Freezing Continuous-Flow Self-Assembly in a Microfluidic Device

Jan 4, 2013 - Andreas Jahn†, Falk Lucas‡, Roger A. Wepf‡, and Petra S. Dittrich*† .... Hatem Fessi , Hemaka C.H. Bandulasena , Richard G. Hold...
0 downloads 0 Views 492KB Size
Article pubs.acs.org/Langmuir

Freezing Continuous-Flow Self-Assembly in a Microfluidic Device: Toward Imaging of Liposome Formation Andreas Jahn,† Falk Lucas,‡ Roger A. Wepf,‡ and Petra S. Dittrich*,† †

Department of Chemistry and Applied Biosciences, ETH Zurich, Switzerland Electron Microscopy (EMEZ), ETH Zurich, Switzerland



S Supporting Information *

ABSTRACT: A new method is described that combines a microfluidic device for the controlled formation of liposomes with instantaneous immobilization by means of ultrarapid cooling. The microfluidic device is composed of capillaries to hydrodynamically focus a stream of lipids dissolved in 2-propanol by two adjacent aqueous buffer streams before rapidly cooling by propane jet-freezing. The capillary containing the frozen sheath-flow is subsequently separated from the flow-focusing unit and trimmed with cryo-ultramicrotomy for imaging with cryo-scanning electron microscopy (SEM). The emergence of liposomes could be visualized by cryo-SEM without the need for chemical fixation or labeling. We demonstrate that the method is capable of revealing in more detail the formation of nonequilibrium liposomes. Partially and completely formed liposomes were observed at the miscible alcohol− buffer interface. The number density of lipid vesicles varied along the focused interface, and we frequently found clusters of liposomes. Additionally, evidence for the formation of disclike transient intermediates is presented. The method is not limited to studying self-assembly processes only. It can be extended to other biochemical reactions, crystallization processes, and even systematic interfacial mixing studies between different solvents.

C

a solid phase while limiting the formation of ice crystals. In combination with cryo-electron microscopy, cryogenic ultrastructural fixation has developed into a powerful technique for structural characterization of biomacromolecular complexes.9,10 Experimental methodologies for performing time-resolved cryo-electron microscopy have been demonstrated recently by combining microfluidic devices with rapid cooling in order to investigate self-assembly. However, most combinations of microfluidics with rapid cooling employ a sequential process, where biochemical reactions are initiated utilizing microfluidic mixing and are subsequently ejected via a jet either onto a cooled surface,11,12 directly into a cryogen,10 or onto a transmission electron microscopy (TEM) grid.13−15 The latter method can be divided broadly into two approaches, where the actual reaction is either occurring on a TEM-grid13 or inside a microfluidic device and subsequently drop-casted or spraycasted onto a TEM-grid14,15 before plunge freezing. However, a limitation of these methods is the fact that drop- or spraycasting a solution onto a TEM-grid strongly perturbs the miscible interface either due to segregated flow as in spraycasting15 or subsequent blotting as in drop-casting.14 Hence, the self-assembly of molecules can be substantially modified by the sample preparation process prior to imaging. Although plunge-freezing is commonly used to preserve biological specimen in their native environment it is limited to layers

ontinuous-flow microfluidic devices allow facile spatial and temporal control of reaction kinetics.1 Reagents are mixed at a junction and observations of the reaction are made along the length of the exit channel. As a result much effort has been undertaken to utilize them in order to synthesize microand nanoparticles at the miscible liquid interfaces. These systems include but are not limited to semiconductor quantum dots,2 metal colloids,3 nanofibers and nanowires,4 polymeric nanoparticles,5 pH-sensitive polymersomes,6 niosomes,7 and liposomes.8 Direct observation of the formation process of unmodified molecules into the supramolecular structure at the mixing interface remains challenging due to high flow velocities, rapid self-assembly, and the limited resolution of light microscopy which is typically used for observations. As microfluidic devices contain all the necessary information about the different aggregation states during the formation of particles it would be advantageous to arrest the self-assembly instantaneously and inhibit subsequent fluid movement and mass-transport; thus allowing the size, shape, orientation, and statistical distribution of the assemblies to be preserved. A promising approach is the instantaneous freezing of the microdevice to temperatures below the melting points of the solvents. Cryogenic ultrastructural fixation has become a routine method to maintain a statistical distribution of solute components and preserve the size and shape of the macromolecules in a hydrated state. Moreover, the association and orientation of the interacting particles are preserved. Very rapid cooling allows the transition of aqueous solutions from liquid to © 2013 American Chemical Society

Received: September 12, 2012 Revised: December 18, 2012 Published: January 4, 2013 1717

dx.doi.org/10.1021/la303675g | Langmuir 2013, 29, 1717−1723

Langmuir

Article

with a thickness not exceeding a few hundred nanometers in order to prevent the formation of ice crystals. Therefore, it is not able to maintain an unperturbed diffusive interface.9 Propane jet-freezing provides an alternative to plunge-freezing with the advantage of freezing tens of micrometer thick specimens without perturbing the mixing interfaces between coflowing liquid streams commonly used in microfluidic devices. Propane jet-freezing developed by Moor et al.16 in 1976 allows for high cooling rates of about 35 000 K/s for thicker samples. Twenty μm thick samples can be vitrified without introducing cryo-protectives such as glycerol or dextran. In jet-freezing, the specimen is loaded within a metal chamber or capillary that protects it from being damaged by a pressurized liquid jet of cryogen. Applying a liquid jet increases the convective heat exchange at the specimen surface to very high values allowing cooling rates up to 30 times faster than those of standard immersion methods and subsequently reducing segregation phenomena.17 We developed a novel method that employs a microcapillary system for liposome formation, rapid freezing by a propane jet and cryo-scanning electron microscopy (cryo-SEM) for imaging of nanoparticles. Our primary aim is the investigation of the formation of nanometer-sized lipid membrane structures (small unilamellar vesicles, SUVs), which is of particular interest due to the omnipresence of liposomes in nature, and their use in various commercially relevant products within the food, cosmetic and pharmaceutical industries.18−20 Formation approaches in continuous systems which include microfluidic devices involves the use of aqueous buffers and alcohols and are advantageous for large-scale production. The formation process is mainly dictated by the molecular properties of the lipids and well-defined flow and solvent conditions. However, the detailed mechanisms are not currently resolved but could provide valuable information for improvements in size, homogeneity, and encapsulation efficiency of the created liposomes. In previous studies, it was hypothesized that the formation occurs via larger intermediate structures such as discs at the alcohol− water interface or following a partitioning period and hence later onset of vesicle formation.21−24 In the following, we introduce the method and present more detailed results obtained for the assembly of phosphocholines at water and isopropyl alcohol (IPA) interfaces that give evidence for disclike intermediate assemblies.

Figure 1. Simulation of the microfluidic hydrodynamic focusing of an alcohol stream with two aqueous buffer streams. Lipids are dissolved in the alcohol stream and as mixing with the adjacent buffer streams progresses, liposomes are formed at the interface. The mechanism of formation, however, is unclear and may involve bilayer disc intermediates which eventually close upon themselves as indicated in the inset.

Device Design. The microfluidic device shown in Figure 2a is composed of three detachable units; (1) capillaries of poly(ether ether ketone) (PEEK) for solution transport, (2) a stainless steel mixer to hydrodynamically focus two miscible liquids, and (3) a brass capillary containing the sheath-flow. PEEK tubes are press-fitted into the stainless steel mixer to deliver aqueous−buffer solution from the side inlets and lipids dissolved in IPA from the far end upstream of the mixing section. An excellent leakage-free seal is obtained with slightly deformable PEEK tubes at the applied flow rates. The mixer is fabricated from a standard HPLC stainless steel tube modified with a hole drilled cross-axially close to one end to provide access for the two aqueous buffer flows. The hydrodynamically focused sheath-flow generated inside the mixer continues into a detachable brass capillary press-fitted to the focusing end of the HPLC tube. After establishing a stable sheath flow, the capillary of the microfluidic device is rapidly cooled with propane jetfreezing and immediately plunged into a receptacle filled with liquid propane (Figure 2b and Supporting Information, Figure 1). The capillary with the frozen sample solution is then removed from the microfluidic mixing unit and trimmed with a diamond knife in a cryo-ultramicrotome at a desired location along the long axis of the capillary (Figure 2c). The capillary is then stored in liquid nitrogen for at least 12 h before imaging with cryo-SEM (Figure 2d). Numerical simulations assist in providing details about the alcohol concentration distribution across the sheath-flow as well as details about the velocity distributions (Figure 2e-f). Rapid Freezing with Liquid Propane. Rapid freezing of the sample inside the capillary is of great importance in order to suppress the formation of ice crystals that would influence the statistical distribution of the solute compounds or destruct the assemblies. The size of ice crystals nucleating during cooling is inversely proportional to the speed of heat removal and hence sensitive to the surface-to-volume ratio, the temperature of liquid cryogens used during freezing, as well as its volumetric heat capacity.9 For dilute solutions the temperature zone between 0 °C and about −100 °C should be passed in less than 0.01 s.17 To realize fast cooling cylindrical brass capillaries with a high thermal conductivity of about 100 Wm−1 K−1 were used.17



RESULTS AND DISCUSSION Principle of the Method. The motivation for this study is observing the self-assembly of molecules in a miniaturized device to elucidate particle growth including nucleation and intermediate assemblies at different positions along the device. In particular, we investigated the formation of liposomes in a flow-focusing device as schematically shown in Figure 1. An IPA stream containing phosphocholine (DMPC) and cholesterol was hydrodynamically focused by two buffer streams of phosphate buffered saline (PBS). As IPA and PBS diffusively mix, liposomes with diameters between 100 and 300 nm are formed.8 It has been hypothesized that transient lipid structures with a disc shape form at the vicinity of the diffusive alcoholbuffer interface and eventually close into vesicles as the solvent quality for the lipids deteriorates as indicated in Figure 1.21,22 To better understand the liposome formation process at the miscible alcohol−buffer, interface we varied the flow focusing of the central alcohol−lipid mixture and investigated the miscible liquid interface at different positions along the capillary. 1718

dx.doi.org/10.1021/la303675g | Langmuir 2013, 29, 1717−1723

Langmuir

Article

Figure 2. Schematics to investigate lipid self-assembly at frozen miscible liquid interfaces. (a) A microfluidic device composed of three detachable units (PEEK tubing, stainless steel mixer, brass capillary). (b) A brass capillary is positioned between two orifices from which liquid propane is shot onto the capillary to arrest lipid self-assembly and freeze the liquid−liquid interface. (c) The released capillary is subsequently trimmed at different locations with a diamond knife in an cryo-ultramicrotome. (d) Cryo-SEM image of the trimmed surface following metal coating. Scale bar, 30 μm. (e) Numerical simulations of a representative model are performed to obtain (f) qualitative concentration and (g) flow velocity profiles. (h) Cross section of the elliptical concentration profile of the focused IPA stream 200 μm downstream of the capillary inlet.

Figure 3. Cryo-SEM images of the radial distribution of lipid vesicle structures within the capillary 3.5 mm downstream of the inlet. Lipid vesicles formed at the diffusive alcohol−buffer interface extend over different regions with respect to the position along the interface. All images also show contaminations and impurities as obtained during the entire process. Therefore, the liposomes are encircled for clarity. (a) Low magnification image of a focused IPA stream (FRR = 5 and VFR = 7 μL/min) traced with white dashed line. The inside capillary wall is traced by a black dashed line. The letters correspond to high magnification images indicated with white rectangles. Scale bar, 10 μm. (b) Clustered spherical and flattened lipid vesicles. Scale bar, 300 nm. (c) Series of stitched images showing a narrow band of vesicles in a region with a high IPA concentration gradient. Scale bar, 300 nm. (d) Image showing the transition from a wide band of vesicles into a narrower band. The region between the two white lines roughly show the expanding region of the vesicles. Scale bar, 1 μm. (e) Stitched images showing numerous vesicles and imprints of vesicles. Note: Scale bar, 300 nm.

1719

dx.doi.org/10.1021/la303675g | Langmuir 2013, 29, 1717−1723

Langmuir

Article

Brass capillaries with an inner diameter of approximately 50 μm and an outer diameter of approximately 200 μm were a good compromise between the pressure necessary for pumping the liquid through the capillary and mechanical stability for handling, subsequent release and trimming. Alternatively, copper and aluminum capillaries with much higher thermal conductivity can be used. Increased cooling rates are achieved with liquid propane due to the large temperature difference of its freezing point (−186 °C) and boiling point (−42 °C) which minimizes the formation of an insulating gas layer.10 Glycerol which is entirely miscible with water and IPA was added at a concentration of 15% w/w to the PBS solution to further limit ice-crystal formation during rapid cooling.10,25 We tested the freezing quality with a solution containing polystyrene spheres with diameters of approximately 25 nm. These spheres were suspended in distilled water and injected through the axial inlet and hydrodynamically focused with PBS. The polystyrene spheres remained homogenously distributed without any sign of segregation due to larger ice crystal formation (Supporting Information, Figure 2). Observation of Liposome Formation. First, vesicle formation was investigated far away from the focusing position where vesicles are presumably formed with high yield. A mixture of 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and cholesterol at a molar ratio of 3:2 dissolved in IPA at a concentration of 150 mM was injected in the center inlet at a volumetric flow rate of 1.4 μL/min and hydrodynamically focused by two adjacent buffer streams at a volumetric flow rate of 5.6 μL/min resulting in a flow rate ratio, FRR (i.e., the fraction of the volumetric flow rate of IPA to the total volumetric flow rate of PBS and IPA), of 5 and a total volumetric flow rate (VFR) of 7.0 μL/min. After freezing, the brass capillary is cut to a length of 3.5 mm and the trimmed edge imaged by cryo-SEM (Figure 3a). In Figure 3a, the focused IPA sheath-flow is highlighted by a white dashed line and appears as a smooth elliptical area in the center of the capillary. A black dashed line indicates the inside wall of the liquid-filled capillary. A lower surface roughness in the region with high concentrations of IPA compared to the surrounding aqueous buffer is a direct result of the different brittleness of IPA and buffer at a temperature of about −170 °C. The large temperature difference between the freezing point of buffer containing 15% w/w glycerol (about −5 °C) and the trimming temperature in the cryo-ultramicrotome resulted in strong fracturing. In contrast, the low freezing point of IPA (about −90 °C) makes the IPA-rich region less brittle and therefore resulted in a smoother surface topography. It was important to perform the trimming of the capillary at ≤ −170 °C to prevent smearing of IPA across the trimmed surface and obstructing the underlying features. The number density of lipid structures is heterogeneously distributed around the circumference of the elliptically focused IPA stream in Figure 3a. Large areas with high number densities of liposomes were visible at the diffusive interface across the long axis of the focused stream (Figure 3b, e). The mainly spherical liposomes had diameters varying from about 100−200 nm. Furthermore, the hollow structures visible at the surface are most likely the imprints of liposomes. The occurrence of lipid structures along the short axis of the elliptical focused alcohol stream was limited to a narrow band at the interface (Figure 3c). Figure 3d depicts the transition from a wide band of vesicles at the long axis to a narrow band toward the short axis of the ellipse. This finding correlates with the

varying concentration profiles along the different axis of the elliptical alcohol stream. A cross-sectional view of the simulated alcohol concentration profile depicted in Figure 2h qualitatively shows a shallower concentration gradient along the long axis and a steeper gradient along the short axis of the ellipse. Figure 3e corresponds to the extended region along the long axis of the ellipse with a low IPA concentration gradient as can be seen qualitatively in Figure 3h. The narrow band of lipid structures shown in Figure 3c corresponds to a region with a much higher IPA concentration gradient. Additionally, due to the parabolic flow profile in pressure-driven flow the velocity decreases toward the capillary wall with a maximum velocity in the center of the capillary. This velocity distribution results in longer lipid self-assembly times closer to the capillary wall, whereas the flow velocity increases toward the center of the capillary leaving less time for lipid self-assembly. Clusters of spherical and flattened lipid structures were often observed (Figure 3b). These clusters, which contained a large number of lipid structures, appear to be randomly distributed along the circumference of the alcohol−buffer interface. The flattened and spherical shapes of vesicles correspond to previously reported variations in possible shapes of nonequilibrium vesicles.24 However, the reason for clustering of vesicles at such high density remains unclear at the moment. It could be the result of hydrodynamic instabilities due to unstable displacement of fluid with different viscosities known as viscous fingering.26 It is well-known that mixtures of IPA and water result in complex concentration dependent viscosity changes.27 While segregation due to large ice crystal formation cannot be entirely excluded, it nevertheless seems unlikely considering the addition of glycerol, high salt concentration in the hydration buffer solution, and the large amount of IPA further suppressing the formation of ice crystals. Observation of Intermediates. Figure 4 shows multiple images of the trimmed surface of the same capillary used before, but at a distance of approximately 500 μm downstream of the capillary inlet. Since this section is much closer to the hydrodynamic mixing region in the HPLC tube, the IPA concentration in the center is higher. Pure IPA can be highly viscous at −130 °C (currently the lowest temperature for our cryo-SEM setup), and therefore, the central region of the focused IPA stream breaks through the surface metallization layer during imaging in the cryo-SEM at −130 °C (Figure 4a). Figure 4b−d shows magnifications of the region indicated in Figure 4a, and shows multiple flat aggregate structures but very few liposomes. The numerous white spherical structures and white aggregates were contaminations most likely from the liquid nitrogen bath during sample transfer. In a region with a lower IPA concentration (see magnification in Figure 4b), few complete liposomes were observed. The incomplete vesicles could be the result of intact vesicles which were fractured during cryo-ultramicrotomy or are vesicles that are in the process of closing. Nearby we found flat structures with either smooth, round-shaped or more diffuse edges. We also found structures with ellipsoid shape (e.g., in Figure 4c), indicating different orientations of the structures along the channel. Figure 4c and d shows higher magnification images of these disclike structures which are localized to a specific region at the buffer/ IPA interface, that is, at both ends of the long axis of the ellipsoid (compare to Figure 2h). They reduce in number density with higher buffer concentrations (toward the bottom left corner of Figure 4a) and higher IPA concentration (toward the center in Figure 4a). 1720

dx.doi.org/10.1021/la303675g | Langmuir 2013, 29, 1717−1723

Langmuir

Article

equally distributed over the cross-section of the capillary and seems to correlate to the alcohol concentration gradient. Due to the two-dimensional focusing of the alcohol stream, this gradient is elliptic as depicted in Figure 2h. As most liposomes were formed along the long axis of the concentration ellipsoid, particularly toward the outer edge, we conclude that a low alcohol gradient favors liposome formation. However, there are other influences on the formation process that should be considered. In particular, one might presume that shear stress given by the fluid viscosity and the velocity gradient have an effect on the formation efficiency of the liposomes. However, a previous study provided evidence that shear stress appears to be only a minor contributing factor.22 It is more likely that velocity distributions and hence mixing time across the microchannel diameter cause significant differences in the spatiotemporal location of liposomes. These variables together with the concentration distribution of the lipids including solubility and viscosities will be investigated in future. However, some improvements of the method are required to perform more systematic studies such as a better prevention of ice crystal formation, so that liposomes could be clearly identified, reliably quantified and specified, including their diameter.



Figure 4. Cryo-SEM images of the radial distribution of lipid vesicle structures about 500 μm downstream of the inlet of the 5 mm long capillary. (a) Low magnification image of a focused IPA stream (FRR = 5 and VFR = 7 μL/min) traced with a white broken line. The inside capillary wall is traced by a black broken line. Reduced mixing and hence higher alcohol concentration result in a highly viscous region that broke through the metal surface coating during cryo-SEM imaging. The letters correspond to high magnification images indicated by a star and image series indicated by black rectangles. Scale bar, 10 μm. (b) High magnification image showing a mixture of spherical and fractured vesicles (yellow circles) as well as disclike lipid structures (red circles, dashed lines). Scale bar, 300 nm. (c) High magnification image showing encircled multiple disc-like lipid structures from (b) in more detail. Scale bar, 100 nm. (d) High magnification image of fractured or incomplete vesicles (circles with full lines) and flat aggregate structures (circles with dashed lines). Scale bar, 100 nm.

CONCLUSION In conclusion, we have described a new method to investigate molecular self-assembly at the interface of coflowing, diffusively mixing streams in a microfluidic device by means of rapid freezing, cryo-ultramicrotomy, and cryo-SEM imaging. We investigated the formation of liposomes and transient intermediate lipid structures formed at the interface of lipid− alcohol and buffer. Furthermore, the unexpected formation of randomly distributed clusters with high number densities of lipid vesicles was observed, which could be the result of viscous fingering between liquids of different viscosities. Hence, our technique provides a means to investigate molecular self-assembly without altering the physical characteristics of the molecules of interest, as is often the case with fluorescence and other measurements. We anticipate the extension of our approach to the investigation of interfacial assembly of other microfluidic systems, including metal−organic frameworks28 or polymeric systems.6 Furthermore, this method can potentially allow the investigation of mixing at the submicrometer scale in order to study cluster formation between different miscible liquids as is presumed to occur in alcohol−water systems.29,30 Other applications could include studies of the kinetics and reassociation of large protein constructs such as ribosomal subunits.

The absence of these flat aggregate structures further downstream of the capillary (Figure 3) suggests that these disclike structures are transient nonequilibrium structures which are most likely intermediate structures of aggregated lipid molecules. Closed liposomes are formed at much higher number densities at a later time suggesting that a certain time is required for lipid molecules or micelles to diffuse and aggregate. It should be noted that the concentration gradient of IPA could have a significant effect on the formation process as well. The distribution of vesicles as shown in Figure 3 indicates that gentle gradients are favorable for a high yield of liposomes. In this regard, it should be mentioned that preliminary experiments under conditions with lower FFR of 3.5 were performed. This generated a steeper gradient of IPA and only a low yield of liposomes could be found in the cryo-SEM images (Supporting Information, Figure 3). Taking the observations together, we found strong indications for the hypothesized, certainly still simplified, mechanism depicted in Figure 1. We observed disclike intermediate structures close to the point where the alcohol stream is focused by the aqueous solutions, and we could clearly image numerous fully closed, spherical liposomes further downstream. Thereby, the number of liposomes was not



MATERIALS AND METHODS

Device Fabrication. A 20 mm long standard stainless steel HPLC tube with an outer diameter (O.D.) of 1.6 mm and an inner diameter (I.D.) of 190 μm was modified with a stepped hole perpendicular to the tubular axis. First, a 380 μm through-hole is drilled at a 1 mm distance from the exit end of the HPLC tube. In order to attach PEEK tubes (1542, Ercatech AG, Bern, Switzerland) with an O.D. of 510 μm and an I.D. of 255 μm to the HPLC tube, three of the access holes are expanded to 500 μm and chamfered to about 540 μm while the exit end is manually expanded with a conical reamer to slightly more than 200 μm. Custom fabricated 5 mm long brass microcapillaries (Hessmer & Wercker Microtubes, Berlin, Germany) with an O.D. of about 200 μm and I.D. of 50 μm serve as specimen microtubes. The PEEK tubes and the brass microcapillary are press-fitted into the respective openings. 1721

dx.doi.org/10.1021/la303675g | Langmuir 2013, 29, 1717−1723

Langmuir

Article

Lipid Mixture and Buffer Preparation. 1,2-Dimyristoyl-snglycero-3-phosphocholine (DMPC) was obtained from Avanti Polar Lipids (Alabaster, AL), and cholesterol was obtained from Sigma Aldrich (Buchs, Switzerland). DMPC and cholesterol in a molar ratio of 3:2 were dissolved in dehydrated and degassed isopropyl alcohol (IPA) to a total lipid concentration of 150 mM. Phosphate buffered saline (PBS) (10 mmol/L phosphate, 2.7 mmol/L potassium chloride, 138 mmol/L sodium chloride, pH 7.2−7.6) was purchased from Sigma Aldrich and used as a hydration buffer solution with the addition of 15% w/w glycerol (99% grade, ABCR GmbH & Co.KG, Karlsruhe, Germany). Specimen Microtube Preparation. A lipid mixture dissolved in IPA was injected from the center axis inlet of the HPLC tube and hydrodynamically focused with two aqueous PBS streams and transferred into a press-fitted brass capillary. Fluidic reagents were introduced with gastight glass syringes (BGB Analytik AG, Boeckten, Switzerland) using a syringe pump (Nexus 3000, Chemyx Inc., Stafford, TX). All fluids were filtered (pore size: 0.2 μm, Sartorius Stedim Biotech, Goettingen, Germany) to prevent particulate contamination and clogging of the flow-focusing device. The flow rate ratio (FRR) was varied between 3.5 and 5 at a constant total VFR of 7.0 μL/min to obtain a stable mixing interface between IPA and the buffer solution. The sheath-flow inside the brass capillary was frozen in a liquid propane jet freezer (Bal-Tec JFD 030, Balzer, Liechtenstein) with a cooling rate of 35 000 K/s to avoid water crystallization. A modified plunger housing was used to position the capillary between two opposite nozzles, through which liquid propane at a temperature of −180 °C is shot onto the samples at a pressure of approximately 6 bar for 0.7 s.16 Immediately after the propane jet subsided, the frozen sample capillary was plunged into a cooled receptacle filled with liquid propane cooled to −150 °C. The immersed brass capillary is subsequently removed from the HPLC tube-mixer and stored in liquid nitrogen before trimming with a cryo-ultramicrotome. Cryo-Ultramicrotomy. Trimmed surfaces of the frozen sample in the brass capillary were prepared with a Leica EM FC6 ultramicrotome equipped with an EM UC6 cryo-chamber (Leica Microsystems, Balgach, Switzerland). The temperatures of the knife, the sample holder, and the cryo-chamber were set to −170 °C. The frozen sample microtube was fixed in a custom-made chuck in the microtome and then trimmed using a 45° diamond cryo-knife (Diatome, Biel, Switzerland) with a clearance angle of 6°. Coarse trimming of the capillary was performed with a nominal cutting feed set in the range of 200−300 nm and a cutting speed of 100 mm/s. Fine trimming was performed at a cutting feed set between 50 and 100 nm and a cutting speed in the range of 20 to 50 mm/s. An ionizer antistatic device (Diatome, Biel, Switzerland) was used to reduce contamination of the freshly trimmed sample surface. The samples were kept in a custommade holder in liquid nitrogen before observation with a cryoscanning electron microscope. Metal Coating. After trimming the surface of the capillary, the capillary was mounted under liquid nitrogen onto a precooled cryostage and transferred under inert gas in a cryo-high vacuum airlock (Bal-Tec VCT010, Balzer, Liechtenstein) to a precooled metal deposition device at −145 °C (Bal-Tec BAF060, Balzer, Liechtenstein) for conductive surface layer coating. Tungsten was deposited at an elevation angle of 45° to a total thickness of 2 nm at −130 °C followed by an additional 2 nm with a continuously varying angle between 90° and 45°. The vacuum of the BAF 060 was held at 3 × 10−9 bar and the stage was cooled to a temperature of −145 °C. The second deposition is needed to minimize charging artifacts of the strongly corrugated surface during imaging that may compromise image stability at high magnification. Cryo-SEM Imaging. After coating, frozen sample microcapillaries are quickly transferred for imaging under high vacuum (