|Home | About | Journals | Submit | Contact Us | Français|
Three-dimensional live cell imaging of the interaction of T cells with antigen presenting cells (APC) visualizes the subcellular distributions of signaling intermediates during T cell activation at thousands of resolved positions within a cell. These information-rich maps of local protein concentrations are a valuable resource in understanding T cell signaling. Here, we describe a protocol for the efficient acquisition of such imaging data and their computational processing to create four-dimensional maps of local concentrations. This protocol allows quantitative analysis of T cell signaling as it occurs inside live cells with resolution in time and space across thousands of cells.
T cell activation is regulated by the complex interactions of dozens of signaling intermediates. One of the great, general challenges in current biology is to understand how dozens of proteins function as an integrated system. As a key resource to elucidate such complexity, signaling intermediates are not evenly distributed through an activating T cell but enrich at particular locations at distinct times [1, 2]. Concurrent enrichment of two proteins increases their interaction probability and thus the local efficiency of the signaling step mediated by their interaction. Thus at the system scale uneven signaling distributions govern the information flow through signaling networks [3–6]. Supporting the importance of subcellular location for function in T cell activation, loss-of-function mutations of various signaling intermediates consistently yield diminished localization [7–11]. In addition, through the imaging of large numbers of signaling intermediates, subcellular structures mediating signal integration have emerged [12, 13]. Thus a quantitative determination of subcellular signaling distributions of a large number of T cell signaling intermediates is a powerful resource to discover functional signaling connectivity. Here we describe the methodology for such characterization across thousands of cells. Three-dimensional live cell imaging of the interaction of T cells with antigen presenting cells (APC) visualizes the subcellular distributions of signaling intermediates at thousands of resolved positions within a cell, i.e. it yields detailed three-dimensional maps of local concentrations. We describe a protocol for efficient acquisition of such data. Voxel-by-Voxel quantification of such imaging data provides full access to the three-dimensional maps of local concentrations. We describe a computational analysis routine to generate maps that are aligned across all T cells to be analyzed. Such maps form a powerful foundation for the computational analysis of signaling connectivity and the modeling of T cell signal transduction.
The original research upon which these protocols are based was supported in part by National Institutes of Health grant P41 GM103712, by National Science Foundation grants MCB1121793 and MCB1121919, and by European Research Council grant PCIG11-GA-2012-321554.
1The use of EDTA alone rather than the usual Trypsin/EDTA combination and limiting the time of EDTA on the NX-E cells optimizes NX-E cell health.
2We find that maintaining the parental NX-E stock under continuous selection ensures consistently high viral titers over months. Nevertheless, the stock is replaced from freezes about every 6 to 9 months.
3For each batch of HBS the exact amount of NaOH to be added is determined in a dose response series of NaOH. The NaOH amount giving the smallest visible calcium phosphate precipitate is chosen.
4By reducing the medium volume from 6 to 3.2 ml the concentration of virus in the supernatant is effectively doubled.
5Our most widely used TCR transgene is the 5C.C7 TCR. However, this protocol has worked equally well with a number of other TCR transgenic mouse strains, including DO11.10, OTII, AND, Tg4, P14, CL4, and HY [14–18].
6We commonly use combined inguinal, axillary, submandibular, and mesenteric lymph nodes. Spleen can be used instead, yet requires red blood cell lysis.
7The centrifuge is heated by air friction. Therefore, pre-spin until 32°C are reached.
8The LLMV long terminal repeat functions as a very strong promoter in the NX-E cells and resulting viral mRNA is efficiently translated. The presence of imaging sensor in the NX-E cells thus is an indirect yet convenient readout of the amount of viral mRNA produced.
9The 1 to 1.5 log range above cellular autofluorescence background corresponds to the minimal fluorescence level detectable by sensitive microscope systems, a GFP concentration of 2.5µM . The use of minimal sensor concentrations optimizes physiological relevance of the imaging data by minimizing interference of the imaging sensor with the signaling event to be studied. Using this sorting strategy across a range of actin regulators, the combined concentration of endogenous protein plus its retrovirally expressed GFP-tagged version as an imaging sensor was commonly within the 5 to 95 percentile of the endogenous protein concentration .
10Transduction efficiencies vary as a function of (I) the TCR transgene and (II) the protein to be expressed. (I) The more vigorous the initial T cell proliferation in tissue culture is, the higher the transduction efficiency becomes. Therefore MHCI-restricted CD8 TCR transgenes generally yield higher efficiencies that can easily exceed 50 %. (II) The expression of a sensor is tightly linked to the endogenous concentration of the protein it is derived from. Thus Actin-GFP expresses substantially better than e.g. Itk-GFP. Sensors based on synthetic constructs or isolated protein domains, such as the F-tractin peptide of the PLCδ-PH domain, generally express well. Transduction efficiencies as low as 0.01 % can be used for subsequent imaging.
11We find that homogeneously high sensor expression at the time of sorting becomes variable already on the time scale of days.
12We have used a wide variety of APCs, from transfected CHO cells, various B and T cell lymphoma lines, dendritic cells to tumor target cells. In general non-adherent cells allow for easier detection of cell coupling as the T cell:APC interface forms perpendicular to the cover slip and thus is readily detectable in the DIC bright field images.
13We have used magnifications from 20× to 100×. Higher magnifications trade better resolution for dimmer images and lower number of cells and thus cell couples in the field. Therefore, the majority of our experiments are done with 40× magnification. For cell coupling efficiencies below 10 % even lower magnification is desirable to capture an adequate number of cell couples.
14Biochemically detectable T cell signaling activity in T cell:APC couples peaks in the first five minutes of cell coupling [20, 21]. NFAT and NFκB translocate into the nucleus within about 3 and 7 min respectively [13, 18]. Therefore, 15 minutes imaging times captures the peak of T cell signaling activity. Nevertheless, we have imaged as long as 16 h. Any imaging beyond 30 min will require measures to minimize buffer evaporation.
15Binning of camera pixels trades resolution in x and y against signal-to noise ratio. With the chosen 2×2 binning on the 40× objective resolution is roughly equivalent in all three dimensions.
16In response to strong stimulus, e.g. a high concentration of agonist peptide in the presence of costimulation, cell coupling can be virtually instantaneous. As cell couples that form before the onset of data acquisition cannot be accurately timed relative to the time of cell couple formation and thus have to be excluded from analysis, a rapid decision on the imaging field is essential.
17Data can be saved in any format. However, archiving images as .tiff files ensure great compatibility with various analysis packages and ready exchange with colleagues.
18Autoscaling, the process where the dimmest pixel in an image is set to black and the brightest one is set to white, is helpful in image acquisition to get an immediate first impression of the data. However in image analysis, in particular in the comparison of images across a time series, we display all images on the same scale, as thus apparent differences in brightness reflect actual differences in fluorescence intensity rather than potential differences in scaling.
19We don’t use fluorescence data in the determination of cell coupling. This would set up a circular argument where fluorescence data would be used to determine time of cell coupling to then be analyzed relative to this time.
20When imaging data are acquired with a 40× objective and 2×2 binning each T cell is represented by >5,000 voxels. This paragraph describes how to generate voxel-by-voxel-resolved models of these distributions. Importantly, the MATLAB script creates these models in a cell shape-normalized fashion such that each position in one T cell model can directly be compared to an equivalent position in any other model .
21Once cell shape-normalized models of the three-dimensional protein distributions have been generated they can be computationally processed in whatever form desired. In this paragraph we list four examples for visualization and quantitative analysis.