Monomeric agonist peptide/MHCII complexes activate T‐cells in an autonomous fashion

Abstract Molecular crowding of agonist peptide/MHC class II complexes (pMHCIIs) with structurally similar, yet per se non‐stimulatory endogenous pMHCIIs is postulated to sensitize T‐cells for the recognition of single antigens on the surface of dendritic cells and B‐cells. When testing this premise with the use of advanced live cell microscopy, we observe pMHCIIs as monomeric, randomly distributed entities diffusing rapidly after entering the APC surface. Synaptic TCR engagement of highly abundant endogenous pMHCIIs is low or non‐existent and affects neither TCR engagement of rare agonist pMHCII in early and advanced synapses nor agonist‐induced TCR‐proximal signaling. Our findings highlight the capacity of single freely diffusing agonist pMHCIIs to elicit the full T‐cell response in an autonomous and peptide‐specific fashion with consequences for adaptive immunity and immunotherapeutic approaches.


Introduction
The recognition of antigenic peptides by T-cells relies on transient interactions with antigen-presenting cells or target cells (APC).The formation of this specialized cell contact-termed the immunological synapse-is driven by specific T-cell antigen receptors (TCRs) on the T-cell side engaging peptide-loaded MHC complex molecules (pMHCs; Bromley et al, 2001;Huppa & Davis, 2003).In this setting, T-cells can detect the presence of even a single stimulatory pMHC molecule among thousands of structurally similar yet nonstimulatory or endogenous pMHCs on a densely packed APC surface (Irvine et al, 2002;Li et al, 2004;Purbhoo et al, 2004;Huang et al, 2013).The molecular, biophysical, and cellular mechanisms underlying this phenomenon are still unclear, in particular when considering that most T-cell antigen receptors (TCR) engage their nominal ligands with a low-to-moderate affinity (1-100 μM), at least when measured in vitro (Matsui et al, 1991;Corr et al, 1994;Davis et al, 1998).Previous studies have suggested that ligand recognition may not only depend on the intrinsic binding properties of TCRs and pMHCs but also the quality and quantity of accessory interactions (i.e., CD28/B7-1, CD2/CD48, LFA-1/ICAM-1, and CD4 and CD8 co-receptors), as well as cell biological parameters such as TCR subunit or pMHC stoichiometry, local molecular crowding, and cytoskeletal forces (reviewed in Platzer & Huppa, 2020;Gohring et al, 2022).
Key to understanding the molecular foundation of T-cell antigen recognition is to examine TCR-pMHC engagement in situ, as conditions for binding within the synapse differ in many regards from those found in solution (Huang et al, 2010;Huppa et al, 2010).If positioned in sufficient proximity, TCRs and pMHCs engage each other on opposed plasma membranes in a pre-oriented fashion which appears to accelerate the binding on-rate (Huppa et al, 2010;Huang et al, 2013;Weikl et al, 2016).Synaptic TCR-pMHC lifetimes are furthermore affected by simultaneous engagement of coreceptors with the same pMHC (Huppa et al, 2010;Jiang et al, 2011) as well as cellular forces imposed on the binding partners (Liu et al, 2014;Feng et al, 2017;Sibener et al, 2018;Gohring et al, 2021;Pettmann et al, 2023).Furthermore, both pMHCI and pMHCII have been suggested to be organized in clusters (Lu et al, 2012) where they may undergo lateral interactions, which could alter binding and signaling dynamics.In fact, nanoscale MHCII clustering through interaction with tetraspanins or partitioning into glycosphingolipid rafts has been considered critical for accelerated T-cell scanning and sensitized antigen recognition (Anderson et al, 2000;Bosch et al, 2013;Zuidscherwoude et al, 2015).pMHCII enrichment has furthermore been suggested to drive the formation of pMHCII higher order structures, in particular in view of the tendency of some pMHCIIs to crystallize as dimers (Brown et al, 1993;Fremont et al, 2002).Consistent with this, soluble agonist pMHCIIs activate antigen-experienced T-cells only when tethered to a second agonist pMHCII or an endogenous pMHCII capable of TCR binding (Cochran et al, 2000;Krogsgaard et al, 2005).Rather than confounding the process of antigen recognition, highly abundant endogenous pMHCs have hence been proposed to facilitate the detection of rare agonist pMHCs in a so-termed "pseudodimer" arrangement (Wulfing et al, 2002;Krogsgaard et al, 2005;Anikeeva et al, 2012) whereby such pMHCs may act-without being able to stimulate T-cells on their own-as so-termed "co-agonists." We have recently shown via molecular live cell imaging that TCRs are randomly distributed prior to pMHCII engagement (Rossboth et al, 2018;Platzer et al, 2020) and function as monomeric antigen-binding and -signaling entities (Brameshuber et al, 2018).Therefore, any co-agonistic effect mediated by a pseudodimer arrangement would have to result from agonist pMHCIIs being either displayed as part of the same higher order MHCII structure or densely packed in membrane nanodomains.However, to this date, neither the presence of MHCII-enriched membrane domains nor the existence of pMHCII dimers or higher order structures has been demonstrated on the surface of living APCs, and hence the relevance of pMHCII clustering for T-cell sensitization remains to be determined (Nagy et al, 2001;Gascoigne, 2008;Ma et al, 2008;Anikeeva et al, 2012;Hoerter et al, 2013).
To do so we have conducted minimally invasive single-molecule imaging on living professional APCs with sub-diffraction resolution.We failed to obtain any evidence for the existence or formation of higher order pMHCII structures prior to or in the course of antigen recognition and ruled out any relevance of such entities for antigen recognition if generated experimentally.Instead, we observed pMHCIIs as monomeric entities uniformly distributed at high densities across the plasma membrane and diffusing rapidly immediately after entering the surface of bone marrow-derived dendritic cells and B-cells.When measuring synaptic TCR-pMHCII interactions with the use of a Fo ¨rster resonance energy transfer (FRET)-based assay, we did not find the presence of highly abundant endogenous bystander pMHCIIs to be relevant for (i) the TCR engagement of rare agonist pMHCII at early and later stages of synapse formation and (ii) ensuing TCR-proximal signaling.Instead, we witnessed lowlevel calcium signaling in T-cells facing "co-agonistic" altered peptide ligands (APLs) at supraphysiological densities (above 1,000 molecules μm -2 ), yet without contributing to T-cell sensitization toward rare antigenic pMHCIIs.In summary, our findings challenge the concept of pMHCII co-agonism and underscore the stimulatory potency of single, freely diffusing agonist pMHCIIs, which activate cognate T-cells in a peptide-specific manner without any significant contribution from endogenous bystander pMHCIIs.

Dendritic cells and B-cells express pMHCIIs at high numbers and densities
We first determined the number of pMHCIIs that T-cells can in principle encounter while scanning the surface of professional APCs.For precise quantitation, we employed a monovalent site-specifically labeled and in selected cases also biotinylated recombinant singlechain antibody fragment (scF V ) derived from the I-E k -reactive 14.4.4SmAb (Appendix Fig S1A -E, for details, see Materials and Methods section).As verified by flow cytometry, 14.4.4scF V binding to I-E k expressed on CH27 B-cells was specific, efficient, and stable: when employed at a concentration of 5 μg ml À1 14.4.4scF V conjugated to Alexa Fluor 647 or Alexa Fluor 488 (14.4.4 scF V -AF647 or -AF488) decorated more than 95% of surface-expressed I-E k with half-lives of staining amounting to 41 min at 37°C, 80 min at 25°C, and 16 h on ice (Appendix Fig S1C-E).To further confirm target specificity of the 14.4.4scF V , we co-cultured 5c.c7 TCR transgenic T-cells with moth cytochrome c (MCC 88-103) peptide-pulsed BMDCs at 37°C in the presence of 20 μg ml À1 14.4.4scF V .T-cells no longer secreted interferon-γ (IFN-γ) and tumor necrosis factor (TNF-α).EC50 values derived from a peptide titration series in the presence or absence of the 14.4.4scF V indicated an epitope coverage of 98.7% over a period of 4 h (Fig 1A and Appendix Fig S2A).
Total I-E k surface expression was quantitated based on flow cytometric median fluorescence intensities (MFIs) of 14.4.4scF V -AF647stained BMDCs and B-cells, which we then normalized to measured fluorescence intensities of AF647 quantification beads serving as reference (Appendix Fig S2B).As shown in Fig 1B, non-activated days 6 to 7 CD11c + I-E k+ BMDCs expressed an average of 0.5 million copies of I-E k at the surface per cell, which increased to 4-5 million copies per BMDC upon activation with lipopolysaccharide (LPS).Indicative of partial BMDC maturation prior to LPS exposure, I-E k expression levels of non-activated BMDCs (7 days after bone marrow isolation) were highly heterogeneous (Appendix Fig S3A), with CD11c + CD11b + I-E k hi BMDCs representing mainly mature DCs (Helft et al, 2015).At least in part owing to their significantly smaller size, CD19 + B-cells displayed about 10-to 50-fold fewer copies of I-E k , that is, on average $1 × 10 5 molecules per non-activated and $5 × 10 5 molecules per B-cell stimulated with LPS for 48 h (Fig 1C,and Appendix Fig S3B and C).
To account for all surface-resident pMHCIIs that are in principle available for T-cell scanning, we also determined the total number of displayed I-A k copies indirectly through comparison of MFIs obtained from APCs stained with saturating quantities of the 14.4.4mAb and the I-A k -reactive 11-5.2 mAb (Fig 1D).After normalizing MFIs for degrees of antibody labeling (DOL; 14.4.4mAb = 2.1; 11-5.2 mAb = 2.4), we arrived at similar numbers of I-A k and I-E k that were accessible to antibody staining on the surface of BMDCs and B-cells.
To assess whether endogenous pMHCIIs exerted any influence on the detection of rare agonist ligands, we sought to determine the number of bystander pMHCII in close vicinity to any scanning or ligated TCR.Of note, both BMDCs and B-cells gain considerably in surface area upon activation, while B-cells increase in overall size (blast formation), the surface of BMDCs changes from microvilli to veils (Verdijk et al, 2004).As a result, total cellular expression levels cannot easily be converted into molecular densities due to varying cell size and plasma membrane convolutions.We hence stained BMDCs and B-cells with saturating amounts of 14.4.4scF V sitespecifically conjugated with Alexa Fluor 647 or Alexa Fluor 568, and then allowed the cells to spread on intercellular adhesion molecule 1 (ICAM-1)-coated glass slides as a means to directly measure surface I-E k densities using total internal reflection fluorescence (TIRF) microscopy ( Fig We then determined average I-E k densities within a central region of interest (ROI) randomly picked from the cell-glass contact region.
I-E k molecules fail to associate with detergent-resistant membranes and are randomly distributed on the surface of activated BMDCs and B-cells pMHCII have previously been described to be sequestered to sotermed lipid rafts, a class of small and highly dynamic membrane domains consisting of cholesterol, sphingolipids, saturated fatty acids, and glycosylphosphatidylinositol (GPI)-anchored and cysteinepalmitoylated proteins (Anderson et al, 2000;Kusumi et al, 2012;Sevcsik et al, 2015), the existence of which is, however, controversially debated.To assess possible mechanisms underlying lipid raftdependent T-cell sensitization as proposed by others (Anderson et al, 2000;Bosch et al, 2013), we quantitated the extent to which pMHCIIs on the surface of BMDCs and B-cells are found in so-termed detergent-resistant membranes (DRMs).Protein presence in DRMs has previously been regarded as the biochemical equivalent of lipid raft residence.To this end, we conducted lysis gradient centrifugation (LGC) affording comprehensive results with a sample size of fewer than 10,000 cells when combined with a flow cytometric readout (Schatzlmaier et al, 2015).The LGC methodology rests on quantitative co-isolation of DRMs and their associated proteins with the detergent-resistant nuclei of their parental detergent-treated cells (Appendix Fig S4A).As shown in Appendix Fig S4B and C, we analyzed surface-resident I-E k or I-A k molecules on non-activated and activated BMDCs and B-cells as well as the GPI-anchored DRM markers CD44 and CD55 as positive control, and the non-DRM resident CD71 as negative control.As expected, CD44 and CD55 decorated with fluorescently labeled mAbs were predominantly present within the DRM fraction while CD71 was absent, testifying to the technical integrity of the experiment.Strikingly, I-E k labeled with either the 14.4.4-scFV -AF647 or 14.4.4mAb-AF647 was hardly detectable in DRMs (< 1%, see Appendix Fig S4C ), a behavior we could also associate with I-A k decorated with the 11-5.2mAb-AF647 on the surface of both BMDCs and B-cells.Contrasting previous reports, these results indicate that MHCIIs, at least when present on the surface of primary B-cells and BMDCs, are predominantly absent from or only very weakly associated with DRMs (< 1%).
To examine whether pMHCIIs are enriched in micro-or nanoclusters other than lipid rafts, we stained living activated BMDCs and B-cells with randomly AF647-conjugated I-E k and I-A kreactive mAbs or site-specifically AF647-conjugated 14.4.4scF V s and analyzed cellular mid-sections via fluorescence microscopy.
We continued our study by focusing primarily on activated BMDCs given the pivotal role in T-cell priming.For comparative analysis, we also included activated B-cells in our study.Using the 14.4.4scF V or 14.4.4mAb (or 11-5.2 mAb) as probe, deconvolved images revealed uniform staining of I-E k or I-A k .As is consistent ◀ Figure 1.Dendritic cells and B-cells express pMHCIIs at high numbers and densities.
A Efficient blockade of T-cell antigen recognition by the 14.4.4scF V. Shown is the flow cytometric analysis (refer to Appendix Fig S2A ) of antigen-induced TNF-α and IFN-γ expression in 5c.c7 TCR-transgenic T-cells co-cultured with BMDCs which had been prepulsed with MCC peptide antigen (from 0 to 1 μM) in the absence or presence of the 14.4.4scF V with a representative dot plot (0.05 μM MCC peptide) depicted on the left.Flow cytometry analysis was conducted with n > 11,000 cells per condition.B I-E k is expressed at high numbers on activated CD11c + CD11b + BMDCs.Left panel: staining procedure employed for quantitation.Middle panel: the median number of I-E k molecules displayed on the surface of non-activated and activated BMDCs (n > 10,000 cells per analysis), gated as CD11c + I-E k+ or CD11c with a previous study (Bosch et al, 2013), we observed a punctate staining pattern on BMDCs after exposing surface-resident mAbengaged I-E k molecules with secondary AF647-conjugated IgG mAbs.Taken together, we concluded that pMHCIIs are not organized in plasma membrane microdomains that are large enough to be resolvable with diffraction-limited microscopy (Appendix Fig S5A and B).
To determine whether pMHCIIs are enriched in plasma membrane nanodomains, which can no longer be identified by conventional fluorescence microscopy, we analyzed the surface distribution of I-E k on living and paraformaldehyde (PFA)-fixed BMDCs and B-cells with sub-diffraction resolution applying highspeed photoactivation localization microscopy (hsPALM).We included cell fixation in half of our analyses since the high lateral diffusion of pMHCII may compromise cluster detection via hsPALM.Of note, PALM relies on the stochastic photo-switching of suitable fluorophores between a blue-shifted fluorescent or dark state and a red-shifted or bright fluorescent state and the subsequent reconstruction of diffraction-limited but well-separated single-molecule events into a localization map (Betzig et al, 2006;Hess et al, 2006).For visualization of I-E k , we site-specifically conjugated the 14.4.4scF V with AF647 and biotin (14.4.4 scF V -AF647-biotin, Fig 2A and Appendix Fig S1B).Cell surface-resident 14.4.4scF V -AF647-biotin was then detected in hsPALM with the use of monovalent streptavidin (mSav) site-specifically conjugated with recombinant PS-CFP2 (mSav-cc-PS-CFP2, Fig 2A and Appendix Fig S6A -E), a photoswitchable fluorophore with a moderate repetitive blinking behavior (Platzer et al, 2020).
To allow for noise-reduced imaging in TIRF mode, we seeded living BMDCs and B-cells labeled with 14.4.4scF V -AF647-biotin and mSav-cc-PS-CFP2 on ICAM-1-coated glass slides.Prior to conducting hsPALM, attachment of each cell was verified through acquisition of a single AF647 ensemble fluorescence image, which also served to assess surface I-E k distribution at a diffraction-limited level (Fig 2B).For hsPALM, we recorded 3,050 image frames that displayed singlediffraction-limited PS-CFP2 fluorescence events within a timeframe of 7 s.Diffraction-limited single-molecule signals were then fitted with a Gaussian intensity profile accounting for the point spread function, spatially positioned with an accuracy of 15 to 108 (mean value 51 AE 12) nm (Appendix Fig S6F) and then superimposed to give rise to a composite localization map.
Of note, the interpretation of single-molecule localization microscopy (SMLM) data is typically complicated by overcounting and undercounting artifacts, which result from insufficient photobleaching (on-off blinking) and variability in psCFP maturation.Fluorophore blinking can give rise to false-positive nanocluster detections in the reconstructed localization map (Annibale et al, 2011;Baumgart et al, 2016).To distinguish true protein nanoclusters from overcounting artifacts, we first employed an imaging platform to quantitatively determine the blinking signature of  prior to PALM measurements on cells (Platzer et al, 2020).To promote the appearance of PS-CFP2 in a single imaging frame only, we employed a 488 nm laser power density of 3.5 kW cm À2 and recorded the images in stream mode (488 nm laser is continuously illuminating) with an illumination time of 2 ms.Applying a Monte-Carlo simulation-based approach, we quantitatively related cell-derived PALM images to localization maps, which had been computed from simulated random We hence carried out artifact-free label density variation SMLM (Baumgart et al, 2016) to quantitate the extent to which I-E k is organized in nanoscale entities within I-E k localization maps derived from PFA-fixed activated BMDCs and B-cells.To this end and based on image analysis, we assigned a spatial uncertainty (σ) of 35 nm to each PS-CFP2 localization and merged overlapping Gaussian profiles to three-dimensional cluster maps (X, Y position, intensity, Fig 2E).We calculated localization densities within an assigned clustered area (ρ) and the relative clustered area to the total area (η) for a given threshold (1.0) and plotted the density of localizations in the clustered area μm À2 (ρ) against η (clustered area divided by total area).In case of a uniform fluorophore distribution, the relative area covered by artificial clusters resulting from fluorophore blinking will increase steadily with an increasing degree of labeling, while the density of localizations per cluster is expected to remain constant.Conversely, if molecules are in fact clustered, the relative area will saturate, and the density of localizations per cluster will increase with the degree of labeling.Results correlated best with a random distribution of I-E k on activated PFA-fixed BMDCs and B-cells, in particular, because the localization density within the clustered area did not deviate with higher staining degree from a simulated random distribution (solid line), unlike what would be expected from a clustered distribution (Fig 2F).In contrast and consistent with previous studies, label density variation SMLM revealed clustering of CD205 on BMDCs, with an increase in the localization density within clusters at higher expression or staining levels (Fig 2F).
We further refined and advanced our analysis by applying a model-based correction (MBC) workflow on localization maps derived from living and PFA-fixed cell regions.MBC is based on the calibration-free estimation of PS-CFP2 blinking dynamics followed by a hierarchical clustering step to give rise to a corrected set of localization coordinates without fluorescence blinking (Bohrer et (Appendix Fig S8E and G).Consequently, only 0 to 6 % of I-E k molecules (average $1%) appeared in zero to two clusters per region with a maximum of zero to seven molecules per cluster on (living and PFA-fixed) activated BMDCs and B-cells (Appendix Fig S8E and G).This was in stark contrast to CD205, with $39% of CD205 molecules (26-58%) present in clusters of up to 16 molecules per cluster (average $9) and 2 to 16 clusters per region (average $8), which was further confirmed by means of the summarized Ripley's K function as being clearly above the simulated random distribution (Appendix Fig S8B,D and F).Lastly, to avoid any issues potentially arising from SMLM and data interpretation, we employed stimulated emission depletion (STED) microscopy.STED is not susceptible to artifacts resulting from overcounting, as it resolves structures below the diffraction limit of visible light by reducing the width of the point spread function to several tens of nanometers through high-powered stimulated emission depletion (Sahl et al, 2017).For this purpose, we stained activated BMDCs and B-cells with a 14.4.4-scFV which had been site-specifically conjugated with abSTAR635P (14.4.4-scFV -abSTAR635P), a STED-compatible fluorophore, and allowed the cells to spread on ICAM-1-coated glass slides prior to PFA-mediated fixation (Appendix Figs S10A and B,and S11).To assess the degree of pMHCII clustering in STED microscopy images, we compared the STED recordings with simulated images of randomly distributed molecules obtained with the fitted parameters of single fluorescence emitters derived from activated BMDCs and B-cells labeled with single 14.4.4scF V -abSTAR635P molecules (Appendix Fig S10A).To this end, we employed image autocorrelation function (ACF) analysis, which calculates the probability that two pixels separated by a distance r feature similar brightness, and thus enables quantification of the spatial distribution of fluorophores in the recorded STED images (Robertson & George, 2012).The ACF amplitude scales with the inverse of the number of molecules per pixel, thus, small surface densities as well as clustering of molecules yielded a pronounced increase in the ACF amplitude (Sengupta et al, 2011;Rossboth et al, 2018).ACF analysis of the recorded STED images of fixed activated BMDCs (Appendix Fig S10B) and B-cells (Appendix Fig S11) decorated with 14.4.4scF V -abSTAR635P did not reveal any difference in the ACF curves when compared to analyses of simulated random pMHCII distributions.
In summary, the use of two independent super-resolution methodologies and three autonomous cluster analysis tools leaves no conclusion other than that the majority of surface I-E k molecules (> 90%) are not organized in nanoclusters but instead uniformly distributed on the surface of living as well as PFA-fixed activated BMDCs and B-cells.
Surface-resident I-E k molecules pMHCIIs are to 99% mobile with 60-70% displaying fast diffusion immediately after entering the cell surface The lateral mobility of pMHCII may influence T-cell antigen detection as it affects (i) the speed with which antigens become distributed on the APC surface, (ii) the frequency of serial TCR engagement, (iii) the dynamics of lateral pMHCII interactions, and also (iv) the mechanical forces exerted on pMHCII-engaged TCRs (Gohring et al, 2021) with possible consequences for the ensuing TCR-proximal signaling response (Hsu et al, 2012).To measure the lateral mobility of I-E k by single-molecule tracking (Vrljic et al, 2002;Umemura et al, 2008), we decorated I-E k molecules in low abundance on non-activated as well as activated BMDCs and B-cells with Alexa Fluor 555-conjugated 14.4.4scF V (14.4.4 scF V -AF555) prior to allowing the cells to spread on ICAM-1-coated glass slides for TIRF imaging at 37°C (Fig 3A).We recorded traces of single I-E k molecules over several hundred ◀ Figure 2. I-E k molecules are randomly distributed on the surface of activated BMDCs and B-cells.

A
Staining procedure to image the surface distribution of I-E k on living professional APCs below the diffraction limit of visible light using hsPALM.B BMDCs and B-cells were stained with 14.4.4scF V -AF647-biotin and mSav-cc-PS-CFP2 and seeded onto ICAM-1-coated glass slides for imaging in TIRF mode.The presence of each cell was verified through acquisition of a white light image (first image, scale bar 10 μm), a single AF647 ensemble fluorescence image (second image), followed by 3,050 image frames displaying single PS-CFP2 fluorescence events acquired within 7 s to be assembled as a localization map (third image), which was compared to a simulated random I-E k localization map (fourth image, scale bar 1 μm).Ripley's K analysis was used to compare BMDC-derived I-E k distributions (red) or B-cell-derived I-E k distributions (blue) to 10 simulations of randomized distributions with corresponding molecular densities, diffusion parameters, and blinking statistics (gray).Statistics: mean AE standard deviation.C, D Quantification of the surface distribution of (C) CD205 and (D) CD18 on activated BMDCs.Left panels: Staining and imaging procedure.Right panels: BMDCs were stained with the indicated mAbs (NLDC-145 mAb-AF647-biotin or M18/2 mAb-AF647-biotin) and mSav-cc-PS-CFP2 for hsPALM imaging.Large scale bars, 10 μm.Small scale bars, 1 μm.Ripley's K analysis was used to compare cell-derived CD205 distributions (red) to 10 simulations of randomized distributions with corresponding molecular densities, diffusion parameters, and blinking statistics (gray  (Wieser & Schutz, 2008).We arrived at similar I-E k diffusion parameters when comparing the binary data fit performed on individual cells (Fig 3F).
Following 24 to 48 h of LPS treatment, the fraction of rapidly diffusing I-E k entities increased marginally in B-cells, but not in BMDCs, and the mobility of rapidly moving I-E k molecules increased slightly on both activated BMDCs and B-cells (Appendix Fig S12A -F).Furthermore, the average diffusion constant of the slowly moving fraction did not change after LPS treatment and was similar for BMDC-and Bcell-resident I-E k molecules (Appendix Fig S12B and D).
We next visualized the dynamics with which I-E k molecules dispersed at any given time on the APC surface.For this, we decorated activated BMDCs with a site-specifically biotinylated 14.4.4scF V (14.4.4 scF V -biotin) for the detection with mSav.The latter had been site-specifically conjugated with abCAGE635, a photo-switchable and photo-stable bright fluorophore (Fig 3G and Appendix Fig S6C).Stained BMDCs and B-cells were seeded on ICAM-1-coated glass slides for imaging at 37°C in TIRF mode.As shown in Fig 3H and I and consistent with high lateral mobility, we witnessed a rapid dispersion of I-E k -associated photo-uncaged fluorescence over the entire surface of BMDCs and B-cells within 150 s after applying a single UV pulse to photo-converge a small, selected area.
Extracellular antigens are present at high concentrations in the MIIC or endosomal peptide-loading compartments (Pierre et al, 1997;Turley et al, 2000).Antigen proximity could in principle be spatially and temporally conserved on the cell surface after exosome fusion with the plasma membrane.If true, this may in turn enhance T-cell antigen detection.To test whether pMHCIIs are locally enriched within the first minutes after entering the cell surface, we loaded I-E k with a traceable peptide.To this end, we exposed non-activated BMDCs (day 6) and activated B-cells (day 2) for 0 to 36 h to a C-terminally biotinylated MCC peptide 88-103 derivative (MCC-PEG 2 -biotin) in the presence of LPS.We first determined the proportion of surface-resident MCC-PEG 2 -biotin-loaded I-E k molecules by flow cytometry of activated BMDCs (up to 17%) and B-cells (up to 25%) with the use of site-specifically AF647conjugated monovalent streptavidin (Fig 4A and B,and Appendix Fig S13A).We next tracked surface I-E k /MCC-PEG 2 -biotin complexes 6-12 h after addition of the peptide to the APC culture, that is, 0-12 h prior to maximal I-E k /MCC-PEG 2 -biotin surface expression (Fig 4A).To discriminate newly arriving I-E k /MCC-PEG 2 -biotin complexes from all other previously arrived I-E k /MCC-PEG 2 -biotin complexes, we first blocked surface-exposed biotin residues quantitatively using an excess of unlabeled monovalent mSav (500 μg ml À1 ; for more details, refer to Materials and Methods section).After removing unbound mSav species as part of a rapid washing step conducted at 4°C to avoid fusion of MHCII-loaded anterograde transport vesicles with the plasma membrane, we allowed newly assembled I-E k /MCC-PEG 2 -biotin complexes to arrive at the cell surface for 10 min at 37°C, where they were directly conjugated to site-specifically abCAGE635-labeled mSav (or AF555labeled mSav) and subsequently tracked as single molecules after photo-uncaging by TIRF microscopy at 37°C (Fig 4C).After preblocking and staining at 4°C, only a small number of I-E k /MCC-PEG 2 -biotin complexes were still accessible for staining with abCAGE635-labeled mSav (with an average of 70 trajectories per BMDC and 15 trajectories per B-cell).This number increased around 10-fold after incubating the cells at 37°C, indicating that more than 90% of the observable I-E k /MCC-PEG 2 -biotin complexes had arrived at the cell surface within the recovery time of 10 min (Appendix Fig S13B and C).
As shown in Fig 4C, newly arriving I-E k molecules exhibited a similar, if not even slightly elevated lateral mobility when compared to all other I-E k molecules on mature BMDCs (D new = 0.534 AE 0.008 μm 2 s À1 ; D pool = 0.448 AE 0.005 μm 2 s À1 ) and activated B-cells (D new = 0.251 AE 0.005 μm 2 s À1 ; D pool = 0.237 AE 0.003 μm 2 s À1 ).The proportion of newly arrived rapidly moving I-E k molecules remained unchanged on activated BMDCs (Fig 4D and Appendix Fig S13D) and was marginally decreased on activated B-cells (Fig 4E and Appendix Fig S13E).
Taken together, our single-molecule tracking analysis rendered a scenario, in which newly arriving antigen-loaded MHCIIs conserve their spatial proximity observed in MIIC compartments and anterograde transport vesicles, rather unlikely, at least within a physiologically meaningful timeframe.Still to be considered is, however, the possibility that such molecules undergo confined diffusion over a prolonged period of time together with other colocalizing antigenloaded MHCIIs derived from the same vesicle.Yet inconsistent with this notion is the observation that the labeling procedure employed gave exclusively rise to diffraction-limited fluorescence events which featured brightness values matching those of single fluorophores.

Surface-resident I-Ek molecules are monomeric
In large part motivated by our above findings, we next sought to assess in a definitive manner the possibility that pMHCII forms higher order structures on the surface of living APCs.In fact, the observation that some pMHCII variants crystalize as dimeric entities (Brown et al, 1993) could in principle result from an inherent propensity of pMHCII to assemble into complexes of larger size.If true, such behavior may sensitize T-cell antigen detection, as had been previously proposed by the "TCR-pseudodimer" model, which postulates simultaneous binding of two adjacent TCRs to a pMHCII agonist and a physically linked pMHCII co-agonist (Krogsgaard et al, 2005).
I-E k surface densities on living activated BMDCs (i.e., median 514 molecules μm -2 ) and activated B-cells (median 413 molecules μm -2 ) are too high to allow for nanoscopic assessment of individual I-E k mono-or multimers via diffraction-limited microscopy.To score for the presence of I-E k oligomers on living APCs, we hence resorted to the "Thinning Out Clusters While Conserving the Stoichiometry of Labeling" (TOCCSL) methodology, a live cell-compatible singlemolecule fluorescence recovery after photobleaching (FRAP) approach, which affords identification and quantification of laterally mobile higher order entities at high surface densities (Moertelmaier et al, 2005;Brameshuber et al, 2010Brameshuber et al, , 2018)).For TOCCSL measurements, we stained activated BMDCs and B-cells with saturating quantities of the 14.4.4scF V -AF647 and then allowed the cells to spread on ICAM-1-coated glass slides for imaging in TIRF mode.To ensure quantitative decoration of cell surface-resident I-E k with the 14.4.4-scFV , we conducted TOCCSL experiments at 25°C (a temperature at which 14.4.4-scFV exhibited a half-life of binding of $80 min) and only between 2 and 20 min after the cells had landed on the glass slide.As is illustrated in To also account for the less abundant yet more slowly diffusing pMHCIIs, which might have escaped detection by the TOCCSL method, we employed FRET as a readout for spatial proximity between surface-resident pMHCII.To this end, we quantitatively decorated I-E k molecules on BMDCs, B-cells, and functionalized planar glass-supported lipid bilayers (PLBs) with 14.4.4scF V s, which had been site-specifically conjugated with either AF555 or AF647 serving as FRET donor and acceptor, respectively.For most direct FRET yield quantitation, we recorded FRET donor recovery after FRET acceptor photobleaching (FRET DRAAP).We anticipated FRET yields to augment with increasing FRET acceptor abundance, yet we also had to factor in decreasing signal-to-noise ratios with decreasing FRET donor surface densities.We therefore applied ◀ Figure 3. Surface-resident I-E k molecules are to 99% mobile with 60-75% displaying fast diffusion immediately after entering the cell surface.AF555-and AF647-conjugated 14.4.4scF V s in our staining procedures at molar ratios of 1:1, 1:2, 1:3, and 1:4 with expected FRET detection efficiencies of 66.7,80,85.7,and 88.9% and signal recovery of 50,44.4,37.5,and 32%,.To quantitate the I-E k density on BMDCs and B-cells with and without the use of FRET acceptor probe (14.4.4 scF V -AF647), we measured the density of I-E k bound to the FRET donor probe (14.4.4 scF V -AF555) for every scenario and related loss of the FRET donor intensity with the gain in FRET acceptor intensity.As shown in Appendix Fig S14B, the use of FRET donor and FRET acceptor probes premixed at a ratio of 1:1 led to a twofold reduction in FRET donor intensity (as determined after FRET acceptor ablation), confirming that the density of all I-E k was twofold higher than the density measured for FRET donor probe-associated I-E k .
We next compared FRET efficiencies measured on cells and PLBs via FRET donor recovery after acceptor photobleaching with simulated FRET efficiencies we had calculated from monomeric and randomly distributed molecules present at comparable densities.As shown in Fig 5D and E, and Appendix Fig S14C and D, we failed to observe in our cell-based measurements significant deviations from FRET efficiencies as expected from random molecular collisions.Yet, validating the use of our FRET-based approach to score for higher order I-E k entities, we witnessed a considerable increase in FRET efficiency after crosslinking I-E k -bound biotinylated 14.4.4scF V (14.4.4 scF V -AF555-biotin and 14.4.4scF V -AF647-biotin premixed in a 1:1 ratio) on activated BMDCs via divalent streptavidin (Fig 5F and G).
In summary, both TOCCSL-and FRET-based analyses rendered the existence of mobile and immobile I-E k higher order structures, at least at frequencies that would be relevant for T-cell detection, highly unlikely.

TCR interactions by endogenous bystander pMHCIIs are indistinguishable from random molecular collisions
We conjectured that, if they exist, co-agonist pMHCIIs engage TCRs within the immunological synapse as a means to sensitize T-cells for rare antigens.To test this, we quantitated synaptic TCR engagement of previously reported co-agonist pMHCIIs with the use of a FRET-based imaging approach previously employed to deduce the molecular dynamics of TCR-pMHCII engagement in situ at the ensemble and single molecule level (Huppa et al, 2010).For this, we confronted primary TCR-transgenic T-cells with proteinfunctionalized PLBs, which served as surrogate APCs and recapitulated scenarios of free diffusion and random distribution of pMHCIs, in line with our observations concerning living APCs described above.Of note, the use of PLBs did not only provide full control over the identity and quantity of synaptic T-cell-binding partners but was also compatible with the use of TIRF microscopy, which afforded considerably improved signal-to-noise ratios when recording individual synaptic TCR-pMHCII interactions via single-molecule FRET (Huppa et al, 2010).
For FRET measurements, we first decorated 5c.c7 TCR-transgenic T-cells with a site-specifically AF555-conjugated TCRβ-reactive H57-scF V serving as FRET donor and confronted them with PLBs functionalized with the costimulatory molecule B7-1, the adhesion molecule ICAM-1 and I-E k complexed with either the agonist moth cytochrome C peptide 88-103 (MCC), MCC derivatives, or endogenous peptides for T-cell recognition (Fig 6A, first panel).Prior to MHC loading, presented peptides have been C-terminally conjugated with AF647 to act as FRET acceptors (for details, see Materials and Methods).With an AF555/AF647-Fo ¨rster radius of $5.1 nm, the inter-dye distance giving rise to half-maximal energy transfer, we reasoned that our labeling procedure resulted in measurable FRET only when TCR and pMHCII were engaged with one another ( The latter is proportional to the TCR occupancy, the proportion of ◀ Figure 4. Newly arriving I-E k molecules are not confined and feature a similar diffusion behavior as the total pool of I-E k molecules.

A
Scheme illustrating the experimental procedure applied.(i) Activated BMDCs and B-cells were pulsed with MCC-PEG 2 -biotin.(ii) Resulting I-E k /MCC-PEG 2 -biotin molecules on the cell surface were blocked with unlabeled mSav after the initial peptide pulse (6-12 h).(iii) Following a recovery lasting 10 min at 37°C, newly arrived and surface-exposed I-E k /MCC-PEG 2 -biotin molecules were stained with mSav-abCAGE635 at room temperature, washed, and (iv) directly subjected to single-molecule tracking analysis.B To determine the duration of peptide-loading most adequately for tracking experiments (analysis in C-E), we quantitated the degree of I-E k -peptide loading on activated BMDCs and B-cells by flow cytometric analysis.Cells were pulsed with MCC-PEG 2 -biotin for up to 36 h and then stained with either 14.4.4scF V -AF647 (reflecting all surface-resident I-E k molecules) or mSav-AF647 (marking I-E k loaded with MCC-PEG 2 -biotin).Ratios were built from MFIs measured for either scF V -AF647-or mSav-AF647-decorated cells to determine the proportion of surface-accessible I-E k molecules loaded with MCC-PEG 2 -biotin.Statistics: mean and standard deviation of two technical replicates.C Activated BMDCs and B-cells pulsed with MCC-PEG 2 -biotin for 12 and 6 h, respectively, were stained with mSav-abCAGE635 and seeded on ICAM-1-coated glass slides for imaging in TIRF mode at 37°C.  Figure 6C displays TCR occupancies as they had been measured for 5c.c7TCR transgenic T-cells recognizing I-E k loaded with the agonist MCC peptide, the weak agonistic T102S, the antagonistic K99R, or the reported co-agonist K99A MCC peptide variant (Matsui et al, 1994;Wulfing et al, 1997Wulfing et al, , 2002)).We also included measurements involving the recognition of I-E k complexed with the endogenous peptides ER60 and β2m (Ebert et al, 2009).I-E k displaying the triple-null MCC mutant, which no longer binds the 5c.c7TCR (ANERAELIAYLTQAAK, null peptide), was employed as a negative control to determine FRET yields resulting solely from molecular crowding.
Consistent with their stimulatory potency and the TCR occupancies measured, the agonist I-E k /MCC and weak agonist I-E k /T102S rapidly accumulated underneath T-cells, which was followed by their recruitment to the central supramolecular activation cluster.In contrast, the antagonist I-E k /K99R became only slightly enriched, while non-stimulatory I-E k /K99A, I-E k /ER60, I-E k /β2m, and I-E k / null failed altogether to accumulate in the synapse and to induce the formation of a mature immune synapse (Appendix Fig S16A and B).
As shown in Fig 6C, synaptic TCRs became ligand engaged already at low densities of I-E k /MCC and, to a lesser extent, I-E k / T102S.Expectedly, we observed much reduced synaptic binding of I-E k /K99R and especially of I-E k /K99A, with median TCR occupancies of $5% at high densities (i.e., 1,390 I-E k /K99A μm -2 or 1,611 I-E k /K99R μm -2 ).Of note, we failed to detect any synaptic binding of the endogenous ligands I-E k /ER60 and I-E k /β2m even when these had been present in numbers comparable to pMHCII expression levels on activated BMDCs and B-cells (median 400-500 I-E k μm -2 , see Fig 1E).Challenging the notion of any co-agonistic potency associated with I-E k /ER60 and I-E k /β2m, measured FRET levels could merely be attributed to molecular crowding rather than TCRbinding, as they were indistinguishable from those determined for I-E k /null.

Competition rather than cooperativity among agonist, "co-agonist", and weak agonist ligands defines synaptic pMHCII-TCR binding
To further test the conceptual validity of co-agonism, we assessed whether co-presentation of more weakly binding TCR ligands promotes or rather suppresses TCR binding of low-abundant agonist ligands.
Among all investigated co-agonist candidates, only I-E k /K99A engaged a measurable fraction of synaptic TCRs, at least when present at $1,400 molecules μm -2 or higher.To evaluate whether I-E k / K99A affected initial TCR binding to the lowly abundant agonist I-E k /MCC, we loaded I-E k with an AF647-conjugated derivative of the MCC peptide, which had been modified at the central lysine residue to contain a (4,5-dimethoxy-2-nitrophenyl)methyl carbonochloridate (NVOC) moiety (DeMond et al, 2006;Huse et al, 2007) To assess any influence of co-agonists on the TCR engagement by agonist ligands, we titrated I-E k /K99A-AF488 in high abundance (417 AE 11 molecules μm -2 ) to I-E k /MCC[NVOC]-AF647 (92 AE 4 molecules μm -2 ) and quantified I-E k /MCC-TCR engagement by ◀ Figure 5. Surface-resident I-E k molecules are monomeric.

A
Principle of TOCCSL to study the stoichiometry of I-E k on living BMDCs and B-cells.Cells were quantitatively labeled with the 14.4.4scF V -AF647, placed on an ICAM-1-coated glass slide for imaging, and subjected to an illumination sequence of (i) a prebleach image followed by a high-powered laser pulse to quantitatively ablate the AF647 fluorescence of a particular cell area (as defined by the slit aperture).(ii) Complete photobleaching of AF647 was verified by a control image 100-750 ms after the bleach pulse.(iii) Following a recovery phase of 2-20 s, we next recorded TOCCSL images to detect single diffraction-limited events resulting from laterally mobile I-E k molecules, which had moved from the masked into the unmasked area.Finally, the brightness of individual fluorescence events was determined to assess I-E k -complex stoichiometry (shown in B, C).Scale bar, 5 μm.

B
Quantitative brightness analysis revealed a monomeric contribution of recovered I-E k molecules on BMDCs (n = 471 molecules on 59 cells) and B-cells (n = 187 molecules on 52 cells).As monomer control, we stained the cells with a cocktail of 14.4.4scF V -AF647-biotin and 14.To assess whether bystander pMHCIIs influenced synaptic TCR binding of rare agonist pMHCIIs at later stages of synapse formation (i.e., between 2 and 12 min), we measured via FRET TCR-agonist binding on PLBs featuring no or increasing levels of I-E k /K99A.As is shown in Fig 6G , synaptic 5c.c7TCR-I-E k /MCC binding was not affected by the presence of up to 942 I-E k /K99A μm À2 , even when agonist ligands were outnumbered by the co-agonist candidates up to 40-fold.Vice versa, 5c.c7 TCR binding to I-E k /K99A-AF647 present in high numbers (1,188 AE 65 molecules μm -2 ) dropped appreciably with increasing numbers of co-presented agonist I-E k /MCC-AF488 (Fig 6H).This proved also true for synaptic engagement of the weak agonist I-E k /T102S-AF647 (present at a density of 1,053 AE 55 I-E k /T102S), which diminished with increasing densities of co-presented agonist I-E k /MCC (ranging from 0 to 202 I-E k /MCC μm -2 , Appendix Fig S17A).In a similar fashion, 5c.c7 TCR binding to I-E k /MCC-AF647 (presented at 36 AE 1 I-E k /MCC) dropped substantially with increasing densities of I-E k /T102S-AF488 (up to 1,235 I-E k /T102S μm -2 , Appendix Fig S17B ).We observed a similar trend for the antagonist I-E k /K99R, which predictably competed more strongly with agonist binding than I-E k /K99A yet to a lesser extent than I-E k /T102S (Appendix Fig S17C and D).Of note, high densities of I-E k /K99A (1,060 I-E k /K99A-AF488 μm À2 ) did not affect the synaptic lifetime of agonistic TCR:I-E k /MCC interactions as measured via smFRET imaging (Fig 6I and Appendix Fig S17E).
Taken together, our FRET-based synaptic binding studies render the notion that TCR-engagement of low-affinity pMHCII ligands promotes the binding of rare agonist ligands highly improbable.Instead, and as would be expected from applying the law of mass action, high-affinity ligands outcompeted low-affinity ligands for TCR binding, especially under conditions where pMHCIIs outnumbered TCRs.Our findings are best explained with a scenario in which pMHCIIs engage TCRs in an autonomous fashion.

Highly abundant bystander pMHCIIs do not sensitize T-cells for low levels of agonist pMHCIIs
As shown above, we have detected higher order I-E k structures on the surface of activated BMDCs and B-cells only after their specific and experimentally defined crosslinking.In addition, we did not detect any influence of bystander pMHCIIs on the kinetics of synaptic TCR engagement of rare agonist pMHCIIs.To be able to arrive at a definitive evaluation of endogenous or (low-affinity) bystander pMHCIIs with regard to their role in sensitized agonist detection, we stimulated T-cells with a defined APC mimetic biointerface (Fig 7A).The latter involved PLB-embedded and laterally mobile yet structurally rigid DNA origami platforms, which we had engineered to control for pMHCII identities and inter-pMHCII distances at the nanoscale level and which allowed for the spatially defined display of I-E k molecules loaded with agonist (MCC) and co-agonist (K99A) ◀ Figure 6.TCR-agonist binding is not affected by the presence of the reported co-agonist I-E k /K99A.
A Scheme of a FRET-based assay to visualize TCR-pMHCII interactions in situ.B FRET yields resulting from 5c.c7 TCR-transgenic T-cells stained with H57 scF V -AF555 and interacting with PLBs featuring I-E k /MCC-AF647 as determined via FRET DRAAP and FRET-sensitized emission.The FRET yield is directly proportional to the TCR occupancy, the ratio of bound TCRs to total TCRs.Scale bar, 5 μm.C Synaptic TCR-ligand engagement was quantified at 25°C via FRET observed for H57 scF V -AF555-decorated 5c.c7 TCR-transgenic T-cells engaging PLBs featuring agonist (MCC), weak agonist (T102S), antagonist (K99R), potential co-agonist (K99A, β2m, ER60), and non-binding (null) I-E k molecules displayed at increasing densities.ICAM-1 and B7-1 were present at densities of $100 molecules μm -2 .The TCR occupancy of the entire synapse was calculated according to TCR occupancy = E FRET × 1.21.S1A, we arrived at a functionalization efficiency of $60% for a single modification site; for two modification sites, we detected a mixed population of DNA origami structures featuring one ($20%), two ($50%), and zero ligands ($10%).
To gauge the impact of homo-or heterodimerization of pMHCIIs on T-cell activation, we loaded 5c.c7 TCR transgenic T-cells with the ratiometric calcium-sensitive dye Fura-2-acetoxymethyl (AM) ester (Fura-2) and allowed the cells to engage PLBs displaying functionalized DNA origami platforms at increasing densities and ICAM-1 and B7-1 at densities of $100 molecules μm -2 (Fig 7B).Ligand densities on PLBs were quantitated by dividing the average fluorescence brightness per area by the average brightness of single fluorophores.TCR-agonist pMHCII binding resulted in TCR-proximal signaling as witnessed by increasing levels of intracellular calcium acting as second messenger to promote T-cell activation.
For quantitative analysis, we calculated Fura-2 intensity ratios for each T-cell trace and classified the T-cells as activated or nonactivated based on a receiver operating curve between a positive control (i.e., > 10 I-E k /MCC μm À2 , ICAM-1 and B7-1 at 100 μm À2 ) and a negative control (ICAM-1 and B7-1 at 100 μm À2 ).For better display of low-level calcium signaling in individual T-cells, we identified the maximum Fura-2 ratio in each cell trace and plotted the median value of the peak Fura-2 ratio (i.e., median of the peak Fura-2 ratio plus 10 imaging frames) in a probability density plot (see Materials and Methods and Appendix Fig S18A).Non-activated T-cells featured median Fura-2 intensity ratios of 1.0 (after normalization to the median value representing the negative control), whereas oscillatory or activated T-cells exhibited median Fura-2 ratios ranging from 1.3 to 4.0 (Appendix Fig S18A).In Fig 7B and C, we plotted the fraction of activated T-cells as dose-response curves as a means to determine activation thresholds, that is, ligand densities that had given rise to half-maximum responses.To this end, we fitted the data with a four-parameter dose-response curve (for more information, see Materials and Methods, equation ( 12); all fit parameters are listed in Appendix Table S1B).
When confronted with I-E k /MCC-AF555 tethered to monovalent streptavidin, T-cells responded at density of 0.33 molecules μm -2 (Fig 7C ).Attachment of either a single or two I-E k /MCC-AF555 molecules to a DNA origami platform did not lower but rather slightly increased the activation threshold to 0.4 and 0.48 molecules μm -2 , respectively.Likewise, confronting T-cells with DNA origami platforms that featured a single I-E k /MCC-AF555 as well as a single I-E k /K99A-AF647 molecule spatially separated by either 10 or 20 nm did not significantly affect activation threshold with values amounting to 0.34 and 0.46 molecules μm -2 , respectively.Of note, T-cells failed to respond to I-E k /K99A when displayed in the absence of the agonist ligand I-E k /MCC at $100 molecules μm -2 (Fig 7B ).
Taken together, pre-organization of two agonist pMHCIIs or one agonist pMHCII and one co-agonist pMHCII did not measurably alter the activation threshold of agonist pMHCIIs in 5c.c7 TCR transgenic T-cells (see Appendix Table S1C for significance table).We hence conclude that T-cell activation was independent of ligand spacing.
We next assessed whether bystander pMHCIIs in high abundance can trigger low-level calcium signaling in T-cells, which may ultimately affect overall TCR-proximal signaling in the presence of limiting agonist ligands.To test this, we recorded changes in intracellular calcium in 5c.c7 TCR transgenic T-cells in response to increasing densities of I-E k loaded with the agonist peptide MCC, the weak agonist T102S, the co-agonist K99A, the endogenous peptides ER60 and β2m, the antagonist K99R, as well as the null peptide (Fig 7D and E).
As shown in Appendix Fig S18B, the endogenous ligands I-E k / ER60, I-E k /β2m, and I-E k /null failed to trigger T-cell calcium signaling, even when displayed on the PLB at densities higher than 1,000 molecules μm -2 , as Fura-2 ratio values were comparable to those recorded in the absence of pMHCIIs.The MCC-based altered peptide ligands (APLs) K99A and K99R induced low-level oscillatory calcium signaling patterns when presented at supraphysiological densities above 1,000 molecules μm -2 (Appendix Fig S18B).
To assess whether TCR-proximal signaling in response to limiting numbers of agonist ligands was affected by bystander pMHCIIs copresented in high abundance, we exposed T-cells to PLBs functionalized with I-E k /MCC in low numbers ($1 molecule μm -2 ) together with ICAM-1, B7-1, and high numbers of I-E k loaded with T102S, K99A, K99R, ER60, β2m, or the non-binding null peptide.As expected, I-E k /T102S acted as weak agonist: when present in high abundance, it increased the number of fully activated T-cells as well as the overall calcium signaling capacity (Fura-2 ratio) and accelerated furthermore the onset of signaling.Quite in contrast, endogenous I-E k /ER60 and I-E k /β2m, as well as I-E k /K99A, I-E k /K99R, and I-E k /null, did not affect calcium response parameters even when present in more than 1,000-fold excess (Fig 7E and Appendix Fig S18C).Furthermore, T-cells displayed comparable, if not identical dose-response curves toward the agonist I-E k /MCC, regardless of the absence or presence of I-E k /K99A, I-E k /ER60, or I-E k /null (Fig 7F,and Appendix Fig S19A and B).
In summary, our findings challenge the very concept of pMHCII co-agonism and underscore the stimulatory potency of single, freely diffusing agonist pMHCIIs, as these activated T-cells in a peptidespecific manner and without contributions from endogenous bystander pMHCIIs.

Discussion
Advances in our understanding of T-cell antigen recognition and the extent to which it might be sensitized by pMHCII nanoscale  organization on professional APCs have been stalled for the last decade, in large part because the readout of most biochemical and imaging approaches employed did not support quantitative conclusions at the molecular level in living cells.We have therefore combined minimally invasive molecular live-cell microscopy with the use of custom-prepared imaging probes to arrive at a meaningful spatiotemporal resolution in the millisecond and nanometer regimes without adversely affecting cellular integrity.Following this strategy, we demonstrated that APCs express pMHCIIs at high numbers which become quickly distributed over the entire cell surface as single stimulatory entities as soon as they arrive at the plasma membrane.Highly abundant endogenous pMHCIIs did not enhance TCR binding and T-cell detection of rare agonist pMHCIIs at early or later stages of synapse formation.Endogenous pMHCIIs need to engage at least a fraction of TCRs in order to induce low-level signaling, which we observed only for MCC-derived APLs when these were presented to T-cells at exceptionally high densities.Together, this leaves us with the conclusion that single-agonist pMHCII entities elicit the full response of the scanning T-cell in a truly autonomous fashion.Following this concept, most efficient detection of rare agonist pMHCIIs would hence be best promoted by their rapid redistribution as soon as they arrive on the APC surface.
Central to the robustness of our analysis was the use of a monovalent and site-specifically fluorophore-conjugated (and optionally biotinylated) 14.4.4mAb-derived scF V , which targeted the pMHCII molecule I-E k in a highly specific and stable manner.This allowed for tagging surface-resident I-E k molecules with a single fluorescence emitter and supported quantitative analyses.We can exclude the possibility that we missed surface-resident I-E k in significant numbers as antigen-pulsed APCs preabsorbed with 14.4.4-scFV effectively blocked T-cell activation at 37°C due to their direct and effective competition with TCRs for pMHCII binding.
Underscoring the validity of employing monovalent and sitespecifically labeled scF V s for quantitation, we detected 40 times more I-E k molecules per APC (17 times more on B-cells and 40 times more on BMDCs) and also four-to fivefold higher molecular densities than previously reported with the use of phycoerythrin (PE)-conjugated full mAbs for cell staining (Unternaehrer et al, 2007).Given their large size, in particular when linked to PE, and their divalent nature, mAbs may be in part sterically hindered or bind instead of a single two pMHCIIs, which is plausible especially when facing exceedingly high epitope densities.Other possibilities may include non-quantitative PE conjugation or mAb linkage to photobleached PE, which together contrast the use of verifiably site-specifically dye-conjugated monomeric scF V s, as done in our study.
It was speculated that the high density of pMHCIIs present in MIIC-derived vesicles on their anterograde transport to the cell surface may be preserved at least transiently after vesicle fusion with the plasma membrane to support sensitized T-cell detection (Bosch et al, 2013).Diffusion measurements of pulse-chased I-E k as done in our study render such a scenario unlikely since 99% of newly arriving pMHCII molecules were mobile and dispersed over the entire APC surface within < 1 min.Single particle tracking revealed furthermore that 60-80% of pMHCIIs diffuse rapidly without any evidence of confinement.We examined whether the remaining fraction of more slowly diffusing pMHCIIs (20-40%) was associated with membrane entities such as sphingolipid rafts as had been previously reported (Anderson et al, 2000;Hiltbold et al, 2003;Knorr et al, 2009;Bosch et al, 2013).However, unlike Bosch et al, we failed to recover any surface I-E k in DRMs to an extent that was larger than 1%, when applying a highly quantitative LGC centrifugation-and flowcytometry-based method.This implies that such association is either non-existent on the cell surface or too weak to be conserved when using the non-ionic detergent NP40 employed to discriminate between DRM-resident proteins (e.g., CD44 and CD55) and solubilized membrane proteins (e.g., CD71).Since we failed to co-isolate surface I-E k with DRMs, as would be expected from raft-resident proteins, we consider transient I-E k association with tetraspanins (Zuidscherwoude et al, 2014) or reinternalization (Ma et al, 2012) a likely cause for the observed reduction in lateral mobility.◀ Figure 7. Rare antigenic pMHCIIs are recognized by T-cells independently of highly abundant bystander pMHCIIs.
A Strategies for the site-specific functionalization of DNA origami structures (70 × 100 nm) with biotinylated I-E k /MCC and I-E k /K99A molecules to investigate the effect of ligand spacing (D 10 = 10 nm, D 20 = 20 nm) on T-cell activation.B Dose-response curves for T-cell activation mediated by I-E k /MCC displayed in the absence or presence of the co-agonist I-E k /K99A and in the context of the mSav platform or different DNA origami constructs.Each data point corresponds to the percentage of activated T-cells determined in an individual experiment at a specific I-E k /MCC density.Data were normalized to activation levels recorded for positive control (=100%) involving the use of PLBs featuring I-E k /MCC at a density of 150 molecules μm -2 together with ICAM-1 and B7-1 at 100 molecules μm -2 .Data are representative of >2 biological replicates with the use of antigen-experienced T-cells isolated from 3-4 different 5c.c7 TCR-transgenic mice (Appendix Table S1B).C Dose-response curves were fitted with equation ( 12) to extract the activation threshold (T A , i.e., pMHCII density at half-maximum response, EC50).Statistics: mean and standard deviation (from a bootstrapping analysis).For each dose-response curve, data were derived from at least two independent experiments involving T-cells isolated from at least three different mice.Results of the fit and the significance test of the T A are summarized in Appendix Table S1B and C Of note, living BMDCs and B-cells decorated with monovalent, site-specifically labeled 14.4.4-scFV did not feature in our hands any form of I-E k clustering as was previously reported (Bosch et al, 2013), both above and below the diffraction limit.Decoration of living BMDCs and B-cells with randomly labeled I-E k and I-A kreactive mAbs showed homogenous staining, whereas co-staining with secondary IgG mAbs produced a punctate appearance (Appendix Fig S5A and B).In view of our diffusion and TOCCSL analysis, we consider it likely that the use of divalent mAbs, a temperature shift from 37 to 0°C, and PFA fixation had in fact caused the reported dotted appearance of I-E k .
Unlike what had been reported earlier (Turley et al, 2000;Bosch et al, 2013;Zuidscherwoude et al, 2015), neither hsPALM on living cells nor hsPALM and STED microscopy on fixed cells provided in our hands convincing evidence for clustered pMHCIIs on activated BMDCs and B-cells.Previous studies might have suffered from protein crosslinking as caused by antibodies applied at insufficient concentrations as well as by fixation artifacts impeding an unperturbed assessment of the pMHCII surface distribution.We circumvented these issues altogether by using monomeric and monovalent scF V to avoid receptor crosslinking, even when scF V s had been applied at non-saturating conditions, and to allow for non-invasive labeling and imaging strategies.Moreover, we implemented an unbiased cluster analysis protocol that included every single localization and improved the detection of true nanoscale enrichments from falsepositive clusters emerging as a result of continuously blinking photo-switchable fluorescent proteins (Baumgart et al, 2016;Platzer et al, 2020;Jensen et al, 2022).Contrary to our observations, a study involving STED microscopy reported exclusively pMHCII clusters on immature and activated APCs (Zuidscherwoude et al, 2015).It should be noted, however, that neither (i) the extent to which such reported pMHCII clusters differed in intensity from single molecule emitters nor (ii) the frequencies of their occurrence had been further compared to that of single pMHCIIs.By visually inspecting STED images of I-E k present on activated BMDCs and B-cells, we did not observe a clustered appearance of pMHCIIs.In contrast, especially in areas with low surface density on the APC cell membrane, we detected most I-E k molecules (i.e., 14.4.4scF V -abSTAR635P) as single events which gave rise to the same mean intensity and FWHM values as single fluorescence emitters.
A dimeric arrangement of pMHCIIs was suggested to be a potent mediator of sensitized detection of antigen by T-cells (Krogsgaard et al, 2005), yet without direct confirmation of the existence or the formation of higher order pMHCII structures within the plasma membrane of living APCs.In fact, despite using all available means we were unable to identify such entities on the surface of activated B-cells and BMDCs.This pertained to their detection via TOCCSL, for which they had to be stable for no longer than 2-20 s, and via inter-pMHCII FRET-based measurements, which as a whole did not provide any evidence for their existence.The latter experiments were performed to account in particular for pMHCIIs diffusing slowly on the cell surface.FRET-based measurements relied on FRET dye combination featuring a Fo ¨rster radius of 5.1 nm (AF555 and AF647) and site-specifically labeled probes present at ratios of 1:1 to 1:4 for quantitative readout.An earlier study employing FITC and TRITC-conjugated 14.4.4monoclonal antibodies (Fo ¨rster radius: 5.5 nm) to reveal homo-association of pMHCII on B lymphoma cells arrived at an average FRET efficiency of 6.5% without further description of the donor to acceptor ratios applied and the numbers of fluorophores attached to the 14.4.4mAb (Gombos et al, 2004).In contrast, our quantitative analysis, which included the use of sitespecifically conjugated scF V and FRET simulations involving monomeric and randomly distributed FRET pairs, failed to provide any evidence for pMHCII homodimerization.Instead, we showed with the use of PLB-anchored pMHCIIs that random encounters of freely diffusing monomeric entities can give rise to FRET values of such magnitude.
With this information, we set out to recapitulate antigen recognition by T-cells in contact with PLBs featuring pMHCIIs as well as ICAM-1 and B7-1 at physiological densities.In preceding T-cell imaging studies involving fluorescence and calcium measurements, it had been reported that co-agonist pMHCIIs contributed in part to the response to agonist pMHCIIs (Wulfing et al, 2002;Krogsgaard et al, 2005;Anikeeva et al, 2012).Yet other studies featuring similar techniques questioned the role of co-agonists (Ma et al, 2008) or concluded that CD8 co-receptor affinities dictated the coagonistic effect (Gascoigne, 2008;Hoerter et al, 2013).In this study, we found TCR engagement of all endogenous pMHCIIs with previously attributed co-agonist qualities at the level of random molecular collisions.Only for the altered peptide ligand I-E k /K99A (and I-E k /K99R, a proposed antagonist), we detected specific TCR binding, which amounted to fewer than 500 interactions per synapse at any given time when these pMHCIIs had been present at a density of 1,390 molecules μm -2 .Assuming that the T-cell synapse covers an area of 100 μm 2 with a TCR density of 100 TCRs μm -2 , this number indicates that on average fewer than 0.4% of available I-E k /K99A and only 1% of all synaptic TCRs had been occupied.Not surprisingly, highly abundant I-E k /K99A (outnumbering TCRs in our example 14:1) did not alter TCR binding of rare agonist pMHCIIs in early and mature synapses.In fact, the ensuing TCRproximal signaling response was unaffected, even when I-E k /K99A was co-presented on DNA origami-based platforms in a spatially highly defined fashion with the agonist ligand I-E k /MCC.Interestingly, our observations contrast previous findings we had based on the use of DNA origami-tethered anti-TCRβ-scF V s as a means of Tcell stimulation (Hellmeier et al, 2021a).Here, we identified distances separating individual stimulatory scF V s that were smaller than 20 nm as permissive for T-cell activation, a notion that did not hold up for the recognition of nominal pMHCII antigens.A possible explanation for such divergent behavior may include liganddependent differences in stimulatory potency as well as the highly transient nature of TCR-pMHC engagement, which enables single antigens to trigger more than one TCR/CD3 complex in rather close proximity.
It should also be noted that instead of promoting or functionally enhancing rare agonist-TCR interactions, highly abundant bystander pMHCIIs rather competed with agonist pMHCIIs for available TCRs with increasing TCR affinities.We also observed that unlike endogenous I-E k /ER60 and I-E k /β2m, the MCC-based APLs I-E k /K99A and I-E k /K99R promoted low-level signaling when present at supraphysiological pMHCII densities (> 1,000 molecules μm -2 ), yet without sensitizing T-cells for agonist pMHCIIs.
In summary, our results leave us with no plausible conclusion other than that single agonist pMHCII entities elicit the full 20 of 39 this realization places a particular emphasis on a deep subcellular and molecular understanding of effective, faithful, and fail-safe TCR-proximal signal amplification, which appears to rely solely on the distinct nature of synaptic TCR-pMHCII interactions and CD4 co-receptor engagement within the context of plasma membrane biophysics and physiology.Our findings imply furthermore that rapid redistribution of newly arriving pMHCIIs rather than clustering underlies highly efficient detection by scanning CD4 + T-cells.It will be highly informative to reveal communalities and differences concerning the organization of pMHCI on the surface of target cells with possible consequences for CD8 + T-cell recognition.

Materials and Methods
Reagents and Tools
After expansion, T-cells were subjected to Histopaque-1119 (Merck) density gradient centrifugation to remove dead cells.Antigen-experienced T-cells were used for experiments from day 8 to 10. B-cells were isolated from lymph nodes of 5c.c7 αβ TCRtransgenic mice and enriched using a murine B-cell isolation kit in order to deplete activated B-cells and non-B-cells (B-cell Isolation Kit mouse, Miltenyi Biotec).B-cells were cultured at 37°C and 5% CO 2 in 1640 RPMI (Gibco) supplemented with 10% FBS (Biowest), 100 U ml À1 penicillin, 100 μg ml À1 streptomycin (Gibco), 2 mM Lglutamine (Gibco) and 50 μM β-mercaptoethanol (Gibco).B-cells were either used for experiments directly after isolation and enrichment as naive (non-activated) B-cells or activated with 50 μg ml À1 LPS derived from E. coli serotype O55:B5 (Merck) for 24-48 h to induce B-cell blast formation.
All cell lines were tested for mycoplasma contamination on a biweekly basis using Hoechst staining of supernatant-treated Vero cells as an indicator.

Protein expression, refolding, and preparation of imaging tools
In order to genetically assemble the 14.4.4scF V , we obtained the sequence information of the 14.4.4mAb VH and VL chain by isolating total mRNA from a 14.4.4mAb-producing hybridoma cell line with the use of TRIzol (Thermo Fisher Scientific) followed by reversely transcribing the isolated mRNA into cDNA (First-strand cDNA Kit, Roche).
We genetically combined the C-terminus of the VH domain and the N-terminus of the VL domain with a [GGGGS] 3 linker to obtain a functional scF V transcript, which was cloned into the pet21a(+) expression vector using the restriction enzymes NdeI and HindIII (Appendix Fig S20A and B).To create a target for site-specific modifications using maleimide-based click chemistry, we substituted in the VH domain the serine at position 120 to an unpaired cysteine (14.4.4 scF V ) via site-directed mutagenesis (QuikChange II, Agilent).In addition, we extended the 14.4.4scF V C-terminally with the BirA biotin ligase (Avidity) recognition site (GLNDIFEAQKIEWHE) to allow for a site-specific attachment of biotin and fluorophores (e.g., 14.4.4scF V -AF647-biotin).
The I-E k -specific 14.4.4mAb was purified from hybridoma supernatant using a protein A/G agarose column (Thermo Fisher Scientific) and gel filtration (Superdex-200 10/300 GL, Cytiva).The 14.4.4mAb and the I-A k -reactive 11-5.2 mAb were randomly labeled via available lysine residues with Alexa Fluor 647 carboxylic acid succinimidyl ester (Thermo Fisher Scientific) according to the manufacturer's instructions and purified via gel filtration (Superdex-200 10/300 GL, Cytiva).Protein-to-dye ratio of the 14.4.4mAb and 11-5.2 mAb were 2.1 and 2.4, respectively.
mSav was prepared with some modifications as described (Howarth et al, 2006;Fairhead et al, 2014).The pET21a(+) expression vectors encoding either for the "dead" (i.e., biotin non-binding) or the "alive" (i.e., biotin binding) streptavidin subunit were a kind gift from Alice Ting (Stanford University, USA).We substituted the hexa-histidine tag on the "alive" subunit with a cleavable hexaglutamate tag to allow for purification via anion exchange chromatography (Mono Q 5/50 GL, Cytiva) followed by a recognition site of the 3C protease for optimal removal of the tag (Platzer et al, 2020).To produce mSav for conjugation with maleimidelinked fluorescent dyes or maleimide-tetrazine (or ÀTCO) for orthogonal click chemistry-based protein conjugation, we mutated the alanine residue at position 106 (A106C) for a cysteine residue in the "alive" subunit.The mSav platform mSav-STAR635 was prepared as described (Platzer et al, 2020).
The placeholder peptide can be substituted with any peptide that binds into the peptide-binding cleft of I-E k under acidic conditions (pH adjusted to 5.1 with 0.2 M citric acid buffer) over a period of 3 days (Xie et al, 2012).
The same loading strategy was used to produce I-E k /MCC [NVOC], which contains a UV-cleavable 6-nitroveratryloxycarbonyl (NVOC) attached to the lysine residue in the central position of the MCC peptide (ANERADLIAYLK[NVOC]QATKGGSC) facing the TCR (DeMond et al, 2006).
All peptides were ordered as lyophilized products, purified via reversed-phase chromatography (Pursuit XRs C18 column, Agilent), and verified via MALDI-TOF (Bruker) mass spectrometry.For sitespecific labeling, peptides were dissolved at a concentration of 1 to 5 mg ml À1 in 1× PBS and incubated with TCEP agarose CL4-B (Merck) for 1 h to reduce paired cysteines.The peptide sample was filtered through a 0.22 μm filter to remove TCEP agarose and incubated with a 1.2-fold excess of maleimide-conjugated dye (Alexa Fluor 647 C2 Maleimide, Alexa Fluor 488 C5 Maleimide, or Alexa Fluor 555 C2 Maleimide) for 2 h in 1× PBS supplemented with 10 mM HEPES, pH 6.8.
In addition, we produced an MCC peptide derivative (ANERA-DLIAYLKQATKdSdSdC, d refers to D enantiomer; underlined amino acids are recognized by the T-cell) conjugated with EZ-Link Maleimide-PEG 2 -Biotin (Thermo Fisher Scientific) according to the manufacturer's instructions.
For intracellular cell-staining, we first labeled cell surface proteins (e.g., TCR) in FACS buffer for 30 min before washing with 1× PBS and fixation with BD Cytofix/Cytoperm (BD Biosciences) for 15 min.Samples were then washed again with BD Perm/Wash (BD Biosciences) and subsequently stained for intracellular proteins (e.g., mAbs binding to TNF-α or IFN-γ, both 1:200) in the presence of BD Perm/Wash for 30 min.After another washing step with BD Perm/Wash and FACS buffer, samples were directly subjected to flow cytometric analysis.
Samples were analyzed on the LSR II flow cytometer (BD Biosciences) or the Cytek Aurora (Cytek Biosciences).Data derived from flow cytometry measurements were analyzed with the FlowJo v10 software (BD Biosciences).
Fluorescence intensities of intact cells and DRM-coated LGCderived nuclei were analyzed with an LSR II flow cytometer (BD Biosciences).Appropriate single-color staining of cells and OneComp eBeads (Thermo Fisher Scientific) were performed to compensate for spectral overlap.Unstained cells and unstained isolated nuclei were used to evaluate background fluorescence.The FlowJo v10 software (BD Biosciences) was employed to determine mean fluorescence intensities (MFIs) from cellular and nuclear single-event populations per sample.DRM association for mAbstained cell surface molecules was calculated in percent as [(MFI nuclei À MFI nuclear background )/(MFI intact cells À MFI cellular background )] × 100.

BMDC and T-cell co-culture assays
Activated BMDCs were pulsed with 0.008 to 1.0 μM MCC (88-103) peptide in the presence of 100 ng ml À1 LPS for 1 h followed by an epitope blocking step with 0 or 20 μg ml À1 14.4.4scF V on ice for 30 min.BMDCs were then washed once in T-cell medium (see above) and transferred to a 96-well plate together with 5c.c7 TCR transgenic T-cells at a 1:1 ratio (60,000 cells in total) in the presence or absence of 20 μg ml À1 14.4.4scF V (total volume 100 μl).The coculture was maintained in T-cell medium supplemented with BD GolgiStop (1 μl ml À1 ) and BD GolgiPlug (0.7 μl ml À1 ) at 37°C in an atmosphere of 5% CO 2 for 4 h prior to intracellular staining and flow cytometric analysis.
Preparation of DNA origami DNA origami structures were assembled in a single folding reaction followed by purification and step-wise functionalization as previously described (Hellmeier et al, 2021a).At sites chosen for ligand attachment, staple strands were elongated at their 3 0 -end with 21 bases.At sites chosen for cholesterol anchor attachment, staple strands were elongated at their 3 0 -end with 25 bases.DNA origami was stored for up to 2 weeks at À20°C.DNA origami decorated with a single pMHCII species was functionalized as described (Hellmeier et al, 2021a).Hetero-dimeric configurations on DNA origami were created by pre-incubating biotinylated pMHCIIs with a 3× molar excess of DNA-conjugated mSav (mSav X-strand with I-E k /K99A-AF647-biotin, and mSav V-strand with I-E k /MCC-AF555-biotin) for 30 min at 24°C.Unbound proteins were removed by spin column purification (Amicon Ultra-2 centrifugal filters, 100 kDa cut-off (Merck)).As a last step, DNA-conjugated mSav-bound biotinylated pMHCIIs (I-E k /K99A-AF647, I-E k /MCC-AF555) were hybridized to the elongated staple strands of the DNA origami at a 5× molar excess for 60 min at 24°C, followed by spin column purification (100 kDa cut-off).Functionalized DNA origami platforms were seeded onto mobile PLBs decorated with B7-1 and ICAM-1 at 100 μm À2 and were used for experiments on the same day.Functionalization efficiency of DNA origami was determined via two-color colocalization microscopy as previously described (Hellmeier et al, 2021b).

Microscopy setup
For molecular live-cell imaging, we operated setup #1 and setup #2 at the Medical University of Vienna and setup #3 at the Technical University of Vienna.Setup #1 was a Leica DMI 4000B inverted microscope, which was used to image samples in epifluorescence or total internal reflection fluorescence mode (TIRF) with a 100× objective (HCX PL APO 100×, NA = 1.47,Leica) and a 20× objective (HCX PL Fluotar 20×, NA = 0.5, Leica).As setup #2, we operated an Eclipse Ti-E (Nikon) inverted microscope system that was equipped with a 100× objective (Nikon SR APO TIRF 100×, NA = 1.49) and a 20× objective (Nikon S Fluor, NA = 0.75).
Emission light was recorded with the EMCCD camera (Andor iXon Ultra 897) or the CMOS prime95B (Teledyne Photometrics).To program and apply timing protocols and regulate hardware components, both setups were either operated with an 8-channel DAQ-card (National Instruments) or the USB-3114 analog and digital output system (Measurement Computing) controlled by the microscopy automation and image analysis software Metamorph (Molecular Devices).Setup #1 was additionally operated with the Leica software (Leica Suite) and μManager (Open Imaging).All setups were enclosed with a heating chamber (Leica (setup #1), Pecon (setup #2)) and equipped with a heating unit to adjust the temperature within the box.
Setup #3 was equipped with a monochromatic light source (Polychrome V, TILL Photonics) coupled to a Zeiss Axiovert 200 M, which was equipped with a 10× objective (UplanFL N 10×, NA = 0.3, Olympus), a 1.6 × tube lens, and the EMCCD camera Andor iXon Ultra 897.In the emission path, we installed a long-pass filter (T400lp, Chroma) and an emission filter (510/80 ET, Chroma).Precise temperature control was enabled by an in-house-built incubator equipped with a heating unit.
Mid-sections of BMDCs or B-cells were recorded as z-stacks (50 images with 0.3 μm spacing, total depth of z-stack: 15 μm) and deconvolved with the ImageJ plugin iterative deconvolve 3D, which is based on the DAMAS (Deconvolution Approach for the Mapping of Acoustic Sources) algorithm (Dougherty, 2005).The point spread function in x, y, and z used to deconvolve the images was measured with single fluorescent beads (TetraSpeck Fluorescent Microspheres).

Quantification of protein densities within cell membranes and PLBs
Rectangular regions of a cell or planar-supported lipid bilayer were specified for further calculations.Next, the camera background was subtracted from the intensity of the selected images and divided by the single-molecule intensity of the corresponding fluorophore.The intensity of single fluorescence emitters was obtained from cells or planar-supported lipid bilayers decorated with single molecules (e.g., 14.4.4scF V -AF647 or I-E k /MCC-AF647) and imaged at a power density of 0.1-0.4kW cm À2 and 10-20 ms illumination times.To avoid saturation of the camera when imaging highly fluorescent PLBs, we reduced either the power density by a factor of 10 using neutral density filters in the beam path (OD 1.0, Thorlabs) or reduced the linear EM gain of the EMCCD camera by a predetermined factor resulting in a tenfold reduction in pixel counts.Protein densities were specified as number of molecules μm -2 .
High-speed photoactivation localization microscopy (hsPALM) 0.1-0.5 × 10 6 BMDCs or B-cells were labeled with 20 μg ml À1 14.4.4scF V -AF647-biotin for 30 min on ice, washed twice with 1× HBSS (Gibco) supplemented with 1% FBS (Biowest), 2 mM CaCl 2 , and 2 mM MgCl 2 (Merck), and afterward incubated with excess amounts of mSav-cc-PS-CFP2 for 30 min on ice.Next, cells were washed twice and either kept on ice or allowed to spread on ICAM-1-coated glass slides for 1 to 5 min and were then directly subjected to live PALM imaging.Optionally, cell samples were fixed with 4% PFA (Thermo Fisher Scientific) in 1× PBS for 15 min at room temperature, washed, stored at 4°C (to complete fixation), and imaged either immediately after or the next day.PALM images were recorded in TIRF mode at 25°C.We recorded hsPALM image stacks only from adherent cells, a condition that was determined by one red image (637 nm laser excitation (OBIS-637-140), 705/72 emission filter) of the AF647-conjugated 14.4.4scF V -biotin seconds before streaming the hsPALM image stack.With the intention to excite and bleach PS-CFP2 in one single frame (with 2 ms illumination time and 0.3 ms readout time), we used the 488 nm laser line (OBIS-488-150-LS) with a power density of 3.5 kW cm À2 .After a short prebleach stream with 488 nm laser excitation (50 frames), we used continuous 405 nm laser excitation (OBIS-405-100) operated at a very low power density (1-10 W cm À2 ) to photo-convert PS-CFP2 from cyan to green emission and filtered the emission channel with the bandpass filters ET525/36 (Leica) or ET525/50 (Chroma).With the use of the Andor iXon Ultra 897 EMCCD camera, we collected at a frame rate of 435 Hz 3,050 image frames (50 × 200 pixel) sufficient for one reconstructed superresolution image in 7 s.
We obtained the blinking statistics of PS-CFP2 under the same experimental conditions (PFA fixed or unfixed) and with the same illumination settings and 488 nm power densities following a procedure we had previously published (Platzer et al, 2020).More specifically, we decorated immobile PLBs with spatially wellseparated mSav-abSTAR635P in complex with PS-CFP2 molecules which had been site-specifically biotinylated via an AVI-tag.Imaging of hundreds of different PLB positions was conducted at room temperature and in TIRF mode as described (Platzer et al, 2020).To visualize the position of single mSav-STAR635-PS-CFP2 molecules, we used the 637 nm laser line (OBIS-637-140) with a power density of 0.5-1.0kW cm À2 and the emission bandpass filter ET700/75 (Chroma).We recorded 20 images to localize (and photo bleach) the mSav-STAR635 platform prior to streaming 3,050 frames to create the hsPALM stack for the PS-CFP2 blinking analysis.
Single-molecule blinking analysis of PS-CFP2 conjugated to the mSav-STAR635 platform was carried out as described (Platzer et al, 2020) with the following modifications: The position of mSav-STAR635 molecules was averaged by using the localization merging algorithm implemented in ThunderSTORM (Ovesny et al, 2014) with the following parameters: maximum t off = 20 frames and maximum displacement = 100 nm.A localization cluster threshold was set to 30 to filter out extreme outliers.Data were pooled from two independent experiments.PALM image analysis via Ripley's K was performed as described in Platzer et al (2020).We first selected for each cell a rectangular region of interest (ROI) with a maximum size specified by the boundaries of the cell (e.g., 3,000 × 3,000 nm).Single-molecule PS-CFP2 signals within the ROI were fitted with a Gaussian intensity distribution by maximum-likelihood estimation using the ImageJ plug-in ThunderSTORM and filtered for sigma (< 320 nm) and positional accuracy (< 150 nm).We obtained a positional accuracy of median of 51 nm for single PS-CFP2 signals with the illumination settings described above (Appendix Fig S6F).We then performed a qualitative Ripley's K analysis of a region of interest and compared the calculated maxima of the Ripley's K function of the cell-derived localization map with a simulated localization map, which represented a random distribution of molecules present at the same molecular density and exhibited features such as blinking behavior of PS-CFP2 and in selected cases diffusion parameters of respective molecules (Appendix Fig S8A and B).Of note, Ripley's K analysis of a ground truth localization map (after MBC) was compared to a simulated localization map, which represented only a random distribution of molecules present at the same molecular density (without blinking or diffusion parameters).Localization maps with an L(r)-r maxima mainly above the standard deviation of the associated simulated random distribution (within r = 250 nm) were registered to contain (partly or exclusively) clusters (Platzer et al, 2020).
PALM image analysis via label density variation was carried out as described by Baumgart et al (2016).After obtaining the X and Y coordinates for every localization as described above, we assigned a Gauss function with a sigma (σ) of 35 nm (that was cut at 2σ) to every localization within the sample regions.Briefly, overlapping Gauss functions were summed subpixel-wise after applying a threshold (from 0.5 to 4.5; here 1.0) to generate a three-dimensional cluster map (Fig 2E).Next, we calculated the density within the selected cluster regions (ρ; localizations μm -2 ) and the relative area of selected cluster regions (η; cluster area divided by total area).The pair-wise numbers obtained for each sample were plotted in a ρ/η diagram and compared to a reference curve for randomly distributed localizations (equation 1).The reference curve was determined by simulating randomly distributed molecules under different conditions (variations in blinking statistics and thresholds) and applying a polynomial function that fits every scenario, as described by Baumgart et al (2016).
Model-based correction (MBC) and DBSCAN analysis of PALM data We corrected for multiple blinking artifacts in our PALM data with the MBC workflow introduced by Jensen et al (2022).To this end, we selected 3,000 × 3,000 nm regions (Appendix Fig S8A and B) within the recorded cells and saved the respective localization maps as .csvfiles containing the information for x (nm), y (nm), frame, and positional accuracy (nm).MBC was performed on the selected ROIs with default parameters (as suggested) and a framerate of 435 fps (2.3 ms time lag) using the provided R script.After MBC, we subjected the corrected localization maps, now containing ground truth fluorophore localizations, to Ripley's K analysis (to verify clusters over spatial randomness as described above), and DBSCAN analysis to characterize potential clusters (Ester et al, 1996), both implemented in Matlab.DBSCAN analysis is based on the density distribution within a dataset and requires two parameters, epsilon (threshold for a neighborhood search radius) and minpts (minimum number of neighbors), to identify core points and assign them to clusters.All unconnected points are considered noise points.To facilitate the identification of clusters over noise in our datasets, we found minpts = 5 to be suitable for all investigated cell regions.The adequate choice of epsilon is at a threshold just large enough to minimize the number of noise points incorrectly assigned to clusters.We estimated epsilon for every cell region with a k-distance graph (Ester et al, 1996) and decreased this value (depending on the density of points) in steps of 10 until only noise points (i.e., percentage of molecules in clusters $1%) were detected in the associated simulated random distribution as recommended by Endesfelder et al (2013).With this approach, we were able to determine the percentage of molecules within clusters, the number of molecules per cluster, and the number of detected clusters of I-E k on activated BMDCs and B-cells and CD205 on activated BMDCs.

Single-molecule tracking
BMDCs and B-cells were stained with a single molecule dilution of the 14.4.4scF V -AF555 on ice for 30 min and washed twice with imaging buffer containing 1× HBSS (Gibco) supplemented with 1% FBS (Biowest), 2 mM CaCl 2 and 2 mM MgCl 2 (Merck).Cells were kept on ice or seeded onto ICAM-1-coated glass slides for imaging at 37°C and in TIRF mode.AF555 was excited with the use of a 532 nm laser (OBIS-532-100-LS), and the emission channel was cleaned up with the bandpass filter ET605/52.We recorded movies of single I-E k (AF555-14.4.4 scF V ) localizations over 500 frames with a power density of 0.25-0.5 kW cm À2 , an illumination time of 10 ms, and a total time lag of 10.26-10.5 ms between two adjacent images (frame transfer mode).
For I-E k recovery experiments, BMDCs and B-cells were pulsed for 12 and 6 h, respectively, with 50 μM MCC-PEG 2 -biotin.We divided each cell sample into two aliquots.The first aliquot represented the total pool of I-E k /MCC-PEG 2 -biotin and was kept on ice.The second one was blocked with unlabeled monovalent A minimum of 180 brightness values was applied to calculate ρ 1 (B) and ρ N (B) from all TOCCSL images of multiple cells of one experimental condition.A least-square fit with equation ( 6) was employed to determine the weights of the individual pdfs, α N , with ∑ Nmax N¼1 α N ¼ 1.
Stained T-cells were allowed to settle for 2 min on functionalized PLBs, which were equipped with unlabeled ICAM-1, B7-1 (100 molecules μm -2 ), and I-E k /peptide-AF647 (FRET acceptor) at different densities.Synaptic FRET was recorded as a consequence of TCR-pMHCII binding for up to 10 min (i.e., 2-12 min after cells had made first bilayer contact).
For all FRET experiments, the emission beam path was separated into two spectral channels employing an Optosplit 2 device (Cairn Research) which was equipped with a beam splitter (ZT640rdc, Chroma) and adequate emission bandpass filters (ET575/50 or ET640SP, ET655LP, Chroma).The following alternating excitation protocol was used to record FRET donor and FRET acceptor channels: a low-intensity AF647 acquisition (to monitor the AF647 signal, OBIS-637-140, power < 0.01 kW cm À2 ) was succeeded by a low-intensity AF555/FRET acquisition (to record the AF555 and FRET signal before FRET acceptor photobleaching, OBIS-532-100-LS, power $0.01 kW cm À2 ), followed by a high-intensity 640-nm bleach pulse (to ablate AF647, power > 2 kW cm À2 ), a second low-intensity AF647 acquisition (to verify AF647 ablation, power $0.01 kW cm À2 ), and finally, a second AF555/FRET acquisition (to record the AF555 signal after FRET acceptor bleaching and verify FRET ablation, power $ 0.01 kW cm À2 ).Time lapsing between the two AF555/FRET acquisitions was less than 500 ms.
Quantitative imaging was ensured by minimizing photobleaching at a given illumination time of 30 ms.To this end, the power levels of the 532 nm FRET donor excitation laser were adjusted to keep photobleaching between consecutive images below 1%.I-E k /peptide-AF488 was excited with the 488 nm laser (OBIS-488-150-LS) at a power density of 0.1-0.2kW cm À2 and the emission was cleaned up employing the ET525/50 (Chroma) filter.
For FRET-sensitized emission analysis, donor and acceptor channels were superimposed using an affine transformation matrix (including shift, stretch, rotation, and tilt) and intensity registration, both generated from a fluorescent bead reference (TetraSpeck Fluorescent Microspheres Size Kit, Thermo Fisher Scientific).Net fluorescence intensities were obtained by pixel-wise subtraction of the camera background from the raw-intensity images.To compensate for laser illumination profiles, we performed a pixel-wise flatfield correction by means of median averaging 10 separately recorded PLB samples representing a similar fluorophore density as the sample employed for FRET measurements.
FRET donor recovery after FRET acceptor photobleaching analysis was performed as described (Huppa et al, 2010).After background subtraction, the AF555 intensity (FRET donor) before FRET acceptor photobleaching (DDpre) was subtracted from the AF555 intensity after FRET acceptor photobleaching (DDpost).The difference was then divided by DDpost to obtain the FRET yield according to equation ( 7).
The TCR occupancy, that is, the ratio of ligand-engaged and total TCRs was calculated as shown in equation ( 9 The experimentally determined conversion factor C = 1.21 was comparable with the estimated values of C = 1.29 (for R filter = 1.07) and C = 1.28 (obtained from the maximal FRET efficiency of the FRET system max E FRET = 1/0.78= 1.28).

Single-molecule FRET imaging
Single-molecule FRET recordings were conducted with the same experimental setup, hardware settings, and filters as used for recording ensemble FRET.To attenuate bleed through from the FRET donor fluorophore into the FRET acceptor channel, we opted to stain approximately 10% of the TCRs on T-cells with the H57 scF V -AF555 by premixing one part of H57 scF V -AF555 with nine parts of unlabeled H57 scF V prior to cell staining (total 10 μg ml À1 ).We also adjusted I-E k /MCC-AF647 densities on PLBs to 5-30 I-E k μm À2 , to reduce cross-excitation of the FRET acceptor fluorophore.
To visualize earliest single TCR-pMHCII-binding events, we employed, in addition, the caged MCC peptide derivative ANERADLIAYLK[NVOC]QATKGGSC-AF647 loaded onto I-E k (I-E k / MCC[NVOC]-AF647) which represents a null ligand unless it becomes uncaged by means of a short 405 nm laser pulse (0.17 kW cm À2 , 100 ms).1-2 seconds after the 405 nm pulse, individual TCR-pMHCII-binding events appeared as single-molecule FRET events, disappeared in single steps, and aligned with single FRET acceptor fluorophores (I-E k /MCC-AF647).We recorded smFRET events with an illumination time of 20 ms and applied a 532 nm power density of 0.1 kW cm À2 (OBIS-532-100-LS or LCX-532-500) to excite the donor fluorophore AF555 and a 640 nm power density of 0.05 kW cm À2 (OBIS-637-140 or LCX-640-500) to excite the acceptor fluorophore AF647.I-E k /K99A-AF488 was excited with the 488 nm laser (LBX-488-200) at a power density of 0.1-0.2kW cm À2 and the emission was filtered employing an ET525/50 (Chroma) filter.
Intensity and positional accuracy of single-molecule donor and single-molecule FRET events as a consequence of TCR-pMHCII binding were calculated with the ImageJ plugin ThunderSTORM (Ovesny et al, 2014) and filtered for sigma (< 320 nm) and positional accuracy (< 150 nm).
To measure the synaptic off-rate of 5c.c7 TCR:I-E k /MCC interactions at 26°C in the presence or absence of I-E k /K99A, we adjusted the I-E k /MCC-AF647 density on PLBs to 11 I-E k /MCC-AF647 μm À2 and ICAM-1 and B7-1 to 100 μm À2 .I-E k /K99A-AF488 was either completely absent or present at a density of 1,060 I-E k /K99A μm À2 .For off-rate determination, we recorded the appearance of smFRET traces at different time lags (56, 84, 252, 476, 924, and 1,820 ms) to correct for photobleaching.We then listed the smFRET traces according to their trajectory length and created a normalized inverse cumulative decay function for each time lag as described in Axmann et al (2015).The resulting decay function was fitted with a single exponential decay function (start 1.0, plateau 0) to calculate (app t off = 1/app k off ) the apparent lifetime (Appendix Fig S17E).Lastly, we plotted the calculated apparent lifetimes against the respective time lags and fitted the lifetime (t off ) and the bleach constant c bleach based on equation (11).app t off ¼ t off = t off c bleach þ t lag (11)

Calcium imaging
Intracellular calcium levels were measured with the ratio-metric dye Fura-2 (Thermo Fischer Scientific) as published (Roe et al, 1990).1-2 × 10 6 antigen-experienced 5c.c7 TCR-transgenic T-cells were incubated with 2 μM Fura-2 in T-cell growth medium for 15-20 min at room temperature and subsequently washed with warm (room temperature) imaging buffer (14 ml of 1× HBSS (Gibco) supplemented with 1% FBS (Biowest), and 2 mM CaCl 2 and 2 mM MgCl 2 (Merck)).Immediately afterward, T-cells were seeded onto functionalized PLBs featuring increasing densities of I-E k in complex with agonist or non-agonist (bystander) peptides and unlabeled B7-1 and ICAM-1 (100 μm À2 ).The usage of 16-well Nunc LabTek chambers (Thermo Fisher Scientific) enabled us to record up to 12 different conditions (i.e., various or different concentrations of I-E k ) in a single experimental run.Calcium experiments were carried out on setup #1 or setup #3 at 37°C.On setup #1, Fura-2 was excited using a monochromatic light source (Leica) coupled into an inverted Leica DMI4000B microscope that was equipped with a 20× objective (HCX PL Fluotar 20×, NA = 0.5, Leica), the dichroic beamsplitter FU2 (Leica), and the bandpass filter ET525/36 (Leica).Calcium signaling was measured ratiometrically by continuous switching between 340 and 387 nm excitation, which was achieved with the use of a fast excitation filter wheel (Leica) featuring the excitation bandpass filters 340/26 and 387/11 (Leica).The total recording time was 10-20 min with an image recorded every 15-20 s.An automated XY stage (Leica) allowed fast changes between different positions (i.e., LabTek chamber wells).Precise temperature control was carried out with a heating unit and a box enclosing the microscope (Leica).
For calcium imaging of T-cells approaching DNA origami structures, we used setup #3.Fura-2 was excited with a monochromatic light source (Polychrome V, TILL Photonics) coupled to a Zeiss Axiovert 200 M which was equipped with a 10× objective (UplanFL N 10×, NA = 0.3, Olympus), a 1.6× tube lens, a long-pass filter (T400lp, Chroma), the emission filter ET510/80 (Chroma), and the EMCCD camera Andor iXon Ultra 897.Imaging was performed with excitation at 340 and 380 nm with illumination times of 50 and 10 ms, respectively.The total recording time was 10 min with one image per second.Experiments were carried out at 37°C using an in-house-built incubator equipped with a heating unit for precise temperature control.
Calcium image analysis was carried out either with a custom code written in Matlab (based on Salles et al, 2013) or as described in Hellmeier et al (2021b).In detail, we first generated the ratio image of the 340 nm / 380 nm image for every frame.Individual Tcells were then tracked using an in-house Matlab algorithm based on Gao & Kilfoil (2009) and the positional information of each cell was used to extract the intensity value from the ratio image stack.Fura-2 intensity traces were either normalized to the starting value at timepoint 0 or the median of all Fura-2 intensity traces of a negative control (ICAM-1,100 μm À2 , B7-1,100 μm À2 ).Traces were categorized into "activating," oscillatory, and "non-activating" based on an activation threshold defined by a receiver operating curve between a negative control and a positive control (I-E k / MCC > 10 μm À2 , ICAM-1,100 μm À2 , and B7-1,100 μm À2 ).T-cells were categorized as "activating" when the Fura-2 intensity value remained 80% of the entire trace above the threshold, "nonactivating" when it remained 80% of the trace below the threshold, otherwise the cell was classified as oscillatory.The generation of dose-response curves and the bootstrap analysis of EC50 data is best described in Hellmeier et al (2021b).The calcium imaging data were fitted to a three-(hillslope n = 1.0) or four-parameter doseresponse curve with equation ( 12) to extract the activation threshold (T A ) at half-maximum response.Statistical significance between obtained values T A,1 and T A,2 for two different data sets was determined via a bootstrap ratio test (Appendix Table S1C) as described in MacKinnon ( 2009) and Wheeler et al (2006).

STED microscopy
BMDCs and B-cells (0.25 × 10 6 to 0.5 × 10 6 ) were stained with 20 μg ml À1 14.4.4scF V -abSTAR635P for 30 min on ice and washed with imaging buffer (1× HBSS (Gibco) supplemented with 1% FBS (Biowest), 2 mM CaCl 2 and 2 mM MgCl 2 (Merck)).Cells were allowed to adhere on ICAM-1-coated glass slides for 10 min at 25°C before fixation.Cells were stained with limited amounts of 14.4.4scF V abSTAR635P to visualize single fluorescent emitters.Samples were fixed for 15 min at room temperature applying 4% PFA (Thermo Fisher Scientific), washed with 1× PBS and imaging buffer, mounted in Abberior Mount Solid (Abberior), and stored at 4°C until imaging.STED imaging was performed on a Zeiss Axio Imager A2 upright microscope equipped with a 100× objective (alpha-Plan-Apochromat 100×, NA = 1.46,Oil DIC M27, Zeiss) and the Stedycon super-resolution microscopy unit (Abberior Instruments).Scanning was performed with the built-in QUAD beam scanner with a pixel dwell time of 250 ms at a pixel size of 25 nm.A pulsed 640 nm diode laser with an output of 1.2 mW at 40 MHz (max.5 mW) and > 20 pJ per pulse (< 150 ps) was used to excite abSTAR635P molecules, and a pulsed 775 nm STED depletion laser with an output of 1.25 W and > 30 nJ per pulse (jitter < 100 ps) was used for stimulated emission depletion.Time gating delay was set to 1 ns and gate width to 6 ns.The Stedycon was controlled via a Chrome browser-based software provided by the manufacturer (Abberior Instruments).
Simulations and software used for analysis Microscopy images were processed and analyzed with the opensource image processing package Fiji (Schindelin et al, 2012).More extensive image analysis was carried out with custom-written algorithms in Matlab (MathWorks).Frame, XY localization, intensity, and positional accuracy of single molecules were calculated with the ImageJ plugin ThunderSTORM (Ovesny et al, 2014).Matlab, GraphPad Prism 10 (GraphPad), and R (r project) were used for statistical data analysis and evaluation.

STED autocorrelation analysis
To judge the randomness of pMHCII in STED microscopy data, we calculated ACFs for four different 2.5 × 2.5 μm regions of interest per cell and compared these experimental samples with ACFs of simulated STED images of randomly distributed molecules obtained with experimentally determined parameters (Sengupta et al, 2011).
To characterize the point-spread function of the imaging system, we sparsely labeled activated BMDCs and B-cells with the 14.4.4scF V -abSTAR65P and fitted the recorded single molecules signals using the ImageJ plugin ThunderSTORM.The background was determined from the mean fluorescence signal of non-fluorescent areas within each region of interest.I-E k densities were calculated by dividing the background subtracted mean intensity in fully labeled cells by the average single-molecule intensity, which yielded 15-108 I-E k molecules μm -2 on activated BMDCs and 24-44 I-E k molecules μm -2 on activated B-cells.Simulated images of random signal distributions were then generated using the parameters intensity (I), and sigma (σ) derived from single-molecule signals and background and molecular density from the cell regions of interest using an in-house written Matlab code (for details see Rossboth et al, 2018).
Simulation of random distributions for blinking analysis via Ripley's K A region of interest (e.g., 3,000 × 3,000 nm) featuring randomly distributed molecules at a specified density was simulated as described (Rossboth et al, 2018;Platzer et al, 2020).To arrive at simulations to be used for comparison with cell-associated PALM data, the approximate expression levels of I-E k (14.4.4 scF V -AF647-biotin + mSav-cc-PSCFP2), CD205 (NLDC-145 mAb-AF647-biotin + mSav-cc-PSCFP2), and CD18 (M18/2 mAb-AF647biotin + mSav-cc-PS-CFP2) within a region of interest were determined by dividing the number of PS-CFP2 localizations with the mean number of detections N per fluorescence molecule as obtained from the blinking analysis (Appendix Fig S7A).The experimentally derived probability distribution of the number of detections N was randomly assigned to every simulated localization to include the blinking of PS-CFP2.To arrive at simulations to be used for comparison with PALM data after MBC, we determined the number of PS-CFP2 localizations from the corrected region and set the mean number of detections N per fluorescence molecule to N = 1.Finally, all positions were shifted into a random direction by a distance drawn from a normal distribution with an average value of zero and localization precision (51 AE 12 nm) as standard deviation to simulate a PFA-fixed scenario.In order to simulate live-cell scenarios, we additionally shifted the localizations based on the diffusion parameters F (fast diffusing fraction Monte-Carlo simulation of cluster detection using PALM Ripley's K functions were determined from Monte-Carlo simulations of nanoclusters with a cluster radius of 20, 40, 60, or 100 nm, a fraction of molecules residing inside clusters between 20 and 100%, 3, 5, 10, 15, and 20 clusters μm -2 , an average molecular density of 100 molecules μm -2 , and the experimentally determined blinking statistics of PS-CFP2 determined under 4% PFA-fixed and non-fixed imaging conditions (Appendix Fig S7A).Functions were compared to those calculated from random localization distributions and categorized as "Resolvable," "Borderline," or "Not resolvable" as described in Platzer et al (2020).
Monte-Carlo simulation of TOCCSL TOCCSL results were simulated based on experimentally derived parameters using a Monte-Carlo-based approach as outlined in detail in Bodner et al (2023).In brief, a cell size of 10.5 × 10.5 μm 2 containing 500 molecules μm -2 diffusing with an average diffusion constant of 0.244 μm 2 s À1 (D = 0.83*0.284+ 0.17*0.046μm 2 s À1 , see Appendix Fig S7B) was simulated.An area of 7 × 7 μm 2 was photobleached for 750 ms with bleaching parameters derived from photobleaching curves of AF647 illuminated with a laser power density of 3 kW cm À2 and a diffraction-affected laser intensity profile as determined experimentally.For TOCCSL simulations, the fraction of molecular dimers was varied between 3 and 100% while keeping the overall density of molecules constant.For each dimer fraction, the fraction of detected dimers after a recovery time of 2 s was determined from the simulations.Simulations were repeated 1,000 times for each dimer fraction.The 5% detection level of the TOCCSL method is described in Brameshuber & Schutz (2012).
Monte-Carlo simulation of FRET efficiency FRET efficiencies of randomly distributed donor and acceptor molecules at increasing densities were simulated as described (Brameshuber et al, 2018).We used algorithms custom written in Matlab to simulate random X and Y positions of pMHCII molecules for increasing surface densities between 20 and 1,000 molecules μm -2 (increment 20).Due to higher simulated surface densities compared to our previous study (Brameshuber et al, 2018), we considered not only the nearest neighbor of every donor molecule for calculation of the FRET efficiency but also the second nearest neighbor.To this end, the distances d 1,2 between every donor molecule and the two nearest acceptor molecules were calculated and used to determine the FRET efficiencies E 1,2 , according to equation ( 13), with R 0 = 5.1 nm, the Fo ¨rster radius for the AF555/AF647 dye pair.
The overall FRET efficiency for a donor with two acceptor molecules (Koushik et al, 2009) was calculated according to equation (14).
For every surface density, the simulation was repeated 20 times to arrive at a mean and range of FRET efficiencies.FRET efficiencies were plotted against the total density (sum of donor and acceptor molecules).
as analyzed through quantitative flow cytometry.Statistics: median and Mann-Whitney U-test.Data were pooled from two biological replicates each containing n = 2 (BMDCs day 7) and n = 2-4 (activated BMDCs) technical replicates.Right panel: histogram depicting the distribution of surface-expressed I-E k molecules on nonactivated (dashed) and activated (solid) CD11c + BMDCs (binning = 0.15 (log scale), n = 14,585 non-activated BMDCs, n = 24,214 activated BMDCs).Data are representative of two biological replicates.P-value format: P > 0.05 (NS), P ≤ 0.01 (**).C I-E k is expressed at high numbers on non-activated and activated CD19 + B-cells.Left panel: the median number of I-E k molecules displayed on the surface of nonactivated and activated CD19 + B-cells (n > 10,000 cells per analysis) as analyzed through quantitative flow cytometry.Statistics: median and Mann-Whitney U-test.Data were pooled from three biological replicates each containing n = 2-3 technical replicates.Right panel: histogram depicting the distribution of I-E k molecules expressed on non-activated and activated CD19 + B-cells (binning = 0.15 (log scale), n = 14,032 non-activated CD19 + B-cells, n = 10,010 activated CD19 + B-cells).Data are representative of three biological replicates.P-value format: P ≤ 0.001 (***).D I-A k and I-E k molecules are expressed on the surface of activated CD11c + BMDCs and activated CD19 + B-cells at similar levels.Left panel: staining procedure employed for quantitation.Right panel: I-A k and I-E k expression levels as determined on the basis of I-E k surface expression after correcting for the degree of fluorophore conjugation of the antibodies employed.The mAb titration curves were fitted to a one-site-specific binding model to estimate B max and the 95% confidence interval.Goodness of fit R 2 > 0.99.Median fluorescence intensity of n > 10,000 cells shown.These data are representative of one experiment.E High I-E k densities as measured via TIRF microscopy on the surface of activated BMDCs and B-cells.Left panel: Scheme illustrating 14.4.4scF V -AF647-based staining as well as interference reflection microscopy (IRM)-and total internal reflection fluorescence (TIRF)-based microscopy readout to confirm attachment and spreading of cells on the ICAM-1-coated glass surface and density quantitation, respectively.Middle panel: Example of IRM and TIRF images of indicated cells on ICAM-1functionalized glass slides.Upper scale bar, 10 μm.Lower scale bar, 5 μm.Right panel: Dot plot of I-E k surface densities on non-activated BMDCs (median, n = 38 cells), activated BMDCs (median, n = 113 cells), non-activated B-cells (median, n = 141 cells), and activated B-cells (median, n = 188 cells).Data were pooled from two (BMDCs day 7), three (activated BMDCs), and five (non-activated or activated B-cells) biological replicates.Statistics: median, unpaired two-tailed Student's t-test.P-value format: P ≤ 0.001 (***).Source data are available online for this figure.
distributions featuring cell-derived molecular densities, the diffusion parameters of I-E k at 25°C (Appendix Fig S7B), and the respective blinking signature of PS-CFP2 (Appendix Fig S8A and B, for more details, see Materials and Methods).As shown in Fig 2B and summarized in Appendix Fig S8C, comparing respective Ripley's K functions of recorded I-E k localization maps with those derived from simulated random localization maps did not reveal Ripley's K maxima above the simulation (mean AE standard deviation) for most of the investigated cell regions, which let us conclude that pMHCIIs are uniformly distributed on the surface of activated BMDCs and Bcells, at least when judged within the cluster detection sensitivity of our approach (Appendix Fig S7A).Importantly and in contrast, we could detect local surface enrichments for the internalization receptor CD205 and for the integrin beta-chain CD18 (Fig 2C and D, and Appendix Fig S7C and D) on all investigated cell regions (Appendix Fig S8C).Confirming the sensitivity and robustness of our imaging and analysis methodologies, background localizations registered only in negligible numbers in the absence of PS-CFP2 (Appendix Fig S7E).To circumvent potential complications arising from molecular diffusion and cellular motility, we performed analogous imaging and analysis procedures on PFA-fixed activated BMDCs and Bcells.As shown in Appendix Fig S9A and B and summarized in Appendix Fig S8C, we frequently identified inhomogeneities and thus small deviations from a normal random distribution of I-E k on PFA-fixed cell regions, which could have also arisen from the fixation procedure itself.
Fig 5A,  we subjected cells to a laser-based TOCCSL illumination sequence, which included a highpowered laser pulse (750-1,000 ms) to quantitatively ablate the AF647 fluorescence of a spatially defined area, to be succeeded by the acquisition of a control image 100-750 ms after bleaching for verification of complete AF647 photobleaching(Fig 5A, center  panel).Following a recovery phase of 2-20 s, we recorded a sotermed TOCCSL image representing isolated diffraction-limited spots.These originated from the masked area of the cell surface as laterally mobile I-E k entities and had therefore not undergone bleaching(Fig 5A, right panel).To enumerate the number of 14.4.4-scFV -AF647-decorated I-E k molecules per diffraction-limited signal, we compared the obtained distribution of AF647 fluorescence brightness values ρ(B) with the brightness values obtained from recorded cells decorated with AF647-conjugated and unconjugated 14.4.4-scFV premixed at a ratio of 15:1 as monomer control (ρ1).A linear combination of multimeric n-mer contributions, ρn, derived from ρ1 and applied to fit the distribution ρ(B) yielded exclusively I-E k monomers (Fig5B).To determine the overall capacity of the TOCCSL procedure to reveal the existence of diffraction-limited entities exhibiting more than one fluorophore, we imaged cells that we had decorated with 14.4.4-scFV -AF647-biotin preconjugated with mSav-AF647.This regimen gave invariably rise to events associated with two AF647 dyes, and as shown in Fig5B, 19 and 30% of the events recorded in this fashion on activated BMDCs and activated Bcells, respectively, reflected dimers.The moderate detection efficiency, especially in the case of activated BMDCs, resulted in all likelihood from fast lateral I-E k diffusion during the application of the TOCCSL bleach pulse, which appeared to have led to a relatively high number of partially ablated dimeric events at the interface of the bleached and the masked area, and which we confirmed in Monte-Carlo-based TOCCSL simulations(Appendix Fig S14A).Similarly, crosslinking of I-E k -associated biotinylated 14.4.4scF V -AF647 with 125 nM divalent streptavidin revealed a dimer fraction of 20% on activated BMDCs (Fig5C).

A
Figure 4.
Fig 6A, second panel).Given the calculated inter-dye distance of approximately 4.1 nm within the 5c.c7TCR-I-E k /peptide complex (Fig 6A), we expected TCR-pMHCII engagement to result in a FRET yield of about 78% (Fig 6B and Appendix Fig S15A), which would be sufficiently high to visualize synaptic TCR-pMHCII interactions and quantitate their lifetime at the single molecule level (Appendix Fig S15B).In fact, ablation of AF647 gave indeed rise to an increase in AF555 FRET donor fluorescence as a direct measure for the FRET yield (Fig 6B and equations 7-10-see Materials and Methods part).
Single I-E k /MCC-PEG 2 -biotin molecules were tracked over time with a time lag of 10.5 ms.The graphs demonstrate the mean square displacement of the total pool and newly arrived I-E k molecules on activated BMDCs (10,051 trajectories on 14 cells, total pool; 8,607 trajectories on 13 cells, 10 min recovery) and activated B-cells (11,312 trajectories on 16 cells, total pool; 8,974 trajectories on 15 cells, 10 min recovery).Data are representative of one (BMDCs) and two (B-cells) biological replicates.Diffusion constants D were calculated based on the first two data points.Fractions were calculated by fitting the first 10 data points.Statistics: mean and standard deviation.D, E Fast-moving fraction of newly arriving and the total pool of I-E k molecules on activated BMDCs (D) and activated B-cells (E) were calculated by fitting the recorded I-E k trajectories to a binary diffusion model (left panel).Mean square displacement plot of fast-moving I-E k molecules (middle panel) and slowly moving I-E k molecules (right panel) are plotted for the indicated time lags employed in the tracking experiments.Statistics: mean and standard deviation.F = fraction of fastmoving molecules.D 1 = diffusion constant of fast-moving fraction.D 2 = diffusion constant of slowly moving fraction.Diffusion constants were calculated by fitting the first two data points.Fractions were calculated by fitting the first 10 data points.Source data are available online for this figure.
. As schematically illustrated in Fig 6D, NVOC sterically hinders I-E k binding to the 5c.c7TCR, unless it becomes ablated on demand through a brief violet (λ = 405 nm) laser pulse to release I-E k /MCC for TCR engagement (DeMond et al, 2006).As is demonstrated in Fig 6D-F, T-cells did not engage the provided I-E k /MCC[NVOC] until photo-uncaging had occurred.
Statistics: mean and standard deviation of n = 13-81 T-cell synapses recorded between 2 and 12 min after establishing PLB contact.Data were pooled from 1-2 biological replicates.D Scheme of a FRET-based assay to analyze I-E k /MCC-TCR-binding events in the presence or absence of bystander pMHCIIs with and without the use of a caged peptide (MCC[NVOC]).E Relative synaptic TCR occupancy of T-cells approaching functionalized PLBs displaying the agonist I-E k /MCC[NVOC]-AF647 together with I-E k /K99A-AF488 as well as ICAM-1 and B7-1 (both 100 molecules μm À2 ).Binding of the TCR to I-E k /MCC-AF647 was quantitated 25 s after photo-uncaging via FRET DRAAP.Scale bar, 5 μm.Statistics: mean and standard deviation and unpaired two-tailed Student's t-test.Each gray circle represents a T-cell synapse (n = 10-13 cells) of one experiment.P-value format: P > 0.05 (NS), P ≤ 0.001 (***).F Single-molecule FRET analysis of T-cells approaching PLBs functionalized with the photoactivatable agonist I-E k /MCC[NVOC]-AF647 in low abundance and I-E k /K99A-AF488 in high abundance together with ICAM-1 and B7-1 (100 μm À1 ).Ten seconds after the uncaging pulse, single-molecule FRET events were recorded and counted as individual TCR-I-E k /MCC-binding events (white arrow) if they had disappeared within one frame during recording.Scale bar, 10 μm.Statistics: mean, standard deviation, and unpaired two-tailed Student's t-test.Each gray circle represents a T-cell synapse (n = 10-23 cells) of one experiment.P-value format: P > 0.05 (NS), P ≤ 0.001 (***).G Synaptic TCR occupancy of T-cells engaging I-E k /MCC-AF647 in the presence of increasing densities of I-E k /K99A-AF488.Statistics: mean, standard deviation, and unpaired two-tailed Student's t-test.Each gray circle represents a T-cell synapse (n = 30-52 cells) of one experiment.P-value format: P > 0.05 (NS), P ≤ 0.001 (***).H Synaptic TCR occupancy of T-cells engaging I-E k /K99A-AF647 in the presence of increasing I-E k /MCC-AF488 densities.Statistics: mean, standard deviation, and unpaired two-tailed Student's t-test.These data are representative of two experiments.Each gray circle represents a T-cell synapse (n = 46-61 cells) of one experiment.P-value format: P > 0.05 (NS), P ≤ 0.05 (*), P ≤ 0.01 (**), P ≤ 0.001 (***).I Synaptic lifetime of I-E k /MCC-TCR interactions in the presence or absence of 1,060 I-E k /K99A-AF488 μm À2 as measured via single-molecule FRET imaging at 26°C.Number of smFRET traces (n = 135-227) to calculate the apparent lifetime (app t off ) for each time lag (56, 84, 252, 476, 924, and 1,820 ms) are summarized in Appendix Fig S17E.Source data are available online for this figure.
peptides(Hellmeier et al, 2021a(Hellmeier et al, , 2021b)).More specifically and as shown in Fig 7A,we built rectangular DNA origami tiles of 100 × 70 nm in size, with either a single or two functionalization sites at 10 and 20 nm distance on the top side and attachment sites for cholesterol-DNA at the bottom side for PLB anchorage (for details, please see Materials and Methods section).To create defined pMHCII homodimers, DNA origami structures were functionalized with biotinylated oligonucleotides, divalent streptavidin, and biotinylated pMHCII; for heterodimers, we preassembled DNAconjugated monovalent streptavidin and biotinylated pMHCIIs prior to DNA origami decoration.As shown in Fig7A, both strategies allowed for site-specific DNA origami attachment of I-E k /MCC-AF555 or I-E k /K99A-AF647.Biotinylated I-E k /MCC-AF555, which had been attached to poly-histidinetagged monovalent streptavidin, served as control for monomeric pMHCII.Engineered ligand occupancies of DNA origami-based platforms were verified by comparing the brightness of single DNA origami platforms to single fluorescence emitters recorded via TIRFbased microscopy.As shown in Appendix Table
. D Scheme illustrating the means by which intracellular calcium levels were recorded to assess the impact of highly abundant bystander pMHCIIs on the recognition of agonist pMHCIIs present in low abundance.E Calcium response of T-cells in contact with PLBs featuring the agonist I-E k /MCC at a density of 0.2-0.6I-E k /MCC μm À2 , bystander pMHCIIs at high densities (763 to 1,264 I-E k μm À2 ), as well as ICAM-1 and B7-1 (both present with 100 molecules μm À2 ).Shown are percentage of activated T-cells (upper left), probability density plot of Fura-2 ratios of individual T-cells (upper right), mean Fura-2 ratios of activated T-cells over time (lower left), and mean Fura-2 ratios of the entire T-cell population synchronized to onset of signaling (indicated by the highest Fura-2 value within a given track; lower right).Data are representative of >3 biological replicates with the use of antigen-experienced T-cells (n = 233-814 cells per condition) and isolated from >3 different 5c.c7 TCR-transgenic mice (Appendix Fig S18B and C).F Calcium response of antigen-experienced 5c.c7 TCR-transgenic T-cells (n = 62-441 cells per condition) exposed to increasing levels of the agonist I-E k /MCC (n = 8-9 conditions) presented together at indicated densities with I-E k /K99A, I-E k /ER60, or I-E k /null.Three-parameter dose-response curve fit is summarized in the source data.Source data are available online for this figure.Ó 2023 The Authors EMBO reports 24: e57842 | 2023 scanning T-cell in a fully autonomous fashion.In view of future improved designs of T-cell-based immunotherapies, ), D 1 (diffusion constant of fast-moving fraction), and D 2 (diffusion constant of slowly moving fraction) of I-E k on BMDCs (F = 0.83, D 1 = 0.284 μm 2 s À1 , D 2 = 0.046 μm 2 s À1 , Appendix Fig S7B), I-E k on B-cells (F = 0.77, D 1 = 0.150 μm 2 s À1 , D 2 = 0.024 μm 2 s À1 , Appendix Fig S7B), CD205 (F = 0.9, D 1 = 0.065 μm 2 s À1 , D 2 = 0.0006 μm 2 s À1 (Manzo et al, 2012)), and CD18 (F = 0.3, D 1 = 0.03 μm 2 s À1 , D 2 = 0.0 μm 2 s À1 (Mukherjee et al, 1998)), and a time lag of 2.3 ms.

table Reagent
, and the guidelines of the ethics committees of the Medical University of Vienna.Mice were kept and bred in the animal facility of the Medical University of Vienna.
(Axmann et al, 2015)ncy is directly proportional to the FRET yield(Axmann et al, 2015), we established a relationship between these two parameters by plotting the synaptic FRET yields obtained from FRET DRAAP measurements against TCR occupancies obtained from the same T-cell synapses(Appendix Fig S15F).By fitting a linear regression, we experimentally determined the