Cholesterol and Sphingomyelin Drive Ligand-independent T-cell Antigen Receptor Nanoclustering*

Background: The TCR forms nanoclusters in the plasma membrane independent of ligand binding. Results: Membrane cholesterol and sphingomyelin facilitate TCR nanoclustering, thereby enhancing the avidity toward the ligand. Conclusion: The membrane lipid composition regulates the degree of TCR nanoclustering and thus T-cell sensitivity. Significance: This work contributes to the understanding of the consequences of specific lipid-membrane protein interactions. The T-cell antigen receptor (TCR) exists in monomeric and nanoclustered forms independently of antigen binding. Although the clustering is involved in the regulation of T-cell sensitivity, it is unknown how the TCR nanoclusters form. We show that cholesterol is required for TCR nanoclustering in T cells and that this clustering enhances the avidity but not the affinity of the TCR-antigen interaction. Investigating the mechanism of the nanoclustering, we found that radioactive photocholesterol specifically binds to the TCRβ chain in vivo. In order to reduce the complexity of cellular membranes, we used a synthetic biology approach and reconstituted the TCR in liposomes of defined lipid composition. Both cholesterol and sphingomyelin were required for the formation of TCR dimers in phosphatidylcholine-containing large unilamellar vesicles. Further, the TCR was localized in the liquid disordered phase in giant unilamellar vesicles. We propose a model in which cholesterol and sphingomyelin binding to the TCRβ chain causes TCR dimerization. The lipid-induced TCR nanoclustering enhances the avidity to antigen and thus might be involved in enhanced sensitivity of memory compared with naive T cells. Our work contributes to the understanding of the function of specific nonannular lipid-membrane protein interactions.


SUMMARY
The T-cell antigen receptor (TCR) exists in monomeric and nanoclustered forms independently of antigen binding. Although the clustering is involved in the regulation of T-cell sensitivity, it is unknown how the TCR nanoclusters form. We show that cholesterol is required for TCR nanoclustering in T cells, and that this clustering enhances the avidity but not the affinity of the TCR-antigen interaction. Investigating the mechanism of the nanoclustering, we found that radioactive photocholesterol specifically binds to the TCRβ chain in vivo. In order to reduce the complexity of cellular membranes, we used a synthetic biology approach and reconstituted the TCR in liposomes of defined lipid composition. Both cholesterol and sphingomyelin (SM) were required for the formation of TCR dimers in phosphatidylcholine-containing large unilamellar vesicles. Further, the TCR was localized in the liquid disordered phase in giant unilamellar vesicles. We propose a model in which cholesterol-and SM-binding to the TCRβ chain causes TCR dimerization. The lipid-induced TCR nanoclustering enhances the avidity to antigen, and thus might be involved in enhanced sensitivity of memory compared to naïve T cells. Our work contributes to the understanding of the function of specific nonannular lipid-membrane protein interactions.
The T-cell antigen receptor (TCR) is a multisubunit transmembrane protein complex responsible for the triggering of a T-cell mediated adaptive immune response. It consists of the antigen-recognizing TCRαβ (or TCRγδ) heterodimer and the signaltransducing CD3 dimers: CD3εγ, CD3εδ and ζζ (1,2). The basic functional unit of the TCR − defined as the monomeric TCR − has a TCRαβCD3δεγεζζ stoichiometry (3)(4)(5)(6). TCR nanoclusters have been detected by Blue Native (BN)-PAGE analysis, gel filtration and co-immunopurification of two different TCRs co-expressed on the same cell (4,7,8) as well as by immuno-gold electron microscopy (4,9,10), high-speed photoactivated localization microscopy and dual-color fluorescence cross-correlation spectroscopy (10). TCR nanoclusters have a size of 2 to 30 TCRs and are co-expressed with TCR monomers (4,10). Using bioluminescence resonance energy transfer and two-color coincidence detection microscopy it was shown that up to 7-10% of the TCRs could be nanoclustered (11,12). Here we use the term "nanocluster" to distinguish TCR nanoclustering, which is independent of MHC-peptide (MHCp)binding, from activation-induced microcluster formation of 20-70 TCRs (13). TCR nanoclusters increase the sensitivity of T cells to antigenic stimulation (9), possibly because TCR nanoclusters can be stimulated at lower antigen concentrations than monomeric TCRs (4,14). However, the molecular mechanism by which T cells regulate the extent of TCR nanoclustering is unknown. The mechanisms that contribute to the lateral segregation of proteins and lipids are subject of intense research. It was hypothesized that in cells at least two distinct membrane microdomains exist: the cholesterol-and sphingomyelin (SM)-rich lipid rafts and the phospholipid enriched non-rafts (15). Due to post-translational lipid modifications, a number of proteins partition into the raft domain (16), where they might form clusters (17). In artificial membranes containing cholesterol, SM and phospholipids, the formation of lateral liquid ordered (l o ) and liquid disordered (l d ) phases occurs (18), which might correspond to the raft and nonraft domains. This phase separation is facilitated by the interaction of cholesterol with SM (19). The actin cytoskeleton (20,21) and protein-protein interactions (22) can also be involved in the clustering of membrane proteins.
In this report, we study the role of the lipid environment in the formation of TCR nanoclusters and the physiological relevance of TCR nanoclustering, i.e. the avidity of the TCR-antigen interaction.
Generation of expression plasmids and cell lines -To generate the expression vector pcDNA3_mζ-SBP, the DNA fragment coding for the mouse ζ chain C-terminally linked to the streptavidin-binding peptide (SBP) purification tag (23) was amplified by PCR and cloned into the EcoRI/XhoI site of the pcDNA3 vector (Invitrogen). pcDNA3_mζ-SBP was transfected into the mouse 2B4-derived ζ-deficient line MA5.8 to yield M.mζ-SBP. The cDNA of the human TfR C-terminally linked to the SBP tag was amplified by PCR and inserted into the BglII/XhoI site of the pMIG-based expression vector, pMItom (provided by R.Y. Tsien). pMItomTfR-SBP was transfected into MA5.8 cells, to yield the M.hTfR-SBP cell line. To obtain the expression vectors for the BiFC assay, cDNA coding for the mouse ζ chain was C-terminally linked to enhanced GFP, the N terminal part (residues 1-172, YN) of a yellow fluorescent protein (venus), and the C terminal part (residues 155-238, CC) of enhanced cyan fluorescent protein (eCFP) (both from Clontech), amplified by PCR, and cloned into the BglII/XhoI site of pMItom. The vectors were transfected into M.mζ-SBP cells yielding the M.mζ-SBP/mζ-GFP, M.mζ-SBP/mζ-YN and M.mζ-SBP/mζ-CC by guest on July 16, 2017 http://www.jbc.org/ Downloaded from cell lines. The human T cell line 31-13.scTCRβ has been described (24). All cells were cultured in complete RPMI-1640 media supplemented with 5% fetal calf serum.

Treatments, cell lysis, immunoprecipitation and
immunoblotting -For actin depolymerization, 1 or 5 µg/ml Latrunculin A was used at 37°C for 30 min. For cholesterol depletion and loading, treatments with 2 mM methyl-β-cyclodextrin (mβCD) for 2 min or 20 µg/ml cholesterol complexed to mβCD for 3 h (both at 37°C) were performed. The cholesterol concentration in lysates was measured using the Amplex-Red Cholesterol Assay Kit (Invitrogen). Serial lysis was performed by re-solubilizing the cellular and membrane material after each 15 min lysis and 15 min centrifugation steps (14000 g) three times by 1% saponin, and subsequently by 0.5% Brij96 in the lysis buffer containing 20 mM TrisHCl (pH8), 137 mM NaCl, 2 mM EDTA, 10% glycerol, 10 µg/ml leupeptin, 10 µg/ml aprotinin and 1 mM PMSF. Pooled saponin lysates represented the nanoclustered TCR, while the rest of the TCR solubilized in Brij96 was the monomeric (Fig. 1A). Anti-TCR IPs were performed with 4 µg anti-CD3ε (145-2C11), anti-TCRβ (H57) or anti-TCRζ (448) antibodies at 4°C for 4 h. Native, purified TCR preparations were separated on BN-PAGE as described (25). SDS-PAGE and immunoblotting were performed by conventional methods. Western Blot quantifications were done with the Li-Cor Odyssey infrared imager system or with the ImageJ program after chemoluminescence detection.
MHCp-binding experiments -To quantify the binding of MHCp-tetramers and MHCpmonomers to the different TCR forms, we used PbCS peptide(ABA)-H2-Kd as the MHC molecule and T1.4 hybridoma T cells that do not contain the co-receptor CD8 (26). T1.4 cells were treated with 2 mM mβCD for 30 minutes at 37ºC or with cholesterol complexed to mβCD as above, and then serially lysed in first saponin and then digitonin-containing buffer, in order to test for the disassembly of the TCR nanoclusters. TCR IP was performed with anti-CD3ε, and samples were subjected to SDS-PAGE, anti-ζ WB and detection with Li-Cor Odyssey infrared imager. In parallel, treated and nontreated T1.4 cells were coupled with 250 nM PE-labeled PbCS peptide(ABA)-H2-Kd tetramers or PbCS peptide(ABA)-H2-Kd monomers for 2 hours on ice. Subsequently, bound MHCp molecules were covalently cross-linked to the TCR by UV irradiation. In case of the MHCp monomer, the cells were stained with streptavidin-PE. Surface TCR expression was measured by staining with FITC-labeled anti-TCRβ. Fluorescence was measured by flow cytometry with a Calibur flow cytometer (Becton-Dickinson). Statistical analysis was done with the Prism4 software.
Protein purification -Proteins were purified from T or B cells expressing the appropriate construct. 5x10 8 cells were lysed and affinity purifications were performed using streptavidin-conjugated agarose (GE Healthcare) in case of SBP-linked constructs, or nitrophenol (NP)-conjugated agarose (Biosearch Technologies) in case of the scTCRβ-containing TCR and the NP-specific BCR. Elution was performed by incubation for 30 min at 4°C with 2 mM free biotin, or 2 mM nitro-iodo phenol respectively, in BN lysis buffer containing 20 mM Bis-Tris pH7.0, 500 mM ε-aminocaproic acid, 20 mM NaCl, 2 mM EDTA, 10% glycerol and detergent as indicated.
TCR reconstitution in LUVs -Liposomes with different membrane compositions using soybean phosphatidylcholine, egg sphingomyelin (Lipoid) and cholesterol (Sigma-Aldrich) were prepared with the thin film method (27 Approximately 0.1 µg of the purified TCR in 100 µl 0.02% Triton X-100 containing buffer was mixed with 100 µl of 2 mM LUV preparation, and 40 µg Triton X-100 was added. Samples were agitated for 30 min at 4°C, and the detergent was removed by adsorption to 3 mg of BioBeads SM-2 (BioRad) per sample at 4°C over night. The same procedure was used for the generation of TfR-or BCR-containing proteoliposomes.
Flow cytometry and BiFC -Flow cytometry for analyzing TCR expression was performed by conventional methods with a Calibur flow cytometer.
For BiFC experiments proteoliposomes were formed by separately adding the TCRs bearing the C-and the Nterminal part of the fluorophore to the preformed vesicles. As a control, only one half fluorophore, and a GFP-linked TCR was reconstituted. After TCR reconstitution the liposomes were lysed in lysis buffer containing 1% digitonin and an IP was performed with anti-CD3ε (145-2C11)coupled to carboxylate-modified latex beads (Invitrogen). Fluorescence was measured by flow cytometry and analyzed with the FowJo 8.2 software. Statistical analysis was done with the Prism4 software.
Preparation of GUVs -Giant unilamellar vesicles (GUVs) were made by the electroformation technique. Proteoliposomes (3 g/l) in droplets of 2 µl were deposited on indium tin oxide (ITO) covered glass slide. The film was partially dried overnight in a desiccator under saturated vapor pressure of a saturated NaCl solution. The ITO cover slip was assembled together with a second ITO coverslip into a flow-chamber of homemade design, which was filled with a buffer containing 1mM HEPES, 1mM NaCl, pH 7.4. Alternating electric field (400 V/m, 10 Hz) was applied for 4 h at room temperature. GUVs were observed in the same chamber. Images were collected with Zeiss 510 ConfoCor3 microscope using water C-Apochromat 40x, NA 1.2 objective and avalanche photodiodes as detectors.

RESULTS
Cholesterol stabilizes TCR nanoclusters -We established an efficient affinity purification procedure for the isolation of the native TCR from cellular lysates. We used mouse M.mζ-SBP cells derived from the 2B4 T cell hybridoma, expressing a streptavidin-binding peptide (SBP) purification tag C-terminally fused to the ζ chain, and human 31-13.scTCRβ cells derived from the Jurkat T cell line, which contain a single chain variable fragment of an anti-nitrophenol (NP) antibody linked to the TCRβ chain (scTCRβ) (4,24). We analyzed the effect of different detergents on the integrity of TCR nanoclusters using Blue Native (BN)-PAGE (4,25) in case of both human and mouse T cells (Fig. 1A). 1% digitonin extracted the mouse TCR in monomeric form (lane 1), whereas 0.5% Brij96 extracted TCR monomers and nanoclusters (lane 2) as reported before (4,5). In contrast, 1% saponin extracted only the nanoclusters (lane 3). The same held true for the human TCR (lanes 4, 5 and 6). Brij96 extracted the remaining monomeric TCR from the saponin-insoluble membranes (lane 7). As controls, digitoninor Brij96-extracted TCRs did not aggregate when kept in a saponin-containing buffer by guest on July 16, 2017 http://www.jbc.org/ Downloaded from (Fig. 1B, lanes 1 and 2), while saponin-and Brij96-solubilized TCR nanoclusters broke down to monomeric TCR when digitonin was added (lanes 3 and 4). The detergentdependence of the TCR nanocluster stability suggested that membrane lipids play a role in the formation of the nanoclusters. Previously, using harsh conditions to extract cholesterol from cells, we suggested that cholesterol might stabilize the TCR nanoclusters (4). To test whether cholesterol is involved in TCR nanoclustering, we reduced the amount of cholesterol using a short-term low-dose methyl-β-cyclodextrin (mβCD) treatment (2 mM for 2 minutes at 37°C), which does not extract membrane proteins (29). To increase the cholesterol content of the membranes, mβCD-complexed cholesterol was used. The change of the cholesterol content of total cell lysates was measured with Amplex-Red cholesterol assay kit (Fig. 1C). To quantify the ratio of nanoclustered to monomeric TCRs, we lysed the cells serially in 1% saponin and 0.5% Brij96 (as in Fig. 1A). Anti-CD3ε immunopurification and SDS-PAGE were performed on the two fractions, and the amount of assembled TCR was quantified by anti-ζ WB (Fig.  1D). The nanocluster/monomer ratio decreased upon cholesterol depletion, and increased upon cholesterol loading. The treatments did not extract TCRs from the membrane (Fig. 1E). These results show that TCR nanoclustering is reversible and dependent on the cholesterol content of the plasma membrane. Disrupting actin filaments by Latrunculin A-treatment did not disassemble TCR nanoclusters (Fig.  1F) indicating that the actin cytoskeleton is dispensable for the maintenance of the TCR nanoclusters.
The TCRβ subunit binds to cholesterol in living T cells -Since we found that TCR nanoclusters depend on the presence of cholesterol, we used a biochemical approach to test whether the TCR can bind to cholesterol in vivo. To this end, we used a photoactivatable radioactive analogue of cholesterol (28). Photocholesterol mimics cholesterol in several assays, and has been used to identify proteins that specifically interact with cholesterol (30,31). Here, Jurkat cells were grown in the presence of photocholesterol and upon activation by UV light, photocholesterol cross-linked to molecules in close proximity. The TCR and non-assembled CD3 dimers were immunoprecipitated with anti-CD3 antibodies from cellular lysates and analyzed by non-reducing SDS-PAGE and autoradiography ( Fig. 2A). A large number of proteins were cross-linked to cholesterol in the lysate, but only two bands were detected in case of the purified TCR, corresponding to the disulfide-linked TCRαβ dimer and CD3δ and/or CD3ε. When isolated from the same cellular lysates, the highly abundant cell surface protein CD45 did not bind cholesterol (lane 3). Using the same method, we studied the interaction between cholesterol and the antigen receptor of A20 B cells, and we found that cholesterol did not cross-link to the BCR (Fig. 2B).
To identify which subunits of the TCR bound cholesterol, we had to distinguish between the TCRα and TCRβ, and the CD3δ and CD3ε chains. First, we used 31-13.scTCRβ cells, which bear the larger scTCRβ chain (4,24) and Jurkat T cells expressing a wt TCR. After labeling with photocholesterol, a TCR IP was performed with anti-CD3ε and samples were subjected to reducing SDS-PAGE and autoradiography. In addition to the signal from CD3δ and/or CD3ε, a protein with the apparent molecular weight of 45 kDa (in case of Jurkat) and 80 kDa (in case of the 31-13.scTCRβ cells; corresponding to the scTCRβ) was detected (Fig. 2C). No additional band corresponding to the size of TCRα was detected, indicating that it was TCRβ that bound to photocholesterol. CD3ε, in contrast to CD3δ, is not glycosylated. Therefore, we used a deglycosylation assay to distinguish between CD3ε and CD3δ. An IP was performed from photocholesterollabeled lysates of Jurkat cells using anti-TCRβ, anti-CD3ε or anti-ζ antibodies (Fig.  2D). Each sample was left untreated or deglycosylated using N-glycosidase F. The increased mobility of the deglycosylated TCRβ in the autoradiogram indicates efficient deglycosylation. The mobility of the low molecular weight radioactive band did not change, showing that it represents CD3ε. However, cholesterol-labeled CD3ε was only present in the anti-CD3ε IPs (lanes 3 and 4), although CD3ε was present in all IPs, as seen in the anti-CD3ε WB. This suggests that CD3ε bound to cholesterol was not the part of the TCR complex and was probably located intracellularly. Therefore, within the TCR complex, TCRβ is the only subunit, which associates with cholesterol in living T cells.
TCR nanoclusters increase the avidity towards multivalent MHCp -Next, we investigated whether TCR nanoclustering influences MHCp-binding to the TCR. We measured the binding of fluorescent PbCS peptide(ABA)-H2-K d (MHCp) to CD8negative T1.4 T cells, as this system allows exact quantifications due to the fact that the MHCp can be covalently cross-linked to the TCR (24,26). As above, we altered the TCR nanocluster/monomer ratio by extracting or loading cholesterol (Fig. 3A). The total amount of TCR in each treatment remained unchanged (Fig. 3B, left panel). However, there was a significant reduction in MHCp tetramer binding to cholesterol-depleted cells and an increased binding to cells where cholesterol was added (middle panel). Importantly, the binding of MHCp monomers was unaffected (right panel). Moreover, re-addition of cholesterol after previous extraction reversed the effect of mβCD treatment on MHCp tetramer binding. When after treating the T1.4 cells with mβCD, exogenous cholesterol was added, TCR nanoclusters were restored in cells reloaded with cholesterol (Fig. 3C), while the amount of surface TCR remained unchanged (data not shown). A dose-response curve of MHCp tetramer-binding to the cells shows that cholesterol-treatment increased the tetramer-binding capacity four to five fold (Fig. 3D, left panel). Again, the binding of MHCp monomers was unchanged (right panel). In conclusion, TCR nanoclustering significantly enhanced the TCR-MHCp avidity, represented by the MHCp tetramer binding (Fig. 3E), but did not change the TCR-MHCp affinity as measured by MHCp monomer binding (Fig. 3F).
Reconstitution of the TCR in large unilamellar vesicles -To learn more about the role of lipids in TCR nanoclustering, we reduced the complexity of cellular membranes using a synthetic biology approach. The native digitonin-solubilized TCR from M.mζ-SBP or 31-13.scTCRβ cells was purified (Fig. 4A), and it was reconstituted in large unilamellar vesicles (LUVs) with different compositions of phosphatidylcholine (PC), cholesterol and sphingomyelin (SM) (Fig. 4B). The orientation of the integrated TCR was determined by proteinase K digestion. For WB analysis, we used an anti-TCRβ antibody, which binds to the extracellular part of TCRβ, and anti-CD3ε and anti-ζ antibodies to detect the intracellular domains of CD3ε and ζ. Proteinase K treatment abolished the TCRβ signal (Fig. 4C), indicating that the extracellular part of the TCR is located extra-liposomally. The size of CD3ε decreased upon proteinase K treatment due to the loss of the extracellular domain, and a substantial fraction of ζ was resistant to digestion. Without integration into LUVs (lane 8) and after integration and subsequent lysis of the LUVs (lanes 3 and 6) the TCR was completely digested. These data show that most of the TCR was integrated into the liposomes in the same orientation as on the cell surface; i.e. the extracellular parts are extra-liposomal (Fig. 4B). Furthermore, we found that the majority of the TCR was integrated in the lipid bilayer and the reconstitution process did not alter the lipid content of the LUVs (Fig. 4D and E).
The TCR forms dimers in LUVs composed of PC, cholesterol and sphingomyelin -To study TCR nanoclustering, we chose the SBP-tagged TCR for its high purity and yield (Fig. 4A). The proteoliposomes of different lipid composition were sedimented, lysed in 1% saponin supplemented with 0.5% Brij96 to maintain, or with 1% digitonin to disrupt the TCR nanoclusters and analyzed by BN-PAGE. In liposomes containing a natural mixture of PC, the TCR remained monomeric (Fig. 5A, lane 1 7 and 8). To exclude the possibility that the TCR dimers were the result of TCR aggregation after the lysis of the proteoliposomes, we mixed the TCR with LUVs, but did not allow integration before lysing the vesicles and performing BN PAGE. As expected, TCR dimers did not form without integration of the TCR into the bilayer of the PC/chol/SM-LUVs (Fig. 5B). To assess whether dimerization in PC/chol/SM LUVs observed for the TCR is a general feature of TM proteins, we reconstituted the native purified transferrin receptor (TfR) purified from the parental cell line of the M.mζ-SBP cells and the purified B-cell antigen receptor (BCR) in PC or PC/chol/SM (40:30:30 mol%) LUVs. Neither TfR nor BCR multimers were detected (Fig.  5C). We also studied the effect of the temperature on the TCR clustering. Although we usually performed the reconstitution at 4°C, we obtained the same results at 37°C (Fig. 5D). In this particular experiment, the LUVs composed of PC contained more TCR than those composed of PC/chol/SM. The fact that TCR dimers formed only in case of the ternary mixture indicated, that TCR dimerization did not occur due to the congestion of a higher amount of proteins in the l d phase. Furthermore, we applied bifluorescence complementation (BiFC) as a detergentindependent read-out of TCR dimerization (32,33). To the C-terminus of the ζ chain, we fused either the N-terminal part of Venus or the C-terminal part of eCFP. The fusion proteins were individually expressed in M.mζ-SBP cells, which already expressed SBP-tagged ζ. Due to ζ dimer formation in the TCR complex, a portion of TCRs contained a half of the fluorophore after TCR purification via the SBP tag. Monomeric TCRs were purified from cellular lysates, and mixed at equimolar concentrations for reconstitution into LUVs. The formation of TCR dimers in the LUVs allows the assembly of the fluorescent domain (Fig. 5E). Significantly higher BiFC fluorescence was detected from PC/chol/SM LUVs (40:30:30 mol%) as compared to PC LUVs ( Fig. 5F and G). These results confirm that the presence of cholesterol and SM in PC-liposomes is sufficient for the formation of TCR dimers.
As a control, we show that the tags appended to ζ do not influence the nanoclustering of the respective TCRs (Fig. 5H).
The TCR localizes in the liquid disordered phase of GUVs -The PC and SM used in this study consisted of a spectrum of acyl chain lengths. We determined the phase behavior of the bilayers in GUVs by confocal imaging. No phase separation occurred in case of PC or PC and cholesterol (70:30 mol%), whereas in the mixture of PC, cholesterol and SM (40:30:30 mol%) we found a liquid ordered (l o ) and a liquid disordered (l d ) phase (Fig.  6A). To determine TCR localization in the artificial membranes, we reconstituted the purified GFP-coupled TCR in PC or PC/chol/SM (40:30:30 mol%) LUVs supplemented with 0.05 mol% DiD, a dye staining the l d domain, and grew GUVs from the proteoliposomes. Confocal imaging revealed that the distribution of the TCR was homogenous in PC liposomes, whereas in PC/chol/SM liposomes the TCR colocalized with the l d domain (Fig. 6B). A quantification of the distribution of the GFP-coupled TCR showed that 83±11 % of the TCRs was present in the l d domain in PC/chol/SM GUVs (Fig. 6C).

DISCUSSION
The molecular mechanism of TCR nanoclustering is poorly understood. In this study, we used T cells and a synthetic biology approach to investigate the role of lipids in antigen-independent TCR dimerization. We established a procedure to purify the complete TCR complex in native form and to reconstitute it in LUVs of different lipid composition. We found that TCR dimers formed in PC/chol/SM liposomes, but not in binary mixtures or in PC alone. The effect was specific to the TCR, since the TfR and BCR remained monomeric under all conditions. As the proteoliposomes did not contain other proteins than the TCR, we concluded that the lipid environment induced dimer formation. A number of specific lipid-protein interactions have been revealed by x-ray crystallography (34,35), radioactive photolipids (28,36) and mutagenesis analyses (37,38). Ordered cholesterol molecules were by guest on July 16, 2017 http://www.jbc.org/ Downloaded from shown in the structure of metarhodopsin (39) and of the β 2 -adrenergic G protein-coupled receptor (40,41). Therefore, we considered that a direct interaction with cholesterol might cause TCR dimerization. Indeed, in live T cells photoactivatable cholesterol (28) cross-linked to the TCRβ chain, but not to any other subunits of the assembled TCR. It also did not cross-link to the BCR or to CD45. This suggests that the TCRcholesterol interaction is remarkably sitespecific and that cholesterol is a nonannular lipid-binding partner of the TCR, i.e. cholesterol might bind stably to the TCR. Annular lipids were suggested to mediate intra-and intermolecular interactions between the TM regions involved (35,42). In analogy, we propose a mechanism of TCR dimerization, in which cholesterol and SM serve as structural components of a TCR dimer (Fig. 6D). In most (43,44) but not all (45) reports the TCR was found in the nonraft phase in resting T cells. Likewise, we show that the TCR is localized in the l d phase in our liposomes. Thus, we suggest that cholesterol binds to TCRβ in the l d phase of the plasma membrane.
Since SM preferentially interacts with cholesterol (46), cholesterol recruits SM to the TCR TM surface (Fig. 6D). The subsequent formation of TCR dimers is energetically favored, as it leads to the shielding of cholesterol-SM from the l d phase. In addition, protein-protein interactions between TCR subunits might stabilize the TCR-TCR association (9,47). This model is supported by our findings that cholesterol is required for the maintenance of TCR nanoclusters. Earlier, we studied the effect of a high-dose (4) and here of a mild mβCD treatment and cholesterol loading of T cells, and we found a correlation between cholesterol concentration in the membranes and TCR nanoclustering. Furthermore, disrupting the TCR-cholesterol interaction with digitonin led to disassembly of the TCR nanoclusters. Cholesterol amounts were measured in total cell lysates, thus the relevant concentrations of cholesterol in the plasma membrane are unknown. However, we measured an effect of cholesterol addition or removal on the ligand-binding activity of the surface TCR (Fig. 3), demonstrating that cholesterol levels were changed in the plasma membrane. The TCR in LUVs formed dimers but not multimers. In contrast, on the surface of T cells, larger nanoclusters are present (4,10). Since both dimers (in LUVs) and nanoclusters (in cells) are cholesterol dependent, we suggest that the nanoclusters derive from TCR dimers. It is unclear if the lack of nanoclusters in LUVs is due to the experimental settings (e.g. not more than two TCRs are present in one LUV), or if TCR dimers and nanoclusters form along a different mechanism. We propose that TCR dimerization is a dynamic process in which the equilibrium between the monomers and dimers is regulated by the concentration of cholesterol and SM (Fig. 6D). The cholesterol and sphingolipid content in activated T cells is higher than in naïve T cells (48). This might contribute to increased TCR nanoclustering in activated T cells as compared to naïve T cells (9). Activated T cells possess enhanced avidity, but not affinity, to pMHC tetramers when compared to naïve T cells (49). Increased avidity was dependent on cholesterol (45,49), but the underlying mechanism was unknown. Using MHCp tetramers and monomers, we show that cholesterol-mediated TCR nanoclustering translates into a higher antigen-TCR avidity. This suggests that upon activation and differentiation, naïve T cells upregulate their cholesterol and SM content, thereby expressing more nanoclustered TCRs. As a result, activated and memory T cells show enhanced avidity to MHCp (avidity maturation). In fact, activated and memory T cells posses increased sensitivity to low antigen levels as compared to naïve T cells (50). Our data and the new model for TCR dimerization presented here contribute to the understanding of TM protein clustering as well as to the consequences of specific binding of lipids to TM proteins.    The bands corresponding to PC, cholesterol and SM along with relative band intensity are indicated. The quantification demonstrates that the ratios of PC to SM and PC to cholesterol are unchanged after the TCR reconstitution. (E) Sucrose gradient centrifugation was performed with TCR containing proteoliposomes of pure PC, and the fractions were subjected to SDS PAGE followed by an anti-ε and anti-ζ WB. The fractions 2 and 3 contain the LUVs and the TCR, indicating that the TCR was integrated into the LUVs.