Reconstitution and real-time quantification of membrane remodeling by single proteins and protein complexes

Abstract

Cellular membrane processes, from signal transduction to membrane fusion and fission, depend on acute membrane deformations produced by small and short-lived protein complexes working in conditions far from equilibrium. Real-time monitoring and quantitative assessment of such deformations are challenging; hence, mechanistic analyses of the protein action are commonly based on ensemble averaging, which masks important mechanistic details of the action. In this protocol, we describe how to reconstruct and quantify membrane remodeling by individual proteins and small protein complexes in vitro, using an ultra-short (80- to 400-nm) lipid nanotube (usNT) template. We use the luminal conductance of the usNT as the real-time reporter of the protein interaction(s) with the usNT. We explain how to make and calibrate the usNT template to achieve subnanometer precision in the geometrical assessment of the molecular footprints on the nanotube membrane. We next demonstrate how membrane deformations driven by purified proteins implicated in cellular membrane remodeling can be analyzed at a single-molecule level. The preparation of one usNT takes ~1 h, and the shortest procedure yielding the basic geometrical parameters of a small protein complex takes 10 h.

Introduction

Membrane curvature is an essential regulator of membrane structure and function, often bridging the two¹. Curvature critically affects membrane stability and hence its barrier function², influences self-assembly of functional membrane domains and structures^3-6 and modulates membrane motility and trans-membrane signal transduction⁷. Curvature of cellular membranes is generally highly dynamic, with large variations of the membrane curvature required for membrane remodeling: fusion, fission and poration². The curvature is made and maintained by dedicated protein complexes specific to each intracellular organelle and deeply integrated into intracellular metabolic and signaling pathways¹. Such complexes share a set of membrane-residing motifs, or domains, specifically designed for curvature sensing and creation⁸. All those domains interact directly with lipids and locally perturb the lipid bilayer. Although several general mechanisms of such perturbation(s) leading to induction of membrane curvature have been identified⁹, the corresponding local nanoscale membrane deformations have rarely been observed directly. Mechanistic insights generally come from analyses of collective protein effects on the membrane, such as the creation of visible shapes¹⁰ or downstream effects of the local deformations, such as protein sorting in curvature gradients¹¹. However, how the individual proteins and protein complexes act and cooperate at the nano- and mesoscales often remain uncertain and await direct observation^12,13. Hence, experimental methodologies are needed to monitor the dynamics of nanoconfined membrane remodeling with adequate temporal resolution.

Mechanistic analyses of protein-driven membrane deformations tremendously benefit from in vitro modeling, in which membrane transformations by individual protein actors can be reconstructed and analyzed in a controlled environment¹⁴. For this purpose, researchers have developed various lipid membrane templates, ranging from giant lipid vesicles to lipid bilayers supported by silicon nanowires and carbon nanotubes^15-18. Importantly, similar nanostructure platforms are now available and can be used to perform precise nanoscale manipulations of membrane curvature in cells¹⁹. However, rigid supports might critically restrict protein-driven membrane deformations. By contrast, resolving membrane deformations produced by individual proteins on soft, free-standing membrane templates is difficult because large curvature and shape fluctuations intrinsic to such templates²⁰ impede high-resolution analyses. We recently demonstrated that highly curved lipid membrane nanocylinders (ultra-short lipid nanotubes (usNTs; ref. ²¹) can provide a compromise solution. Although the usNT membrane retains bulk softness^22,23, the fluctuations of the tube shape are damped so that the tube geometry can be controlled with subnanometer precision by lateral membrane tension²¹. We demonstrated that ionic conductance of the usNT lumen could be measured by the conventional patch-clamp technique²⁴. Protein adsorption causes deformations of the usNT membrane, leading to changes in the luminal conductance, analogous to partial blockage events in conventional nanopores and protein channels²⁵. Upon proper calibration, the conductance changes can report the geometry of nanoscale membrane deformations caused by protein binding²¹. Furthermore, part of the electric field used in the luminal conductance measurements is applied across the usNT wall. This field could drive rapid subnanometer changes of the usNT shape, enabling perturbative analysis of the elastic parameters of the nanotube²⁶.

Here, we describe how to create, calibrate and manipulate usNTs and further use these templates to assess protein activities. We further explain how to use electric field for quantitative manipulations of the NT geometry at the nanoscale and how to use such nanoscale perturbations of the usNT shape for quantitative analyses of the protein action. Finally, we describe the simple data treatment needed to extract molecular ‘footprints’ of protein–membrane interactions.

Development of the protocol

Lipid membrane nanotubes have been widely used for studying viscoelastic properties of the lipid bilayer and, recently, mechanistic analyses of proteo-lipid interactions^27,28. More than a decade ago, we attempted to use NTs for studying mechanisms of membrane fission²⁴. We needed to create an assay for real-time monitoring of the connectedness of the NT lumen in the pre-fission state(s). Inspired by the electrophysiological patch-clamp assays applied in cellular systems to monitor dynamics of membrane transformations in fusion and fission^24,29,30, we adapted the patch-clamp technique to planar lipid bilayers and, recently, to giant unilamellar vesicle templates³¹. Briefly, to make a lipid NT, a glass patch-clamp pipette is brought into close contact with the lipid membrane template of choice. As in cellular systems, a tight contact is formed between the pipette rim and the membrane. Similar to establishing a ‘whole-cell’ configuration²⁴, breaking the patch inside the pipette does not alter this tight contact. Moving the pipette away from the membrane creates the NT (Fig. 1). In this protocol, we focus on the formation of usNTs from planar lipid bilayers ‘painted’ over a mesh mounted on the coverslip of an observation chamber (Fig. 1a). We found that this system provides maximum stability and robustness of the usNTs, with their lengths ranging between 80 and 400 nm and their luminal radius between 3 and 10 nm. As depicted in Fig. 1b, the lumen of such a small usNT closely resembles a nanopore, connecting the pipette interior and the media in the observation chamber. The ionic conductance through such a pore is measured by applying a fixed potential difference between the ground electrode placed in the observation chamber and the measuring electrode in the pipette (²⁴, Fig. 1). By analyzing the signal-to-noise ratio (SNR), we found that the usNT conductance is extremely sensitive to changes of the tube shape caused, for example, by protein adsorption. With appropriate calibration, the conductance changes reported the geometry of nanoscale membrane deformations induced by individual protein units²¹. We continue to develop the method toward its integration with single-molecule fluorescence microscopy for correlated stoichiometric analyses of membrane remodeling processes³¹.

Fig. 1 ∣. Main geometric and electrical parameters of usNT.

a, A schematic representation of a usNT pulled from a planar lipid membrane formed over a hole in a TEM grid mesh by a patch-clamp pipette. b, The equivalent electrical circuit of the usNT consists of three main elements: the electrical resistance of the pipette $({\text{Re}}_{\mathrm{p}})$, generally negligible; the electrical resistance of the gigaseal $({\text{Re}}_{\text{seal}})$; and the electrical resistance of the NT lumen $({\text{Re}}_{\text{NT}})$. ${\text{Re}}_{\text{NT}}$ depends the NT length $(L_{\text{NT}})$; the length changes $(\Delta L)$ are controlled by a precise nanoactuator.

Molecular-sensing capabilities of the usNT system inspired creation of a local perfusion system for delivering extremely small amounts of a protein of interest directly to the usNT (Fig. 2). The perfusion system utilizes glass patch-clamp pipettes with a minute luminal volume (a fraction of microliter), greatly diminishing the amount of protein used in the assay. The protein concentration in the vicinity of a usNT is proportional to the distance between the delivery pipette and the tube, which, in the case of reversible protein adsorption, enables real-time manipulation of the stationary protein coverage of the usNT membrane. Such manipulation is instrumental in the analysis of both collective and individual protein activities, as described in detail below.

Fig. 2 ∣. Experimental setup for usNT production and the protein delivery system.

a, Photograph of the experimental setup depicting the experimental chamber with the inserted ground electrode set on the stage of an inverted microscope. The patch-clamp and delivery pipettes, located above the chamber, are moved in 3D by two independent coarse xyz micromanipulators (not shown). The Tygon tubing is connected to the open port of the patch-clamp pipette. b, Micrograph showing the tips of the patch-clamp and delivery pipettes located proximal to the lipid membrane formed over a hole in the TEM grid. c, Schematic representation of local protein application through the tip of a delivery pipette located in the vicinity of a usNT.

In the elaboration of the usNT protocol, we noticed that, besides driving ionic current through the usNT lumen, the electric field could be used to perturb the NT shape by effective renormalization of the lateral membrane tension due to the Lippmann effect³². The tension decrease leads to the NT widening in proportion to the field power; the effect was quantified to obtain the elastic moduli of the usNT membrane^26,33. We further advanced the method to monitor the effect of bulk protein adsorption on membrane elasticity. As we demonstrate here, such measurements characterize the collective protein effect in terms of the intrinsic curvature of the proteo-lipid membrane. Thus, the method potentially allows direct comparison of individual and collective activity of membrane-remodeling proteins in a single experiment.

We envision that further development of the method would involve partial to complete automation of the usNT production, enabling parallel analyses of multiple usNTs. Eventually, the assay could be converted to a high-throughput tool for single-molecule and mesoscopic analyses of protein and proteo-lipid machineries.

Overview of the procedure

The protocol has a linear structure and consists of relatively few sequential steps. First, planar lipid membranes suitable for NT pulling are generated (Steps 1–6). A grid mesh support is used to enable simultaneous production of many isolated membranes (Fig. 1a). The system greatly facilitates the Sequential production of multiple usNTs, which are required to characterize the geometry of membrane deformations by individual protein species. Next, an NT is pulled from one of the reservoir membranes (Steps 7–18). A multistep calibration protocol is applied (Step 19) to obtain the NT membrane curvature, mechanical parameters of the NT membrane (tension and bending modulus $(k)$) and the position of the nanoactuator corresponding to zero NT length. Next, the SNR of the NT system is checked at a range of lengths to verify that short (80- to 400-nm) NTs (usNTs) provide the best SNR and to test the long-term stability of the usNT conductance (Step 20). Finally, the protein-delivery system is installed and tested (Fig. 2, Steps 21–31). At this stage, everything is ready to analyze membrane curvature creation by the protein of choice.

Various scenarios of the protein actions on usNTs can be envisioned: the proteins can cause no effect on usNT conductance, destabilize the NT by membrane rupture or fission, encage the NT within a rigid protein scaffold or cause NT constriction or expansion. This protocol is designed to resolve the activity of individual proteins or protein complexes sparsely covering the usNT and reservoir membranes. In such a regime, small, and often reversible, changes in the usNT conductance report local membrane deformations by the proteins. Basic geometric characteristics of the deformations, such as size and mean curvature, can be obtained from the dependence of the amplitude of the conductance changes on the NT radius. Unless a protein of interest has known patterns of curvature activity, one begins from a preliminary characterization of the protein’s effect(s) (Steps 32–39). First, one assesses whether the protein destabilizes the NT, restricts its geometry or interferes with stable conductance measurements. If the protein produces robust constriction of the usNT, the local characteristics of such activity is verified by assessing the dependence of the occurrence of the conductance changes on the usNT’s length (Step 40A and B). Finally, the dependence of the occurrence and amplitude of the conductance changes on the usNT membrane curvature is measured. The obtained dependence is fitted with a model equation describing the local geometry of membrane deformation (Steps 41–49).

In an additional section, we also describe the analysis of the usNT constriction at higher protein coverage and specifically discuss how to obtain quantitative information about the protein adsorption and elastic properties of the proteo-lipid membrane (Step 50). This information can be further used to extract averaged geometric parameters of the protein (e.g., spontaneous curvature) and compare them with those obtained earlier by analysis of individual deformations. In the ‘Experimental design’ section, we subdivide the workflow into several distinct blocks and provide additional details and explanations for each one.

Comparison with other methods

Studies of membrane transformations at nanoscale rely heavily on electron microscopy (EM), primarily because of its exceptional spatial resolution. However, protein-driven membrane transformation can be ultra-fast²⁶, making sequential EM sampling difficult and thus limiting the analyses to stationary or artificially stabilized membrane structures^34,35. An emerging alternative method for quantitative characterization of nanoscale membrane remodeling is high-speed atomic force microscopy (HS-AFM), the technique of choice for real-time imaging of biomolecules³⁶. Application of HS-AFM to membrane remodeling, realized already by several groups, revealed previously unseen dynamic rearrangements of the protein machineries implicated in intracellular membrane fission^37,38. However, HS-AFM had been primarily used to resolve the protein part of the membrane remodeling machinery, leaving it uncertain as to whether the subtle changes of membrane curvature hidden beneath the protein complex could be assessed as well.

At larger length scales, membrane curvature dynamics can be quantitatively assessed by fluorescence microscopy. Even with cutting-edge real-time super-resolution technology³⁹, fast subnanometer changes of membrane geometry by individual protein complexes remain out of reach. However, the curvature of elementary membrane shapes, such as spheres and cylinders, can be resolved with 1- to 10-nm precision if the fluorescence signal from a membrane label is calibrated, enabling ensemble measurements of membrane curvature creation and sensing by proteins^40,41. Microns-long NTs pulled from different membrane reservoirs have been widely used in such measurements^2,27,28. Fluorescence microscopy observations of membrane NTs can be seamlessly combined with axial force measurements by optical tweezers, enabling quantitative assessment of the protein effects on the viscoelastic properties of the nanotube membrane^42,43. Our approach takes advantage of the NT system and supplements it with single-molecule sensing capabilities intrinsic to nanopore systems, thus enabling the detection and quantification of the membrane deformations produced by single protein species.

Limitations

The main experimental readout of this method is the usNT conductance; interpretation of its changes due to protein-induced membrane deformations is necessarily limited by the molecular model of the deformations. In this protocol, a simple and intuitive approximation is used in which the local deformation is parameterized by its area and mean curvature. Although extremely coarse, this approximation suffices for basic characterization of the membrane remodeling activity of a single protein molecule and, importantly, allows straightforward linking of single-molecule and ensemble measurements. Further elaborations of the model would be needed to assess the topology of the deformation, as well as the effects of protein shape and charge. As in nanopore-based molecular sensing, rational protein mutagenesis could help relate conductance phenotypes to the deformation structure and, ultimately, discriminate different protein species using the deformation footprints. Although the method provides label-free molecular-sensing capabilities, it benefits from an additional readout verifying the stoichiometry of protein oligomers. In this sense, we recently showed that the method could be integrated with fluorescence microscopy³¹. The method seems to be limited to curvature-active proteins (i.e., those with large deformation footprints). However, it depends on the fine details of the proteo-lipid interactions and, more importantly, on lipid susceptibility to bending. Softer usNTs could provide better molecular-sensing capabilities. Overall, the full capacity and limitations of the method are yet to be determined.

Applications of the method

Initially, we designed the method to resolve and characterize acute membrane constriction by dynamint (Dyn1), a large GTPase orchestrating membrane fission during synaptic endocytosis⁴⁴. Dyn1 is the founding member of the dynamin superfamily of large GTPases, which is implicated in intracellular membrane remodeling, fusion and fission⁴⁵. Using purified Dyn1 and usNTs, we unraveled how small Dyn1 oligomers constrict and destabilize the usNT membrane, using the energy of GTP hydrolysis^21,26,46. We identified the size of the minimal Dyn1 complex capable of severing usNTs and determined the pathway of membrane rearrangement during the scission^21,26,46. One of the major methodological breakthroughs in these studies was the real-time detection of minute and reversible membrane deformations synchronized with the GTPase activity of small Dyn1 oligomers. These findings highlight the main advantages of the method in quantitative analyses of nanoscale non-equilibrium membrane deformations driven by active, transient protein complexes such as fusion and fission machineries.

However, the method is not bounded by such relatively exotic and complex protein actors. We demonstrate here that similar analyses can be applied to simpler protein species (e.g., the Epsin N-terminal homology (ENTH) domain widely used as a prototype membrane curvature creator)^2,47, indicating a possible usage of the method for label-free detection and quantification of protein adsorption. In this respect, we believe that future applications will include integration of the usNT in molecular-sensing devices with a broad spectrum of analytical capabilities.

The precision and predictive power of the method are expected to greatly benefit from preliminary ultra-structural analyses of the protein complexes of interest, especially in the membrane-bound form. If supplemented by fluorescence microscopy and EM³¹, as well as by rational protein muta-genesis, the method should become a powerful tool for studying dynamic and reversible membrane deformations at the nanoscale.

Apart from the detection and characterization of elementary membrane deformations, the method allows straightforward quantification of the usNT constriction by multiple protein species, as demonstrated here for the ENTH domain. Direct comparison of individual and collective protein action is expected to provide critical insights into cooperative protein action at the mesoscale, for example, during protein oligomerization, heterogeneous polymerization and formation of membrane nanodomains. Finally, the method remains one of the few tools enabling real-time quantification of membrane deformations causing instabilities, such as poration, and fission.

Experimental design

Making the planar lipid membrane reservoir (Steps 1–6)

The procedure begins with the assembly of an observation chamber with a grid mesh for the lipid membrane formation. Semi-cured PDMS spacers are made for the grid, the grid is mounted, and the PDMS is firmly cured (‘Equipment setup’, ‘Preparation of the observation chamber’). This protocol describes an ‘open chamber’ scheme in which the lipid bilayer is suspended within a single compartment instead of separated into two compartments as in traditional designs^26,48. This scheme increases membrane stability, thus facilitating the production of durable usNTs with stable conductance (see Step 19B). Lipid bilayers are prepared on the grid mesh using the ‘painting’ technique. The goal is to form multiple independent membranes on distinct holes of the grid (Fig. 1a). Depending on the lipid composition, 50–90% of the holes will contain a lipid bilayer (Fig. 2b).

Production and calibration of the usNT (Steps 7–20)

The production begins with pulling of a conventional, micrometer-long NT and assessment of its mechanical parameters and geometry²⁶. To pull the NT, a glass patch-clamp pipette is used. The pulling is accompanied by electrical measurements that provide essential readouts at different stages of the NT production (Steps 7–18). Upon NT formation, the dependence of the ionic conductance of the NT lumen on the NT length is used to find the luminal radius characterizing the mean curvature of the NT membrane (Step 19A). The long-term stability of the usNT is subsequently assessed (Step 19B). Next, the lateral tension and bending modulus $(k)$ of the NT membrane are recalculated from nonlinear current/voltage characteristics of the luminal conductance (Step 19C). Following the quality control step, the SNR characteristic of the usNT conductance is measured (Step 20). Of note, for various lipid compositions tested here, the bending moduli match published values. In this protocol, the value of the bending modulus is considered to be a quality control parameter that confirms the bilayer structure of the NT and reservoir membranes. Whereas $k$ is a material constant, the lateral tension of usNT membranes varies from grid to grid, explaining the wide range of the usNT radius.

Assessment of the protein delivery system (Steps 21–31)

Although the volume of the observation chamber can be made small, bulk protein addition seems impractical when the action of only a few molecules is to be tested. Unless concentration dependence of the protein action is to be assessed, it is preferable to add the protein locally, via free diffusion from the 1-μm-wide tip of a delivery pipette brought into proximity of the usNT (Fig. 2). The local concentration and thus, membrane coverage can be altered by changing the distance between the NT position and the delivery pipette (Fig. 2c). To verify the delivery protocol, reversible NT squeezing by a hyperosmotic solution can be performed.

Characterization of protein–membrane interaction (Steps 32–42)

The primary goal of this protocol is to make usNTs work as molecular sensors to resolve the adsorption of individual protein molecules. This is achieved by optimizing the usNT geometry and adjusting the position of the protein delivery pipette (Steps 32–40). The optimization procedure is explained by comparing the conductance patterns obtained with wild-type Dyn1, the Dyn1 mutant defective in oligomerization, and the ENTH domain. Next, the dependence of the amplitude of the individual conductance changes on the usNT curvature is characterized (Steps 41–49). Fitting of this dependence with a prototype model equation yields the geometric parameters of a distinct membrane deformation.

The method is primarily aimed at characterization of local usNT deformations by individual proteins, such as the reversible usNT constriction by small Dyn1 oligomers in the presence of GTP²¹. Importantly, the geometric parameters of the local deformation extracted by the method for Dyn1 match those predicted by cryo-EM analyses⁴⁹. However, the effect of the protein adsorption on the bulk mechanical properties of the usNT membrane can be analyzed as well, albeit at the limit of sparse protein coverage (Step 50), Importantly, molecular parameters, such as intrinsic curvature of a proteo-lipid complex, can be extracted from such ensemble measurements.

Materials

Reagents

Organic solvents

! CAUTION Special care should be taken while handling organic solvents. Carefully read and follow all the precautionary instructions and handling directions given by the supplier ▲CRITICAL The purity of all the organic solvents must be >99% (e.g., gas chromatography, HPLC or analytical grade) to prevent trace contamination effects on the bilayer lipid membrane (BLM).

• Squalene (Sigma-Aldrich, cat. no. 442785)

• Octane (Sigma-Aldrich, cat. no. 74820)

• Decane (Sigma-Aldrich, cat. no. 30540)

• Chloroform (Sigma-Aldrich, cat. no. 1.02445) ! CAUTION Chloroform is highly toxic and a suspected cancer hazard. Carefully read and follow all the precautionary instructions and handling directions given by the supplier

• Ethanol (Merck Millipore, cat. no. 100983)

• Methanol (VWR International, cat. no. 83966.320P)

Gases

• Argon compressed gas

• Nitrogen compressed gas

Lipid stocks

! CAUTION Special care should be taken while handling organic solvents. Carefully read and follow all the precautionary instructions and handling directions given by the supplier. ! CAUTION Chloroform is highly toxic and a suspected cancer hazard. Carefully read and follow all the precautionary instructions and handling directions given by the supplier ▲CRITICAL Store lipid stocks in chloroform or chloroform/methanol at −20 °C in amber glass vials with PTFE (polytetrafluoroethylene) caps in a vertical position. Follow the supplier’s instructions for the maximum storage time allowed for each lipid species. Before opening the storage vial, allow it to reach room temperature (RT: 15–25 °C) to avoid water precipitation into the lipid stock. To prevent lipid oxidation, purge the air from the vial with a gentle stream of argon gas for 2–3 s before closing the vial. Protect the vial from accidental opening by wrapping a Parafilm strip around the vial’s cap in the direction of the cap closure. Be careful to remove all the Parafilm before opening the vial again.

• Chloroform stock of DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine; Avanti Polar Lipids, cat. no. 850375)

• Chloroform stock of DOPE (1,2-dioleleoyl-sn-glycero-3-phosphoethanolamine; Avanti Polar Lipids, cat. no. 850725)

• Chloroform stock of DOPS (1,2-dioleoyl-sn-glycero-3-phospho-l-serine; Avanti Polar Lipids, cat. no. 840035)

• Chloroform/methanol stock of brain PI(4,5)P₂ (l-α-phosphatidylinositol-4,5-bisphosphate (brain, porcine); Avanti Polar Lipids, cat. no. 840046)

• Cholesterol (Avanti Polar Lipids, cat. no. 700100)

Reagents for buffer preparation

• KCl (Sigma-Aldrich, cat. no. P9541)

• EDTA (Sigma-Aldrich, cat. no. EDS)

• HEPES (Sigma-Aldrich, cat. no. 83264)

• HCl (Sigma-Aldrich, cat. no. H1758)

• NaOH (Sigma-Aldrich, cat. no. 415413)

• Milli-Q water (18.2 MΩ·cm; Millipore)

• MgCl₂ solution in H₂O (1 M; Sigma-Aldrich, cat. no. 63069)

• KOH (Sigma-Aldrich, cat. no. 417661)

Other reagents

• Polydimethylsiloxane (PDMS) elastomer (Sylgard 184 Silicone Elastomer Kit; Dow)

• Ficoll PM 70 (Sigma-Aldrich, cat. no. F2878)

• Purified proteins or other molecules of interest. In the example described in this protocol, we use Dyn1 (wild type and R399A mutant) and epsin N-terminal homology domain (ENTH), which were purified as described in^21,26,47

• (Optional) For Dyn1: high-purity guanosine 5′-triphosphate (GTP) lithium salt (Sigma-Aldrich, cat. no. G5884)

• Bleach

Equipment

Preparation of planar lipid membrane reservoirs

• Amber glass threaded vials (4 mL) with attached PTFE caps (Fisher Scientific, cat. no. 11583542) ▲CRITICAL Before use with lipid stocks, the vials should be cleaned with ethanol and then rinsed with chloroform to avoid dust and organic matter contamination.

• Gilder TEM (transmission EM) grids (200 mesh, 3.05-mm o.d., copper; Ted Pella)

• Round glass coverslips (25 mm; no. 1 thickness; Karl Hecht, cat. no. 41001125)

• Low-profile microscopy chamber for 25-mm round cover glasses (Warner Instruments, cat. no. RC-40LP)

• Red sable brushes (3/0; Raphaël Kolinsky, cat. no. 8404) ▲CRITICAL Natural bristles (such as Kolinsky sable) are the choice for BLM painting because of their relative stability in organic solvents. Avoid using synthetic bristles.

• Vacuum pump (Vacuubrand, model no. MZ 2C NT+AK+EK)

• Crystallization dish (90 mm diameter; Pyrex 90 × 50 mm)

• Common cutter knife

• Hotplate or hair dryer

• Sticky tape

• High-precision forceps with Teflon coating for TEM grid handling (Electron Microscopy Sciences, cat. no. 78325-SSA)

• Cover glass forceps (PTFE; Electron Microscopy Sciences, cat. no. 72977)

• Gastight Hamilton syringes of different volumes for lipid handling (Hamilton, e.g., Gastight 1700 series)

Fabrication and filling of patch-clamp micropipettes

▲CRITICAL Select the proper glass type for each membrane target. For details, see ‘Equipment setup‘.

• Micropipette puller (Sutter Instruments, model no. P-2000)

• Micropipette storage jar (World Precision Instruments, cat. no. E215)

• Inverted microscope with 20× or 40× air objective lenses (Zeiss Instruments, model no. ID 03)

• Disposable syringes (1 mL; Sigma-Aldrich, cat. no. Z683647)

• Syringe filters (0.2-μm pore size; GE Healthcare, cat. no. 6780-2502)

• MicroFil syringe needles (28-gauge and 34-gauge; World Precision Instruments, cat. nos. MF 28G-5 and MF 34G-5)

• Bath sonicator (Fisherbrand, model no. FB15051)

• Plain silica beads (5 μm; Microspheres-Nanospheres, cat. no. 140226)

• Fluorescence beads with sizes <20 nm and fluorescence spectrum compatible with the existing fluorescence cubes of the microscope (Molecular Probes, cat. no. F8888)

• Thin-wall borosilicate glass (Science Products, cat. no. GB150-10 or World Precision Instruments, cat. no. TW150F-4

• Quartz capillary glass (Sutter Instruments, cat. no. Q150-110-10)

Patch-clamping and NT pulling

• Antivibration table (Newport, model no. VH3036W-OPT)

• Inverted microscope with 60× air objective (Olympus, model no. IX-81)

• Patch-clamp amplifier (Molecular Devices, model no. Axopatch 200B)

• Amplifier headstage (Molecular Devices)

• Micropipette holder (with port, for 1.5-mm glass capillaries, 45° angle; Science Products, model no. HUW-P1.5-M45) ▲CRITICAL Use bent micropipette holders (see Fig. 2a) to make the patch clamp-pipette orientation as close as possible to vertical.

• Flexible Tygon tubing with a 2-mm i.d., matching that of the port of the micropipette holder (Saint-Gobain, cat. no. ACF1S1502). The tubing is to be attached to the holder port as shown in Fig. 2a.

• Two-way valve and connectors for Tygon tubing (Büerkle)

• Ag/AgCl pellet electrode (Science Products, cat. no. 550015)

• Silver chloride bath electrode prepared from Teflon-coated silver wire, 75-μm uncoated diameter (Science Products, cat. no. AG-3T)

• Adjustable miniature magnetic holders (Bioscience Tools, cat. no. MA-MTH) or similar

• Motorized coarse micromanipulation system (referred to as a ‘micromanipulator’ below) consisting of three linear stages (Newport, model no. 461-XYZ) moved by three NewStep stepper motors (Newport, model no. NSA12) controlled by a NewStep controller (Newport, model no. NSC-SB) ▲CRITICAL Precise, remotely controlled micromanipulators greatly simplify patch-clamp experiments. Routinely test the system for long-term position stability and precision of fine movements, using high-resolution microscopy.

• Precise nanopositioning system (referred to as a ‘nanoactuator’) consisting of a calibrated piezo linear actuator and controller (piezo linear actuator and actuator controller; Newport, model no. ESA-CSA or later) ! CAUTION Do not overload the piezo-actuator. The maximum load should not be more than the value declared by the manufacturer ▲CRITICAL The piezo actuator should be mounted on top of the coarse micromanipulator driver along the z axis of the xyz linear stage. To produce ultra-short NTs, the piezo actuator should have a step size of <30 nm per step, allowing for tiny, controlled movements. The piezo actuator should be calibrated so that its values, being digitized, can be converted to the vertical displacement along the z axis.

• Oscilloscope (Tektronix, model no. TDS 220)

• Signal generator (Good Will Instrument, model no. GOS-620FG)

• Faraday cage (A Faraday cage (approximately 30 × 30 × 30 cm) is needed to house the amplifier headstage; this can be built in-house. As an alternative, aluminum foil can be used to screen the noise around the amplifier headstage locally.)

Protein application

• Micropipette holder (for 1.5-mm glass capillary with port, straight; Science Products, model no. HUW-P1.5)

• Motorized coarse micromanipulation system (referred to as a ‘micromanipulator’ below) consisting of three linear stages (Newport, model no. 461-XYZ) moved by three NewStep stepper motors (Newport, model no. NSA12) controlled by a NewStep controller (Newport, model no. NSC-SB)

Data acquisition and analysis

• PC with data acquisition system (Molecular Devices, model no. Axon Digidata 1550)

• AxoScope software (Molecular Devices: http://mdc.custhelp.com/app/products/detail/p/95)

• Data analysis: pCLAMP software (Molecular Devices): https://www.moleculardevices.com/products/axon-patch-clamp-system/acquisition-and-analysis-software/pclamp-software-suite

• OriginLab software (Origin v.8.1. or higher: https://www.originlab.com/)

Reagent setup

PDMS elastomer preparation

Follow the manufacturer’s instructions to prepare the elastomer for its application as support for the TEM grid. Briefly, mix 5 g of the elastomer base with 0.5 g of the curing agent (or another quantity at a 1:10 base/curing agent ratio) in suitable disposable plasticware and stir well to homogenize the mixture. Next, carefully degas the mixture under vacuum until no bubbles are formed. It may be necessary to open the vacuum to avoid spills at this step. Do so slowly to prevent the container from being knocked over. The mixture of the elastomer base with the curing agent should be prepared immediately before use ! CAUTION Uncured elastomer is very viscous and may be irritate the skin. Use gloves for protection and protect all the surfaces that may enter into contact with the elastomer during its preparation.

Working buffer

We use a standard ionic solution containing 150 mM KCl, 10 mM HEPES and 1 mM EDTA. Adjust the pH of the buffer to 7.0 using 1 M KOH. KCl is routinely used in patch-clamp experiments because the similar mobilities of K⁺ and Cl⁻ ions in water facilitate interpretation of conductance measurements. This solution can be stored at 4 °C for up to 1 month. The base buffer is modified to suit the needs of the proteins used in the assay; for example, 2 mM MgCl₂ can be added to support the GTPase activity of Dyn1⁴⁴ ▲CRITICAL Always use freshly prepared double-distilled (Milli-Q-grade) water for buffer preparation. All solutions must be filtered through 0.2-μm pore filters ▲CRITICAL Modification of the base can affect the lipid bilayer formation and behavior. Always test the long-term stability of usNTs upon changing the buffer.

Solvents for the brush cleaning

Transfer pure methanol, chloroform and Milli-Q water to clean 50-mL vials. It is advisable to have two vials of each to provide a precleaning and a cleaning set of solvents for the bristles of the brushes.

Lipid solution for pretreatment of the grid mesh

To prepare the pretreatment solution, mix lipid stocks in chloroform in a clean glass vial to the desired molar ratio. For example, use DOPC/DOPE/cholesterol/DOPS/PI(4,5)P₂ at a 39:10:30:20:1 (mol/mol/mol/mol/mol) ratio as the reference mixture to study membrane remodeling by Dyn1 or ENTH^21,26,47. Evaporate the solvent using a gentle N₂ gas stream followed by vacuum desiccation for 0.5–1 h. Dissolve the dried lipid film in a 1:1 decane/octane (vol/vol) mixture to a final lipid concentration of 10 g/L. It is preferable to prepare 30–60 μL of the mixture on the day of the experiment, although the mixture can be stored at −20 °C for a week.

Lipid solution for the production of planar bilayer reservoirs by painting technique

Prepare the same lipid mixtures as for the pretreatment solution and evaporate the solvent under an N₂ stream followed by vacuum desiccation for 0.5–1 h. Dissolve the lipid mixture in 50–60 μL of squalene solvent to a final lipid concentration of 10g/L (high tension) and 50g/L (low tension). This painting mixture should be freshly prepared on the day of the experiment ▲CRITICAL Store the lipid solutions for up to 1 week at −20 °C in a closed vial purged with Ar₂ or N₂ gas to prevent lipid oxidation; on the day of the experiment, keep the lipid solutions at 4 °C.

Protein solution

Quickly thaw an aliquot of the protein of interest, dialyze it against the base buffer (or the base buffer modified according to the protein needs), keep on it ice and discard it at the end of the experiment.

Equipment setup

TEM grid cleaning

The TEM grids should be cleaned of possible organic matter by soaking them in ethanol and then rinsing with chloroform. Allow them to air dry while protecting from dust and store in a closed container.

Preparation of the observation chamber

The chamber consists of a TEM grid elevated and firmly fixed to the center of a cover glass of the microscopy chamber. Assembly is as follows:

1. To fix the grid to the cover glass, use the PDMS elastomer as shown in Fig. 3. Attach a ~2.5-mm-wide and ~30-mm-long ribbon of sticky tape to the center of a 25-mm round cover glass.

2. Place a small (<20 μL) drop of uncured PDMS on top of the ribbon in the center of the cover glass.

3. Cure the PDMS at 80 °C, applying heat from below the cover glass with a hair dryer or placing the cover glass on a hotplate. Do this for ~10 min. The PDMS should stop behaving as a viscous fluid and become almost solid. Check this by gently touching the rim of the PDMS drop with a clean pipette tip. If the PDMS attaches and is dragged by the tip, continue curing the PDMS. ! CAUTION If a hair dryer is used at this step, avoid skin burning by handling the cover glass with PTFE covered forceps ▲CRITICAL Uncured PDMS is very viscous. Use a trimmed 200-μL pipette tip to handle it. The drop placed on the sticky tape should be minimal, because the PDMS will expand on the glass surface during the curing procedure, and bigger PDMS quantities may overflow the cover glass.

4. When the PDMS becomes almost completely cured, remove the sticky tape strip by carefully pulling one of the strip’s ends to form a 2-mm channel between two PDMS islands (Fig. 3c).

5. Deposit a few EM grids over a clean surface, such as the inner side of a Parafilm lining, by carefully tapping the EM grid container.

6. Gently take one EM grid by the edge with the high-precision forceps. Use the neoprene O-ring provided with the tweezers to hold the tweezers in the closed position.

7. Place the grid in the center of the channel by attaching it to the PDMS on two opposite rims.

8. Complete the curing of the PDMS with an additional 3–5 min of heating with the hair dryer/hotplate (Fig. 3d) or by leaving at RT for 30 min ▲CRITICAL The grid should be fixed at a height such that it can be observed with the selected microscope objective (usually a 40× or a 60× air objective), while being distant enough from the cover glass bottom to prevent grid–glass contact and lipid spills onto the cover glass during the BLM painting procedure.

9. Finally, mount the cover glass with the attached grid in the microscopy chamber and set it on the inverted microscope stage (Fig. 3e,f). Note that if the PDMS is below the optimal curing point, it could transfer to the grid and interfere with production of the planar membrane reservoir (Procedure, Steps 3–6); if the PDMS is above the optimum curing point, the grid will not be fixed with the required robustness, and it may detach during the membrane painting (Procedure, Step 6; see Troubleshooting section).

Fig. 3 ∣. Step-by-step procedure for the preparation of the experimental chamber.

a, A small drop of PDMS is placed on top of a sticky tape ribbon in the center of a clean cover glass. b, The PDMS drop is heated with a hair dryer. c, An ~2-mm-wide channel is formed by peeling the sticky tape away from the cover glass. d, A TEM grid is placed in the center of the channel and fixed by applying additional heating. e, The cover glass containing the grid is set into the microscopy chamber. f, An example of the final configuration of the chamber set on the microscope stage.

Ground and measuring electrodes

Use Ag/AgCl pellet electrode as a reference electrode (see Fig. 2a). Take proper care of the pellet electrode after each use by rinsing it with EtOH and soaking it in pure water for several minutes. Use thin Teflon-coated silver wire for the measuring electrode. Carefully peel off the protective Teflon film from the end (~5 mm) of the wire and rinse it with EtOH and pure water. Chlorination of the wire should be performed by soaking it in an undiluted bleach solution for 10–15 min. If the wire has been previously treated with chloride, the old coating must be completely removed with diluted HCl. Mount the measuring electrode inside the 45° angle holder for the patch-clamp micropipette.

Patch-clamp and delivery pipettes

Several different types of glass should be tried to determine which is the optimal one for a particular membrane composition. For painting the membranes, we usually use thin-wall borosilicate glass. In experiments with membranes containing high amounts of charged lipids, we suggest using quartz capillary glass. Heat-polishing of the tip of the patch-clamp pipette is optional but can improve the probability of the ‘gigaseal’ formation. Establish a proper protocol for micropipette fabrication by following the guidance from the micropipette puller manufacturer (e.g., Sutter Pipette Cookbook: https://www.sutter.com/PDFs/pipette_cookbook.pdf). The aim is to produce pipettes with a standardized tip diameter. The diameter of the tip can be estimated from the electric resistance of the pipette. Typically, we use patch-clamp pipettes with an electrical resistance of 3–6 MΩ when filled with the base buffer.

The delivery pipettes (required for local protein application to the NTs) are fabricated in the same way as the patch-clamp pipettes. The tip of the delivery pipette should be slightly wider, with the corresponding electrical resistance of 0.5–1 MΩ in the base buffer. Ideally, the pipettes should be fabricated immediately before the experiment. If pipettes are prepared earlier, they should be kept in a sealed jar to minimize the accumulation of contaminants on the pipette tips ▲CRITICAL Fill both the patch-clamp and the delivery pipettes right before using them (Procedure, Steps 7 and 21, respectively). Backfill the pipettes to the level defined by the capillary force to avoid water fluxes through the tip. Extra care should be taken to exactly match this level and avoid water fluxes in/from the delivery pipette, which can interfere with the NT constriction measurements. To determine the capillarity level, dip the capillary tube (used for the patch-clamp pipette preparation) into the bathing solution and measure the height to which the liquid rises inside the tube.

Electrical noise check

All the equipment, especially the microscope body, antivibration table and isolated parts, should be electrically grounded. The setup should also be shielded from the external noise by a Faraday cage. Instead of a large cage enclosing the whole setup, local shielding constructed around the microscope stage can be used. Such shielding protects from nearby sources of noise, such as electrical micromanipulators. Metal sheets or foil, electrically connected to the grounding point, can be used for such local shielding. We recommend using an oscilloscope for the initial inspection of the noise level. If the equipment is correctly grounded and electrically isolated, with an open input of the amplifier (without the micropipette holder), the current output should yield a flat baseline with uniform background noise at high gain (e.g., 100 mV/pA). The random mean square amplitude of the noise should be ~0.1 pA. See Troubleshooting section.

Automated protocols for nanotube conductance measurements

To analyze the ionic permeability (conductance) of the NT, the voltage-clamp mode of the patch-clamp amplifier is used. In this mode, a constant or time-dependent voltage $(U)$ is applied to the measuring electrode (Fig. 1b) by the patch-clamp amplifier. Different voltage signals are used at different stages of the experiment. A triangle-wave voltage signal is applied during the membrane patching (Steps 14–16). Constant $U$, ranging from 50 to 200 mV, is applied to control the NT formation (Steps 17–18), assess the NT geometry and deformations (Steps 19A and B, 20, 24–28, 30 and 31) Voltage ramps and square-wave sequences are applied to obtain current/voltage characteristics of the NT conductance (Step 19C). The corresponding voltage protocols are created using the PClamp software controlling the acquisition board (Digidata 1550) analog output, which is used as an external signal by the patch-clamp Axopatch 200B amplifier. For manually controlled amplifiers, the protocols can be created and applied using a function generator. The current through the measuring electrode placed inside the patch-clamp pipette is collected (filtered and digitized) using the amplifier’s intrinsic filter and the analog-to-digital converter of the Digidata 1550 acquisition board. The acquisition rate is defined by the characteristic times of the membrane remodeling process. The corner frequency of the internal filter is set at approximately half of the rate. Typically, a 2-kHz acquisition rate and a 1-kHz filter corner frequency can be used. A faster acquisition might be required to resolve fast current changes during, for example, NT membrane fission.

Procedure

Pretreatment of the grid mesh for reservoir membrane production ● Timing 10 min

1. Use a suitable dispenser to deposit a 0.5- to 1-μL drop of the grid pretreatment solution (Reagent setup) onto the center of the grid, avoiding the PDMS support. Use a low-magnification mode (20–40×) of the inverted microscope to control the solution application. Apply a weak N₂ or Ar₂ gas stream until a dry lipid film on the grid mesh becomes visible (usually 1–2 min). If needed, another drop of the lipid mixture can be applied.

2. Carefully fill the observation chamber with the working buffer until the grid mesh is completely covered. Check whether the space under the grid is properly filled. If air bubbles are trapped under the grid, pipette them out.

Making multiple reservoir lipid membranes using the painting technique ● Timing 15 min

• 3.Clean the brushes by immersing the bristles sequentially in Milli-Q water, methanol and, finally, chloroform solvents to remove any organic traces.

  ! CAUTION Chloroform is highly toxic and a suspected cancer hazard. Carefully read and follow all the precautionary instructions and handling directions given by the supplier.

  ! CAUTION Special care should be taken while handling organic solvents. Carefully read and follow all the precautionary instructions and handling directions given by the supplier.

  ▲CRITICAL STEP It is important that only the bristles be immersed in the solvents, because the rest of the brush may contaminate the samples because of solubility of its components in organic solvents.

• 4.Dry the bristles under an N₂ gas stream.
• 5.Immerse the brush bristles into the membrane painting mixture (Reagent setup) and take up a small amount.
• 6.Under microscope observation (use low magnification, a 10× or 20× objective), bring the brush above the grid mesh and gently smear the lipid solution over the mesh, avoiding the PDMS support. A thick film of the lipid solution should cover the holes. In some of the holes, the lipid bilayer forms via a nucleation process when the thick film spontaneously thins at a random location, transforming into a small lipid bilayer. The bilayer then expands, pushing the excess of the solution to the rim (Fig. 2b, lipid reservoir). The whole process is readily visible by phase-contrast microscopy. Generally, stable bilayers form in 50–70% of the holes.

  ? TROUBLESHOOTING

Pulling the lipid membrane nanotube ● Timing 10 min

• 7.Fill a patch-clamp pipette with the working buffer (Reagent setup). Use a 1-mL syringe with a MicroFil needle. Insert the needle into the pipette, move it close to the pipette tip, apply pressure and fill the pipette so that the length of the filled part is equal to the capillarity height (see ‘Equipment setup‘, ‘Patch-clamp and delivery pipettes’). Check for air bubbles trapped near the pipette tip. If you detect a bubble, try to remove it by gently tapping on the pipette with a finger.
• 8.Insert the pipette into the 45°-angle patch-clamp pipette holder (Fig. 2a) and fix it tightly.
• 9.Set a low gain on the patch-clamp amplifier (0.1 mV/pA).
• 10.Apply a weak positive pressure through the holder port by blowing into the pipette tip attached to the Tygon tubing (Fig. 2a). Close the two-way valve to hold the pressure.
• 11.Using the coarse micromanipulator, move the pipette above the center of one of the mesh holes containing a lipid bilayer (Fig. 2b). The pipette tip is easily resolved by a bright-field microscope at low magnification (10× or 20×) (Fig. 2b). Bring the pipette close to the membrane but avoid direct contact.
• 12.Compensate for the electrode offset using the standard compensatory circuit of the patch-clamp amplifier.
• 13.Open the two-way valve to release the pipette pressure (see Step 10).
• 14.Apply triangle-wave voltage, −50 mV to 50 mV, 1 V/s. Verify that the measured current $(I_{\mathrm{m}})$ has no capacitive component indicating pipette clogging: the current should have the same triangle-wave pattern as the voltage.
• 15.Using the precise nanoactuator, carefully lower the pipette until its tip touches the membrane, as seen by phase-contrast optics and/or as a small decrease of the measuring current amplitude. Stop the pipette movement. At this point, the tight contact between the membrane and the tip, known as the ‘gigaseal’, should form spontaneously. The gigaseal formation is seen as an abrupt decrease of the current amplitude to zero.

  ▲CRITICAL STEP Check the quality of the patch before pulling the NT. For that, increase the patch-clamp amplifier gain to 50 mV/pA. Apply different potentials, ranging from 50 to 200 mV, to the measuring electrode ($U$, Fig. 1b). Check that the seal conductance $G_{\text{seal}}=I_{\mathrm{m}}∕U$ is in the 10- to 100-pS range and does not depend on $U$. Decrease the gain back to 0.1mV/pA.

  ? TROUBLESHOOTING

• 16.Rupture the membrane patch isolated inside the tip of the patch-clamp pipette by applying either a negative pressure pulse to the patch-clamp pipette interior (through the holder’s port) or a short high-voltage pulse using the ZAP function of the patch-clamp amplifier. Once the membrane patch is ruptured, the triangle-wave current shape is restored.
• 17.Apply constant $U=100\text{mV}$. Using the nanoactuator, slowly (~0.05–0.1 μm/s) move the pipette up while monitoring $I_{\mathrm{m}}$. An abrupt decrease of $I_{\mathrm{m}}$ to approximately background level indicates NT formation. Once the current drop occurs, stop the pipette movement.
• 18.In 10–50% of cases, depending on the lipid composition, the NT breaks during/quickly after formation. To quickly verify that the NT remains, increase the amplifier gain to 100 mV/pA and, using the nanoactuator, slowly move the pipette down closer to the membrane. If the NT remains, a gradual increase of $I_{\mathrm{m}}$ will be seen.

  ? TROUBLESHOOTING

Nanotube calibration

• 19.At this point, identify the geometric and elastic characteristics of the NT. Follow option A for determination of the NT geometry, option B for assessment of the long-term stability of the NT and option C for determination of the mechanical parameters of the NT membrane.

(A). Determination of the NT geometry ● Timing 10 min

1. Activate the main acquisition protocol. The position of the nanoactuator, $U$ and $I_{\mathrm{m}}$ will be recorded simultaneously. The amplifier gain is set at 50–100 mV/pA, $U=100\text{mV}$

   ▲CRITICAL STEP For manually controlled patch-clamp amplifiers, the value of the amplifier gain and the settings of the filters must be recorded.

2. Complete three to five consecutive cycles of NT elongation and shortening, using the nanoactuator. Stop the acquisition.

   ▲CRITICAL STEP When moving the pipette down, avoid touching the lipid membrane. Bringing the pipette too close could trigger spontaneous NT expansion²⁴, seen as an abrupt and large current increase, which generally causes the amplifier to overload. Should that occur, decrease the amplifier gain to 0.1mV/pA and repeat the process, starting from Step 17.

3. Process the data. Select the segments of the $I_{\mathrm{m}}$ record corresponding to NT shortening and elongation, each segment showing a hyperbolic dependence of the current on the NT length. Follow the procedure described in Box 1 to determine the NT radius and the position of the nanoactuator corresponding to zero NT length.

   ▲CRITICAL STEP Make sure that there are no uncontrollable changes of NT length, that is, that the offset of the BLM position measured as described in Box 1 remains unaltered within error limits.

   ? TROUBLESHOOTING

Box 1 ∣. Determination of the radius of the NT lumen ● Timing 20 min.

1. Calculate the sum of the NT conductance $(G_{\text{NT}})$ and the conductance of the gigaseal ($G_{\text{seal}}$, where $G_{\text{NT}}=1∕{\text{Re}}_{\text{NT}}$ and $G_{\text{seal}}=1∕{\text{Re}}_{\text{seal}}$; see Fig. 1b), from the current measured by the patch-clamp amplifier ($I_{\mathrm{m}}$, Fig. 1b): $G_{\mathrm{m}}=I_{\mathrm{m}}∕U=G_{\text{NT}}+G_{\text{seal}}$. According to Ohm’s law for a cylindrical conductor, the NT conductance depends inversely on the tube length, so plot the dependence of $G_{\mathrm{m}}=I_{\mathrm{m}}∕U$ on the vertical displacement of the pipette, $\Delta L$, and find the vertical and horizontal asymptotes of $G_{\mathrm{m}}(\Delta L)$. The horizontal asymptote defines $G_{\text{seal}}$. Subtract $G_{\text{seal}}$ from $G_{\mathrm{m}}$ to obtain $G_{\text{NT}}:G_{\text{NT}}=G_{\mathrm{m}}-G_{\text{seal}}$. The vertical asymptote, $\alpha$, defines the offset between the NT length $(L_{\text{NT}})$ and $\Delta L$ measured by the controller of the nanopositioner moving the patch-clamp pipette in the vertical direction: $L_{\text{NT}}=\Delta L-\alpha$.

   

2. Replot $G_{\mathrm{m}}(\Delta L)$ as $r_{\mathrm{m}}(L_{\text{NT}})$, where $r_{\mathrm{m}}=\sqrt{G_{\text{NT}}{L}_{\text{NT}}{\rho }_{\text{NT}}∕\pi }$ and ${\rho }_{\text{NT}}$ is the specific resistance of the electrolyte in the NT lumen. For NTs containing no charged lipid species, the ion concentration in the lumen is assumed equal to the bulk concentration: ${\rho }_{\text{NT}}={\rho }_0$ (where ${\rho }_0$ is the bulk specific resistance, ${\rho }_0=66.7\Omega \times \text{cm}$ for 150 mM KCl solution). Define the radius of the NT lumen $r_{\text{NT}}$ by fitting $r_{\mathrm{m}}(L_{\text{NT}})$ with a constant; Gaussian distribution of the residuals confirms the quality of the fit (see inset in figure above). The approximation procedure yields $r_{\text{NT}}$ with P <0.001 (generally). Make sure that the fit works robustly when the NT length is reduced to less than a micron.

Note: Negatively charged lipids, such as DOPS and/or PIP², are typically required for recruitment of membrane-remodeling proteins from cytoplasm to the membrane. In the NT lumen, the membrane surface charge is balanced by the net charge of the electrolyte, leading to an increase of the electrolyte concentration in the lumen. To find the specific resistance of the electrolyte inside the lumen of a charged NT, use Eq. 1 (ref. ²¹):

$$\frac{{\rho }_0}{{\rho }_{\text{NT}}}=\sqrt{1+{\left(\frac{\gamma }{{er}_{\text{NT}}{c}_0}\right)}^2}$$ (1)

where $\gamma$ is the surface charge density of the NT internal monolayer, e is the electron charge, and $c_0$ is the bulk electrolyte concentration/activity.

(B). Assessment of the long-term stability of the NT ● Timing 10 min

1. Protein-driven membrane constriction and/or fission, measured at a fixed NT length, can take substantial time (10–100 s) to complete. Verify that, before the protein addition, the current baseline corresponding to the conductance of the pure lipid NT remains stable at longer time scales (hundreds of seconds), showing no drift or transient changes.

   ? TROUBLESHOOTING

(C). Determination of the mechanical parameters of the NT membrane ● Timing 10 min for each new grid

1. Set the NT length at the desired value. Obtain current/voltage characteristics using voltage ramps (from 50 to 200 mV, 10–20 mV/s, Fig. 4a) or 50-mV incremental/decremental steps (Fig. 4b).

   ? TROUBLESHOOTING

2. Repeat Step 19C(i) at a different length, cover the usNT length range (100–400 nm).

3. Follow the procedure described in Box 2 to determine the lateral tension and bending modulus $(k)$ of the NT membrane (Fig. 5).

   ▲CRITICAL STEP Verify that the lateral tension is ~10⁻³N/m. Such a high value and also large variation of lateral tension (Fig. 5, inset) determine the range of NT curvature (3–12 nm, Fig. 5, inset) suitable for detection of protein-driven deformations of NT described below.

   ? TROUBLESHOOTING

Fig. 4 ∣. The nonlinear current-voltage $(\mathit{I}∕\mathit{U})$ characteristics of the usNT lumen.

a, An example of the $I∕U$ characteristics obtained by applying a voltage ramp; the inset shows the nonlinear component of the $I∕U$ curve. b, An example of the $I∕U$ dependence obtained by applying 50-mV voltage steps; the inset shows the corresponding dependence of the normalized NT conductance on the voltage. The conductance increase with voltage indicates the NT widening by the electric field.

Box 2 ∣. Determination of the elastic constants of the NT membrane ● Timing 30 min.

The electric field in the NT lumen affects its shape as shown by Eq. 2 (see ref. ²⁶):

$$\frac{1}{r_{\mathrm{m}}^2}=\frac{2\sigma }{k}-\frac{C_{\text{sp}}{U}^2}{3k}$$ (2)

where $\sigma$ is the lateral tension, $k$ is the mean curvature bending modulus and $C_{\text{sp}}$ is the specific electrical capacitance of the usNT membrane. Eq. 2 is used to fit $r_{\mathrm{m}}^{-2}(U^2)$ (obtained experimentally) and to determine the $\sigma$ and $k$ values.

Procedure

1. From the $I∕U$ characteristics (Fig. 4) obtain $G_{\text{NT}}(U)=G_{\mathrm{m}}(U)-G_{\text{seal}}$ (see Box 1). Verify that $G_{\mathrm{m}}$ does not depend on $U$ at large $L_{NT}$ when $G_m=G_{seal}$ (see the graph in Box 1)

2. Plot $r_{\mathrm{m}}^{-2}(U^2)$, where $r_{\mathrm{m}}(U)=\sqrt{G_{\text{NT}}(U){\rho }_{\text{NT}}{L}_{\text{NT}}∕\pi }+h$ and $h=2\text{nm}$ is the thickness of lipid monolayer. The upper curve corresponds to Fig. 4a (obtained with a voltage ramp), whereas the sequence of voltage steps (as in Fig. 4b) was used to obtain the lower curve. At each constant $U<r_{\mathrm{m}}(U)>$ was found as described in Box 1.

3. Find lateral tension $\sigma$ and bending rigidity $k$ of the NT membrane from the linear approximation of $r_{\mathrm{m}}^{-2}(U^2)$ using Eq. 2.

   ▲CRITICAL STEP The ramp (upper curve) and steps (lower curve) methods should yield similar slopes of $r_{\mathrm{m}}^{-2}(U^2)$. Whereas the ramp method enables real-time monitoring of $k$ and is used to assess dynamic protein effects, the step method is more precise and is used to discriminate small changes of $k$ in stationary situations.

4. Repeat steps 1–3 at different $L_{\text{NT}}$ values to verify that the elastic parameters do not depend on the NT length.

Fig. 5 ∣. Box plot of the bending rigidity moduli $(\mathit{k})$ obtained by the procedure described in Box 2.

For the lipid compositions shown here, the method recovers the published values of the bending modulus for DOPC/Chol and POPC/Chol lipid mixtures^50,51. The inset shows the distribution of NT luminal radius (red) and lateral tension (black) measured for the DOPC/Chol 7:3 molar mixture. The boxes represent IQR (25th–75th percentiles) and the whiskers represent 1–99 percentiles. The number of data points for the main graph n = 16 (blue), n = 21 (black), n = 23 (red); for the inset n = 23 (red), n = 23 (black). Chol, cholesterol. POPC, 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine. Source data

Determination of signal-to-noise ratio ● Timing 40 min for each new grid

• 20.Follow the procedure described in Box 3 and see Fig. 6 to determine the dependence of the SNR on the NT length. Note that SNR depends on the usNT radius; hence, the procedure must be repeated when the membrane elastic constants (Box 2) change.

  ? TROUBLESHOOTING

Box 3 ∣. Obtaining the dependence of the signal-to-noise ratio on the NT length ● Timing 40 min.

Procedure

1. Apply $U=100\text{mV}$ to the measuring electrode.

2. Find the SNR, defined as the ratio of the mean to standard deviation of the NT conductance measured from a 1- to 2-s current recording at a constant length:

   $$\text{SNR}(L_{\text{NT}})={\frac{\langle G_{\text{NT}}\rangle }{\delta G}\mid }_{L=L_{\text{NT}}}$$ (3)

3. Change the NT length by a small step (100–500 nm).

4. Repeat steps 2 and 3.

5. Plot $\text{SNR}(L_{\text{NT}})$ (Fig. 6, black) and find the optimal range of NT length that provides maximum sensitivity (SNR > 10) of NT conductance measurements.

Fig. 6 ∣. Signal-to-noise ratio in the usNT conductance measurements.

The black curve (measured on the lower x axis) shows the $\text{SNR}(G_{\text{NT}}∕\delta G)$ dependence on the usNT length, measured as described in Box 3 for the pure lipid usNT before protein addition. The brown curve (measured on the upper x axis) shows the ${\text{SNR}}_{\mathrm{p}}({\Delta G}_{\text{NT}}∕\delta G$; see Step 40) dependence on the usNT length for the tube constriction driven by 20-nm-long Dyn1 helix²¹. The inset shows the increase of ${\text{SNR}}_{\mathrm{p}}$ with the ratio of the protein and usNT curvatures.

Testing the protein delivery system ● Timing 10–20 min for each new grid

▲CRITICAL At this point, the protein-delivery system (Fig. 2) must be tested.

• 21.Using the procedure described in Step 7, fill the delivery pipette with the working buffer supplemented with a 1 M osmoticant (e.g., Ficoll).

  ▲CRITICAL STEP To diminish volumetric fluxes to/from the pipette due to possible imbalance of hydrostatic pressure at the pipette tip (Step 7), add a small amount of 5 μM plain silica beads to the pipette lumen and let the beads precipitate to the pipette tip.

• 22.Mount the pipette into the delivery pipette holder.
• 23.Set the NT length to 2–3 μm.
• 24.Activate the main acquisition protocol (Step 19A(i)).
• 25.Carefully bring a delivery pipette close to the patch-clamp pipette, aiming at the position shown in Fig. 2b.

  ? TROUBLESHOOTING

• 26.Determine the delivery pipette position where the NT conductance begins to change (mark as position P1).
• 27.Move the pipette closer, aiming at obtaining a 40–50% conductance decrease (mark as position P2). Record the steady-state conductance to verify the baseline stability (Fig. 7a).

  ? TROUBLESHOOTING

• 28.Move the pipette back beyond position P1 and verify that the conductance resumes the initial value (Fig. 7a)
• 29.Stop the acquisition protocol.
• 30.Carefully remove and discard the delivery pipette.

  ▲CRITICAL STEP Use the proper procedure for discarding the glass waste.

Fig. 7 ∣. Representative examples of the usNT conductance behaviors in different experimental conditions.

a, Reversible constriction of the usNT induced by a hyperosmotic (1 M KCl) electrolyte solution. b, Stepwise decrease of the usNT conductance induced by Dyn1 (red curve) and a more gradual usNT constriction R399A mutant of Dyn1 impaired in oligomerization (blue curve). The inset shows the dependence of the number of elementary constriction steps on the usNT length. c,d, Examples of the changes in usNT conductance (blue and red curves) upon variations in the usNT length (black curves) for reversible (c, osmotic stress) and irreversible (d, Dyn1) usNT constriction.

Initial characterization of the protein effect(s) ● Timing 30 min

• 31.Fill the delivery pipette with the protein solution (Reagent setup) as described in Step 7.
• 32.Set the length of the NT to one corresponding to the optimal SNR (Box 3): this creates the usNT.
• 33.Activate the main acquisition protocol as described in Step 19A(i).
• 34.Move the delivery pipette to position P2 as described in Step 27.
• 35.Record the decrease of the usNT conductance until a steady state is achieved. Individual constriction events (steps) are expected to be observed (Fig. 7b).

  ▲CRITICAL STEP Reproducible bulk constriction is required to verify that purified protein species binds to the usNT membrane and, if applicable, remain functional, that is, react to external stimuli such as GTP⁴⁴. Aim for a robust 30–50% conductance decrease.

  ? TROUBLESHOOTING

• 36.Set the usNT length to 400 nm and monitor additional constriction steps.
• 37.Attempt to decrease the usNT length to 100 nm; see whether its conductance follows (Fig. 7c,d). A lack of conductance increase indicates the formation of a large protein scaffold, which restricts the usNT²⁶.

  ▲CRITICAL STEP The radius of reversibly constricted usNTs should not depend on the nanotube length as with a pure lipid nanotube (Box 1). Changes of the radius during elongation/shortening of the usNT indicate polymerization and/or viscoelastic effects⁴².

• 38.Move the delivery pipette beyond position P1 and check whether the conductance returns to that recorded at Step 29 before the protein addition. Full recovery of the conductance (as shown in Fig. 7a for osmotic pressure-driven constriction) confirms reversible usNT constriction, whereas persisting conductance changes indicate irreversible constriction by stable protein scaffolding encaging the usNT membrane.

Preliminary analysis of the conductance steps

• 39.At this point, the usNT system has been installed and calibrated, and discrete conductance changes indicate that the activities of small protein complexes have been recorded. As an initial estimation of the size and localization of membrane deformations causing the conductance steps, assess the dependence of step amplitude and occurrence on usNT length and determine the length with maximal SNR for the step amplitude measurements. For irreversible NT constrictions, follow option A. For reversible NT constrictions, follow option B.

(A). Irreversible NT constriction ● Timing 60 min

▲CRITICAL Here, the new usNT is to be formed from a different membrane on the grid mesh, far from one already used and containing the protein.

1. Carefully remove and discard the patch-clamp pipette.

   ▲CRITICAL STEP Use the proper procedure for discarding the glass waste.

2. Move the observation chamber and find another membrane.

3. Begin the protocol from Step 7 (skip Steps 19C and 20–30 and Box 2).

4. At Step 33 set the usNT length $(L_{\text{NT}})$ to a different value.

5. Stop recording after Step 36.

6. Measure the amplitude $\Delta G$ of the first conductance step (Fig. 8a).

7. Calculate SNR of the $\Delta G$ measurement ${\text{SNR}}_{\mathrm{p}}=(\Delta G∕G_{\text{NT}})\text{SNR}(L_{\text{NT}})$, where $\text{SNR}(L_{\text{NT}})$ is the SNR of the pure lipid usNT (with the same length $L_{\text{NT}}$) obtained as described in Box 3.

8. Repeat Step 39A(i–viii) to cover an 80- to 800-nm length range.

9. Verify that the number of conductance steps is proportional to $L_{\text{NT}}$ (see Fig. 7b, inset).

10. Plot ${\text{SNR}}_{\mathrm{p}}(L_{\text{NT}})$ dependence (see an example in Fig. 6, brown) and find the usNT length corresponding to the maximal ${\text{SNR}}_{\mathrm{p}}$.

Fig. 8 ∣. Representative examples of elementary constriction events produced by individual protein species.

a, Small constriction steps produced by R399A mutant of Dyn1. b, An acute constriction of the usNT by ENTH (blue curve) follows the increase of $U$ from 50 to 200 mV, causing the usNT expansion. The black curve shows the protein-free control. c, Reversible change of the usNT conductance caused by Dyn1 in the presence of GTP²¹. The usNT conductance normalized to its value before the Dyn1 addition ($G_{\mathrm{n}}$; ref. ²¹) is shown. Red lines in a–c represent distinct levels of NT conductance. d, Linear regression of $L_{\text{NT}}(\Delta G∕G)$ on ${(r_{\text{NT}})}^2$ (Eq. 5, Box 4) yields the length (~20 nm) and the curvature $(j_{\mathrm{p}}\sim 0.25{\text{nm}}^{-1})$ of the Dyn1 structure producing the transient membrane deformations shown in c. d adapted with permission from Shnyrova et al.²¹.Source data

(B). Reversible decrease of the NT conductance ● Timing 20 min

▲CRITICAL Because the protein effect is reversible, NT is ‘reset’ to its original, protein-free state upon delivery pipette removal beyond the position P2. Hence, the length dependence can be obtained using a single reservoir membrane.

1. Begin from Step 33 and set the usNT length to a different value.

2. Repeat Steps 33–36 and 39.

3. Measure the amplitude $\Delta G$ of the first conductance step (Fig. 8a,b).

4. Calculate signal-to noise ratio of the $\Delta G$ measurement ${\text{SNR}}_{\mathrm{p}}=(\Delta G∕G_{\text{NT}})\text{SNR}(L_{\text{NT}})$, where $\text{SNR}(L_{\text{NT}})$ is the SNR of a pure lipid usNT (with the same length, $L_{NT}$) obtained as described in Box 3.

5. Repeat the steps from Step 39B(i) and cover an 80- to 800-nm length range.

6. Plot ${\text{SNR}}_{\mathrm{p}}(L_{\text{NT}})$ dependence (see an example in Fig. 6, brown) and find the usNT length corresponding to the maximal ${\text{SNR}}_{\mathrm{p}}$.

   ▲CRITICAL STEP Active proteins can produce time-dependent constriction (ref. ²⁶, Fig. 8c). Determine the characteristic time for the protein action (e.g., GTP hydrolysis) and modify the usNT length to much larger time scale.

Characterization of membrane deformations by single protein complexes ● Timing 3–4 h

▲CRITICAL At this moment, the system is optimized for collecting and analyzing conductance changes caused by individual protein actors. For these, the dependence of the amplitude of elementary constriction steps on the usNT radius must be analyzed.

• 40.Check that the delivery pipette is situated away from the grid, beyond the position P1.
• 41.Carefully remove and discard the patch-clamp pipette

  CRITICAL STEP Use the proper procedure for discarding the glass waste.

• 42.Move the observation chamber and find another membrane.
• 43.Begin the protocol from Step 7 (skip Steps 19C and 20–30 and Box 2).
• 44.At Step 33, set the usNT length to the value ensuring the maximal SNR determined in Step 39C(vi).
• 45.Stop recording after Step 36.
• 46.Repeat Steps 41–46, aiming at recording 20–30 constriction events from 10 different usNTs.
• 47.Analyze the constriction events using the procedure described in Box 4 and Fig. 8.

  ▲CRITICAL STEP The analysis can be greatly improved if complementary ultra-structural information is available and rational mutagenesis is performed (e.g., see Fig. 7b, blue and red traces).

• 48.48 Repeat Steps 41–46 if more data points are needed.

  ? TROUBLESHOOTING

Box 4 ∣. Approximation of the geometry of membrane deformations from conductance changes ● Timing 20 min.

Deformation of the usNT membrane by a protein changes the luminal volume of the tube. The volume change $v$ can be approximated using the area $a_{\mathrm{p}}$ of the protein footprint in the lipid bilayer and the intrinsic mean curvature of the proteo-lipid complex, $j_{\mathrm{p}}$, defined as the radius $(r_{\mathrm{p}}=j_{\mathrm{p}}^{-1})$ of the membrane cylinder fully covered by the protein:

$$v=\frac{a_{\mathrm{p}}{r}_{\mathrm{p}}}{2}\left({\left(\frac{r_{\text{NT}}}{r_{\mathrm{p}}}\right)}^2-1\right)$$ (4)

For the deformation caused by the protein ring of width $I_{\mathrm{p}}$, the change of the usNT conductance, $\Delta G=G_0-G$, can be obtained using Eq. 4 and the Ohm’s equation for cylinder²⁶:

$$\frac{\Delta G}{G}=\frac{I_{\mathrm{p}}}{L_{\text{NT}}}\left({\left(\frac{r_{\text{NT}}}{r_{\mathrm{p}}}\right)}^2-1\right)=\frac{v}{{\pi r}_{\mathrm{p}}^2{L}_{\text{NT}}}$$ (5)

For small membrane deformation produced by an individual protein, Eq. 5 can be rewritten as:

$$\Delta G\approx \frac{v}{{\rho }_{\text{NT}}{L}_{\text{NT}}^2}$$ (6)

where ${\rho }_{\text{NT}}$ and $L_{\text{NT}}$ are defined in Box 1. Linear regression of $\Delta G(r_{\text{NT}}^2)$ dependence yields $a_{\mathrm{p}}$ (or $I_{\mathrm{p}}$) and $r_{\mathrm{p}}$ (an example of such a fit, taken from (ref. ²⁶, is shown in Fig. 8d). Note that $\Delta G$, and hence ${\text{SNR}}_{\mathrm{p}}$ (Box 3), are proportional to ${(r_{\text{NT}}∕r_{\mathrm{p}})}^2$; see Fig. 6, inset.

Analysis of bulk usNT constriction by multiple protein actors ● Timing 30 min

• 49.In most biological situations, membrane remodeling is produced by concerted action of many protein actors. usNT constriction by many proteins is easily resolved, and the collective action of the proteins is seen in changes of the elastic parameters of the usNT membrane (Fig. 9). To quantify such changes and to further use them for characterization of individual protein actors, follow the procedure described in Box 5. The protocol is complementary to the single-event analysis (Step 40) and enables qualitative comparison of individual and collective protein actions.

Fig. 9 ∣. Bending rigidity of the usNT membrane with bound proteins.

Changes in the effective membrane bending rigidity (Box 5) produced by the ENTH domain. DOPC/Chol 7/3 plus ENTH (blue); DOPC/Chol 7/3 without ENTH (red).Source data

Box 5 ∣. Analysis of stationary usNT constriction by multiple proteins ● Timing 30 min.

The parameters $a_{\mathrm{p}}$ and $j_{\mathrm{p}}$ of the protein footprint on the usNT membrane (Box 4) can be obtained from the analysis of stationary usNT constriction by multiple proteins. For non-zero $j_{\mathrm{p}}$, protein sorting between the curved usNT and the flat reservoir membrane affects the effective bending compliance of the usNT. In the linear sorting regime, the usNT radius is defined by Eq. 2 (Box 2) with $k=k_{\text{eff}}$. Here, $k_{\text{eff}}$ is smaller than the initial $k_{\mathsf{l}}$ of the pure lipid bilayer (Fig. 9), reflecting facilitation of the usNT constriction by the protein sorting⁴². The relative decrease of the apparent bending rigidity is: determined by Eq. 7 (refs. ^42,47):

$$\frac{\Delta k}{k_{\text{eff}}}=\frac{k_{\mathsf{l}}{\phi }_0{(a_{\mathrm{p}}{j}_{\mathrm{p}})}^2}{k_{\mathrm{B}}T}$$ (7)

where ${\phi }_0=\frac{n}{S}$ is the protein surface density on the planar bilayer lipid membrane. ${\phi }_0$ is estimated from the number of individual constriction events following the voltage-driven usNT expansion (Box 2).

The voltage increase by $\Delta U$ causes usNT area growth according to Eq. 8 (see panel a of the figure above):

$$\Delta S=\frac{C_{sp}{\Delta U}^2}{6\sigma }2{\pi r}_{NT}{L}_{NT}$$ (8)

The membrane area is transferred from the planar reservoir, bringing on $\text{average}N=\Delta {S\phi }_0$ (for an average usNT with $r_{\text{NT}}=5\text{nm}$, $L_{\text{NT}}=100\text{nm}$, $\sigma ={10}^{-3}\mathrm{N}∕\mathrm{m}$ and ${\phi }_0=0.01\text{nm}-2$, one gets $N\approx 1$ for $\Delta U=50\text{mV}$ and $N\approx 10$ for $\Delta U=150\text{mV}$. Because $N$ is small, the stochastic effects are pronounced, and the probability of detecting the transfer of $n$ proteins is determined by Eq. 9:

$$P(n,\Delta S)=\frac{{(\Delta S{\phi }_0)}^n\text{exp}(-\Delta S{\phi }_0)}{n!}$$ (9)

Panel b of the figure above shows the probability distributions for $\Delta U=50\text{mV}$ (red) and 150 mV (black). The deviation of $n$ from $N$ leads to additional constriction events $(m)$, probabilistically seen upon the usNT expansion (an example is shown in Fig. 8b, blue curve). Fitting the probability of the $m$ events’ occurrence with Eq. 10:

$$P(N-m,\Delta S)=\frac{N^{N-m}\text{exp}(-N)}{(N-m)!}$$ (10)

(panel b, inset, in the figure above), yields $N$ and ${\phi }_0=N∕\Delta S$.

Procedure

1. Once the usNT constriction finishes, measure $k_{\text{eff}}$ (Box 2).

2. Apply the voltage protocol for the surface density measurements ($V$ is switched between 50 and 200 mV every 5 s).

3. Determine the probability of the appearance of exactly $m$ constriction steps upon a 50- to 200-mV increase of $V$ (Fig. 8b).

4. Fit the probability distribution (panel b in the figure above) and determine ${\phi }_0$.

Troubleshooting

Troubleshooting advice can be found in Table 1.

Table 1 ∣.

Troubleshooting table

Step	Problem	Possible reason	Solution
‘Equipment setup’, ‘Preparation of the observation chamber’	PDMS support is not properly cured	Wrong curing time	Find the optimal PDMS curing time by repeating the procedure several times. As an alternative to a PDMS-made support, small ribbons of biocompatible double-sided adhesives can be used (such as the GraceBio SecureSeal Adhesive). However, chamber contamination by the adhesives during membrane painting can occur
‘Equipment setup’, ‘Electrical noise check’	Excessive electrical noise	Improper shielding and/or grounding	Check that you have connected all the metallic parts of the setup to grounding point(s); avoid ‘ground loops’; ensure that all the equipment’s ground outlets are connected to a single grounding point
6	Thick lipid films fail to spontaneously convert to lipid bilayers	Excessive amount of lipids on the grid mesh	Clean the brush (Step 3); use the clean brush to remove excess lipid material
	Lipid bilayers break spontaneously during the process of/shortly after painting	Impurities are triggering nucleation of pores and membrane rupture	Verify that lipid reservoirs (Fig. 2b) are not in contact with the PDMS support. Increase the lipid concentration in squalene. If nothing above works, change your lipids and/or solvent
15	Bad gigaseal. $G_{seal}$ (Fig. 1b) is >100 pS and/or unstable	Contaminated pipette tip	Use freshly made pipettes and store pipettes in a closed jar to protect the tips from dust. Fire-polish the pipette tip right before use. Verify that a constant positive pressure is applied to the pipette during manipulations in solution; constant efflux for the tip diminishes the tip contamination risk
		Inappropriate type of glass used	Change the type of the glass capillary you use (see ‘Equipment setup‘, ‘Patch-clamp and delivery pipettes’); if conventional patch-clamp and borosilicate capillary fail, consider using a thick-walled capillary or quartz
18	The measuring current does not increase with the pipette movement toward the membrane reservoir	The NT is broken or its lumen is inaccessible or clogged	Remove and discard the patch-clamp pipette. Repeat one to three times from Step 7; an NT should form in 50–90% of the attempts Decrease the deviation of the pipette axis from the vertical Decrease the speed of the pipette movement. Consider raising the pipette incrementally, in small steps See Troubleshooting advice for Step 6
19A(iii)	The dependence of $I_{\mathrm{m}}$ on changes of the NT length is not hyperbolic; the data fitting (Box 1) is not reliable	Nano-positioner is not calibrated	Label the tip of a patch-clamp pipette with a fluorescent marker (e.g., soak the pipette in a solution of 100-nm fluorescent microspheres). Using this pipette as a marker, make a calibration curve relating the position of the tip (in the vertical direction) and the reading of the nanoposition controller. Verify that the curve is reproducible in repetitive raising/lowering of the pipette
		The NT conductance is unstable at constant length (drift)	See Troubleshooting advice for Step 19B(i)
19B(i)	Conductance changes at a fixed position of the nanopositioner	Patch-clamp pipette and/or grid mesh move slowly	Using phase-contrast optics (40× or 60× objective) determine whether the pipette and/or grid move. Identify and eliminate the reason of the drift (loose attachment of the grid/pipette, rotation of the pipette holder, mechanical drift of the micromanipulator and/or microscope stage)
19C(i)	$I∕U$ characteristic is either linear or not symmetric	Wrong voltage range used	Increase $U$. Chose the nanotube with high $r_{\text{NT}}$ value (low tension) $I∕U$ asymmetry is seen with charged membrane due to electrophoresis in the inner lipid monolayer. Reduce $U$, apply fast voltage ramps
19C(iii)	$k$ is highly variable; the distribution of the $k$ values is not a sharp Gaussian	Poor membrane material	Change lipid stocks
20	Low SNR	Vibrations of the patch-clamp pipette	Check that your setup is properly protected against vibrations
25	Displacements of the delivery pipette cause acute and irreproducible changes in the NT conductance	Mechanical interference or large efflux from the delivery pipette	Check that the delivery pipette is mechanically isolated from the patch-clamp pipette (look for, e.g., entangled cords, physical contact between the micromanipulators; verify that the delivery pipette’s controller is not causing vibrations). Reduce the approach speed. Trace the efflux, for example, by adding fluorescently labeled nanospheres to the delivery pipette; verify that the spheres escape the pipette slowly
27	Upward stepwise changes of $I_{\mathrm{m}}$	Membrane poration	Change the osmoticant. If the problem persists, it indicates that the poration is caused by the curvature stress; the effect depends on the lipid composition used. Change the lipid composition or avoid high NT constriction
	Acute drop of the NT conductance to the background level	NT scission	Avoid extreme NT constrictions causing scission
35	No NT constriction	Delivery pipette is clogged	Change the delivery pipette, using a low-magnification microscope verify that the pipette tip is clean
	Gradual conductance decrease; no steps are observed	Low SNR	Decrease the NT length. Verify the SNR curve (Step 20). Use NTs with higher $r_{\text{NT}}$ values (lower lateral tension). To further increase SNR high-ionic-strength (1 M) solution can be used in the patch-clamp pipette
	Upward stepwise changes of $I_{\mathrm{m}}$	Membrane poration by the protein	Use dialysis to remove detergent from the protein solution. Decrease the protein concentration. To further assess the effect, the protein can be added to the patch-clamp pipette at Step 7. Continue the protocol until Step 15; do not rupture the patch. Apply the main acquisition protocol (Step 19A(i)), monitor the patch conductance, and look for upward channel-like events. The channel-like activity, if detected, can be minimized by optimizing lipid composition, for example, by adding cholesterol or lipids promoting inverted hexagonal phase appearance
48	The usNT radius variations from membrane to membrane are small	Membranes on the different holes of the grid mesh share the same lipid reservoir	Prepare another grid and repeat the protocol

Timing

Steps 1 and 2, pretreatment of the grid mesh for reservoir membrane production: 10 min

Steps 3–6, making multiple reservoir lipid membranes using the painting technique: 15 min

Steps 7–18, pulling the lipid membrane NT: 10 min

Step 19A, determination of the NT geometry: 10 min

Step 19B, assessment of the long-term stability of the NT: 10 min

Step 19C, determination of the mechanical parameters of the NT membrane: 10 min

Step 20, determination of SNR: 40 min for each new grid

Steps 21–30, testing the protein delivery system: 10–20 min for each new grid

Steps 31–38, initial characterization of the protein effect(s): 30 min

Step 39A, irreversible NT constriction: 60 min

Step 39B, reversible decrease of the NT conductance: 20 min

Steps 40–48, characterization of membrane deformations by single protein complexes: 3–4 h

Step 49, analysis of bulk usNT constriction by multiple protein actors: 30 min

Box 1, determination of the radius of the NT lumen: 20 min

Box 2, determination of the elastic constants of the NT membrane: 30 min

Box 3, obtaining the dependence of the SNR on the NT length: 40 min

Box 4, approximation of the geometry of membrane deformations from conductance changes: 20 min

Box 5, analysis of the stationary usNT constriction by multiple proteins: 30 min

Anticipated results

The method described in this protocol enables the detection and characterization of membrane deformations driven by individual proteins and protein complexes. One of the illustrative examples of the direct application of the methods is the quantification of transient mechano–chemical action of small oligomers of Dyn1²¹. Figure 8c shows a small and transient change of the usNT normalized conductance $({\Delta G}_{\mathrm{n}})$ produced by Dyn1 in the presence of GTP. The conductance changes are recurrent, their periodicity set by the characteristic time of GTP hydrolysis by the protein²¹. Analysis of the ${\Delta G}_{\mathrm{n}}$ dependence on the nanotube radius reveals that membrane deformations are produced by short structures (Fig. 10a, gray) associated with ~1- to 2-rung Dyn1 helices²¹.

Fig. 10 ∣. Self-limited character of membrane deformations produced by Dyn1 in the presence of GTP.

a, The axial length (Box 4) of the membrane deformation (the footprint, Box 4) produced by Dyn1 in the presence (gray) and the absence (black) of GTP. Black and red curves represent the lognormal approximation of the redistributions. b, The number of Dyn1-GFP molecules in small protein clusters formed on lipid nanotubes in the presence of GTP (left). Right, representative fluorescence microscopy images showing an example of the cluster (green) and lipid nanotube labeled by a lipid fluorescence probe (red). Scale bar, 2 μm.Source data

Small reversible conductance changes contrast with stepwise irreversible constriction of the usNT observed in the absence of GTP (²⁶, Fig. 7b, red trace). Here, much longer Dyn1 helices form (Fig. 10a, black), apparently restricting the usNT transformations (Fig. 7d) and precluding manipulations and mechanistic analysis of the non-physiological rigid end state of the system^26,46. Altogether, these observations imply that GTP hydrolysis prevents the formation of the rigid end-state structures via depolymerization of Dyn1, explaining the limited size of Dyn1 oligomers in the presence of the nucleotide (Fig. 10a,²¹). Interestingly, fluorescence microscopy observation of GFP-Dyn1 also confirms that the size of the protein oligomers in the presence of GTP is self-limited. The stoichiometry of GFP-Dyn1 oligomers (Fig. 10b) corresponds to that of 1- to 2-rung Dyn1 helix^44,49, in good agreement with the conductance measurements (Fig. 10a, gray). Importantly, the SNR in these measurements exceeded 10 for certain geometries of usNT (see Fig. 6), indicating that the method described here is capable of resolving membrane deformations produced by individual Dyn1 tetramers. Indeed, small constriction steps produced by the R399A mutant of Dyn1, which is defective in high-order oligomerization, can be resolved (Fig. 8a).

For smaller, non-oligomerizing protein species, such as the ENTH domain, the end state of the constriction process can be further analyzed. The usNT can be quantitatively expanded/contracted by an electric field (Box 5). Such fast changes of the usNT curvature allow exploration of curvature-driven sorting of the ENTH between the usNT and reservoir membranes⁴⁷. Analyses of individual constriction events (Fig. 8b) and changes in the usNT effective bending modulus by ENTH (Fig. 9) enable comparison of ENTH-driven local membrane deformations at different membrane coverage of the protein, assaying for cooperative behavior. Of note, a similar analysis can be applied to explore nanoscale elastic properties of multicomponent lipid bilayers. Interestingly, changes in the usNT bending compliance by ENTH point to a novel mechanism of the usNT constriction coupled to the protein redistribution. Instead of coercing usNT to a specific curvature defined by the protein concentration in the bulk, ENTH acts as softener (Fig. 9), facilitating the lateral tension–driven constriction of the usNT to a curvature defined by the tensile force.

Finally, conductance measurements in usNT remain the method of choice for analyzing membrane instabilities, pore formation and fission produced by local membrane constriction². Importantly, because those instabilities often develop at a 1- to 10-s timescale^2,21,26, our method enables fast sampling of the mechanical properties of the usNT membrane before the instability onset. Protein-driven deformations of usNT can be correlated with the appearance of the instabilities, thus allowing quantitative characterization of the geometry and dynamics of the protein machinery responsible for the instability.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The data that support the findings of this study are available from the corresponding authors upon reasonable request. The source data underlying Figs. 5, 7b, 8d, 9 and 10 are provided as Source Data files.