Generation and molecular recognition of melanoma-associated antigen-specific human γδ T cells

See allHide authors and affiliations

Science Immunology  14 Dec 2018:
Vol. 3, Issue 30, eaav4036
DOI: 10.1126/sciimmunol.aav4036

γδ T cells come to the fore

Studies on T cells within the tumor microenvironment (TME) have largely focused on T cells that express αβ T cell receptors (TCRs). Using an in vitro culture system to expand melanoma antigen–specific T cells, Benveniste et al. have identified γδ T cells that recognize melanoma antigens in a class I MHC–restricted manner. To understand how these T cells recognize antigens in an MHC-restricted fashion, they crystallized one of these γδ TCRs complexed with cognate peptide, MHC, and β2 microglobulin. Given the success of T cell–centric therapies in cancer, the results presented here call for a closer examination of the role of γδ T cells in the TME of melanomas and other cancers.


Antigen recognition by T cells bearing αβ T cell receptors (TCRs) is restricted by major histocompatibility complex (MHC). However, how antigens are recognized by T cells bearing γδ TCRs remains unclear. Although γδ T cells can recognize nonclassical MHC, it is generally thought that recognition of antigens is not MHC restricted. Here, we took advantage of an in vitro system to generate antigen-specific human T cells and show that melanoma-associated antigens, MART-1 and gp100, can be recognized by γδ T cells in an MHC-restricted fashion. Cloning and transferring of MART-1–specific γδ TCRs restored the specific recognition of the initial antigen MHC/peptide reactivity and conferred antigen-specific functional responses. A crystal structure of a MART-1–specific γδ TCR, together with MHC/peptide, revealed distinctive but similar docking properties to those previously reported for αβ TCRs, recognizing MART-1 on HLA-A*0201. Our work shows that antigen-specific and MHC-restricted γδ T cells can be generated in vitro and that MART-1–specific γδ T cells can also be found and cloned from the naïve repertoire. These findings reveal that classical MHC-restricted human γδ TCRs exist in the periphery and have the potential to be used in developing of new TCR-based immunotherapeutic approaches.


Successful adaptive immune responses rely on the appropriate recognition of antigens by T lymphocytes (1). In the case of αβ T cells, their T cell receptors (TCRs) recognize processed antigen peptides presented by the major histocompatibility complex (MHC) (2). In contrast, γδ TCRs can recognize unprocessed and nonpeptide antigens that are not presented by MHC (1, 3, 4), and it has been proposed that antigen recognition by some γδ TCRs resembles that of immunoglobulin (1, 5).

Most γδ T cells present in human blood express a Vγ9/Vδ2 TCR (1), which can recognize tumor cells and microbial infections via nonpeptide phosphoantigens (pAgs) generated by the mevalonate pathway (1, 4). However, pAg-mediated stimulation of Vγ9/Vδ2 T cells was shown to be through binding of pAg to the internal domain of the butyrophilin-3A1 protein, rather than directly to the TCR (6). On the other hand, Vδ1-expressing γδ T cells are rare in the blood, representing less than 10% of all γδ T cells (7), but are found at higher frequency mostly in mucosal sites and can engage nonclassical MHC molecules on the surface of virally infected or tumor cells (1, 4, 8-10). γδ T cells, expressing either Vγ9/Vδ2 or Vδ1 TCRs, have been found within tumor-infiltrating lymphocytes (TILs) in melanomas (1, 7, 11-13). The significance of γδ T cells within TILs and the role of each subset in the evolution of the disease are largely unknown.

Here, we investigated whether human γδ T cells could recognize melanoma-associated antigens (MAAs) and achieve this specificity by engaging MHC-presented peptide antigens. To this end, we used an in vitro system to generate antigen-specific human T cells, which revealed that MAAs MART-1 and gp100 can be recognized by γδ T cells in an MHC-restricted fashion. Cloning and transferring of MART-1–specific γδ TCRs to T cells restored the specific recognition of the initial antigen MHC/peptide (MHCp) reactivity and conferred antigen-specific functional responses. A three-dimensional crystal structure of the complex between a MART-1–specific γδ TCR and HLA-A2 (human leukocyte antigen–A2)/MART-1 revealed distinctive but similar properties to those previously reported for αβ TCRs. Our work shows that MHC-restricted antigen-specific γδ T cells can be generated in vitro and that, similarly, MART-1–specific γδ T cells can also be found and cloned from the naïve repertoire of peripheral T cells. These findings reveal that classical MHC-restricted human γδ TCRs are present in the periphery and have the potential to be used in developing of new TCR-based immunotherapeutic approaches.


Human γδ T cells generated in vitro recognize MHCp antigens

To investigate whether γδ T cells can recognize tumor-associated antigens, we took advantage of an in vitro system to generate human T cells from hematopoietic stem/progenitor cells (HSPCs) (14-16), so that both αβ and γδ lineages of T cells could be interrogated. We reasoned that T cells generated in vitro from HSPCs, using the OP9-DL system (14, 17), would include cells reactive to self-antigens, such as MAAs (18). As previously reported (15), HSPCs cultured on OP9-DL4 cells can differentiate into functional CD8+ T cells expressing a broad TCR repertoire (16); therefore, MAA-specific T cells would be expected to be found in these cultures. CD34+ CD38 HLA-A*0201+ (HLA-A2) cord blood HSPCs cultured on OP9-DL4 cells differentiated into T-lineage cells, and CD7+ CD5+ CD1a+ cells were observed from day 19 and onward, with their frequency increasing over the culture period (fig. S1, top row). CD4+ CD8+ double-positive (DP) cells emerged on day 29, and by day 49, more than half of the cells present in the cultures were DP (fig. S1, middle row). CD4 CD8+ single-positive (SP) T cells became apparent by day 39, with some of these cells also expressing TCR/CD3 (fig. S1, bottom rows).

We directly examined for the presence of T cells that can recognize MAAs; in particular, we looked for MART-1 (Melan-A) or gp100 reactivity. Analysis of day 55 cocultures revealed the presence of HLA-A2/MART-1 (heteroclitic peptide)–specific cells, within the DP and CD8+ SP subsets (Fig. 1A, left panels, first column). These cells were observed after an enrichment for CD8α+ cells (Fig. 1A, right panels, second column). These findings show that MAA-specific CD8+ SP T cells can be generated from HSPCs in vitro.

Fig. 1 Generation of MAA-specific CD8

+ γδ T cells from cord blood HSPCs cultured on OP9-DL4 cells. (A) Flow cytometric analysis of day 55 HSPC/OP9-DL4 cell cocultures for the cell surface expression of CD4 and CD8α (top row) and enrichment by magnetic cell sorting of CD8+ cells is shown (right column). Analyses for CD8α expression and HLA-A*0201/MART-1 [heteroclitic (H)] dextramer staining of cells are shown in both total and CD8α-enriched populations, gated for either DP (middle row) or SP CD8 cells (bottom row). (B) Flow cytometric analysis of day 59 HSPC/OP9-DL4 cell cocultures for cell surface expression of CD4 and CD8α (top row) and enrichment by magnetic cell sorting of CD8+ cells is shown (right) for cells analyzed for αβ TCR, γδ TCR, and CD8β expression. Analyses for HLA-A*0201/MART-1 (heteroclitic) and HLA-A*0201/gp100 dextramer staining of CD8α-enriched populations, as well as for CD8β and γδ TCR expression, are shown. The above results are representative of at least 10 independent cocultures. (C) Cord blood cells were enriched for γδ TCR+ CD4 T cells by magnetic cell sorting from HLA-A2+ samples and stimulated for 18 days in vitro, as described in Material and Methods. Flow cytometric analysis of cultures for expression of Vδ1, Vγ9, γδ TCR, and HLA-A*0201/MART-1 (heteroclitic and WT), as indicated, shows cells gated for the expression of γδ TCR or αβ TCR. These results are representative of at least three independent cultures.

We further characterized the in vitro–generated CD8α+ T cells and observed that both αβ and γδ TCR–bearing cells were present (Fig. 1B, top row), with CD8β being exclusively expressed on αβ TCR+ cells. Analysis of CD8β+ T cells showed that a few of these cells were reactive to the MAAs MART-1 and gp100 (Fig. 1B). We also noticed the presence of γδ TCR+ cells that showed reactivity for these MAAs (Fig. 1B and fig. S2). These findings revealed that the repertoire of both αβ and γδ T cells generated in vitro had reactivity to HLA/peptide-dependent antigens.

Peripheral human γδ T cells recognize MHCp antigens

The generation of HLA-A2/MART-1–specific γδ T cells from HSPCs in vitro was unexpected and raised the key question as to whether γδ T cells with similar HLA/peptide reactivity are present in the periphery. To this end, we isolated naïve T cells from HLA-A2+ umbilical cord blood and, after enrichment for CD4 γδ TCR–expressing cells, stimulated them with MART-1 peptide–pulsed antigen-presenting cells (Fig. 1C). After 18 days, analysis of the cultures for the presence of γδ T cells showed that a majority were Vδ1+, whereas Vγ9+ made up a small fraction that did not appear to show any reactivity to HLA-A2/MART-1. However, among the γδ T cells, a clear but small subset of HLA-A2/MART-1–reactive cells was present. Similarly, αβ T cells also contained a small subset of HLA-A2/MART-1–specific cells. These findings show that γδ (Vδ1) T cells present in the normal naïve repertoire show specificity for HLA/peptide antigen recognition.

Cloning of human γδ T cells that recognize MHCp antigens

Classical MHC-restricted αβ T cell clones have been extensively characterized both structurally and functionally. Although, recognition of classical MHCp by γδ T cells has been reported (19), it is not understood how this interaction is mediated. Thus, the ability to grow and maintain γδ T cell clones with HLA-A2/MART-1 specificity was necessary to characterize how these cells recognize antigen and to validate whether this reactivity was TCR dependent (20). To this end, we generated MART-1–specific γδ T cell clones from both CD8α+ SP and DP subsets, as indicated in Fig. 1A (red circles). After single cell sorting, MART-1–specific T cell clones were derived from CD8+ SP at a higher plating efficiency than from DP cells (table S1). A total of 33 MART-1–specific γδ T cell clones were generated from three independent experiments; all of these showed positive staining with HLA-A2/MART-1 dextramers, but not with either gp100 or HIV peptide dextramers (Fig. 2A and fig. S3). All 33 clones expressed Vδ1, and none expressed Vδ2 and/or Vγ9 (Fig. 2B, showing four representative clones, and fig. S3, showing seven additional clones from three independent experiments). In summary, we were successful in identifying MART-1–specific γδ T cell clones from HPSC-derived T cells generated in vitro that belonged to the Vδ1 subset of γδ T cells. In addition, we were also able to expand MART-1–specific γδ T cells from adult peripheral blood mononuclear cells (PBMCs) (fig. S4) and cord blood naïve PBMCs (figs. S5 to S7), which were then used to derive Vδ1+ MART-1–specific γδ T cell clones (fig. S7 and table S2), with most clones displaying a CD8α+ phenotype. Furthermore, functional testing of MART-1–specific γδ T cells, either in vitro–generated clones (Fig. 2A) or naïve PMBC-derived lines (figs. S5E and S6), produced interferon-γ, but not interleukin-17 (IL-17), after stimulation (fig. S8A). Effector function proteins were expressed (granzyme B) and detected (CD107a) on MART-1–specific γδ T cells after stimulation (fig. S8B).

Fig. 2 Specificity of CD8

+ γδ T cell clones and their TCRs generated from cord blood HSPCs cultured on OP9-DL4 cells. (A) Flow cytometric analysis of CD8+ clones derived from HSPC/OP9-DL4 cell cocultures for the cell surface expression of HLA-A*0201/MART-1 (heteroclitic), HLA-A*0201/gp100, and/or HLA-A*0201/HIV dextramers, as indicated. HLA-A*0201/MART-1–specific αβ TCR cell clone (IC2) derived from ex vivo cord blood CD8 T cells was used as a positive control (top row). A total of 33 γδ T cell clones were analyzed, and 4 representative clones are shown. (B) Flow cytometric analysis of CD8+ T clones derived from HSPC/OP9-DL4 cell cocultures for the cell surface expression of γδ TCR, αβ TCR, Vδ1, and Vδ2, as indicated. (C) γδ TCR chains obtained from HSPC/OP9-DL4 coculture-generated γδ T cell clones (3C2 and 5F3) were retrovirally transduced into J76 (left column) or J76CD8α (right column). αβ TCR chains obtained from an ex vivo–derived T cell clone (49) were transduced into J76 (left column) or J76CD8αβ (right column). Flow cytometric analysis for HLA-A*0201/MART-1 dextramer staining is shown as histograms for each set of transduced cells, as well as untransduced controls, as indicated. RCN, relative cell number. Shown in untransduced histogram are unstained cells (light gray) and cells stained for HLA-A*0201/MART-1 dextramer (dark gray); for the other histograms, untransduced (light gray) and transduced (dark gray) cells stained for HLA-A*0201/MART-1 dextramer are shown. Frequencies of each populations based on unstained gated cells are indicated. Four γδ TCR clones derived from two independent experiments were analyzed. Each TCR chain was sequenced, and CDR3s are shown in table S2. (D) Specificity of transduced J76 T cells (3C2 or 49) was evaluated using flow cytometry with the indicated HLA dextramers for untransduced cells (top row), γδ TCR (middle row), and αβ TCR (bottom row). Shown for the untransduced panel are unstained cells (light gray) and cells stained as indicated (dark gray); for the other panels, untransduced (light gray) and transduced (dark gray) cells stained for the indicated HLA dextramers are shown. Representative analysis of one of the four transduced γδ TCR chains is shown (n ≥ 3). (E) Cytotoxic T cell assay of sorted activated cord blood PBMCs transduced with 3C2 γδ TCR, as described in fig. S12, showing specific percent lysis at the indicated effector to target ratios of MART-1– or gp100-pulsed T2 cell targets. Representative analysis of γδ TCR–transduced PBMCs is shown (n = 3). MART-1–speficic αβ T cell line, gp100-specific αβ T cell line, and nontransduced PBMCs are also shown.

Human γδ TCRs transfer recognition of MHCp antigens

To conclusively show that HLA-A2/MART-1 reactivity by the γδ T cells was TCR dependent, we cloned and sequenced the TCR-γ and TCR-δ chains of three independent CD8+ T cell clones generated in vitro; all TCR-δ chains contained Vδ1 segments, whereas two of the TCR-γ chains were Vγ8, and one was Vγ4 (table S3). Of note, the complementarity determining region 3 (CDR3) regions showed that all three TCRs were unique. Retroviral expression constructs encoding individual TCR-δ and TCR-γ chains from two γδ T cell clones (3C2 and 5F3) were transduced into Jurkat 76 (J76) or Jurkat 76 CD8α+ (J76CD8α) T cells (2). As shown in Fig. 2C, HLA-A2/MART-1 reactivity was seen by both 3C2 and 5F3 γδ TCR–transduced J76, as well as J76CD8α T cells, showing that the γδ TCRs were sufficient to confer the original MART-1 recognition to the J76 T cells.

As expected, untransduced J76 T cells only showed background reactivity to HLA-A2/MART-1 dextramers, whereas a control αβ T cell clone (49) showed the anticipated specific MART-1 reactivity (Fig. 2C). Although transferring the γδ TCRs to J76 or J76CD8α T cells restored HLA-A2/MART-1 reactivity, we further evaluated their specificity by testing for reactivity against a panel of different HLA/peptide dextramers. The γδ TCR (3C2)–reconstituted J76 (Fig. 2D) or J76CD8α (fig. S9) T cells were specific for HLA-A2/MART-1 but did not recognize either gp100 or influenza matrix protein (MP) peptides presented on HLA-A2. In addition, MART-1 [wild-type (WT) peptide] presented on HLA-B*3501 was not recognized by the γδ TCR–transduced J76 T cells. We also showed that the addition of HLA-A2/MART-1 dextramers to two T cell clones (3C2 γδ TCR and 1C2 αβ TCR) lowered the staining levels for TCR expression (fig. S10). These data suggest that recognition of MART-1 by γδ TCRs was specific for the HLA-A2/MART-1 complex and may be similarly constrained as the recognition by αβ TCRs, and compatible with the notion that recognition of MART-1 by γδ TCRs appears to be MHC restricted.

Human γδ TCRs transfer functional recognition MHCp antigens

Although the analysis using HLA dextramers strongly suggested that MAA-specific γδ TCRs recognize MART-1 in the context of HLA-A2, we further tested whether this recognition was sufficient to functionally activate T cells bearing these TCRs. γδ TCR–transduced J76 T cells specifically produced IL-2 in response to MART-1–pulsed T2 cells (HLA-A*0201+), whereas no response was apparent with other peptides tested (fig. S11). Similarly, αβ TCR–transduced J76 T cell showed the same specificity for MART-1, albeit higher levels of IL-2 were produced by these cells. Furthermore, we compared γδ TCR (3C2)–transduced activated cord blood PBMCs (fig. S12) to established control αβ TCR T cell lines (specific for MART-1 or gp100) and showed that 3C2-transduced T cells were able to lyse only MART-1 peptide–pulsed T2 cells, whereas gp100-pulsed T2 cells showed background killing (Fig. 2E). Thus, MART-1–specific γδ TCRs ectopically expressed by T cells (J76 or PBMCs) can transfer functional responsiveness in an HLA/peptide-dependent manner.

Molecular interaction of human γδ TCRs with MHCp antigens

To understand the molecular basis of the interaction of the 5F3 and 3C2 TCRs with HLA-A2/MART-1, we expressed recombinant forms of the extracellular domains of the TCRs and HLA-A2 either in insect cells or by refolding Escherichia coli–expressed inclusion bodies. Using biolayer interferometry (BLI), we demonstrated a direct interaction between the 5F3 and 3C2 TCRs with HLA-A2/MART-1, but no detectable interaction between HLA-A2 complexed with the control gp100 peptide (fig. S13). The 5F3 and 3C2 TCRs bind the refolded HLA-A2/β2M/MART-1 construct with distinctly different Kd values of 2.9 and 71 μM, respectively, exhibiting an about 20-fold difference in affinity. To probe the contribution of the TCR chains to the HLA-A2 interaction, we analyzed a TCR (named 1A6) generated from the δ chain of 5F3 and the γ chain of 3C2 and measured the affinity of this hybrid construct to refolded HLA-A2 using BLI. 1A6 bound HLA-A2/MART-1 with an affinity (9 μM) similar to that of 5F3, suggesting that the δ chain shared between 1A6 and 5F3 contributes most of the binding energy of the interaction with HLA-A2/MART-1. Had the γ chain predominated energetically, we would anticipate that the affinity of the 1A6 TCR would more closely resemble the 3C2 TCR. These affinities are within the range of those previously measured for Vδ1 γδ TCRs/ligand interactions with nonclassical MHC molecules CD1c and CD1d (21-23) and comparable to αβ TCR affinity to HLA-A2/β2M/MART-1 (24, 25). Furthermore, the apparent bias toward use of the Vδ1 domain for binding and docking, as revealed by the chain swapping experiment, is reminiscent of the Vδ1 domain bias in recognition of CD1d, wherein the γ chain contributes only minimally or not at all (22, 23).

Unexpectedly, affinity measurements of the TCRs with the insect cell–expressed, trimeric single-chain constructs of HLA-A2/β2M/MART-1 showed that whereas the 5F3 and 1A6 TCRs recognize this construct similarly to that of the refolded HLA-A2/β2M/MART-1, the 3C2 TCR does not (fig. S13C). This trimeric construct has the MART-1 peptide covalently linked to β2M that is then linked to the HLA-A2 heavy chain with glycine-serine linkers (26); these linkers are absent in the refolded HLA-A2 construct. We hypothesize that the failure of the 3C2 TCR to bind to the trimeric construct is due to obstruction by the glycine-serine linker connecting the MART-1 peptide and β2M, which would extend up and out of the HLA-A2 groove. This finding suggests that the docking orientation of the 3C2 TCR differs from that of the 5F3 TCR, which binds to both refolded and trimeric HLA-A2 constructs equally. A similar pattern was found in the binding results of the hybrid 1A6 TCR, which binds the trimeric version and refolded HLA-A2/MART-1 with similar affinities, suggesting that not only does the shared δ chain of 5F3 TCR contributes most of the binding energy but also it dictates the docking footprint. All three γδ TCRs did not bind CD1c, a known ligand for a subset of Vδ1 γδ T cells (21, 27), reflecting their specificity for HLA-A2. In addition, these TCRs show explicit binding specificity for the MART-1 peptide, because they did not interact with the irrelevant gp100 peptide (fig. S13, D and F), reflecting our results from functional assays presented earlier. Overall, our binding experiments provide conclusive biochemical evidence that these γδ TCRs recognize HLA-A2/MART-1 directly and specifically.

Crystal structure of a human γδ TCR with MHCp antigen

Generally, the structural information related to ligand recognition by γδ TCR is very limited in comparison to αβ TCR. Only three complex structures with nonclassical MHC-like molecules (T22 and CD1d) are available (22, 23, 28). To understand the molecular basis of γδ TCR recognition of classical MHC, here, HLA-A*0201, we crystallized the 5F3 TCR in complex with single-chain trimeric HLA-A2/β2M/MART-1 and solved the structure to 2.75 Å (table S4). The TCR and HLA-A2/MART-1 proteins were well resolved in the electron density maps and revealed the TCR docked onto HLA-A2 in a diagonal binding orientation, where the two TCR chains straddle the MART-1 peptide, reminiscent of the classical docking mode of αβ TCRs on MHCp (29), similar to that of the HLA-A2/MART-1–specific MEL5 TCR, but distinctly different from a Vδ1 γδ TCR recognition of CD1d/sulfatide complex (Fig. 3A) (22). The docking angle difference between 5F3 and MEL5 is ~22°, well within the range of docking angles noted for αβ TCRs that recognize HLA-A2/MART-1 (Fig. 3B). Curiously, in the 5F3/HLA-A2/MART-1 complex structure, the CDR1 and CDR2 loops of the δ chain are positioned over the α1 helix of HLA-A2, and the CDR1 and CDR2 loops of the γ chain are positioned over the α2 helix, with both CDR3 loops centrally positioned over the peptide. Although direct homology between the γ, δ, α, and β gene segments is unclear, the α and δ gene segments are embedded in the same chromosomal locus (30), sometimes share V segments (31), and therefore are generally considered to be homologous. With this interpretation, the docking orientation of the 5F3 would then be considered to be in a flipped (more than 180°) orientation to that of a classical CD8+ αβ TCR, which docks generally with the β chain over the α1 helix and the α chain over the α2 helix. However, interpreting the structure from a different perspective, both δ and β chains use D segments during rearrangement, which generates a highly diverse (and typically longer) CDR3 loop in these chains. The orientation that we observe in the 5F3 TCR positions the highly diverse CDR3δ in a similar position to that of the CDR3β loop, suggesting that the 5F3 orientation may be following docking principles inherent in αβ TCR recognition of MHC.

Fig. 3 Three-dimensional complex structure of the 5F3 TCR with HLA-A*02/MART-1.

(A) Overall crystal structure of 5F3/HLA-A*02/β2M/MART-1 peptide complex (upper left) with the γ and δ chains of the 5F3 TCR colored cyan and magenta, respectively. HLA-A*02, β2M, and MART-1 are colored white, wheat, and yellow, respectively. Shown also is the complex structure of HLA-A*02/β2M/MART-1 with the αβ TCR MEL5 (PDB code: 3HG1; middle) with the α and β chains of the MEL5 TCR colored light green and dark brown. The right panel is the complex structure of CD1d-sulfatide with the γδ TCR DP10.7 (PDB code: 4MNG), where γ and δ chains are colored cyan and magenta, respectively. The lower panels show the positions of the TCR CDR loops on HLA-A*02/MART-1 with the squares indicating the docking orientation. The dashed square in the case of CD1d/DP10.7 denotes that the γ chain does not contact CD1d or the sulfatide ligand. (B) Comparison of the positioning of the CDR loops of the 5F3 γδ TCR and MEL5 αβ TCR on HLA-A*02/MART-1 (left), coloring scheme same as in (A); the numbers 1, 2, and 3 denote CDR loops. Right panel shows comparison of the docking angle of the 5F3 γδ TCR with that of the HLA-A*02/MART-1–reactive αβ TCRs MEL5, DMF4, and DMF5. Spheres represent the position of the conserved intrachain disulfide cysteine residues in the TCR variable domains, and the dashed black lines represent the vectors connecting them. (C) MART-1 peptide recognition by the CDR1γ (cyan) and CDR3δ (magenta) loops. TCR interacting residues are showing as sticks and are labeled. Peptide residues contacted by CDR3δ only are colored yellow, which contacted by both CDR1γ and 3δ loops, and interacting residue is colored orange. Zoomed in view of MART-1 recognition is shown in the inset. 2Fo-Fc electron density maps (contoured at 1σ) for Y32, W98, and D99 are shown as blue mesh.

All six CDR loops of the 5F3 TCR are used in engaging the HLA-A2/MART-1 composite surface (Fig. 3B and table S5). In the contacts with HLA-A2, the CDR loops of δ chain exclusively interact with the α1 helix, and contacts of the γ chain are biased toward the α2 helix. The total buried surface area (BSA) of the TCR/HLA-A2/MART-1 interface is ~2127 Å2, a value that is in the higher range of αβ TCR/MHC complexes, with the BSA distributed between the individual δ and γ chains (59 and 41% BSA, respectively). From the MHCp, nearly one-fifth of the BSA (19%) was contributed by the peptide, whereas HLA-A2 accounted for 81%. In the complexes of the HLA-A2/MART-1–reactive TCRs, MEL5, DMF4, and DMF5 (24, 25), there is more BSA with the peptide, ranging from 29 to 34% (table S6), suggesting the 5F3 TCR may be slightly less peptide centric than these αβ TCRs.

Of the six CDR loops of the 5F3 TCR, only the CDR3δ and CDR1γ loops make contacts with the peptide. The junctionally encoded CDR3δ residues W98 and D99 are involved in multiple contacts (Fig. 3C and fig. S14), with the aromatic side chain of W98 providing an anchoring role through insertion between the centrally twisted region of MART-1 peptide and the α1 helix of HLA-A2. The W98 side chain forms van der Waals contacts with the A3, G4, and I7 residues of MART-1, and D99 is positioned centrally over the peptide and forms one hydrogen bond with G4 and van der Waals contacts with I5 and G4 residues (Fig. 3C, inset). Tyrosine-32 is the only residue from the γ chain that contributes to peptide recognition through a van der Waals interaction with I5 (Fig. 3C, fig. S14, and table S5). Similarly, in the MEL5 αβ TCR/HLA-A2 complex, an analogous CDR1α residue, Q31, interacts extensively with the MART-1 peptide.

Mutational footprinting analysis of human γδ TCRs with MHCp antigens

Both the 5F3 and 3C2 TCRs use the Vδ1 and Vγ8 domains in their TCRs, only differing in the sequences of their CDR3 loops. To compare the docking strategies of these two TCRs, we determined their energetic footprints by using alanine-scanning mutagenesis of the residues in the HLA-A2 α helices. Seventeen amino acid residues that are fully exposed on the α1 and α2 helices of HLA-A2 were individually mutated to alanine, and these mutants were refolded with the MART-1 peptide and used in BLI measurements with the γδ TCRs. WT HLA-A2 was also used as a control, and the effect of the mutation was calculated as a ratio of mutant Kd to WT Kd. This analysis indicated that the 5F3 and 3C2 TCRs interact differently with HLA-A2 (Fig. 4). Residues R65, K68, R75, and V76 are critical for 5F3 TCR as alanine mutation of these residues resulted in more than fourfold reductions of affinities, consistent with the contacts observed in the crystal structure. Q72A and E154A mutants have more than a twofold effect on affinities, owing to their multiple interactions with the 5F3 TCR (Fig. 4A, left panel). The mutational analysis resulted in a different profile of interactions for the 3C2 TCR, where mutants R75A, Y84A, and H151A resulted in more than twofold reduction, V76A resulted in more than threefold reduction, and K146A and T163A resulted in more than fourfold reduction (Fig. 4A, right panel).

Fig. 4 Footprint mapping of the 5F3 and 3C2 γδ TCRs via alanine-scanning mutagenesis of HLA-A*02 α1 and α2 helices.

(A) Graph showing the binding results of alanine-scanning mutagenesis. The bars represent the fold change in affinity between the 5F3 TCR (left) and 3C2 TCR (right) with mutant and WT HLA-A*02 (Mut Kd/ WT Kd). Mutations with more than threefold reduction in affinity are colored red, mutations with two- to threefold reduction in affinities are colored orange, mutations with one- to twofold reduction in affinity are colored yellow, mutations with no effect on binding are colored blue, and mutations with less than onefold changes are colored green. Plus sign indicates more than fourfold reduction. Binding analysis was performed twice, and average Kd was used for analysis. (B) The results from (A) are mapped onto a surface representation of the platform domain of HLA-A*02 for the 5F3 TCR (left) and 3C2 TCR (right). HLA-A*02 is colored white, and MART-1 is colored yellow. The mutant residues in α1 and α2 helices are mapped, labeled, and color-coded as shown in (A) on the basis of fold changes in binding due to alanine mutation. (C) Proposed model of the footprints of 5F3 and 3C2 TCRs are shown on the HLA-A*02 as colored ellipses as predicted from the alanine-scanning mutagenesis. Dashed black line indicates the position of linker connecting the MART-1 and β2M in the single-chain construct.

Mapping of a color-coded representation of these mutational results onto a surface representation of HLA-A2 provides a view of the docking footprints of the 3C2 TCR in relation to that of 5F3 (Fig. 4B). The energetic hotspot of 5F3 TCR is mostly focused on the α1 helix, reflecting the dominant role played by the Vδ1 domain of 5F3 TCR in MHC recognition (Fig. 4B, left panel). The energetic hotspot of 3C2 TCR differs from that of the 5F3 TCR and is instead distributed across the α1 and α2 helices near the C terminus of the peptide (Fig. 4B, right panel). This finding is consistent with the inability of the 3C2 TCR to engage with the trimeric HLA-A2/β2M/MART-1 construct, because the linker between the C terminus of the peptide and β2M as hypothesized would disrupt the 3C2 TCR interaction (Fig. 4C). Our TCR footprinting analysis indicates that these γδ TCRs, despite using identical V gene segments in their TCRs, use different docking strategies to recognize the same HLA-A2/peptide antigen complex.


Human T cells can be generated from HSPCs in vitro, and these cultures are able to give rise to both αβ and γδ T cell lineages. Here, we show that antigen-specific human γδ T cells can be generated in vitro, which can recognize MAAs MART-1 and gp100 in an MHC-restricted fashion. Cloned γδ TCRs were able to transfer MART-1 specificity to T cells, allowing for the specific recognition of the initial antigen MHCp reactivity and conferring antigen-specific functional responses. Structural analysis of crystals from a MART-1–specific γδ TCR, together with MHCp, revealed distinctive binding properties to those previously reported for αβ TCRs. This work establishes that MHC-restricted antigen-specific γδ T cells can be generated in vitro and that, similarly, MART-1–specific γδ T cells can also be found and cloned from the naïve repertoire of peripheral T cells.

Our molecular analyses demonstrate a direct interaction between γδ TCRs and HLA-A2/MART-1 with affinity ranges within that documented for agonist αβ TCR/MHCp interactions. The crystal structure of the MART-1–specific γδ TCR 5F3 provides a molecular snapshot of how γδ T cells can recognize classical MHCp. The docking orientation is notably similar to αβ TCR recognition of the same MHCp complex, with a convergent focus of the diverse, D segment containing CDR3δ and β loops on the same location of the MART-1 peptide. Our footprinting analyses of the 5F3 and 3C2 TCRs show that multiple docking strategies are used in γδ TCR binding despite bearing the same Vγ and Vδ domains, suggesting that the sequences of the CDR3 loops factor strongly into how the TCRs engage the MHCp composite surface. Whether the 3C2 TCR adopts a variation on the diagonal docking orientation awaits further structural analysis.

Although our work revealed the presence of MHC-restricted HLA-A2/MART-1–specific γδ T cells in both naïve cord blood and adult PBMCs, it is still unclear as to when these cells normally arise, and to which tissues, if any, these T cells traffic and/or reside remains to be elucidated. In addition, whether γδ T cells that recognize MHCp-dependent antigens have a role in immune surveillance of melanomas (or other cancers), and importantly whether they have immunomodulatory functions, also remains to be examined.

In the thymus and in the periphery, αβ T cells recognize antigens in the context of self-MHC—a property named MHC restriction by Doherty and Zinkernagel (32-34). Although this phenomenon has been deemed unique to αβ T cells, the exact nature of antigen recognition by γδ T cells is still largely unknown. It is unclear how many defined γδ T populations exist in humans and how these cells recognize and respond to their ligands through their TCR (1, 4). Documented cases of γδ T cell recognition reveal that γδ TCRs can recognize antigen directly, but other examples demonstrate the requirement for MHC molecules (nonclassical or MHC-like), either alone or with presented antigen, such as the case with the lipid-presenting MHC-like proteins CD1c and CD1d. Cases for recognition of classical MHCp are few, usually limited to a single T cell clone, and none have been characterized at the molecular level. Here, we show that γδ T cells with MHC-restricted reactivity to MAAs can not only be generated from HSPC in vitro but are also present among cord blood and adult PBMCs. Elucidating whether MHC restriction applies more broadly to γδ T cells has critical implications for our general understanding of the surveillance function of these cells and for designing potential cell-based immunotherapies. To this end, the in vitro generation of γδ T cells with MAA specificity offers an alternative approach for the application of tumor-targeted T cells.


Umbilical cord blood samples and OP9-DL4 culture system

OP9-DL4 cells were generated and maintained as previously described (17). Human umbilical cord blood samples were obtained by syringe extraction and collected in a blood-pack unit containing citrate phosphate dextrose anticoagulant (Baxter Healthcare, Deerfield, IL) from consenting mothers after delivery in accordance to approved guidelines established by the Research Ethics Board of Sunnybrook Health Sciences Centre. Within 24 hours of collection, cord blood mononuclear cells were isolated as previously described (35). Sorted CD34+ CD38 cord blood stem cells were plated on OP9-DL4 cells and cultured as previously described (35).

Generation of MART-1–reactive clones

Sorted CD34+ CD38 cord blood stem cells cultured on OP9-DL4 cells for 45 to 57 days. At this time, cells were first CD8α-enriched by magnetic-activated cell sorting (MACS; Miltenyi) and subsequently stained with heteroclitic MART-1 26-35 (26 ELAGIGILTV 35/HLA-A*0201) dextramer. Positive cells were sort-deposited for at one cell per well into 96-well plates. Cells were stimulated with 8 × 104 irradiated PBMCs from five different normal donors, 3 × 104 K21 and K80 (provided by C. June) (36), artificial antigen-presenting cells (aAPC), phytoagglutinin (PHA; 5 μg/ml), and IL-2 (100 IU/ml). Wells were examined for cell growth on days 18 to 21 after plating, and cells were expanded using stimulators described above every 3 to 4 weeks. Heteroclitic MART-1 TCR αβ clones and cell lines as well as gp100-specific CD8+ T cell lines were generated from ex vivo cord blood samples following the same approach.

Generation of γδ TCR cells from cord blood naïve or adult PBMCs

Twelve individual HLA-A*201+ cord blood samples, or two sets of 10 pooled HLA-A*201+ cord blood PBMCs, and one individual HLA-A*201+ adult PBMC, were first analyzed for the presence of cells expressing γδ TCR, Vδ1, and MART-1 or gp100. On day 0, PBMCs were then enriched for γδ TCR+ using Miltenyi (γδ TCR microbeads kit, catalog number 130-050-701). Enriched γδ TCR cells (60 to 89%) were then stimulated with irradiated MART-1 (10 μM)–pulsed T2 cells (1:1 ratio), plus irradiated K21 (4:1) and irradiated K80 (4:1) aAPCs in the presence of IL-7 (5 ng/ml) and IL-2 (100 IU/ml). After 3 weeks, growing cells were analyzed for γδ TCR–, Vδ1-, MART-1–, or gp100-expressing cells, as well as CD4 and CD8α expression, and CD4+ cells were removed by MACS. The cells were then sorted twice, sequentially for Vδ1+ and MART-1+ cells by MACS, and cultured for an additional 5 to 10 days either as a line or manually plated at 0.3 cells per well in 96-well plates. Cloning and expansion of Vδ1+ MART-1+ cells were achieved culturing the cells with irradiated PBMCs derived from at least five different donors (8 × 104 cells per well), irradiated K21 and K80 aAPCs (3 × 104 cells per well), and PHA in the presence of IL-7 (5 ng/ml) and IL-2 (100 IU/ml). Wells were refed after 2 weeks, and clones were analyzed by flow cytometry for the presence of Vδ1+ and MART-1+ cells after 3 weeks. Positive clones were then used for functional assays.

Crystallization, x-ray crystallographic data collection, molecular replacement, and refinement

Single-chain HLA-A*02/β2M/MART-1 and the 5F3 TCR were expressed in insect cells as described above. The purified HLA-A*02/β2M/MART-1 protein was treated with endoglycosidase-F3 at 37°C for 2.5 hours to remove N-linked glycosylation. Histidine-tagged HLA-A*02/β2M/MART-1 was purified by Ni–nitrilotriacetic acid (NTA) agarose column, and endoglycosidase-F3 was removed by washing with Hepes buffer saline (HBS). In next step, the 8× histidine tag was removed by treating with carboxypeptidase (Sigma-Aldrich) for overnight at room temperature (22°C). EDTA (1 mM) was added to protein sample to quench the protease activity, and last, the protein was purified with Superdex 200 size-exclusion chromatography. 5F3 TCR was expressed in Hi5 cells as described earlier. After Ni-NTA purification, TCR was treated with 3C protease overnight at 4°C to remove zipper and histidine tag. TCR was passed through the Ni-NTA column to remove any undigested TCR. TCR was treated with endoglycosidase-F3 to remove glycosylation for 2.5 hours at 37°C. Last, TCR was purified using Mono S cation exchange and Superdex 200 size-exclusion chromatography (GE Healthcare). HLA-A*02 and 5F3 fractions were checked in SDS–polyacrylamide gel electrophoresis for purity. HLA-A*02/β2M/MART-1 and 5F3 TCR were mixed together at an equimolar concentration. Proteins were concentrated to 5.5 mg/ml at 4°C and screened for crystals using sitting drop vapor diffusion method combining 0.5 μl of protein with 0.5 μl of reservoir solution in a 96-well plate (Douglas Scientific). Diamond-shaped crystals were obtained in 0.1 M tris (pH 8.5) and 23% (v/v) poly(ethylene glycol) methyl ether 550. Crystals were run on a gel to confirm for HLA-A*02/β2M/MART-1 and 5F3 complex. Crystals were cryocooled in mother liquor solution supplemented with 20% glycerol before cryocooling at 100 K liquid nitrogen. X-ray datasets were collected on Pilitus detector at beamline 24ID-C, Advanced Photon Sources, Argonne National Laboratory. Data were indexed, integrated, and scaled with HKL2000. HLA-A*02 [Protein Data Bank (PDB) code: 2AV7] without any ligand and 9C2 TCR (PDB code: 4LHU) without the CDR sequences were used as search models for molecular replacement in Phaser in CCP4i (37, 38). The solution generated one HLA-A*02 molecule adjacent to one 5F3 TCR molecule, and symmetry analysis showed that 5F3 TCR is on top of HLA-A*02 molecule and recognizing the peptide through the CDR loops. Initially, rigid body and restrained refinements were carried out in Phenix (39). The iterative model building was performed in Coot (40), peptide and N-acetylglucosamine (NAG) were introduced in the model, and refinement including translation/liberation/screw was carried out for the final model. All the refinements were performed by taking random 5% reflections and excluding them for Rfree. Contacts were analyzed with Contact/N-contact program in CCP4. Buried surface was analyzed with AREAIMOL in CCP4i and PDBePISA ( All the figures were generated with PyMOL (Schrödinger Scientific).


Materials and Methods

Fig. S1. Generation of human CD8 T cells from cord blood OP9-DL4 cell cocultures.

Fig. S2. Generation of human MART-1+ γδ TCR T cells from cord blood HSPCs cocultured with OP9-DL4 cells.

Fig. S3. Generation of human MART-1+ γδ TCR cells and clones from cord blood HSPCs cocultured with OP9-DL4 cells.

Fig. S4. Derivation of human MART-1+ γδ TCR T cells from adult PBMCs.

Fig. S5. Derivation of human MART-1+ γδ TCR T cells from cord blood naïve PBMCs.

Fig. S6. Derivation of human MART-1+ γδ TCR T cell line from pooled cord blood naïve PBMCs.

Fig. S7. Derivation of human MART-1+ γδ TCR T cell clones from individual cord blood PBMCs.

Fig. S8. Functional characterization of MART-1+ γδ TCR T cell clones generated from cord blood HSPCs cocultured on OP9-DL4 cells and from MART-1+ γδ TCR T cell lines derived from cord blood naïve PBMCs.

Fig. S9. Specificity of γδ TCR–transduced J76CD8α T cells.

Fig. S10. Cold target inhibition of γδ TCR staining by HLA/peptide dextramers.

Fig. S11. Functional characterization of TCR-transduced J76 T cells.

Fig. S12. Transduction and isolation of PBMCs ectopically expressing 3C2 γδ TCR.

Fig. S13. Binding analysis of human γδ TCRs to MHCp using BLI.

Fig. S14. Contribution of germline and recombined residues of the 5F3 γδ TCR in recognition of HLA-A*02/MART-1.

Table S1. Generation of γδ TCR CD8α+ T cells clones with specificity for HLA-A*0201/MART-1 from HSPC/OP9-DL4 cell cocultures.

Table S2. Derivation of γδ TCR CD8α+ T cells clones with specificity for HLA-A*0201/MART-1 from naïve PBMCs.

Table S3. Cloning and sequencing of three in vitro–derived γδ TCR T cell clones.

Table S4. Data collection and refinement statistics of HLA-A*02-MART-1/5F3 TCR complex.

Table S5. Contacts residues between γδ TCR and HLA-A*02.

Table S6. Percentage of BSA contribution of complex components.


Acknowledgments: We are grateful to M. Fyrsta for expert assistance with the cytotoxic T lymphocyte (CTL) assays and to E. Kobylecky for assistance with cell cultures. We are thankful to G. Awong for knowledgeable assistance with flow cytometric cell sorting. We thank the staff of the beamline for use and assistance with x-ray beamline and, in particular, K. Rajashankar for help and advice during data collection. Funding: This work was supported by grants from the Canadian Institutes of Health (CIHR; MOP-42387 and FDN-154332), NIH (P01AI102853), and the Krembil Foundation to J.C.Z.-P. and by NIH (RO1AI073922) to E.J.A. J.C.Z.-P. is supported by a Canada Research Chair in Developmental Immunology. X-ray data were collected at the Northeastern Collaborative Access Team beamlines (P41 GM103403), with a Pilatus 6 M detector on 24-ID-C beamline (S10 RR029205) at the Argonne National Laboratory (DE-AC02-06CH11357). Author contributions: P.M.B. designed and performed all experiments related to the generation and characterization of human γδ T cells and wrote parts of the paper. S.R. designed and performed all experiments related to the molecular and structural characterization of human γδ TCRs and wrote parts of the paper. M.N. generated the retroviral TCR expression constructs and helped with Jurkat experiments. E.L.Y.C. performed cytokine secretion experiments. L.N. and D.G.M. performed CTL assays. P.S.O. and N.H. provided expertise and critical reagents. E.J.A. and J.C.Z.-P. provided funding, wrote the paper, conceptualized the experimental design, and guided the work. Competing interests: The authors declare that they have no competing interests. A provisional patent will be filed describing the generation and cloning of the human γδ TCRs. Data and materials availability: The coordinate and structure factor files of the HLA-A*02/β2M/MART-1/5F3 TCR complex structure have been deposited in the PDB (code: 6D7G).
View Abstract

Stay Connected to Science Immunology

Editor's Blog

Navigate This Article