Structure of MHC-Independent TCRs and Their Recognition of Native Antigen CD155

After the initial activation by evolutionarily conserved innate immunity components, the immune systems of humans and other mammals further develop robust, specific, and memory-type responses against various pathogens. These acquired immunities bifurcate into two equally important and mutually reacting arms, B cell–mediated humoral responses and T cell–mediated cellular responses. To confront a large array of potential pathogens, B and T cells use similar gene recombination machineries to generate diverse number of AgR. Although BCR recognize both linear and three-dimensional antigenic epitopes, peripheral T cells normally recognize only linear peptides presented by major histocompatibility Ags (MHCs), a feature referred to as “MHC restriction.” A wealth of structural studies have shown that αβ TCRs adopt a common binding geometry on peptide-MHC, with germline-derived CDR1 and CDR2 loops primarily interacting with MHC helices and CDR3 engaging the peptide. The fixed TCR docking geometry on MHC was interpreted as coevolution between TCR and MHC. However, in CD4/CD8 coreceptor-deficient mice that are also MHC-deficient Quad^KO mice (B2m^−/−H-2Ab1^−/−CD4^−/−CD8a^−/−), the thymus generates mature T cells expressing MHC-independent TCRs that bind to conformational epitopes on native ligands independently of MHC. Recently, the molecular and sequence signatures of the MHC-independent TCR repertoire from Quad^KO mice was found to be distinct from the MHC-restricted TCR repertoire of conventional B6 mice, especially in their hypervariable CDR3 regions (1). For example, both TCRα and β sequences of the Quad^KO showed increased Cys frequency in their hypervariable CDR3 regions compared with their MHC-sufficient littermates. In addition, the average CDR3α and CDR3β are longer in the Quad^KO repertoires than those of MHC-restricted repertoires. Whether the T cells from Quad^KO mice are genuine non–MHC-restricted T cells that lack specific binding to peptide-MHC and instead bind specifically to non-MHC ligands or whether they are a subset of MHC-restricted T cells exhibiting cross- reactivities to non-MHC ligands is intensely debated, as their ligands are mostly unknown. To further characterize MHC-independent TCRs, we cloned two TCRs from the Quad^KO T cells, A11 and B12A, and identified CD155 as their selecting ligand (2). To date, there is no structural information available on MHC-independent TCR, and how they recognize non-MHC ligands remains unknown.

To our knowledge in this study, we report the first crystal structures of two MHC-independent αβ TCRs and characterized their conformational epitopes on CD155. Specifically, we attempted to address if MHC-independent TCRs possess unique structural features and determined their binding epitopes on CD155.

Full-length TCR cDNA for TCRα and TCRβ were cloned with PCR primers specific to sequences 5′ of the start and 3′ of the stop codon of rearranged TCR chains. TCR expression constructs were cloned into murine stem cell virus–based retroviral plasmids with pMX-IRES-GFP (provided by R. Bosselut, National Cancer Institute) for TCRβ and MSCV-IRES-tNGFR (provided by W. Pear, University of Pennsylvania) for TCRα, and they were transfected into PlatE cells separately to produce retrovirus containing supernatants. TCR-negative 4G4 T cells were retrovirally transduced with both TCRα- and TCRβ-containing retroviruses in the presence of 2 μg/ml of polybrene. The resulting B12A-4G4 or A11-4G4 cells (5 × 10⁴ cells per well) were incubated overnight at 37°C with B6 or MHC^KO stimulator spleen cells that had been treated overnight with LPS (10 μg/ml), irradiated with 3000R, and T depleted. Purified B10.A lymph node T cells that were alloreactive against H-2^b stimulators were used as control responder T cells. MHC blocking Abs used were a mixture of mAbs specific for MHC class I anti–H-2K^b/H-2D^b (28-8-6) and MHC class II Y3P (I-A), M5/114 (I-A/I-E) at 10 μg/ml each. Culture supernatants of stimulated B12A- and A11-4G4 cells were collected, and IL-2 concentrations were measured by ELISA (R&D Systems).

Chimeric DNA encoding TCR A11 or B12A α and β variable regions and corresponding human constant regions (Cα and Cβ) were synthesized and cloned into pET30a vectors (Genescript, Piscataway, NJ), as previously described. The soluble TCR A11 and B12A were produced as disulfide-linked heterodimers by a rapid dilution refolding procedure as previously described (3). In brief, each receptor chain was expressed separately as inclusion bodies in BL21 (DE3) cells. Then, purified inclusion bodies of TCR α­- and β-chains were mixed at ∼1:1 M ratio and dissolved in 8 M urea, 20 mM Tris (pH 8), 1.5 M guanidine-HCl, 5 mM EDTA, and 5 mM sodium acetate supplemented with 5 mM DTT. The dissolved inclusion bodies were injected into the refolding buffer containing 0.4 M arginine, 100 mM Tris (pH 8.5), 5 mM reduced glutathione, 0.5 mM oxidized glutathione, and 2 mM EDTA at 4°C. Two days later, refolding mixture was dialyzed against 10 mM Tris (pH 8) and 100 mM urea. Soluble TCRs were purified by anion-exchange chromatography (HiTrap Q FF; GE Healthcare), followed by size exclusion chromatography (Superdex200 16/60; GE Healthcare) in 10 mM HEPES (pH 7.4) and 0.15 M NaCl. The V domain–only single-chain construct of B12A consists of aa 1–116 from Vα, 1–116 from Vβ with a (GGGGS)₃ linker between the variable regions and a C-terminal six-histidine tag. The inclusion body of and B12A single-chain variable domains were refolded and dialyzed the same as described for the heterodimers. Soluble monomeric single-chain variable domains were purified by Ni-NTA affinity chromatography and size-exclusion chromatography.

TCR A11 and B12A proteins were concentrated to an A_{280 nm} of 12.2 and 10.5, respectively. TCRA11 was crystallized in 20% PEG 4000, 0.1 M Tris (pH 8.5), and 10 mM MgCl₂ and optimized by streak seeding. TCR B12A was grown in 25% PEG 4000, 0.1 M sodium acetate (pH 4.6), and 0.2 M (NH₄)₂SO₄. The crystals were immersed in cryoprotectant containing the respective mother liquor plus 15% glycerol prior to flash cooling in liquid nitrogen. All x-ray data were collected at SER-CAT beamlines and processed with HKL2000 (4) (Table II). The structures of TCR A11 and B12A were solved by a molecular replacement method with the program Phaser (5) in CCP4 packages (6) using the TCR B3K506 structure (Protein Data Bank [PDB] identifier [ID]: 3C5Z) as the search model. The structure refinement was subsequently carried out by autoBUSTER (7) with repeated cycles of rebuilding in COOT (8). Data collection and model statistics are summarized in Table II. Figures were prepared with PyMOL (9). The structural numbering of residues does not include the signal peptide region of TCR and thus is different from the sequencing numbering. In particular, CDR3β is between Cys 89 and Phe 100 for A11, Cys 90 and Phe 105 for B12A, and Cys 104 and Phe 118 for all repertoire sequences.

Surface plasmon resonance measurements were performed using a BIAcore 3000 instrument and analyzed with BIAevaluation 4.1 software (Biacore). Murine CD155 and human CD155 were obtained from Sino Biological and R&D Systems, respectively. To measure the affinity, TCRs or CD155 proteins were immobilized on carboxylated dextran CM5 chips (Biacore) to 200–500 response units using a primary amine-coupling and 5 μl/min flow rate in 10 mM sodium acetate (pH 4). The analyte consisted of serial dilutions of TCRs or CD155 between 0.3 and 2.5 μM in a buffer containing 10 mM HEPES (pH 7.4) and 0.15 M NaCl. The dissociation constants were obtained by kinetic curve fitting using BIAevaluation 4.1 (Biacore).

rTCR B12A and CD155 (Sino Biological) proteins were dialyzed against HEPES buffer (10 mM HEPES, 150 mM NaCl [pH 7.5]), respectively. To prepare TCR B12A–CD155 complex samples, equal volumes of TCR B12A (4 mg/ml) and CD155 (4 mg/ml) were mixed and chemically cross-linked with 0.04% glutaraldehyde for 30 min on ice. The cross-linking reaction was quenched by addition of Tris buffer (pH 8) to a final concentration of 10 mM. The mixture was subsequently subjected to size-exclusion chromatography (Superdex 200 10/300; GE Healthcare), and proteins corresponding to molecular mass of ∼100 kDa were collected to fractions of 0.5 ml. To prepare samples for negative-stained electron microscopy, each fraction was adsorbed to plasma-cleaned (Solarus Model 950 cleaner; Gatan) electron microscope grids coated with continuous carbon film, which were subsequently washed with HEPES buffer and stained with 0.75% uranyl formate. Images were collected using EPU software (FEI) on a Tecnai T12 electron microscope (FEI) fitted with a 4K charge-coupled device camera (Gatan) at an effective pixel size of 0.18 nm in the specimen plane, with the magnification of 110K or 67K.

Wild-type or chimeric CD155 constructs were synthesized (GeneScript) and cloned into SPORT6 or pIRES2-ZsGreen1 expression vectors. All constructs (300–400 ng of DNA) were transfected in 293T cells (10⁵ cells per well) using Lipofectamine 2000 (Invitrogen) in 96-well flat-bottom plates and left overnight. Mouse CD155 surface expression levels were analyzed 24 h posttransfection by flow cytometry. Transduced 4G4 responder cells (10⁵ cells per well) transduced with B12A αβ-chain as described previously were cocultured with transfected 293T cells for 24 h (2) and mIL-2 amounts in the supernatants were subsequently determined by ELISA (R&D Systems).

The structural coordinates of MHC-independent TCRs A11 and B12A have been deposited to the PDB (www.rcsb.org) under accession number 6C68 and 6C61, respectively. The sequence data described in the manuscript are freely accessible through ImmuneAccess (http://clients.adaptivebiotech.com/pub/lu-2019-natcomms; doi.org/10.21417/JL2019) and are available upon request.

In contrast to MHC-deficient animals that are devoid of mature αβ T cells (10–13), Quad-deficient mice developed ∼25% of mature αβ T cells, which was further increased to that similar to B6 in the presence of Bcl2 transgene (14). Two of these T hybridomas, A11 and B12A, recognized mouse CD155, a murine poliovirus receptor homolog (2). Moreover, transgenic T cells expressing A11 or B12A failed to proliferate in the CD155^−/−, but MHC sufficient, host, but developed normally in CD155^+/+ host mice (15), suggesting these TCRs used CD155 as their selection ligand during thymic development. To further confirm that A11 and B12A TCRs have no intrinsic MHC reactivity, we retrovirally expressed each αβ TCR in a TCR-negative 4G4 T cell line (16). Both A11- and B12A-expressing 4G4 cells responded to anti-TCR stimulation as measured by their IL-2 secretion (Fig. 1), demonstrating a functional reconstitution of their αβ TCR on 4G4 cells. However, unlike MHC-restricted B10.A T cells, which reacted only against H-2^b expressing but not against MHC-deficient stimulators (Fig. 1A), A11 and B12A T cells responded to both B6 and MHC^KO(b2m^−/−Ab^b−/−) splenocytes (Fig. 1B, 1C). Comparable amounts of IL-2 were secreted in the presence and absence of MHC, demonstrating that the reactivity of A11- or B12A-expressing 4G4 T cells was independent of MHC. To further rule out if their response to MHC^KO stimulators resulted from cross-reactivity of TCRs to non-MHC ligands, we stimulated B12A-expressing 4G4 T cells in the presence of MHC-blocking Abs. The results showed that whereas blocking MHC abolished the B6 stimulation of MHC-restricted B10.A lymph node T cells, it did not affect the B12A and A11 response to B6 stimulation (Fig. 1B, 1C). Thus, these TCRs have no intrinsic MHC reactivity and are functional independent of MHC.

Previously, we showed that the solution binding affinity of B12A to CD155 was 230 nM, much higher than a typical TCR–MHC binding (2). We also measured the binding affinity of A11 to recombinant soluble CD155 and obtained 280 nM dissociation constant K_D (Fig. 2A). Thus, both A11 and B12A have high affinity binding of CD155. Conventional αβ TCRs recognize short peptidyl- or glyco-Ags presented by MHC or MHC-like molecules. Their antigenic interactions are mediated by TCR variable CDR regions (17, 18). Unlike the conventional MHC-restricted αβ TCRs, both A11 and B12A bound native CD155 without MHC. However, the structural requirement for TCR recognition of non-MHC dependent Ag remains unknown. A previous mutational work of A11 TCR showed the involvement Tyr 46 and Tyr 48 of CDR2β in recognition of CD155 (2). To further address if the recognition of CD155 is mediated by the receptor variable domain, we generated a V-domain–only single-chain construct of B12A and obtained its solution binding affinity of 400 nM to soluble CD155, slightly less but similar to that of V-C two domain B12A (Fig. 2B). Thus, B12A V-domain alone is sufficient to recognize CD155.

Although both A11 and B12A recognize CD155, their TCR sequences are derived from different variable gene segments. The V region of A11 consists of TRAV12-1 (Vα8) and TRBV13-3*3 (Vβ8.1) alleles, and B12A is composed of TRAV6D-6*02 (Vα4) and TRBV26*01 (Vβ3) segments (Table I). The CDR3α and β of A11 contain 13 and 10 aa, whereas those of B12A contain 13 and 14 residues, respectively. To investigate the structures of MHC-independent TCRs and their recognition of CD155, we expressed soluble αβ TCR A11 and B12A heterodimers, fused to the C region of human TCR, with an engineered interchain disulfide bond between Cα146 and Cβ154 as described for MHC-restricted αβ TCRs (19). The crystal structures of A11 and B12A were determined to resolutions of 2.6 and 2.4 Å, respectively (Fig. 3A, 3B, Table II). All CDR regions of A11 and B12A are well ordered. The overall structure of A11 is nearly identical to that of B12A with a 0.5 Å root-mean-square deviation (rmsd) among all Cα carbons (Fig. 3C). In fact, both A11 and B12A exhibit canonical TCR structures with an rmsd of 1 Å when superimposed to the structure of an MHC-restricted TCR (3PQY) (Fig. 3D).

As both A11 and B12A exhibit canonical TCR structures, we then examined their CDR sequences for possible reasons of discriminating against MHC binding. Despite recognizing the same CD155 ligand, A11 and B12A use different V-gene segments. A11 and B12A α-chains CDR1 and 2 are derived from TRAV12-1 (Vα8) and TRAV6D-6*2 (Vα4), respectively, whereas their β-chains are derived from TRBV13-3 and TRBV26-1, respectively. There is no sequence conservation in A11 and B12A CDR1, CDR2 (Table I). To address if these V-genes are favored for MHC-independent ligand binding, we compared the usage frequencies of these V-genes between MHC-restricted (B6 and its Bcl2 transgenic littermate B6Bcl2^tg) and MHC-independent (Quad^KO and its Bcl2 transgenic littermate Quad^KOBcl2^tg) TCR sequence repertoires (1). The results showed that although the frequencies of both A11 and B12A Vα-genes were slightly higher in MHC-independent TCR sequences, their Vβ-gene frequencies were similar or lower in MHC-independent TCR sequences (Fig. 4A). Indeed, the same Vα- and Vβ-genes of A11 are also found in the structures of MHC-restricted TCR 1LP9 and 1NFD, respectively (Fig. 4B, 4C). Structure comparison showed that both CDR1 and 2 of A11 α- and β-chains adopted near identical conformations as those of MHC-restricted ones (Fig. 4B, 4C). Similarly, the CDR1 and 2 conformations of Vβ-gene observed in B12A are also in close agreement with those that are MHC restricted (Fig. 4D, 4E). Thus, A11 and B12A V-genes do not discriminate against the MHC-restricted ligand.

Conventional MHC-restricted TCRs favor CDR3 length of 8–13 aa (1). To further address if the length of A11 and B12A CDR3 is less optimal for binding to MHC, we compiled CDR3α and CDR3β length distributions from 132 published TCR–MHC complex structures in PDB. The majority of their CDR3α and β are 9–12 aa long, with average lengths of 11.5 and 11.7 aa for CDR3α and β, respectively (Fig. 5A). Although the lengths of A11 CDR3α and β are within the optimal length of those MHC-restricted TCRs, those of B12A are slightly longer than most of MHC-restricted TCRs (Fig. 5A). In the cases of TCRs that contained a longer CDR3α or β, the longer CDR3 often results in decreased MHC contacts from CDR1 and 2 (1). The longer CDR3 found in CD155-specific B12A, however, is consistent with those from γδ TCRs as well as those present in the preselection repertoire of MHC-restricted animals (20).

To further assess the uniqueness of A11 and B12A CDR3 sequences, we examined the occurrence of A11 and B12A CDR3 sequences in various TCR sequence repertoires, including preselection double negative and double positive (B6_DN, DP), mature MHC-restricted (B6), and mature MHC-independent sequences (Quad^KO). The 10-residue A11 CDR3β has an optimum length of MHC-restricted CDR3, and its sequence is found in all three repertoires as well as in B6Bcl-2^tg repertoires (Fig. 5B) (1). In fact, A11 CDR3β occurs rather frequently in B6, with some in the top 1000 sequences out of ∼300,000 sequences, albeit it is sometimes associated with different V-genes. In addition, the intact TCRβ of A11, the recombination of TRBV13-3 with CDR3β (CDR1, 2, and 3β), can be found in all four repertoires (Fig. 5B). Thus, TCRβ of A11 is not unique to MHC-independent TCR. However, TCRα of A11 is not found in any of the MHC-restricted repertoires, suggesting the non-MHC ligand specificity of A11 is derived from the unique pairing between its α/β-chains during thymic selection. In contrast, the length of the 14-residue B12A CDR3β is less optimal for MHC-restricted TCRs (1), and its sequence occurred rarely in various repertoires but is observed in the B6 repertoire (Fig. 5B). The TCRβ of B12A (recombination of TRBV26 with its CDR3β), however, is only found in B6Bcl2^tg repertoire but not in B6. Consistent with non-MHC ligand specificity derived from the pairing between TCR α/β-chains, the α- but not β-chain of B12A was found in MHC-restricted repertoires.

Previous domain swapping of murine and human CD155 showed that the activation of A11 and B12A required D1-D2 domains of CD155 (2). To further assess the mode of TCR-CD155 recognition, we acquired negative staining electron microscopy images of recombinant B12A in complex with soluble CD155. B12A TCR alone appears as a donut-shaped structure, and the TCR–CD155 complex resembles the shape of a tadpole with TCR as its head and CD155 as its tail in negatively stained electron microscopy images (Fig. 6A). Although the resolution of the electron microscopy images is not sufficient to determine the precise molecular interface, the structural images are best fit with a binding mode in which the TCR is orientated with its CDR regions docked onto the D1 domain of CD155 (Fig. 6B). Because both A11 and B12A responded to murine but not human CD155, we used a mutational approach to further map the TCR binding region on CD155 D1 domain. In all, there are 25 variant surface residues on CD155 D1 domain between human and mouse (Supplemental Fig. 1A). We generated 12 cluster mutations based on the crystal structure of human CD155 (PDB entry 3UDW) to replace mouse CD155 residues with either human counterparts or alanine residues (Fig. 6C, Supplemental Fig. 1B). The activation of A11 or B12A-transduced 4G4 T cells was measured by IL-2 production in the presence of wild-type or mutant CD155-expressing HEK293T cells (2). Both the mutants and wild-type exhibited similar surface expressions of CD155 (Supplemental Fig. 2). Although a majority of mutations stimulated either A11 or B12A TCR to produce comparable level of IL-2 as the wild-type CD155, two separate cluster mutations resulted significantly reduced IL-2 production from either A11 or B12A-transduced T cells despite their similar surface expressions of CD155 as the wild-type (Fig. 7). Specifically, cluster mutations 6 and 7, located at C′-strand and C′C″-loop, impaired A11 activation, and mutations 5 and 6, located at CC′-loop, abolished B12A activation. Clusters 6 and 7 and 5 and 6 define two continuous surface patches on CD155 structure. These mutational results show that A11 and B12A TCR recognize two closely related but different epitopes on D1 domain of CD155. To further confirm the mutational result, we aggregated the five activating mutations, 2, 4, 7, 9, and 12, to generate a penta-cluster (Y36T/N57E/K83T/K84Q/E92K/Q124L) mutation. The penta-cluster mutant activated both A11 and B12A like wild-type CD155 (Fig. 7). Interestingly, clusters 6 and 7 are located in the human CD155 region identified for poliovirus and Tactile (CD96) binding (21, 22). The docking of B12A on CD155 showed the close contact of both CDR2 and CDR3 of B12A TCR with the ligand. This is consistent with previous Ala mutation of A11 CDR2β, which resulted in decreased IL-2 production in response to CD155 (2). The involvement of CDR3 is further supported by the failure of B12F TCR to respond to CD155 stimulation (2). B12F differs from B12A only in its CDR3α residues. Thus, unlike conventional TCRs that only recognize short peptide fragments presented by MHC molecules, A11 and B12A recognize distinct native surface patches on D1 domain of CD155, consistent with their different surface electrostatic potential distributions of their CDR loop regions (Fig. 8). The combination of high affinity binding and different surface shape of A11 and B12A epitopes are characteristic of Ab recognition. It suggests non-MHC restricted TCRs may recognize a broad spectrum of conformational Ags.

We showed previously that TCRs can recognize non-MHC ligands (2). In this study, we showed A11 and B12A TCRs produced IL-2 in response to MHC-deficient stimulators, and blocking of MHC did not affect the IL-2 production from MHC-sufficient stimulators (B6) (Fig. 1), demonstrating the functional independence of A11 and B12A to MHC expression. To our knowledge, we determined the first structure of these MHC-independent TCRs and mapped their binding epitope on CD155. The structures of A11 and B12A resemble those of MHC-restricted TCR. We showed in this study that unlike conventional αβ TCRs recognizing short peptide fragments presented by MHC molecules, A11 and B12A TCRs recognize distinct native surface patches on D1 domain of CD155, consistent with their distinct amino acid and surface electrostatic potential distributions of their CDR loop regions (Fig. 8). The combination of high affinity binding and different surface shape of A11 and B12A epitopes are characteristic of Ab recognition.

However, MHC-independent TCRs A11 and B12A bear normal germline genes that are also used in MHC-restricted αβ TCRs. Through deep sequencing, we could identify their respective Vβ region including CDR3β among preselected, MHC-restricted, and MHC-independent mature T cell repertoires, suggesting individual V genes were not sufficient to predetermine the ligand specificity of TCRs. Recent publication of Krovi et al. (23) showed that the stimulation of unselected αβ TCRs by MHC-expressing APC was largely inhibited by anti-MHC blockade. It is not clear, however, if these randomized unselected TCRs would fail to respond to MHC-deficient stimulation. Although the unselected T cells necessarily contained many MHC-restricted T cells, it is not clear if they are sufficient for the response. The presence of a minor population of non–MHC-restricted T cells in the pool of unselected T cells could not be ruled out without testing their response to MHC-deficient Ag stimulation. In this study, both A11 and B12A responded to MHC-deficient APC, whereas MHC-restricted B10.A allogenic T cells did not (Fig. 1). Furthermore, unlike the TCR response from Krovi et al., the response of B12A-4G4 T cells was not affected in the presence of MHC blockade. Indeed, substantial αβ TCRs underwent positive selection and maturation in quad-deficient mice (B2m^−/−H-2Ab1^−/−CD4^−/−CD8a^−/−), demonstrating αβ TCRs do not necessarily have intrinsic activity to MHC molecules. This also highlights the importance of thymic selection and its intricate selection mechanism in determining the ligand specificity of αβ TCRs.

Our previous study also revealed that transgenic TCRs A11 and B12A could only initiate positive selection through CD155 in the thymus, but MHC-independent TCR B12F that was identical to B12A except the CDR3α residues could not undergo positive selection through CD155, demonstrating that MHC-independent reactivity was TCRαβ pair specific. Through VDJ recombination, the adaptive immune system generates huge numbers of TCR α- and β-chains. Although structural studies have glinted important information for individual α- or β-chain in the contribution of MHC and Ag recognition, the functional role of their pairing remains to be further investigated. A recent study using TCR β-chain (Vβ8.2 or TRBV13-2) of IA^b-3k-reactive YAe62 TCR as a transgene demonstrated that TCRs paired with this transgene and different Vα-chains resulted in exclusive MHC class II specificity, or cross-reactivity with classical or nonclassical MHC class I ligands upon thymic selection on different MHC background, revealing a role of differential TCR αβ pairing in directing MHC ligand specificity (24). The influence of TCR αβ pairing on the ligand specificity of TCR repertoire was also examined by high-throughput single-cell sequencing, which showed little overlap between CD4 and CD8 TCR repertoires (25). Therefore, αβ pairing showed strong association with T cell lineage and their corresponding MHC–Ag specificities. In our previous deep sequencing–based repertoire comparisons, we showed that MHC-restricted and MHC-independent TCR repertoires display different molecular constraints as MHC-restricted repertoires have shorter CDR3 length and limit the usage of positively charged and hydrophobic amino acids in CDR3β. In the current study, we showed that two MHC-independent TCRs, A11 and B12A, form canonical TCR structures, whose CDR3s can also be found in preselection and MHC-restricted repertoire, but often were recombined with other germline V genes (Fig. 5B). Interestingly, each MHC-restricted animal may contain either matched α- or β-chain but not both chains of A11 and B12A TCR sequences, suggesting the pairing of A11 and B12A is unique for MHC-independent ligand recognition. With the emerging high throughput TCR pairing based on single-cell RNA sequencing, the functional importance of TCR αβ pairing in directing their ligand specificity would expand our current understanding of T cell repertoires at different states such as infection, tumor and autoimmunity.

Our current study showed that TCRs may recognize a broad spectrum of conformational Ags other than MHC molecules, which could expand our current dogma of T cell–mediated biology. A recent example showed the presence of tumor-specific αβ T cells in a patient with renal cancer recognized intact TRAIL-DR4 Ag independent of class I and II MHC on human renal cell carcinomas (26, 27). The recognition of a newly formed conformational tumor Ag by a naturally occurring TCR exemplifies that TCRs could have broader ligand specificity not as expected. In addition, the Ab like recognition by MHC-independent A11 and B12A could transform our view of how TCRs interact with their Ags. As T cells from Quad^KOBcl-2^tg animals displayed activation or memory phenotype with some animals exhibited autoreactive T cell infiltrations in lung and pancreas (14), it is intriguing to speculate that self-reactive MHC-independent T cells are present in the preselection stage but are normally deleted in thymus. Failure to remove them, however, may result in autoimmunity. Recently a diverse of human γδ T cells were found to be autoreactive to the monomorphic MHC-related protein 1 (MR1) in peripheral blood and tissues. Instead of recognizing the MR1–Ag binding groove, these γδ T cells bind underneath the Ag-binding platform of MR1 like Abs (28). Ignoring the potential link between MHC-independent T cell–Ag recognition and autoimmune diseases may impede our search for cure. Therefore, the current study not only broadens our knowledge of TCR ligand specificity but also highlights an intriguing question whether TCRs recognizing non-MHC ligand would be of clinical implications.

We thank Kira A. Podolsky and Dr. Siriram Subramaniam for the support of negative staining electron microscopy at the Center for Cancer Research, National Cancer Institute, National Institutes of Health.

The authors have no financial conflicts of interest.