Expansion and long-range differentiation of the NKT cell lineage in mice expressing CD1d exclusively on cortical thymocytes

To restrict CD1d expression on the T cell lineage, we constructed transgenic vector using the proximal Lck promoter (29) to drive CD1d cDNA in a β-globin expression cassette (Fig. 1 A). Five founders were established by injection into fertilized oocytes of inbred B6 (n = 2) or nonobese diabetic (NOD) strains (n = 3). The B6 founders subsequently were bred to the B6.CD1d KO background, and the NOD founders were bred to NOD.CD1d KO mice followed by a cross to B6.CD1d KO mice to obtain (NODxB6) F1.CD1d KO mice. The transgenic lines in the CD1d homozygous KO background are thereafter termed pLck-CD1d mice. For in-depth analysis of CD1d expression, we selected a pLck/low and a pLck/high as low and high expressors, respectively, as well as a pLck/var exhibiting a variegated pattern of expression on ∼10–20% of thymocytes. All pLck-CD1d lines showed faithful expression of CD1d in the T cell lineage, i.e., cortical thymocytes and mature T cells. Examples of expression pattern in pLck-CD1d mice are shown in Fig. 1 B and Fig. 2 and are summarized in Table I. Our breeding scheme allowed each transgenic mouse to be analyzed side by side with a nontransgenic CD1d^+/− (CD1d Het) littermate for optimal comparison purposes.

In thymic frozen sections (Fig. 1 B), CD1d Het controls showed broad expression of CD1d (red) not only on all cortical thymocytes, but also on all MHC class II I-A^b+ cells (green), including cortical and medullary epithelial cells, CD11c⁺ DCs, and CD11b⁺ macrophages, as previously shown (27). In contrast, pLck/low, pLck/high, and pLck/var expressed CD1d on cortical and, to a lesser extent, medullary thymocytes, but not on any I-A^b+ cells.

A more quantitative assessment of CD1d expression by different cell types was performed using flow cytometry (summarized in Table I for pLck/low and pLck/high and shown in Fig. 2 for pLck/var). Although CD1d expression was restricted absolutely to the T cell lineage, both in the thymus and periphery, the level of CD1d expression varied in different transgenic lines, ranging from five to 20 times the level expressed by CD1d^+/− heterozygous controls, i.e., 2.5 to 10 times the levels of WT CD1d^+/+. The pLck-CD1d transgenic mice exhibited a marked drop in CD1d expression between the immature cortical DP and the single-positive mature T cell stage, as expected.

Next, we examined the NKT cell developmental stages in the pLck-CD1d mice using flow cytometry with various stage-specific markers and α-galactosylceramide (α-GalCer)–loaded CD1d tetramers specific for the canonical Vα14-Jα18 TCR. Remarkably, all lines of pLck-CD1d mice expressed substantial percentages of NKT cells (Fig. 1 C). Despite their reduced proportion of DP thymocytes expressing CD1d, the pLck/var mice (C57BL/6 background) expressed the same frequencies and absolute numbers of NKT cells in the thymus and spleen as CD1d^+/− littermate controls at the adult age (Fig. 1 C) as well at the younger ages of 2 and 4 wk (unpublished data). In the C57BL/6 background, CD1d^+/− and CD1d^+/+ mice have the same number of NKT cells (unpublished data). The pLck/low and pLck/high mice (NODxC57BL/6 F1) expressed 1.5 to three times more NKT cells than CD1d^+/− controls (Fig. 1 C), or about the same as CD1d^+/+ mice in the NODxC57BL/6 background (unpublished data). Similar results were found for liver lymphocyte samples (unpublished data).

NKT cells in pLck/var mice expressed the same Vβ repertoire composed of three families, Vβ8>Vβ7>Vβ2 (Fig. 3 A), and passed through the same CD44^low, CD44^high, and NK1.1⁺ stages as WT mice with no difference in their double-negative (DN)/CD4 ratio (Fig. 3 B). However, one consistent defect observed in the three pLck-CD1d transgenic lines examined was an ∼30% lower ratio of NK1.1⁺/NK1.1⁻ cells in peripheral tissues (spleen and liver) (Fig. 3 B and unpublished data). These mature NK1.1⁺ cells, however, expressed the whole complement of NK lineage receptors examined, with intensity and frequency similar to those seen in WT mice (Fig. 3 C), including NKG2D, Ly49A, C/I, G2, CD94, NKG2A/C/E, and other characteristic markers such as DX5, CD44, CD62L, CD69, and CD122.

We next measured the rate of cell divisions previously reported at the HSA^lowNK1.1⁻ stage of NKT cell thymic development. 4 hr after an i.p. injection of 1 mg bromodeoxyuridine (BrdU), 11–12% of NK1.1⁻ thymic NKT cells were labeled in WT, as in pLck/var mice (Fig. 4), demonstrating that the massive lineage expansion that is a hallmark of this NKT cell thymic stage occurred normally.

Further analysis of pLck/var mice with a variegated pattern of transgene expression provided an unexpected clue to the ongoing interactions between the NKT cell precursors and CD1d-expressing cells. We made the surprising observation that the frequency of CD1d expression among positively selected HSA^lowCD44^low NKT cells (40.1%) was two times higher than expected from the frequency among DP thymocytes (18.6%) (Fig. 5 A). This increased frequency was particularly evident at the early HSA^lowCD44^low NKT cell stage after positive selection from the HSA^high stage, when pLck-driven CD1d expression is not yet extinguished. This surprising but consistent result might suggest that developing NKT cells use their own CD1d to enhance positive selection. To evaluate further the general significance of this unexpected result, we generated mixed WT + CD1d KO into CD1d KO radiation BM chimeras in which CD1d expression was confined to a fraction of thymocytes. Here again, the proportion of NKT cells expressing CD1d was approximately twice that of DP or mainstream T cells (Fig. 5 B). Thus, the data suggest that, although not essential, intrinsic expression of CD1d significantly enhances the selection and development of NKT cell precursors.

Although the quasinormal selection, differentiation, and maturation of NKT cells in pLck-CD1d mice rules out an essential contribution by DCs, macrophages, or thymic epithelial cells, it was important to investigate whether transgenic expression of CD1d on mature T cells inside and outside the thymus was contributing to NKT cell development. Indeed, even though the proximal Lck promoter is substantially extinguished after the DP stage, residual gene expression combined with the long lifespan of CD1d proteins provides a window of opportunity for mature HSA^low NKT thymocytes to interact with CD1d (including CD1d expressed on NKT cells themselves, as described before) after they exit the cortex. Such interaction might account for the rounds of cell division and the effector/memory formation that characterize this developmental stage, as well as terminal NK differentiation in the periphery. To abrogate this residual expression of CD1d in the thymic medulla and in the periphery, we generated mixed BM radiation chimeras using a mixture of pLck/low.Cα KO BM (as antigen-presenting component, unable to generate T cells) and CD1d KO BM (as responding component, generating the T cell but unable to present antigen). The hosts were irradiated CD1d KO mice. In these chimeras, therefore, CD1d was expressed only on cortical thymocytes that were themselves unable to mature because of the Cα mutation, whereas NKT cells could only arise from cells that were themselves CD1d KO. Surprisingly, NKT cells were quasinormally generated in such chimeras by comparison with control CD1d KO + Cα KO mixed chimeras (Fig. 6).

NKT cells were identical to controls in terms of numbers, Vβ repertoire, and DN/CD4 ratio, as well as pattern of NK receptor expression (unpublished data), although the proportion of NK-lineage receptor-expressing cells was substantially decreased in the periphery (Fig. 6). Similar results were observed using pLck/var.Cα KO cells instead of pLck/low.Cα KO BM cells (unpublished data). Thus, these mixed chimeras firmly establish that, even with CD1d expression strictly confined to the thymic cortex, NKT cell development proceeds through the sequential stages leading to the final NKT cell product weeks after emigration to the periphery. The partial defect in peripheral NK1.1 acquisition previously noted in the pLck-CD1d mice as compared with WT mice was observed again when compared with control chimeras. In fact, in these mixed chimeras, the relative defect tended to be even more severe (50% reduction) than when pLck-CD1d mice were compared with WT littermates (30% reduction), possibly because of the more restricted distribution of CD1d expression.

To study the functional properties of NKT cells developing in pLck-CD1d mice, in which CD1d expression by peripheral APCs is absent, we first injected DCs loaded with the agonist NKT cell ligand α-GalCer intravenously and measured cytokines released in the sera. Peak IFN-γ secretion not only was preserved but was, in fact, consistently three to four times higher in the two pLck-CD1d lines examined, pLck/var and pLck/low, than in WT cells (Fig. 7 A). IL-4 was not detected in this assay. However, the early IL-4 release induced by anti-CD3ε i.v. injection, a characteristic functional property of effector-stage NKT cells (24), was unimpaired and perhaps was increased slightly in pLck/var mice compared with CD1d Het littermate controls (Fig. 7 B). In an in vitro assay combining sorted splenic CD1d-α-GalCer tetramer⁺NK1.1⁺ NKT cells and α-GalCer-pulsed DCs, NKT cells from pLck-CD1d mice secreted both IFN-γ and IL-4, again significantly more than WT cells (Fig. 7 C).

We next crossed the pLck/var mouse with the Eα-CD1d transgenic mouse, which expresses CD1d exclusively under the control of the MHC class II Eα promoter. The Eα-CD1d transgenic mouse previously was shown to express CD1d faithfully according to the MHC class II pattern, i.e., in the thymic epithelium as well as in thymus and peripheral APCs, including DCs, macrophages, and B cells (27). Together, the pLck and Eα promoters reconstitute the natural distribution of CD1d (with the exception of hepatocytes). Strikingly, the splenic NKT cell responses in these double-pLck/var.Eα-CD1d transgenic mice were normalized, comparable to WT B6 mice (Fig. 7 C). In addition, the relative defect in the acquisition of NK receptors by peripheral NKT cells was corrected (Fig. 7 D), demonstrating the concomitant influence of peripheral CD1d-expressing APCs in promoting the terminal differentiation. Similar findings were observed when the pLck/var mice were crossed with WT mice (Fig. 7, C and D), further demonstrating that the altered tissue expression pattern of CD1d, restricted to T cells, accounted for the relative decrease in terminal NKT cell differentiation and concomitant increase in cytokine secretion.

In an attempt to account for the complex, orchestrated sequence of thymic and peripheral events leading to the natural NKT cell differentiation in response to self-agonist self ligands such as iGb3, this study focused on elucidating the nature of the CD1d-expressing cell types involved at different stages of the process, from positive selection in the thymic cortex to clonal expansion and cytokine gene locus activation in the medulla and to the unique expression of the NK program in the periphery. Previous studies had indicated that the entire sequence of events could be directed by CD1d-expressing BM-derived cells (14–17), that CD1d-expressing DP thymocytes were required for the process (17), and that DCs transgenically overexpressing CD1d could induce negative selection (21). Thus, although the survival of terminally differentiated NKT cells was shown to be independent of CD1d (23), it was logically anticipated that the different stages of NKT cell development and differentiation occurring after positive selection and leading to the terminal NKT cell stage would require additional interactions between the NKT cell, autoreactive TCR, and agonist iGb3 or other ligands in different cell types of the thymus and periphery.

The present findings, however, clearly establish that, to a large extent, a single, temporo-spatially restricted TCR interaction between the immature Vα14⁺ T cell precursor and CD1d-expressing DP T cells in the thymic cortex is sufficient to induce the whole sequence of distinct events that characterize NKT cell differentiation, even weeks after immigration to the periphery. Careful analysis by flow cytometry and tissue immunohistochemistry of the pattern of CD1d expression in the pLck-CD1d transgenic mice excluded cryptic or aberrant expression of CD1d. Furthermore, because analysis of the pLck/var mouse demonstrated that expression of CD1d by the developing NKT cell itself could have a substantial impact on its differentiation, we generated mixed chimeras in which we could observe the development of CD1d KO thymocytes in response to pLck-CD1d BM cells developmentally arrested at the DP stage. In this experimental set up, in which CD1d could be presented only in trans by cortical thymocytes, NKT cells were selected, amplified, and differentiated largely as in WT. Thus, we conclude that the transient interaction of the immature Vα14⁺ thymocyte with CD1d expressed by neighboring cortical thymocytes is sufficient to switch on the long-range expansion and differentiation program that characterizes the NKT cell lineage.

This conclusion is surprising for several reasons. First, TCR engagement in the cortex has not been shown previously to induce entry into cell cycle, nor do DP thymocytes express the B7 costimulatory molecules normally required to induce naive T cell expansion. Second, because CD1d expression was largely extinguished in the periphery of pLck-CD1d mice and completely absent in the mixed chimeras, how did NKT cells, unlike other T cells (30), survive and maintain an effector status in the absence of TCR signals? Perhaps the induction of a memory phenotype and expression of the IL-15 receptor CD122 chain in the thymus, before emigration, constitutes the solution to this enigma, because memory cells expressing IL-15 receptor are largely independent of TCR engagement for survival (23). Third, if NK receptors serve the purpose of balancing TCR autoreactivity (18), how could they be selected in the absence of TCR signals? Our findings indicate that the genetic program of NK lineage differentiation, although it unfolds after emigration to the periphery, is activated in the thymus. Although we found no changes in the expression of the individual NK receptors examined, others may be altered, and it is possible that more NKT cells have mismatched NK receptors in pLck-CD1d mice than in WT mice, explaining their greater responses to TCR stimuli. This greater mismatch might explain the tendency of NKT cells from pLck-CD1d mice to exhibit higher TCR reactivity than their WT counterparts, an appealing possibility that could not be tested in this transgenic system. Alternatively, inhibitory receptors may selected based on their ability to antagonize activating NK receptors such as NK1.1 and NKG2D. This plasticity in the developmental scheme may also reflect plasticity in functional activation, because evidence suggests that NKT cells can suppress type I diabetes in a CD1d-independent manner (unpublished data).

Interestingly, although the transition to the NK stage was preserved in the pLck-CD1d mice, it seemed to be somewhat less efficient than in WT mice. This observation was confirmed further in the mixed BM chimeras, an experimental system in which CD1d could be restricted to a subset of developmentally arrested Cα KO thymocytes. Together with the higher reactivity of NKT cells in these mice, these observations suggest that, in WT mice, additional interactions with other types of CD1d-expressing cells may shape further some of their functional properties and terminal differentiation. Although one cannot rule out the possibility that these changes might be related in part to the transgenic system, our results indicated that both the hyperreactivity and the delayed NK receptor expression were corrected after introduction of the Eα-CD1d transgene (driving CD1d with an MHC class II promoter) or the WT CD1d in these pLck-CD1d mice. These findings support the notion that additional interactions with CD1d-expressing APCs fine-tune NKT cell development and function. The mechanisms involved in these tuning events remain to be elucidated.

Agonist-driven thymocyte development has been reported in a number of transgenic systems (31–37), resulting in altered differentiation programs leading to the NKT cell, CD4⁺CD25⁺, or CD8αα lineages with various memory/effector phenotypes. Although both the epithelial and the BM-derived compartments of the thymus could support the differentiation of such autoreactive cells, the specific nature of the signaling events involved has remained elusive. The crucial role of Fyn in Vα14⁺ NKT cell development (9, 10) is particularly intriguing in light of the recent discovery of its upstream connections to signaling lymphocytic activation molecule (SLAM)–associated protein (SAP) and SLAM-type molecules and downstream activation of NF-κB (38, 39). Indeed, NF-κB mutant mice exhibit major NKT cell developmental defects (11–13), and we have found that SAP-deficient mice, like Fyn KO mice, also completely lacked NKT cells (unpublished data). SLAM-type molecules and SAP are prominently expressed by cortical thymocytes (40), suggesting that homophilic interactions between SLAM family members expressed by developing Vα14⁺ T cells and their fellow CD1d-expressing DP thymocytes might be essential to NKT cell development and differentiation, providing the costimulation needed for expansion and perhaps also for rescue from negative selection. It is striking that, unlike CD1d molecules, MHC class I molecules are transiently down-regulated during the DP stage, both in mouse and human (41). It is intriguing, therefore, to speculate that these conspicuous differences underlie fundamentally different pathways of T cell development for the recognition of different categories of antigens, i.e., peptides versus lipids. Future experiments are in progress to test the hypothesis that restricted expression of MHC class I on DP thymocytes may lead to the development of MHC class I–specific NK-type T cells.

CD1d KO mice were bred for more than 14 generations onto either NOD or C57BL/6 background in our laboratory. C57BL/6 mice and TCR Cα KO mice in the C57BL/6 background were purchased from Jackson ImmunoResearch Laboratories. C57BL/6.Eα-CD1d mice expressing a CD1d1 transgene driven by the MHC II Eα promoter in a CD1d KO background were previously described (27). To generate pLck-CD1d Tg mice, we replaced the Eα promoter in the previously described Eα-CD1d construct with the proximal Lck promoter (29) excised (NotI and BamHI) from the p1017 transgenic expression cassette (provided by Nigel Killeen, University of California, San Francisco), as described in Fig. 1 A. This construct contains exons, introns, and poly-adenylation sequence for optimal expression. The linearized (SpeI and XhoI) construct was injected into fertilized C57BL/6 oocytes and into fertilized NOD oocytes, and the injected oocytes were implanted into pseudopregnant CD-1 (VAF+) outbred females. Transgenic mice were screened using PCR (forward primer: 5′-TCCTGTGAACTTGGTGCTTGAG-3′; reverse primer: 5′-CCAAAATGATGAGACAGCACAAC-3′) and were bred to C57BL/6.CD1d KO or NOD.CD1d KO to obtain pLck-CD1d mice expressing the pLck-CD1d transgene in the C57BL/6.CD1d KO (n = 2) or to NOD.CD1d KO (n = 3) background, respectively. Because the NOD strain has a partial, recessive defect in NKT cell development (42), the NOD.pLck-CD1d mice were further crossed with C57BL/6.CD1d KO mice to generate NODxC57BL/6 F1.pLck-CD1d mice. All mice were housed in a specific pathogen-free facility at the University of Chicago according to the guidelines of the institutional Animal Care and Use Committee.

Fluorochrome-conjugated mAbs anti-CD1d, CD3ε, CD4, CD8α, CD11c, CD11b, CD24 (HSA), CD44, CD45, CD62L, CD69, CD94, CD122, DX5, Gr-1, Ly49A, Ly49C/I, Ly49G2, NKG2A/C/E, NK1.1, TCR Vβ2, Vβ6, Vβ7, and Vβ8.1+8.2, were purchased from BD Biosciences. Anti-F4/80, B220, NKG2D mAbs, and streptavidin conjugated with allophycocyanin or R-phycoerythrin were purchased from eBioscience. Anti-I-A^b mAb (clone Y3P) was conjugated to FITC in our laboratory. CD1d tetramers loaded with α-GalCer were prepared and used in staining as described previously (43). BrdU staining of cells extracted from mice (after one i.p. injection of 1 mg BrdU) was done according to manufacturer's instructions in BrdU Flow Kits (BD Biosciences). Flow cytometry analysis was performed on a four-color FACSCalibur with CELLQuest software (BD Biosciences).

Thymi were harvested, rinsed with PBS, embedded in OCT freezing medium (Tissue-Tek, Sakura Finetek), frozen in a mixture of 2-methylbutane (Sigma-Aldrich) and dry ice, and stored at –20°C. Thymic sections were cut with a cryostat (10 μm), mounted on glass slides (Fisher Scientific International), dried, fixed with acetone (Sigma-Aldrich), and covered with glass coverslips (Fisher Scientific International). Thymic sections were first stained with a ∼5 μg/ml of a mixture of three purified rat anti-CD1d mAbs (clones 19G11, 17F5, and 16G9 in 1:1:1 ratio) developed in our laboratory. Biotinylated goat anti–rat IgG, streptavidin-Texas red, and FITC-conjugated Y3P anti–I-A^b antibodies were used in sequential order with washes in between. Fluorescence images were acquired with an Axiovert 200 microscope (Carl Zeiss MicroImaging, Inc.) and processed with OpenLab (Improvision Inc.) and Adobe Photoshop (Adobe Systems Inc.) software.

Single-cell suspension of splenocytes was first enriched by depleting B220⁺ cells using B220-conjugated magnetic microbeads with AutoMACS (Miltenyi Biotec) according to the manufacturer's instructions. CD1d-α-GalCer tetramer⁺NK1.1⁺ cells were then sorted with FACSAria (BD Biosciences), and cultured with DCs in ER 10% medium, a 1:1 mixture of Eagle's Ham's amino acid and RPMI 1640 (Biofluids) enriched with 10% heat-inactivated FCS, glutamine, antibiotics, and 5 × 10⁻⁵ M of 2-mercaptoethanol. The DCs were prepared from cultured BM in ER10% medium supplemented with mouse GM-CSF (Biosource International) at 2 ng/ml for 6 d and pulsed with α-GalCer (100 ng/ml) overnight before washing and culture at 10⁵ per well with titrated numbers of NKT cells in 96 flat-bottom microwell plates (Corning Costar Corporation). Stimulation assays were incubated for 48 h before assaying IL-4 and IFN-γ release with OptEIA ELISA sets (BD Biosciences) according to manufacturer's instructions. For in vivo experiments, mice were injected i.v. with 5 × 10⁵ α-GalCer-pulsed DCs in PBS. The spleen fragment assay measuring IL-4 release in vitro after 1.25 μg anti-CD3ε injection in vivo was performed as described previously (24).

C57BL/6.CD1d KO mice were irradiated (900 rad) 2 h before BM reconstitution. BM cells were harvested from mouse femurs and depleted of T cells with FITC-conjugated anti-CD3ε mAb and anti-FITC magnetic microbeads on AutoMACS (Miltenyi Biotec). BM cells from various donor combinations were mixed in a 1:1 ratio, and a dose of 10⁷ mixed BM cells was injected i.v. Chimeras were analyzed 6–8 wk after injection.

We thank L. Degenstein for oocyte injections, T. Li for help with immunohistochemistry; J. Marvin, R. Duggan, and B. Eisfelder for cell sorting; and all Bendelac laboratory members for expert advice and discussion.

This work is supported by grants from the New Jersey Committee on Cancer Research (to D.G. Wei), the Korea Research Foundation grant no. 2001-015-DS0063 (to S.-H. Park), the French Ministry of Research (to A. Lehuen), and the National Institutes of Health no. RO1 AI38339 (to A. Bendelac).

The authors have no conflicting financial interests.