Allosteric activation of T-cell antigen receptor signalling by quaternary structure relaxation

The mechanism of T cell antigen receptor (TCR-CD3) signalling remains elusive. Here, we identified mutations in the transmembrane region of TCRβ or CD3ζ that augmented pMHC-induced signalling, not explicable by enhanced ligand binding, lateral diffusion, clustering or co-receptor function. Using a novel biochemical assay and molecular dynamics simulation, we demonstrated that the gain-of-function mutations loosened interaction between TCRαβ and CD3ζ. We found that, similar to the activating mutations, pMHC binding reduced TCRαβ cohesion with CD3ζ. This event occurred prior to CD3ζ phosphorylation and at 0°C. Moreover, we demonstrated that soluble monovalent pMHC alone induced signalling and reduced TCRαβ cohesion with CD3ζ in membrane-bound or solubilised TCR-CD3. Our data provide compelling evidence that pMHC binding suffices to activate allosteric changes propagating from TCRαβ to the CD3 subunits, reconfiguring interchain transmembrane region interactions. These dynamic modifications could change the arrangement of TCR-CD3 boundary lipids to licence CD3ζ phosphorylation and initiate signal propagation.


Introduction
Signalling through the TCR-CD3 complex drives thymocyte maturation into immunocompetent T cells and T cell response to foreign antigens (Stritesky et al., 2012).
These processes initiate upon TCR-CD3 ligation by highly polymorphic major histocompatibility complex (MHC) proteins carrying short peptides (p) originated from the degradation of self and foreign proteins. TCR-CD3 allows T cells to respond with exceptional specificity and sensitivity  to membrane-bound pMHC ligands of a virtual continuum of weak Kd (0.1-100 μM) and t1/2 of < 0.5 to several seconds (Aleksic et al., 2012;Cole et al., 2007;Stone et al., 2009) and ligand-receptor interfaces of diverse shape and chemical reactivity. To accomplish this task, TCR-CD3 employs a clonally distributed ab disulphide-linked dimer (TCR) with Ig-like variable domains, Va and Vb. VaVb contains the pMHC binding site composed of six loops homologous to antibody complementarity determining regions (CDRs) 1, 2 and 3 (Garboczi et al., 1996;Garcia et al., 1996). Germlineencoded CDR1 and CDR2 have limited variability, while CDR3s are hypervariable. VaVb orientates diagonally relative to the long axis of the peptide-binding groove (Garboczi et al., 1996;Garcia et al., 1996), with CDR3s contacting mainly the peptide and CDR1s, and CDR2s contacting primarily the MHC Garcia et al., 2012;Marrack et al., 2008).
Va and Vb are joined to Ig-like constant domains, Ca and Cb, that are linked to transmembrane regions (TMRs) via a stalk connecting peptide (CP). pMHC binding is signalled intracellularly by four non-covalently associated subunits (g, d, e and z), called CD3, organised into three dimers, ge, de and zz, the latter disulphide-linked (Call et al., 2002). e, g and d each exhibits an Ig-like extracellular domain (ECD) connected to their TMRs by short CPs, while z features a ≈ 10 residue-long ECD. A recent TCR-CD3 cryo-electron microscopy (EM) structure at 3.7 Å (Dong et al., 2019) largely reconciles with mutational and NMR studies (Call et al., 2002;He et al., 2015;Mariuzza et al., 2020;Natarajan et al., 2016) but reveals also unsuspected features. VaVb projects forward while Ca interfaces with CD3d ECD, Cb interfaces with CD3ge and CD3d ECDs, and CD3g and CD3e (of de) ECDs contact each other. Whilst the TMRs of both zz subunits (z1z2) and of ab interact with each other, de is contacted by a and z1, and ge is contacted by b and z2. The CPs of a, d, and the ECD of z1 are stabilised via polar interactions. This highly interlaced structure suggests a mutualistic contribution of each dimer to TCR-CD3 topology and cohesion. The intrinsically disordered intracellular tails of e, g, d and z, invisible in the cryo-EM structure, contain immunoreceptor tyrosine-based activation motifs (ITAMs) that become phosphorylated by constitutively active Lck kinase (Nika et al., 2010) within ≤ 1 sec after pMHC binding (Acuto et al., 2008;Huse et al., 2007). The CD3 tails are anchored to the plasma membrane (PM) via basic amino acid residues and ITAM tyrosines that interact with negatively charged lipids and hydrophobic interactions, respectively (DeFord-Watts et al., 2009;Xu et al., 2008), perhaps preventing ITAM phosphorylation of unliganded receptor by Lck. Early studies indicated that agonist anti-CD3 Ab induces exposure of CD3 cytoplasmic tails, presumably by conformational changes triggered by the Ab binding (Gil et al., 2002). However, crystallographic studies of pMHC bound to isolated TCRab ECD found considerable conformation changes in the CDRs Garcia et al., 2012) but no unambiguous or consistent changes beyond the TCRab binding site. This led to put forward signalling models independent of conformational changes or in which pMHC binding alone was insufficient to induce conformational changes of TCR-CD3. These models have proposed that pMHC-induced TCR-CD3 clustering (Cochran et al., 2001;Yokosuka et al., 2005), coreceptors (CD8/CD4) recruitment (Delon et al., 1998) or segregation of the tyrosine phosphatase CD45 (Davis and van der Merwe, 2006) initiated ITAMs phosphorylation and T cell activation. Alternatively, mechanosensing-based models have suggested that force generated by PM movements acts on pMHC-bound TCR-CD3 to induce conformational changes and signalling (Kim et al., 2009;Liu et al., 2014). Finally, it was proposed that clustering by pre-existing pMHC dimers drives conformational changes in CD3e, but not directly in TCRab (Gil et al., 2002;Minguet et al., 2007). Nevertheless, one crystal structure (Kjer-Nielsen et al., 2003) and a fluorescence-based study (Beddoe et al., 2009) provided evidence that pMHC binding induced a conformational change in a Ca loop. Moreover, deuterium-exchange  and recent NMR investigations (Natarajan et al., 2017;Rangarajan et al., 2018) have inferred changes in conformational dynamics of soluble TCRab ECD bound to pMHC. These changes mapped to where Ca and Cb interface with the ECD of the CD3 subunits (He et al., 2015;Natarajan et al., 2016). Although of great appeal, these studies do not rule in or out models proposed thus far; nor do they prove that allosteric effects propagate from ab to the CD3 subunits for signalling to occur. To challenge this impasse, we conceived a genetic perturbation analysis that should help discriminate first between models requiring or not molecular flexibility (i.e., conformational changes).
Towards this goal, we questioned the functional role of ab TMR, as a key portion of the entire complex establishing physical connection between the pMHC-binding module and the CD3 subunits governing signal delivery into the intracellular milieu. If TMRs are exclusively required for TCR-CD3 solvation within the lipid bilayer and for quaternary structure topology, mutations should not change TCR-CD3 intrinsic signalling capability. In contrast, this could happen in mechanisms based on allosteric interaction or force. We gathered compelling evidence for TMR mutations in TCRb and CD3z that slightly modified the quaternary structure cohesion and augmented intrinsic signalling output. We also found that cohesion changes in TCR-CD3 quaternary structure and signal transduction were induced by soluble monomeric agonist pMHC, independently of co-receptor, clustering or force. We propose that allosteric activation of the T cell antigen receptor by pMHC binding is the prime mover of T cell activation.

Gain-of-function mutations in b TMR
To question whether structural alterations in the TMR of the TCRab ligand-binding module affected signalling, we employed 1G4, an HLA-A2-restricted TCR specific for the 157-165 peptide from the NY-ESO-1 tumor antigen (Chen et al., 2005). Most residues of TCRb TMR were individually replaced by alanine or leucine and the corresponding mutants tested for reconstituting TCR-CD3 surface expression in the TCRb-deficient 31.13 Jurkat cell line (J31.13) (Fig. 1A). As reported earlier, bK287 mutation substantially reduced TCR-CD3 surface expression (Alcover et al., 1990). However, alanine substitution at bY281, bL285, bG286, bT289, bL290, bY291, bS296 and Leu at bA292 showed only a ≈ 20 % to 40 % decrease of surface expression. Next, the majority of mutants showing 0 % to 40 % reduction of surface expression were co-expressed together with WT 1G4 TCRa in J31.13 and Erk activation (pErk) was monitored after stimulation with 6V-HLA-A2 tetramer (6V-A2)4 ( Fig. 1B). While no mutation significantly reduced Erk activation, both bA290 and bA291 significantly increased pErk. A gain-of-function was unexpected, even more so as bA290 and bA291 reduced TCR-CD3 surface expression (the data in Fig. 1B are not normalised for TCR-CD3 surface expression).

bA291 heightens basal and ligand-induced signalling
To validate this apparently paradoxical observation, we focused on bA291 and modified the experimental set up to improve data robustness. Thus, a and b of 1G4 were expressed as a single self-cleavable polypeptide (Fig. S1A) from a doxycycline (dox)-inducible promoter in J76, a TCRab-deficient Jurkat cell line (STAR Methods). J76 expressed maximum levels of surface TCR-CD3 after 16-18 h of dox treatment and were tested soon after to reduce potential risk of phenotypic drift of cells expressing 1G4 carrying bA291 (hereafter, referred to as 1G4-bA291). As in 31.13 cells, 1G4-bA291 expressed in J76 showed reduced surface expression (≈ 30 %) (cf. Fig. 1A with Fig. S1B). However, in most experiments we lowered the dox concentration when inducing 1G4-WT in order to reduce the difference in surface expression with 1G4-bA291 (to < 5 %) (Fig. S1C). Moreover, in most flow-cytometry analyses, J76 expressing 1G4-WT or mutant were bar-coded by labelling with CellTrace™ violet, mixed before stimulation and analysed simultaneously. These stratagems considerably simplified and made more robust the computation of differences in signalling output between WT and mutant. Erk activation was retained as a sensitive and reliable read-out of TCR-CD3 signal transduction and propagation as it reports the occurrence of a cascade of early signalling steps, including ITAM phosphorylation, ZAP-70 activation, LAT signalosome assembly and PLCg1 activation that generates IP3 (for intracellular [Ca 2+ ] increase) and DAG required for Ras activation by Ras-GRP (Acuto et al., 2008). Titration of (6V-A2)4 showed a shift in pErk response by 1G4-bA291 towards higher sensitivity, but also revealed a significant higher Erk activation (Fig. 1C). This was not due to a higher Erk activation ceiling in 1G4-bA291expressing cells (Fig. S1D) nor to augmented binding of (6V-A2)4 to 1G4-bA291 (Fig. 1D, upper panel and lower panels), but it was consistent with the dose-response plot showing unchanged EC50 between 1G4-bA291 and 1G4-WT ( Fig. 1C and see Method for computation). The higher maximal response of 1G4-bA291 was compatible with a faster proofreading rate (kp) for a receptor operating in a kinetic proofreading regimen (McKeithan, 1995). Indeed, fitting the data of Fig. 1C into a minimal model of kinetic proofreading (Dushek et al., 2011) showed the kp for 1G4-bA291 was considerably higher than 1G4-bWT ( Fig. S1E and STAR Methods), consistent with bA291 enhancing TCR-CD3 intrinsic signalling capability (i.e., enhancing ligand potency). Note that the gain-of-function was observed in CD8-deficient J76 compared to CD8-deficient J76 expressing WT (Fig. 1C), ruling out that the bA291 mutation enhanced TCR-CD3 interaction with co-receptor. Augmented signalling was also evident for z phosphorylation (pz) (Figs. 1E and S1F), the earliest intracellular signalling event. Remarkably, anti-CD3e (UCHT1) Ab stimulation of 1G4-bA291 also heightened pz (Fig.  S1G), a triggering modality that by-passes pMHC binding, further supporting that bA291 enhanced TCR-CD3 signalling output. These data suggested that bA291 might increase constitutive TCR-CD3 signalling that can be detected by measuring pz in non-stimulated cells. Indeed, pz was significantly higher basally in cells expressing 1G4-bA291 as compared to 1G4-WT (Figs. 1F and S1H) and it was TCR-signal specific as it disappeared after treatment by A770041 (Stachlewitz et al., 2005), a potent and highly specific inhibitor of Lck (Fig. S1I).
We then asked if bA291 increased signalling by influencing TCR-CD3 lateral diffusion and/or distribution. However, fluorescence recovery after photo-bleaching (FRAP) found no significant difference in the diffusion coefficient (D) between 1G4-bA291 and 1G4-WT ( Fig.   1G, left panel), which remained unchanged after A770041 treatment (Fig. 1G, right panel).
dSTORM super-resolution microscopy found no statistically significant difference in the cluster size distribution formed by 1G4-bA291 and 1G4-WT (histograms in Fig. 1H). Although not statistically significant, the reproducible small increase of larger cluster frequency for 1G4-bA291 disappeared after A770041 treatment (cf. auto-correlation function plots in left and right panels of Fig. 1H), indicating it to be secondary to 1G4-bA291 heightened basal signalling (Fig. 1F), rather than bA291 causing it. Finally, we questioned the potential cause(s) of mildly reduced 1G4-bA291 surface expression. We excluded that bA291 reduced b protein expression (Fig. S1J) and considered that heightened basal signalling might decrease receptor surface expression by increasing its down-regulation rate. However, exposure to A770041 for several hours increased surface expression of both 1G4-bA291 and 1G4-WT in similar proportion (≈ 20 %) but did not significantly reduce their difference (Fig.   S1K). These data led us to consider if bA291 modified the stability of TCR-CD3 quaternary structure that could mildly reduce export to the PM due to increased negative triage of mutant vs. WT by protein quality control systems (Feige and Hendershot, 2013). bY291 contribution to TCR-CD3 quaternary structure cohesion Non-ionic detergents used at high concentration to quantitatively extract TCR-CD3 can dissociate TCRab from the CD3 modules (Testi et al., 1989). Presumably, this can be attributed to the substitution of natural boundary lipids by the detergent, with possible interference with TMR inter-helical interactions that are critical for TCR-CD3 quaternary structure cohesion (Alcover et al., 1990;Call et al., 2002;Dong et al., 2019). However, 0.5 % of the non-ionic detergent n-Dodecyl-b-D-Maltopyranoside (DDM) allows quantitative extraction of stoichiometrically intact TCR-CD3 (Swamy et al., 2008) (Fig. S2A). Thus, if bA291 altered TCRab cohesion with CD3 by unsettling TMR inter-helical interactions, 0.5 % DDM extraction may show lower recovery of intact 1G4-bA291 with respect to 1G4-WT.  Fig. 2A). b3 was the endo-H-sensitive ER-resident b isoform (Fig. S2C) that is assembled with a, ge, de but not with zz (Alcover et al., 2018), as confirmed by b3 being undetected in CD3z PD (Fig. S2D). b1 and b2 were both endo-Hresistant ( Fig. S2C), though b2 was the only b isoform associated with zz ( Fig. S2D). Thus, to evaluate the effect of bA291 on TCR-CD3 complex cohesion we used the IB signals for the b isoforms, zz and e (including ge and de). When z/b2 was set equal to 1 for 1G4-WT (i.e., 100 % recovery of intact TCR-CD3), reduced cohesion between zz and ab in 1G4-bA291 should result in z/b2 < 1 (Fig. S2B). The DSA showed that z/b2 for 1G4-bA291 was 0.2, indicating only 20 % recovery of intact TCR-CD3 (or 80 % loss of zz recovery) after DDM solubilisation ( Fig.   2A). To determine the effect of bA291 on ge and de cohesion with ab, we used instead the sum of b1, b2 and b3 (or total b (bT) IB signals that represented cytoplasmic and PM ab, most of which is associated with e (Alcover et al., 2018). e/bT for 1G4-bA291 was ≈ 0.5 indicating ≈ 50 % reduced recovery of e ( Fig. 2A). These ratios did not change after A770041 treatment during dox induction of ab expression (Fig. S2E), excluding that reduced recovery concerned the pool of 1G4-bA291 with increased z basal phosphorylation (Fig. 1F). The considerable reduction of z (80%) and e (50%) recovery for 1G4-bA291 could not be the consequence of severance of the same magnitude of ab from zz (or from de and ge dimers) before export to the PM and/or at the PM, when considering a mere 20-30% reduction of TCR-CD3 expression observed by flow cytometry. And, even more so compatible with increased pMHC-induced signalling. Indeed, when TCRab is no longer in contact zz, TCRabdege alone cannot be exported to the T cell surface ( Fig. S2F and (Alcover et al., 2018). Therefore, in the PM natural lipid environment, bA291 only slightly perturbed TCR-CD3 quaternary structure cohesion (as the MDS shows, see below), minimally reducing surface expression. However, substitution of natural boundary lipids by DDM severely corroded TCR-CD3 cohesion in 1G4-bA29 and provoked partial physical detachment of z and e from ab during the solubilisation.
IB for g and d revealed that bA291 affected both ge and de cohesion with the rest of the complex though asymmetrically, as it reduced ge and de recovery of 40 % and 10 %, respectively (Figs. 2B and 2C). In the cryo-EM structure, bY291 (note that Dong et al. (Dong et al., 2019) refer to bY291 as bY292) contacts mostly ge and therefore bA291 can be expected to affect primarily the interaction between ab and ge in accordance with the DSA.
However, bY291 makes no contacts with zz and de (Dong et al., 2019) (and see also MDS below). Therefore, the DSA revealed a more complex picture, with bA291 presumably affecting indirectly the interaction of both zz and de with the rest of the complex. To further understand the structural role of bA291, we used the TMRs' atomic coordinates of the cryo-EM structure of the TCR-CD3 octamer (PDB: 6JXR) (Dong et al., 2019) to carry out all-atom molecular dynamics simulations (MDS) with bWT and bA291 in an asymmetric lipid bilayer, mimicking the lipid environment of TCR-CD3 (see STAR Methods for details) and adding dynamical insight into TCR-CD3 cohesion. Simulations for 1250 ns confirmed considerable contacts of b WT with e (of ge), g, a and zz but not with de (Figs. 2D and S2G) and revealed one new contact of b with zz as well as significant reduction in six b contacts with e (ge), five with g and four with a (Fig. S2G). Specifically, during the simulations, significant contacts of bY291 with aN263, aT267, gL129, gG132, and eL145 were observed (Figs. 2D right panel and S2H) and also with gV133 and gI136, though not considered significant on the normalised scale (Fig. S2H). No contacts of bY291 with zz were seen (Fig. S2G). Simulations of the TMR octamer carrying bA291 indicated new and augmented contacts of b with e (ge) and g (Fig.   S2I). In addition, while bA291 still contacted gL129, it completely lost interaction with gG132, gV133 and gI136 (Fig. S2J, middle panel). Likely, these changes were secondary to spatial readjustments due to the loss of the bulky tyrosine side chain. No contacts of bA291 with zz were observed. Overall, the simulations suggested that bA291 reshuffled contacts with ge, with the net effect of increasing local compaction (Fig. S2K), as also indicated by a stabilisation of their a-helices crossing angle (Fig. S2L). This result seemed to contradict the DSA data of bA291 severely affecting zz interaction with the rest of the complex. Although 1250 ns time-scale is relatively long for all-atom simulations of membrane proteins, it might be insufficient to capture re-adjustments of interchain contacts that possibly occur at larger time-scales. bA291 might affect the role of interfacial lipids in cementing a-helices interactions (Gupta et al., 2017) that when challenged with DDM could cause crumbling of TMRs' cohesion in the mutant, despite augmented compaction by bA291 elsewhere.
However, reduced export to the PM was a good indicator that bA291 (and other b and zTMR mutants, see below) promoted some instability of the complex, causing dynamical exposure of hydrophobic site and/or retention signals, detected and negatively triaged by protein quality control systems (Feige and Hendershot, 2013). Comprehensively, these data suggested a positive correlation between reduced quaternary structure cohesion of TCR-CD3 and signal transduction activation.

Loosening z association enhances signalling
To corroborate this hypothesis, we investigated the phenotype of additional mutations in b and z TMRs. We found that similar to bA291, also bF291 and bL291 mildly reduced TCR-CD3 surface expression, despite no decrease in b expression (Fig. 3A). Both mutations reduced interaction of b with z and e (Fig. 3B) and augmented pErk maximal response to (6V-A2)4 ( Fig. 3C and 3D), whose binding remained unchanged (Fig. S3A). These three readouts ranked according to: bL291 ≥ bA291 > bF291 > WT, presumably reflecting conservative or non-conservative replacements, hence indicating a direct correlation between increased quaternary structure loosening and heightened signalling. We then tested the effect of bA291 in 2H5, an HLA-A2-restricted TCR specific for the MART-1 tumour antigen (MART-1 (Circosta et al., 2009). Similar to 1G4-bA291, 2H5-bA291 showed reduced surface expression bY291 did not contact z, but its mutation augmented basal (Fig. 1F) and ligand-induced z phosphorylation (Fig. 1E) and signal propagation. This was reminiscent of allosteric interaction revealed by mutations (Changeux and Christopoulos, 2016;Volkman et al., 2001) -e.g., mutations at bY291 induced local re-adjustments but also distal functional effects, such as favouring exposure of z cytosolic tail to active-Lck. To investigate this possibility, we tested whether mutations in z TMR residues susceptible to loosen zz contacts with subunits other than ab phenocopied mutations at bY291. TCR-CD3 cryo-EM structure and MDS indicated that z1 and z2 TMRs contacted only the N-terminal moiety of b TMR (Figs. 4A and S4A) and a TMR throughout (Figs. 4B and S4B). However, z2 and z1 contacted also g (Figs. 4C and S4C) and e (of the de) (Figs. 4D and S4D), respectively. Specifically, MDS revealed that z1I38 contacted two residues of e (of de) (Figs. 4D and S4E, left panel) and z2I41 contacted two residues of g (Figs. 4C and S4F, left panel), whereas z2I38 and z1I41 bulged toward the membrane lipids and made no contact with the complex. Thus, z1I38 and z2I41 were deemed capable of partially disturbing z1 and z2 interactions with e (of de) and g, but perhaps not with ab. To verify this prediction, 1250 ns all-atom simulations of TCR-CD3 octamer TMRs composed of z WT, zA41 and zA38 mutants were carried out. At the end of the simulations, alignment of snapshots of the mutated and WT TMRs showed distortion in the contacts of z1 with de (Fig. 4E) and z2 with g (Fig. 4G). As a consequence, zz containing z1A38 (3 out of 3 simulations) or z2A41 (2 out of 3 simulations) increased fluctuation relative to ab as compared to zz WT. This can be appreciated from the average spatial distribution plots of the Ca atoms of zz relative to the Ca atoms of ab that showed broader density for both mutants (Figs. 4F and 4H), though more pronounced for z1A38. These results were indicative of zA38 and zA41 increasing zz flexibility relative to ab. Both mutants maintained some zz contacts with the rest of the complex (Figs. S4G -S4N). These results prompted us to test if, similar to the bA291 mutations, also these z mutations showed reduced surface expression, complex cohesion by DSA and enhanced signalling. The data showed that zA38 or zA41 reduced 1G4 surface expression by » 30 %, for similar z expression (Fig. 5A). The DSA showed that zA38 and zA41 reduced z/b2 ratio to » 0.05 and » 0.25 (95 % and 75 % loss of z recovery), respectively, without apparently affecting e cohesion with ab. (Figs. 5B and   S5A). Thus, the DSA agreed with the loosening of zz interaction with ab, as predicted by the atomistic simulations. We surmise that by weakening direct interactions with CD3 TMRs and causing higher zz mobility, zA38 and zA41 indirectly loosened zz interaction also with ab, as revealed by the DDM extraction. Simulations of larger time-scales may provide clearer insights into dynamic alterations of zz by bA291 that presumably first produced a direct effect on ge and later on zz. Importantly, similar to bA291, both z mutations conferred to 1G4 heightened pErk response to (6V-A2)4, with a higher maximum compared to 1G4-WT (Figs. 5C and 5D) for equal (6V-A2)4 binding (Figs. S5B and S5C). We concluded that reduced cohesion between ab and zz caused heightened signalling, rather than the mutations of bY291 per se. This strengthened the idea that reducing TCR-CD3 cohesion populated the active signalling state of TCR-CD3 -i.e., it lowered the activation energy between two presumed functional states: inactive and active, the latter initiating transmembrane signalling. These data made unlikely that TCR-CD3 TMRs are just structural elements required for TCR-CD3 membrane solvation and architecture, as conformational changeindependent models would imply. Rather, by analogy with allosterically regulated proteins that can be switched on or off by mutations distal from their active site(s) (Changeux and Christopoulos, 2016;Volkman et al., 2001) and considering recent NMR studies (He et al., 2020;Natarajan et al., 2016;Natarajan et al., 2017;Rangarajan et al., 2018), our data suggested that pMHC binding could activate an allosteric cascade that loosened TCR-CD3 cohesion including TMR interactions with zz TMRs serving as a second-to-last relay before licencing z ITAM phosphorylation. These considerations prompted us to investigate this possibility. VaVb, almost indistinguishable from 1G4-WT (Fig. S6B). (9V-A2)4-His induced robust wtc51mediated Erk activation (Fig. 6A, left panel) and allowed specific capture of engaged wtc51 (Fig. 6A, middle panel, lanes 2 and 4) to be compared with unliganded wtc51 isolated by anti-HA b pull-down (Fig. 6A, middle panel, lanes 1 and 3). b2 was the only isoform bound to (9V-A2)4-His (Fig. 6A, middle panel, lanes 2 and 4), consistent with it being the only one associated to z and present at the cell surface (Fig. S2D). Therefore, z/b2 ratio was used to assess if pMHC binding had reduced cohesion of z within TCR-CD3 (Fig. 6A, right panel). The data showed that z/b2 in liganded wtc51 was 0.5 (Fig. 6A, right panel), in agreement with pMHC binding causing TCR-CD3 relaxation. pMHC-induced reduced cohesion of TCR-CD3 was independent of z phosphorylation, as identical results were obtained after A770041 treatment (Fig. 6A middle and right panels). Allosteric interaction typically occurs in the µs to few ms time-scale (Volkman et al., 2001), similar to the time required by pMHC binding to induce conformational changes in Cb loops (Natarajan et al., 2017). Because pMHC binding dwell-times are of a much longer time-scale (e.g., hundreds of ms to min), allostericallyinduced conformational changes should be observable at non-physiological lower temperatures. Consistently, almost identical reduction of z/b2 ratio was observed when (9V-A2)4 was reacted with cells at 0 o C (Fig. 6B middle and right panels). To exclude that our observations were biased by the particular mutations introduced in bCDR2 and/or by the non-physiological affinity of wtc51, we used QM-a TCR, a 1G4 variant carrying mutations in aCDR2, bCDR2 and bCDR3 (Fig. S6B) (Irving et al., 2012), which confer a 140 nM Kd (koff, 0.015 sec -1 at 25 o C) for NY-ESO-1157-165-HLA-A2 (Irving et al., 2012), which is within the physiological range of TCR-pMHC binding affinity (Aleksic et al., 2012;Cole et al., 2017;Cole et al., 2007;Stone et al., 2009). Molecular modelling showed that that QM-a and 1G4-WT have superimposable canonical orientation when bound to NY-ESO-1157-165-HLA-A2 (Fig.   S6B). Figure 6C showed that binding of (9V-A2)4-His to QM-a induced strong Erk activation ( Fig. 6C, left panel) and reduced z/b2 ratio (40 %), which remained unchanged after A770041 treatment (Fig. S6C) Binding of tetramerised ligand (6I-A2)4-His stimulated strong Erk activation (Fig. 6D, left panel) and weakened 868 quaternary structure cohesion, as shown by the reduced z/b2 ratio (Fig. 6D, middle and right panels). The occurrence of the same effect (i.e., structural changes) in three different TCRs by mere coincidence is highly unlikely but it is rather the consequence of the same cause: ligand-induced conformational changes that modify critical contacts maintaining TCR-CD3 complex cohesion (Alcover et al., 1990;Call et al., 2002;Dong et al., 2019). Reduced z cohesion was also observed in wtc51 when expressed in primary human T cells stimulated with (9V-A2)4-His (Fig. 6E), excluding non-physiological behaviour of TCR-CD3 in the PM of Jurkat cells. To date, it remains unclear whether anti-CD3e Abs used in clinical settings activate TCR-CD3 by mechanisms distinct from that of pMHC. To address this question, we slightly modified the DSA (STAR Methods). We employed mono-biotinylated Fab' of UCHT1 anti-CD3e as a proxy for minimally-or non-stimulated receptor and mono-biotinylated UCHT1 Ab to stimulate and capture TCR-CD3 with streptavidin for IB analysis. Since TCR-CD3 was captured via CD3e, b/e and z/e ratios were used to assess TCR-CD3 cohesion. We found that UCHT1 Ab binding reduced b/e and z/e ratios, hence the cohesion of e with b but less so with z (Fig. 6F). Similar observations were made if cells were pre-treated with A770041 (Figs. 6G and S6D) or reacted with UCHT1 at 0 o C (Figs. 6H and   S6E). Taken together, these observations and the TMR mutants' phenotype strongly suggested that TCR-CD3 signals intracellularly by an allosteric interaction propagating from the ab binding site to the CD3 subunits and modifying critical contacts within the TMRs.

Monovalent pMHC in solution triggers TCR-CD3 untying and intracellular signalling
pMHC tetramers induced conformational change and signalling without applying force.
However, pMHC tetramers necessarily induced fast TCR-CD3 clustering and therefore cannot allow to discern if receptor aggregation was responsible for allosteric activation, as previously suggested (Minguet et al., 2007). We therefore reacted wtc51-expressing J76 with biotinylated soluble monovalent (sm)-9V-A2 (Fig. 7A). Mono-dispersion of (sm)-9V-A2 was controlled by size-exclusion chromatography-multi-angle-light scattering (SEC-MALS), just before use (Fig. S7B, note that fractions within the sm-9V-A2 peak were used). Following DDM solubilisation, sm-9V-A2-bound TCR-CD3 was captured by His-Streptavidin/His-Cobalt beads (Fig. S7A) and z recovery examined. The data showed that z/b2 ratio was considerably reduced in sm-9V-A2-bound vs. unbound wtc51 and was unaffected by A770041 treatment (Fig. 7A, IBs and histograms) or by the absence of CD8 co-receptor (Fig. 7B). To definitively exclude potential sm-pMHC cross-linking after solubilisation by streptavidin used for capturing ligand-bound TCR-CD3, we used instead an Avidin monomer (mAv). However, this condition did not change the result (Fig. S7D). Similar z/b2 reduction was observed for 868 TCR reacted with monodispersed sm-6I-A2 (Figs. 7C for DSA and S7C for SEC). Figure 7E (IBs and histograms) shows that sm-9V-A2 reduced z recovery also in wtc51 expressed in primary T cells, excluding a bias of Jurkat cell PM. A more stringent test for TCR-CD3 allosteric regulation was to assess whether sm-pMHC promotes quaternary structure untying after solubilisation. Sm-9V-A2 bound to wtc51 TCR at 0 o C in post-nuclear lysates with 0.5 % DDM, as revealed by streptavidin IB (Fig. S7E), considerably reduced z/b2 ratio (Fig. 7F). This indicated that TCR-CD3 complex loosening by pMHC binding did not require intact PM and therefore, as it should be expected for an allosteric change, it relied essentially on proteinprotein interactions. Moreover, because it occurred in isolated TCR-CD3, these data further corroborated the idea that the allosteric change was independent of force, clustering and co-receptor. If the conformational change induced by sm-pMHC was functionally relevant, it should also induce intracellular signalling. Previous work could not demonstrate that binding of sm-pMHC in solution elicited [Ca 2+ ]i increase unless co-receptor was expressed (Delon et al., 1998). However, we found that sm-9V-A2, controlled by SEC-MALS for being monodispersed (Fig. S7B), did induce robust pErk in both CD8-efficient (Fig. 7A) and CD8-deficient ( Fig. 7B) J76 cells expressing wtc51, that was abolished by A770041 (Fig. 7A). Erk activation by sm-9V-A2 was dose-dependent (Fig. S7F), with as little as 3 nM inducing 50 % of the maximum and occurred at 2 min after sm-9V-A2 addition (Fig. S7G), similar to (9V-A2)4 stimulation of 1G4-WT (Paster et al., 2015), though (9V-A2)4 generally induced stronger pErk response. We obtained similar data with 868 in presence or absence of CD8 co-receptor (Figs. 7C, 7D and S7C) and with QM-a without CD8 (Fig. S7H). Non-specific adsorption of sm-pMHC onto J76 cell membrane during the stimulation assay was negligible even at the highest sm-9V-A2 concentration (Fig. S7I). This made unlikely that signalling by sm-9V-A2 was the consequence of non-specific adsorption to the cells surface resulting in cell-to-cell ligand cross-presentation rather than direct stimulation by soluble sm-9V-A2. Moreover, we experimentally tested whether even this negligible amount of non-specifically bound sm-9V-A2 on J76 cells could be stimulatory. However, we did not detect any Erk activation (Fig.  S7J). Multiple reasons can explain why our data apparently contradict previous observations. First and foremost, we used TCRs of reduced koff (higher-affinity range) for pMHC, including a natural one (868). sm-pMHC ligands of low-medium affinity range (µM) can be expected to induce low/non-sustained [Ca 2+ ]i increase, whose ramp-up requires a more robust and complex cascade of additional events (Irvine et al., 2002;Lewis, 2019), including co-receptor implication (Delon et al., 1998;Minguet et al., 2007). Also, sm-pMHC engages TCR-CD3 without immediately clustering it, contrary to pMHC tetramers that provide this critical signalling-reinforcing effect (see Discussion). Moreover, membrane-tethered pMHC has lower degree of freedom than soluble pMHC, a property that sensibly increases pMHC onrate (Huppa et al., 2010;O'Donoghue et al., 2013). Comprehensively, our genetic, biochemical, MDS and functional data constitute substantial evidence that TCR-CD3 is a genuine allosteric device. We name this model "TCR-CD3 allosteric relaxation" (Fig. S7K) as a mechanism sufficient to incite initial T cell activation solely by pMHC binding.

Discussion
Allostery governs signal transduction of many membrane receptors (Changeux and Christopoulos, 2016). However, this hard-wired, tuneable and fast interaction mode exploiting protein conformational flexibility has not gained sufficient traction for the elucidation of TCR-CD3 signalling mechanism . To gather insight into TCR-CD3 signalling mode, we used a genetic perturbation approach and uncovered a previously unnoticed allosteric property of the entire TCR-CD3 complex. We found that TMR mutations that loosened cohesion between TCRab and CD3z populated TCR-CD3 activated state and increased agonist's potency. This gain-of-function phenotype mimicked pMHC agonist binding that also reduced cohesion between TCRab and CD3z, independently of CD3 ITAM phosphorylation and at 0 o C. These convergent results suggested that weakening of TCR-CD3 TMR contacts is a key step in an allosteric mechanism initiated by pMHC binding and culminating in ITAM phosphorylation. We favour the idea that conformational changes occurring at the pMHC binding site propagate to CaCb ECDs, where they contact the CD3 subunits as indicated by several investigations (Beddoe et al., 2009;He et al., 2020;Natarajan et al., 2017;Rangarajan et al., 2018). The ECDs and TMRs of TCRab, CD3de and CD3ge show extended intra-dimer interface, yet less so between dimers (Dong et al., 2019), suggesting that each dimer retains flexibility vis-à-vis the other dimers. Moreover, CD3de and CD3ge interact much more with TCRab than with each other. It is therefore conceivable that pMHC binding can induce reshuffling of contacts between ECD dimers, making the CD3de and CD3ge acquiring a higher degree of freedom vis-à-vis TCRab. This may result in slight rotation and/or translation of CD3de and/or CD3ge vis-à-vis TCRab. The mechanical rigidity conferred to the CPs of e, d and g by the Cys-Cys loop (Alcover et al., 2018) could transmit these movements to the respective TMRs, resulting in local rearrangements of helix-helix packing, similar to those caused by the activating TMR mutations studied here and perhaps of interfacial lipids (Gupta et al., 2017) (Fig. S7K). Consistently, mutations of e Cys-Cys loop affect TCR-CD3 signalling (Wang et al., 2009). The relaxed quaternary structure of ligand-activated TCR-CD3 could reduce contacts between the TMRs of ab with zz (the most loosely attached dimer), making zz prone to detach from the rest of the complex, due to further erosion of TMR contacts by DDM during membrane extraction. As our data suggest, TMR quaternary structure relaxation activated by pMHC (or anti-CD3 Ab) or by TMR mutations promotes ITAM accessibility by active-Lck (Nika et al., 2010), which would require conformational changes of membrane-tethered CD3 intracellular tails (Xu et al., 2008). However, the membrane-juxtaposed segments of all CD3 subunits are intrinsically disordered, hence they may lack mechanical rigidity required to respond to TMR movements. We suggest instead that subtle untying of TMRs may indirectly reduce the grip of CD3 intracellular tails to the membrane and favour ITAM tyrosine exposure (Fig. S7K).
Phosphatidylserine (PS) (Xu et al., 2008) and PIP2 (Chouaki-Benmansour et al., 2018) are thought to keep the CD3 tails retracted onto the PM. An attractive possibility is that local TMR octamer rearrangement permits fast exchange of PIP2 and PS with neutral lipids that may reduce CD3z and e tails interaction with the lipid bilayer (Fig. S7K), gradually augmenting the exposure of ITAM tyrosines to active-Lck. Changes in cholesterol interacting with TMR helices (Swamy et al., 2016;Wang et al., 2016) and/or with the tyrosines of the ITAMs might be part of this mechanism. Agonist anti-CD3e mAb produced similar gain-offunction in TMR mutants as well as quaternary structure untying, in agreement with CD3e ECD lying on the conformational trajectory activated by pMHC binding.
A key finding of our investigation is the formal evidence that binding of soluble monovalent and mono-dispersed pMHC (sm-pMHC) alone to membrane-bound or detergent-solubilised TCR-CD3 suffices to induce TCR-CD3 quaternary structure relaxation and signal transduction.
Stimulation of TCR-CD3 by sm-pMHC alone agrees with a genuine allosteric mechanism, as hinted by our genetic perturbation analysis, much like membrane receptors activated by soluble ligands. Allosteric activation occurred without co-receptor or TCR-CD3 clustering or force, making extrinsic energy source, such as actomyosin-induced membrane movements required for mechanosensing (Das et al., 2015;Kim et al., 2009;Liu et al., 2014) dispensable for igniting TCR-CD3 signalling. A Kd of 7 µM was insufficient to analyse biochemically quaternary structure cohesion of 9V-A2-bound vs. free 1G4 TCR. However, we succeeded by using 1G4 variants wtc51 and QM-a capable of binding 9V-A2 with a Kd of 15 nM and 140 nM, respectively, the latter within the physiological Kd range for pMHC agonists (Aleksic et al., 2012;Cole et al., 2007;Stone et al., 2009). A third example was 868, a CTL-derived anti-HIV TCR, whose Kd was 53 nM, slightly higher than the lower limits of » 100 nM observed for naturally occurring pMHC-TCRs binding. This was due to a single natural mutation in the antigen peptide (6V ® 6I) that raised the Kd towards the unmutated epitope of 170 nM (Aleksic et al., 2012;Cole et al., 2017). All these TCRs showed ligand-induced quaternary structure loosening and intracellular signalling upon pMHC engagement. There is no evidence for high range affinity TCRs to induce T cell signalling and activation different from ligands of one-two-digit µM affinities and proven valid for T-cell adoptive immunotherapy.
Thus, it is unlikely that higher-than-normal affinity for pMHC confers to TCR-CD3 an allosterically regulated signalling while lower affinities do not. Indeed, allosteric changes in TCRab have been also demonstrated with pMHC binding with µM Kd (Beddoe et al., 2009;He et al., 2020;Natarajan et al., 2017;Rangarajan et al., 2018).
Early work could not detect [Ca 2+ ]i increase with sm-pMHC, unless CD8 was co-engaged (Delon et al., 1998). Others also did not observe TCR-CD3 signalling using pMHC monomer (Boniface et al., 1998). This apparent conflict with our data can be reconciled by considering differences in sensitivity of the signalling outputs measured (i.e., [Ca 2+ ]i vs. pErk) and increased ligand dwell-time in our experiments (in absence of co-receptor). Moreover, membrane-tethered pMHC considerably increases kon (and little or no increase of the koff) as compared to in solution measures (Huppa et al., 2010;O'Donoghue et al., 2013), suggesting more effective entropically-driven signalling by the former. Moreover, [Ca 2+ ]i rise of high amplitude and duration requires sustained TCRab engagement (Irvine et al., 2002;Lewis, 2019), likely achieved by higher lateral ordering of cell-surface and signalling complexes in micro-clusters and immunological synapse (IS) (Varma et al., 2006). The combination of these conditions (e.g., co-receptors and clustering), that sets in motion robust intracellular signalling mechanism (i.e. sustained signalling for full T-cell activation), may not be required for just igniting TCR-CD3 signal transduction as sm-pMHC alone does.
Mechanical forces play multiple roles in T cell activation at the molecular and cellular levels . Reduced koff of TCR-pMHC interaction is observed when subjected to ≈ 10 -20 pN pulling force, which means that "catch-bond" can be formed (Liu et al., 2014).
However, mechanical force (pulling/pushing) generated by membrane fluctuations and/or dedicated actin protrusions are observed in the time-scale of seconds, much slower (at least three order of magnitude) than allosteric changes propagating occurring in µs to ms (Kern and Zuiderweg, 2003;Natarajan et al., 2017;Rangarajan et al., 2018). Thus, pMHC-induced signal transduction and initial intracellular signalling could occur without the need of active mechanical force. It is also interesting that recent biophysical data suggest that force developed between membrane-tethered TCR-CD3 and pMHC is of fairly low magnitude (~ 2 pN) (Göhring et al., 2020). Perhaps, this low-amplitude force may play a role in extending the purely allosterically-induced interactions supported by our study.
Changes in conformational dynamics can have long-range consequences of functional relevance, a mechanism known as dynamic allostery (Tzeng and Kalodimos, 2012), which relies on changes in conformational entropy only. Conformational entropy cannot be frequently observed in a protein's crystalline state, unlikely to capture a protein higherenergy (activated) state. However, NMR can correlate very fast local conformational changes (ps, ns) occurring at distant sites over time-scales of µs to ms, compatible with allosteric regulation (Kern and Zuiderweg, 2003;Natarajan et al., 2017). Dynamic allostery may therefore apply to TCRab, as pMHC binding induces changes in conformational dynamics at distal H3, H4 helices and FG loop of Cb and Ca AB loop (Beddoe et al., 2009;He et al., 2020;Natarajan et al., 2017;Rangarajan et al., 2018), the latter having been captured only in a single crystal structure (Kjer-Nielsen et al., 2003) of many solved to date (http://atlas.wenglab.org/web/index.php. High affinity TCR-CD3-pMHC interacting by noncanonical orientation cannot induce signalling (Adams et al., 2011). In the light of our data, allosteric activation of TCR-CD3 occurs perhaps optimally only with ligands interacting in a diagonal canonical orientation.
Our data should help reconcile controversies about TCR-CD3 signalling mechanism. Thus, pMHC co-engagement by TCR and co-receptor has been found to be conditional on initial TCR-CD3 signalling (Casas et al., 2014;Jiang et al., 2011) and catch-bonding was contrasted by inhibiting Lck (Hong et al., 2018). We found that enhanced basal signalling by 1G4-bA291 induced weak but detectable clustering that was erased by Lck inhibition. These data suggest that all these events are instigated by an initial mechanism of allosteric nature induced solely by pMHC binding. Thus, co-receptor engagement, receptor clustering, shielding from PTPs and actomyosin-driven mechanical force (facilitating clustering and catch-bonding) may stabilise and potentiate initial allosterically-induced signalling. They may help reduce physical and chemical noise during receptor signalling, augmenting and stabilising signals of narrow amplitude and duration initiated by sparse engagement of pMHC monomers (Brameshuber et al., 2018).
The fast time-scale by which allosteric interaction propagates should ensure that ITAMs' exposure to active-Lck relies on pMHC binding dwell-time compatible with both very weak (self) and agonist ligands (e.g., hundreds of ms to sec). However, allosteric activation for receptors that recognise multiple ligands raises the possibility of "biased agonism" (Freed et al., 2017;Furness et al., 2016;Lane et al., 2017) whereby different ligand-induced conformational changes and/or ligand-binding kinetics correlate with distinct functional outputs. In principle, allosteric interaction activated by different TCRab CDR loop conformational changes upon canonical orientations over pMHC might follow different conformational trajectories propagating along TCR-CD3. In this case, ligand potency may result from both binding kinetics and variable conformational trajectories. Alternatively, binding of diverse pMHC-TCRab interfaces in canonical orientation may generate equivalent allosteric changes making ligand binding kinetics the unique factor governing ligand potency.
Future studies on TCR-CD3 using native nanodiscs, cryo-EM, MDS and genetic perturbation should help to further clarify these questions. We anticipate that the novel data for TCR-CD3 signalling reported here should spark interest for innovative strategies to harness TCR-CD3 signalling for immunotherapy. suggestions and for reading the manuscript; C.R. and J.C. and Ana Maria Vallés for manuscript editing. We apologise to colleagues whose work could not be adequately cited and commented herein because of space limitation.

Declaration of Interests
"The authors declare no competing interests"