T cell receptor reversed polarity recognition of a self-antigen major histocompatibility complex

Central to adaptive immunity is the interaction between the αβ T cell receptor (TCR) and peptide presented by the major histocompatibility complex (MHC) molecule. Presumably reflecting TCR-MHC bias and T cell signaling constraints, the TCR universally adopts a canonical polarity atop the MHC. We report the structures of two TCRs, derived from human induced T regulatory (iTreg) cells, complexed to an MHC class II molecule presenting a proinsulin-derived peptide. The ternary complexes revealed a 180° polarity reversal compared to all other TCR-peptide-MHC complex structures. Namely, the iTreg TCR α-chain and β-chain are overlaid with the α-chain and β-chain of MHC class II, respectively. Nevertheless, this TCR interaction elicited a peptide-reactive, MHC-restricted T cell signal. Thus TCRs are not 'hardwired' to interact with MHC molecules in a stereotypic manner to elicit a T cell signal, a finding that fundamentally challenges our understanding of TCR recognition.

1 1 5 4 VOLUME 16 NUMBER 11 NOVEMBER 2015 nature immunology A r t i c l e s is thought to reflect co-receptor-mediated T cell signaling steric constraints and a predisposed TCR bias towards the MHC 5,6 .
Although the factors governing the consensus TCR-pMHC polarity are unclear, this observation has nevertheless represented a key argument, which spans decades, in support of TCRs being 'hardwired' for the MHC. Here, the TCR is envisioned to be 'locked in' to bind onto the MHC molecule using conserved regions ('codons') 4,7,8 . Key to this theory is that the germline-encoded regions of the TCR bind to the MHC, whereas the hypervariable CDR3 loops bind predominantly to the peptide. Support for TCR-MHC bias has come from data showing that TCRs possessing a common TCR β-chain (Vβ8.2) dock on different MHC class I and class II molecules in a stereotypic manner, principally mediated by two tyrosine residues originating from the CDR2β loop 6,7,9,10 . However, the 'rules' underpinning evolutionarily based TCR-pMHC contacts are not always maintained. For example, variations in the peptide or in CDR3 usage can result in notably different binding footprints on the pMHC complex 2 . Thus, whether the TCR is hardwired for the MHC remains unresolved. Nevertheless, the consensus TCR-pMHC docking polarity is consistent with the MHC bias theory.
Little is known at present about the antigen specificity or the underlying TCR-pMHC interaction of T reg cells 11 . T reg cells can act through diverse suppressive mechanisms, including killing of antigen-presenting cells (APCs), production of inhibitory factors and modulation of dendritic cells (DCs), thereby generating enthusiasm for diverse clinical applications 12 . T reg cells can be thymus-derived (tT reg cells), peripherally derived (pT reg cells) or in vitro-induced T reg (iT reg cells) 13 . Here, we characterize a population of human proinsulin-reactive iT reg cells, which were derived from mature conventional naive T cells that were educated and selected in the thymus on the basis of self-MHC restriction. These T cells were activated to become proinsulin-reactive iT reg by tolerogenic dendritic cells pulsed with proinsulin peptide C19-A3 (GSLQPLALEGSLQKRGIV). Functionally and phenotypically these iT reg cells are indistinguishable from pT reg cells derived directly from the blood of healthy individuals 14 . Moreover, these iT reg cells are clinically relevant in that T reg cells reactive to human leukocyte antigen (HLA)-DR4 presenting the proinsulin peptide (HLA-DR4 proinsulin ) have been found in patients with late-onset type 1 diabetes and in healthy individuals with increased MHC-associated risk for this disease 15 . The combination of proinsulin peptide and tolerogenic dendritic cells is currently being assessed for clinical use in the treatment of type 1 diabetes.
We determined the structures of two iT reg TCR-pMHC complexes, which revealed a reversed TCR-pMHC docking topology. Accordingly, our findings, which stem from naturally selected TCRs, demonstrate a 180° reversal of this TCR-pMHC docking polarity. Despite this reversed docking topology, the TCR interaction nevertheless elicited a T cell signal in a peptide-reactive and MHC-restricted manner. Thus our findings directly challenge the TCR-MHC hardwiring theory and the requirement for geometric T cell signaling constraints, which collectively has major implications for T cell biology in general.

RESULTS
A proinsulin-reactive iT reg cell population T reg cells can be subdivided into various subsets, including tT reg , pT reg and iT reg cells 13 . Although tT reg cells are uniformly Foxp3 + CD25 hi , in humans these markers do not distinguish pT reg or iT reg cells from activated conventional CD4 + effector T cells due to a transient upregulation of both Foxp3 and CD25 upon TCR-dependent stimulation that is associated with a methylated region within the Foxp3 promoter (the T reg -specific demethylated region, TSDR). Human pT reg cells 14 and iT reg cells such as type 1 regulatory T cells (Tr1) 16 are defined by TCR activation-dependent secretion of high levels of IL-10 and varying amounts of interferon-γ (IFN-γ). The most reliable definition of human T reg cells is their functional ability to suppress T cell responses in an MHC-restricted manner. Antigen-reactive iT reg cells can be propagated in vitro by stimulation with self-antigen presented by tolerogenic dendritic cells (tDCs) 17 . Vitamin D 3 induces tolerogenic human DCs by altering a metabolic pathway that subsequently leads to the downregulation of HLA class II on the cell surface, and this suboptimal stimulatory capacity can then convert conventional naive T cells to iT reg cells 18 . We used HLA-DR4 proinsulin tDCs to generate iT reg cells reactive against a self-antigen. Vitamin D 3 -modulated tDCs induce antigen-specific iT reg cells that suppress both naive and effector T cells through their ability to modify the function of proinflammatory DCs and kill APCs through the action of granzyme B 17 . We cloned proinsulin-specific T reg cells induced by tDCs and investigated their function in relation to phenotype, the mechanisms of suppression used by different iT reg cells, and defining markers that act as immune correlates of inhibitory function of antigen-specific regulatory responses ( Fig. 1 17,19 with a methylated TSDR within the FOXP3 gene locus, a property that distinguished these from tT reg cells but was reminiscent of other antigen-specific T reg cells types, such as Tr1 and pT reg cells 14,16 . Principal-component analysis of the variation in surface phenotype clearly segregated iT reg clones from proinsulin reactive effector T cells (T eff ) (Fig. 1a). iT reg cells showed lower surface expression of CD4, αβTCR, CD28, CD62L and CD25 and higher expression of the T reg cell-associated surface marker GITR (Supplementary Fig. 2a).
We next investigated functional capacities of proinsulin-reactive iT reg clones. Nine iT reg clones inhibited proliferation of naive CD4 + T cells (Fig. 1b). Of these, seven clones also suppressed the proliferation of autoreactive type 1 T helper (T H 1) cells specific for a diabetogenic glutamic acid decarboxylase (GAD)-derived peptide in an antigen-dependent manner (Fig. 1c). Stimulation of iT reg clones with increasing concentrations of proinsulin peptide produced a dose-dependent production of IL-10 ( Supplementary Fig. 2c). iT reg clones killed monocytes, albeit with variable efficiency, similarly to islet autoantigen-specific T reg cells isolated from blood and IL-10-induced antigen-specific T reg cells 14,16 (Fig. 1d,e). Unsupervised clustering of the intracellular expression of molecules associated with regulatory function (granzyme B, CTLA-4, IFN-γ and IL-10) and cytokine production segregated clones into three groups, one of which closely resembled previously described Tr1 cells defined by high cytokine secretion capacity (IL-10, IFN-γ, TNF and IL-13) (Fig. 1e, left) 14,16 . Two clones from the Tr1-like cluster (FS17 and FS18) displayed a proinsulin-reactive proliferative response and IL-10 production (Supplementary Fig. 2b,c), suppression of naive T cells with differential capacity to suppress T H 1 cells in an antigen-dependent fashion and a good capacity to kill proinsulin-loaded monocytes ( Fig. 1b-d). Accordingly, the clones described here are bona fide T reg cells, derived from mature human T cells that have been selected in vivo and drawn from the natural T cell repertoire.
Reversed-polarity TCR-self MHC class II recognition Next, we aimed to understand how TCRs from iT reg clones interacted with the HLA-DR4 proinsulin complex. To gain insight into the Tr1-like cluster, we expressed, refolded and purified the TCRs from the clones FS17 and FS18. Notably, the FS17 and FS18 TCRs use the same TRAV and TRBV gene segments (TRAV29 and TRBV6-2) npg (Supplementary Table 1). Both the FS17 and FS18 TCRs showed chromatographic properties very similar to those of other purified TCRs and reacted with a conformation-specific anti-TCR monoclonal antibody (data not shown). We generated the HLA-DR4 proinsulin complex by transient overexpression in N-acetylglucosaminyltransferase I (GnTI)-deficient HEK293S cells, purified it and determined it to be correctly folded by its chromatographic profile and reactivity towards an anti-HLA-DR monoclonal antibody (data not shown). We then crystallized the FS17 and FS18 TCR-HLA-DR4 proinsulin ternary complexes. We determined the FS18 TCR-HLA-DR4 proinsulin structure to 2.5 Å resolution (R fac and R free of 16.0% and 19.7%, respectively) ( Table 1). We also determined the structure of the FS17 TCR-HLA-DR4 proinsulin ternary complex to 4.0 Å resolution. Although the FS17 TCR ternary complex was of moderate resolution, it refined well (R fac and R free of 22.5% and 28.3%, respectively) and unambiguously allowed us to ascertain the TCR-pMHC docking topology in a manner similar to that of the initial low-resolution TCR-pMHC class I complex 20 . Unless otherwise stated, the detailed structural analyses will be confined to the higher resolution FS18 TCR-HLA-DR4 proinsulin ternary complex.
The electron density for the entire FS18 TCR ternary complex, and in particular at the TCR-pMHC interface, was very clear ( Supplementary  Figs. 3 and 4), thereby permitting detailed structural analysis of the FS18 TCR-HLA-DR4 proinsulin interaction (Supplementary Table 2). Without exception, in all previous studies the TCR has been universally observed to adopt a consensus polarity above the pMHC 2 ( Fig. 2a-d). Here, the TCR α-chain and TCR β-chain were invariably positioned over the α2-helix and α1-helix of MHC class I or the β-chain and α-chain of MHC class II, respectively (Fig. 2a). However, the FS18 and FS17 TCRs docked with the TCR α-chain positioned over the HLA-DR4 α-chain, whereas the TCR β-chain was located over the HLA-DR4 β-chain (Fig. 2a,e-g). The mode of FS18 TCR-pMHC class II docking was very similar to that of the FS17 TCR ternary complex (Supplementary Fig. 5). Thus, the orientation of the FS18 and FS17 TCRs are reversed through 180° compared to all previously determined ternary TCR-pMHC complexes. Given that the previous >120 TCR-pMHC structures determined adopt the consensus    iT reg TCR-self peptide MHC interactions The FS18 and FS17 TCRs docked at 80° across the long axis of the antigen-binding cleft, and most unusually, was tilted 60° relative to the HLA-DR4, leaving the TCR α-chain remotely positioned from the interface (Fig. 2e,f). Indeed, the TCR α-chain did not contact HLA-DR4 proinsulin . Accordingly, the FS18 and FS17 TCRs made limited contacts with the HLA-DR4 molecule (Fig. 2e, In relation to the FS18 TCR ternary complex, the TCR β-chain solely contacted HLA-DR4 proinsulin , ligating to a few residues on the HLA-DR4α chain and HLA-DR4β chain (Fig. 2e,g and Supplementary  Table 2). Here, the CDR1β, CDR2β and CDR3β loops contributed 5%, 33% and 11% of the BSA at the interface, respectively. The CDR3β loop BSA contribution represents the lowest reported value compared to all other TCR-pMHC class II ternary structures (median value 28.6% BSA). Conversely, the contribution of the CDR2β loop to the BSA (33%) represented the highest value as compared to all other TCR-pMHC class II complexes (median value 10.2% BSA) 2 . Notably, because of the 60° tilt of the FS18 TCR atop HLA-DR4 proinsulin , 49% of the BSA arose from the TRBV gene-encoded framework region. Thus, atypical relative contributions from the CDRβ loops and the TCR β-chain framework regions underpin the FS18 TCR interaction with HLA-DR4 proinsulin (Fig. 3a-d).
Interactions through the CDR1β loop were limited to the aliphatic moiety of Glu37β packing against the aliphatic moiety of Gln57α from HLA-DR4 and the side chain of Gln57α packing against the side chain of Tyr38β (Fig. 3b). This germline-encoded loop was therefore not ideally disposed to ligate HLA-DR4. Val57β of the CDR2β loop formed van der Waals contacts with Ala61α, whereas its main chain hydrogen-bonded to Gln57α from HLA-DR4 (Fig. 3b). Gly64β and Glu63β were principally involved in peptide-mediated contacts (discussed below), with Glu63β also packing against Glu55α and Gly58α of HLA-DR4 (Supplementary Table 2). Thr65β of the CDR2β loop made van der Waals contacts with Ala61α and Asn62α of HLA-DR4 (Supplementary Table 2). The solvent-exposed CDR3β loop was located at the periphery of the HLA-DR4 α-chain, whereupon the non-germline-encoded Arg109β pointed toward, and contacted, Gln57α and Lys39α (Fig. 3b). The TRBV framework-mediated contacts with HLA-DR4 were dispersed throughout the TCR β-chain and located at a number of discrete sites. Many of these frameworkderived contact regions have not been observed in previous TCR-pMHC complexes 2 . Namely, Lys68β hydrogen-bonded to Gln64β from HLA-DR4, whereas Asn77β slotted between Gln70β and Asp66β of HLA-DR4 (Fig. 3c). Lys84β and Glu92β formed salt bridges with Glu55α and Lys65β from HLA-DR4, respectively (Fig. 3c). In addition to the 180° reversal in docking geometry, many unique features therefore distinguish the FS18 TCR-HLA-DR4 interaction from canonical TCR-pMHC class II interactions 2 .
The proinsulin peptide contributed 26% of the BSA at the interface, with the FS18 TCR contacting the N-terminal region of the epitope, namely positions P1, P3 and P5 (Fig. 3d). These interactions were mediated through the CDR2β loop and a TCR β-chain framework residue ( Fig. 3d and Supplementary Table 2). However, the peptide interactions seemed to be suboptimal and nonspecific, with Lys83β packing against P-1-Pro and the remainder of the contacts involving main-chain interactions from either the peptide or the TCR. Namely, Glu63β hydrogen-bonded to the main chain of P1-Leu, whereas Gly64β packed against the backbone of P3-Leu and P5-Gly (Fig. 3d). The lack of extensive and specific contacts with the proinsulin peptide was consistent with the HLA-DR4-restricted autoreactivity of the FS17 and FS18 iT reg cells, as described below. Thus atypical interactions underpinned this TCR-pHLA interaction.
Same gene usage equates to differing TCR-pMHC docking modes The variable (V) gene usage of the FS18 and FS17 TCRs has been observed previously in TCR-pMHC class I crystal structures. Namely, the B7 TCR (TRAV29-TRBV6-5*01) is restricted to HLA-A2 for all data except as indicated in footnote c. c 5% of data was used for the R free calculation.
Values in parentheses refer to the highest-resolution bin.

T cell signaling
To determine whether the presentation of the proinsulin peptide to FS17 and FS18 T cells in the context of HLA-DR4 caused TCR signaling comparable to that achieved with a canonical TCR-pMHC interaction, we retrovirally transduced SKW3 cells with the FS17 TCR, FS18 TCR and a control SP3.4 TCR (SKW3.FS17, SKW3. FS18 and SKW3.SP3.4, respectively). The SP3.4 TCR is restricted to HLA-DQ8 presenting a gliadin-derived determinant (DQ8 glia-α1 ) 29 . We incubated the SKW3.FS17, SKW3.FS18 and SKW3.SP3.4 cells overnight with their corresponding cognate peptides loaded on BLCL9031 (HLA-DR4 + DQ8 + ) APCs. The SKW3.SP3.4 clone showed a specific HLA-DQ8 glia-α1 dose-dependent upregulation of CD69 (a marker of T cell activation) in response to peptide, which was blocked by an anti-HLA-DQ but not by an anti-HLA-DR monoclonal antibody (Supplementary Fig. 6a). The SKW3.FS17 and SKW3.FS18 clones showed similar activation in the presence of the BLCL9031 APCs, independent of the addition of proinsulin peptide ( Supplementary  Figs. 6b and 7). Stimulation still occurred in an HLA-DR4-restricted manner, as evidenced by blocking experiments with anti-HLA-DR and anti-MHC class II monoclonal antibodies ( Supplementary  Figs. 6b and 7). This observation suggests that the FS17 and FS18 TCRs show inherent autoreactivity towards HLA-DR4. We therefore decided to look at earlier markers of activation given that the upregulation of CD69 on the SKW3 cells in the absence of peptide might have resulted from extended culture in the presence of a weakly cross-reactive HLA-DR4-peptide complex. Src-family kinases (SFKs) are activated when the TCR engages pMHC and allow for CD3 phosphorylation and downstream mitogen-activated protein kinase Erk1/2 signaling 30 . Accordingly, we assessed phosphorylation and activation npg A r t i c l e s of SFKs, CD3 and Erk1/2 by monitoring for Src Tyr418, CD3ζ Tyr142 and Erk1/2 Thr202 and Tyr204 phosphorylation in SKW3.FS17 or SKW3.FS18 cells incubated with cognate peptide loaded onto either BLCL9031 APCs (Fig. 4a,b) or human HLA-DR4 + DCs (Fig. 4c,d). Incubation of SKW3.FS17 cells or SKW3.FS18 cells with BLCL9031 APCs (Fig. 4a,b), or of SKW3.FS18 cells with human HLA-DR4 + DCs (Fig. 4c,d), resulted in TCR signaling that was enhanced by the proinsulin peptide. After prolonged exposure, TCR signaling in the absence of peptide was maximal (Fig. 4b,c). Much as we observed with CD69 expression, FS18 TCR signaling in the presence or absence of cognate peptide was HLA-DR4 restricted, as demonstrated by its suppression with an anti-HLA-DR monoclonal antibody (Fig. 4b-d). TCR signaling in the absence of cognate peptide ligand was not seen with SKW3.SP3.4 cells or SKW3 cells bearing a TCR specific for the hemagglutinin (HA) peptide (Supplementary Fig. 8). Thus, despite the reversed FS17 and FS18 TCR-pMHC class II docking mode, engagement of the FS17 and FS18 TCRs elicited a T cell signal in an HLA-DR4-autoreactive and proinsulin-reactive manner.

Self-peptide and TCR mutagenesis
As the reversed TCR docking topology was unexpected, we validated the docking mode through targeted mutagenesis of the self-peptide and the TCR using the crystal structure as a guide. Moreover, given the extent of HLA-DR4 autoreactivity exhibited by the FS18 and FS17 iT reg clones, and the observation that the FS18 TCR and FS17 TCRs made limited contacts with the proinsulin peptide, we also established the relative importance of the TCR-peptide interactions in enabling a productive T cell signal. Specifically, we created a series of mutants of the proinsulin epitope and examined the effects of these mutants on SKW3.FS18 and SKW3.FS17 activation (Fig. 5a,b). As expected, mutation of a peptide residue (Gln8Ala) that did not contact the FS18 TCR or FS17 TCR had no impact on the T cell signaling. Next we created mutants (Leu3Trp and Gly5Arg) of peptide residues that do contact the FS18 and FS17 TCRs. These single-site bulky mutants reduced FS18 and FS17 T cell activation, and the double mutants (L3R-G5R and L3W-G5R) caused a further reduction in T cell signaling. The same responsiveness of the FS18 and FS17 TCRs to the peptide mutants highlights the common proinsulin reactivity. Moreover, the impact of these peptide mutants further underscored the common reversed docking topology of these two TCRs. Next, we assessed the role of the TCR residues in mediating the interaction with HLA-DR4 proinsulin , naturally focusing on TCR β-chain residues involved in the interaction. It should be noted, however, that although the FS17 and FS18 TCRs share the same TRAV and TRBV gene usage, they differ in their CDR3α and CDR3β loops (Supplementary Table 1). For example, the CDR3α loop differs in two positions, including a non-conservative tyrosine (FS18 TCR) to serine (FS17 TCR) substitution. Thus, a naturally occurring variation within the CDR3α loop does not affect the interaction with HLA-DR4 proinsulin , as would be expected on the basis of the FS17 and FS18 TCR ternary complexes. Next, we created seven single-site mutants within the TCR β-chain (Lys68βAla, Asn77βAla, Leu81βAla, Lys83βAla, Lys84βAla, Glu92βAla, Arg109βAla) and stably transduced them into the SKW3 T cell line, matching TCR cell surface expression levels ( Supplementary  Fig. 9). We examined the impact of these TCR β-chain mutants on CD69 upregulation when co-incubated with BLCL9031 APCs in the absence and presence of the proinsulin peptide. Leu81βAla was chosen as a control mutant because Leu81 did not contact HLA-DR4, and, as expected, this mutant did not affect T cell activation. Notably, the effect of the mutations were very similar in the presence or absence of the proinsulin peptide, thereby indicating that a common docking mode underpins the TCR interaction towards HLA-DR4 when bound to an endogenous peptide(s) or the proinsulin peptide. Not all residues observed in the TCR-HLA-DR4 interaction, including residues mediating polar contacts, were energetically critical for the interaction. This observation is in line with previous mutagenesis studies that have shown that although the TCR-MHC, TCR-CD1 and TCR-MR1 interfaces can be extensive and involve a large number of polar and nonpolar contacts, only a few key contact points ('hot spots') can underpin the interaction 2 . Here, mutations at positions 68, 77, 83 and 92 did not affect T cell signaling, whereas mutations at Lys84 and Arg109 markedly reduced CD69 upregulation (Fig. 5c). Lys84β is a framework-encoded residue that formed an intimate salt bridge with Glu55 from the HLA-DR4 α-chain, whereas Arg109β is within the CDR3β loop and was observed to pack against Lys39 and Gln57 as well as to hydrogen bond to Gln57 from the HLA-DR4 α-chain. Indeed, both the Lys84 and Arg109 FS18 TCR residues converge upon a similar spot on the HLA-DR4 molecule (namely Gln57), revealing that this region of the TCR and the HLA-DR4 molecule defines the 'energetic hot spot' for the interaction (Fig. 5d). Accordingly, our peptide mutational data, the natural variability in the FS17 and FS18 TCRs, and our TCR β-chain mutagenesis data are npg fully consistent with our observed reversed TCR docking topology.

DISCUSSION
In MHC-restricted immunity, a universal consensus TCR-pMHC docking polarity has been observed in the more than 120 TCR-pMHC structures determined to date. Specifically, the Vα chain and Vβ chains of the TCR are invariably positioned over the α2-helix and α1-helix of MHC class I, respectively (and the β-chain and α-chain of MHC class II, respectively) 2 . Although some autoreactive TCR-pMHC class II complexes and 'super-bulged' TCR-pMHC class I complexes have adopted unusual topologies 31,32 , they nevertheless conform to the universal TCR-pMHC polarity. A key question has been: what drives this consensus TCR-pMHC docking topology? One likely answer, which is a view widely held, is that this consensus docking mode represents a primordial feature underpinning the inherent TCR-MHC bias 6,8,33 . From this consensus TCR-pMHC docking polarity has arisen the hypothesis of 'interaction codons' between the TCR and the MHC, whereby specific germline-encoded regions of the TCR are thought to interact with specific regions of a given MHC allotype 4 . Our findings directly challenge the TCR-MHC hardwiring theory and favor an alternative paradigm of TCR recognition. Taking a broader view, it is pertinent to reflect that the αβ T cell repertoire is vast, and that this diversity has enabled αβ TCRs to recognize peptides, lipids and vitamin B metabolites through diverse docking strategies 2 . Indeed, the presence of non-MHC-restricted TCRs in co-receptordeficient mice, suggests that the TCR is not inherently configured to specifically interact with MHC 34,35 . Our work, from two naturally selected TCRs, demonstrates, for the first time, an 180° reversal of the TCR-pMHC docking polarity. Clearly it remains to be determined whether unusual TCR-pMHC docking modes will typify other T reg cell TCR interactions or be associated with T reg development. Nevertheless, it is intriguing that of many unique TCR-pMHC complexes determined 2 , the first example of a reversed TCR docking polarity has arisen from TCRs that are expressed on the surface of iT reg cells. This represents a less than 1/10,000 (P < 0.0001) possibility of these unusual iT reg TCR-pMHC docking modes occurring by chance. Although FS17 and the FS18 TCRs are similar in sequence, the FS17 and FS18 T cell clones were independently derived and exhibit different functional profiles, and thus they should be considered independent. Even taking the FS18 TCR ternary complex in isolation, the probability of this iT reg TCR adopting a reversed docking topology by chance is P < 0.0082. Could the topology of the TCRs isolated from iT reg cells somehow be 'flipped' when in a naive T cell, or perhaps be attributable to the methodology used in generating the iT reg cell? Clearly the answer to this question is no. The docking mode of a TCR must be an intrinsic feature of the TCR and independent of the cell type it finds itself in. Notably, these iT reg cells TCRs shared germline-encoded TRAV and TRBV elements that have been observed previously in TCR-pMHC class I complex structures, but they nevertheless adopted a reversed docking topology. Thus, from TCR gene usage considerations alone, one cannot predict the TCR-pMHC docking topology, and we speculate that the naive T cell repertoire possesses a wide range of TCR-pMHC docking modes.  Many TCRs, including those arising from autoreactive and tumor antigen-specific T cells, can possess extremely low affinity for their pMHC class II complex, with the affinity value falling at the cusp of, or below, the threshold of sensitivity for the SPR-based measurements 2,22 . Thus, our observation that the TCRs described here do not have a measurable affinity via SPR is not exceptional and is fully consistent with the very limited TCR-pMHC class II interface. We chose to examine TCR docking topologies from the human, antigen-specific iT reg cell described here because at present the antigen specificity of any tT reg cell has not been established in a natural, authentic setting 36 , therefore making it impossible, at this stage, to evaluate the tT reg TCR-MHC docking mode. Although there are a number of models regarding T reg cell development (for example, low dose of a high-affinity ligand), we are not aware of any data that support this directly (such as T reg cell TCR interactions measured via SPR against a defined pMHC complex). We speculate that induction of T reg cells may be a feature more of the reversed docking mode and ensuing signals than of affinity-based considerations. In line with this, previous findings 37,38 suggest that T reg cell TCRs have different signaling requirements compared to effector TCRs. Thus the reversed docking topology might be a potential mechanism underpinning iT reg fate, or perhaps be a common feature of iT reg and tT reg cell TCR interactions, although many more structural and functional studies will be required to test this hypothesis.
The consensus TCR-pMHC docking topology has also been suggested to reflect the geometric constraints of productive TCR signaling 5,6 . However, our work shows that TCR signaling is much more permissive than previously considered and that docking geometry alone cannot dictate signaling outcome. Indeed, the factors that enable productive signaling upon TCR engagement remain poorly understood and will ultimately require TCR-CD3 structural studies 39 to fully appreciate this critical signaling event.
Our primary observation suggests that the T cell repertoire expresses an array of functionally competent TCRs capable of docking in radically different ways with pMHC-antigen complexes. We provide the first insight into how a TCR adopted a reversed polarity docking mode yet nevertheless transmitted a productive T cell signal. Our findings challenge a longstanding dogma, namely the predisposed bias of the TCR for the MHC. Accordingly, our findings represent a paradigm shift in thinking about T cell biology and should hopefully serve as a catalyst for further research in the field.

METHODS
Methods and any associated references are available in the online version of the paper. FS18 mutants and peptide mutants with an effect on SKW3 cell activation and/or signaling are plotted in red (FS18β-K84A, FS18β-R109A, proinsulin pL3W and proinsulin pG5R), and mutants with no effect are plotted in blue (FS18β-K68A, FS18β-N77A, FS18β-L81A, FS18β-L83A, FS18β-E92A and proinsulin pQ8A). FS18 contact residues that were not tested are plotted in yellow. npg