-assembly | DNA gel | lipid droplet | liposome

Liposomes have been used as artificial cell models to un-derstand cell shape, membrane protein function, and lipid−

protein interaction, among other biological functions (1–3). Inaddition, liposomes have been used as a platform for biosensingand as drug delivery systems (DDS) because of their excellentbiocompatibility and biodegradability (4). However, liposomescollapse easily against environmental shifts and mechanicalforces because of their low bending modulus. The fragility ofliposomes causes uncontrolled leakage of the entrapped com-pounds and thus inhibits their use in biomedical applications andartificial cells experiments.In contrast, cell membranes are tolerant against environmen-

tal shifts and mechanical forces. The stability of cell membranearises from the cytoskeleton underneath the membrane. Themajor component of cytoskeletons is actin (5). Actin gels showhigh elasticity (6), which ensures the stability of cell membranesagainst various forces. For liposomes, the use of actin filamentsas a cytoskeleton is not an optimal strategy for the followingthree reasons: First, although actin bundles and actomyosin ringshave been reconstituted in artificial cells (7, 8), formation of anactin cortex underneath artificial membranes has been stillchallenging. Second, actin is hard to modify by chemical andgenetic means because of its essentiality for cell growth. Third,the physicochemical properties of actin gels are still unclear (9,10). Hence, the cytoskeleton of liposomes should be constructedwith defined and designable materials. To accomplish this aim,DNA nanotechnology, which uses limited components with highdesignability in a nanometer scale (11), is a feasible candidate toconstruct cytoskeleton structures in artificial cells.DNA nanostructures form three- or four-way junctions, con-

nect with each other through their complementary sticky ends,and are reported to become gels by forming network structures(11–14). DNA gels have several advantages as the cytoskeleton

of liposomes compared with other polymers: (i) the gelationtemperature of DNA gels is easily controlled by changing the DNAsequences and nanostructures, (ii) the interaction between DNAgels and lipid membranes can be controlled because the negativelycharged DNA has an attractive force with the positively chargedmembrane (15), and (iii) DNA sequence is easily designed withcomputer software [such as Nucleic Acid Package (NUPACK)(16) and Cadnano (17)] and easily chemically modified (11).Here, we implemented a DNA cytoskeleton by enforced lo-

calization of Y-shaped DNA nanostructures (hereafter referredto as Y-motif DNA) on lipid membranes using the attractiveinteraction between DNA and positively charged lipids. Wesucceeded in preparing droplets and liposomes that have a shell-like structure of self-assembled DNA underneath the membrane.The DNA shell in the gel phase increased the mechanicalstrength of the membrane, such as the interfacial tension of thedroplet surface and the critical pressure where the dropletscollapse. Therefore, the artificial cytoskeleton made from theDNA gel shell works to enhance the membrane stability akin tothe cytoskeleton in cells.

Materials and MethodsMaterials. To prepare the artificial cells, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC; Avanti Polar Lipids), positively charged lipid 1,2-dioleoyl-3-trimethylammonium-propane (DOTAP) (chloride salt; Avanti PolarLipids) and mineral oil (Nacalai Tesque) were used to prepare artificial cells.We used a fluorescently labeled lipid, 1,2-dimyristoyl-sn-glycero-3-phos-phoethanolamine-N-(lissamine rhodamine B sulfonyl) (Rh-PE) (ammoniumsalt; Avanti Polar Lipids) to visualize the lipid membrane. We designed DNAoligomers that assemble to form nanostructures, both with sticky ends

Significance

Although liposomes and lipid droplets have been used fornumerous applications, the fragility of the lipid membranecauses an unintentional collapse, which is problematic for ad-vanced applications. To solve this problem, we constructed anartificial cytoskeleton with DNA nanotechnology (a DNA cyto-skeleton). The DNA cytoskeleton is a DNA network formedunderneath the membrane of positively charged lipids throughelectrostatic interactions without the need for special handling.The DNA cytoskeleton significantly improves mechanical sta-bility and, therefore, confers tolerance against osmotic shockto liposomes like the cytoskeleton in live cells. Because of itsbiocompatibility and the easiness of implementing designchanges, the DNA cytoskeleton could become a tool forgreat stabilizer of liposomes and lipid droplets.

Author contributions: M.T. and M.Y. designed research; C.K., K.F., M.M., I.K., Y.K., A.S.,Y.M., S.-i.M.N., S.M., M.T., and M.Y. performed research; C.K., K.F., M.T., and M.Y. ana-lyzed data; and C.K., K.F., M.T., and M.Y. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. D.L. is a guest editor invited by the EditorialBoard.1To whom correspondence may be addressed. Email: myanagi@cc.tuat.ac.jp or takinoue@c.titech.ac.jp.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1702208114/-/DCSupplemental.

7228–7233 | PNAS | July 11, 2017 | vol. 114 | no. 28 www.pnas.org/cgi/doi/10.1073/pnas.1702208114

Dow

nloa

ded

by g

uest

on

June

25,

202

0

(G1 to G3) and without sticky ends (S1 to S3) (Fig. 1A), and obtained them aslyophilized powders (salt-free grade purification; Eurofins Genomics). Tovisualize DNA with sticky ends (G1 to G3), we replaced 1/120 of G3 with6-carboxy-fluorescein (FAM)-labeled DNA G3, where the FAM was attachedto the 3′ end of G3. In the case of DNA without sticky ends (S1 to S3), wemixed the oligomers with double-stranded DNA (dsDNA)-binding dye SYBR-Green I (Takara Bio), diluted with distilled water to 1:30,000. The designedDNA nanostructures were dissolved in distilled water to a concentration of2 mM and stored at −20 °C until use. All materials were used withoutfurther purification.

Hybridization of DNA in Bulk. DNA oligomers were dissolved in a buffer[20 mM Tris·HCl (pH 8.0) and 350 mM NaCl] at room temperature (∼20 °C).Equimolar solutions of DNA were mixed with dyes (1/120 of total oligo-deoxynucleotide amount) to an approximate concentration of 18 μM. Afterheating at 80 °C for 10 min, the DNA solution was gradually cooled to 10 °Cat a rate of 0.01 °C/s using a thermal cycler (T-Gradient; Biometra). The hy-bridized DNA with sticky ends (G1−G3) has two melting points; G1 toG3 forms a Y motif below Tm1, and the Y-motif units form duplexes at Tm2

(Fig. 1B). On the other hand, DNA without sticky ends (S1 to S3) has only onemelting point at Tm1. In other words, DNA S1−S3 cannot form a networkstructure like DNA G1 to G3. The melting temperatures were estimated to beTm1 = 69 °C and Tm2 = 49 °C, according to previous reports (18).

Preparation of Droplets and Liposomes. Microdroplets covered with a lipidlayer were prepared as reported previously (1, 19). Briefly, dry lipid films withRh-PE (0.1 mol% of total lipid amount) were prepared on the bottom of aglass tube. These lipids were dissolved in mineral oil to a concentration of∼2 mM by ultrasonication for 90 min using a sonicator with an output powerof 90 W (Branson). The solution of hybridized DNA (i.e., DNA microspheresof G1 to G3 or DNA solution of S1 to S3) was dispersed in the lipids-in-oil(∼1/30 of total volume) by pipetting to prepare droplets covered with a lipidlayer. An aliquot containing the droplet-encapsulated DNA was placed on asilicone-coated cover glass (Matsunami Glass) to prevent the droplets fromsticking to the glass plate.

We used a droplet transfer method, as reported previously, to prepare theliposomes (20–22). POPC in mineral oil at a concentration of ∼2 mMwas usedto form the outer membrane. The outer solution of the liposomes was thesame as the inner buffer solution. These procedures allowed us to obtain

relatively abundant and large liposomes. The liposome solution was pouredbetween two glass plates with a 0.5-mm silicone spacer for observation.

Fluorescence Observation of Artificial Cells with DNA. Liposomes and dropletswith DNA were observed by confocal laser scanning fluorescence microscopy(CLSM) (Olympus IX83 with FV1200; Olympus). Fluorescent images of Rh-PEand FAM, which were excited by a mercury lamp or a laser (488 nm and473 nm for DNA and lipid membrane, respectively), were obtained usingfluorescence filter sets (U-FGW and U-FBNA for mercury lamp; 530 nm to550 nm and 470 nm to 495 nm for laser; Olympus). The pinhole size was fixedto be ∼1 μm. The obtained images were analyzed by National Institutes ofHealth Image J software. We evaluated the fluidity of DNA assemblies withfluorescence recovery after photobleaching (FRAP) experiments using CLSM.After photobleaching (bleach radius 1 μm to 3 μm, bleach time 1 s to 2 s forDNA and 10 min for lipid, laser power 100%), fluorescence recovery wasrecorded according to the manufacturer’s instructions (frame rate,120 frame/s) and analyzed to determine the diffusion constant D by usingsoftware (Diffusion measurement package; Olympus) based on a 2D diffu-sion equation, described by Axelrod et al. (23).

Micropipette Aspiration of Artificial Cells. To investigate how the DNA gel,underneath the artificial cell membrane, affects the mechanical properties ofthemembrane,weaspirated the dropletswith theDNAbyusing amicropipettemanipulator system (MMO-202ND andMN-4; Narishige) and microinjector (IM-11-2; Narishige) with a differential pressure transducer (DP15; Validyne), whichwas equippedwith amicroscope (Axiovert 40CFL; Carl Zeiss).We prepared glassmicropipettes with the inner diameter of the tip being ∼5 μm by pulling glasscapillaries (inner diameter 0.5 mm, GC-1; Narishige) with a puller (PC-10;Narishige) and a microforge (MF-900; Narishige).

Results and DiscussionSpontaneous Formation of DNA Spheres in Bulk. As a material forthe artificial cytoskeleton, we designed three sequences of DNAoligomers that assemble to form a Y motif. To ensure the flex-ibility for efficient Y-motif formation, TT sequences withouthybridization pairs were inserted at the center of the Y motif(Fig. 1). We designed Y-motif DNA to have sticky ends toconnect with each other to form a gel-like network structure.Before preparing DNA gel network inside artificial cells, we

observed the hybridization of Y-motif DNAs in a bulk solutionby using CLSM. Fig. 2A shows a typical CLSM image of hy-bridized Y-motif DNA with sticky ends (sequences G1 to G3).Similar to previous reports (14), micrometer-sized fluorescentspheres were observed. Because the fluorescence dye (SYBR-Green I) binds dsDNA, the green fluorescence indicates thatthe microspheres incorporated dsDNA structure. The distribu-tion of the average diameter (d) from the binarized image (Fig.2B) shows the average size is 3.1 μm. In testing hybridized DNAwithout sticky ends (sequences S1 to S3), no microspheres wereobserved (Fig. 2C).

DNA Shell in DOTAP Droplets. To localize DNA structures nearlipid membranes, we used electrostatic interactions betweennegatively charged DNA and the positively charged lipid DOTAP.

Fig. 1. DNA sequences used to form Y-motif DNA. (A) DNA sequences withor without sticky ends, which are marked as orange. Each color (red, green,and blue) shows its complementary sequence. (B) Schematic illustration ofhybridization of DNA with sticky ends (orange) by decreasing the temper-ature (T). The three different DNA oligomers form a Y motif of dsDNA withtheir complementary regions below Tm1, and the Y-motif DNAs form anetwork structure by annealing with their sticky ends below Tm2.

Fig. 2. (A) CLSM image of the microspheres incorporating hybridized DNAoligomers with sticky ends. (Scale bar, 50 μm.) (B) Histogram of the averagediameter of the DNA microspheres. (C) CLSM image of the microspheresincorporating hybridized DNA oligomers without sticky ends. (Scale bar,50 μm.)

Kurokawa et al. PNAS | July 11, 2017 | vol. 114 | no. 28 | 7229

APP

LIED

PHYS

ICAL

SCIENCE

SBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

25,

202

0

First, we encapsulated the solution of the hybridized sticky-endDNA into droplets coated with a DOTAP monolayer. Fig. 3Ashows a typical cross-section image of the DOTAP droplet, wherethe DOTAP membrane and DNA are shown in red and green,respectively. Intensity profiles demonstrate that the DNA micro-spheres spontaneously accumulated near the DOTAP membraneto cover the interior surface of the droplet like a shell (DNA

shell). The DNA shell had a homogeneous thickness of ∼1 μm,which is smaller than the diameter of the DNA microspheres inbulk (∼3 μm, Fig. 2B). We found the thickness of DNA micro-spheres deposited on the DOTAP membrane to be 1 μm to 2 μm,analyzed by atomic force microscopy (AFM) (Fig. S1 and Topologyand Elasticity of DNA Microgels on Lipid Membrane). This value issimilar to the thickness of DNA shell (∼1 μm). If the accumulationof the DNAmicrospheres is due to electrostatic interaction with theDOTAP membrane rather than the DNA gelation, Y-motif DNAswithout sticky ends (S1 to S3) should also accumulate on themembrane. In fact, the sticky-end–less DNA accumulation wasobserved, although the level was lower than that observed whenusing DNA microspheres with sticky ends (Fig. 3A). This differencemay result from the fact that high concentrations of DNA moleculeshave a large entropy and that DNA molecules diffusing in the centerof the lipid droplet are not influenced by the membrane. This DNAshell was not observed for droplets covered with a zwitterionic PCmembrane (Fig. 3C), which is moderately negatively charged in thisbuffer condition (24). The DNA microspheres formed inside thedroplet, not on its surface (Fig. 3C). These results strongly suggestthat the electrostatic interaction between the DNAmicrospheres andthe lipid membrane is useful to construct the DNA shell structureinside the droplet covered with the lipid layer.To confirm adhesion among microgels on the DOTAP mem-

brane, we analyzed the size of the collected DNA shells by cen-trifugation after droplet rapture (Fig. S2 and Collection of DNAShells from Droplets). The collected shells had a larger size than theinitial DNA microspheres in bulk, ∼3 μm. Therefore, we concludedthat the encapsulated DNA microspheres fused with one anotherand spread out on the DOTAP membrane (Figs. S1 and S2).

Fluidity of DNA Shell in DOTAP Droplets. The thin structure of theDNA shell in DOTAP droplets raises concerns that the DNAshell is not in gel phase. To test this possibility, we performedFRAP experiments to analyze the fluidity of the assembled DNAand membrane lipids. In the case of sticky-end DNA shell, thefluorescence intensity did not recover 10 min after photo-bleaching (Fig. 4A, Left), indicating that fluidity of the DNAshell was extremely low. Therefore, we concluded that the DNAshell was in gel phase. However, the fluorescence intensity ofmembrane lipids was recovered soon after the photobleaching,and the estimated diffusion constant (D) was on the order of10 μm2/s (Fig. 4A, Right). This diffusion coefficient is comparableto that of lipids without DNA, as reported previously (25, 26).FRAP experiments for the DNA without sticky ends support

the gel structure of the DNA shell originated from the sticky-endarrangement (Fig. 4B). The fluorescent intensity of the DNAwithout sticky ends recovered soon after photobleaching, in-dicating that the sticky-end–less DNA was not in gel phase. Theestimated D of the DNA without sticky ends was on the order of1 μm2/s (Fig. 4B, Left). The D of lipids of the droplets did notdepend on whether the DNA is in gel phase or not (Fig. 4B,Right), suggesting that the DNA molecules without sticky endsare just localized at the membrane. Taken together, the observedslow fluidity of the DNA gel shell likely originates from the DNAgel shell structure via adhesion at the sticky ends, and the en-capsulated DNA microspheres may connect to each other on themembrane. In addition, similar to the cytoskeleton, the DNA gelshell does not impede lipid diffusion.

Micropipette Aspiration of the Droplets with DNA. We showed thatthe DNA microspheres spontaneously form shell structure in gelphase at the interior surface of DOTAP droplets. To investigatethe degree to which the DNA shell mechanically stabilizes thelipid membrane, interfacial tension at the droplet surface wasevaluated by micropipette aspiration (27–29). As illustrated inFig. 5A, a spherical droplet with a radius R0 (∼30 μm to 50 μm)was aspirated into the micropipette with a radius Rp (∼2.5 μm,

Fig. 3. (A and B) Cross-section images of DOTAP droplets encapsulatingDNA (A) with and (B) without sticky ends. The DNA and lipid membraneare shown in green and red, respectively. Fluorescence intensity profilesalong the broken lines are shown. (Scale bars, 20 μm.) (C ) Cross-sectionimage of PC droplets after the encapsulation of DNA microspheres withsticky ends. Fluorescence intensity profile along the broken line is shown.(Scale bar, 30 μm.) a.u., arbitrary unit.

7230 | www.pnas.org/cgi/doi/10.1073/pnas.1702208114 Kurokawa et al.

Dow

nloa

ded

by g

uest

on

June

25,

202

0

Rp << R0) under an aspiration pressure (ΔP). When the aspi-ration length (ΔL) is small enough (ΔL/Rp < 5), the ΔL/Rp in-creases linearly with ΔP (Fig. 5B). Above a critical pressure(ΔP*), the droplet collapses and flows into the micropipette. Thevalue of ΔP* for droplets with sticky-end DNA shell was greaterthan 10 kPa, which was much higher than that of droplets withsticky-end–less DNA and droplets without DNA, which were lessthan 0.2 kPa. Under the condition ΔL/Rp = 1, the surface tension(γ) of the droplet is derived from the Young−Laplace law,

ΔP= 2γ�1�Rp − 1=R

�,

where R is radius of the droplet. The surface tension of dropletswith the DNA gel shell (γG) was 4 ± 2 mN/m, which was about4 times greater than that of droplets with sticky-end–less DNA inthe sol phase (S1 to S3) (γS = 1 ± 0.5 mN/m), and of dropletswithout DNA (γ0 = 1 ± 0.7 mN/m). A control experimentshowed that the effect of DOTAP adsorption on the glass cap-illary was negligible for the measurement. The values of γS andγ0 were similar to the values of PC droplets without DNA mea-sured by the Wilhelmy plate method (19) and the micropipetteaspiration method (27). These values are summarized in Table 1.Accordingly, we derived the shear modulus (μ) and the areaexpansion modulus (dilation elasticity, κ) based on the followingequations (30, 31) under the condition of 1 < ΔL/Rp < 5 (linearresponse region in Fig. 5B), respectively:

ΔP= μ�Rp

�2L

�Rp − 1+ log

�2L

�Rp

��

k= γðA0=ΔAÞ,

where A0 is the initial surface area of spherical droplets with R0,and ΔA is the increase in surface area from the aspiration. Theobtained κ values and μ values of droplets with the DNA gel shellwere higher than those of droplets with sticky-end–less DNA insol phase and without DNA, as summarized in Table 1. Theseresults mean that the DNA shell mechanically supports the lipidmembrane. In addition, the ratio κ/μ is much greater than 1,

which means that the DNA gel shell droplets easily deform withshearing force, but are not likely to change surface area.

Stability of Liposomes with the DNA Gel Shells. By using the droplettransfer method, we created liposomes with a DNA gel shell.Zwitterionic PC lipid was used for the outer leaflet of the lipo-some because positively charged DOTAP strongly adsorbs onglass (15). Fig. 6A clearly shows successfully prepared asym-metric liposome with the DNA gel shells (green), because thefluorescent-labeled lipids were only contained in the PC layer,that is, the outer membrane was maintained. To analyze stabilityof the liposomes with the DNA gel shell, an osmotic stress wasapplied by addition of hypertonic sorbitol as an osmolyte. Thehypertonic treatment resulted in collapse of almost all liposomescontaining sticky-end–less DNA or without DNA (Fig. 6B, cen-ter and right). However, most liposomes with DNA gel shellsurvived and retained their spherical shapes after the hypertonictreatment (Fig. 6B, left). In addition, the DNase did not pene-trate the membrane (Fig. S3), suggesting that the liposomemembrane is sufficiently stable to isolate its inner solution fromthe surrounding solution. When a much higher concentration ofsorbitol was added (2 times higher pressure), micrometer-sizedpores appeared on the liposome membrane and on the DNAshell at the same position (Fig. 6C). The pores maintained theirsize, and the membrane did not rupture. This result suggests thatthe liposome membrane and the DNA shell are strongly boundto each other, as illustrated in Fig. 6C, Far Right. In general,osmotically stressed liposomes cause strong fluctuations in themembrane and result in membrane rupture (32). Encapsulationof polymer gels inside the liposomes prevents membrane rap-ture, but results in liposome shrinkage and inhibits release of theencapsulated materials (33). These results strongly suggest thatthe DNA shell stabilizes liposomes. Therefore, the presence ofthe DNA gel shell underneath the liposomal membranemechanically strengthens the membrane structure and preventsthe membrane rupture, similar to the cytoskeleton in live cells.To compare the mechanical strength of the DNA shell with that

of polymer gels and with that of cells with elastic modulus rangingfrom 100 Pa to 10 kPa (34), we measured the elasticity of DNAmicrogels deposited on the DOTAP membrane by AFM (Fig.S1C). The average elasticity of the gels was 155 ± 28 kPa (average ±SE; n = 7). This degree of elasticity is comparable to that of 10 wt%polyacrylamide (35, 36) and 2 mol/L agarose (37) gels.

Estimation of Strength of DNA Cytoskeleton from the Aspect ofEnthalpy of DNA Gel Shell. Enthalpy is a major factor in de-termining the strength of the DNA cytoskeleton. When the as-piration length increased up to the critical value ΔLc (>Rp),DOTAP droplets with the DNA cytoskeleton collapsed and

Fig. 4. FRAP curves of DNA shell (green circle) and lipids (red square).(A) DOTAP droplet with shell structure of DNA with sticky ends. The twoimages show before and after photobleaching. (Scale bar, 20 μm.) (B) DOTAPdroplet encapsulating DNA without sticky ends. a.u., arbitrary unit.

Fig. 5. Micropipette aspiration of a droplet. (A) Sketch of an aspirated dropletinto a micropipette with an inner radius Rp under an aspiration pressure ΔP.(B) Relationship between the ΔP and the normalized aspiration length ΔL/Rp.

Kurokawa et al. PNAS | July 11, 2017 | vol. 114 | no. 28 | 7231

APP

LIED

PHYS

ICAL

SCIENCE

SBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

25,

202

0

flowed into the pipette. Experimentally, the value of ΔLc was∼7.5 μm for droplets with R = 20 μm. We assume that a part ofDNA gel shell collapsed into Y-motif DNA by detaching thesticky ends. In other words, an increase in the interfacial energy ofthe DNA cytoskeleton with increasing surface area (ΔEDNA) ishypothesized to be equal to enthalpy of the sticky ends of thecollapsed DNA shell (ΔHDNA), that is, ΔEDNA = ΔHDNA. Whenthe droplets aspirate into the micropipette with the aspirated lengthΔLc (as illustrated in Fig. 5A), the interfacial energy of DNA shellincreases according to the equation

ΔEDNA = ðγ0 − γGÞΔA≈ ðγ0 − γGÞ�2πRpΔLc

�,

where γG and γ0 are interfacial tensions of droplets with andwithout DNA, respectively, and γG = 4 mN/m and γ0 = 1 mN/m.The increase in the interfacial energy (ΔEDNA) was estimated tobe 3.5 × 10─13 J from the experimentally obtained values Rp =2.5 μm and ΔLc = 7.5 μm. The basic structure of the DNA gelnetwork is a six-member ring (Fig. 7). Therefore, one DNA ringwould collapse with energy equal to triple the enthalpy of thesticky end, that is, 3ΔHsticky end. When a number (Nring) of DNArings collapse into the Y motif, the total enthalpy of the DNA ringsis expressed as

ΔHDNA = 3NringΔHsticky  end.

From a previous study (18), the ΔHsticky end is reported to be 12 ×10─19 J. When the ΔEDNA was equal to the ΔHDNA, the numberof DNA rings (Nring) would be 2.5 × 105.

To test the validity of these calculations based on the estimatednumber of DNA rings, we derived the number of DNA rings(Nring) geometrically from a slide of the DNA ring, a = 14.28 nm(G1 to G3 are 42 base pairs) and diameter of the dsDNA, h =2 nm. The minimum volume of the DNA rings occupied (Vring) forNring = 105 is estimated as

Vring =Nring · h · ð2aÞ2 ≈ 0.4  μm3.

Because the thickness of DNA shell (d) was ∼1 μm (Fig. 3A), theincreased ΔEDNA can collapse a part of the DNA gel networkwith an area of 0.4 μm2. If energy is used not only to collapse thedouble helical structure of DNA but also to resolve the entan-glement and the electrostatic interaction among DNA rings, thesize of the collapsed area will be smaller. In other words, theappearance of such a small pore on the DNA gel network mightlead the droplet flow into the pipette above the critical length(ΔLc). This phenomenon corresponds to the fact that the DNAshell has a small shear modulus (μ) and a large area expansionmodulus (κ), i.e., κ/μ >> 1 (Table 1), which is similar to cells withcytoskeletons, such as red blood cells and white cells (30, 31, 38).

ConclusionWe successfully prepared cell-sized liposomes and lipid dropletswith an artificial cytoskeleton by using DNA nanostructures. TheDNA oligomers formed Y motifs and connected to each otherwith their sticky ends to form network structure in gel phase. Thehybridized DNA formed microsphere gels in bulk. By encapsu-lating the DNA microgels into droplets coated by a cationic lipidmembrane, DNA spontaneously formed a shell structure un-derneath the membrane via electrostatic interactions. The DNAgel network increased interfacial tension and area expansionmodulus of the droplet surface to a degree similar to the en-thalpy of the sticky ends of the DNA. Such drastic changes weredetected only with DNA skeletons in the gel phase. Further-more, we demonstrate that liposomes with the DNA skeleton aresubstantially stabilized and tolerant against osmotic stress. Theseresults clearly demonstrate that the DNA gel skeleton is a sta-bilizer of the artificial cells. Yang et al. (39) reported DNAnanorings as templates to prepare nanometer-sized liposomesinside. Meanwhile, our DNA cytoskeleton mechanically sup-ported the micrometer-sized liposomes from the inside. Thus,our method will contribute to the extension of DNA−liposomehybrid systems.The DNA gel network structures on inner membranes show

highly similar traits to cytoskeletons of live cells. Thus, the DNAgel shell can be characterized as a cytoskeleton for artificial cells,including liposomes and lipid droplets. The use of a DNA cy-toskeleton has many remarkable merits. (i) DNA has excellentbiocompatibility and biodegradability, which is important for DDS.We confirmed that the DNA gels have little effect on viability of

Table 1. Comparison of observed mechanical parameters ofdroplets with/without DNA gel

Droplet typesInterfacial

tension, mN/mArea expansionmodulus, mN/m

Shearmodulus, mN/m

Sticky-end DNA inDOTAP droplets

γG = 4 ± 2 κG = 640 ± 390 μG = 0.8 ± 0.1

Sticky-end–less DNAin DOTAP droplets

γS = 1 ± 0.5 κS < 220 μS < 0.5

DOTAP dropletswithout DNA

γ0 = 1 ± 0.7 κ0 < 15 μ0 < 0.4

Fig. 6. (A and C) Liposomes with DNA (A) before and (C) after the osmoticshock treatment. Confocal images show (Left) DNA (green), (Center) lipids ofouter leaflet (red), and (Right) the merged images. Schematic illustrations of theconfocal images are shown together (Far Right). Arrows in C indicate location ofthe pores. (Scale bars, 10 μm.) (B) Survival rates after osmotic treatment.

Fig. 7. Schematic images of DOTAP droplet and the DNA gel network.

7232 | www.pnas.org/cgi/doi/10.1073/pnas.1702208114 Kurokawa et al.

Dow

nloa

ded

by g

uest

on

June

25,

202

0

living cells (Fig. S4 and Biocompatibility), and that they are de-gradable by the addition of DNase solution (Fig. S2C). In addition,previous studies reported that DOTAP also do not show harmfuleffects on live cells (40–42). (ii) The DNA nanostructure is easy toassemble and dissemble by temperature changes and illumination viaDNA sequence design (43). (iii) The mechanical properties of DNAgels are also controllable via DNA sequence design by changing thenumber and length of arms (sticky ends) in DNA gels (11), which isdifficult to achieve with biopolymer gels. (iv) DNA has versatilepotential by combining logic gates (44), actuators (45), and switches(46), which may also be applicable for our artificial cell systems.(v) Because DNA gels are not crystallized, as reported by Rovigattiet al. (47), the DNA gels can rearrange their structures in response

to their surrounding circumstances. Furthermore, our proposed li-posomes with DNA cytoskeletons are much closer to the living cellscompared with the conventional liposomes without skeletal struc-tures. Therefore, liposomes with a DNA cytoskeleton have the po-tential to become an ideal platform to understand how cells regulatetheir shapes with cytoskeleton from physicochemical point of view.

ACKNOWLEDGMENTS. This work was supported by Japan Society for thePromotion of Science (JSPS) KAKENHI Grants 16812285 and 14J10002 (to M.M.);24104005, 22220001, and 15H01715 (to S.M.); 24104002 and 26280097 (to M.T.);and 15H05463, 15KT0081, 16H00796, and N16H01443 (to M.Y.); and by JapanAgency for Medical Research and Development-Core Research for EvolutionalScience and Technology (AMED-CREST), Japan Science and Technology Agency(JST) Grant 16gm0810001h0102 (to S.-i.M.N.).

1. Yanagisawa M, Sakaue T, Yoshikawa K (2014) Characteristic behavior of crowdingmacromolecules confined in cell-sized droplets. Int Rev Cell Mol Biol 307:175–204.

2. Noireaux V, Maeda YT, Libchaber A (2011) Development of an artificial cell, from self-organization to computation and self-reproduction. Proc Natl Acad Sci USA 108:3473–3480.

3. Stano P, Luisi PL (2013) Semi-synthetic minimal cells: Origin and recent developments.Curr Opin Biotechnol 24:633–638.

4. Lian T, Ho RJ (2001) Trends and developments in liposome drug delivery systems.J Pharm Sci 90:667–680.

5. Noireaux V, et al. (2000) Growing an actin gel on spherical surfaces. Biophys J 78:1643–1654.

6. Janmey PA, et al. (1994) The mechanical properties of actin gels. Elastic modulus andfilament motions. J Biol Chem 269:32503–32513.

7. Takiguchi K, Yamada A, Negishi M, Tanaka-Takiguchi Y, Yoshikawa K (2008) En-trapping desired amounts of actin filaments and molecular motor proteins in giantliposomes. Langmuir 24:11323–11326.

8. Miyazaki M, Chiba M, Eguchi H, Ohki T, Ishiwata S (2015) Cell-sized spherical con-finement induces the spontaneous formation of contractile actomyosin rings in vitro.Nat Cell Biol 17:480–489.

9. Takiguchi K, et al. (2009) Chapter 3 - Construction of cell-sized liposomes encapsu-lating actin and actin-cross-linking proteins. Methods Enzymol 464:31–53.

10. Pontani LL, et al. (2009) Reconstitution of an actin cortex inside a liposome. Biophys J96:192–198.

11. Roh YH, Ruiz RC, Peng S, Lee JB, Luo D (2011) Engineering DNA-based functionalmaterials. Chem Soc Rev 40:5730–5744.

12. Ma S, et al. (2012) Fabrication of microgel particles with complex shape via selectivepolymerization of aqueous two-phase systems. Small 8:2356–2360.

13. Liu J (2011) Oligonucleotide-functionalized hydrogels as stimuli responsive materialsand biosensors. Soft Matter 7:6757–6767.

14. Matsuura K, Masumoto K, Igami Y, Fujioka T, Kimizuka N (2007) In situ observation ofspherical DNA assembly in water and the controlled release of bound dyes.Biomacromolecules 8:2726–2732.

15. Helwa Y, Dave N, Liu J (2013) Electrostatically directed liposome adsorption, in-ternalization and fusion on hydrogel microparticles. Soft Matter 9:6151–6158.

16. Zadeh JN, et al. (2011) NUPACK: Analysis and design of nucleic acid systems. J ComputChem 32:170–173.

17. Douglas SM, et al. (2009) Rapid prototyping of 3D DNA-origami shapes with caDNAno.Nucleic Acids Res 37:5001–5006.

18. SantaLucia J, Jr (1998) A unified view of polymer, dumbbell, and oligonucleotide DNAnearest-neighbor thermodynamics. Proc Natl Acad Sci USA 95:1460–1465.

19. Yanagisawa M, Yoshida T, Furuta M, Nakata S, Tokita M (2013) Adhesive force be-tween paired microdroplets coated with lipid monolayers. Soft Matter 9:5891–5897.

20. Yanagisawa M, Iwamoto M, Kato A, Yoshikawa K, Oiki S (2011) Oriented re-constitution of a membrane protein in a giant unilamellar vesicle: Experimentalverification with the potassium channel KcsA. J Am Chem Soc 133:11774–11779.

21. Fujiwara K, Yanagisawa M (2014) Generation of giant unilamellar liposomes con-taining biomacromolecules at physiological intracellular concentrations using hy-pertonic conditions. ACS Synth Biol 3:870–874.

22. Morita T, Narita T, Mukai S, Yanagisawa M, Tokita M (2013) Phase behaviors ofagarose gel. AIP Adv 3:042128.

23. Axelrod D, Koppel DE, Schlessinger J, Elson E, Webb WW (1976) Mobility measure-ment by analysis of fluorescence photobleaching recovery kinetics. Biophys J 16:1055–1069.

24. Egawa H, Furusawa K (1999) Liposome adhesion on mica surface studied by atomicforce microscopy. Langmuir 15:1660–1666.

25. Negishi M, Sakaue T, Yoshikawa K (2010) Mismatch of bulk viscosity reduces in-terfacial diffusivity at an aqueous-oil system. Phys Rev E Stat Nonlin Soft Matter Phys81:020901.

26. Walder RB, Honciuc A, Schwartz DK (2010) Phospholipid diffusion at the oil-waterinterface. J Phys Chem B 114:11484–11488.

27. Delacotte J, Gourier C, Pincet F (2014) Interfacial pressure and phospholipid density atemulsion droplet interface using fluorescence microscopy. Colloids Surf B Biointerfaces117:545–548.

28. Tsamantakis C, Masliyah J, Yeung A, Gentzis T (2005) Investigation of the interfacialproperties of water-in-diluted-bitumen emulsions using micropipette techniques.J Colloid Interface Sci 284:176–183.

29. Yeung A, Dabros T, Masliyah J, Czarnecki J (2000) Micropipette: A new technique inemulsion research. Colloids Surf A Physicochem Eng Asp 174:169–181.

30. Merkel R, et al. (2000) A micromechanic study of cell polarity and plasma membranecell body coupling in Dictyostelium. Biophys J 79:707–719.

31. Lelievre J, Bucherer C, Geiger S, Lacombe C, Vereycken V (1995) Blood cell bio-mechanics evaluated by the single-cell micromanipulation. J Phys III 5:1689–1706.

32. Ohno M, Hamada T, Takiguchi K, Homma M (2009) Dynamic behavior of giant lipo-somes at desired osmotic pressures. Langmuir 25:11680–11685.

33. Campillo C, Pepin-Donat B, Viallat A (2007) Responsive viscoelastic giant lipid vesi-cles filled with a poly(N-isopropylacrylamide) artificial cytoskeleton. Soft Matter3:1421–1427.

34. Levental I, Georges PC, Janmey PA (2007) Soft biological materials and their impact oncell function. Soft Matter 3:299–306.

35. Wyss HM, Franke T, Mele E, Weitz DA (2010) Capillary micromechanics: Measuring theelasticity of microscopic soft objects. Soft Matter 6:4550–4555.

36. Denisin AK, Pruitt BL (2016) Tuning the range of polyacrylamide gel stiffness formechanobiology applications. ACS Appl Mater Interfaces 8:21893–21902.

37. Normand V, Lootens DL, Amici E, Plucknett KP, Aymard P (2000) New insight intoagarose gel mechanical properties. Biomacromolecules 1:730–738.

38. Lenormand G, Hénon S, Richert A, Siméon J, Gallet F (2001) Direct measurement ofthe area expansion and shear moduli of the human red blood cell membrane skel-eton. Biophys J 81:43–56.

39. Yang Y, et al. (2016) Self-assembly of size-controlled liposomes on DNA nano-templates. Nat Chem 8:476–483.

40. Xu Y, Szoka FC, Jr (1996) Mechanism of DNA release from cationic liposome/DNAcomplexes used in cell transfection. Biochemistry 35:5616–5623.

41. Porteous DJ, et al. (1997) Evidence for safety and efficacy of DOTAP cationic liposomemediated CFTR gene transfer to the nasal epithelium of patients with cystic fibrosis.Gene Ther 4:210–218.

42. Wrobel I, Collins D (1995) Fusion of cationic liposomes with mammalian cells occursafter endocytosis. Biochim Biophys Acta Biomembranes 1235:296–304.

43. Bomboi F, et al. (2016) Re-entrant DNA gels. Nat Commun 7:13191.44. Takinoue M, Kiga D, Shohda K, Suyama A (2008) Experiments and simulation models

of a basic computation element of an autonomous molecular computing system. PhysRev E Stat Nonlin Soft Matter Phys 78:041921.

45. Simmel FC, Yurke B (2001) Using DNA to construct and power a nanoactuator. PhysRev E Stat Nonlin Soft Matter Phys 63:041913.

46. Wang F, Liu X, Willner I (2015) DNA switches: from principles to applications. AngewChem Int Ed Engl 54:1098–1129.

47. Rovigatti L, Smallenburg F, Romano F, Sciortino F (2014) Gels of DNA nanostars nevercrystallize. ACS Nano 8:3567–3574.

48. Radmacher M, Fritz M, Hansma PK (1995) Imaging soft samples with the atomic forcemicroscope: Gelatin in water and propanol. Biophys J 69:264–270.

49. Sheppard S, Sweet S, Benedict AJ (1922) Elasticity of purified gelatin jellies as afunction of hydrogen-ion concentration. J Am Chem Soc 44:1857–1866.

50. Boudou T, Ohayon J, Picart C, Tracqui P (2006) An extended relationship for the char-acterization of Young’s modulus and Poisson’s ratio of tunable polyacrylamide gels.Biorheology 43:721–728.

Kurokawa et al. PNAS | July 11, 2017 | vol. 114 | no. 28 | 7233

APP

LIED

PHYS

ICAL

SCIENCE

SBIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

June

25,

202

0