Lipidated Polyaza Crown Ethers as Membrane Anchors for DNA-Controlled Content Mixing between Liposomes

The ability to manipulate and fuse nano-compartmentalized volumes addresses a demand for spatiotemporal control in the field of synthetic biology, for example in the bottom-up construction of (bio)chemical nanoreactors and for the interrogation of enzymatic reactions in confined space. Herein, we mix entrapped sub-attoliter volumes of liposomes (~135 nm diameter) via lipid bilayer fusion, facilitated by the hybridization of membrane-anchored lipidated oligonucleotides. We report on an improved synthesis of the membrane-anchor phosphoramidites that allows for a flexible choice of lipophilic moiety. Lipid-nucleic acid conjugates (LiNAs) with and without triethylene glycol spacers between anchor and the 17 nt binding sequence were synthesized and their fusogenic potential evaluated. A fluorescence-based content mixing assay was employed for kinetic monitoring of fusion of the bulk liposome populations at different temperatures. Data obtained at 50 °C indicated a quantitative conversion of the limiting liposome population into fused liposomes and an unprecedently high initial fusion rate was observed. For most conditions and designs only low leakage during fusion was observed. These results consolidate LiNA-mediated membrane fusion as a robust platform for programming compartmentalized chemical and enzymatic reactions.


Results and Discussion
Synthesis of double-chain anchor building blocks for solid-phase DnA synthesis. We have previously reported on a lipid functionalized macrocyclic anchor building block that was inserted into DNA oligomers to facilitate DNA-mediated assembly of liposomes 3 . Rohr et al. described an early-stage insertion of lipid moieties into the polyaza crown ether scaffold, i.e., at the level of compound 1 (Fig. 3). Herein, we report on an improved synthetic strategy that allows for later-stage lipid anchor functionalization. The syntheses of benzylated diamine 1 and dialdehyde 3 have been described previously 37 . Hence, starting from 1 (see Fig. 3), the final phosphoramidite 5a was obtained over four steps. First, treating 1 with palladium hydroxide in methanol at room temperature provided the deprotected diamine 2. Next, reacting 2 with dialdehyde 3 in a standard reductive amination under dry conditions afforded the desired cyclized polyaza crown ether 4, which was subsequently condensed with hexadecanal to give 5. Finally, phosphitylation of 5 provided the desired lipid anchor building block 5a primed for standard automated solid-phase oligonucleotide synthesis of LiNAs 1-4 ( Fig. 4).
Design, synthesis and characterization of LinA fusogens. Previous studies used a design of DNA oligomers that are able to fold up into a zipper-like geometry (Fig. 4) giving rise to a close distance between bilayers upon hybridization 8,10,34 . In this study we synthesized two pairs of modified DNA oligonucleotides that hybridize in the same manner. The first pair contained a single T overhang and one insertion of the 5a building block within the phosphodiester backbone (designated X E ) at the 5′ or 3′ end, respectively (LiNA-1 and LiNA-2). The single nucleotide overhang was introduced as in previous studies 3,4 to prevent the self-aggregation of LiNA molecules into micelles in aqueous phase, driven by hydrophobic interactions between the anchors 38 . In presence of the extra nucleotide, two negatively charged phosphate groups flank the anchor building block electrostatically and sterically disfavoring micelle formation (absence of foaming of aqueous LiNA stock solutions). The second set contained a triethylene glycol phosphate spacer (P3, see Fig. 4) inserted in between the anchor X E and the DNA-based zipper unit adding approximately 3 nm (1.5 nm for each spacer in the duplex, similar to the ones described for coiled-coil fusogens by Daudey et al.) 39 to the maximal separation between docked bilayers (LiNA-3 and LiNA-4). Previously, using LiNAs based on the 3-amino-1,2-propanediol backbone, the presence of a P3 spacer increased fusion efficiency 12 . In these LiNAs the distance between the 5′-OH and the N-atom bearing the aliphatic anchors was only four bonds, compared to the nine bonds in the present design (X E anchor). We hypothesized that relative to LiNA-1 and LiNA-2, which do not contain spacer P3, less strain-release would be expected upon liposomal fusion mediated by LiNA-3 and LiNA-4. In this study, these two pairs serve to evaluate the effect of linker size and flexibility on LiNA-mediated fusion of liposomes.
The oligonucleotides listed in Fig. 4B were purified on a reverse phase HPLC as previously described 34 . Circular dichroism spectroscopy revealed that both pairs (LiNA-1/2 and LiNA-3/4, Fig. 4B) formed standard B-type DNA duplexes (see the Supplementary Information). Thermal denaturation and circular dichroism experiments of the complementary strands confirmed that each X E or X E -P3-modified DNA could bind to unmodified complementary sequences (Supplementary Information, Table S2 and Figure S1 and S2). The T m of such duplexes Figure 2. Anchor design. Different anchor scaffold structures and proposed modes of anchoring (A) Anchor based on aza-crown ether scaffold has a 6-bond spacing over a rigid aromatic system giving the two lipophilic chains an approximate spacing of 0.7 nm and allowing each chain to interact with a different subset of lipids in the bilayer. (B) In the anchor based on 3-amino-1,2-propanediol scaffold both chains are attached to the same atom, favoring intramolecular interactions of the two chains, effectively resulting in a bigger lipophilic moiety which is expected to interact with the membrane in a concerted manner (C) Molecular representation of average lipid chain distances (red lines) for relaxed palmityl chains in aza crown ether and amino propanediol membrane anchors (hydrogens not shown for clarity) and top view of the aza crown ether macrocycle with hydrogen bonded (green dotted lines) ammonium head group of aminoethanol lipids from the lipid bilayer.
(56 to 58 °C) were comparable to the unmodified reference duplex (55 °C), however, duplexes of LiNA-1/2 and LiNA-3/4 exhibited an increased T m (77 ± 1 °C). This increase was attributed to the hydrophobic interactions between aligned alkyl chains. In the presence of liposomes, the lipid anchors are already embedded into the nonpolar environment of different phospholipid bilayers and cannot exhibit the abovementioned stabilization. When carrying out the fusion experiments at 50 °C, we argue that this is approx. 6-8 °C below the actual T m of the system. No fusion above the T m of the LiNA/DNA duplexes was observed at 60 °C (data not shown).
LinA-Liposome binding using Surface plasmon Resonance (SpR). The spontaneous anchoring of LiNAs into liposome bilayers was assayed using surface plasmon resonance, a method that generates a signal based on the bulk of material present a few hundred nanometers above a gold surface. Liposomes were immobilized onto a sensor chip coated with an alkyl-modified dextran matrix (Biacore L1 sensor chip) in the instrument flow cell (Fig. 5A) 40 . LiNAs were injected slowly (contact time 0-600 s, 2 µl/min, Fig. 5B) giving rise to a fast binding response, saturating the surface of the immobilized liposomes within 60 seconds. After the initial contact time, a continuous flow of 2 µl/min HBS resulted quick removal of loosely associated LiNA, but thereafter the signal approached a new baseline due to the remaining anchored LiNAs (+200-300 response units after 1000 s). LiNA presence was corroborated by testing their ability to form duplexes with complementary DNA strands (non-modified), which were likewise injected slowly (contact time 1600-2100 s, 2 µl/min) giving rise to approximately a doubling of the response at the end of contact time. This is expected for an efficient hybridization as the surface bound mass approximately doubles when LiNAs hybridize with their complementary strand. After contact of the complementary DNA, a slow signal decrease was observed during HBS flow over the surface, which must either be due to de-hybridization of DNA or "de-anchoring" of LiNA/DNA duplexes (anchoring must become less effective after nearly doubling the molecular mass by a second hybridized DNA). The fusion efficiency was assayed by measuring the content mixing (CM) of the two liposome populations using a well-established fluorescence assay. One population encapsulates Sulforhodamine B (SRB) at a self-quenching concentration (20 mM) 21,41 while the other is filled with buffer (unlabeled). In brief, the assay works as follows: upon opening of a fusion pore between docked liposomes, the content volumes rapidly mix and SRB is diluted causing an increase in fluorescence signal intensity (dequenching). SRB efflux and water influx www.nature.com/scientificreports www.nature.com/scientificreports/ (in the following termed "leakage") also lead to signal increase. Such leakage could either occur passively or as a result of membrane lesions during fusion. Thus, we refer to this experiment as "apparent CM" (Fig. 6, green curve) and for each experiment a leakage control was performed in parallel. For these controls both populations contained SRB at the same concentration and content mixing thus gives rise to zero dilution. Any observed signal increase must, in this case, be due to SRB-dilution via leakage processes (Fig. 6, grey curve). In another control, the measurement of the apparent CM was repeated with non-complementary LiNAs (LiNA-1 + LiNA-1, Fig. 6, blue curve). All fusion experiments were carried out at 20, 37 and 50 °C and in independent duplicates. The approx. number of LiNA strands on the SRB-labeled and unlabeled liposomes were 195 and 65, respectively, and the populations were mixed in a 1:3 ratio. This allows each SRB-labeled liposome to statistically fuse with up to three unlabeled ones, diluting encapsulated [SRB] from 20 mM to 10 mM, 6.7 mM and 5 mM, after the first, second and third round, respectively.

LinA-induced liposome fusion and content mixing.
In addition to measuring CM, the particle size distribution was measured at intervals throughout the course of the experiment. This was done by taking aliquots from a fusion experiment using only unlabeled liposomes and diluting these approx. 50-fold in buffer at room temperature. The samples were then analyzed using nanoparticle tracking analysis (NTA, see Figures S4 and S5).
As previously observed, the LiNAs mediated fusion most efficiently at 50 °C as summarized in Fig. 6. The inset illustrates the amount of leakage that contributed to the apparent CM signal, from which it is clear, that in absence of LiNAs the signal increase was due to background processes entirely (see Figure S6 for time course overlay with control experiments). At the same time, the presence of non-complementary LiNAs only gave rise to a minute signal increase (Fig. 6, blue curve) suggesting that engrafting liposomes with polyaza crown ether modified LiNAs had the effect of decreasing passive SRB permeability as observed previously for the anchors with a 3-aminopropane-backbone 12,13 . As hypothesized earlier, this effect is likely due to the charge repulsion between liposomes by the polyanionic DNA on the liposome surface 12,42 .
In the current study, elevated leakage was observed for LiNAs with a P3 spacer (LiNA-3/4), where both leakage and content mixing scaled with temperature, in comparison to LiNA-1/2, where leakage remained low. In previous LiNA systems, based on a simpler anchor structure (3-aminopropane-1,2-diol) 12 , the presence of a P3 spacer did not lead to an increased leakage. Based on this result, we speculate, that the spacer can affect membrane stability.
After correcting for leakage contribution to the apparent CM, the resulting net signal was used to estimate the percentage of full fusion events of SRB-filled liposomes relative to the total number of SRB-filled liposomes, hereafter termed fusion yield. To this end, a calibration curve relating the net I/I 0 to the remaining fraction of liposomal SRB-concentration (χ SRB ) follows:  Table S2).
www.nature.com/scientificreports www.nature.com/scientificreports/ where I I / 1 χ χ= is the intensity of a series of samples with different χ SRB (χ SRB = 0.8, 0.5, 0.33; corresponding to [SRB] 16 mM, 10 mM or 6.7 mM, respectively) divided by the intensity at the initial SRB concentration ([SRB] 0 = 20 mM); a and b are linear coefficients and were found to be independent on the temperature at which the samples were recorded (see Methods and Ref. 13 ). The fusion yield was then defined as following: As indicated above, yields >100% can occur, which signifies that labeled liposomes on average fused with more than one unlabeled liposome. The fusion yields at 50 °C are plotted as a function of time in Fig. 7A, and the yields at 30 min are summarized in Fig. 7B.
The temperature dependence of LiNA-induced membrane fusion; ~5% fusion yield at 20 °C, 20-40% at 37 °C and 95-170% at 50 °C (Fig. 7B); is congruent with the trend of increasing spontaneous fusion with temperature 29 . Temperature increase has a two-fold effect: i) increasing the energy of thermal collisions between membranes brought in proximity by the hybridized LiNAs and ii) decreasing the interfacial tension of the membrane due to an increased surface area -each lipid simply has a higher effective surface area 43 . The consequence of this effect was demonstrated by Parolini et al. using giant liposomes that were tethered to each other via long oligonucleotides: microscopy images showed that the contact area between the liposomes increased with temperature, without changing the number of available tethers 44 . For LiNA induced fusion, this finding suggests that higher temperatures support membrane deformation, increasing inter-liposome contact area and the number of LiNA duplexes acting cooperatively, prolonging contact time, again amplifying the number thermal collisions between liposomes. Furthermore, decreased surface tension is supportive of extreme membrane curvature, i.e., negative during fusion stalk formation and highly positive during pore opening 45,46 . A similar trend was observed by Sadek et al. in the membrane fusion mediated by a β-peptide nucleic acid based SNARE-mimic 47 . www.nature.com/scientificreports www.nature.com/scientificreports/ In the present study, the more rigid pair LiNA-1/2 exhibited a higher maximal fusion yield (~170%) when compared to LiNA-3/4 (~95%), a result that might be explained by the shorter intermembrane distance LiNA-1/2 system, than in presence of the two P3 spacers. In contrast, the apparent CM signal was comparable between these  www.nature.com/scientificreports www.nature.com/scientificreports/ systems, which prompts us to speculate whether the triethylene glycol linkers in the LiNA-3/4 play a role in the process from docking to fusion. If the linkers are innocent, the observed increase in leakage could be explained by a prolonged docked state in a configuration that is not quite as inducive to fusion than in the case of LiNA-1/2. If, however, the ethylene glycol interacts strongly with the lipid headgroups of the membrane, it might increase fusogenicity and permeability by disturbing the hydration layer around the membrane. We looked at the initial fusion rates (dYield/dt t=0 ) of the studied systems and compared them with other LiNA pairs from previous studies (see Table 1). The data suggests that the addition of a spacer slowed down the observed fusion rate both in case of the aza-crown ether and the aminopropane backbone.
The initial fusion rate for LiNA-1/2 was found to be very high (100% fusion yield within ~2 minutes). In the presence of the P3 linker (LiNA-3/4), the initial rate was slower. A similar trend was observed for LiNAs with the aminopropane backbone, that is, for LiNA-ref1/ref2 and LiNA-ref3/ref4, respectively ( Table 1). The system with P3 and aminopropane backbone (LiNA-ref3/ref4), however, showed the highest fusion yield, i.e. the SRB-labeled population in average fused with 2.8 of 3 equivalents of unlabeled liposomes, as reported earlier, despite the lower initial rate. For the best aza-crown ether anchored setup (LiNA-1/2), the labeled liposomes fused with 1.7 equivalents of unlabeled liposomes on average. Based on these and the current results, we were intrigued to further understand the parameters controlling fusion kinetics. Different anchor designs may benefit from different optimal linker. On the other hand, linker length is inversely proportional to fusion rate. In an elegant array of coiled-coil based fusogens with different linker lengths, Daudey et al 39 . were able to suggest optimal linker lengths for phospholipid and cholesterol anchors, respectively. Also, for cholesterol-anchored peptides, a slightly shorter linker on one of the fusogens resulted in increased initial rate. Our previous work tested many combinations of linker positions, lengths and anchor structure based on the aminopropane design, were similar trends in fusion efficiency were observed. On the other hand, it is more challenging to rationalize the influence of backbone structure as well as the spacing between and structure of anchor-moiety on fusion efficiency. Nonetheless, the fact that structurally rather different anchor designs could both lead to very efficient fusion underlines the robustness of using LiNAs as a tool for programmable liposome fusion.

conclusion
Herein we describe the development of new lipid-nucleic acid conjugates (LiNAs) able to efficiently induce liposomal fusion of inner membrane leaflets and subsequently enable efficient content mixing. An improved synthesis of anchor building blocks based on a polyaza crown ether scaffold with flexible late stage introduction of lipid moieties is reported. LiNAs anchored into the outer liposomes leaflet, these anchors provided strong fusogens. When labeled liposomes were fused with 3:1 excess of unlabeled ones, rapid and efficient fusion was observed. The best system provided complete turnover of the labeled population with low leakage after only 2 min when fusion was carried out at 50 °C. While anchors used in this study are structurally rather different from our previously reported LiNAs, the high fusion efficiency at 37 and 50 °C reported in our earlier study was reproduced. The findings underline LiNA-based fusogens as a robust tool for the fusion of biological membranes. LiNA-1/2 T-X E -17nt 170 ± 25 110 ± 10

General liposome preparation.
To produce unlabeled (DOPC/DOPE/Chol; 2:1:1, molar ratio) or membrane-labeled (additionally 0.25% Lissamine-Rhodamine-DPPE) lipid films, stock solutions in chloroform or methanol were mixed and evaporated under a stream of N 2 and further dried under high vacuum. The lipid films were rehydrated in buffer or SRB solution (below). The suspension was vortexed and extruded 21x through a 100 nm polycarbonate membrane (Whatman Nucleopore Track-Etched Membranes) using a hand-extruder (Avanti Polar Lipids) to form unilamellar liposomes (stored at room temperature and used within 36 h. The mean diameter was determined to be 134 ± 3 nm by Nanoparticle Tracking Analysis (NTA). encapsulation of Sulforhodamine B (SRB). Liposomes for content mixing and leakage experiments were obtained by rehydrating lipid films HBS containing 20 mM SRB, placed in a bath sonicator for 10 min at 50 °C, extruded as above and used within 36 h (mean particle diameter 135 ± 4 nm). To remove unentrapped dye, 50 µL batches of the suspensions were purified by spin-column size exclusion chromatography (GE Healthcare Illustra Microspin S-200 HR Columns) pre-equilibrated with HBS. Relative concentrations after spin-column purification was determined by fluorimetry using liposomes labeled with 0.25 mol% DPPE-Lissamine Rhodamine. After size exclusion, the following aliquots were mixed and filled to 100 µl with HBS: blank (10 µL Triton X-100 (1 wt%)), reference (10 µL of untreated liposome dispersion, 10 µL Triton X-100 (1 wt%)) and sample (10 µL of spin-column eluate, 10 µL Triton X-100 (1 wt%)). Average fluorescence was measured from five replicates (excitation wavelength (λ ex ), 560 nm; emission wavelength (λ em ), 583 nm; emission (em.) filter open, excitation/emission (ex./em). slits 5 nm, PMT 550V) and concentration determined from the average intensity (I) between three independent preparations using the formula: c = (I sample − I blank )/(I reference − I blank ) × c stock. Measurement in presence of detergent Triton X-100 (i.e. measuring a non-liposomal dispersion) gave the best reproducibility when using 384-well plates).

Sulforhodamine B (SRB) content mixing assay. Upon content mixing (CM) with unlabeled liposomes
-or leakage into the outer medium -the dye is diluted, leading to an increase in fluorescence. The measured fluorescence increase is thus based on both content mixing and leakage (Apparent CM) was corrected by a leakage (L) control, where both liposome populations contain the SRB and fluorescence increase must stem from leakage. calibration of SRB assay. A calibration curve with standard samples with different fractions of the starting concentration of entrapped SRB (χ SRB ) were measured. When normalizing the signal to the intensity measured at the starting point for content mixing (i.e., I/I 0 = Iχ/Iχ =1 ) a linear relationship to χ SRB was obtained. The normalized signal was largely independent on temperature (20, 37, 50 °C) and current liposome concentration (80, 138 and 275 µM total lipid concentration). The giving the effective χ SRB was calculated for each sample based on the SRB concentration after lysis in 0.1% w/v Triton X-100 (obtained against the standard curve of unentrapped SRB). The linear regression in the main text was calculated from a concatenate plot of three samples for each χ SRB = 1, 0.8, 0.5, 0.33; corresponding to [SRB] 20 mM,16 mM, 10 mM or 6.7 mM, respectively.