Invariant NKT (iNKT) cells develop and differentiate in the thymus, segregating into iNKT1/2/17 subsets akin to Th1/2/17 classical CD4+ T cells; however, iNKT TCRs recognize Ags in a fundamentally different way. How the biophysical parameters of iNKT TCRs influence signal strength in vivo and how such signals affect the development and differentiation of these cells are unknown. In this study, we manipulated TCRs in vivo to generate clonotypic iNKT cells using TCR retrogenic chimeras. We report that the biophysical properties of CD1d–lipid–TCR interactions differentially impacted the development and effector differentiation of iNKT cells. Whereas selection efficiency strongly correlated with TCR avidity, TCR signaling, cell–cell conjugate formation, and iNKT effector differentiation correlated with the half-life of CD1d–lipid–TCR interactions. TCR binding properties, however, did not modulate Ag-induced iNKT cytokine production. Our work establishes that discrete TCR interaction kinetics influence iNKT cell development and central priming.
Invariant NKT (iNKT) cells are innate lymphocytes that express semi-invariant TCRs that recognize (glyco)lipid Ags presented by the MHC class Ib molecule CD1d (1). Their ability to secrete large amounts of cytokines and chemokines within minutes following stimulation allows them to boost innate and adaptive responses. Consequently, iNKT cells play protective or deleterious functions in diseases such as microbial infection, autoimmunity, allergy, and cancer (1). iNKT cells develop in the thymus where they are selected on double-positive (DP) thymocytes through agonist TCR signals and Slam/SAP-mediated signals, which lead them to acquire a memory-like phenotype and fast/potent effector functions (2). Upon or after positive selection, iNKT cells are primed and differentiate into iNKT1, iNKT2, and iNKT17 subsets in the thymus (3), as opposed to classical Th1/2/17 CD4+ T cells that differentiate in peripheral lymphoid tissues upon Ag encounter. Although molecules such as mechanistic target of rapamycin, lymphoid enhancer factor 1, and inhibitor of DNA binding proteins have been shown to influence the frequency of these subsets, the mechanisms of thymic effector differentiation remain largely unknown (2, 4).
iNKT cells potently release both Th1 and Th2 cytokines independently of IL-4R and IFN-γR expression (5), yet certain stimuli such as IL-12 can induce IFN-γ production by iNKT cells independently of the TCR (6, 7). Therefore, whether iNKT cells get polarized and the mechanisms for this remain unclear. Several glycolipid ligands can skew the overall immune response toward Th1 or Th2. However, these ligands do not polarize the cytokines released by iNKT cells per se, and none of the biophysical properties of the TCR–ligand interaction correlates with this bias (8–12).
iNKT cell development and function are highly dependent on signals through their TCRs, which are composed of an invariant Vα-chain (Vα14-Jα18 in mice or Vα24-Jα18 in humans), paired with a diverse set of Vβ-chains (Vβ8s, Vβ7, and Vβ2 in mice or Vβ11 in humans). Unlike classical αβ TCRs, iNKT TCRs dock parallel and above the C-terminal side of the CD1dα helices such that only the invariant Vα-chain makes contact with the lipid Ag (13, 14). The Vβ-chain only makes contact with CD1d, mainly through CDR1β and CDR2β. The hypervariable CDR3β loop is not essential for binding but it has been shown to make contacts with CD1d in some instances and to compensate for low-affinity CDR1/2β (13, 14). This mode of interaction is very well conserved among different species, and it likely contributes to the promiscuous ligand recognition of iNKT cells (15).
The kinetics of TCR–ligand interactions play an important role in T cell selection (16–18), activation (19), effector differentiation (20), and memory formation (21, 22); however, the mechanisms of such processes are not fully understood. Several models have been proposed, with the fundamental ones being the occupancy and kinetic proofreading models. The occupancy model predicts that signal strength will reflect the number of TCRs that are engaged, and therefore will correlate with the dissociation constant (KD) of TCR–ligand interactions (23). Alternatively, the kinetic proofreading model predicts that TCR signaling correlates with the dissociation rate (KOFF) such that the amount of signaling will reflect the time a TCR is engaged to its ligand, only reaching downstream events when it remains bound for a long enough time (23, 24).
Many studies have modeled T cell function based on their TCR kinetics, yet whether and how properties of CD1d–lipid–TCR interactions lead to differential signal strength in vivo and their functional implications on iNKT cell development and effector differentiation remain elusive. A link between TCR signal strength and iNKT cell development has only been suggested following the observations that the avidity conferred by certain Vβ frameworks correlated with their prevalence within the iNKT repertoire (25), and that iNKT2s express higher levels of promyelocytic leukemia zinc finger (PLZF), Nur77, and Vβ7 (3, 26, 27). In the present study, we directly addressed how CD1d–lipid–TCR interactions influence iNKT cell fate in vivo. We manipulated iNKT TCRs in vitro and in vivo through the generation of TCRβ retrogenic chimeras. We show that iNKT cell selection is governed by the avidity of CD1d–lipid–TCR interactions and is not restricted by the thymic niche. Alternatively, TCR signaling, the ability to form cell–cell conjugates, and iNKT cell effector differentiation are dictated by CD1d–lipid–TCR half-life. Interestingly, despite the modulation of effector fate, TCR binding parameters did not influence the quality of the cytokine response in short term in vivo stimulation. Our results show that distinct biophysical parameters of TCR-ligand interactions control the selection and effector differentiation of iNKT cells.
Materials and Methods
Mice and reagents
Mice were used between 5 and 8 wk of age. C57BL/6 wild-type (WT), TCRβ−/−, and Nr4a1-eGFP/Cre mice were purchased from The Jackson Laboratory. CD1d−/− mice were generated and provided by Dr. Chyung-Ru Wang (Northwestern University) (28). Jα18-deficient mice were provided by Dr. Laurent Gapin (University of Colorado Denver). All strains were housed at the Division of Comparative Medicine, University of Toronto animal facility under specific-pathogen free conditions, and animal procedures were approved by the Faculty of Medicine and Pharmacy Animal Care Committee at the University of Toronto (animal use protocols 9003, 9571, 10135, 10715, and 11113).
α-galactosylceramide (αGalCer; KRN7000) was purchased from DiagnoCine. Abs used were purchased from eBioscience, BioLegend, or BD Biosciences. PBS57-loaded and unloaded biotinylated CD1d monomers were obtained from the National Institutes of Health Tetramer Core Facility. PBS44, PBS218, and PBS221 tetramers were made as previously described (29). Unloaded monomers were loaded with lipids (PBS44, PBS218, PBS221) in 0.2 M malonate, 0.2% Tween 20 overnight at 37°C in the presence of 0.2 mg/ml saposin B (provided by Dr. Luc Teyton). Monomers were tetramerized by addition of fluorochrome-conjugated streptavidin. For stimulation assays, mCD1d monomers were purified from the culture supernatant of transduced HEK293 cells lines obtained from the National Institutes of Health Tetramer Core facility, using affinity chromatography.
PBS44 was synthesized as described previously (30). For PBS218 and PBS221, trimethylsilyl-protected donors 1 and 2 were prepared in quantitative yields by treating d-glucose and d-galactose, respectively, with chlorotrimethylsilane in the presence of imidazole. Glycosylation of acceptor 3 or 4 was performed according to the reported procedure (31). Crude products were sequentially treated with acidic hydrogen resin and sodium methoxide. Silica gel chromatography afforded the desired products PBS218, PBS221, and PBS44. Structures and purities were verified using proton and carbon nuclear magnetic resonance and high-resolution mass spectrometry.
Cloning, cell lines, and retroviral infection
TCR constructs were generated by overlapping PCR according to published methods (32, 33) and cloned into mouse stem cell virus-based plasmids with an internal ribosome entry site plus sequence encoding for GFP (MIGR1) or ametrine (pMIAII) (a gift from Dr. Dario Vignali). CD3 multicistronic vectors were obtained from Dr. Dario Vignali (33). The TCRαβ−/− 5KC-78.3.20 hybridoma was used to generate a CD4− and CD1dlo 6KC hybridoma. 6KC hybridomas and mouse WT3 fibroblasts (34) were transduced with different retroviral constructs by spinfection at 4500 × g, 37°C for 90 min. Transduced cells were sorted for similar levels of TCR expression. Mouse Bcl2 and human BclXL templates were obtained from Dr. Cynthia Guidos (The Hospital for Sick Children, University of Toronto) and cloned downstream of a cleavable 2A peptide DNA fragment by overlapping PCR. Retroviruses were generated using jetPRIME (Polyplus Transfection) transfection reagent and used to spinfect cells at 4500 × g for 90 min at 37°C in retrovirus-containing supernatants supplemented with Polybrene (8 μg/ml). Cells were sorted on a FACSAria II (BD Biosciences) based on expression of TCR and the reporter gene.
Tetramer equilibrium and decay measurements
These assays were done as previously described (35). For equilibrium staining, 1 × 105 cells were stained with anti-TCRβ Abs and different concentrations of CD1d–lipid tetramers at 22°C for 3 h in PBS containing 0.5% FBS, 2 mM EDTA, and 0.1% sodium azide. Cells were then washed and fixed using Cytofix/Cytoperm (BD Biosciences) for 30 min at 4°C and analyzed by flow cytometry. For the tetramer decay assay, 0.5 × 105 WT3 cells were stained with CD1d–lipid tetramers for 45 min at 22°C. Cells were then washed and resuspended in 100 μl of 100 μg/ml anti-CD1d purified Ab 1B1 (BioLegend) for various amounts of time. Cells were then fixed and analyzed by flow cytometry on an LSRFortessa X-20 (BD Biosciences).
Cell–cell conjugation assay
CD1d-sufficient and -deficient thymocytes were stained with either PKH26 red fluorescent cell linker (Sigma-Aldrich) or eFluor 450 proliferation dye (eBioscience) in PBS for 20 min at 22°C. Thymocytes were washed twice with complete RPMI medium and mixed in a 1:1 ratio. Total thymocytes (2 × 106) were incubated with 0.1 × 106 hybridomas expressing a specific TCR in complete RPMI supplemented with 10% FBS, centrifuged for 1 min, and incubated at 37°C for 30 min. Cells were immediately fixed with addition of 2 vol Cytofix/Cytoperm and incubated at room temperature for 30 min, washed, and analyzed by flow cytometry on an LSRFortessa X-20.
Hybridoma stimulation assays
Flat-bottom 96-well plates were coated with 10 μg/ml purified mCD1d in PBS at 37°C for 1 h, washed with PBS, and the lipid ligands were then added at different concentrations in 0.1 M malonate and 5 μg/ml saposin B and incubated overnight at 37°C. Bone marrow (BM)–derived dendritic cells (BMDCs) were differentiated for 8 d in GM-CSF–conditioned medium. For cell stimulation assays, lipid Ags were added to the medium for 4 h and cells were washed prior to stimulation. Hybridomas (0.1 × 106) were stimulated with different amounts of C57BL/6 WT thymocytes, BMDCs, 0.1 million sorted thymocytes, or immobilized CD1d–lipid complexes for 2 h at 37°C. Stimulation was stopped by addition of cold buffer and the cells were stained for TCRβ, fixed using a Foxp3 staining kit (eBioscience), stained for Nur77, and analyzed by flow cytometry.
The protocol was based on previously described methods (32). TCRβ−/−, TCRβ−/− Nr4a1-eGFP/Cre, or TCRβ−/−CD1d1/d2−/− mice were injected with 5-fluorouracil (0.15 mg/g weight). BM was collected 4 d later from the femur and tibia and cultured overnight in complete IMDM supplemented with stem cell factor, IL-6, IL-3, and Flt-3 ligand. Cells were transduced with retrovirus-containing supernatants supplemented with Polybrene (8 μg/ml) by two successive spinfections at 650 × g for 90 min at 34°C, with a 30-min incubation at 37°C between spin infections. The following day, transduction efficiency was assessed by flow cytometry, and a minimum of 0.3 × 106 transduced cells were injected i.v. into lethally irradiated (2 × 550 rads) Jα18−/− recipient mice. Mice were analyzed 6–8 wk after reconstitution.
In vivo stimulations
Retrogenic mice were injected with 0.5 μg of αGalCer, and spleen and liver cells were collected after 90 min. Cells were stained for extracellular proteins, fixed and permeabilized using Cytofix/Cytoperm buffer, stained for cytokines, and analyzed by flow cytometry on an LSRFortessa X-20.
Flow cytometry data were analyzed using FlowJo (Tree Star). Statistical analysis was performed using Prism (GraphPad Software). Statistical tests are indicated for each figure and were selected based on the normality test for each data set. TCRβ sequencing results were analyzed using an immunoSEQ analyzer (Adaptive Biotechnologies).
Interaction kinetics of iNKT TCRs toward CD1d–lipid complexes display differential hierarchy of avidity and half-life
To address the role of CD1d–Ag–TCR interaction in iNKT cell development and function, we selected six TCRs that have been described to interact with CD1d–αGalCer complexes with a wide range of avidities (25, 36). Four TCRs, Vβ8.2 DO-11.10 (hereafter referred to as 8-DO) (37), 2C12 (8-2C) (38), DN32.D3 (8-DN) (39), and 24.8.A (8-24) (40), contain the TCR Vβ8.2 framework and differed only in the CDR3β-Jβ region (Table I). The additional two TCRs contain the same CDR3β-Jβ region as 8-DO, but in the context of Vβ7 or Vβ2, and are referred to as 7-DO and 2-DO, respectively (Table I). Of note, the 8-DN, 8-2C, and 8-24 TCRs were originally identified from the natural iNKT cell repertoire, whereas 8-DO, 7-DO, and 2-DO were engineered (25, 36). We reasoned that this selection would allow us to perform an in-depth analysis of the effect of the TCRβ-chains on iNKT cell development and function.
|.||CDR1β .||CDR2β .||CDR3β .||Jβ .||References .|
|8-DO||NNHNN||SYGAGS||GSGTTN||NTEVFFGKGTRLTVV||Vβ8.2 DO-11.10 (37)|
|7-DO||MSHET||SYDVDS||GSGTTN||NTEVFFGKGTRLTVV||Vβ7 DO-11.10 (25)|
|2-DO||NSQYPW||LRSPGD||GSGTTN||NTEVFFGKGTRLTVV||Vβ2 DO-11.10 (25)|
|.||CDR1β .||CDR2β .||CDR3β .||Jβ .||References .|
|8-DO||NNHNN||SYGAGS||GSGTTN||NTEVFFGKGTRLTVV||Vβ8.2 DO-11.10 (37)|
|7-DO||MSHET||SYDVDS||GSGTTN||NTEVFFGKGTRLTVV||Vβ7 DO-11.10 (25)|
|2-DO||NSQYPW||LRSPGD||GSGTTN||NTEVFFGKGTRLTVV||Vβ2 DO-11.10 (25)|
To characterize the kinetics of CD1d–lipid–TCR interaction, the TCRs were expressed in a TCRα−β− hybridoma or WT3 fibroblasts expressing the CD3 complex and sorted for similar TCR expression. Using CD1d tetramers loaded with the prototypical ligand PBS57, we calculated the avidity as the 1/EC50 of equilibrium staining (Fig. 1A, left panels) and the half-life as the t1/2 of tetramer binding (Fig. 1A, right panels). These assays have been shown to reflect the KD and KOFF of TCR–ligand interactions, respectively (35). As expected, we found that the TCRβ-chain greatly modulated the strength of interaction between the TCRs (Fig. 1A). However, the avidity of the TCRs was not a straight reflection of their half-life for PBS57 tetramer binding (Fig. 1A). For example, 8-DN showed relatively high avidity but dissociated much faster than did most other TCRs (Fig. 1A). Alternatively, 8-24 had one of the lowest avidity but an intermediate half-life (Fig. 1A). Of note, 8-2C, but none of the other TCRs, was autoreactive, as it bound to CD1d tetramers without addition of exogenous Ags (data not shown).
Previous work has proposed that the variability in the CDR2β and CDR3β loops broadly modulates the strength of CD1d–lipid–TCR interaction (25). A recent study, however, has challenged this idea and proposed that pairing of specific Vβ-Jβ segments allows the TCR to discriminate between ligands rather than fine-tuning the overall strength of recognition (41). To address this, we determined the avidity and half-life of the TCRs for CD1d tetramers loaded with additional ligands (Fig. 1B). We selected one additional αGalCer (PBS44) and two α-glucosylceramides (PBS218 and PBS221), which have recently been identified as the major endogenous α-linked glycolipids in mammals (29, 42) (Fig. 1B). PBS218 and PBS221, which have the same glucose headgroup but different lipid tails, had very similar binding parameters, indicating that the stability of the ligands within CD1d did not play a major role (Fig. 1B). In contrast, lipids with galactose headgroups generally had higher avidity and longer half-life than those with glucose headgroups (Fig. 1B). Importantly, the hierarchy of binding of the TCRs was the same regardless of the ligand, indicating that the TCRs did not exhibit preferential ligand recognition (Fig. 1B). Overall, this shows that the panel of TCRs recognizes each ligand in the same relative order of strength, but the hierarchy of avidity (8-2C > 8-DO ≥ 7-DO ≥ 8-DN > 8-24 ≥ 2-DO) differs from that of dissociation (8-2C > 8-DO > 8-24 ≥ 7-DO > 8-DN > 2-DO) (Fig. 1).
TCR avidity controls iNKT cell selection efficiency in vivo
To assess the impact of the TCR on iNKT cell development, we expressed the TCRβ-chain of each TCR in vivo using the retrogenic chimera approach (32, 33) (Supplemental Fig. 1A). We used TCRβ−/− BM cells as donors and Jα18−/− recipient mice, which allowed natural rearrangement of the TCRα-chain and ensured that iNKT cells arise only from the grafted BM (Supplemental Fig. 1A). After reconstitution, retrogenic mice had robust production of DP and single-positive (SP) thymocytes, indicating normal T cell development (Fig. 2A, Supplemental Fig. 1B–H).
All chimeras contained iNKT cells, albeit in various amounts (Fig. 2A–C). Classical T cell selection has been proposed to follow a general bell-shaped distribution with a sharp threshold for negative selection (18). Some studies attributed the affinity/KD as the defining parameter for selection (43), whereas others have suggested that the half-life/KOFF dictates selection (16). Both positive selection and negative selection have been shown to occur for iNKT cells (25, 44). Given that all ligands tested induced the same hierarchy of binding, we used PBS57–CD1d tetramers as representative of all ligands. For each TCR we plotted the frequency of thymic and splenic iNKT cells against the avidity or the half-life of CD1d–PBS57–TCR interaction (from Fig. 1A) and fit the data to a log-Gaussian distribution (Fig. 2D, 2E, Supplemental Fig. 2A, 2B). This revealed that iNKT cell selection associated better with avidity than half-life, suggesting that the overall numbers of TCRs that get engaged play a more important role in iNKT cell selection than does the duration of TCR engagement.
Of note, the relative frequency of iNKT cells with a given TCR was similar in the liver as well as inguinal and mesenteric lymph nodes, indicating that the TCR did not affect the migration of the cells in the periphery (Supplemental Fig. 2C). Importantly, no iNKT cells developed in any of the TCRβ chimeras that used TCRβ−/−CD1d−/− BM (Supplemental Fig. 3A), and DPs in all chimeras had similar expression of CD1d and SLAMF1/F6 (Supplemental Fig. 3B). Because forced expression of an autoreactive TCRβ-chain can lead to the development of naive iNKT cells with Vα14-Jα18 TCRα variants (44), we sequenced the Vα14-Jα18 joining region from single sorted splenic TCRβ+tetramer+ cells from all chimeras and found that all retrogenic iNKT cells contained the canonical Vα14-Jα18 chain (data not shown).
Overall, these experiments confirmed that bona fide iNKT cells develop in retrogenic mice, and their selection is governed more prominently by the avidity, rather than the half-life, of the TCR for CD1d–lipid complexes.
Frequency of clonotypic iNKT cells does not depend on survival, proliferation, or competition
Selection events during T cell development are closely linked with survival (45). Hence, we investigated whether low iNKT cell output was a consequence of increased apoptosis by staining ex vivo cells with Annexin V. iNKT cells in all chimeras showed similar levels of Annexin V staining regardless of their TCR (Fig. 3A). In parallel, we generated retrogenics in which the antiapoptotic factors Bcl-2 or BclXL were overexpressed together with the low avidity TCRs 8-24 or 2-DO. Although Bcl-2– and BclXL-containing constructs increased the frequency of classical T cells in these chimeras, they did not improve iNKT cell selection efficiency (Fig. 3B). Taken together, these results suggest that the lower iNKT cell thymic output observed for low-avidity TCRs is not due to increased cell death. Of note, using the same approach with the autoreactive 8-2C TCR revealed that BclXL overexpression (but not Bcl-2) improved thymic iNKT cell frequencies (data not shown), supporting the notion that these iNKT cells start to undergo negative selection.
Newly selected iNKT cells have been shown to undergo a rapid wave of expansion (46). Thus, we assessed whether TCR avidity influenced proliferation by staining for Ki-67, a marker of cells in the active phase of the cell cycle (47). iNKT cells in all chimeras had comparable Ki-67 expression, with the exception of 8-DN relative to 8-24 chimeras (Fig. 3C). However, the 8-DN TCR selects iNKT cells efficiently (Fig. 2B). Therefore, we conclude that the lower iNKT cell output in our chimeras is not mediated by impaired proliferation.
Similar to iNKT cells, regulatory T cells develop through agonist selection. It has been shown that their selection is instructed by the TCR and limited by a small thymic niche (48, 49). We considered the possibility that iNKT cells with low-avidity TCRs only developed due to absence of competition. We tested this by creating dual retrogenic mice through transfer of TCRβ−/− BM expressing the 8-24 (low avidity) or 8-DO (high avidity) TCRs and Ametrine or GFP as reporter, respectively, into Jα18−/− recipient mice. We found that 8-DO iNKT cells did not outcompete 8-24 iNKT cells, and the frequency of iNKT cells relative to Ametrine or GFP+ cells was similar to the noncompetitive setting (Fig. 3D). From this, we conclude that iNKT cell clones expressing TCRs of varying avidity do not compete for a limited thymic niche.
TCR half-life governs signal strength of iNKT TCRs
We next analyzed how TCR interactions influence signaling on iNKT cells in vivo. For this, we assessed expression of CD5 and Nur77, both of which reflect TCR signal strength (50, 51). To measure expression of Nur77, we generated TCRβ−/−Nur77-GFP mice and used them as donors for retrogenic chimeras. Both CD5 and Nur77 expression showed differences among the TCRs (Fig. 4A–D, left panels). In contrast to the association between selection efficiency and TCR avidity, CD5 expression correlated with the TCR half-life but not with TCR avidity (Fig. 4A, 4B, middle and right panels). Nur77 expression correlated better with half-life than with avidity of tetramer binding, yet it did not reach significance with either (Fig. 4C, 4D, middle and right panels). Because TCR expression level can affect the signaling outcome, we analyzed TCRβ expression in retrogenic mice. This showed that only 8-24 thymic iNKT cells had elevated levels of TCR (Fig. 4E). To correct for this, we reanalyzed the correlations in the thymus excluding this TCR (Fig. 4F, 4G). In this scenario, although CD5 levels correlated with binding avidity, the correlation with half-life was stronger (Fig. 4F). Nur77 expression only correlated significantly with half-life (Fig. 4G). Taken together, our results indicate that the half-life of CD1d–lipid–TCR interactions contributes to downstream signaling more prominently than does avidity.
TCR half-life governs effector differentiation of iNKT cells
We determined the distribution of iNKT cell subsets in TCRβ retrogenic mice by staining for PLZF and retinoic acid–related orphan receptor γt (3). Although the three subsets were found in all chimeras, we observed clear shifts in their distribution (Fig. 5A–C). In line with previous studies (3, 50), we found that the frequency of iNKT2 correlated with CD5 and Nur77 expression levels in the thymus (Fig. 5D). In the spleen, CD5 but not Nur77 expression correlated with the frequency of iNKT2 cells (Fig. 5E).
We then correlated both the avidity and half-life of tetramer binding with the frequency of iNKT1 and iNKT2 cells in the thymus and spleen (Fig. 5C–I). As with CD5 and Nur77, t1/2 of tetramer binding showed a stronger positive and negative correlation for iNKT2 and iNKT1, respectively, than did 1/EC50 (Fig. 5F–I), although in the thymus this correlation was only significant when we corrected for unequal TCR expression by excluding 8-24 retrogenic cells (Fig. 5J, 5K). Although, iNKT17 frequency did not correlate with avidity or half-life, the 8-DN and 2-DO TCRs, which have the shortest half-life, contained the most prevalent iNKT17 populations (Fig. 5B, 5C, Supplemental Fig. 4A).
A former classification of iNKT cells defined a linear progression from stage 1 (CD44−NK1.1−) to stage 2 (CD44+NK1.1−) and stage 3 (CD44+NK1.1+) (52), where stage 3 is equivalent to iNKT1 and stages 1 and 2 are a mixture of iNKT2, iNKT17, and progenitors (3). As expected, 8-2C and 8-24 iNKT cells had a higher frequency of stage 1 and 2 whereas 8-DO, 8-DN, and 2-DO had a higher frequency of stage 3 (Supplemental Fig. 4B). Importantly, the relative proportion of stage 3 between 8-DO and 8-24 was similar in single and dual (competitive) chimeras (Supplemental Fig. 4C), suggesting that intrinsic factors drive iNKT cell effector differentiation. Expression of CD69, CD122, and NKG2D, which are expressed at higher levels on iNKT1 cells (stage 3), were elevated in the 8-DO, 8-DN, 7-DO, and 2-DO chimeras (Supplemental Fig. 4D–F). The fact that expression of iNKT stages and other markers agrees with the distribution of subsets in retrogenic mice substantiates an important, cell-intrinsic role for the TCR in the differentiation of iNKT cells.
Because the 8-2C and 8-24 TCRs shared the Jβ2.5 gene segment, we considered the possibility that sequence elements of some Jβ segments could bias differentiation toward iNKT2 cells. To explore this, we sorted iNKT cells subsets from pooled WT thymi based on differential expression of CD4, CD27, and NK1.1 and sequenced their TCRβ repertoire. We observed increased Vβ7 (TRBV29) usage by iNKT2 cells, as previously reported (3), yet there was no particular bias in Jβ2.5 (data not shown). Interestingly, the TCR containing the Vβ7 framework in this study did not have the highest iNKT2 cell frequency, reinforcing the importance of the CDR3β in the selection of Vβ7+ TCRs with elevated TCR half-life in the WT repertoire (data not shown).
Overall, these data show that iNKT cell effector differentiation depends on the amount of time their TCRs are engaged.
TCR activation in vitro is influenced by the context of lipid presentation
We hypothesized that two non–mutually exclusive mechanisms could explain the different TCR biophysical parameters involved in selection and differentiation. First, TCR sensitivity for self-antigens could change from pre- to postselected stages. Second, selection and differentiation could be mediated by different self-ligands in the thymus. Because of the overall low frequency of iNKT cells, it was not feasible to test our first hypothesis on retrogenic iNKT cells. However, we explored the second hypothesis with the TCR hybridomas used for kinetic binding assays. Although these cells may not entirely reflect physiological signaling of iNKT cells, they can be used to understand how changes in stimuli affect activation outcomes.
We first measured Nur77 upregulation following stimulation with PBS44, PBS218, and PBS221 loaded on plate-bound CD1d, ligands for which we had calculated the avidity and half-life (Fig. 6A). Only 8-2C (high avidity and long half-life) and 2-DO (low avidity and short half-life) TCRs showed significantly higher and lower 1/EC50, respectively (Fig. 6A). All TCRs with intermediate avidity and half-life upregulated Nur77 to the same extent, indicating that upon stimulation with immobilized CD1d–glycolipid complexes, high avidity can compensate for short half-life and vice versa (Fig. 6A). Importantly, as with avidity and half-life measurements (Fig. 1), the hierarchy of TCR potency was the same for all ligands used, supporting that none of the TCRs preferentially recognized a particular ligand.
The potency of ligands is influenced by their kinetics of trafficking within APCs. We loaded either WT thymocytes or BMDCs with a high concentration (5 μg/ml) of PBS44, PBS218, or PBS221 and used increasing APC/hybridoma ratios to stimulate TCR-expressing hybridomas (Fig. 6B, 6C). In contrast to the plate-bound setting, the hierarchy of activation was not the same for every ligand. PBS44 was much more potent than either PBS218 or PBS221 with all TCRs except 8-24, reaching a similarly high 1/EC50 (Fig. 6B, 6C). Because PBS44 shares the same lipid tail with PBS221, this difference is likely mediated by the interaction with the TCR rather than stability of the lipid in CD1d. In support of this, TCR activation displayed essentially the same pattern for both α-glucosylceramides (PBS218 and PBS221), with the exception of increased 7-DO activation toward PBS221-loaded thymocytes (Fig. 6B, 6C). This suggests that presentation in the context of cells could have dramatic effects on iNKT cell activation by directly affecting the CD1d–lipid–TCR interface.
TCR binding and activation by self-lipids reflects the half-life of CD1d–lipid–TCR
Based on the above results, we assessed the TCR binding strength and potency toward CD1d–self-antigen complexes presented by developmentally relevant cells and tested whether it recapitulated the hierarchy of avidity or half-life seen in vivo. First, we measured the ability of each TCR-expressing hybridoma to form conjugates with fluorescently labeled CD1d+/+ and CD1d−/− thymocytes (Fig. 7A). The extent of conjugate formation correlated well with the half-life but not the avidity of the TCR for PBS57–CD1d complexes (Fig. 7A). Second, we used thymocytes as APCs without addition of exogenous Ags as an unbiased approach to measure signal strength toward self-antigens (Fig. 7B). As with the conjugate formation, the potency of the TCR toward thymocytes correlated better with TCR half-life than avidity for CD1d–PBS57 (Fig. 7B). This indicates that half-life better predicts TCR-mediated cell–cell interactions and TCR signal strength.
Given that most iNKT2 cells are present in the medulla of the thymus (53), we entertained the possibility that selection and differentiation are compartmentalized events subjected to unique sets of ligands. When bulk thymocytes are used for Ag presentation, these signals might be diluted out. Therefore, we sorted TCRβ−/lo and TCRβhi thymocytes as well as medullary DCs, which are found in the cortex, inner cortex/medulla, and medulla, respectively. Each cell type induced different levels of activation, with TCRβ+ cells being the most potent stimulators (Fig. 7C). In each case, Nur77 upregulation correlated better with TCR half-life than avidity (Fig. 7D, 7E). This suggests that the discrepancy between selection efficiency and effector differentiation of iNKT cells cannot be explained by activation through sets of ligands expressed in different thymic compartments.
The iNKT TCR does not control the quality of the response in vivo
Although iNKT cell subsets preferentially secrete Th1, Th2, or Th17 cytokines, this response is not absolute given that many iNKT cells produce both IL-4 and IFN-γ, and the extent of their activation is influenced by their location within lymphoid organs (53). Given that our set of TCRs induced different frequency of iNKT cell subsets, we predicted that they would also polarize the iNKT cytokine response upon Ag stimulation. To test this, we challenged the 8-DO, 8-DN, 8-24, and 7-DO retrogenic chimeras, which express TCRs with intermediate avidity and half-life, with αGalCer or PBS221 and analyzed production of IL-4 and IFN-γ after 90 min. All splenic (Fig. 8) and hepatic (data not shown) iNKT cells produced high amounts of cytokines, with a prominent population of IFN-γ and IL-4 double-producing cells. However, we did not observe any bias toward either IL-4 or IFN-γ with either ligand, indicating that although the TCR can influence the differentiation of the cells, it does not skew them toward a particular cytokine in short-term stimulation (Fig. 8).
To the best of our knowledge, this study provides the first comprehensive in vivo characterization of how TCR binding parameters influence iNKT cell development and effector differentiation. iNKT cell development is constrained by positive and negative selection events (25, 44). Our work shows that within this developmental window, the TCR binding avidity and half-life differently control these biological outcomes. However, TCR binding kinetics did not influence iNKT cell subset tissue distribution or cytokine production by iNKT cells, indicating that the TCR controls iNKT cell central priming but not Ag-mediated polarization in the periphery (8–12).
Although some studies investigated the impact of differential TCR Vβ gene usage (3, 12, 25, 26), they were confounded by the repertoire variability found within the CDR3β region, making it difficult to draw conclusions. We accounted for this by using a selection of TCR variants in both the CDR2β and CDR3β loops, as well as displaying a wide range of avidities and half-lives for CD1d–lipid complexes. We found that both the CDR2β and CDR3β regions influence iNKT cell selection efficiency as a result of differential TCR avidity, supporting previous findings (25). Although one study found that the CDR3β was dispensable for positive selection, the binding parameters of the TCRs used had not been assessed (54).
Our study supports the view that iNKT TCRs exhibit extensive ligand promiscuity (25, 37). Indeed, the comparison of ligands with differences in both the acyl chain and headgroup revealed that regardless of the ligand, the relative TCR avidity, half-life, and hybridoma activation were the same for the six TCRs tested. Of note, the activation hierarchy of these TCRs was different between cell-free and cell-mediated stimulation. This suggests that properties of cell–cell interactions have an important role in iNKT cell activation. In agreement with this idea, a recent study showed that actin-dependent formation of large CD1d nanoclusters on APCs is crucial for iNKT cell autoreactivity (55). This may also reconcile the identification of some ligand-specific iNKT TCRs in a recent study (41) in which lipid-loaded APCs were used for TCR identification. In our study, binding strength and potency of TCRs toward self-antigens presented by developmentally relevant cells were reflective of the half-life of interaction. Therefore, although presentation of Ags by cells can alter the hierarchy of TCR potency, half-life was a good predictor of activation by self-lipids.
Although support for both the occupancy model and the kinetic proofreading models is found in the literature, a concrete consensus is still missing and many other models have emerged to explain experimental discrepancies (23). Surprisingly, our experiments show that whereas selection is reflective of the avidity of the TCR, effector differentiation correlated better with half-life. Therefore, using iNKT cells we demonstrated that TCR signaling is not universally dependent on one kinetic parameter. Modeling TCR signaling for classical T cells has been challenging due to the inability to directly compare unique TCR–pMHC interactions with structural differences in docking modes, involvement of the CD4/8 coreceptors, as well as the use of ligands or TCRs where the KOFF directly correlates with the KD. Indeed, Palmer and colleagues (56) recently suggested that the rate-limiting step during TCR triggering involves the recruitment of active lck-coupled CD4/8 coreceptors. In contrast to classical T cells, iNKT TCRs are promiscuous, the TCR docking strategy onto CD1d–lipid complexes is strikingly conserved, and iNKT cells are considered coreceptor-independent (13, 14). Therefore, alternative models are needed to explain TCR signal strength on iNKTcells. It has been proposed that extensive rebinding takes place at the TCR synapse, and hence a model where the dwell time of interaction takes into account the association constant (KON) is more appropriate (57, 58). It is unlikely that this model is suitable for both selection and differentiation of iNKT cells given that the hierarchy of TCRs is different for each process. Additionally, selection and differentiation are likely not dependent on discrete sets of ligands because the hierarchy of TCR strength was the same toward different populations of thymic cells. Rather, we propose that iNKT cell selection and effector differentiation are subjected to differential TCR sensitivity. Conventional T cell positive selection signals involve slow Ca2+ accumulation over time from transient interactions of developing thymocytes with cortical thymic epithelial cells, whereas negative selection results from sustained interactions with DCs accompanied with strong/fast Ca2+ burst (18, 59). Furthermore, changes in TCR sensitivity have been reported as a result of changes in surface levels and clustering of the TCR (60, 61) as well as expression of phosphatases (62) and other components of the TCR signaling pathway (63, 64). It is conceivable that preselected iNKT cells experience short-lived interactions with DPs and these signaling events rely more on the avidity of CD1d–lipid–TCR complexes, which ultimately reflects the number of receptors engaged. Alternatively, differentiation likely relies on long cell–cell interactions and is more sensitive to differences in TCR half-life. Our work contributes to the understanding of how TCR signals induce fate decisions, which is crucial for the rational design of T cell therapies.
We thank Dr. Heather Melichar, Dr. Manfred Brigl, Dr. Tania Watts, and Dr. Michele K. Anderson for the critical review of the manuscript.
This work was supported by Canadian Institutes of Health Research Grant MOP -114911 and by Canada Foundation for Innovation Physical Infrastructure Grant 29186 (to T.M.). M.C.T. was supported by a Natural Sciences and Engineering Research Council of Canada doctoral research award. T.M. is supported by a Canada Research Chair in NKT Cell Immunobiology.
The online version of this article contains supplemental material.
The authors have no financial conflicts of interest.