Composition and structure of synaptic ectosomes exporting antigen receptor linked to functional CD40 ligand from helper T cells

Planar supported lipid bilayers (PSLB) presenting T cell receptor (TCR) ligands and ICAM-1 induce budding of extracellular microvesicles enriched in functional TCR, defined here as synaptic ectosomes (SE), from helper T cells. SE bind peptide-MHC directly exporting TCR into the synaptic cleft, but incorporation of other effectors is unknown. Here, we utilized bead supported lipid bilayers (BSLB) to capture SE from single immunological synapses (IS), determined SE composition by immunofluorescence flow cytometry and enriched SE for proteomic analysis by particle sorting. We demonstrate selective enrichment of CD40L and ICOS in SE in response to addition of CD40 and ICOSL, respectively, to SLB presenting TCR ligands and ICAM-1. SE are enriched in tetraspanins, BST-2, TCR signaling and ESCRT proteins. Super-resolution microscopy demonstrated that CD40L is present in microclusters within CD81 defined SE that are spatially segregated from TCR/ICOS/BST-2. CD40L+ SE retain the capacity to induce dendritic cell maturation and cytokine production.

ectosomes (SE), from helper T cells. SE bind peptide-MHC directly exporting TCR into the synaptic cleft, but their ability to incorporate other effectors is unknown. Here, we utilized bead supported lipid bilayers (BSLB) to capture SE from single immunological synapses (IS), determined SE composition by immunofluorescence flow cytometry and enriched SE for proteomic analysis by particle sorting. Our results demonstrate selective enrichment of CD40 10 ligand (CD40L) and inducible T-cell costimulator (ICOS) in SE in response to addition of CD40 and ICOS ligand (ICOSL), respectively, to SLB presenting TCR ligands and ICAM-1. TCR triggering mobilized intracellular CD40L to the T cells surface at the IS, where it engaged CD40 to enable sorting into SE. SEs were enriched in tetraspanins and bone marrow stromal cell antigen 2 (BST-2) by immunofluorescence and TCR signalling and endosomal sorting 15 complexes required for transport by proteomics. Super-resolution microscopy demonstrated that CD40L is present in microclusters within CD81 defined SE that are spatially segregated from TCR/ICOS/BST-2 microclusters. CD40L in SE retains the capacity to induce dendritic cell (DC) maturation and cytokine production. SE enabled helper T cells to release effectors physically linked to TCR. 20

Introduction:
Immune response communication depends on intercellular interactions of surface receptors expressed on T cells and antigen presenting cells (APC) via immunological synapses (IS), kinapses or stabilized microvilli (1, 2). In model IS, receptor-ligand pairs organize into radially symmetric supramolecular activation clusters (SMACs). The central (c)SMAC incorporates a 5 secretory synaptic cleft, TCR interaction with peptide-major histocompatibility complex (pMHC) and costimulatory receptor-ligand interactions and is surrounded by the peripheral (p)SMAC enriched in LFA-1 (T cell side) interaction with ICAM-1 (APC side) enriched peripheral (p)SMAC (3). The dynamics of IS formation involves initial contacts through microvilli that trigger cytoplasmic Ca 2+ elevation leading to rapid spreading and formation of 10 SMACs through inward directed cytoskeletal transport (4,5). Once the IS matures, TCR-pMHC pairs form in the distal (d)SMAC and segregate into microclusters (MCs) that integrate signaling as they centripetally migrate to the cSMAC where signaling is terminated (6). TCR MCs are a common feature of IS, kinapses and stabilized microvilli (1, 7). However, the IS is not only a platform for signal integration, but also enables polarized delivery of effector function. These 15 include the polarized delivery of cytokines (8), nucleic acid containing exosomes (9), and TCR enriched extracellular vesicles that bud directly into the synaptic cleft from the T cell side of the IS (10). "Ectosomes" are extracellular vesicles released from the plasma membrane (11). Therefore, we define directly TCR elicited extracellular vesicles that are formed and simultaneously exported across the IS as synaptic ectosomes (SE). 20 CD40 ligand (CD40L, CD154) is a 39 kDa glycoprotein expressed by CD4 + T cells (12) and is one of the key effectors delivered by helper T cells through the IS (13,14). Inducible T cell costimulator (ICOS, also known at CD278) interaction with ICOSL promotes CD40L-CD40 interactions in the IS (15,16). CD40L is transferred to antigen presenting cells in vitro (17).
Trimeric CD40L released by proteolysis by ADAM10 is a partial agonist of CD40, suggesting the fully active CD40 must remain membrane anchored to sufficiently crosslink CD40 for full agonist function (18,19). How helper T cells achieve this high level of crosslinking in the IS is not established. 5 In this study we set out to determine the protein composition and mechanism of SE release in the synaptic cleft by helper T cells. To this aim we develop technologies for isolation of SE released by T cells directly at the IS on BSLB (20) and integrate complementary flow cytometry, mass spectrometry and super resolution microscopy data. We show that the polarized transfer of T cell derived SE is determined by selective sorting processes directly in the IS and depends on both 10 the presence of ligands on the SLB and their segregation into the synaptic cleft, as shown for TCR complex:anti-CD3ε/pHLA-DR complexes, CD40L:CD40 and ICOS:ICOSL, but not LFA-1:ICAM-1 bound pairs. Other components, such as tetraspanins and BST-2, are enriched in SE without being engaged with a ligand. Quantitative mass spectrometry of SE revealed members of the core ESCRT machinery and adaptor proteins responsible for the scission of SE at the IS. 15 Using direct stochastic optical reconstruction microscopy (dSTORM) we further demonstrate that SE contain discrete TCR/ICOS/BST-2 and CD40L microclusters. T cell budding of SE, therefore, provides a strategy to generate antigen specific and effector armed SE. 20 Results:

CD40L is recruited to the IS and left by kinapses in a CD40 dependent manner
CD40L is stored in intracellular compartments within CD4 + effector cells and mobilized to IS where it engages CD40 in the IS upon activation through the TCR (21, 22). To mimic the APC surface and stimulate IS formation, the PSLB presented the adhesion molecule ICAM-1 and a 5 Fab fragment of the anti-CD3ε mAb UCHT1 (UCHT1-Fab) (10), which functions like a strong agonist pMHC (23) (Figure 1A). Due to challenges with fluorescent protein tagging of CD40L, we detected it in the IS using an anti-CD40L mAb, which has the caveat that it competes with CD40, but nonetheless detects recruitment of CD40L to the IS (16). To determine the impact of CD40 density in the PSLB on detection of CD40L by this method we allowed IS to form on 10 PSLBs presenting ICAM-1 and UCHT1-Fab over the physiological range of CD40 densities from 0 -500 molec./µm 2 . The anti-CD40L signal was imaged by total internal reflection fluorescence microscopy (TIRFM) that only illuminates up to 200 nm into the sample, and thus restricts detection to the IS. Minimal IS CD40L was detected in the absence of CD40 as previously reported (16) and near uniformly increased anti-CD40L was detected at 10, 50 and 15 100 CD40 molec./µm 2 with a reduction in signal at 500 CD40 molec./µm 2 ( Figure 1B). Thus, whether this loss of signal at high CD40 density is due to competition between CD40 and the anti-CD40L mAb or some other process, we conclude that CD40L can be detected and localized over the entire physiological range of CD40 densities using anti-CD40L antibody. To investigate the cellular localization of all CD40L, T cells were incubated on the PSLB with ICAM-1 and 20 UCHT1-Fab without or with 50 CD40/µm 2 for 30 minutes, fixed, permeabilized and stained with anti-CD40L (magenta) and CELL MASK ® (gray) to track cell membranes and 3D images generated by super-resolution Airyscan® confocal microscopy. In the absence of CD40, CD40L puncta were polarized toward the IS, but were mostly 1-3 µm above the PSLB within the T cell ( Figure 1C). In the presence of CD40, most of the CD40L was concentrated at the T cells-PSLB interface in a dense patch of bright puncta ( Figure 1C). Live microscopy demonstrated that CD40L was detected in the IS just after and together with TCR-UCHT1-Fab complexes (Movie S1). It was not clear if these bright puncta of CD40L immunoreactivity were present in 5 microclusters in the plasma membrane or if they were released into the synaptic cleft in some form. When T cells break IS symmetry and transition to migratory kinapses they leave trails of SE behind on the substrate (10) ( Figure 1D). To investigate the fate of CD40L, we imaged IS and kinapses by TIRFM to avoid detection of intracellular CD40L. The cSMAC of IS contained partially overlapping signals for UCHT1-Fab and anti-CD40L staining as previously described 10 (16) ( Figure 1E; top panel; Movie S1 and S2). Upon kinapse formation by T cells, a trail of UCHT1-Fab was present without or with CD40 in the bilayer as expected (10) ( Figure 1E; bottom panel, Movie S3). In the absence of CD40 in the PLSB, the few anti-CD40L reactive puncta that were detected remained associated with the migrating T cell ( Figure 1F). Kinapses formed in the presence of CD40 resulted in a trail of anti-CD40L reactive puncta that remained 15 tethered to the PSLB ( Figure 1F). These images and quantification are consistent with CD40L release in SE along with or in parallel with the TCR.

Selective transfer of CD40L and ICOS into SE
Uptake by APCs makes determining the composition of SE released by T cells challenging (10).
In addition, the bidirectional transfer of membrane derived molecules between APCs and T cells engaged in the bipartite IS confounds the detection of loss of molecules from the T cell (10, 24).
We developed a BSLB platform using the same compositions as on PSLB ( Figure 2A) with 15 similar synaptic accumulation of TCR as viewed by Airyscan® confocal microscopy ( Figure 2B vs Figure 1B, and Movie S4). Kinapse formation would not necessarily allow T cell to separate from BSLB, so we used low temperature and divalent cation chelation to inactive LFA-1-ICAM-1 interactions and gentle shearing forces provided by pipetting to separate the T cells from the BSLB. This enabled the quantitative profiling of gain of T cell molecules on BSLB and loss of 20 molecules from the T cell surface by flow cytometry with a gating strategy based on light scattering and a fluorescent lipid signal to assure analysis of single BSLB (Figure 1 -figure supplement 1). As with the PSLB system, we needed to assess how CD40 on the BSLB would impact detection of CD40L in the IS using the anti-CD40L antibody. We evaluated this by flow cytometry over a range from 0-500 CD40 molec./µm 2 ( Figure 2C). We determined the relative 25 IS transfer of CD40L (%) between T cells and BSLB using isotype control-corrected geometric mean fluorescence intensities (GMFI) as follows (GMFI BSLB ÷ (GMFI BSLB + GMFI T cells )) x 100 after 1.5 h incubation at 37°C and release of BSLB:T cell conjugates with ice cold PBS/EDTA. In the BSLB system, the decreased detection of CD40L using anti-CD40L mAb was not as prominent such that 500 CD40 molec/µm 2 was used for most experiments ( Figure 2C). We next tested a larger panel of T cell surface molecules and calculated a relative IS transfer as above to assess selectivity of the sorting process leading to transfer of TCR, CD40L and other surface proteins by human CD4 + T cell blasts to BLSB presenting ICAM-1, UCHT1-Fab with or  Table S1). Transfer of TCR was ligand dose dependent and independent of ICOSL and CD40. ICOS was weakly transferred in response to ICOSL without UCHT1-Fab in the BSLB, but the efficiency was increased by TCR engagement. CD40L was transferred only when UCHT1-Fab and CD40 was present in the BSLB, and we were not able to detect a 10 substantial increase when ICOSL was also included in the BSLB in contrast to prior results with with other proteins including CD2, CD6 and CD49D). LFA-1, CD38 and LAMP-1 (CD107a) 20 and CD4 were represented at <1% and were not enriched above the ordinary levels found in the plasma membrane ( Figure 2D and  BST2 was also co-localized in the TCR in an antigen dependent manner ( Figure 3C). As with the polyclonal CD4 + T cells, some ICOS transferred to BSLB with ICOSL with control HLA-5 DRB1*09:01:CLIP or in the absence of any MHC molecules ( Figure 2E), a phenomenon which is accompanied by the recruitment of ICOSL to an TCR independent cSMAC-like structure on the PSLB ( Figure 3C,D). This ICOSL driven TCR independent synapse may exert some control over migration of T cells, but it did not lead to CD40L transfer in any setting and thus does not appear to directly elicit delivery of T cell help.
10 Activated T cells have been shown to transfer CD40L to B cells that expressed CD40, but lacked cognate peptide-MHC in vitro (17). We thus wanted to ask if activated human T cell were also capable of transferring CD40L to BSLB that lack UCHT1-Fab, but present CD40. We prepared UCHT1-Fab presenting BSLB Atto-488 and UCHT1-Fab negative BSLB Atto-565 in the 4 possible combinations where each either presents or does not present CD40 ( Figure 4A). TCR and 15 CD40L were readily detected on the UCHT1-Fab and CD40 bearing BSLB at 1.5 h and 24 h ( Figure 4A). The surface expression of CD40L on the T cell was detectable at 1.5 h and 24 h and was decreased when CD40 was also present on the BSLB with UCHT1-Fab. No CD40L was detected on BSLB when CD40 was not present ( Figure 4A). This high degree of specificity suggests that CD40L transfer is tightly linked to IS formation, but we wanted to further 20 investigate strongly general activation could trigger CD40L transfer to ICAM-1 and CD40 bearing SLB in the absence of TCR engagement. We followed two approaches. First, we incubated T cells with phorbol myristate acetate (PMA) and ionomycin for 30-min to expose 14 CD40L on the surface and then for another 90 min in the presence of BSLBs with ICAM-1 and CD40 only or ICAM-1, UCHT1-Fab and CD40. PMA-ionomycin significantly increased the relative transfer of CD40L to BSLBs with 0 or 20/µm 2 UCHT1-Fab ( Figure 4B). This demonstrates that TCR engagement is not absolutely necessary for CD40L transfer. In this context, we also asked if CD40L transfer requires free ubiquitin, like TCR transfer into SE (10). 5 We depleted free ubiquitin by pre-treatment with the proteasome inhibitor MG132 (50 µM for 0.5 hours). MG132 pretreatment reduced CD40L transfer significantly, but had only a marginal effect on TCR transfer ( Figure 4B, C). Thus, we conclude that human CD4 + T cells restrict free ubiquitin dependent transfer of CD40L to antigen specific IS, but this can be overcome by strongly upregulating CD40L using a powerful pharmacological agonist.

Mass Spectrometry (MS) of SEs reveals enrichment of ESCRT proteins and TCR
signaling.
TCR-enriched SE are released through a TSG101 and VPS4-dependent plasma membrane budding process (10). Both TSG101 and VPS4 form part of the Endosomal Sorting Complex Required for transport (ESCRT). Specifically, TSG101 (an ESCRT-I member) is required for 5 TCR sorting into membrane buds (6) 27) and cross-correlation (28). TCR clusters were colocalized with ICOS or BST2 within a search radius of 50 nm, and segregated from CD40L clusters ( Figure 7C). There was a high correlation between TCR with ICOS or BST2 at inter-particle distances less than 50 nm, whereas TCR and CD40L showed the highest correlation at a distance of 150 nm ( Figure 7D

Released SE induce DC maturation and cytokine production
We next evaluated whether T cell isolated SE have an activating effect on monocyte-derived dendritic cells (moDCs). We therefore, first generated T cell derived SE on PSLBs coated with a combination of different T cell accessory ligands. After removal of T cells, immature moDCs were incubated with the released SEs for 24 hours followed by analysis of DC maturation markers and secreted factors. Conditions A1 and C0 had no SE, B1 had CD40L low SE, and C1 had CD40L high SE. CD40L high SE triggered high expression levels of HLA-DR and CD83  (Table S3). As we have noted in the dSTORM imaging, these free CD40L 20 trimers are further organized in even higher density microclusters. This provides a mechanism for SE to promote extensive CD40 cross-linking on the APCs surface to robustly activate downstream signals. To test this hypothesis, we developed synthetic unilamellar vesicles (SUVs) 25 loaded with increasing quantities of N-terminal His-tagged hCD40L (>200 molec./µm 2 , Figure   8D, Tables S3 and S4)

Discussion:
In this study, we systematically profiled composition, nanoscale structure and function of transsynaptic SE that were produced and captured in a model IS. Importantly, we adopted BSLB as a scalable platform to analyze SE by flow cytometry and mass spectroscopy. BSLB have been used previously to demonstrate activity of purified MHC proteins (30), detect weak protein 5 interactions (20) and determine biophysical requirements for phagocytosis (31). We have also used BSLB to calibrate site densities of recombinant proteins attached to SLB by flow cytometry (32) and for bulk functional assays (6). Here, we allowed T cells to form IS with BSLB at a 1:1 ratio and then disrupted the conjugates to allow analysis of material transferred from the T cell to the BSLB. The strength of this approach is that each BSLB carries the output from one T cell, 10 on average, such that the output of single IS can be quantified and calibrated to number of antibody binding sites and compared to average levels of the same molecules on the donor T cells. Another strength is that BSLB can then be isolated by particle sorting, which enabled MS analyses with minimal cellular contamination. A limitation of the approach is that while we can readily analyze both the cells and BSLB at a population level, we cannot link the cellular donor 15 to a BSLB recipient specifically, which would require a different approach-perhaps with microfluidics. For these studies we used ~5 µm beads, which are near cell size and are easily analyzed on commercial flow cytometers. The results we obtained are qualitatively comparable to results with PSLB, but we cannot rule out some impact of the surface curvature of the BSLB.
We also had some concern that the methods used to disrupt conjugates (divalent cation chelation 20 and low temperature) might result in physical tearing off of membranes that might not normally be transferred to APC, and would contaminate SE. However, there is a high degree of enrichment for ligated cargo like ICOS and CD40L (>10%), tetraspannins (33) and BST2 (34) (~10%), and proteins with known tetraspannin association, like MHC class I (1-10%). In contrast, most other proteins are transferred at a level similar to the area of membrane transferred (~0.2%). This suggests a high degree of selectivity of the transfer process and minimal contamination from random pieces of plasma membrane. We previously had failed to detect tetraspanin CD63 enrichment in SE (10), but find here tetraspanin CD81 in the cSMAC, and 5 enrichment and uniform distribution in TCR and/or CD40L positive SE by dSTORM. BST2 may tether some SE to the T cell membrane and account for dragging of SE behind kinapses, which also has implications for signaling of SE induced signaling to the T cell (34, 35). We observed formation of TCR engagement independent IS-like structures and SE-like vesicle transfer stimulated by ICOSL, but these did not elicit CD40L transfer even when CD40 was also 10 present in the BSLB. TCR engagement independent IS are stimulated in CD8 effector T cells by NKG2D ligands, but similarly don't stimulate release of cytolytic effects (36,37). TCR engagement independent IS-like structures may allow T cells to more closely inspect APC for presence of relevant pMHC. We expect that BSLB will be a useful approach to study synaptic transfer in other biological systems. 15 We have demonstrated that SE incorporate CD40L as long as CD40 is included in the SLB in addition to a TCR agonist and ICAM-1. This has important implications for SE function in T cell help. CD40L is primarily expressed on CD4 + T cells, but low-level mRNA and/or protein expression has also been reported on B cells, basophils, eosonophils, NK cells, macrophages, 20 dendritic cells, smooth muscle cells (38,39), and platelets (40). Extracellular vesicles from platelets possess CD40L and these vesicles have CD40L dependent adjuvant-like function when injected into mice (40). However, conditional knockout of CD40L in CD4 T cells appears to fully account for the classical defects in antibody class switching and T cell help associated with CD40L deficiency (41). Instead, platelet CD40L appears to function as an integrin ligand in blood clotting such that its physiological function is not CD40 dependent, but integrin β3 dependent (42). In contrast, T cell SEs are generated in an antigen dependent manner and TCR and CD40L are often present in the same SE, thus adding another level of specificity of this 5 post-T cell product. It doesn't escape our attention that SE may provide an explanation for reports of antigen specific helper factors (43).  20 Both we (16) and Gardell and Parker (17) utilized anti-CD40L mAb to detect CD40L transfer, which comes with caveats. Since these mAb are function blocking, we naturally were concerned that CD40 in the SLB might compete with the fluorescently tagged anti-CD40L mAb leading to underestimation of transfer. High densities of CD40 in the SLB did suppress the signal with anti-CD40L, but CD40L was still equally depleted from T cells on average, so our measurements of CD40L are likely an underestimate, but over the entire physiological range of CD40 densities we could detect transfer with anti-CD40L mAb. The dSTORM analysis of anti-CD40L localization in SE revealed that CD40L was detected in microclusters distinct from the TCR/ICOS/BST2 5 microcluster that must correspond to the site of attachment between the SE and the PSLB. This physical segregation of CD40L from the attachment site to the antigen-presenting surface may in part explain its availability to both the anti-CD40L antibody used for detection and CD40 on

T cells transfer CD40L to B cells in an agonist
DCs, which responded functionally to the CD40L on SE. We also noted a CD40/CD40L independent, DC stimulatory activity associated with released SE, which may be explained by 10 recently reported release of mitochondrial DNA associated with T cell exosomes (46), released in parallel with SE.
MS analysis provided a number of new candidates for SE generation and function. We detected many new ESCRT and vesicle trafficking components in SE. We identified two ubiquitin 15 recognizing ESCRT-0 components HRS (Hepatocyte growth factor-regulated tyrosine kinsase substrate; an ESCRT-0 component) and EPN1 (47), which may explain why we previously found that HRS is not essential for cSMAC formation (6). We previously could find no role for classical ESCRT-II proteins in SE formation (6) Table S1. Isotype controls matching the relevant fluorescent dyes were used for background correction and gating. Other mAb or affinity purified antibodies are described with specific methods below.

Small unilamellar vesicles (SUVs)
SUV are defined as vesicles in the 20-100 nm range. SUV were formed by extrusion as described using the Avanti Miniextruder with a 100 nm filter (53). When SUV were used to mimic SE, all lipids were combined prior to SUV formation, whereas BSLB and PSLB

Nanoparticle Tracking Analysis
A 10 µL aliquot of SUVs or eluted SE preparation was re-suspended in PBS in a 1:100 dilution and kept on ice for Nanoparticle Tracking Analysis (NTA). The instrument used for NTA was 20 Nanosight NS300 (Malvern Instruments Ltd) set on light scattering mode and instrument sensitivity of 15. Measurements were taken with the aid of a syringe pump to improve reproducibility. Three sequential recordings of 60 seconds each were obtained per sample and NTA 3.2 software was used to process and average the three recordings to determine the mean size.

moDC activation by CD40L on SUV.
His-tagged recombinant soluble CD40L (sCD40L, BioLegend) was incubated with NTA-SUV or 5 plain SUV, at ratios designed to match CD40L densities found on SE for 20 min at 24 o C prior to addition to the moDCs. After 24 h, moDCs were recovered by spinning down plates at 1500 rpm

Calibration of flow cytometry data
T cells and BSLB were analyzed using antibodies with known AF647:Ab ratio (Table S1)
The Argon laser at 488 nm and diode laser at 561 nm were used as excitation sources, with 5 power setting of ~1% and ~6%, respectively, which is equivalent to 1 mW and 10 mW. The powers were set in this range in order to achieve the comparable strength of fluorescent signal for both channels. Fluorescence emission was collected at around 515 nm and 653 nm for the green and magenta channels, respectively, with the following filters BP420-480+BP495-550 (green) and BP555-620+LP645 (magenta). The emission signals were collected on the 32 10 channel GaAsP-PMT Airy detector. The datasets were acquired as Z-stacks with 43.5 nm pixel size and 185 nm axial steps, which correspond to ~50-55 slices per 3D data set. ZEN Airyscan software (Zeiss) was used to process the acquired data sets. This software processes each of the 32 Airy detector channels separately by performing filtering, deconvolution and pixel reassignment in order to obtain images with enhanced resolution and improved signal to noise 15 ratio. The value of Wiener filter in ZEN software was chosen in accordance with the value in "auto" reconstruction modality and was set around 7, to ensure the absence of deconvolution artefacts (54). Drift was corrected using the MultiStackReg plug-in of ImageJ (National Institute of Health). Rendering was performed in Imaris software (Bitplane). blocked for Fc receptors with 5% HSA and 5% goat or donkey serum for 1 h at 24 o C.

Total internal reflection fluorescence microscopy (TIRFM)
TIRFM was performed on an Olympus IX83 inverted microscope equipped with a 4-line (405 nm, 488 nm, 561 nm, and 640 nm laser) illumination system. The system was fitted with an Olympus UApON 150x 1.45 numerical aperture objective, and a Photomertrics Evolve delta EMCCD camera to provide Nyquist sampling. Live experiments were performed with an 5 incubator box maintaining 37 o C and a continuous autofocus mechanism. Quantification of fluorescence intensity was performed with ImageJ (National Institute of Health).

dSTORM imaging and data analysis
For three colour dSTORM imaging extracellular vesicles were stained using either wheat germ localizations that appeared within one pixel in five consecutive frames were merged together and fitted as one localization. The final images were rendered by representing the x and y positions 15 of the localizations as a Gaussian with a width that corresponds to the determined localization precision. Sample drift during acquisition was calculated and subtracted by reconstructing dSTORM images from subsets of frames (500 frames) and correlating these images to a reference frame (the initial time segment). ImageJ was used to merge rendered high-resolution images (National Institute of Health). 20

CBC analysis
Coordinate-based colocalization (CBC) mediated analysis between two proteins was performed using an ImageJ (National Institute of Health) plug-in (60) based on an algorithm described previously (27). To assess the correlation function for each localization, the x-y coordinate list from 488-nm and 640-nm dSTORM channels was used. For each localization from the 488-nm 5 channel, the correlation function to each localization from the 640-nm channel was calculated.
This parameter can vary from −1 (perfectly segregated) to 0 (uncorrelated distributions) to +1 (perfectly colocalized). The correlation coefficients were plotted as a histogram of occurrences with a 0.1 binning. The Nearest-neighbor distance (NND) between each localization from the 488-nm channel and its closest localization from the 640-nm channel was measured and plotted 10 as the median NND between localizations per cell.

Cross-correlation analysis
Cross correlation analysis is independent of the number of localizations and is not susceptible to over-counting artifacts related to fluorescent dye re-blinking and the complements other 15 approaches (28). Cross-correlation analysis between two proteins was performed using MATLAB software provided by Sarah Shelby and Sarah Veatch from University of Michigan.
Regions containing cells were masked by region of interest and the cross-correlation function from x-y coordinate list from 488-nm and 640-nm dSTORM channels was computed from these regions using an algorithm described previously (28,61,62). Cross-correlation functions, C(r,q), 20 were firstly tabulated by computing the distances between pairs of localized molecules, then C(r) is obtained by averaging over angles. Generally, C(r) is tabulated from ungrouped images, meaning that localizations detected within a small radius in sequential frames are counted independently. Finally, a normalized histogram with these distances was constructed into discrete bins covering radial distances up to 1000 nm. Cross-correlation functions only indicate significant correlations when the spatial distribution of the first probe influences the spatial distribution of the second probe, even when one or both of the probes are clustered themselves.
Error bars are estimated using the variance within the radial average of the two dimensional C(r, The MS data set is available to reviewers prior to publication: ProteomeXchange Consortium via the PRIDE (64) partner repository with the dataset identifier PXD007988 5 Upon publication the dataset will be publically available.

Statistical analysis
All statistical analyses were performed using SigmaPlot 13.0 (Systat Software Inc), OriginPro  Figure 2D. P values <0.05 were considering significant. Repeat Measures ANOVA with Geisser-Greenhouse correction was performed. Significance was calculated comparing in each bilayer composition group (either No  5 accessory signal, CD40 + ICOSL, CD40 or ICOSL) at 500 and 5,000 molec./µm 2 of UCHT1-Fab to No UCHT1-Fab. P values: *<0.05, **<0.005; ***<0.0005; ****<0.0001; n.s.= non-significant.  The increased ratios of markers on BSLBs compared to the T cell PM is another form to depict the degree of protein 10 enrichment in SE, but compared to the normal protein to TCRαβ ratio found on the PM of a non-activated T cell (shown are ratios for the mean values from 6 different donors). (D) CD4 + T cells were stained for the absolute quantification of surface markers using antibodies with known fluorescent dye to protein (F/P) ratios (See Table S1). test with false discovery rate of 1%. (C) Using a multicolor flow cytometry panel, the relative transfer of proteins (% of transfer) in response to increasing densities of ICOSL was determined. T cells were stimulated with BSLBs containing 200 molec./µm 2 of ICAM-1, 150 molec./µm 2 of UCHT1-Fab and 500 molec./µm 2 of CD40. After 1h, cell: bead conjugates were dissociated with cold 50 mM EDTA-PBS, and then stained with a multicolor panel using the same antibody clones described in Table S1. We used CD4 as a control protein whose relative (% of total CD4 signal) transfer was not enriched in our single-color experiments (see Figure 2D). Each heat map square represent 5 mean +/-SEM of data collected from 5 donors across 2 independent experiments. Shapiro-Wilk normality test. Non-Normally distributed data was analyzed using no matching or pairing Kruskal-Wallis test with Dunn´s multiple comparisons to the rank of conditions with no ICOSL. Normally distributed data was analyzed using one-way ANOVA with Holm-Sidak´s multiple comparison test to the mean of the no ICOSL group. P < 0.03 was considered significant. *<.03; **<.002; ***<.0002; ****<.0001. (D) Representative off set histograms of BSLBs (white 10 background) and T cells (grey background) analyzed after dissociation of conjugates. Histograms for controls stained with antibodies of appropriate isotype conjugated with the relevant fluorescent dye are shown in grey.    Figure 8C. ns, not significant; *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001, Kruskal-Wallis with Newman-Keuls post-hoc test.