DNA micelle flares: a study of the basic properties that contribute to enhanced stability and binding affinity in complex biological systems

DMFs are spherical DNA–diacyllipid nanostructures formed by hydrophobic effects between lipid tails coupled to single-stranded DNAs.


Introduction
The hydrophobic effect of DNA-related amphiphiles has recently attracted considerable attention in the areas of bioanalysis and biomedicine. Spherical structures, such as spherical nucleic acids (SNAs) and DNA block copolymers (DBCs) are signicant forerunners in the detection and treatment of cancer cells, and, more importantly, the generation of many other innovative structures. DBCs are oligonucleotides coupled with hydrophobic polymers that can self-assemble into threedimensional micelle nanostructures. Compared to other biopolymer analogues, DBCs have such advantages as nucleotide precision and controllable supramolecular structure, leading to applications ranging from biomarker discovery to drug delivery. 1,2 In aqueous solution, DBCs self-assemble as a result of microphase separation, while a reversible hydrophobic effect keeps the polymeric chains together. 3 This uniquely dynamic micelle structure has proven particularly useful as both carrier for anticancer drugs and scaffold for specic chemical reactions. 4,5 For example, in the DBC micellebased drug delivery system, a hydrophobic anticancer drug was loaded into the DBC micelle core, and further functionalization was achieved through folic acid-modied DNA strands complementary to DNA sequences on the surface of the DBC micelles. 6 Despite many successful applications, DBCs still have such disadvantages as complicated preparation and high critical micelle concentrations (CMC). To simplify preparation and achieve maximum assembly efficiency, our group reported a DNA micelle are nanostructure consisting of a diacyllipid core and a single-stranded DNA corona. 7 DMFs share many unique properties with DBCs, including the ability to selfassemble through the hydrophobic effect and carry oligonucleotides outside, as well as the capacity to be doped with certain hydrophobic dyes or drugs. However, DMFs can be further characterized by their outstanding cell permeability, which is facilitated by the similarity between the intermolecular forces of the diacyllipid in DNA micelle ares and those of the dynamic phospholipid bilayers in cell membranes, both of which have hydrophobic and hydrophilic portions. Here, strong van der Waals forces result in efficient transport of antisense oligonucleotides and drug molecules to live cells. [8][9][10] For example, hairpin-shaped DMFs successfully entered living cells and emitted uorescent signals upon binding to specic target molecules as a result of opening of the hairpin and increasing the distance between the uorophore and quencher (Fig. S2 †). 11 When incorporated with a targeting aptamer, DNA micelle ares can selectively recognize specic types of cancer cells. 7 Thus, with their combined cell internalization and targeting capabilities, aptamer-micelles can nd many applications in biomedicine, including drug delivery via endocytosis and gene therapy through the inhibition of cancer-related mRNA expression. 11,12 To undertake a study of the basic properties associated with DNA micelle ares, one may look to the pioneering work of Mirkin et al. for guidance. For the sensitive detection of mRNA in living cells, nanoares (NFs) "exhibit high signaling, have low background uorescence, and are sensitive to changes in the number of RNA transcripts present in cells". 13 Similar to DMFs, nanoares present densely functionalized DNA on the surface of gold nanoparticles (NPs). NFs also have signicant chemical and physical advantages that are distinct from DNA and NPs, enabling them to act as scaffolding for the selfassembly of oligonucleotides. The gold core of nanoares can be dissolved aer fabrication to make hollow spherical nucleic acids (SNAs). These SNAs show stronger complementary nucleic acid hybridization and more efficient cell membrane transfection than NFs having gold cores. 14 Moreover, the properties of NFs have been systematically investigated, including measurement of melting transition, binding strength and biostability. The results have inspired the innovative design of many bioprogrammable nanostructures that have become potential tools in biomedicine and areas related to energy conversion.
Inspired by the extensive studies of NFs, the present work investigated the properties of DNA micelle ares that lead to enhanced stability and binding ability. These structures are comprised of either plain ssDNA or hairpin-shaped (molecular beacon) segments. In particular, molecular beacon micelle ares (MBMFs), which are easily prepared through diacyllipid phosphoramidite chemistry using an automated DNA synthesizer, undergo a signicant uorescence signal change when binding to targets. 12 In the absence of target cDNA, MBMFs exhibit very low uorescence because the uorophore and quencher are spatially close. However, upon binding to target DNA, uorophore and quencher are separated, resulting in the restoration of uorescence signal (Scheme 1). In this work, studies of complementary binding assays and biostability are based on this concept. In complex biological media, such as nuclease-containing systems or cell lysate, this structure offers resistance towards enzymatic digestion and maintains its size, suggesting the utility of this tool in cell detection, gene silencing and drug delivery.

Compositional study
MBMFs must rst hybridize with mRNA as a precondition for use in gene therapy. The probe then acts either by blocking translation of the targeted mRNA or by forming a DNA/RNA hybrid with the target mRNA. Either mRNA or the DNA/RNA hybrid can be degraded by the enzyme RNase H. Using these mechanisms, MBMFs can be used for imaging guided gene therapy. 15 Therefore, as a potential tool for cancer gene therapy, increasing the DNA loading capacity of micelle ares would greatly improve therapeutic effect based on the interaction between micelle ares and intracellular cancer-related mRNAs. As such, an essential clinical parameter is aggregation number (N agg ), i.e., the number of monomers present in a micelle are. Previous research has shown that the sizes and aggregation behavior of polymers can be measured with dynamic light scattering (DLS) and static light scattering (SLS), respectively, 1,15 including measurements related to diffusion coefficients (D), hydrodynamic radius (R h ) and aggregation phenomena of molecules in aqueous solution. Therefore DLS/SLS offers the most straightforward technique for the study of intramicelle interactions in aqueous solution. 16 In this study, DMFs with 60-thymine base oligonucleotide sequences and a diacyllipid tail (lipo-T 60 ) were characterized. First, the number-average apparent molecular weight (M n,RI ) of a micelle monomer, diacyllipid-T 60 conjugate, was determined using polyethylene glycol (PEG) standards in aqueous size exclusion chromatography (SEC). The observed M n,RI was 1.25 Â 10 4 g mol À1 , which was lower than the theoretical molecular weight (M n,theory ) of 18 941 g mol À1 . This low molecular weight value can be attributed to the highly compact DNA self-assembled nanostructure, leading to a small hydrodynamic radius. For better accuracy, we chose the observed M n,RI for further calculation, and multi-angle light scattering (MALS) was used to measure the scattered light intensity at various detection angles. ASTRA 6 soware was used to evaluate the absolute Scheme 1 Illustration of a molecular beacon micelle flare nanostructure. Hairpin-shaped DNA-diacyllipid conjugates self-assemble into a spherical micelle flare nanostructure, in which the hairpinshaped DNA corona can lead to an ON/OFF transition upon target binding, temperature change or degradation. molecular weight of the molecule (Fig. S1 †). The absolute number-average molecular weight (M n,SLS ) for a micelle in a 20/80 mixture of acetonitrile/0.05 M Na 2 SO 4 was found to be 6.8 Â 10 6 g mol À1 , according to SLS-SEC measurements. This result conrmed that the proposed architecture results from a large number of self-assembled DNA-diacyllipid amphiphiles.
Next, the average aggregation number for the DNA micelles was calculated from the measured molecular weights according to eqn (1) to give 6.8 Â 10 6 /1.25 Â 10 4 ¼ 544 for lipo-T 60 . This number corresponds to the maximum DNA loading of DNA nanoares with minimal addition of linkers and ions. 17 To analyze the biostability of DNA micelle ares, a key parameter is the critical micelle concentration, which is dened as the concentration of monomers above which DNA micelle ares form and all additional monomers added to the system go to micelles. Here, a unique uorophore, pyrene, was incorporated between the DNA corona and the lipid core (Fig. 1a). A pyrene unit is oen used to probe aggregation behavior in various conditions based on its special uorescence characteristics. 18 Specically, when monomers aggregate, multiple pyrene molecules will cluster together and generate an excimertype uorescence signal with larger Stokes shi. 19 To determine the CMC of DNA micelle ares, three different lengths of DNA strands with 20, 40 and 60 thymine bases (lipo-T 20 , lipo-T 40 , and lipo-T 60 ) were synthesized and modied with pyrene and diacyllipid, respectively. The micelle ares were then incubated in phosphate-buffered saline (PBS) and characterized using steady-state uorescence spectroscopy. As shown in Fig. 1b, the lipo-T 60 micelle ares had a very low CMC value (about 10 nM). The uorescence intensities have also been plotted against the concentration of DMFs and a linear t of the signal has been observed (Fig. S4c †). Further study of lipo-T 20 and lipo-T 40 showed CMC values similar to that of lipo-T 60 , indicating that the formation of DNA micelle ares is sizeindependent ( Fig. S4a and b †). 1 The CMC is a key parameter characterizing surfactants and, hence, is unrelated to hydrophilic DNA length based on the identical diacyllipid core, which accounts for the aggregative nature of DNA micelle ares. Nevertheless, this low CMC does suggest the greater stability of DMFs, even in diluted solution or in vivo environments, when used for cancer treatment.

Enhanced binding affinity of DNA micelle ares
The molecular recognition of nucleic acids to target DNA/RNA provides sensitive detection and promotes gene silencing, both useful in cancer theranostics. As such, the binding affinity of DMFs to target was compared to that of the DNA probe without diacyllipid conjugation. In this experiment, TAMRA (carboxytetramethylrhodamine) and Dabcyl (dimethylaminoazobenzenesulfonic acid) were chosen as uorophore and quencher, respectively, and coupled on the ends of the hairpin-shaped DNA. To evaluate the dissociation constant (K d ) of MBMFs, we compared the binding affinity of MBMFs and molecular beacons (MBs) with the same target, the DNA analogue of partial c-raf-1 mRNA, a cancer biomarker and antisense therapeutic target. The uorescence titration assay was used to quantify the dissociation constant with increased concentrations of target DNA. Opening of the hairpin structure resulted in the separation of TAMRA and Dabcyl, which, in turn, caused a change in peak intensity, thus forming the basis of our measurement. GraphPad Prism 5.0 soware was used to construct a binding isotherm, and K d was calculated by analyzing the binding curve according to eqn (2) 20 where B max stands for the binding affinity at maximum uorescence intensity. As shown in Fig. 2, the K d s of MBMFs and MBs were calculated to be 52 and 95 nM, respectively, demonstrating that MBMFs have stronger binding affinities with their cDNA compared to MBs. This phenomenon can be explained by the density of DNA oligonucleotides on the surface of DNA micelles, which makes it difficult for cDNA to dissociate from the probes. This will greatly lower the cellular detection limit in that it gives a considerable level of signal, even for mRNA with low-expression level, thus improving the efficiency of gene therapy when antisense DNA micelles are being used. The addition of heat to the hybridization complex between molecular beacons and their target cDNAs breaks the hydrogen bonds between base pairs, resulting in dissociation of the complexes and restoration of the hairpin structure. Thus, a decrease of uorescence signal is observed by the resumption of the quenching effect, as shown in Scheme 1, when the system reaches the melting temperature (T m ). Here, the melting temperatures were quantied using a real-time polymerase chain reaction (qPCR) instrument that can monitor the unfolding procedure of the hairpin structures for each 1 C increment by labeling each DNA probe with a uorophore and a quencher. In this assay, MBMF-cDNA and MB-cDNA at the same concentration were heated, followed by monitoring the uorescence intensity. 21 The temperature range of 60 to 80 C was chosen according to the theoretical T m of oligonucleotide hybridization, 69.7 C.
Compared to MBs, MBMFs exhibited a sharper transition and slightly higher melting temperature, as shown in Fig. 3. The sharpness most likely results from the cooperative binding of oligonucleotides, while the higher melting temperature occurs as a result of enhanced interparticle connections endowed by the higher density of oligonucleotides on the surface of micelle ares. 13 Since MBMFs can reach greater effective concentration, they can give a dramatically decreased uorescence signal. By observing the change of uorescence in both situations and determining the point where the slope reached a peak value, the T m s were calculated to be 69 C for MBMFs and 68 C for MBs. Like nanoares, DNA micelle ares exhibit sharper melting temperature increases with their cDNA targets compared to MBs, which implies an enhanced binding affinity and excellent thermal stability (Fig. S5 †). 21 Biostability of DNA micelle ares in biological environments As a drug carrier, DMFs must maintain stability in intracellular environments containing various nucleases. Therefore, the nuclease resistance of MBMFs was evaluated relative to that of MBs for two nucleases, deoxyribonuclease I (DNase I) and exonuclease III (Exo III), by monitoring the uorescence signal change upon the addition of each enzyme. DNase I cleaves DNA in a nonspecic manner, and Exo III primarily recognizes blunt ends or recessed 3 0 hydroxyls and digests towards the 5 0 -end. 22,23 Normally, in an aqueous solution, such as PBS, neither MBMFs nor MBs will have a uorescence change according to the incubation time ( Fig. 4a and b). If the hairpin structure on the surface of the micelle are is cleaved, no quenching between TAMRA and Dabcyl will occur, leading to an increased uorescence signal. 24 As shown in Fig. 4, the intensities were set to 1 aer initial addition of nucleases. In contrast to the MBs, the MBMFs exhibited signicantly less increase in uorescence intensity, indicating the enhanced resistance of MBMFs towards both DNase I and Exo III. This phenomenon agrees well with the comparison of nanoares and MBs, where nanoares showed a smaller degradation rate. 13 Initial degradation rates for three concentrations of DNase I are presented in Fig. 4e, showing much lower rates for MBMFs compared to MBs. Thus, MBMFs can maintain the integrity of their nanostructure when treated with nucleases. In addition, the antinuclease property of DMFs was also comparable to that of NFs under the same conditions. 13 To further conrm that  DMFs can maintain their assembled structures, as a key indicator of stability, assays were performed in CCRF-CEM (human T cell lymphoblast-like cell line) cell lysate. Dynamic light scattering data for MBMFs in PBS before and aer addition of CEM lysate indicated a hydrodynamic radius of 20 nm, which was in accordance with the length of DNA sequence in micelle are nanostructures. By comparison to Fig. 4f, it can also be seen that MBMFs presented much higher resistance towards enzymatic digestion. These results clearly showed that DMFs can also maintain their assembled structures in cell lysate, providing an opportunity for their application in cellular and even in vivo environments, since DMFs can permeate cell membranes and interact with biomolecules inside cells. 11 DMFs, as applied to gene recognition and endocytosis in biological environments As envisioned by the Tan group, it was found that "the hybridization of MBMFs to the target mRNA can specically inhibit gene expression through different mechanisms, including translational arrest by steric hindrance of ribosomal activity and the induction of RNase H endonuclease activity leading to the suppression of cancer cell growth," thus establishing the function of DMFs in gene therapy. However, while MBMFs show strong resistance towards nuclease degradation when compared to MBs, as demonstrated in the present paper, biostability relative to their real-time use in cancer cell recognition and gene therapy has not yet been demonstrated. Therefore, another experiment was designed to conrm the enhanced mRNA binding behavior of MBMFs, and aer MBMFs and MBs were individually incubated in CEM lysate for 2 hours, their mRNA binding affinity could be evaluated. Since binding events, which can cause the unfolding of the hairpin structures, may occur only with the addition of cDNA solution, uorescence spectroscopy is a dependable method for comparing the binding capacity of MBMFs and MBs. A higher uorescence signal would suggest an enhanced binding affinity resulting from higher concentration of intact structures that were not digested during the 2 hour incubation time.
As shown in Fig. 5a and b, by titrating cDNA solution into MBMF or MB/lysate buffer, an increasing uorescence signal at 580 nm was detected in both systems, indicating the disturbance of the hairpin structure. Compared with the MBs' spectrum, MBMFs could bind with cDNAs in very low concentrations and the uorescence intensity displayed a stronger cDNA concentration dependency. To further quantify the limits of detection (LODs) of both MBs and MBMFs, uorescence characterization was carried out using a cDNA concentration  (Fig. 5c and d). The LODs were calculated according to eqn (3) where s is standard deviation of the uorescence intensity at low cDNA concentration, and S is slope of the calibration line. The LOD of cDNA by MBMFs was calculated to be 37 nM, whereas that by MBs could not be calculated. Therefore, when combined with strong resistance to nuclease digestion (Fig. 4f), this proven enhanced cDNA binding affinity further demonstrates the biostability required for practical applications in genetic engineering, e.g., carriers for gene transplantation. Dening cancer cell endocytosis pathways of DMFs would be benecial for analyzing cell fusion capability and maximizing oligonucleotide carrying efficiency of the structure. In the absence of such study, the effect could be predicted from timedependent imaging of uorescent library DMFs with HeLa cells. First, an MTS-assay was carried out to prove the bio-compatibility of DMFs. The cell viability of HeLa cells with DMF concentrations from 50 nM to 10 mM was characterized and an average of above 90% was observed (Fig. S6a †). This endocytosis experiment used confocal microscopy and organelle dyes, including LysoSensor Green and transferrin-alexa633 (Fig. 6). LysoSensor is a uorescent dye for lysosomes and transferrin has been widely used to identify the location of endosomes. In this study, as incubation time increased from 30 min to 4 h, the probes gradually escaped from endosomes and entered most of the lysosomes, and nally entered the whole cell except for the nucleus. While the uorescence signals of TAMRA-labeled DMFs (red), LysoSensor (green) and transferrin (blue) overlapped, most red signals were observed inside the cells, suggesting that the major portion of the probes stayed in the cytoplasm. This result shows that DMFs can serve as gene carriers for certain molecular beacons and DNAzymes. On the contrary, for normal uorescence-labeled DNAs, there is minimal cellular entry aer a 2 h incubation ( Fig. S6b and c †).

DNA synthesis
All oligonucleotides were synthesized by an automated ABI 3400 DNA synthesizer (Applied Biosystems, Inc, Foster City, CA) at 1.0 mM scale. Diacyllipid phosphoramidite was then coupled to the oligonucleotides using the same instrument. Micelles with TAMRA modication were deprotected in 3 mL TAMRA deprotection solution (methanol : tert-butylamine : water ¼ 1 : 1 : 2) at 65 C for 4 h, while all others were deprotected in 3 mL AMA solution (ammonium hydroxide aqueous methylamine ¼ 1 : 1) at 65 C for 30 min. Aer precipitation, the oligonucleotide probes were further puried by reversed-phase high-pressure liquid chromatography (HPLC) (ProStar, Varian, Walnut Creek, CA, USA). Oligonucleotide probes with and without diacyllipid groups were puried using a C4 and C18 column, respectively, and the mobile phase was CH 3 CN-TEAA solution. The concentration of probes was quantied using their absorbance at 260 nm, as measured by a Varian Cary 100 UV-vis spectrometer (Agilent Technologies, Santa Clara, CA, USA).

Real-time PCR measurement
All measurements were performed using a MyiQ™ Single-Color Real-Time PCR Detection System. The temperature was increased from 60 to 80 C in 1 degree increments and held for 5 min aer each increment. Fluorescent signal was measured, and the change in relative uorescence units (dRFU) was recorded.
Cell culture CCRF-CEM (human leukaemia) and HeLa (cervical adenocarcinoma) cell lines were purchased from American Type Culture Collection (Manassas, VA). CCRF-CEM cells were cultured in RPMI-1640 medium (American Type Culture Collection), with 10% fetal bovine serum (Invitrogen, Carlsbad, CA) and HeLa cells were cultured in Dulbecco's modication of Eagle's medium (Fisher Scientic) with 15% fetal bovine serum and 0.5 mg mL À1 penicillin-streptomycin (Sigma, St. Louis, MO), both at 37 C in 5% CO 2 atmosphere. HeLa cells were grown to 80% conuency for 48 h before incubation with probes.

Cell lysate preparation
Cells were washed and dispersed in 10 mL of PBS solution. The mixture was lysed with a Branson Sonier 450 sonicator (200 W) for 1 min, and the resulting cell lysate was stored at 4 C.

Cytotoxicity assay
The cytotoxicity of DMFs was evaluated using the CellTiter 96 proliferation assay (Promega, Madison, WI, USA) at 37 C under 5% CO 2 atmosphere. A sample of 3000 HeLa cells in 50 mL fresh cell culture medium was seeded into each test well on a 96-well plate. Aer 12 h, medium was removed from the wells and another 50 mL of fresh medium was added. Then DMFs at the desired concentrations were added to the well. Aer 48 h treatment, 20 mL of CellTiter 96 reagent was added to the well, and the 96-well plate was subjected to absorption measurement at 490 nm using a VersaMax tunable microplate reader (Molecular Devices, Inc., Sunnyvale, CA, USA). Each experiment has been repeated 5 times and the data are shown as the mean AE the standard deviation. A Student t-test was performed and p > 0.05 indicated no signicance.

Confocal uorescence microscopy imaging
HeLa cells were incubated with 1 mM DMFs at 37 C with 5% CO 2 for different time lengths, followed by washing twice with PBS to remove free probes. LysoSensor Green (Thermo Fisher) was used for lysosome staining by incubating with cells at 37 C for 30 min and transferrin from Human Serum, Alexa Fluor 633 Conjugate (ThermoFisher) was incubated with cells for 1 hour. Fluorescence imaging was performed on a Leica TCS SP5 confocal microscope (Leica Microsystems) with a 40Â oilimmersion objective. In most cases, the optical slice thickness was adjusted to 0.5 mm. DMFs with TAMRA were excited at 543 nm, and the uorescence was collected at 570 nm. Lyso-Sensor Green was excited at 443 nm, and the uorescence was collected at 505 nm. Transferrin was excited at 633 nm, and the uorescence was collected at 650 nm.

Fluorescence measurements
A Jobin Yvon uorometer was used for all uorescence measurements. For TAMRA, the solutions were excited at 488 nm and scanned from 550 to 650 nm. The anti-photobleaching mode was used for time-dependent measurement.

Static light scattering
The dn/dc value of DNA-lipid complex is 0.161 mL g À1 . 26 A 2.0 mg sample of DNA micelles was dissolved in 0.7 mL H 2 O/MeOH (90 : 10) and ltered twice before characterization by SEC/Multi-Angle Light Scattering (SEC-MALS) (Agilent Isocratic Pump, Degasser, and Autosampler, SEC-MALS from Wyatt Technology). Light scattering measurements were performed using 3 detectors at 41 , 90 and 139 and a Zimm plot was carried out by ASTRA 6 to obtain the molecular weight of DMFs.

Conclusions
In summary, using nanoares as a comparative guideline, this work investigated several basic properties of DNA micelle ares, including aggregation number, critical micelle concentration, dissociation constant and biostability. The aggregation number of DNA-diacyllipid amphiphiles for the lipo-T 60 micelle is 544, which is comparable to that of nanoares, but signicantly greater than that of DNA block copolymers. 25 DNA micelle ares have much lower CMC values compared to their polymeric analogues, indicating superior stability in low concentrations, making MBMFs suitable for use as carriers for drugs and genes. In addition, DNA micelle ares also present stronger binding affinities and higher melting temperatures toward their cDNA target relative to DNA probes without diacyllipid conjugation. Most importantly, DNA micelle ares exhibit enhanced biostability in complex biological media, further demonstrating their utility in gene recognition and gene therapy. It was also proven that DMFs could enter the entire cell aer 4 hours of incubation with only partial residency in lysosomes. These properties endow DNA micelle ares with the versatility required for bioimaging, drug delivery, and cancer gene therapy, potentially inspiring the functionalization of DNA with multiple oligonucleotides to increase the range of biomedical applications.