Authors: Peter Beemiller, Jordan Jacobelli & Matthew Krummel
Supported lipid bilayers are frequently used to study cell membrane protein dynamics during immune synapse formation by T cells. Here we describe methods for the imaging and analysis of OT1+ T cell activation and T-cell receptor (TCR) dynamics on lipid bilayers.
T cells are activated at immune synapses when TCRs bind agonist ligands on antigen presenting cells (APCs). Glass coverslip–supported lipid bilayers provide a system for in vitro T cell activation and immune synapse formation. In these systems, the supported bilayer acts as a surrogate APC, presenting all the factors needed to trigger TCR signaling and synapse formation. In a minimal activation system, only pMHC and ICAM are incorporated to activate TCRs. Lipid bilayers provide a number of technical advantages over authentic APCs. Coverslip–spanning bilayers can be formed, allowing large numbers of T cells to be deposited and analyzed in parallel. Unlike synapse formation on an authentic APC, the support restricts molecular reorganizations in the synapse membrane to a Euclidean plane. Using a total internal reflection fluorescence (TIRF) microscope restricts the imaging field to within ~100 nm of the synaptic interface. However, care should be taken when extrapolating synapse characteristics seen on bilayers to physiological synapses. In vivo, T cells generate synapses with irregular geometries, as they continuously crawl over APCs and potentially encounter other T cells.
The lipid bilayer system consists of a polyethylene glycol-cushioned lipid bilayer bearing Ni-NTA and biotin modified phospholipids. The PEG5,000 cushion is formed by the inclusion of a small fraction of phospholipids with PEG5,000 polymers covalently attached to the head groups (1,2), which improves the bilayer uniformity and streptavidin mobility. The Ni-NTA– and biotin–modified lipid head groups are used to capture dodecahistidine–sICAM-1 (3) and tetravalent streptavidin, respectively. Captured streptavidin is then used to bind monobiotinylated SIINFEKL:H2-K(b) pMHC complexes, resulting in stimulating bilayers that can activate OT1 TCR signaling (Fig. 1). The bilayers can be standardized by creating silica microsphere supported bilayers and comparing the protein ligand levels to the levels displayed on antigen presenting cells (4). Alternatively, the microsphere bilayer standards can be compared to reference beads to estimate the density in number of molecules per unit area. We standardize stimulating bilayers using bone marrow derived dendritic cells (BMDCs) pulsed with SIINFEKL peptide at a concentration that produces maximum in vitro T cell proliferation as a reference APC.
To image calcium fluxes, we use the updated TIRF microscope in standard epifluorescent mode, employing a DG-4 (Sutter Instruments) with 340x and 380x excitation bandpass filters (Chroma Technology) and a Zeiss 1.3 NA, 40× PlanFluar objective.
OT-I T cell blast preparation
Conditional myosin II knockout
Preparation of cells for imaging
Liposomes and Bilayers
Glass can be cleaned in advance, dried and stored until use.
Lipid bilayer setup:
Analysis of protein motility:
To generate standardized bilayers, measure the loading of his-ICAM and biotinylated pMHC onto bilayers relative to bone marrow derived dendritic cells (BMDCs), a prototypical antigen presenting cell (Fig. 1). BMDCs are loaded with 100 ng/ml SIINFEKL peptide in complete RPMI at 37 C for 30 min, and then rinsed thoroughly.
Bilayer preparation: 2+ hours
Generation of retrovirally transduced T cells: 5 days
Preparation of T cell blasts for imaging: 1 hour
Imaging: 2–4 hours.
Bilayer uniformity: The lipid bilayers should be uniform over many mm2, but occasional discontinuities are expected. If the discontinuities are frequent, this might indicate an issue with the cleanliness of the glass support, or contamination in the the liposome preparation. To test the quality of the naked bilayers, either incorporate a small amount of fluorescently labeled phospolipids into your liposome preparations (e.g., 0.5% Oregon Green 488 DHPE, Invitrogen O-12650), or pre-mix your liposomes with DiO before applying to the glass.
Ligand immobility: Protein ligand immobility is a common issue. You should first ensure that the bilayers are setting up as uniform, continuous sheets (above). In general, it is also best to use the minimum amount of ligand-binding phospholipids (DGS-NTA(Ni) and Biotinyl-CAP-PE) required to achieve sufficient protein loading.
The image analysis routines are performed almost entirely using MATLAB (MathWorks) scripts. The scripts for these analyses can be found as attachments, organized by application (tracking, segmentation, etc.). The functions performed by the scripts are described in general below.
Image arithmetic and cell and TCR microcluster tracking
All image arithmetic operations, for example: filtering, background subtraction, masking, and division, are performed in MATLAB. Cell tracking for analysis of calcium and cell motility is performed in Imaris using fura-2 ratiometric images series calculated and masked in MATALB. To create the fura-2 ratiometric image series, the component images acquired with 340 nm and 380 nm excitation are converted to floating point and the images acquired using 340 nm excitation are divided by the images acquired using 380 nm excitation. Image masks are created using Otsu’s algorithm on the 380 nm component images. Small non-cell debris is removed from the masks, and then the masked ratiometric images are transferred from MATLAB to Imaris for tracking. After tracking, ratiometric intensities for each cell track are normalized to the ratiometric intensity before cell binding to the bilayer. Cell track displacements and normalized ratios are then aligned to the onset of bilayer binding, which typically corresponds to the initiation of calcium fluxes for cells on stimulating bilayers. To calculate synapse parameters, such as mean speed, ratiometric intensities versus distances from the origin, characterization of synapses as high motility, etc., cell intensities and positions are transferred from Imaris to Excel files. The data is then imported from the Excel worksheets into MATLAB for calculation of synapse parameters.
TCR microcluster identification is performed using the polynomial fitting with Gaussian weight method (13). Assignment of identified microclusters to tracks is performed in Imaris (Andor) by transferring the microcluster data through the ImarisXT MATLAB interface. Where necessary, broken microcluster tracks are manually linked to generate completed microcluster tracks. All further track manipulations, such categorization of tracks based on their time of formation, or calculation of movement vectors, are performed after transferring the assembled tracks to data structures in MATLAB.
Conversion of fura-2 ratiometric intensities to calcium values
To calculate the relative amount of elevated calcium signal detected versus the distance that the cell had displaced from its binding site on the bilayer, the fura-2 ratiometric intensity time series data was divided by the sum of the above-baseline ratiometric intensities at all the time points. This converted the ratiometric intensities to values representing the fraction of all calcium flux detected. The values were then graphed versus to the displacement of the cell at the time of the ratiometric intensity measurement, binning the displacement values into 1 μm intervals.
Segmentation of synapses and cSMACs
To define and measure synapse footprints, TCR TIRF image sequences are filtered with a 1‒2 pixel standard deviation Gaussian filter as needed, and then masked with an intensity threshold that coarsely segments the synapse footprint from the background. The appropriate threshold is automatically selected using a minimum cross entropy threshold algorithm, which typically identifies a threshold that represents the full synapse, rather than the bright central region of TCRs. However, all automated segmentation routines should be manually verified for accuracy. In cases where the algorithm fails to identify an appropriate threshold for the synapse, a threshold can be manually selected. Morphological closing, hole-filling and removal of small, unconnected objects are then sequentially performed on each image in the series to yield masks with a single, contiguous region representing the cell footprint over time. In cases where Lifeact-GFP TIRF images are acquired, the GFP image is used to generate synapse masks.
To define and measure cSMACs, a threshold is applied to segment the bright, interior accumulations of TCRs (SMACs) from the dimmer peripheral microclusters. This intensity threshold was manually selected for each cell to accurately reflect the borders of the bright SMACs. This intensity threshold is then applied to all images in the time series to create a preliminary mask of the cSMAC. Occasional peripheral signaling microclusters with above-threshold intensities are then eliminated from the cSMAC mask with a 1 μm2 size filter. Morphological closing and hole-filling of the individual SMACs are then applied to generate the cSMAC mask. To account for loosely collected SMACs, rather than generating a single region, the cSMAC is allowed to be represented by multiple SMAC regions. Therefore, to measure the centroid of cSMACs, the area-weighted centroid of all SMAC regions is calculated.
Calculation of TCR microcluster radial displacement and centralization values
Instantaneous TCR microcluster radial displacements are calculated as the dot product of the microcluster movement vector and the vector from the microcluster base position to the center of the cSMAC. This converts the two-dimensional (xy) movements of the microclusters to one-dimensional (radial) values. Calculating the dot product using the vector from the microcluster to the cSMAC establishes the direction to the cSMAC as the positive flow direction. Microcluster radial displacements are calculated for each movement vector in the microcluster track and then cumulatively summed to generate radial displacement series, which represent the radial displacement of microclusters from their initial position. In these graphs, a microcluster is moving away from the cSMAC as the displacement decreases and moving towards the cSMAC as the displacement increases. To calculate instantaneous edge flow values, at each point in the microcluster track, a line from the center of the cSMAC through the microcluster position and to the synapse edge is constructed. The edge of the synapse is determined from the synapse masks, and the intersection of the edge with the line from the cSMAC through the microcluster position is calculated. This calculation is performed at each position in the microcluster track to create an edge intersection track. Instantaneous edge movement vectors are calculated from these intersection tracks, and edge cumulative radial displacement series are generated as for microclusters.
To measure microcluster centralization while accounting for outward movement during spreading, the centralization value of a microcluster are calculated as the difference between: 1) the distance from the microcluster to the cSMAC when the microcluster reached its greatest separation from the cSMAC and 2) the distance from the microcluster to the cSMAC after it centralized. Therefore, the centralization measures the distances microclusters travelled inward from the point at which inward movement began.
Imaris is used to generate speeds and straightness factors for the TCR microcluster tracks. These values are then transferred to MATLAB, which is used to calculate microcluster track mean speeds and mean straightness factors.
Calculation of synapse areas relative to cell volumes
To quantify synapse sizes relative to cell volumes, OT1+ T cell blasts are labeled with CFSE and Alexa Fluor 568-H57-597 prior to introduction to stimulating bilayers. The cells are fixed with 1% paraformaldehyde, and then imaged by spinning disk confocal microscopy to collect images of the cytoplasmic volume and TCRs at the cell-bilayer interface. The volumes of the cells were estimated by creating isosurfaces in Imaris using the z-series images of the CFSE-marked cell volume. Synapse areas are measured at the synapse image plane by manually applying a threshold to mask the cell. The equivalent radii from both the volumes and areas are then calculated. The equivalent radius calculated from the cell volume is then taken as the ‘expected’ synapse radius—the radius that would be achieved if a cell with the calculated volume spread so that its synapse radius matched the equivalent radius. This volume-derived radius is subtracted from the equivalent radius calculated from the synapse area to calculate the extent to which the synapse outgrew its expected radius.
Segmentation of synapses into edge and interior regions
To segment the synapse into interior and edge regions, the synapse are masked using the Lifeact-GFP TIRF images as described above, and the region of the synapse within 2 μm of the edge identified at each time point. The edge region is removed from whole synapse mask to create a second mask for the interior. The whole synapse and interior masks at every time point are then used to generate Delaunay triangulations of the regions. Microclusters are classified based on whether their initial positions were enclosed within the interior Delaunay triangulation (interior microclusters), or were enclosed within the whole synapse triangulation but not the interior triangulation (edge microclusters). By ensuring that microclusters formed within the synapse triangulation, this analysis excludes microclusters formed in nearby cells that might intrude into the image region of the cell being analyzed. Once microcluster track origins are identified, microcluster radial displacements are calculated as described above.
Calculation of Lifeact-GFP intensity derivatives in the regions around microclusters
To calculate the changes in Lifeact-GFP intensity in the regions through which microclusters moved, 1 μm2 regions centered on the microcluster positions are generated at all points in their tracks after the initial position. The average intensities of Lifeact-GFP in the cluster regions are then calculated when the microcluster was centered within each region. From these intensities, the average intensities at the time points before the microcluster entered the patch are subtracted to calculate the cluster region intensity changes (temporal derivatives). The cluster region intensity changes, therefore, served as a proxy for how much the actin filament density changes as microclusters enter regions. These values are plotted against the instantaneous microcluster flows associated with the movements into each patch to examine the correlation between changes in actin density with the direction of radial microcluster flow.
Statistical analyses are performed in Prism (GraphPad Software). The Mann-Whitney U test is used for nonparametric comparisons. For data that passes the D’Agostino & Pearson omnibus normality test, Student’s t test is used. For comparing multiple groups, 1-way ANOVA (α = 0.05) with Dunnett’s post-test is used.
The polynomial fit Gaussian weight function was written and made available by S. Rogers (University of Manchester). Lifeact-GFP was a generous gift of R.Wedlich-Soldner (Max Planck Institute of Biochemistry).The InterX MATLAB function was written and made available on the MathWorks File Exchange by “NS”. His-ICAM constructs were provided by B. Lillemeier (Salk Institute) and M. Davis (Stanford University). We thank M. Werner and K. Austgen for assistance in preparing His-ICAM. Biotinylated pMHC monomers were provided by J. Altman (NIH Tetramer Facility, Emory University).
Figure 1: A cushioned bilayer system for activating OT1+ T cells
(a) Schematic of the cushioned bilayer system for activating OT1+ T cells. (b,c) Flow cytometric analysis of lipid bilayer standards formed on 5 μm silica microspheres and loaded with a series of concentrations of biotinylated pMHC and his-ICAM protein. Top: microsphere bilayer standards and BMDCs (loaded with 100 ng/ml SIINFEKL peptide) stained with YN1/1.7.4 anti-ICAM (b) and 25D1.16 anti–pMHC (c). Bottom: plots of input protein concentration (log scale) versus the median fluorescence intensities (from the graphs at top) for the bilayer standards and reference BMDCs.
MATLAB Functions: Analysis MATLAB functions and scripts
The zip file includes scripts and functions that can be used to analyze microcluster tracks.
Integration of the movement of signaling microclusters with cellular motility in immunological synapses, Peter Beemiller, Jordan Jacobelli, and Matthew F Krummel. Nature Immunology 13 (8) 787 - 795 doi:10.1038/ni.2364
Peter Beemiller & Matthew Krummel, Krummel Lab, UCSF
Jordan Jacobelli, Unaffiliated
Correspondence to: Peter Beemiller ([email protected])
Source: Protocol Exchange (2012) doi:10.1038/protex.2012.028. Originally published online 4 October 2012.