MicroRNA Detection by DNA‐Mediated Liposome Fusion

Abstract Membrane fusion is a process of fundamental importance in biological systems that involves highly selective recognition mechanisms for the trafficking of molecular and ionic cargos. Mimicking natural membrane fusion mechanisms for the purpose of biosensor development holds great potential for amplified detection because relatively few highly discriminating targets lead to fusion and an accompanied engagement of a large payload of signal‐generating molecules. In this work, sequence‐specific DNA‐mediated liposome fusion is used for the highly selective detection of microRNA. The detection of miR‐29a, a known flu biomarker, is demonstrated down to 18 nm within 30 min with high specificity by using a standard laboratory microplate reader. Furthermore, one order of magnitude improvement in the limit of detection is demonstrated by using a novel imaging technique combined with an intensity fluctuation analysis, which is coined two‐color fluorescence correlation microscopy.

Membrane fusion is ap rocess of fundamentali mportance in biological systems that involves highlys elective recognition mechanisms for the trafficking of molecular andi onic cargos. Mimicking natural membrane fusion mechanismsf or the purpose of biosensor development holds great potentialf or amplified detection because relativelyf ew highly discriminating targets lead to fusion and an accompanied engagement of a large payload of signal-generatingm olecules.I nt his work, sequence-specific DNA-mediated liposomef usion is used for the highly selective detection of microRNA. The detection of miR-29a, ak nown flu biomarker,i sd emonstrated down to 18 nm within 30 min with high specificity by using as tandard laboratory microplate reader. Furthermore, one order of magnitude improvement in the limit of detection is demonstrated by using an ovel imaging technique combined with an intensity fluctuation analysis, which is coined two-color fluorescence correlation microscopy.
The detection of target molecules in as pecific and sensitive manner is of criticali mportance for the development of efficient disease diagnostic devices. Nature is ag reat source of inspirationf or the design of such platforms because throughout the course of evolution highly sensitive and specific sensing or signaling processes have emerged that use refined components made of only af ew molecular buildingb locks. [1] One of these mechanismsi st he fusion of lipid bilayers; an essential process that allows the transfer of chemicals through an otherwise impervious barrier, [2,3] which facilitates inter-and intracellular communication. [4] In the case of neuronal fusion, in which Ca 2 + -triggered neurotransmitter release into the synaptic cleft from av esicle occurs on the sub-millisecond timescale, [5] complex molecular machinery facilitates this highly regulated process, and proteins belongingt ot he SNAREf amily( soluble Nethylmaleimide-sensitive factor attachment protein receptors) have emerged as the key components to facilitate the molecular recognition and fusion of bilayers. [6][7][8] An increased understanding of biological processes and the designp rinciples underlying recognition enables the progression of the field of the design of biomimetic materials, which harness the power and efficiency of natural processes, combined with biological building blocks (e.g.,D NA, RNA, peptides, proteins, lipids) to create hierarchically organized materials, and the development of an ew generation of sensors. [1] Membrane fusion has been extensively studied by using models ystems, such as proteoliposomes, in which native fusogenic proteins are exposed on the surface of artificial lipid vesicles (liposomes). [9][10][11][12][13][14] Another approach involves the designo f biologically inspiredc onstructs to trigger liposome docking and membrane fusion. Simplified synthetic analogues are an excellent tool to increase the understanding of the fusion process at the atomistic level, and each segment of the synthetic construct can be varied to study its role in the occurrence of fusion. [2] Synthetic analogues enabling specificm olecular recognition used to trigger membrane fusion include coiled-coil forming peptides [3,15] and DNA. [16][17][18][19] Cholesterol-terminated double-stranded DNA (dsDNA) has been shown to facilitate efficients elf-insertion into liposomes, [20] and that their hybridization, if designed to occur in az ipper-like fashion, can induce liposome-liposome fusion ( Figure S1 in the Supporting Information). [18,19] DNA, thanks to its high selectivity in sequenceguideds elf-assembly,h as great potentialf or biosensing applications.A ttractive properties of liposomes, such as ease of functionalization andb iocompatibility,e ndow them with promising applicationsi nb iomedical and biotechnology fields. [21][22][23] Furthermore, with relativelyf ew targets leading to fusion and engagemento falarge payloado fs ignaling molecules (e.g.,F RET-active dyes), it is reasonable to assumet hat target-mediated liposomefusion couldlead to highly amplified detection. Inspired by this reasoning, we have engineered a sensing platform for detecting ac linically relevant biomarker for influenzav irus infection, [24] microRNA-29a (miR-29a), in a highly specific and sensitive manner by DNA-mediated liposome membrane fusion. This work, conjointly with ar ecent report demonstrating the possibility of triggering peptiden ucleic acidm ediated liposome fusion with oligonucleotides, [25] constitutes the first example of biomarker-triggered liposome fusion.
Herein, we encapsulated FRET donor-acceptor pairs, DiI and DiD ( Figure S2), in the membrane of DNA-functionalized liposomes.I nt his assay,l iposomes containing either DiI or DiD FRET pairs in the membrane are functionalized with self-inserting, cholesterol-terminated dsDNA ( Figure 1). Ah airpin DNA (H, green strand in Figure 1) is designed to block the sticky end of dsDNA A/B (ds-A/B), which prevents hybridization with dsDNA C/D (ds-C/D) and inhibits liposome docking andf usion. The hairpin DNA is also specially designed to include ar egion containing ac omplementary sequence with miR-29a. Hence, in the presence of target miRNA, hybridization with hairpin DNA reveals the sticky end of ds-A/B, which can then hybridize with ds-C/D to initialize liposomed ocking and membrane fusion. This is then detected by an increase in FRET signal. The reported liposomef usion mechanism for oligonucleotide detection presentss everal key advantages in its experimental design. It is ah omogeneous assay,o perated at room temperature, and involving relativelyf ew experimental steps. Notably,o ther liposome-based assays for oligonucleotide detectionr equire the destruction of liposomes through the addition of membranedisruptinga gents. [21,26] Herein, similarly to assays developed by Jakobsene tal., [27,28] the readouti so btained without the need for additional separation, amplification, and washing steps, which has great potential for further developingt his assay towards point-of-care applications.
In this assay,s chematically illustrated in Figure 1, the hairpin DNA strand (H) is designed to hybridize on the stickye nd of a duplex and to be displaced in the presence of target miRNA. This novel mechanism of ad ual-function hairpin DNA is an advantageouss trategy because it offersauniques ite for target hybridization, whichi si ndependento ft he zippering regions; therefore, its design is universal and can be tailoredt oa dapt to ab road spectrum of target biomarkers.I ti sc rucial to optimize the design of Ht om eet the following three requirements:1 )unbound Hm ust maintain its structurals tem-loop feature (not ar andom coil structure),a nd have at least one arm of the stem that is parto ft he recognition sequences;2 )H must be strongly hybridized to ds-A/B in the absence of target, which prevents the hybridization of Aw ith Ca nd further unzipping of the dsDNAs; and 3) Hm ust be displaced only by the specific target miRNA (T), revealing the sticky end of ds-A/B, which will initiate liposome docking by hybridization with ds-C/D.
We designedt hree hairpin structures:H 6 ,H 7 ,a nd H 13 (sequences in Table S1). The optimizationo ft he hairpin design is summarized in sectionS5( Supporting Information). Briefly, first, we visualized the predicted secondary structures formed by the hybridization of H, A, and B( Figure S3) by using algorithms providedi nt he NUPACK software, [29] which also showed that out of the three predicted constructs, ds-A/B/H 7 had the highest probability of formation (Table S2). Furthermore, the hybridization and displacement of Hi ns olution were characterized by meanso fn ative PAGE analysis ( Figure S4). We demonstrated that only H 7 showed both efficient hybridization with ds-A/B andd isplacementb yT ;t herefore, H 7 was chosen as the hairpin structure for the liposomefusion assay.
Following selection of H 7 ,w et ranslated the hairpin strand displacementm echanism in the presence of target miR-29a into al iposome fusion assay.T he majority of methods for detectingm embrane fusion are based on fluorescenceg eneration or FRET,a nd careful choice of the reporter assay allows selectivediscrimination between lipid mixing( inner or outer lipid leaflet) and contentm ixing. [30] To maximize signal generation during the assay,w es elected al ipid-mixing assay to report fusion eventsb ecause studies have shown that ah igh level of lipid mixing can occur with al imited degree of content mixing. [17] DiI and DiD FRET pairs are based on Cy3 and Cy5 cyanine dyes, respectively ( Figure S2),a nd contain long alkyl chains that render them highly lipophilic;t herefore, they will be included in the membrane of liposomes. Twop opulations of liposomes containing either DiI or DiD were mixed;u pon membrane fusion, these fluorescent dyes mixed,r esultingi n FRET signal generation.
DiD and DiI liposomesw ere functionalized with ds-A/B/H 7 and ds-C/D, respectively,a nd we studied the ability of miR-29a to displace H 7 and trigger liposome fusion (Figure 2). DiD liposomes were incubated with variousq uantitieso fm iR-29a.
Then, equal volumes of DiI and DiD liposomesw eremixed and the evolution of the FRET ratio was measured overt ime (Figures 2A and S5). In the absence of miR-29a, the FRET ratio slightly increased during the first 20 mino fm easurement then plateaued, which showed that H 7 was successfully hybridized on ds-A/B and was able to preventl iposomef usion. As the quantity of miR-29ai ncreased, the FRET ratio and its gradient in the first 20 min increased;t his demonstrated thatm iR-29a was able to displace H 7 and trigger liposome fusion. To study the sensitivity of the assay, we plottedt he dose-response curve obtained after 30 min of incubating DiI and DiD liposomes ( Figure 2B). The limit of detection (LOD) was calculated by first determining the z value (z = blank + 3s), in which s is the standard deviation of the blank.T he LODw as determined to be 18 nm by reporting the z value in the equation of the curve fitted to the dose-response measurements.
Next,t he specificity of the assay was studied (Figure 3). The miR-29f amily members are important regulators of human diseases, [31] and include miR-29a, miR-29b-1, miR-29b-2,a nd miR-29c.I namicroarray assay,t he expression levels of both miR-29a and miR-29bw ere significantly different betweenc ritically ill patients with H1N1 infection and healthy controls. [24] We tested the ability of the assay to discriminate between the fully complementary target, miR-29a,a nd its miR-29 family counterparts, miR-29b and miR-29c. The sequences of miR-29b and miR-29c differ from that of miR-29a by several mismatching and additional nucleotides (TableS1). Therefore, to study the specificity of our assay,w ed esigned artificial sequences, differing from miR-29a by 1, 2, or 3n ucleotides:m iR-29a-1MM, miR-29a-2MM, and miR-29a-3MM( Ta ble S1). DiD liposomes functionalized with ds-A/B/H 7 were incubated with one molar equivalent of different miRNAs. Subsequently,e qualv olumes of DiI andD iD liposomes were mixed,a nd the evolution of the FRET signalw as measured over time (Figures 3A and S6). Only the fully complementary target, miR-29a, was able to trigger liposomefusion, whereas no other sequences showed asignificant increasei nt he FRET ratio, relative to that of the signal measured in the absence of target. Figure 3B summarizes the relative values of the FRET ratios at 30 min of incubation obtained with each miRNA, relative to the control,a nd showed that our assay had ah igh specificity because we were able to discriminate between sequences with one nucleotide mismatch.
This high specificity is attributed to the fact that displacing the hairpin requires high stringency. [32] Althought he LOD  value reported herein is much highert han that obtained with other miRNA detection methods based on enzymatic target amplification, [33,34] this assay remarkably distinguished strong sequence homologies, whichi st ypically difficult to achieve for homogeneous assays that do not incorporate washing, amplification, or separation steps. [35] Additionally,t he DNA-mediated liposomef usion assay reduces the complexity andr isk of errors that would otherwise be associated with enzymatic target amplification. [36] To further the understanding of the assay for miRNA detection and potentially gain in detection limit and/or minimize materialc onsumption,asuspension-based two-color fluorescence correlation microscopy methodw as developed that allowed extraction of FRET signals from imaging low numbers of liposomes (< 50). [37] The experimental setup is presented in section S7 (Supporting Information). In brief, it comprises a regularfluorescence microscope equipped with abeam splitter (Figure S7 a), whiche nables the simultaneous separation of the emitted light originating from DiI liposomes (centered around l = 570 nm;d enoted as the "green channel" in the following) and fused DiI andD iD liposome complexes (centered around l = 670 nm;" red channel") upon direct excitation of DiI liposomesb yal aser source (l = 488 nm).
We exploited the increased sensitivity in terms of liposome concentrationo ft he two-color fluorescencem icroscopy setup and mixed liposomes olutions at 10 pm concentration for incubation with the target, and then diluted the solutions to 1pm before performing the measurements (compared with 1.7 nm liposomec oncentration used for the microplate readers etup).
Due to its large conceptual similarity to that of fluorescence correlation spectroscopy (FCS), [38][39][40] we tried to assess FRET characteristics from the intensity fluctuations of I g (t)a nd I r (t) (intensity traces integrated over the entire green and red channels, respectively). In FCS, the average number, N,o ff luorescent particles within the readout volume (usually ad iffractionlimited spot due to the usage of ac onfocal excitation scheme) can be obtained by calculating the autocorrelation function, g, of the recorded fluorescence intensity trace, and by employing N = 1/g(0). [41] As shown in sectionS7( Supporting Information), the same methodology can be applied to the intensity traces obtainedf rom our microscope setup, if the autocorrelation function of the intensity ratio I r (t)/I g (t)i sc alculated. This normalization was necessary because fluctuations in I r (t)w ere caused by two processes, fused FRETing liposomes and bleed-through of DiI liposomes, making g(0) more sensitive to the number of DiI liposomes than that of fused FRETing DiD liposomes (section S7). Because DiI liposomes cause correlated fluctuations in I g (t)a nd I r (t), the use of I r (t)/I g (t)e liminates the sensitivity of DiI liposomes, but boosts FRET-based fluctuations ( Figure 4A), making g(0) mainly sensitive to the number of fused FRETing DiD liposomes (section S7 and Figure S8). Representative examples for the autocorrelation function of I r (t)/I g (t)a re shown in Figure 4B,c learly indicating ad ecrease in g(0) with increasing miR-29a concentration, whichi s, due to N % 1/g(0), indicative of an increase in the number of fused FRETingD iD liposomes, N FRET ,i nt he field of view.C alibration of this approach with solutionsc ontaining only FRETing vesicles of known bulk concentration, c FRET ( Figure S9), allowed us to extract ad ose-response curve ( Figure 4C)s imilar to that of the microplate reader assay.T he LOD obtained with the twocolor fluorescencec orrelation microscopy setup was determined to be 1.2 nm,w hichw as ao ne order of magnitude improvement in sensitivity compared with that of ensemble measurementso btainedb yu sing the microplate readers etup. Hence, two-color fluorescencec orrelation microscopy has in its current implementation an improved performance relative to that of the microplate readera ssay (in terms of LOD), and consumes three orders of magnitude lessm aterial thanks to its ability to operate at liposome concentrationsa sl ow as 1pm (in contrastto1 .7 nm used in the microplate readerassay).
In summary,w es et out to explore whether DNA-mediated liposomef usion could be controlled by interactions with specific nucleic acid targets,a nd whether this mechanismc ould form the basis of ad etection assay.W es uccessfully demon- Figure 4. A) The I r (t)/I g (t)r atio of the total intensity of the red channelover the green channel. B) Representative autocorrelationf unctions of I r (t)/I g (t)Àg (with g = 0.065 as the average bleed-through factor),f or miR-29a concentrations indicated (1 Tcorresponds to 7.2 10 À10 m). C) Dose-response curve obtained by using the two-color fluorescence correlation microscopy setup (averaged over threeindependent sample sets, coveringa tleast five measurements each). Error bars representSEM. See SectionS7i nthe Supporting Information for details of the entire process.  sensitive and specific detection of miR-29a at nanomolar concentrations by using aF RET-based fluorescenceo utput, and explored two methodso fa nalyzing the fluorescence signal. With as tandard laboratory microplate reader, aL OD of 18 nm was obtained, whereas, with two-color fluorescence correlation microscopy combined with an intensity fluctuation analysis, we observed aL OD improved by one order of magnitude. Liposome and analyte concentrations that were three orders of magnitude lower than those used in the microplate reader assay were employed, and demonstrated the utility of such as etup in both the study of liposome fusion and application in ab iosensing system. Furthermore, our system has potential in the development of new diagnostic platforms targeting miRNA and other nucleic acids. By optimizing the highly promisingc oncept of relativelyf ew targets leadingt of usion and activation of al arge payload of signal-generating molecules, the LOD could be much improved in subsequenti terations. Factors such as assay temperature, dsDNA coverage, and choice of reporter signal are expected to improvet he sensitivity in our assay,a nd will be the subject of furtherr esearch. Also, by tuning the design of the DNA hairpin,d etection of diversem olecular targets with high specificity,c ombined with as imple experimental workflow,i sp ossible.