Actomyosin-Assisted Pulling of Lipid Nanotubes from Lipid Vesicles and Cells

Molecular motors are pivotal for intracellular transport as well as cell motility and have great potential to be put to use outside cells. Here, we exploit engineered motor proteins in combination with self-assembly of actin filaments to actively pull lipid nanotubes from giant unilamellar vesicles (GUVs). In particular, actin filaments are bound to the outer GUV membrane and the GUVs are seeded on a heavy meromyosin-coated substrate. Upon addition of ATP, hollow lipid nanotubes with a length of tens of micrometer are pulled from single GUVs due to the motor activity. We employ the same mechanism to pull lipid nanotubes from different types of cells. We find that the length and number of nanotubes critically depends on the cell type, whereby suspension cells form bigger networks than adherent cells. This suggests that molecular machines can be used to exert forces on living cells to probe membrane-to-cortex attachment.


Actin polymerization
Actin from New Zealand white rabbit skeletal muscle was purified from acetone powder based on the method of Pardee and Spudich, 1 modified after Kron et al.

Heavy meromyosin purification
Myosin isolated from New Zealand white rabbit skeletal muscle was used to prepare HMM based on the method of Margossian and Lowey. 3 To obtain a highly functional motor driven motility, we first remove non-functional myosin heads (rigor heads) via an actin affinity purification. 4 In short, the myosin motor fragments (heavy meromyosin, HMM) are mixed with actin filaments and MgATP in solution, followed by ultracentrifugation at 1 × 10 5 g to pellet any MgATP insensitive motors together with the actin filaments. The high functional HMM then is supplemented with 20 % sucrose and will be stored at −80°C until use.

Particle image velocimetry analysis
The background of timelapse videos was subtracted. Therefore, the minimal intensity of the stack was projected and then subtracted using the Calculator Plus plugin in Fiji. Furthermore, images were rotated in a way that actin filaments move from left to right. The individual images of the stack were loaded into JPIV (https://eguvep.github.io/jpiv/index.html) run in a Python environment. To obtain the vector field, particle image velocimetry was performed on consecutive images using first a 64x64 and then a 16x16 pixel interrogation window with a final vector spacing of 8x8 pixel. The vector fields were batch-filtered by performing a normalized median test and a median filter, where all invalid vectors were excluded. These invalid vectors were replaced by the median to obtain the final vector field.

Velocity correlation length calculation
A custom-written Python script was used to format the JPIV data for further processing.
The velocity correlation length of the actin fibers was calculated in MATLAB using a script described elsewhere. 5,6 In brief, the displacement vectors were divided by the time difference between the two images from which they were generated, resulting in the velocity vector r i,j , which was assigned to the central coordinate (i,j) of each 8x8 window. Since the axial movement is the dominant direction in the described experimental setup, only the lateral component U i,j , perpendicular to the movement direction was used to calculate the velocity fluctuations u i,j as: U mean is the mean velocity. The lateral correlation function C r was calculated as: . .⟩ is the average and r = ∥r i,j ∥ is the norm of r i,j . The first crossing of the threshold 0.01 with the lateral correlation function C r was defined as the velocity correlation length.

Analysis of actin filament velocity
Moving actin filaments were recorded as a time-lapse and tracked using ImageJ plugins: A classifier in the Trainable Weka Segmentation plugin 7 was trained to detect filaments and create a binary map of well defined particles against the background. This improves the ability of the Trackmate plugin 8 to correctly identify, track and return the trajectories of the individual particles. For each trajectory the magnitude of the average velocity vector and the orientation, i.e. the argument of the end-to-end vector were calculated. To display the data in a rose plot the trajectories were binned first by their argument and then within these bins by the magnitude of their average velocity vector. The rose plots were generated using the plotly library (v4.14.3) for python (v3.7.4).

GUV formation
Giant unilamellar vesicles were prepared using the electroformation method 9 using a Vesi-

STED microscopy
Lipid nanotubes were imaged on an Abberior expert line (Abberior Instruments GmbH, Germany) with a pulsed STED line at 775 nm using an excitation laser at 640 nm and spectral detection. The detection window was set between 650-750 nm to detect Atto633-labeled lipid nanotubes. Images were acquired with a 100×/1.4 NA magnification oil immersion lens (Olympus). The pixel size was set to 15-18 nm and the pinhole was set to 0.8 AU for 2D-STED and to 0.6 AU for 3D-STED. Images were analyzed and processed with ImageJ (NIH, brightness and contrast adjusted).

Analysis of lipid nanotube networks for GUVs
GUVs and nanotubes were classified separately using two different classifiers in the trainable Weka Segmentation plugin. In the segmented image the individual GUVs (imaged crosssectional area larger than 6 µm 2 ) are then counted and the nanotubes skeletonized, i.e.
reduced to one-dimensional branches, whose individual lengths can be determined using the Analyze Skeleton plugin. For each micrograph, the sum of all branch lengths, i.e. the network length is calculated, omitting branches smaller than 10 µm to exclude artefacts.
The network length is divided by the respective number of GUVs to obtain the normalized network length per GUV for one micrograph, from which the average and standard deviation displayed in the text were calculated.

Statistical analysis
All the experimental data is reported as mean ± SD from n experiments. The respective value for n is stated in the corresponding figure captions. All experiments were repeated at least twice. To analyze the significance of the data, a Student's t-test with Welch's correction was performed using Prism GraphPad (Version 9.1.2) and p-values correspond to ****: p ≤ 0.0001, ***: p ≤ 0.001, **: p ≤ 0.01, *: p ≤ 0.05 and ns: p ≥ 0.05.

Data availability
The datasets generated during and analyzed during the current study are available from the corresponding author on reasonable request.

Supporting Figures
Supporting Figure S1: SDS-PAGE of HMM and actin        Supporting Video S13: Actin filament dynamics during lipid nanotube pulling of a Jurkat cell (SiR-actin) Confocal time series of a Jurkat cell (membrane labeled with WGA-Alexa488, λ ex = 488 nm) during lipid nanotube formation with labeled SiR-actin (corresponding to Video S12). There does not seem to be any time delay in between the initial lipid nanotube pulling and actin filament presence within the lipid nanotube. Scale bar: 10 µm.