Key residues at third CDR3β position impact structure and antigen recognition of human invariant natural killer T cell receptors

Abstract

The human invariant natural killer T (iNKT) TCR is largely composed of the invariant TCR Vα24-Jα18 chain and semi-variant TCR Vβ11 chains with variable CDR3β sequences. The direct role of CDR3β in antigen recognition has been extensively studied. Although it has been noted that CDR3β can interact with CDR3α, how this interaction might indirectly influence antigen recognition is not fully elucidated. We observed that the third position of Vβ11 CDR3 can encode an arginine (Arg) or serine (Ser) residue as a result of somatic rearrangement. Clonotypic analysis of the two iNKT TCR types with a single amino acid substitution revealed that the staining intensity by anti-Vα24 antibodies depends on whether Ser or Arg is encoded. When stained with an anti-Vα24-Jα18 antibody, human primary iNKT cells could be divided into Vα24 low- and high-intensity subsets, and Arg-encoding TCR Vβ11 chains were more frequently isolated from the Vα24 low-intensity compared with the Vα24 high-intensity subpopulation. The Arg/Ser substitution also influenced antigen recognition as determined by CD1d multimer staining and CD1d-restricted functional responses. Importantly, in silico modeling validated that this Ser to Arg mutation could alter the structure of not only the CDR3β but also the CDR3α loop. Collectively, these results indicate that the Arg/Ser encoded at the third CDR3β residue can effectively modulate the overall structure of and antigen recognition by human iNKT TCRs.

Introduction

Invariant natural killer T (iNKT) cells are a subset of evolutionarily conserved αβ T cells that recognize lipids presented by the MHC class I homolog CD1d. These cells are rapid responders that play immunological roles in various settings, such as autoimmunity, cancer, and infection. Recognition of the canonical glycolipidα-galactosylceramide (α-GalCer) or its analog, PBS-57, presented by CD1d is a defining feature of iNKT cells. In addition to α-GalCer and PBS-57, iNKT cells are activated by various microbial lipids, self-lipids, and synthetic glycolipids such as OCH. Another feature of iNKT cells is their biased TCR repertoire. Human iNKT TCR variable region genes are largely limited to the invariant TCR Vα24-Jα18 (Vα24i) and semi-variant TCR Vβ11 chains. Murine iNKT cells preferentially express an invariant TCR Vα14-Jα18 and TCR Vβ8, 7, and 2 chains. Despite the limited V gene usage, the CDR3β sequences are highly heterogeneous in both species (1–7).

Currently, two methods are widely used to detect human iNKT cells: staining with α-GalCer/PBS-57-loaded CD1d tetramer and co-staining with anti-Vα24 and anti-Vβ11 mAbs. The CD1d tetramer was first described by Benlagha et al (8) and has since been widely adopted in the iNKT biology field. Two anti-Vα24 mAbs are commercially available for detecting human iNKT cells. Clone C15 is a pan-Vα24 mAb that was first described by Padovan et al. and binds to Vα24 regardless of the Jα gene usage (9). Clone 6B11 was developed by Exley et al. to specifically detect human iNKT cells and it recognizes the Vα24-Jα18 CDR3 loop. The 6B11 mAb was generated by immunizing CD1d^−/− mice with a cyclic form of the Vα24i CDR3 peptide. Point mutation at the Vα24-Jα18 junction of an iNKT TCR decreased 6B11 reactivity, thus confirming its cognate epitope (10). Furthermore, the frequency of 6B11 positivity was comparable to that of α-GalCer-loaded CD1d tetramer positive cells when tested with PBMC (10, 11).

Recognition of self-lipids by iNKT cells is important in their thymic development and peripheral activation. The endogenous antigenic lipids for human iNKT cells include glycolipids, phospholipids, and plasmalogens (7, 12–18). The molecular basis of self-recognition is similar to that of α-GalCer/CD1d recognition, where CDR1α, CDR3α, and CDR2β mediate critical interactions. The difference lies in the role of CDR3β, whose direct interaction with CD1d is critical when recognizing self-antigens (19–21). Since CDR3β does not directly contact the ligand, previous studies have highlighted a ligand-non-selective role in its control of the overall affinity of iNKT TCRs for CD1d-lipid complexes (20, 22, 23). We have previously identified three CDR3β amino acid sequence motifs that are associated with greater human iNKT TCR auto-reactivity strength, regardless of the lipid presented by CD1d (24). Interestingly, studies of human iNKT TCR crystal structures have suggested an additional role for CDR3β in influencing antigen recognition indirectly via interactions with the CDR3α loop (21, 25).

In the current study, we provide novel evidence for this indirect mechanism of CDR3β in regulating iNKT TCR antigen recognition. We found that staining with anti-Vα24 mAbs was altered depending on whether the Vβ11 CDR3 sequence encoded serine (Ser) or arginine (Arg) at the third position, which occurs as a result of somatic rearrangement. Herein, the amino acid following the conserved cysteine was defined as the first residue of the CDR3β sequence according to The International Immunogenetics Information System annotation. Furthermore, recognition of self-lipids and OCH differed between the Arg- and Ser-encoding iNKT TCRs. Finally, molecular modeling indicated that a mutation at this particular CDR3β residue is able to influence the conformation of both the CDR3α and CDR3β loops, which could account for the altered anti-Vα24 mAb staining and antigen reactivity. Together, these data highlight a role of CDR3β in regulating the structure of the invariant TCRα chain of human iNKT cells.

Materials and Methods

Cells and reagents

PBMC and thymus samples were obtained with institutional review board approval from the University Health Network and appropriate informed consent. SupT1, Jurkat 76, K562, C1R cells, and their derivatives were cultured in RPMI 1640 supplemented with 10% FCS and gentamicin. α-Galactosylceramide (α-GalCer) was purchased from Axxora (San Diego, CA). Recombinant human IL-2 was purchased from Novartis (New York, NY).

High-throughput sequencing of the CDR3β region of primary human iNKT cells

Primary human iNKT cells were initially purified from PBMCs or thymi using anti-iNKT MicroBeads (clone 6B11, Miltenyi Biotec, Auburn, CA). Subsequently, the cells were stained with anti-Vα24 (clone C15) and anti-Vβ11 mAbs. The double positive population was sorted with a FACSAria (BD Biosciences, Mississauga, Canada). The purity of the sorted cells was consistently >95%. TCRβ sequencing was performed at Adaptive Biotech using the ImmunoSEQ platform (Seattle, WA). This method was used to capture the frequencies of individual TCRs in biologic samples with accurate reproducibility and a sensitivity of 1/100,000 T cells (26).

Flow cytometry analysis

The following mAbs recognizing the indicated antigens were used: human TCR Vα24 (clone C15) and Vβ11 from Beckman Coulter (Mississauga, Canada); human pan TCR (clone BMA031) from Thermo Fisher Scientific (Burlingame, CA); and human TCR Vα24-Jα18 (clone 6B11), human CD3, human CD1d, human CD69, human pan TCR (clone IP26), and isotype controls from BioLegend (San Diego, CA). 7-AAD was used for live/dead staining (BioLegend). Human CD1d monomers, unloaded, OCH-loaded, and PBS-57-loaded, were kindly provided by the NIH Tetramer Core Facility. Unloaded monomers were produced in HEK293 cells and therefore presented HEK293-derived endogenous ligands. CD1d monomers were multimerized with streptavidin-PE (Life Technologies, Grand Island, NY) according to a protocol provided by the Tetramer Core Facility. All data shown were gated on singlets and live cells. Data involving Cl6 transfectants were further gated on CD3⁺ cells unless otherwise specified.

cDNAs

Full-length cDNAs encoding the Vα24i, Vβ11, and CD1d genes were molecularly cloned via RT-PCR using gene-specific primers into the pMX vector (24, 27–29). Nucleotide sequencing was performed at the Centre for Applied Genomics, The Hospital for Sick Children (Toronto, Canada). CDR3β sequences were defined according to IMGT (http://www.imgt.org/).

Generation of TCR transfectants

Using the 293GPG-based retrovirus system (30), TCRα⁻ β⁻ Jurkat 76 cells were transduced with a β₂-microglobulin shRNA (Origene, Rockville, MD) and the Vα24i gene. The Jurkat 76.clone 6 (Cl6) expressing Vα24i and low levels of CD1d was established by the limiting dilution method. Cl.6 cells were further transduced with various clonotypic TCR Vβ11 genes encoded by the 293GPG virus (24, 31). Jurkat 76 cells transduced with the HLA-A2/TAX-restricted TCR clone A6 was used as a control (32). The CD3 expression levels for each transfectant were >90%.

Expansion of CD1d-restricted iNKT cells

Human CD3⁺ T cells purified from healthy donors were plated in 24-well plates at a density of 2x10⁶ cells/well in RPMI 1640 with 10% human AB serum. Then, CD1d-expressing K562-based artificial APCs (aAPCs) pulsed with 500 ng/mL of α-GalCer were irradiated (200 Gray) and added to the responder cells at a responder to stimulator ratio of 20:1 (day 0), as previously described (24). The T cells were restimulated every 7 days and supplemented with 100 IU/ml of IL-2 every three days.

Cytokine ELISPOT analysis

IL-2 ELISPOT assays were conducted as described elsewhere (33–36). Briefly, PVDF plates (Millipore, Etobicoke, ON) were coated with capture mAb (R&D Systems, Minneapolis, MN). One hundred thousand Jurkat 76.Cl6 transfectants were incubated with indicated numbers of target cells per well for 22–24 hours at 37°C. Plates were washed and incubated with biotin-conjugated detection mAb (R&D Systems), followed by washing and incubation with streptavidin-conjugated alkaline phosphatase (Jackson ImmunoResearch, West Grove, PA). Maximal response was determined by stimulating iNKT TCR transfectants with a K562-based aAPC expressing a membranous form of the anti-human CD3 mAb, clone OKT3 (aAPC/mOKT3) (34).

Molecular dynamics simulation

Molecular dynamics (MD) simulations were performed with the MOE program (v2015.10). The atomic coordinates were derived from the crystal structures of the Vα24-Jα18-Vβ11 iNKT TCRs in complex with CD1d-α-GalCer (PDB ID 2PO6), CD1d-β-GalCer (PDB ID 3SDX), or CD1d-lysophosphatidylcholine (PDB ID 3TZV) (19–21). These iNKT TCRs are identical in their CDR except for varying CDR3β sequences. CD1d-lipid molecules were deleted from each model. Missing side chains and hydrogen atoms were generated using QuickPrep and Protonate3D. The iNKT TCRs were then energy minimized using the AMBER10 force field. Sodium ions and water molecules were added as a droplet around the molecule and energy minimized once more. iNKT TCR CDR loops and water molecules were experimentally probed for their dynamics, whereas framework residues were fixed in position. Molecular dynamics simulations were performed using the NPA algorithm on both the wild-type Vα24-Jα18-Vβ11 iNKT TCRs and an Arg95β mutant. During the MD simulations, the sample was heated to 300 K over 100 ps, followed by an equilibration of 100 ps. A production run of 500 ps was then performed followed by a cooling step to 0 K over 100 ps. A time step of 1 fs and no bond constraints were used.

Statistics

Statistical analyses were performed using GraphPad Prism version 6.0. Two-way ANOVA tests with Bonferroni post-hoc tests were employed. Analyses were paired between the Ser and Arg version of each clone, or paired between each effector-target ratio for the ELISPOT experiments. All tests were two-tailed, and p < 0.05 was considered statistically significant.

Results

A natural polymorphism at the third position of CDR3β sequences alters the antigenicity of human iNKT TCRs

We previously isolated 54 unique human TCR Vβ11 genes and individually reconstituted them along with the invariant TCR Vα24 (Vα24i) chain on the human Vα24⁻Vβ11⁻ T cell line, SupT1. Among the 54 SupT1 iNKT TCR transfectants (24), 2 clones, Cl.3007 and Cl.2133, exhibited noticeably lower staining by the pan anti-Vα24 mAb (clone C15), despite expressing comparable levels of CD3 (Supplementary Fig. 1). These two clones encoded CASR at the beginning of CDR3β(with Arg at the third residue) whereas all of the others encoded CASS or CAST. Two clones encoded CAST instead of CASS; however, they did not demonstrate lower staining by the C15 mAb. To confirm that peripheral human iNKT cells naturally encode CASR at the CDR3β loop, we purified Vα24-Jα18⁺Vβ11⁺ cells from PBMCs and thymi of 12 different donors (n = 6 for each), and analyzed their TCRβ sequences by high-throughput sequencing (Fig. 1A). The majority of the TCR Vβ11 genes encoded Ser at the third position. However, CASR sequences were indeed detected in both tissues from all donors (Fig. 1B). These results demonstrate that Arg at the third CDR3β position is a natural polymorphism of human Vβ11 TCRs.

Figure 1. Frequency of peripheral and thymic human iNKT TCRs encoding Ser or Arg at the third CDR3β position.

(A) The sorting strategy is shown for one PBMC donor. Thymic iNKT cells were similarly sorted. (B) Vα24⁺Vβ11⁺ cells were purified from PBMC and thymi samples (six donors each), and the TCRβ repertoire of sorted cells from each sample was analyzed by high throughput sequencing. Frequencies of CASS- or CASR-encoding clones among total or unique sequences are shown. Means ± standard deviations are shown in the bar graphs.

Next, we generated CASR- or CASS-encoding point mutants of Cl.3014, Cl.2037, Cl.3010, and Cl.2119 (CASS to CASR) as well as Cl.3007 and Cl.2133 (CASR to CASS). The parental and mutated sequences are shown in Table I. Note that the third CDR3β position encoding Ser or Arg was the only difference between each pair of TCR Vβ11 genes. Both CASS and CASR versions of each clone were individually reconstituted in the Jurkat 76.clone 6 (Cl6) cell line, which was engineered to stably express the Vα24i chain and low surface CD1d levels (Supplementary Fig. 2). Jurkat 76 cells lack endogenous TCR expression (37). Cl6 with low CD1d expression was used to avoid fraternal activation mediated by the transduced iNKT TCRs. Each transfectant was stained with the two distinct anti-Vα24 mAbs, C15 and 6B11. The C15 mAb stained all CASR-encoding transfectants with lower mean fluorescence intensity (MFI) compared to their respective CASS counterparts, although the differences were smaller with Cl.3014, 2037, and 3010 (Fig. 2A). The 6B11 MFIs were substantially lower for all the CASR-encoding transfectants except Cl.2037, in which the CASR was conversely stained better than the CASS version (Fig. 2B). These data demonstrate that the antigenicity of the Vα24i chain can be modulated by the third amino acid of the CDR3β sequence, which suggests that this residue may alter the structure of the Vα24i chain. Furthermore, our initial high-throughput sequencing analysis may have underestimated the frequency of the CASR-encoding TCR Vβ11 genes because the cells were purified with anti-Vα24 mAbs.

Table I.

CDR3β sequences of human TCR Vβ11 clones

Clone	CASS sequences	CASR sequences
Cl.3014	CASSEFGQSADEQFF	CASREFGQSADEQFF*
Cl.2037	CASSEFDGGQETQYF	CASREFDGGQETQYF*
Cl.3010	CASSGLLTGPDTQYF	CASRGLLTGPDTQYF*
Cl.2119	CASSEPTGLGTDTQYF	CASREPTGLGTDTQYF*
Cl.3007	CASSYYSVQGRTDTQYF*	CASRYYSVQGRTDTQYF
Cl.2133	CASSGQGLGEQYF*	CASRGQGLGEQYF

CASS and CASR encoding TCR Vβ11 sequences for each clone are shown. Third residue of CDR3β is indicated in bold. Asterisks denote the mutated sequence of each clone.

Figure 2. Ser or Arg at the third CDR3β residue impacts antigenicity of the Vα24i chain.

(A) CASS- or CASR-encoding Cl.3014, 2037, 3010, 2119, 3007, or 2133 TCR Vβ11 chain was reconstituted on Cl6 cells. Transfectants were stained with anti-Vα24 clone C15 and anti-CD3 mAbs. (B) Transfectants were similarly analyzed with clone 6B11. MFI of staining with anti-Vα24 mAbs was normalized by respective CD3 MFI. Data are representative of 3 repeated experiments. Means ± standard deviations are shown in the bar graphs. **** p < 0.0001.

CASR-encoding iNKT TCRs are preferentially expressed by human iNKT cells with low 6B11 reactivity

To address whether these in vitro findings apply to human primary iNKT cells, we stimulated peripheral T cells with α-GalCer loaded artificial APCs to polyclonally expand peripheral iNKT cells (24). Without ex vivo expansion, this experiment would not be feasible given the rarity of peripheral iNKT cells, especially 6B11 low iNKT cells, in the majority of donors. After expansion, we observed Vα24-Jα18 high, low, and negative populations among the PBS-57 tetramer positive cells when stained with the 6B11 mAb. All three populations expressed comparable levels of TCR Vβ11 (Fig. 3A). Vα24 negative, α-GalCer reactive human iNKT cells have been previously described (38). Vα24-Jα18 high and low populations were purified by flow cytometry-guided sorting, and the Vβ11 genes were cloned and analyzed from respective populations. Two unique TCRβ sequences encoding CASR were identified in each cohort. Interestingly, the two CASR sequences identified in the 6B11 low population were two of the most frequent within that group, together encompassing over 40% of all Vβ11 genes cloned from that cohort. In contrast, only one copy of each of the CASR sequences in the 6B11 high population was identified from a total of 120 Vβ11 genes (Fig. 3B and C). These results demonstrated that human peripheral Vβ11⁺ iNKT cells with lower 6B11 reactivity preferentially encode CASR within the CDR3β.

Figure 3. Peripheral iNKT cells weakly stained by 6B11 express CASR-encoding Vβ11⁺ TCRs more frequently than those stained strongly by 6B11.

(A) T cells were isolated from PBMC and stimulated by artificial APCs pulsed with α-GalCer at the concentration of 500 ng/ml. After three stimulation, the T cells were stained with PBS-75 CD1d tetramer, anti-Vα24 (clone 6B11) and anti-Vβ11 mAbs. Data for staining with anti-Vα24 and Vβ11 mAbs were shown after gating on tetramer⁺ cells. Data are representative of 3 donors. (B) Vα24 high and Vα24 low cells were isolated from three donors by flow cytometry (purity>95%). After synthesizing cDNA, we molecularly cloned a total of 120 and 187 Vβ11⁺ TCRs from the Vα24 high and Vα24 low populations, respectively, and determined their CDR3β sequences. The frequency of each unique Vβ11⁺ TCR is shown as a section of the respective pie graph. Different colors indicate clones encoding different CASR sequences. (C) CDR3β sequences of CASR-encoding Vβ11 genes are shown. The same color is used to denote identical sequences in B and C.

Encoding Ser or Arg alters recognition of self-lipids and OCH but not PBS-57

Given the dominant role of Vα24i in antigen recognition for human iNKT TCRs, we explored whether the potentially altered Vα24i conformation induced by the CDR3β single amino acid substitution impacted antigen recognition. We first stained each pair of iNKT TCR transfectants with unloaded or PBS-57-loaded CD1d tetramers. The PBS-57 tetramer staining was comparable between the CASR- and CASS-encoding clones (Fig. 4A). When stained with the unloaded tetramer, which present HEK293-derived endogenous lipids, the transfectants expressing CASR-encoding TCRs possessed low tetramer positivity (Fig. 4B). We also compared the auto-reactivity of the CASS- or CASR-encoding transfectants in functional assays. Each Cl6 transfectant was stimulated with K562 or C1R transduced with CD1d, and CD69 upregulation was measured by flow cytometry. Parental K562 and C1R cells, which do not endogenously express surface CD1d, were used as negative controls. Except for Cl.2133, a significant decrease in CD69 expression was observed for the Arg-encoding transfectants compared with those encoding Ser, regardless of the target cells used (Fig. 5). Functional autoreactivity between CASS- and CASR-encoding clones was also compared by IL-2 ELISPOT assays. Number of spot forming units were normalized to the maximal response obtained upon stimulation with an aAPC expressing a membranous form of anti-CD3 mAb (34). At the highest effector-target ratio tested, all CASS-encoding transfectants secreted IL-2 in greater proportions in response to C1R expressing CD1d, compared to CASR clones. Significantly higher cytokine secretion was also detected at limiting target cell numbers by all CASS-encoding iNKT TCRs except Cl.2133 (Fig. 6), which inherently possesses low autoreactivity (Fig. 4 and 5). These data indicate that CASS-encoding iNKT TCRs are more sensitive in recognizing self-antigens than the CASR type.

Figure 4. CASR-encoding iNKT TCR transfectants possess diminished self-lipid tetramer staining.

(A–B) The transfectants in Fig. 2 were stained with 1 μg/mL of PBS-57-loaded CD1d tetramer (A) or 10 μg/mL of unloaded CD1d tetramer (B) along with anti-CD3 mAb. MFI of tetramer staining was normalized by respective CD3 MFI. Data are representative of 3 repeated experiments. Means ± standard deviations are shown in the bar graphs. *** p < 0.001, **** p < 0.0001.

Figure 5. CASR-encoding transfectants possess diminished functional autoreactivity.

(A) The transfectants in Fig. 2 were stimulated with 5:1 effector-target ratio of K562 or K562 expressing CD1d. (B) Transfectants were also stimulated with 10:1 effector-target ratio of C1R or C1R expressing CD1d. After a 4-hour stimulation, cells were stained with anti-CD69 and CD3 mAbs and analyzed by flow cytometry. Raw data for stimulation with CD1d expressing target cells are shown. Data are representative of 3 repeated experiments. Means ± standard deviations are shown in the bar graphs. ** p < 0.01, *** p < 0.001, **** p < 0.0001.

Figure 6. CASR-encoding transfectants possess diminished cytokine secretion at various effector-target ratios.

1x10⁵ cells of each transfectant were stimulated with 5x10⁴ parental C1R or indicated numbers of CD1d-expressing C1R target cells. Responses were measured by IL-2 ELISPOT assays. All spots counted were normalized to a maximal response, which was measured by stimulating each transfectant with 5x10⁴ aAPC/mOKT3. Data are representative of 2 independent experiments performed in triplicates. Means ± standard deviations are shown in the graphs. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

OCH is a synthetic α-linked glycolipid that is also recognized by iNKT cells (39). Although human iNKT TCR-OCH-CD1d trimolecular complexes have yet to be structurally analyzed, structural analysis of a murine iNKT TCR recognizing OCH-CD1d demonstrated that the CDR3β loop did not mediate direct contact with the antigen complex, similar to how α-GalCer is recognized (40). For four of the six clones, the CASR-encoding transfectants were stained with weaker intensities by the OCH-loaded CD1d tetramer compared with the CASS transfectants. Encoding Arg or Ser had no observable effect on OCH-CD1d recognition by Cl.3007 and 2133 (Supplementary Fig. 3). We cannot rule out the possibility that the diminished reactivities could at least partly be due to a CDR3β-intrinsic effect of CASR sequence rather than modulation of Vα24i, since CDR3β can mediate direct contact with the CD1d molecule (20, 21, 23). These results, however, indicate that the third residue of CDR3β also influences the antigen recognition of human iNKT TCRs, possibly by modulating the conformation of other CDR loops within the TCR.

The third amino acid of Vβ11 CDR3β modulates the structure of CDR3α

To determine the potential for an Arg residue in the third position of the CDR3β sequence to alter the structure of the iNKT TCR CDRα loops, we performed molecular dynamics (MD) simulations on the wild-type CASS and its respective Arg95β CASR mutant of three Vα24-Jα18-Vβ11 iNKT TCRs (PDB ID 2PO6, 3SDX and 3TZV. Amino acid numbering based on 2PO6) (19–21). MD simulations of the iNKT TCRs revealed only minor differences between the wild-type and Arg95β mutant models in the CDR1/2 of TCRα and TCRβ chains, with an average main-chain root-mean-square deviation (RMSD) of 0.38 Å and 0.45 Å in the CDR1/2α loops and 0.68 Å and 0.60 Å in the CDR1/2β loops, respectively. In contrast, the CDR3 loops displayed a higher degree of variation, with the wild-type and Arg95β mutant CDR3α and CDR3β loops differing by an average RMSD of 1.13 Å and 2.17 Å, respectively (Fig. 7A and B).

Figure 7. MD simulations of the wild-type (Ser95β) and Arg95β mutated Vα24–Jα18-Vβ11 iNKT TCRs from PDB 2PO6, 3SDX and 3TZV.

Average RMSD difference in (A) CDRα and (B) CDRβ between the wild-type and Arg95β mutated iNKT TCRs models. Means ± standard deviations are shown in the bar graphs. Superposition of the CDR loops of the wild-type and Arg95β iNKT TCR for (C) 2PO6, (D) 3SDX and (E) 3TZV. The CDR3β sequence of the iNKT TCRs is shown. The wild-type iNKT TCRs are colored in grey, while the TCRα and TCRβ chains of the Arg95β mutant iNKT TCRs are colored in green and wheat, respectively. Insets in each figure are the interactions between the CDR3α/β loops of the wild-type and Arg95β mutant iNKT TCRs. Hydrogen bonds are shown as black dashes.

A comparison of the wild-type and mutant iNKT TCR simulated models revealed key differences in the interactions between the CDR3α loop and either the Ser95β- or Arg95β-encoding CDR3β loop. In the wild-type models, Ser95β either formed a lone Van der Waals contact with Thr98α of the CDR3α loop (Fig. 7C) or did not contact the CDR3α loop at all (Fig. 7D and E). In contrast, in two simulated Arg95β models, the Arg95β residue of the mutant models hydrogen bonded extensively with main-chain atoms or residues of the CDR3α loop, which resulted in a shift in the CDR3α loops position compared with the wild-type model (Fig. 7C and D). Conversely, as observed in a third MD simulation, the Arg95β residue could also be prevented from contacting the CDR3α loop by a neighboring Glu97β residue (Fig. 7E), indicating that neighboring residues in the CDR3β loop could potentially prevent Arg95β from contacting CDR3α in some settings.

In an effort to support our MD simulations, we also compared the crystal structure of a Vα10-Vβ8.1 murine iNKT TCR with our MD simulated Arg95β iNKT TCR models (41). The Vα10-Vβ8.1 murine iNKT TCR naturally contains the CASR sequence in its CDR3β loop. Similar to two of our Arg95β iNKT TCR MD simulated models, the third Arg of the CDR3β loop also contacts a main-chain carbonyl group (Supplementary Fig. 4). Based on these in silico observations, we propose that by contacting the CDR3α of the iNKT TCR and slightly re-orientating the CDR3α loop, the Arg at the third position of the CDR3β loop has the potential to affect the ability of human iNKT TCRs to recognize antigens.

Discussion

In this study, we have demonstrated that the third residue of Vβ11 CDR3 influences the antigenicity, antigen recognition, and TCRα chain structure of human iNKT TCRs. Ser or Arg at this position significantly altered the binding by two anti-Vα24 mAbs, clones C15 and 6B11. It is unclear where exactly clone C15 binds on the TCR Vα24 variable region, but clone 6B11 is specific for the Vα24-Jα18 CDR3 loop. Therefore, this result strongly suggests that the Arg-encoding CDR3β alters the conformation of the CDR3α region. Furthermore, the altered 6B11 staining for iNKT TCRs encoding Arg did not appear to be entirely due to masking of the epitope, since Cl.2037 had an increased 6B11 staining intensity upon mutating from Ser to Arg. Nevertheless, staining with these two anti-Vα24 mAbs was generally decreased in the CASR clones. It is important to note that while the C15 staining among CASS clones was fairly consistent, 6B11 staining was variable, even among the clones encoding CASS. Thus, although encoding Ser or Arg at the third position can influence 6B11 staining, other CDR3β residues also likely play a role in dictating the reactivity of this mAb to Vα24 and possibly the conformation of CDR3α.

Based on the modeling from previous human iNKT TCR crystal structures, we observed that the CDR3α and CDR3β loop conformations were affected by whether Ser or Arg was encoded at the third position of CDR3β. Specifically, the Arg residue at this position in the CDR3β was able to make direct interactions with CDR3α in two of the three MD simulated models we performed. Previous structural analyses of human iNKT TCRs also found interactions between the CDR3α and CDR3β residues, although it was not mediated by Arg at the same position as we have described here. Kjer-Nielsen et al described a Van der Waals interaction between Gly96 of CDR3α and Tyr101 of CDR3β (25). In a human iNKT TCR-CD1d-lysophosphatidylcholine trimolecular complex, López-Sagaseta et al observed several contacts between similar regions of CDR3α and CDR3β that were coordinated by water molecules (21). Therefore, CDR3β and CDR3α of human iNKT TCRs do form inter-loop interactions, which influence their respective structures.

Our finding has interesting implications for the indirect role of CDR3β in regulating antigen recognition. Numerous structural and mutational analyses have identified residues within CDR1α, CDR3α, and CDR2β important for the recognition of the α-GalcCer-CD1d complex (20, 23, 42–45). Since CDR3β is the only variable region among iNKT TCRs with the same Vβ usage and does not mediate direct contact with the ligand, it is thought that CDR3β variability is only able to influence the overall affinity toward CD1d-lipid complexes and not lipid specificity (20, 22, 23). Our data suggest that CDR3β could indirectly influence antigen recognition by altering the CDR3α loop structure in a sequence-dependent manner. This may be the mechanism by which certain clonotypic murine Vβ8.2 iNKT TCRs with unique CDR3β sequences can possess lipid-selectivity for human CD1d-self-lipid complexes (31). However, this effect might be only apparent for low-affinity interactions between iNKT TCRs and CD1d presenting weak lipid ligands but not potent agonists, such as α-GalCer and PBS-57.

Detection of iNKT cells using 6B11 and anti-CD3 mAbs tends to yield higher frequencies compared with detection using C15 and anti-Vβ11 mAbs (46, 47), which might be because not all Vα24i pairs with Vβ11. However, the analyses in those studies were not conducted in a pairwise manner. Montoya et al demonstrated that these two methods produced similar results when they were compared within each donor, albeit with a smaller sample size (11). Additionally, it is obvious that 6B11 cannot detect Vα24-Jα18-independent human iNKT cells, which can only be detected by the α-GalCer/PBS-57-loaded CD1d tetramer. In addition to these data, this report suggests that the variability in CDR3β sequences could be another element that affects the accuracy of detecting human iNKT cells using the 6B11 mAb.

Florence et al (42) and Scott-Browne et al (44) previously demonstrated that a glycine (Gly) to alanine (Ala) mutation at the third position of murine Vβ8.2 iNKT TCR CDR3 sequence did not influence α-GalcCer- or PBS-57-mouse CD1d (mCD1d) recognition, as measured by functional response and/or tetramer staining. Murine Vβ8.2 is homologous to human Vβ11, and Gly is typically encoded at this position by mouse Vβ8.2 compared with the Ser encoded by human Vβ11. In the same studies, the Gly to Ala mutant was found to moderately decrease reactivity against mCD1d presenting iGb3, which is a self-lipid for murine iNKT cells. These data are similar to what we observed for human iNKT TCRs when Ser was mutated to Arg. However, it is possible that the Gly to Arg mutation in murine iNKT TCRs may behave differently than Gly to Ala or Ser to Arg in humans.

We previously demonstrated that three CDR3β sequence motifs are associated with strong auto-reactivity of human iNKT TCRs (24). In this study, however, we found that a feature of the CDR3β amino acid sequence, CASR, was associated with weak auto-reactivity of iNKT TCRs. Additionally, the CASR sequence altered binding of Vα24-specific mAbs, especially clone 6B11, which binds the CDR3α region. In silico analysis also highlighted a role for the third CDR3β residue in regulating the CDR3β and CDR3α loop conformations. Taken together, our study provides evidence for the role of CDR3β in modulating the structure of the iNKT TCRα chain, which could influence the function of the overall TCR.

Abbreviations

α-GalCer

α-galactosylceramide

Ala

alanine

Arg

arginine

aAPC

artificial APC

Gly

glycine

iNKT

invariant natural killer T

Vα24i

invariant TCR Vα24-Jα18

MD

molecular dynamics

Ser

serine