αβ T cell receptor germline CDR regions moderate contact with MHC ligands and regulate peptide cross-reactivity

αβ T cells respond to peptide epitopes presented by major histocompatibility complex (MHC) molecules. The role of T cell receptor (TCR) germline complementarity determining regions (CDR1 and 2) in MHC restriction is not well understood. Here, we examine T cell development, MHC restriction and antigen recognition where germline CDR loop structure has been modified by multiple glycine/alanine substitutions. Surprisingly, loss of germline structure increases TCR engagement with MHC ligands leading to excessive loss of immature thymocytes. MHC restriction is, however, strictly maintained. The peripheral T cell repertoire is affected similarly, exhibiting elevated cross-reactivity to foreign peptides. Our findings are consistent with germline TCR structure optimising T cell cross-reactivity and immunity by moderating engagement with MHC ligands. This strategy may operate alongside co-receptor imposed MHC restriction, freeing germline TCR structure to adopt this novel role in the TCR-MHC interface.


The TCR Adopts an antibody-like strategy for ligand recognition
To directly establish whether the TCR uses an antibody-like strategy of ligand recognition, we produced retrogenic mice expressing chimeric TCR/Ig chains composed of TRBV16 (Vβ 11) containing the germline CDR1 and 2 regions from a heavy (H) or light (H) chain immunoglobulin V segment. The IgH and IgL germline CDRs differ in length and composition from TRBV16 and introduce 19 and 15 amino-acid changes respectively (Fig. 1a). To minimise the role of TCR-β CDR3 in ligand engagement, it was reduced to triple glycine in all constructs used in this work (Figs 1a and 2a) 9,10 . Retrogenic mice were produced in FVB/N (H2 q ) TCR β /δ KO mice making T cell development dependent on the exogenous chimeric β chain 11 . Both chimeric chains were able to direct the development of mature thymocytes and peripheral T cell compartments (Fig. 1b,c) consistent with the TCR adopting an antibody-like strategy for ligand recognition 9 . Further, these findings support the hypothesis that the CD4 and CD8 co-receptors play a key role in imprinting MHC restriction and suggest that the germline CDR regions may play a secondary role in 'fine-tuning' TCR engagement with MHC:peptide ligands.

Germline TCR structure restrains ligand engagement during thymic T cell development
To explore this idea, we produced a transgenic strain expressing a mutant TRBV16 chain in which germline structure was 'simplified' by introducing alanine/glycine substitutions across CDR1 and CDR2 (canonical structure group β CDR1.2/β CDR2.2 12 ) ( Fig. 2a; Δ β CDR1-3). Additionally, the semi-conserved Y56 at the start of CDR2, which plays a role in controlling MHC engagement 7 , was deleted (Fig. 2a). As for the retrogenic mice, to minimise the role of TCR-β CDR3 in ligand engagement, it was reduced to triple glycine (Fig. 2a). Founder Δ β CDR1-3 and control TRBV16 with mutated CDR3 only (Δ β CDR3; Fig. 2a) transgenic lines were also produced directly in the FVB/N (H2 q ) TCR β /δ KO strain. The remainder of this paper will focus on these two transgenic strains.
Strikingly, thymic cellularity of the Δ β CDR1-3 transgenic mice was significantly reduced in comparison to the control Δ CDR3 transgenic (Fig. 2b). The numbers of CD4 + CD8 + double positive (DP) thymocytes and single positive (SP)4 and SP8 thymocytes were also significantly reduced in the Δ β CDR1-3 transgenic (Fig. 2c,d). The Δ β CDR1-3 phenotype is consistent with intensified ligand engagement across the DP pre-selection thymocyte repertoire where the majority of negative selection occurs 13 . This hypothesis was examined by looking at markers of TCR engagement. Expression of the negative regulator of TCR signalling CD5 on DP thymocytes parallels the intensity of interaction between the TCR and MHC 14 providing a quantitative readout of MHC engagement. CD5 expression on Δ β CDR1-3 DPs was significantly higher than in the control Δ CDR3 transgenic (Fig. 3a). Similarly, up-modulation of the TCR itself is indicative of competence to undergo positive selection 15 . Again, the Δ β CDR1-3 mutant had an increased proportion of TCR hi DP thymocytes in comparison with the Δ CDR3 control transgenic (Fig. 3b). TCR mean fluorescence intensity (MFI) was comparable in TCR lo cells from Δ β CDR3 and Δ β CDR1-3 mice (Fig. 3c). The activation marker CD69 is induced on DP thymocytes undergoing positive or negative selection 16 and was also expressed on an increased proportion of Δ β CDR1-3 DP thymocytes ( Supplementary Fig. S1). We conclude that the TCR repertoire of Δ β CDR1-3 mutant DP thymocytes has enhanced affinity for self-ligands identifying an unexpected role for germline TCR structure in restraining engagement with MHC ligands.

MHC restriction is imposed independently of germline TCR structure
Wild-type TCRs can engage with MHC and non-MHC ligands although, in the presence of the co-receptors, the latter are generally non-productive interactions 4,5 . We considered that loss of germline TCR structure could also promote recognition of non-MHC ligands, perhaps overriding co-receptor imposed MHC restriction and contributing to functional ligand engagement by the Δ β CDR1-3 TCR repertoire. To identify any such productive, non-MHC ligand engagement, the Δ β CDR1-3 mutant was crossed onto MHC deficient backgrounds. As previously shown 17 , in the absence of both MHC class I and II, few SP4/8 thymocytes (Fig. 4a,b) and peripheral CD4/8 T cells (Fig. 5a,b) are selected from wild-type TCR repertoires. Identical phenotypes were seen when the Δ β CDR1-3 mutant TCR-β was co-expressed (Figs 4b,c and 5b,c). In the absence of MHC I/II, CD5 expression on DP thymocytes is reduced reflecting the lack of TCR signalling 14 . Importantly, in this setting CD5 expression was expressed equivalently in the presence or absence of the Δ β CDR1-3 mutant (Fig. 6) excluding a role for non-MHC ligands in TCR signalling via Δ β CDR1-3 mutant TCRs. Individually, expression of MHC class I and class II gave rise exclusively to SP8/CD8 and SP4/CD4 populations respectively both in the absence and presence of the transgenic Δ β CDR1-3 mutant (Figs 4a-c and 5a-c). Despite Δ β CDR1-3 expression being lower than endogenous TCR in this cross, possibly due to co-expression with endogenous TCR ( Supplementary Fig. S2), its presence significantly increased CD5 expression in the MHC class I + /II + and MHC I − /II + groups (Fig. 6). These data demonstrate that the Δ β CDR1-3 mutant CD4 and CD8 T cell repertoires formed by pairing with endogenous TCR-α chains are strictly MHC restricted.

Germline TCR structure controls peptide cross-reactivity
Given the increased engagement of the Δ β CDR1-3 repertoire with MHC ligands during development, the impact on peripheral T cell-mediated immunity was assessed. The peripheral Δ β CDR1-3 T cell repertoire was distinct from the Δ β CDR3 in having smaller CD4 and CD8 T cell populations (Fig. 7a,b) reflecting the smaller thymic SP4 and SP8 populations (Fig. 2b). Further, surface TCR expression was reduced (Fig. 7c,d) and proportions of activated CD62-L − CD44 + CD4 and CD8 T cells were increased (Fig. 7e,f). Overall, this phenotype is consistent with constitutive TCR engagement promoting T cell differentiation. On-going engagement with MHC ligands as the cause of low surface TCR expression was investigated by adoptive transfer of Δ β CDR1-3 spleen cells into MHC I/II sufficient (FVB/N β /δ KO) or MHC I/II deficient (B6 MHC I/II KO) recipients. Consistent with this explanation, surface TCR was up-modulated in the absence of MHC but remained low in MHC sufficient recipients (Fig. 8).
Antigen-specific T cell responses of Δ β CDR1-3 and control Δ β CDR3 transgenic mice were assessed by IFN-γ elispot following immunisation with hen egg lysozyme protein (HEL). Responses were assayed on pools of 4-5 HEL peptides showing Δ β CDR1-3 mice responded more robustly to 4/10 pools than control Δ β CDR3 mice (Fig. 9a). This observation prompted us to examine peptide cross-reactivity within the two repertoires. HEL immunised mice were assessed for recognition of peptides from the unrelated protein Epstein-Barr virus nuclear antigen 1 (EBNA1). Both strains responded similarly to whole HEL protein (Fig. 9b upper). Strikingly, HEL immunised Δ β CDR1-3, but not Δ β CDR3 mutant mice, also responded to the EBNA1 peptide pool (Fig. 9b  lower). These data demonstrate that TCR cross-reactivity to is regulated by germline TCR structure.

Discussion
Our findings provide further evidence that germline CDRs do not hardwire the TCR for recognition of MHC ligands 4,5,9,16,18,19 . Rather, we show they have the opposite function of moderating engagement with MHC ligands. The Δ β CDR1-3 mutant studied here has germline CDRs with simpler, flexible loop structures through substitution of larger, complex amino-acid side-chains (H, Y, R, N, Q) for glycine and alanine. We hypothesise that the modified loops reduce steric hindrance effects when engaging with MHC ligands increasing the average affinity of the interface accounting for the increase in CD5 expression across the DP pre-selection thymocyte repertoire and the increased TCR hi and CD69 hi components (Fig. 3, Supplementary Fig. S1). Conversely, the more complex and rigid 20 wild-type structures will less frequently find compatible interfaces and thus enrich the repertoire with T cells that generally engage MHC ligands within a lower affinity range. Indeed the germline β CDR1 and 2 are known to undergo the smallest conformational changes on binding MHC ligands 20 . The moderating influence of germline structure on MHC engagement can likely be achieved in many ways consistent with the sequence diversity of CDR1 and 2 across the α and β variable segment repertoires 9 which results in substantial combinatorial complexity.
As thymic positive selection proceeds at extremely low affinities 21 , a degree of steric hindrance may offer an efficient solution to achieving optimal engagement across the set of highly polymorphic MHC I and II ligands. Indeed, chimeric TCR-β chains with γ δ T cell lineage, Ig heavy and Ig light chain CDRs are capable of productively engaging with MHC I and II 9,22 (Fig. 1). In the context of wild-type germline structure, single amino-acid substitutions, in particular tyrosine, can diminish MHC engagement 23 . However, the same residues are also important for engagement with non-MHC ligands 16 and are enriched within Ig germline CDRs 24 suggesting generic properties such as versatile Van der Waals interactions are employed to contribute to interface formation. Strikingly, loss of germline TCR structure did not compromise MHC restriction highlighting the efficiency of co-receptor imposed ligand bias 18 .
The majority of TCR:MHC:peptide co-crystals show the TCR engages MHC ligands with a fixed orientation and constrained geometry 25 which, at least for MHC class I, can promote productive signalling 26 . Our findings do not support the idea that germline TCR structure is crucial for directing docking topology as functional engagement with MHC was increased in the absence of germline structure. Indeed, two TCRs derived from human induced T regulatory cells bound MHC class II with reversed polarity 19 . A mechanistic link between docking topology and signalling competence has been proposed to involve appropriate positioning of co-receptor bound lck with CD3 ITAMs 27 . Whether the modified Δ β CDR1-3 germline CDR loops influence docking topology is unclear but we may infer that the germline CDRs remain MHC-centric 28 rather than peptide-centric as the Δ β CDR1-3 peripheral repertoire is highly peptide cross-reactive, which would not be expected if self-peptides were strongly engaged by the germline CDRs. We propose that increased engagement of MHC by Δ β CDR1-3 TCRs shifts the pre-selection repertoire towards negative selection (Figs 2 and 3) favoring survival of DP thymocytes whose TCR CDR3 loops make only limited contact with self-peptide. In similar fashion, MHC-centric Δ β CDR1-3 TCRs on peripheral T cells would reduce the contribution required from non-self peptide to reach the threshold for T cell activation accounting for their increased peptide cross-reactivity (Fig. 9). On the other hand, wild-type CDR structures, by engaging less well with MHC, place greater emphasis on CDR3 engagement with peptide resulting in lower peptide cross-reactivity. The T cell compartment must be able to recognise virtually any peptide, this is orders of magnitude greater than the number of T cell clones in an individual. TCR cross-reactivity is therefore an essential feature of the T cell repertoire 29 ; indeed, peptide cross-reactivity of several TCRs has been estimated to be typically > 10 6 28,30,31 . We conclude that germline structure governs this crucial aspect of the TCR repertoire by 'fine-tuning' TCR engagement with MHC ligands. In WT TCRs, cross-reactivity is mostly due to tolerance of peptide substitutions to residues outside of the TCR interface and relatively conservative changes to residues that contact the CDR3 loops whilst maintaining germline CDR-MHC docking 28 . Δ β CDR1-3 germline CDRs with simpler, flexible loop structures will have broader docking options providing a mechanism to further increase cross-reactivity.  In summary, the BCR and α β TCR appear to use their germline regions in a functionally analogous manner to achieve their distinct strategies of antigen recognition. Co-receptor mediated sequestration of lck focuses the α β TCR repertoire onto MHC ligands allowing the germline regions to 'fine-tune' the affinity of this engagement and regulate peptide cross-reactivity.
Although both arms of adaptive immunity evolved in early jawed vertebrates with primitive extant species (cartilaginous fish) having the key components (RAG 1/2, BCR, TCR, MHC I/II, CD4 and CD8), their evolutionary relationship is unclear. We speculate that 'antibody-like' recognition of ligand by receptors of both lineages may have allowed variable segments to be more easily 'co-opted' to function as part of the alternative strategy of antigen recognition simplifying the early evolutionary path of adaptive immunity.

Methods
Mice. Mice were housed under specific pathogen-free conditions at Imperial College London. All experimental protocols were approved by the Institutional Animal Welfare and Ethical Review Committee and the Home Office and were performed in accordance with the relevant guidelines and regulations. WT FVB/N (H2 q ) mice were obtained from Harlan Laboratories, Loughborough, UK. FVB/N TCR-β /δ KO were a kind gift from Prof. Daniel Pennington. C57BL/6 MHC class I, II deficient mice 32 were a kind gift from Prof. Matthias Merkenschlager. TCR-β chain variants were synthesised (Mr Gene, Regensburg, Germany). Genetically modified mice (retrogenic and transgenic) were produced directly on an FVB/N TCR-β /δ KO background. Disruption at the trb locus prevents expression of endogenous TCR-β chains and α β T cell development. Developing α β T cells must use the exogenous TCR-β chain 9,11 . For retrogenic expression, TCR-β chains were cloned into the pMigR1 retroviral vector and transduced into FVB/N TCR-β /δ KO haematopoietic stems cells (HSC) which were then injected intravenously into irradiated (600 Rad) FVB/N TCR-β /δ KO recipients as previously described 9,11 . For conventional transgenic production, the T cell specific expression vector pVA-hCD2 33 was used which gives position independent expression directed by the CD2 LCR. For each transgene, two founders were produced which had indistinguishable phenotypes. To breed the Δ β CDR1-3 transgenic strain onto MHC deficient backgrounds (MHC I + II + ; MHC I + II − ; MHC I − II + ; MHC I − II − ) it was back-crossed with MHC class I, II deficient mice (H2 b ) and the offspring typed by flow cytometry for MHC I, MHC II and TRBV16 for the transgenic TCR.
Immunisation. 6 week old female mice were immunised (s.c.) by injection of 200 μ l of HEL (final concentration 7 nM) (Sigma, Poole, UK) emulsified in complete Freund's adjuvant (CFA) and boosted after 7 days with HEL emulsified in incomplete Freund's adjuvant (IFA). Mice were sacrificed after a further 7 days.

IFN-γ ELISpot immunoassay.
An Elispot kit (Genprobe-Diaclone, Besançon, France) was used in accordance with the manufacturer's directions. In brief, PVDF 96-well plates (Millipore, Watford, UK) were activated with 70% ethanol then coated with 100 μ l of 10 μ g/ml of capture antibody overnight (4 °C) then washed 6X with sterile PBS. Plates were then blocked with PBS/2% powdered milk and washed. 10 5 responder cells were added to each well along with antigen and incubated overnight at 37 °C/5% CO 2 . Plates were then washed 6X with 100 μl PBS/0.1% Tween-20 and the biotinylated detection antibody added. After 2 hrs, the plates were washed 6X as before. 100 μ l of streptavidin-alkaline phosphatase was then added (1 hr/37 °C) and the plates developed. Spot forming cells were visualised using the AID ELISpot plate reader (AID Diagnostika, Germany). Antigens: HEL (Sigma, Poole, UK); HEL peptide pool of 24 15mers with 10 amino acid overlap (GL Biochem, Shanghai, China) and EBV EBNA1 peptide pool (JPT Peptide Technologies, Berlin, Germany) with 158 15mer peptides with 11 aa overlap.
Cell preparation. Tissues were harvested from 5-8 week old wild-type transgenic mice and from retrogenic mice 5-7 weeks after reconstitution. Cell suspensions were prepared using a sterile cell strainer and washed twice in 2 ml of phosphate-buffered saline (PBS). Blood and spleen preparations suspended for 30 minutes at room temperature in red blood cell lysis buffer (Qiagen, Hilden, Germany) to lyse red blood cells, centrifuged and resuspended as appropriate for the procedure.
Flow Cytometry. Cell suspensions were stained with fluorochrome-conjugated monoclonal antibodies against CD4, CD8, TCR-β constant region, TRBV16, CD44, CD62-L, CD5, CD69, MHC I, MHC II (all from BD Biosciences). Samples were incubated for twenty minutes at room temperature, washed in 2 ml of PBS and kept at 4 °C until acquisition on a BD FACSCalibur TM flow cytometer using the Cellquest TM Software (BD). A live lymphocyte gate was applied to all samples. Further gating was done on CD4 + and CD8 + as appropriate. Data analysis was performed on FlowJo version 5.7 (Tree Star Inc, US).