Approach to map nanotopography of cell surface receptors

Abstract

Cells communicate with their environment via surface receptors, but nanoscopic receptor organization with respect to complex cell surface morphology remains unclear. This is mainly due to a lack of accessible, robust and high-resolution methods. Here, we present an approach for mapping the topography of receptors at the cell surface with nanometer precision. The method involves coating glass coverslips with glycine, which preserves the fine membrane morphology while allowing immobilized cells to be positioned close to the optical surface. We developed an advanced and simplified algorithm for the analysis of single-molecule localization data acquired in a biplane detection scheme. These advancements enable direct and quantitative mapping of protein distribution on ruffled plasma membranes with near isotropic 3D nanometer resolution. As demonstrated successfully for CD4 and CD45 receptors, the described workflow is a straightforward quantitative technique to study molecules and their interactions at the complex surface nanomorphology of differentiated metazoan cells.

Fig. 1. Coating of coverslips with glycine better preserves cell surface morphology and resting state of immobilized T cells.

a, b Schematic illustration of T-cell landing on PLL- (left panels) and glycine-coated coverslips (right panels). The arrows indicate forces influencing the cell on the coated coverslip. The g-force is the sole force affecting T cells on glycine. T cells on PLL are further stretched due to electrostatic interactions of the surface molecules with PLL. The blue stripes represent glass coverslip, the light blue stripes above represent the glycine layer (not to scale). c, d Live-cell TIRF microscopy of CD4-GFP in Jurkat cells landing on PLL- (left panels) or glycine-coated (right panels) coverslips measured at 37 °C. Representative contact morphology is shown for cells 5 min (c) and 10 min (d) after addition of the cell suspension to coverslips. e Graph showing signal homogeneity of cell contacts on PLL- and glycine-coated coverslips as in c and d. Signal homogeneity was quantified according to ref. (see “Methods” and Supplementary Data for more details). In total, 31 cells for PLL_5min, 21 cells for Glycine_5min, 32 cells for PLL_10min and 29 cells for Glycine_10min were analyzed. Blue dots represent values from individual cells, red lines median values, boxes indicate upper and lower quartile of data and outliers are circled. Statistical significance was evaluated using one-way ANOVA test; p_5min < 10⁻⁵, p_10min ≤ 0.0025. f Calcium response induced by the interaction of Jurkat cells with coverslips coated with stimulating antibody (anti-CD3ε; OKT3), PLL or glycine as indicated by the changes in the fluorescence of ultrafast genetically encoded calcium sensor GCaMP6f_u at 5 min (black bars; I_max/I_max+5 min) and 10 min (gray bars; I_max/I_max+10 min) after the maximal stimulation of cells (I_max). The extent of calcium response was calculated from the signal decay after cell spreading on stimulating surfaces (see “Methods” and Supplementary Data for more details). Relative calcium response equal 1 indicates no stimulation. Higher values indicate measured stimulation of cells. Small squares represent values calculated for individual cells; solid lines median values, boxes indicate upper and lower quartile of data and outliers are marked with a red cross. Statistical significance was evaluated using one-way ANOVA test; *p < 0.05 or ***p < 0.001. Five examples of full signal traces together with graphical representation of data processing are shown in Supplementary Fig. S7. g Viability of Jurkat cells interacting with coverslips coated with PLL- (black bars) or glycine-coated (gray bars) coverslips. Dying cells were defined as a fraction of 7-aminoactinomycin-positive cells within the imaged area using wide-field microscopy (see “Methods” and Supplementary Data for more details). Values at 1 min represent the starting point for the analysis—a minimal period required for cells to land at the optical surface. Small squares represent individual measurements, solid lines median values, boxes indicate upper and lower quartile of data and outliers are marked with a red cross. Statistical significance was evaluated using unpaired t-test with unequal variances (Satterthwaite’s approximation); ****p < 0.0001, n.s. (not significant). Scale bars, 5 μm.

Fig. 2. Principle of 3D super-resolution biplane imaging using dTRABI.

a Overview of the experimental setup applied to perform 3D dTRABI imaging of T cells. HILO illumination of the sample (blue beam; only shown in enlarged box) triggers fluorescence emission (orange), which is split by a 50/50 non-polarizing beamsplitter (50/50 BS) to acquire biplane images on two separate EM-CCD cameras. The respective imaging lenses are shifted along the optical axis to induce a relative defocus of the image detection on synchronized cameras. Spots, apparent in both detection planes, are fitted by a Gaussian with identically set FWHM. b Using a piezo stage, the focal plane was linearly moved through the sample plane while imaging a single-molecule surface under dSTORM conditions. Hereby, both cameras were synchronized. Fitting the raw PSFs by independent Gaussians with invariable FWHM yielded axially dependent single-molecule intensity curves (upper panel). The relative change of position of the imaging lens in the reflection path is mirrored by the relative shift of the respective intensity curve (indicated by circles). Data points were spline interpolated to guide the eye (solid lines). An axially precise calibration function γ is derived directly from the raw intensities (I_Camera1, I_Camera2) of corresponding localizations from both cameras as γ(z) = (I_Camera2 − I_Camera1)(I_Camera2 + I_Camera1)⁻¹. The running median of the raw data (gray squares) is fitted with a high-order polynomial (black line) to generate the basis of the axial lookup table (lower panel). c A two-dimensional high-resolution data set is generated from both image stacks (transmission and reflection path) to create a three-dimensional dTRABI data set according to the calibration. Finally, the transmission localization set is used to render a high-resolution, axially color-coded image of the focused target structure. Scale bars, 5 μm.

Fig. 3. Fiducial free drift and tilt correction of dTRABI data.

CD4 WT on Jurkat cells is shown, where the membrane was labeled using Alexa Fluor 647-conjugated primary antibody (OKT4). a Principle of the fiducial-free correction of axial drift. Left: representative, color-coded high-resolution dTRABI image of an axially unstable sample. Due to the temporal change of the axial coordinate, the resulting image does not exhibit axially distinct features. Middle: by extracting the axial localization distribution of the thin membrane layer per frame and tracking it over the entire stack, a spatio-temporal drift trace can be plotted (top). Fitting of the raw data (gray squares) by a high-order polynomial (red line) allows the temporal linearization of the localization data, leading to a stable axial mean value over time (bottom). Right: re-rendering of the drift-corrected localization data reveals a color-coded dTRABI image exhibiting distinct, high-density clusters of CD4 below another, disperse, layer. b Principle of the fiducial-free correction of axial tilt. Left: representative, color-coded high-resolution dTRABI image of an axially tilted sample. Middle: the axial tilt of the sample is extrapolated by fitting a plane to the raw image. Afterwards, the data are linearized by subtracting the local plane-value from the raw localization. Right: the tilt-corrected image exhibits a homogeneous color-code in the lowest data layer, indicating no residual axial tilt. Exemplary drift and tilt were simulated for illustration but based on real data (the displayed cell displayed both significant drift and tilt). Right columns represent experimentally corrected real data on which the drift and tilt were projected. For all experimental data, axial drift correction was performed prior to a tilt test and correction. Scale bars, 5 μm.

Fig. 4. Cell surface receptor nanotopography visualized by 3D TRABI imaging.

a Schematic illustration of sample preparation for a receptor nanotopography imaging using dTRABI. A cell in suspension is first immobilized on a glycine-coated coverslip, fixed and imaged using dTRABI approach. The resulting three-dimensional dTRABI image represents a footprint of a cell on the optical surface with the color-coded axial position of localizations (right panel). b Three-dimensional dTRABI image of T-cell surface receptor CD45 with selected ROIs exhibiting a broad z-distribution of receptor localizations. CD45 was labeled with Alexa Fluor 647-conjugated primary antibody (MEM-28). c Magnified x–y and x–z projections of ROIs 1–5 as in b. Blue arrowheads point to microscopic membrane protrusions at the cell edges, red arrowheads to folded nanoscopic protrusions under the cell body and green arrowhead to membrane invagination. Scale bars, 5 μm in b and 500 nm in c.

Fig. 5. CD4 clusters represent the receptor accumulation at the tips of membrane protrusions.

a Representative three-dimensional dTRABI image of CD4 WT at the surface of Jurkat cell immobilized on a glycine-coated coverslip. b The axial distribution of CD4 WT localizations of the 4 μm² region of interest (ROI) of the cell in a. c Representative 3D dTRABI image of non-palmitoylatable CD4 CS1 mutant at the surface of Jurkat cell immobilized on a glycine-coated coverslip. d The axial distribution of CD4 CS1 localizations of the 4 μm² ROI of the cell in c. e–g Quantitative analysis of the receptor axial (z-axis) distribution on the surface of T cells. Receptors were analyzed by using a bi-Gaussian fit to the axial distribution of localizations for each ROI as in b, d and i (see Supplementary Fig. S11 for more examples and Supplementary Data for the details) and the FWHM range of the two Gaussian functions represent the z-distribution width (z_w) as depicted in e. Black circles in e represent the axial distribution of receptor localizations for a selected ROI, black line the bi-Gaussian fit, which is the sum of two Gaussians as depicted in red and blue, dashed lines in gray depict z_w and lines in light gray depict mean values of the Gaussian distributions. The graphs in f and g represent histograms of z_w obtained from 21 CD4 WT cells with 846 ROIs (black), 18 CD4 CS1 cells with 1044 ROIs (green) and 13 CD45 WT cells with 305 ROIs (magenta). The histograms in f and g show relative and absolute occurrence, respectively. Data points in b, d, f and i were spline interpolated to guide the eye. h Representative 3D dTRABI image of CD45 at the surface of Jurkat cell immobilized on a glycine-coated coverslip. Color-bars in the upper right corner of a, c and h indicate the axial position of the localizations in the image. i The axial distribution of CD45 localizations of the 4 μm² ROI of the cell in h. CD4 variants and CD45 were stained with directly labeled primary antibody as in Figs. 3 and 4, respectively. Scale bars in a and c and h, 5 μm.

Fig. 6. CD45 exclusion zones on the surface of resting Jurkat cells.

a Three-dimensional dTRABI image of CD45 (labeled with antibody as in Fig. 4) on Jurkat cell immobilized on a glycine-coated coverslip. b Magnified ROI as in a with indicated areas lacking CD45 localizations (yellow arrowheads). c 2D SOFI image of CD45 on Jurkat cell immobilized on a glycine-coated coverslip. CD45 was labeled using Alexa Fluor 647-conjugated MEM-28 antibody. Yellow arrowheads indicate cell surface areas lacking CD45 signal. Scale bars, a 5 µm, b 1 µm and c 5 µm.

Fig. 7. Nanoscopic segregation of CD4 WT and CD45 on the surface of resting Jurkat cells.

a Representative two-dimensional images of Jurkat cell surface sequentially analyzed for CD45 (magenta; left panel) and CD4 WT (green; middle panel) by SMLM. The right panel represents an overlaid image. CD4 was visualized as mEos2 fusion protein (PALM) after transient transfection of cells, and the surface CD45 was labeled using Alexa Fluor 647-conjugated MEM-28 antibody (dSTORM). ROIs 1-2 were zoomed to indicate details of proteins’ distribution (right side). b Intensity line-profiles were measured along the transparent gray regions indicated in ROIs 1–2 (as in a). Green line represents CD4 WT and magenta CD45 signals. c Schematic illustration of microvillus with indicated structural segments: tip, shaft and the basis. d Two-dimensional SMLM images of a selected Jurkat cell captured during its association with the optical surface which was sequentially analyzed for CD45 and CD4 WT (as in a). The accumulation of CD4 WT on the tips of large membrane protrusions was observed in 10 out of 32 imaged Jurkat cells. ROIs 3–4 were zoomed to show the details of membrane protrusions with accumulated CD4 WT on their tips (right side). e Intensity line-profiles were measured along the transparent gray regions indicated in ROIs 3–4. The arrows indicate the onset of line-profiles. Scale bars, 5 µm. Images from three independent experiments are shown (n = 32). f Schematic illustration of the organization of CD4 WT, CD4 CS1 and CD45 on the protrusions as indicated by the nanoscopy.