SUPPLEMENTARY INFORMATION PARAGRAPH
A functioning artificial secretory cell
Lisa Simonsson, Michael E. Kurczy, Raphaël Trouillon, Fredrik Hook & Ann-Sofie Cans
Materials.Soybean lecithin (polar lipid extract), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE), cholesterol and 1,2-dioleoyl-sn-glycero-3-phospho-l-serine-N-(7-nitro-2 – 1,3-benzoxadiazol-4-yl) diammonium salt (NBD-PS) were purchased from Avanti Lipids Inc., Alabaster, AL. Lissamine™ rhodamineB 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine (Rhodamine-DHPE) was purchased from Molecular Probes, Eugene, Oregon.Oligonucleotides were purchased from Eurogentec S.A., Seraing, Belgium; sequences: 5’-TGG-ACA-TCA-GAA-ATA-AGG-CAC-GAC-GGA-cholesterol-3’ (a); 5’-cholesterol-TCC-GTC-GTG-CCT-3’ (a’); 5’-ACC-TGT-AGT-CTT-TAT-TCC-GTG-GTC-CCT-cholesterol-3’ (b); 5’-cholesterol-AGG-CAG-CAC-GGA-3’ (b’). Tris(hydroxymethyl)-aminomethane hydrochloride (Tris) was from VWR International, Stockholm, Sweden. Pyrocatechol (catechol), calcium chloride (CaCl2), sodium chloride (NaCl), ethylenediaminetetraacetic acid (EDTA), methanol and chloroform were from Sigma Aldrich, Steinheim, Germany. Hydrogen chloride (HCl) was from Merck, Darmstadt, Germany. Tris(hydroxymethyl)-aminomethane free base (Trisma base), potassium phosphate tribasic (K3PO4), potassium phosphate dibasic (K2HPO4), potassium phosphate monobasic (KH2PO4) and magnesium sulfate (MgSO4) were purchased from Sigma Aldrich, Stockholm Sweden.
CH-DNA strands details.The CH-DNA construct was designed to mimic the SNARE proteins. The hydrophobic CH tag act as the transmembrane part of the SNARE-protein, making it thermodynamically favorable to anchor the construct into the hydrophobic part of the membrane. The construct consists of two different complementary double stranded DNA molecules; a/a’ and b/b’. Double strand b/b’ (red in Figure 1d and e) is inserted into the GUV membrane (t-CH-DNA), by microinjection (first injection). Double strand a/a’ (v-CH-DNA, black in Figure 1d and e) is incubated with the LUVs for 15-30 min, prior to microinjection of the DNA-coated, catechol filled LUVs (second injection). The double strands have a CH tag on each strand, in order to make the insertion into the membrane irreversible.1 Each double strand consists of a 12-mer (a’ and b’) and a 27-mer (a and b) single strand, leaving a sticky end of 15 base pairs. The construct is designed such that the two 27-mers (a black and b red) and the two 12-mers (a’ black and b’ red) are complementary and re-hybridize in a zipper-like fashion (a/b and a’/b’, see Figure 1e), similar to the coiled-coil transition of the SNARE-proteins upon recognition. The DNA construct is designed such that the re-hybridization (closing the zippers) is energetically favorable, yielding a net free energy gain of Δ(ΔG) = 21 kcal/mol.2
Fluorescence resonance energy transfer (FRET) measurement of DNA-induced and Ca2+ triggered lipid mixing of LUVs filled with catechol. The lipid rearrangements taking place as a consequence of fusion of two lipid bilayers were investigated using fluorescence resonance energy transfer (FRET) between donor and acceptor dyes. DNA-modified vesicles containing donor and acceptor dyes (NBD-PS and Rhodamine-DHPE, respectively) were mixed with unlabeled DNA-modified vesicles filled with catechol at a ratio of 1:4; the total lipid concentration was 275 µM. The lipid composition of both vesicle populations was DOPC/DOPE/cholesterol (39:21:40 mol %). Before mixing, the two vesicle populations were functionalized with t-CH-DNA and v-CH-DNA, respectively, at a molar lipid-to-DNA ratio of 658:1 by simply mixing the aqueous vesicle and DNA solutions. Self-incorporation of the DNA strands was allowed to take place for at least 30 min prior to use. As a negative control the experiment was repeated using vesicles where the CH-DNA was omitted. The temperature was set to 37 °C. Merging of labeled and unlabeled lipid bilayers (lipid mixing) causes an increase in donor emission due to the decrease in FRET efficiency. The change in donor intensity is plotted as ID(%) = 100 × (It-I0)/(Itotal-I0), with I0 being the donor intensity at t = 0 before lipid mixing and Itotal the donor intensity after disruption of the vesicles in 0.8% (w/v) Triton X-100. Lipid mixing was previously observed to be spontaneous and immediate upon mixing of DNA-functionalized vesicles of the lipid composition used in this work.2,3 Interestingly, we observe that by the addition of catechol to the vesicle interior, lipid mixing is no longer spontaneous upon mixing of vesicles. Instead we observe that the addition of CaCl2 (10 mM) is necessary in order to trigger lipid mixing, as seen in Supplementary Figure 1.
Fluorometer settings. FRET measurements of lipid mixing were carried out by using a QM-4/2005 spectrofluorometer (Photon Technology International Inc., Birmingham, NJ) and a 3 mL quartz cuvette. The NBD donor was excited at 460 nm, whereas the emission monochromator was set to 525 nm (excitation slits 1 nm, emission slits 10 nm). A 500 nm long pass cutoff filter was used to reject scattered light (ThorLabs).
Quartz crystal microbalance with dissipation (QCM-D). QCM-D instrument (Q-Sense E4) and substrates (AT-cut quartz crystals, f0 = 5 MHz, SiO2-coated) were from Q-Sense AB, Västra Frölunda, Sweden. We monitored the adsorption of DOPC/DOPE/cholesterol (39:21:40 mol %) vesicles using a quartz crystal microbalance with dissipation (QCM-D) setup. Vesicles of this lipid composition are adsorbed more or less intact and do not rupture into a planar supported lipid bilayer (data not shown). We subsequently added a 200 mM catechol buffer solution (200 mM pyrocatechol, 10 mM Tris, 100 mM NaCl, 1 mM EDTA) to the vesicle film and a clean SiO2-coated quartz crystal, as a reference, and monitored the dissipation and frequency response. Catechol buffer addition was followed by rinsing with catechol free buffer (10 mM Tris, 100 mM NaCl, 1 mM EDTA). The QCM-D response is shown in Supplementary Figure 2. The results seen in Supplementary Figure 2a show that there is no strong, irreversible interaction between the lipids and the catechol, since the frequency and dissipation are not changed after rinsing with buffer although an interaction with sub-second kinetics cannot be ruled out. There is, a change in both frequency and dissipation upon addition of catechol buffer to the vesicles, interpreted as a response of the vesicle shape and structure on the change in osmotic pressure, immediately after catechol buffer addition. As the catechol buffer is added, the surface tension of the membrane is reduced leaving the vesicles less rigid, which causes the observed spike in dissipation. The change in surface tension also has the effect of flattening the vesicle so that the center of mass is located deeper within the QCM sensing field, corresponding to a drop in frequency. This drop in frequency is subsequently counteracted by the loss of mass due to the exit of water. Likewise, the loss of water increases the ratio of the relatively rigid lipid molecules to water, thus reversing the increase in dissipation. These two effects eventually reach equilibrium, observed as a stable dissipation and frequency signal. Upon rinsing both dissipation and frequency return to the base line, indicating that the vesicles retain their original shape and structure and that no catechol is left bound to the lipids. These results suggest that the altered fusion behavior of the vesicles filled with catechol is rather an effect of osmotic pressure than strong interactions between catechol and lipids.
Supplementary Figure 2b shows the response in dissipation and frequency as catechol buffer is added to and rinsed off a clean SiO2 surface. The observed changes are due to the differences in viscosity between the two buffers. Since both dissipation and frequency returns to the baseline after rinsing, catechol does not bind to the SiO2 surface. This reference experiment ensures that if there is SiO2 exposed to catechol between the adsorbed vesicles, we do not need to consider any catechol-SiO2 interaction.
Current simulations. The exocytotic current was modelled using Comsol Multiphysics 4.0a on a desktop computer (CPU frequency 3.19 gHz, 3.42 GB of RAM). The system was designed in a cylinder (height: 1.3 µm, radius: 20 µm) containing water. Another cylinder (height: 1 µm, radius: 10 µm) modelling the GUV was positioned at the center of this water volume. A vesicle was placed inside the cell model by merging a sphere (radius 82 nm) with a pore cylinder (diameter: 3 to 6 nm, height vide infra). See Supplementary Figure 3 for schematics of the model used in the simulations. In the case of the bolus model, the vesicle was displaced towards the synaptic cleft, so that the vesicle protrudes by 10 nm out of the membrane, resulting in a “pore” radius of 39.5 nm (Figure 2 in the main text). This model was established to simulate an immediate release from the vesicle, without the formation of a transient pore.
The initial state was established by filling the vesicle with 32 mM of analyte. The diffusion coefficient of this analyte was 6 x 10-10 m2s-1. A disk of 5 µm in diameter was drawn on the face of the cylinder opposite to the vesicle, representing the carbon fiber microelectrode, and the analyte concentration was set at 0 mM at its surface to model steady state amperometry. Finally, diffusion profiles were simulated for 20 ms, with increments of 0.1 ms, and the normal flux of analyte at the surface of this disk was measured.
In the computed model, the length of the pore was chosen to accommodate for the different geometries between the real pore and the modeled exocytotic system (ie. toric vs cylindrical). The purpose is here to model the torus with a cylinder of identical radius by adjusting the length of this cylinder. Fick’s first law in one dimension (here, z), where D is the diffusion coefficient and c is the analyte concentration gives the expression of the analyte diffusive flux
(1)
For a cylindrical (radius r, length L) pore separating two infinite tanks of solutions respectively at concentrations C0 and 0, this expression can be simplified, in absolute value and for steady state, following
(2)
The current flowing out of the pore is the product of this flux and the cross sectional area of the pore
(3)
It is therefore possible to define the diffusive resistance of the poreRDiff = L/(πr2D) so that i = C0/RDiff. In particular, this parameter emphasizes the critical role of the pore radius (quadratic dependence) over the length (linear dependence).
Considering the real toric (inner radius r, curvature H, as shown in Supplementary Figure 4) pore, it is possible to calculate L, the length of the cylinder used to model the pore, so that RDiff remains the same, as shown in Supplementary Figure 4. For a small variation dθ of the angle θ, the diffusive resistance of the thin layer associated to this infinitesimal displacement is
(4)
The total resistance of the pore is obtained by integrating this value over the angle θ
(5)
The length L of the equivalent cylindrical pore of radius r is therefore
(6)
By calculating numerically the integral (using a Riemann sum, increment (π/100) rad) for a typical curvature H of 7.5 nm (total neck length of the lipid pore 15 nm), L was calculated for several values of r, as shown in Supplementary Table 1.
Pore diameter/ nm / 3 / 4 / 5 / 6L/ nm / 3.3 / 3.7 / 4.0 / 4.3
Supplementary Table 1: Equivalent L calculated for different pore diameters.
The simulated peaks obtained from this model are shown in Figure 3 in the main text. The 25% - 75% fall time tfall, the half peak width FWHM and the peak height Ip were extracted from these simulations, and are shown in Supplementary Figure 5. The red arrows shows the average value obtained from the experimental data, indicating that a pore diameter of 3 or 4 nm gives a good approximation of the system. However, a better fit is obtained for tFall for a pore diameter of 5 nm. As tFall is associated with the advanced state of the peak and the pore, it may indicate the expansion of the pore during the course of the peak recording. In this case, our model indicates that, on average, the lipid pores expand from 3 to 5 nm during the course of the artificial exocytotic event.
Calculating fused LUV diameters and estimating encapsulation efficiency of catechol in LUVs.The amount of released catechol is calculated from the total amount of released charge during each measured event (, where Q is the charge and I the current). Faraday’s law (Q = nNF) is used to quantify the mole amount of released catechol, N, where n is the number of electrons exchanged in the oxidation reaction and F is Faraday’s constant. The number of moles of released catechol is related to the volume, VLUV = N/c, where c is the expected concentration of catechol inside the LUV and assuming that all LUV content is released upon fusion. Fused LUV diameters, dLUV, were calculated from VLUV using Equation 1.
(1)
From the average of the recorded values of the amount of catechol and the average volume of the LUVs, the average encapsulation efficiency was estimated to approximately 15 %.This value is reasonable, as encapsulation efficiencies for LUVs have been reported to be 15-60 %4. Supplementary Figure 6 shows the distribution of LUV diameters, calculated from the released molecules from each fusion event (red bars), using the estimated encapsulation efficiency above. The fairly broad distribution of fused LUVs is not surprising when compared to the broad LUV size distribution determined prior to injection (white bars, Supplementary Figure 6).It should be noted that the discrepancy between the size distribution of fused LUVs and that measured prior to injection into the GUVs may also indicate that smaller vesicles are more prone to fuse than larger and/or that full release does not take place, but rather transient fusion events (kiss-and-run) releasing only a small fraction of the total content. This would mean that the encapsulation efficiency of 15 % may be underestimate.
Supplementary Figure 1 FRET experiment measuring DNA-mediated and Ca2+ triggered total lipid mixing of LUVs filled with catechol. At t = 0, two populations of LUVs decorated with complementary CH-DNA (v-CH-DNA and t-CH-DNA, respectively) are mixed in solution (black). A negative control using LUVs free from CH-DNA is also shown (grey). The arrow indicates addition of Ca2+ to a final concentration of 10 mM.
Supplementary Figure 2Change in QCM resonant frequency and dissipation versus time for three different overtones, n = 3, 5 and 7, respectively. (a) Addition of catechol buffer, followed by rinsing(indicated by arrows) to (a) DOPC/DOPE/cholesterol vesicles adsorbed to a SiO2 surface and (b) a clean SiO2 surface.
Supplementary Figure 3Model used in the current simulations. Model is not drawn to scale.
Supplementary Figure 4Notations used for the calculations of the equivalent pore length L allowing for the simulation of diffusion through a toric pore using a cylindrical pore. Model is not drawn to scale.
Supplementary Figure 5Peak parameters (25% - 75% fall timetfall, the half peak width FWHM and the peak height Ip) obtainedfrom the simulated peaks for different pore diameters. The red arrows shows the average value obtained from the experimental data.
Supplementary Figure 6Normalized frequency histogram of LUV diameter (white), measured using a single particle tracking system and fused LUV diameter (red), calculated from the detected amount of released catechol from three different GUVs and using an estimated average encapsulation efficiency of catechol of 15%.
References
1Pfeiffer, I.; Hook, F.Bivalent cholesterol-based coupling of oligonucleotides to lipid membrane assemblies.J. Am. Chem. Soc.2004, 126, 10224-10225.
2Stengel, G.; Simonsson, L.; Campbell, R. A.; Hook, F. Determinants for membrane fusion induced by cholesterol-modified DNA zippers. J Phys Chem B2008, 112, 8264-8274.
3Stengel, G., Zahn, R. & Hook, F. DNA-induced programmable fusion of phospholipid vesicles. J. Am. Chem. Soc.129, 9584 (2007).
4Mayer, L. D., Bally, M. B., Hope, M. J. & Cullis, P. R. Techniques for encapsulating bioactive agents into liposomes. Chem. Phys. Lipids40, 333-345 (1986).
S1