Cell culture
hTERT RPE-1 cells (here referred to as RPE-1) (American Type Culture Collection (ATCC), CRL-4000), HUVEC/TERT2 cells (ATCC, CRL-4053), BJ-5ta cells (ATCC, CRL-4001) and 293T cells (Takara Bio, 632180) were used in this study. RPE-1 cells were maintained in DMEM/F-12, HEPES, no phenol red medium (Gibco, 11039047) supplemented with 10% FBS (Millipore Sigma, F1435). HUVEC/TERT2 cells were cultured in EBM2 basal medium (Lonza, CC-3156) supplemented with EGM2 (Lonza, CC-4176). BJ-5ta fibroblast cells were cultured in a mixture medium with 4 parts DMEM (Thermo Fisher, 11995073) and 1 part Medium 199 (Thermo Fisher, 11150059), supplemented with 10% FBS. 293T cells were maintained in DMEM (Thermo Fisher, 11995073) with 10% FBS for lentivirus preparation. RPE-1 cells were used for most experiments in this study either by generating stable cell lines or transient transfection. HUVEC/TERT2 cells and BJ-5ta cells were used for stable cell line generation for measuring ER–PM contact gradients. All stably transfected cell lines were cultured identical to the original cell line.
Alveole/PRIMO system-based micropatterning
The Alveole/PRIMO system was used to pattern linear stripes on 96-well glass-bottomed plates (Cellvis, p96-1.5H-N) using a standard micropatterning protocol. In brief, a 96-well plate was subjected to plasma cleaning for 2 min and coated with 60 μl poly-l-Lysine (Sigma, P4707) per well for 30 min. After 3 wash steps with DI water (Invitrogen, 10977023), the 96-well plate was heated at 90 °C until dry. Meanwhile, 50 μl 100 mg ml–1 of mPEG-SVA (Laysan Bio, MPEG-SVA-5000-5g) was freshly prepared in 0.1 M HEPES (pH 8.5) and added to each well for 1 h to make the glass non-adhesive for cells (passivation). The plate was washed 4 times using DI water and again dried to complete passivation. At this stage, the coated plated could be stored at 4 °C for several weeks. Before experiments were performed, a gel mixture (978.6 μl DI water, 20 μl PLPP gel (Alveole, B002) and 1.4 μl surfactant (Alveole, B002)) was freshly made and applied to individual wells in the 96-well plate (50 μl per well), and the 96-well plate was put on a heat plate at 90 °C. Once the gel solutions were evaporated, the 96-well plate was moved to a microscope connected to a PRIMO system for photopatterning. The HCS wizard of the Leonardo software was used to automatically control the position of each well, and a previously designed 20-μm wide linear stripe template was imported for laser patterning in each well. After photopatterning, the 96-well plate was immediately washed with 1× PBS 3 times and incubated with 70% ethanol for several minutes, followed by 4 times wash with 1× PBS. Finally, the 96-well plate was coated with 50 μl 10 μg ml–1 fibronectin (Sigma, F1141-5mg) per well for 30 min before 1× PBS rinse. A total of 2,000 cells were seeded into each well 12 h before each experiment.
Stable cell line generation
Stable cell lines were generated by lentivirus infection combined with cell sorting or puromycin selection. In brief, plasmids of interest in a lentiviral transfer plasmid backbone were transfected into low-passage 293T cells together with third-generation lentiviral packaging plasmids, including pMDLg/pRRE (Addgene, 12251), pRSV-rev (Addgene, 12253) and pCMV-VSV-G (Addgene, 8454), using Lipofectamine 2000 (Thermo, 11668019). The viral supernatants were collected at 48 h and 72 h after transfection and pooled together for subsequent filtration using a 0.22 μm filter (Millipore, SCGP00525) and concentrated using a 100 kDa centrifugal filter (Millipore, UFC910024). The concentrated virus was then aliquoted into several cryotubes and stored at −80 °C for future use or directly added to cells in the growth medium with polybrene (EMD Millipore, TR-1003-G). To generate RPE-1 cells with constitutive expression of mTurquoise–CAAX and MAPPER–mVenus, single cells with both fluorescence constructs after virus infection were sorted into individual wells of a 96-well plate and cultured for expansion. After confirming that five clones showed the same observation of the back-to-front ER–PM contact gradient, one of them was chosen for most studies and used as the base cell line to generate other cell lines. We selected cells that expressed this ER–PM contact reporter at a low level to minimize its effect on cell morphology and cell polarization. pLV-iRFP–SEC61β and pLV-PTP1B–mCherry were respectively or simultaneously introduced into stable cell line with mTurquoise–CAAX and MAPPER–mVenus to construct a 3-colour stable cell line (mTurquoise–CAAX, MAPPER–mVenus and iRFP–SEC61β) or a 4-colour stable cell line, (mTurquoise–CAAX, MAPPER–mVenus, PTP1B–mCherry and iRFP–SEC61β). pLV-AKT–PH–mCherry was introduced into mTurquoise–CAAX/MAPPER–mVenus stable cell line to generate mTurquoise–CAAX/MAPPER–mVenus/AKT–PH–mCherry stable cell line. For the two doxycycline-inducible cell lines, pCW-RTN4–mCherry or pCW-CLIMP63–mCherry plasmid was introduced into the 3-colour stable cell line, mTurquoise–CAAX/MAPPER–mVenus/iRFP–SEC61β. Doxycycline (1 μg ml–1) was added to induce RTN4 or CLIMP63 expression at the start of imaging.
siRNA and plasmid transfection
siRNAs from Dharmacon (Supplementary Table 1) were dissolved in Ultrapure DNase/RNase free distilled water (Fisher Scientific, 10-977-023) to prepare 2 μM siRNA stock. Stock solutions were aliquoted into multiple tubes to avoid repeated thawing. For siRNA transfection experiments in 96-well plates, RPE-1 cells were seeded 16 h before transfection and transfected with siRNA using DharmaFECT 1 (Dharmacon, T-382 2001-03) according to the manufacturer’s protocol. In brief, 20 nM siRNA and 0.5 μl DharmaFECT1 were diluted with Opti-MEM medium (Gibco, 31-985-070) to prepare a 10 μl volume system in separate tubes. After 5 min of incubation at room temperature, the two tubes were mixed thoroughly and gently for another 20-min incubation. Then another 80 μl Opti-MEM medium was added into the transfection mixture and transferred to each well for a 6 h transfection before the medium was replaced with the complete growth medium.
For transient transfection of DNA, Lipofectamine 2000 was used following standard protocols. In brief, 2–3 × 103 RPE-1 cells were plated per well and transfected with 0.1–0.2 μg DNA of each plasmid and 0.25 μl Lipofectamine 2000 diluted in Opti-MEM medium. The transfection mix was replaced after 4 h with complete growth medium and cells were imaged 16–24 h after transfection.
DNA plasmids and chemicals
The following plasmids were ordered from Addgene: GFP–MAPPER (117721), iRFP–SEC61β (108125), pPTP1BD181A–mCherry (40270), pHAGE2–mCherry–RTN4a (86683), mCherry–CLIMP63 (136293), pBiFC–VN173 (22010), pBiFC–VC155 (22011), EGFR–GFP (32751) and pcDNA3.1–AKT–PH–mCherry (67301). These plasmids were used as templates to amplify required fragments, which were further assembled into the destination plasmid backbone using the Gibson assembly method (NEB, E2611L). In brief, pLV–MAPPER–mVenus, pLV–iRFP–SEC61β, pLV–PTP1B–mCherry and pLV–AKT–PH–mCherry plasmids were constructed based on the pLV–mTurquoise–CAAX backbone. After cutting this plasmid with AgeI/NotI, several fragments, including mVenus, MAPPER and FKBP for pLV–MAPPER–mVenus, iRFP–SEC61β for pLV–RFP–SEC61β, PTP1B and mCherry for pLV-PTP1B–mCherry as well as AKT–PH–mCherry for pLV–AKT–PH–mCherry, were ligated to replace the mTurquoise–CAAX insert using the recombinant cloning method. mCherry–RTN4a and mCherry–CLIMP63 were respectively amplified from their template plasmids and inserted into the pCW backbone (derived from pCW–Cas9, a gift from E. Lander and D. Sabatini, Addgene, plasmid 50661) to generate pCW–mCherry–RTN4a and pCW–mCherry–CLIMP63 plasmid. EGFR–YN plasmid was engineered from the EGFR–GFP plasmid by replacing GFP with the VN173 fragment, which was amplified from pBiFC–VN173. The PTP1B(D181A)–YC plasmid was constructed based on the pPTP1BD181A–mCherry plasmid by replacing mCherry with the VC155 fragment, amplified from the pBiFC–VC155 plasmid.
Drugs used in the study were dissolved into DMSO (Santa Cruz, sc-358801) to prepare stock solutions, including the PI3K inhibitor LY294002 (Cayman, 70920), the PTP1B inhibitors CAS765317-72-4 (EMD Millipore, 539741-5mg) and MSI-1436 (MedChem Express, HY-12219A), and doxycycline hyclate (Sigma, D9891). All drugs were handled according to their datasheets and aliquoted to avoid repeated thawing process. The working concentrations for each drug are indicated in the corresponding experiments.
Antibodies and Immunofluorescence
pTy (P-Tyr-1000) multiMab rabbit monoclonal antibody mix (1:500, Cell Signaling Technology, 8954), anti-ESYT1 antibody (1:200, Sigma, HPA076926) and CLIMP63 monoclonal antibody (G1/296) (1:500, Enzo Life Sciences, ENZ-ABS669-0100) were used as primary antibodies for immunostaining experiments. Secondary antibodies included goat anti-rabbit IgG(H+L) Alexa Fluor 568, Invitrogen (1:2,000, Thermo Scientific, A-11011), goat anti-rabbit IgG(H+L) Alexa Fluor 647, Invitrogen (1:2,000, Thermo Scientific, A-21245) and goat anti-rabbit IgG(H+L) Alexa Fluor 700, Invitrogen (1:2,000, Thermo Scientific, A-21038). Anti-ESYT2 antibody (1:1,000, Sigma, HPA002132), PTP1B antibody (1:1,000, BD Bioscience, 610139) and RTN4 antibody (1:1,000, Thermo Scientific, MA5-32763) were used as primary antibodies for western blotting experiments.
Cells were seeded in a 96-well glass-bottomed plate with pre-patterned linear stripes for immunostaining experiments. After siRNA transfection or inhibitor treatment, cells were fixed using 4% paraformaldehyde in PBS for 10 min at room temperature and washed with PBS. Cells were then permeabilized with 0.1% Triton-X 100 for 10 min, followed by a PBS wash and blocking buffer incubation for 1 h (10% FBS, 1% BSA, 0.1% Triton X-100 and 0.01% NaN3 in PBS). Cells were then incubated with primary antibodies overnight in blocking buffer at 4 °C, followed by a PBS wash and secondary antibody incubation for 1 h at room temperature. Cells were washed with PBS again before imaging.
Microscopy
Automated epifluorescence microscopy
Automated epifluorescence microscopy was used to perform live-cell time-lapse imaging to track cell migration (Figs. 2g,h, 3e,f and Extended Data Figs. 3c, f–i and 7d,e). Cells were seeded in a 96-well glass bottomed plate, coated with collagen (Advanced Biomatrix, 5005-B, 30 μg ml–1 dilution for at least 1 h) and stained with 0.1 μg ml–1 Hoechst 33342 (Invitrogen, H3570) in growth medium or Opti-MEM medium for 30 min at 37 °C immediately before imaging. The 96-well plates were transferred into a live-cell chamber with 37 °C, 5% CO2 environment for 24 h long-term imaging by a Ti2-E inverted microscope (Nikon) equipped with a LED light source (Lumencor Spectra X) and Hamamatsu ORCA-Flash4.0 V3 sCMOS camera. The following acquisition parameters were used: interval, every 12 min; ×20 (Nikon CFI Plan Apo Lambda, 0.75 NA) objective lens; 89903-ET491 BV421/BV480/AF488/AF568/AF647 Quinta Band set (Chroma Technology) for multichannel fluorescent images or BV421 only for Hoechst imaging; 16-bit mode with 2 × 2 binning; and 80 ms exposure time at 5% light strength to reduce the light toxicity for Hoechst imaging. Raw images were shading-corrected through NIS-element software to correct for uneven sample illumination, using wells full of imaging medium without cells as the background autofluorescence subtraction.
Spinning-disk confocal microscopy
Unless otherwise indicated, imaging was performed using a SoRa spinning-disk confocal microscope (Marianas system, 3i), equipped with a Zeiss Axio Observer 7 stand, ORCA-Fusion BT sCMOS camera (Hamamatsu), CSU-W1 SoRa confocal scanner unit (Yokogawa), and 405, 445, 488, 514, 561 and 637 nm LaserStack (3i). For mTurquoise, mVenus and mCherry 3-colour live-cell imaging, the CSU-W1 dichroic for 445, 515 and 561 nm excitation was used. Images were acquired every 5 s, 30 s or 2 min based on experiments. The bottom plane was set as the focus plane with the maximal ER–PM contact signals. Definite Focus 2 function was used for long-term focus control. For mTurquoise, mVenus, mCherry and iRFP 4-colour live-cell imaging, the CSU-W1 dichroic for 445, 515 and 561 nm excitation was used for mTurquoise, mVenus and mCherry imaging, whereas the CSU-W1 dichroic for 405, 488, 561 and 640 nm excitation was used for iRFP imaging. For fixed cell imaging, z stack images were captured with a step size of 0.27 μm. Usually, 9 or 13 optical slices were captured based on experiments. For optogenetic experiments, a 3i spinning-disk confocal microscope equipped with an extra ‘Vector’ photomanipulation device was used. Front regions of migrating cells were manually defined to be illuminated by a 488 nm laser at 1% power every 2 s for continuous activation. Every experiment was pre-imaged for 20 min every 2 min before blue light illumination for another 60 min of imaging.
Total internal reflection microscopy
Imaging was performed on a total internal reflection microscope (Nikon) equipped with 488, 561 and 647 nm lasers and high numerical aperture objective at the Rockefeller University’s Bio-Imaging Resource Center. Cells stably expressing iRFP–CAAX and eGFP–SEC61β were seeded in a 96-well glass-bottomed plate with pre-patterned linear stripes and fixed for E-Syt1 immunostaining as mentioned above. Signals from the evanescent field were captured and used for quantification.
Cryo-ET sample preparation and data collection
Sample preparation and cryo-ET imaging were performed at the New York Structural Biology Center (NYSBC). In brief, 1–2 × 102 wild-type RPE-1 cells were seeded onto Quantifoil R1/2, 200 mesh gold EM grids with pre-patterned linear stripes prepared as described above. About 12–16 h later after seeding, dishes with grids were taken out of the incubator, and EM grids were blotted at opposite sides of the cells for 1–2 s and plunged frozen with a Leica GP2 (Leica Microsystems). Cryo-patterned grids were visualized using a Titan Krios electron microscope (Thermo Fisher scientific) equipped with a field emission gun, a GIF Quantum LS postcolumn energy filter (Gatan) and a K3 summit electron detector (Gatan). The electron microscope was operated at 300 kv in nanoprobe mode at a magnification of ×19,500 (pixel size of 4.53 Å at the specimen level). Cryo-ET tilt-series were collected using a dose symmetric scheme49 with a tilt range of −52° to +52° at a target defocus of −8 µm and 3° increments in SerialEM for a total dose of 126 e Å–2.
3D collagen gel migration
PureCol (Advanced Biomatrix, 5005-B, 3 mg ml–1) was used for 3D collagen gel migration assays. Before gelation, all collagen-related steps were performed on ice. Eight parts of collagen solution was slowly mixed with 1 part of chilled 10× PBS to prepare 2.4 mg ml–1 collagen stock solution, the pH of which was adjusted to 7.0–7.5 using sterile 0.1 M NaOH. Next, 60 μl of 0.5 mg ml–1 collagen solution (stock solution was diluted into cold 1× PBS) was added to a 96-well plate and left for 2 h at 37 °C for polymerization. The gel was washed with PBS for 3 times, and 50 μl of RPE-1 cells at a density of 2 × 104 cells per ml was added into each well for 2–3 h attachment. After incubation, 40 μl of medium was removed from the well and 60 μl of collagen was added. The 96-well plate with collagen on top of the cells was incubated at 37 °C for another 2–3 h and washed again with complete growth medium 6 times and cultured in a 37 °C incubator for 24 h before imaging.
Quantification and image analysis
Automated analysis of time-lapse imaging was performed using a custom Matlab R2021 pipeline based on previous work9,50. Details and parameters regarding how to quantify signals are summarized as below.
Cell segmentation and time-lapse tracking
Cells were automatically segmented from either the nuclear signal or the PM signal based on experiments. For 2D cell migration assays, nuclear signals (Hoechst staining) captured using an automated epifluorescence microscope were used for segmentation based on Laplacian of Gaussian algorithms. The detected nuclei were tracked using a nearest-neighbour algorithm between the current frame and its previous frame. To increase tracking accuracy, the nuclear mass (the product of nuclear intensity and nuclear area) was also used as a constant metric to adjust matching between two neighbouring frames. After tracking, x and y coordinates of each nucleus were exported for cell speed and MSD calculation. Cell speed was quantified as the value of total travelling distance over travelling time. MSD was quantified as the mean value of (xy(t) – xy(0))2, where xy(t) is the position of cell at time point t and xy(0) is the initial position. For 1D cell migration assays, the PM signal (CAAX fluorescence intensity), captured by spinning disk confocal microscope, was used for segmentation based on a modified version of Otsu’s algorithm. With the help of linear stripes, only isolated single cells were selected for subsequent analysis. Nearest-neighbour algorithm between subsequent frames was used for cell tracking. Cell velocity was determined based on the overall centroid distance travelled over a fixed time. A first-order polynomial function was used to fit cell trajectory based on xy coordinates to determine cell direction, which was smoothed to minimize the effects of random centroid movements.
Time lapse of MAPPER puncta tracking
To track ER–PM contacts, the MAPPER reporter signal was captured every 20 s and subjected to tophat filtering (3 pixel radius), followed by cell segmentation based on the CAAX signal as discussed above. The threshold for puncta segmentation was determined by the MAPPER signals located at 40% from the cell front using a modified version of Otsu’s algorithm. This facilitated identification of ER–PM contacts at the cell front, which were much dimmer than those at the back. A nearest-neighbour algorithm was used to connect puncta between subsequent frames, and the MAPPER mass was set as a constant variable to adjust mismatched puncta. The speed of puncta was determined by their travelling distance over duration. The lifetimes of MAPPER puncta were calculated as the duration from its appearance to disappearance or the last frame. The MAPPER mass was quantified as the product of MAPPER intensity and MAPPER area.
Kymograph plots and gradient profiles in 1D migrating cells
A schematic graph of the kymograph and average gradient analysis are shown in Extended Data Fig. 1d. In brief, the CAAX signal at each time point was used to mask individual cells and a 3 μm ring around the cell periphery. The latter was used to include a region with higher relative PM contribution. To minimize the effect of cell shape differences among cells, each cell mask and ring mask were divided into 20 equal segments from the cell front to the back. The cell front was automatically identified as the tip end, which has the farthest distance from the centroid of the cell mask. Each pixel value of the normalized biosensor intensity, biosensor size or biosensor mass in the cell and ring mask were assigned to 1 of the 20 segments. An averaged value in each segment from the front to back was calculated to create profiles at each time point that were then represented in a kymograph. The average of the gradient profiles at the different time points or from different cells is also shown in most figures. Unless otherwise stated, the time series of kymograph profiles is shown as a heatmap, with the x axis representing time and each vertical line representing the 20 bins of the spatial distribution of a specific parameter at a specific time. For the kymograph profiles of the MAPPER sensor, the full cell mask was used, whereas the ring mask was used for the pan-Tyr signal.
Growth rate analysis and slope calculation
For the MAPPER growth rate analysis (Fig. 4h), the CAAX signal was used to define the cell boundary and a region of interest (ROI) was chosen near the cell front as indicated (Extended Data Fig. 5f). As the cell migrated, the ROI moved and became increasingly located towards the back. The MAPPER to CAAX ratio of each MAPPER puncta in the ROI was quantified and plotted versus the ROI position. To compare the growth rate differences in the front versus back in Fig. 4i, the averaged curve of growth rate was split into the front part and the back part, which were separately fitted by a linear regression function to calculate the slope. The front segment was defined as the relation position from 0.2 to 0.5, whereas the back part was from 0.5 to the position where the MAPPER puncta showed the maximal average intensity.
Polarity steepness
In Figs. 2d and 4f and Extended Data Fig. 5b,e, polarity steepness was calculated from normalized MAPPER to CAAX or MAPPER to SEC61β profiles in migrating cells. The average polarity score in Fig. 2d was quantified by normalizing the absolute value of (mean (front 20%) – mean (back 20%)) to the mean (back 20%). In Fig. 4f and Extended Data Fig. 5b,e the MAPPER signals between the back and front 50% of cell were measured and the ratio of front to back was calculated to represent the polarity steepness.
Cryo-ET data processing
The raw tilt movies were motion-corrected and CTF-corrected in Warp (v.1.09)51. The tilt-series stack was exported from Warp, in which the tilt series were aligned with the AreTomo software package52. The aligned tilt series were reconstructed with either weighted-back projection in AreTomo or with using Tomo3D53. To enhance the contrast of the tomograms for visualization, weighted-back projected and simultaneous iterative reconstructive technique tomograms were CTF-deconvolved with IsoNet54. Initial segmentations were manually performed on a few tomographic slices in DragonFly (v.2022.2) for training a neural network in DragonFly (v.2022.2)55 or by using Tardis-Pytorch56 on the deconvolved tomograms. Segmentations were corrected and manually labelled using DragonFly. The segmentations were subsequently processed and analysed using the surface morphometrics analysis toolkit57. Curvedness was calculated for the front and back ER using the surface morphometrics analysis toolkit and was chosen as it is an unassigned combination of the two principal components of curvature and is used because the surface normal vectors do not have a sign.
Statistical analysis
Statistical results were analysed using GraphPad Prism 8.0 or Matlab R2021 and shown as the mean ± s.d. or mean ± s.e.m. as indicated. Comparisons were made between groups using unpaired two-tailed Student’s t-test or one-way ANOVA and Scheffe’s/Dunnett post hoc comparison as indicated. For the pTyr to CAAX or the MAPPER to CAAX gradient profile significance test, 20% front and back ratio was used in Figs. 1b,c and 2f,i and Extended Data Fig. 8a,d,g, whereas the index of peak ratio was used in Figs. 3c,d and 5i and Extended Data Fig. 7f,h. For all analyses, *P < 0.05, **P < 0.01 and ***P < 0.001 were considered significant. NS indicates statistical non-significance with P > 0.05. Each experiment was performed at least three independent times.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.