Distinct Requirements for CD1d Intracellular Transport for Development of Vα14 iNKT Cells¹

Ag presentation via CD1 molecules is important in the protection against viral and bacterial pathogens, against autoimmune diseases and for effective development of antitumor responses (1, 2, 3). CD1 molecules are assembled in the endoplasmic reticulum from CD1 H chain, β2-microglobulin and endogenous lipids. Incorporated lipids are thought to stabilize the CD1/β2-microglobulin complexes until exchange for antigenic lipids is accomplished, analogous to the role of invariant chain (Ii)³-derived peptide in class II MHC assembly (4, 5, 6). Lipid-loaded CD1 arrives at the plasma membrane within 30 min after assembly and is rapidly introduced into the endocytic pathway, mediated by a tyrosine-based sorting motif present in the CD1 cytosolic tail or through interaction with invariant chain/class II MHC complexes (7, 8, 9). Recycling through the endocytic pathway from the cell surface results in exchange of lipid Ags, as CD1 molecules transit back and forth between the plasma membrane and lysosomes.

CD1d controls the development and function of a subset of strictly thymus-derived lymphocytes called NKT cells. Among the CD1d-restricted NKT cells are those that express a relatively invariant (i) TCR α-chain (Vα14-Jα18 in mice; Vα24-Jα18 in humans) coupled with a restricted subset of TCR β-chains (Vβ8.2, Vβ7, and Vβ2 in mice, and Vβ11 in humans), and those that express diverse TCRs. These α-galactosylceramide (αGalCer)- reactive Vα14 iNKT cells constitute most of the NK1.1⁺ T cell population in the spleen and represent >80% of the NK1.1⁺ T cells in the thymus and liver (10, 11). The development of NKT cell precursors diverges from that of mainstream thymocytes at the CD4⁺CD8⁺ double-positive (DP) stage (12, 13, 14, 15, 16). Upon random rearrangement and expression of the canonical Vα14 chain coupled with a restricted TCR β receptor, these DP thymocytes can be positively selected on neighboring CD1d-expressing CD4⁺CD8⁺ cortical thymocytes (17, 18, 19, 20). Such positively selected Vα14 iNKT cell precursors are prompted to mature following a sequence from CD24⁺CD44^lowNK1.1⁻ via CD24⁻CD44^lowNK1.1⁻ and CD24⁻CD44⁺NK1.1⁻ to CD24⁻CD44⁺NK1.1⁺ (12, 15, 16, 21, 22). Studies have shown that the periphery is primarily seeded with immature CD44⁺NK1.1⁻ Vα14 iNKT cells, which eventually mature outside the thymus (12, 16). At the final maturation step NK1.1 and other NK lineage receptors (such as Ly49G2, Ly49C, Ly49I, CD94, NKG2D, Ly6C, and 2B4) are expressed (12, 15, 16, 21, 22). The transition from NK1.1⁻ to NK1.1⁺ Vα14 iNKT cells is important for terminal differentiation of iNKT cells and requires the presence of CD1d in the periphery (20, 23). A developmental block at this stage is linked to major deficiencies in Vα14 iNKT cell function, as for example, in diabetes-prone NOD mice in which selective deficiency of NK1.1⁺ Vα14 iNKT cells predisposes to onset of autoimmune diabetes (24, 25).

Peripheral iNKT cell maturation involves professional APCs, in which CD1d endocytosis may occur via the CD1d intrinsic tyrosine motif or via association with Ii. Although Vα14 iNKT cells fully depend on an intact CD1d cytosolic tail for self-Ag recognition during thymic selection (9, 26), in the periphery, Vα14 iNKT cells are exposed to CD1d-positive cells that express invariant chain (i.e., professional APCs, including dendritic cells (DCs)), and cells that do not express invariant chain, in nonprofessional APCs such as peripheral T cells. Strong reactivity of iNKT cells for CD1d/self-glycolipid complexes results in their removal from the iNKT cell repertoire, as was shown using transgenic mice that overexpress CD1d and in fetal thymic organ cultures containing the high affinity ligand αGalCer (27, 28). In those experiments, the CD1d endosomal trafficking requirements for peripheral maturation could not be judged because postselection thymocytes did not survive. Mutant mice in which CD1d is expressed only on cortical thymocytes (driven by the Lck promotor) and professional APCs (driven by the I-Eα promotor) develop Vα14 iNKT cells that undergo normal selection and terminal differentiation (20). In those mice, postselection maturation of iNKT cells in the periphery could be contributed by either CD1d-tail-mediated endocytosis or Ii-assisted endocytosis. We generated a new mouse model in which all CD1d is expressed as CD1d-enhanced yellow fluorescent fusion protein (EYFP) fusion proteins. CD1d-EYFP molecules exhibit normal subcellular distribution and CD1d-EYFP molecules can support development of Vα14 iNKT cells, albeit at a reduced rate. Using these mice, we clarified the roles that the CD1d-intrinsic sorting motif plays in CD1d-mediated positive selection, postselection terminal maturation, and stimulation of matured Vα14 iNKT cells.

Six- to eight-week-old mice on C57 Bl/6 background were used throughout the study. Wild-type (WT) mice were purchased from The Jackson Laboratory. CD1d-deficient mice were described previously (29). CD1d-EYFP/EYFP mice were generated using homologous recombination in embryonic stem cells (Bruce 4 line). The genomic sequence encoding exons 1–6 and part of the 3′ untranslated region were used as homologous regions. A linker sequence between exon 6 and EYFP was inserted using the following primer: ACC GGT CCT CCT CCA GCA GGA CTC CTG GGA CAG CCG GTC GCC ACC.

For mixed bone-marrow chimeras, 6-wk-old CD1d^−/− hosts were subjected to two consecutive doses of 600 Rads irradiation with 2 h in between. Within 8 h after the second irradiation, irradiated mice were injected retro-orbitally with 1 × 10⁷ bone marrow cells isolated from femurs of donor mice (1:1 mix of CD45.1 CD1d^−/− : CD45.2 CD1d^+/+ or CD45.1 CD1d^−/− : CD45.2 CD1d-EYFP/EYFP). All chimeras showed a ratio of CD45.1:CD45.2 expressing thymocytes and splenocytes between 0.3 and 0.7, demonstrating the comparable capacity of CD1d^−/−, CD1d^+/+ and CD1d-EYFP/EYFP bone marrow cells to reconstitute general lymphocyte population. After 6 wk of grafting, the mice were analyzed for the presence and development of Vα14 iNKT cells. All mice were bred and maintained in a barrier facility and studies were performed according to institutional guidelines for animal use and care.

CD1d:PBS-57 (αGalCer) tetramers were provided by the National Institute of Allergy and Infectious Disease MHC Tetramer Core facility, Atlanta, GA. Abs were acquired from eBiosciences, BD Biosciences, BioLegend, and Invitrogen. Annexin V was acquired from BD Biosciences. Dr. A. Bendelac (Howard Hughes Medical Institute, University of Chicago, Chicago, IL) provided cells excreting rat-anti-mouse CD1d (19G11), Dr. G. Besra (School of Biosciences, University of Birmingham, Birmingham, U.K.) provided αGalCer and Gal(α1→2)GalCer and Dr. H. Ploegh provided polyclonal rabbit anti-EGFP Ab.

Thymocytes were lysed with NP40 lysis buffer (2% NP40, 150 mM NaCL, 1 mM EDTA, 0.02% NaN3, 0.5 mM Trisbase (pH 10.4)) and protease inhibitor (Roche). Lysates were boiled, separated on a 4–20% Tris-HCL gel, and transferred to a polyvinylidene fluoride membrane (Bio-Rad). Membranes were blotted using a polyclonal anti-mCD1d Ab (19G11) or a polyclonal anti-EGFP Ab, which cross-reacts with EYFP.

Mouse liver was perfused with PBS and homogenized trough a stainless steel mesh. After 5 min, centrifugation at 2000 rpm, the liver cells were taken up in 9 ml of PBS and mixed with 5 ml of 80% Percoll. This suspension was overlayed on 3 ml 80% Percoll and the gradient was spun for 20 min at 2000 rpm. The interface containing the liver lymphocytes was collected, washed with PBS, and RBC were lysed using ammoniumchloride.

Cells were stained with the relevant mAb on ice in the presence of Fc-block (2.4G2). Vα14 iNKT cells and their different developmental stages were detected by staining with αGalCer-loaded CD1d-tetramers and Abs against TCRβ, CD44, and NK1.1, unless mentioned otherwise. Incubating thymic Vα14 iNKT cells at 37°C for 30 min induced apoptosis. The amount of apoptosis was determined by staining with Annexin V following the manufacturer’s suggestions. For the endocytosis assay, thymocytes were stained for 1 h on ice in DMEM with PE-conjugated CD1d mAb in the presence of 10 μg/ml cycloheximide to block de novo synthesis. Cells were washed and subsequently incubated for various times at 37°C. Surface-bound mAb was stripped using 300 mM glycine/1% FCS solution (pH 2, 3 min) before neutralization (300 mM glycine/1% FCS solution, pH 7), staining for surface markers, and fixation (10 min, 4% paraformaldehyde). CD1d endocytosis was determined by the relative intensity of internalized anti-CD1d mAb. Normalized residual surface CD1d was calculated by setting surface CD1d-PE fluorescence at 100% and subtracting the percentage of internalized CD1d-PE.

CD11c⁺ MACS-purified spleen DCs were stimulated for 4 h with 100 ng/ml αGalCer or Gal(α1→2)GalCer. After two washes, freshly isolated liver Vα14 iNKT cells were added to the culture for 24 (for IL-4) or 48 (for IFN-γ) h. Vα14 iNKT cells in enriched leukocytes from CD1d-EYFP/EYFP liver were 2-fold reduced compared with WT and were corrected for by addition of twice the number of Vα14 iNKT cell-enriched leukocytes to the Ag-laden DC cultures. IFN-γ and IL-4 secretion was measured by ELISA following the manufacturer’s suggestions (eBiosciences).

Data are shown as mean ± SEM. Unpaired two-tailed t test was used to compare two groups. A p-value of at least 0.05 was considered statistically significant. Analysis was performed using Prism 4.0 for Mac software (GraphPad Software).

Mice have only two CD1 genes, CD1d1 and CD1d2. Expression of CD1d2 protein in mice is restricted to thymocytes and is considered nonfunctional (30, 31). We focused on CD1d1, hereafter referred to as CD1d, and generated knock-in mice in which the CD1d locus was replaced by a version encoding CD1d-EYFP fusion protein by homologous recombination (Fig. 1,A). The phosphoglycerate kinase promoter-driven neomycin resistance gene in the targeting vector, flanked by loxP sites, was deleted by breeding with cre-deleter mice (Fig. 1,A). Development of CD4 and CD8 T cells, B cells, and DCs was unaffected by the knock-in mutation (data not shown). Thymocytes from CD1d-EYFP/EYFP and WT mice show similar protein levels of CD1d (Fig. 1,B, left; polyclonal anti-CD1d blotting Ab). The presence and abundance of CD1d-EYFP polypeptide in whole thymus lysate was also comparable in amount to MHC class II-EGFP β-chain polypeptide (32) (Fig. 1,B, right; polyclonal anti-EGFP Ab, which also recognizes EYFP protein). Thymocytes showed the expected 49 kDa CD1d product in WT mice, and a 76 kDa fusion protein product in CD1d-EYFP/EYFP mice (composed of the 49 kDa CD1d together with 27 kDa EYFP polypeptide) (Fig. 1,B). No free EYFP was detected. Therefore, EYFP detected by flow cytometry (Fig. 1,C) or visualized by microscopy (Fig. 1,D) represents CD1d molecules labeled with EYFP (CD1d-EYFP), and CD1d-EYFP molecules are stable in the cellular environment. Using flow cytometry on fresh peripheral blood B lymphocytes, we showed that CD1d-EYFP/EYFP mice express double the amount of EYFP fluorescence compared with CD1d-EYFP/+ mice that harbor one CD1d-EYFP and one untagged CD1d allele (Fig. 1 C).

CD1d molecules acquire their antigenic cargo in the lysosomal pathway. CD1d molecules therefore predominantly localize to the cell surface and the lysosomal pathway (26). We purified DCs from bone marrow and determined the subcellular distribution of CD1d-EYFP molecules by confocal microscopy. DCs were analyzed immediately after magnetic cell sorting based on CD11c-expression to prevent maturation and DC attachment to coverslips. Using Z-stack analysis, CD1d-EYFP fluorescence was found in intracellular vesicular structures of the endo/lysosomal pathway, as well as on the cell surface in DCs (Fig. 1 D).

The positive selection and terminal differentiation of Vα14 iNKT cells requires CD1d-mediated Ag presentation by both cortical thymocytes and professional APCs (20, 23). To determine whether CD1d-EYFP can fully substitute for its unlabeled counterpart, we asked whether CD1d-EYFP molecules can support Vα14 iNKT cell development in vivo. Vα14 iNKT cells were characterized by staining with anti-TCRβ and CD1d-tetramers loaded with the pan-iNKT cell ligand αGalCer and analyzed by flow cytometry. CD1d-EYFP/EYFP mice showed a ∼60% reduction in tetramer-positive Vα14 iNKT cells in the thymus, and ∼80% reduction in spleen and liver compared with WT (Fig. 2). No Vα14 iNKT cells were detected in CD1d^−/− mice, as previously described (29). Thus, CD1d-EYFP molecules in vivo can support Vα14 iNKT cell development, albeit with reduced efficiency compared with WT CD1d molecules.

The reduced Vα14 NKT cell number in CD1d-EYFP/EYFP mice indicates a defect in the functioning of CD1d-EYFP. Alternatively, this observation could be caused by a reduced expression of CD1d-EYFP compared with WT, as CD1d-EYFP/EYFP thymocytes display less CD1d at the cell surface than do CD1d^+/+ thymocytes (p < 0.05) (Fig. 3^,A). However, CD1d-EYFP/+ heterozygous mice exhibited normal Vα14 iNKT cell numbers, as do CD1d^+/− heterozygous mice that express 50% less CD1d at the cell surface (Fig. 3, A and B). Thus, the expression of a single WT CD1d allele suffices to support normal thymic Vα14 iNKT cell selection, and the decreased Vα14 iNKT cell development in CD1d-EYFP/EYFP mice was unlikely to result from the ∼25% decrease in CD1d surface expression. CD1d-EYFP/EYFP mice moreover exhibited normal percentages of DP thymocytes (Fig. S1).⁴

The thymic selection of Vβ7⁺, but not Vβ8.2⁺, Vα14 iNKT cells is favored in situations where endogenous CD1d ligand concentration are suboptimal, as was shown in CD1d^+/− mice (33), while Vβ7⁺ Vα14 iNKT cells are relatively diminished upon CD1d overexpression (28). We asked whether endogenous Ag presentation by CD1d-EYFP molecules is suboptimal, as an explanation for decreased Vα14 iNKT cell development. To this end, we examined the TCR Vβ repertoire of Vα14 iNKT cells in CD1d-EYFP knock-in mice. We found that thymic Vα14 iNKT cells from CD1d-EYFP/EYFP mice displayed a bias in their Vβ repertoire, with an overrepresentation of Vβ7⁺ usage that was more pronounced than was earlier shown in CD1d^+/− mice (p < 0.001) (Fig. 3 C). Use of the TCR Vβ domain shapes the selection of Vα14 iNKT cells by endogenous ligand displayed by CD1d (33). Therefore, our data suggests that presentation of endogenous lipid Ags supporting thymic selection of Vα14 iNKT cells is suboptimal in CD1d-EYFP/EYFP mice when compared with WT mice.

We next asked how suboptimal endogenous Ag presentation during positive selection affects thymic development of Vα14 iNKT cells. To this end, we generated mixed bone marrow chimeras in which we tracked CD1d^−/− bone marrow precursors for their ability to give rise to Vα14 iNKT cells when transferred into CD1d-expressing recipient hosts (15), and measured the expression of the maturation markers CD44 and NK1.1 on Vα14 iNKT cells. In these chimeras, CD45.1 CD1d^−/− Vα14 iNKT cell precursors were selected on DP thymocytes expressing either WT CD45.2 CD1d (CD1d^−/− : CD1d^+/+ chimeras) or CD45.2 CD1d-EYFP (CD1d^−/− : CD1d-EYFP/EYFP chimeras). CD1d^−/− hosts were lethally irradiated and reconstituted with a 1:1 mix of bone marrow cells, distinguishable by expression of separate CD45 isoforms. After 6 wk of engraftment, the mice were analyzed for the presence and development of Vα14 iNKT cells. CD1d-EYFP/EYFP chimeras supported thymic development of fewer Vα14 iNKT cells when compared with CD1d^+/+ chimeras (p < 0.0001), causing a reduction of ∼80% in Vα14 iNKT cells (Fig. 4,A). We next analyzed three successive Vα14 iNKT cell thymic developmental stages, CD44⁻NK1.1⁻, CD44⁺NK1.1⁻, and CD44⁺NK1.1⁺ by flow cytometry, as described in previous studies (12, 15, 16, 22). The majority of thymic Vα14 iNKT cells extracted from CD1d-EYFP/EYFP chimeras were present in the earlier stages (predominantly CD44⁺NK1.1⁻) of development, with fewer than 15% of cells exhibiting the mature CD44⁺NK1.1⁺ stage. In CD1d^+/+ chimeras, Vα14 iNKT cells were more mature overall, with 40% of cells being found in the CD44⁺NK1.1⁺ stage (p < 0.05) (Fig. 4 B). Thus, CD1d-EYFP molecules are defective at supporting thymic selection and thymic maturation of Vα14 iNKT cells.

After birth, the Vα14 iNKT cell compartment continues to expand and mature, with a cellular increase with age up to 50 days (12, 15, 16, 34) after which thymic Vα14 iNKT cell numbers drop (35). In newborn mice, the majority of thymic Vα14 iNKT cells are CD44⁻NK1.1⁻, which gradually acquire CD44 and NK1.1 expression (12, 15). Also in the periphery, CD44⁺NK1.1⁻ thymic emigrants progressively develop into CD44⁺NK1.1⁺ Vα14 iNKT cells (35). We analyzed thymic Vα14 iNKT cells from three age cohorts of CD1d-EYFP/EYFP and control mice accordingly. Throughout time, thymic Vα14 iNKT cells from CD1d-EYFP/EYFP mice significantly remain in the earlier stages of development (predominantly CD44⁻NK1.1⁻) compared with CD1d^+/+ mice (p < 0.05) (Fig. 5,A). To establish further that Vα14 iNKT cells are inhibited in their thymic maturation when selected on CD1d-EYFP molecules, we analyzed for down-regulation of the developmental markers Slamf1 (CD150) and CD69. Developing iNKT cells temporarily display high levels of Slamf1 expression on DP Vα14 iNKT cell precursors (CD24⁺CD44⁻NK1.1⁻) (36). CD69 is also temporarily expressed following positive selection of MHC-restricted T cells (37) and after positive selection of Vα14 iNKT cells (12, 15, 16, 21, 22). In 6 -to 9-wk-old mice, we found that Vα14 iNKT cells that were selected on CD1d-EYFP molecules retained significantly high surface levels of Slamf1 (p < 0.006) (Fig. 5,B) and CD69 (p < 0.0001) (Fig. 5 C) up until the last stage of development compared with CD1d^+/+ mice. Similar data was found in mice 2–5 wk and 10–13 wk of age (data not shown).

Furthermore, thymic Vα14 iNKT cells from CD1d-EYFP/EYFP mice exhibited increased levels of apoptosis, as measured by flow cytometric analysis of Annexin V staining, which could partially account for the reduction of Vα14 iNKT cells in CD1d-EYFP/EYFP mice (Fig. S2, A and B). This increase in apoptosis was found in each age group and in each of the three stages of development (p < 0.05). Taken together, our data shows that suboptimal endogenous Ag presentation by CD1d-EYFP molecules results in decreased thymic maturation of developing Vα14 iNKT cells.

The majority of Vα14 iNKT cells exits the thymus in immature stage (CD44⁺NK1.1⁻) and matures in the periphery (12, 16). Acquisition of NK1.1 by these immature Vα14 iNKT cells and acquisition of full effector function requires interaction with CD1d in the periphery (23) by professional MHC class II-expressing APCs (20). We next asked whether CD1d-EYFP molecules can mediate postselection maturation of Vα14 iNKT cells. Vα14 iNKT cells from spleens of CD1d-EYFP/EYFP and CD1d^+/+ mice showed comparable CD44⁺NK1.1⁺ profiles, suggesting that CD1d-EYFP molecules support peripheral maturation of Vα14 iNKT cells (Fig. 6,A). Do phenotypically mature Vα14 iNKT cells selected on CD1d-EYFP molecules produce normal amounts of cytokine upon activation? WT spleen DCs were allowed to stimulate Vα14 iNKT cells extracted from either WT or CD1d-EYFP/EYFP livers. In addition to αGalCer glycolipid Ag, we used Gal(α1→2)GalCer, which requires internalization and removal of the terminal sugars by the lysosomal enzyme, α-galactosidase A, for CD1d-mediated presentation to occur (38). CD1d-EYFP/EYFP Vα14 iNKT cells produced normal levels of IFN-γ upon αGalCer and Gal(α1→2)GalCer presentation by WT spleen DCs (Fig. 6, B and D). However, IL-4 production by CD1d-EYFP/EYFP Vα14 iNKT cells was nearly absent compared with their WT counterparts (Fig. 6, C and E). In addition, IL-2 production by CD1d-EYFP/EYFP Vα14 iNKT cells was reduced upon αGalCer and Gal(α1→2)GalCer presentation by WT spleen DCs (p < 0.05) (Fig. S3, A and B). Thus, endogenous Ag presentation by CD1d-EYFP molecules during thymic selection results in Vα14 iNKT cells that produce normal levels of IFN-γ, decreased levels of IL-2, and no IL-4.

We had noticed that in DP thymocytes and spleen CD4⁺ T cells of CD1d-EYFP/EYFP mice, the surface expression of CD1d measured by flow cytometry appears as narrower mean fluorescence intensity histogram peaks compared with WT (Fig. S4A). In the DP thymocytes, this tighter surface CD1d distribution resulted in a significant decrease in mean fluorescence intensity compared with WT (p < 0.05), while in spleen, CD4⁺ T cells from homozygous CD1d-EYFP mice, the number of surface-expressed CD1d molecules was increased (reflected in a narrower histogram peak that is shifted more to the right, p < 0.05). On professional APCs (spleen DCs) from CD1d-EYFP/EYFP mice, however, CD1d surface expression was comparable to their WT counterparts (Fig. S4A).

To determine whether the increase in surface-expressed CD1d related to decreased endocytosis of CD1d-EYFP, we next analyzed the rate of Ab-binding-induced endocytosis in thymocytes, CD4⁺ T cells, and DCs from CD1d^+/+, CD1d-EYFP/EYFP, and CD1d^−/− mice. Cells were labeled on ice using PE-conjugated anti-CD1d in the presence of cycloheximide to block protein synthesis. Aliquots of cells were transferred to 37°C, and collected at sequential times thereafter. Ab bound to surface CD1d was stripped by acid treatment, and the extent of internalized CD1d bound by PE-conjugated anti-CD1d Ab was visualized by flow cytometry (39). The internalization rate of CD1d-EYFP molecules in DP thymocytes was significantly reduced compared with WT (p < 0.05) (Fig. 7, A and B). Control thymocytes from CD1d-deficient mice show no CD1d staining at any of the time points (Fig. S4B). Similar to their thymic counterparts, endocytosis of CD1d-EYFP molecules in CD4⁺ T cells was significantly reduced compared with WT CD1d molecules (p < 0.05) (Fig. 7,A). Compared with DP thymocytes, Ab-binding-induced CD1d endocytosis in spleen-derived DCs is much reduced and without significant difference between CD1d-EYFP/EYFP and WT DCs (p > 0.05) (Fig. 7, A and B).

Thymocytes do not express Ii, and in thymocytes CD1d therefore strictly relies on its own CD1d-encoded tyrosine-based sorting motif for localization in lysosomes. In professional APCs, however, endosomal sorting of CD1d could be mediated by the CD1d-encoded tyrosine-based sorting motif, or via association of CD1d with the class II MHC-chaperone Ii (7, 8, 9). We asked whether in professional APCs, the CD1d-encoded tyrosine-based sorting motif and Ii-encoded sorting motif can supplement for each other in mediating lysosomal localization, Ag acquisition, and presentation by CD1d. To this end, we tested whether Gal(α1→2)GalCer can be processed and presented by spleen-derived DCs, when either CD1d-EYFP is expressed, or Ii is absent, or in DCs from mice that harbor both mutations (Ii^−/−CD1d-EYFP/EYFP mice).

DCs from indicated mouse strains were loaded with graded amounts of Gal(α1→2)GalCer for 4 h, washed, and allowed to stimulate WT Vα14 iNKT cells. WT DCs presenting 100 pg/ml αGalCer to WT Vα14 iNKT cells were included as a control to determine the response potential of the Vα14 iNKT cells in this assay. WT DCs strongly induced IFN-γ production and IL-4 production in Vα14 iNKT cells (Fig. 8, A and B). DCs from CD1d-EYFP/EYFP mice, Ii^−/− mice, and the combined Ii^−/−CD1d-EYFP/EYFP mice did not induce IFN- γ production by WT Vα14 iNKT cells (p < 0.001), while also IL-4 production was significantly reduced (p < 0.05). DCs from CD1d^−/− mice did not induce cytokine production in WT Vα14 iNKT cells. Similarly, presentation of Gal(α1→2)GalCer by CD1d-EYFP/EYFP spleen DCs resulted in significantly less IL-2 production by WT Vα14 iNKT cells (p < 0.05) (third bar in Fig. S3B). Taken together, endosomal sorting of CD1d relies on both its self-encoded tyrosine-based motif, and on association with Ii in professional APCs in the periphery, to stimulate Vα14 iNKT cells. Thus, both the CD1d-encoded and Ii-encoded endosomal sorting motifs are necessary to induce activation of Vα14 iNKT cells by professional APCs in spleen.

A remaining question in the field of CD1d-mediated Ag presentation and Vα14 iNKT cell development concerns the intracellular endosomal trafficking routes of CD1d during Vα14 iNKT cell development. In contrast to conventional, MHC-selected CD4 and CD8 T cells, Vα14 iNKT cells require Ag presentation by both nonprofessional and professional APCs to obtain terminal maturation. Although thymic selection occurs on DP thymocytes presenting CD1d, final CD1d-mediated maturation events after their thymic emigration, by interaction with professional APCs (17, 18, 19, 20, 23). Endosomal tyrosine-based sorting motifs in the CD1d-tail and in Ii mediate endosomal localization of CD1d (7, 8, 9), which therefore could result in exposure to nonoverlapping Ags due to differential localization in non-Ii-expressing cells (i.e., DP cortical thymocytes) and Ii-expressing cells (i.e., professional APCs). In this study, we investigated the role of the CD1d-encoded endosomal sorting motif in thymic development, peripheral maturation, and activation of mature Vα14 iNKT cells.

A comparative study for the role of professional APCs and other CD1d-expressing cells in peripheral maturation of Vα14 iNKT cells is complicated, because in most mutant mice in which CD1d-mediated presentation is affected, both thymic development and peripheral maturation of Vα14 iNKT cells is hampered, resulting in thymic deletion of Vα14 iNKT cells during thymic development and inability for their study during peripheral maturation. The importance of peripheral maturation events for Vα14 iNKT cell function, however, is clear. In NOD mice, for example, Vα14 iNKT cells do develop but are unable to fully mature, which results in a predisposition to insulitis development (24, 25). We generated a new mouse strain in which all CD1d is expressed as CD1d-EYFP fusion proteins, in which we investigated the role of CD1d endosomal sorting and Vα14 iNKT cell development and maturation.

In the homozygous CD1d-EYFP mice, a sizable fraction of Vα14 iNKT cells of 40% pass successfully through positive selection. This is in contrast with CD1d tail-deleted mice, in which Vα14 iNKT cells were decreased to ∼10% (9). In CD1d-EYFP/EYFP mice, the thymic selection defect probably differs from the defect in CD1d tail-deleted mice in which CD1d access to endosomal compartments is blocked. In CD1d-EYFP/EYFP mice, CD1d-EYFP molecules do localize to endosomal compartments, albeit at a different kinetic rate. Using CD1d^+/− mice, it was shown that in situations where endogenous CD1d ligand concentration are suboptimal, thymic selection of Vβ7⁺, but not Vβ8.2⁺, Vα14 iNKT cells is favored (33), while Vβ7⁺ Vα14 iNKT cells are relatively diminished upon CD1d overexpression (28). The hierarchy of Vβ-chain usage (Vβ8, Vβ7, and Vβ2) is already established during positive selection and is not altered thereafter through preferential cellular expansion (21, 33). CD1d-EYFP/EYFP mice also show an overrepresentation of Vβ7 usage by Vα14 iNKT cells, suggesting that endogenous Ag presentation by CD1d-EYFP molecules is suboptimal. Endogenous Ag presentation during thymic selection, however, is not absent as in CD1d-tail deleted mice, and CD1d-EYFP/EYFP mice are still able to support development of 40% of Vα14 iNKT cells.

Maturation of Vα14 iNKT cells in both the thymus and the periphery can be monitored by acquisition of NK1.1 and CD44 (12, 16) and is CD1d-dependent (20, 23). In thymus of CD1d-EYFP/EYFP mice, acquisition of NK1.1 and CD44 was delayed, but was compensated for in full during peripheral maturation of Vα14 iNKT cells capable of seeding in the periphery. NK1.1 expression signals a maturation of the cytokine response toward high IFN-γ production (12, 16, 22). We confirm this data, as terminally matured Vα14 iNKT cells in CD1-EYFP/EYFP mice were fully capable of producing IFN-γ, but not IL-4. In humans, invariant NKT cells (characterized by Vα24/Vβ11 TCR junctions) are readily distinguished based on expression of CD4: CD4⁺ Vα24 iNKT cells are the exclusive producers IL-4 and IL-13 upon primary stimulation, but could also produce Th1 cytokines, whereas the double negative (DN) Vα24 iNKT cells had a strict Th1 profile producing IFN-γ and TNF-α (40, 41). It is yet unclear whether in mice a similar divide of CD4⁺ and DN Vα14 iNKT cells in production of Th1 and Th2 responses exists (37, 42, 43, 44, 45). In CD1d-EYFP/EYFP mice, it is a possibility that IL-4-producing Vα14 iNKT cells are specifically deleted during thymic selection on CD1d-EYFP molecules. In mice, CD4⁺Vα14 iNKT cells constitute the majority of the Vα14 iNKT cells in the periphery. Both CD4⁺ and DN Vα14 iNKT cell subsets showed comparable decreases in numbers present in CD1d-EYFP/EYFP spleen. There was no selective depletion of DN Vα14 iNKT cell subsets that could explain the absence of IL-4 production (data not shown).

CD1d-EYFP/EYFP mice exhibit a narrower mean fluorescence peak in CD1d surface expression in DP thymocytes and CD4⁺ T cells, representing a more homogenous population of cells expressing similar surface levels of CD1d. This tighter distribution of surface CD1d-EYFP could be caused by the delay in internalization rate. Tail-deleted CD1d molecules exhibit a slower internalization rate than WT CD1d, but recycling (reappearance from intracellular compartments) is undisturbed (8). For CD1d-EYFP molecules, a slower endocytosis rate resulted in accumulation of CD1d at the cell surface in peripheral CD4⁺ T cells, but not for CD4⁺CD8⁺ thymocytes. Recycling assays using primary cells were unsuccessful in our hands. The rate of CD1d-EYFP recycling could possibly be faster in CD4⁺ T cells than thymocytes, which would explain the accumulation on the surface in CD4⁺ T cells and not in CD4⁺CD8⁺ thymocytes.

The homeostasis of Vα14 iNKT cells in the periphery is independent of the presence of CD1d in the periphery: CD1d does not affect the survival or expansion rate of iNKT cells in the periphery (23, 46). CD1d-EYFP/EYFP mice exhibit a decrease in Vα14 NKT cell numbers in spleen and liver. Because thymic output of postselection Vα14 iNKT cells is decreased in CD1d-EYFP/EYFP mice, fewer recent thymic emigrants are seeded in the periphery, which is a likely cause for the observed lower set-point for Vα14 iNKT cell homeostasis in the periphery.

Stimulation of terminally differentiated Vα14 iNKT cells by presentation of lysosomal Ag on professional APCs required unmodified CD1d and presence of Ii. Thus, while CD1d-EYFP molecules were capable of mediating peripheral maturation of Vα14 iNKT cells, they did not succeed in provoking IFN-γ and IL-4 production in Vα14 iNKT cells. Taken together, in this study, we describe a new mouse model, in which all CD1d is expressed as CD1d-EYFP fusion proteins. An unexpected phenotype of these mice clarified the role of CD1d endosomal sorting motifs in Vα14 iNKT cell selection and peripheral iNKT cell maturation, during thymic selection and by peripheral Ii-expressing professional APCs. Endosomal sorting mediated by the CD1d-intrinsic tyrosine motif is critical for thymic positive selection. In the periphery, CD1d endosomal sorting does not require the intrinsic CD1d-encoded sorting motif for terminal maturation of Vα14 iNKT cells. Activation of differentiated Vα14 iNKT cells, however, does require both the CD1d-tail encoded sorting motif, and cannot be compensated for by the class II MHC-associated Ii.

We thank all members of the Boes laboratory; Dr. S. Behar and his laboratory members; and Dr. M. Nowak from the M. Exley laboratory for helpful discussions. We are also thankful to Dr. H. Ploegh for the anti-EGFP Ab; Dr. M. Exley for providing CD1d^−/− mice; Dr. A. Bendelac for the hybridoma excreting rat-anti-mouse CD1d (19G11); Dr. G. Besra for αGalCer and Gal(α1→2(GalCer)); and The National Institutes of Health Tetramer Facility for supplying the CD1d-tetramers.

The authors have no financial conflict of interest.