Cd1-Reactive Natural Killer T Cells Are Required for Development of Systemic Tolerance through an Immune-Privileged Site

Female, 8–10-wk-old mice were used in all experiments. C57BL/6 (B6) mice were obtained from Taconic Farms. (B6 × 129/Sv) F1 (F1) mice were obtained from The Jackson Laboratory. CD1 KO mice were generated in the Transgenic Facility, Harvard Medical School, Boston, MA (Exley, M., manuscript submitted for publication). In brief, the CD1 (both CD1.1 and CD1.2) mutation was created in strain 129/Sv-derived embryonic stem (ES) cells. Mutant ES cell clones were injected into B6 blastocysts to obtain chimeric mice. Heterozygous mutant animals were intracrossed in brother–sister mating to obtain (B6 × 129/Sv)F2 (F2) homozygous mutants. In most cases, control wild-type (WT) mice were F2 mice (The Jackson Laboratory), but F1 cells were used as WT in reconstitution experiments. A confirmatory experiment was performed in B6 CD1 KO mice. During the time the experiments were performed, the CD1 mutation was being backcrossed to the B6 parent for five generations (N5). Progeny that lacked the CD1 gene as determined by DNA analyses were chosen for breeders. The animals were maintained on food and water ad libitum until they reached the desired weight (20–24 g). All animals were treated humanely and in accordance with the Schepens Eye Research Institute–Boston Biomedical Research Institute Animal Care and Use Committee and National Institutes of Health guidelines.

ACAID was induced in mice by inoculating OVA (50 μg/2 μl in HBSS; Sigma Chemical Co.) into the ac 10 7 d before sensitizing subcutaneously for DTH. Intravenously induced immune deviation was induced by inoculation of the antigen (OVA, 50 μg/100 μl in HBSS) into the tail vein with a 30-gauge needle 7 d before immunizing for DTH. To induce DTH, mice received a subcutaneous inoculation with OVA (100 μg/ml in HBSS, 50 μl) emulsified in CFA (50 μl) and 1 wk later were tested for the development of DTH by an intradermal inoculation of OVA-pulsed peritoneal exudate cells (PECs; 2 × 10⁵/10 μl HBSS) into the right ear pinnae. Ear swelling was measured 24 and 48 h later with an engineer's micrometer (Mitutoyo/MTI).

To test for the efferent-regulatory cell of ACAID, a modified local adoptive transfer (LAT) assay was performed as described elsewhere 14. In brief, T (effector) cells were generated in B6 mice or F1 3,4,5 by immunizing subcutaneously with OVA in HBSS and CFA. 7 d later the primed T cells were enriched from dissociated spleen cells by removing B cells and macrophages using IMMULAN™ columns (no. BL7020; Biotecx Laboratories, Inc.). Regulator cells were similarly enriched on IMMUNLAN™ columns from spleen cells of ACAID mice 7 d after ac inoculation of OVA. Stimulator cells were OVA-pulsed PECs as described below. Effector (5 × 10⁵), stimulator (5 × 10⁵), and regulator (5 × 10⁵) cells were mixed and resuspended in 10 μl HBSS for intradermal inoculation into the right ear pinnae of naive mice. Ear swelling was measured with an engineer's micrometer at 24 and 48 h. As a negative control, naive T cells from nonmanipulated mice were used as effector cells and regulator cells. Primed T cells were used as effector cells, and naive T cells from nonmanipulated mice were used as regulator cells for positive control.

PECs were obtained from peritoneal washes of mice 3 d after they received an intraperitoneal inoculation of 2.5 ml of 3% aged thioglycolate solution (Sigma Chemical Co.). After counting, PECs were cultured with OVA (5 mg/ml) in a 24-well culture plate in serum-free medium (RPMI 1640 medium, 10 mM Hepes, 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, 100 U/ml penicillin, 100 μg/ml streptomycin (BioWhittaker), and supplemented with 0.1% bovine serum albumin (Sigma Chemical Co.), ITS⁺ culture supplement (1 μg/ml iron-free transferrin, 10 ng/ml linoleic acid, 0.3 ng/ml Na₂Se, and 0.2 μg/ml Fe(NO₃)₃; Collaborative Biomedical Products). Nonadherent cells were removed from the cultures after 18 h by three washes, and the remaining adherent cells were collected by vigorous pipetting with cold medium (4°C) before washing (three times in HBSS) to remove free OVA.

The Abs used for flow cytometry analysis were as follows: Fc Block™ (anti–mouse FcRγ II/III mAb, 2.4G2), biotin or FITC-conjugated anti-NK1.1 mAb (PK136), biotin-conjugated anti-Ly49C (5E6), biotin-conjugated anti-CD1 mAb (1B1), FITC-conjugated anti-CD3 mAb (145-2C11), CyChrome 5–conjugated anti–TCR-β mAb (H57-597) were all purchased from PharMingen. PE-conjugated anti-B220 mAb (RA3-6B2) and PE-conjugated Mac-1 (M1/70.15) were purchased from CALTAG Laboratories. Rabbit antiasialo GM1 Ab (RαAsGM1) was purchased from Wako Chemicals USA, Inc.; streptavidin-PE was purchased from Jackson ImmunoResearch Labs, Inc. and FITC-conjugated goat anti–rabbit IgG was purchased from Sigma Chemical Co.

The Abs used for in vivo treatment were as follows: anti-NK1.1 mAb (PK136, mouse IgG2a) and anti-Ly49C (5E6, mouse IgG2a) were purified from mouse ascites using protein A column chromatography (GIBCO BRL) in our laboratory. Purified mouse IgG was purchased from Sigma Chemical Co. for use as control for anti-NK1.1 mAb and anti-Ly49C. RαAsGM1 was purchased from Wako Chemicals USA, Inc., and purified rabbit IgG was purchased from Sigma Chemical Co. Anti-CD1 mAb (3C11, rat IgM) was also purified from mouse ascites using protein A columns. Purified rat IgM isotype control (R4-22) was purchased from PharMingen and used as control for anti-CD1 mAb.

Splenic NK and NKT cells were analyzed by flow cytometry. RBCs were lysed by adding Tris-buffered ammonium chloride to a cell pellet of spleen cells. Staining was performed in the presence of saturating concentration of Fc Block™ (blocks FcRγ II/IIIs). Cells were stained with the following three reagents and colors (using concentrations recommended by the manufacturer): biotin-conjugated anti-NK1.1 mAb counterstained with streptavidin-PE; CyChrome 5–conjugated anti–TCR-β chain mAb; and FITC-conjugated anti-CD3 mAb. In some experiments the cells were stained with RαAsGM1 Ab and counterstained with FITC-conjugated goat anti–rabbit IgG and with CyChrome 5–conjugated anti–TCR β chain mAb. Stained cells were analyzed on an EPICS XL flow cytometer (Coulter). The absolute number of splenic NKT cells detected in flow cytometry was calculated from the percent of NKT cells in the number of viable cells. The total number of viable cells harvested from the spleens before staining was determined by the trypan blue exclusion method.

To deplete NK cells in vivo, 100 μl of PBS containing one of the following RαAsGM1 (250 μg), mouse anti-NK1.1 Ab (50 μg), rabbit IgG (250 μg), or mouse IgG (50 μg) was injected into the tail vein of B6 mice. To deplete both NK and NKT cells in vivo, a mixture of anti-NK1.1 mAb and anti-Ly49C (50 μg + 50 μg) or mouse IgG (100 μg) was injected into the tail vein of B6 mice. 24 h later, spleen cells from Ab-treated animals were monitored for the presence of NK or NKT cells by flow cytometry using Abs that detected an NK marker that was different from the target of the Ab used in the depletion treatment. 24 h after the cell depletion treatments, B6 mice were inoculated ac with OVA (50 μg/2 μl). 7 d after ac inoculation, the ability to suppress a primed DTH response was tested in a LAT assay. Before enriching for the regulator T cells from spleens from the ac-inoculated mice, the NK and NKT cells were monitored again to confirm their absence.

Purified anti-CD1 mAb (3C11) or control rat IgM mAb (50 μg in 100 μl PBS) was injected into the tail vein of B6 mice to block the interaction of NKT cells with CD1. It is reported that 3C11 blocks the NKT cell–CD1 interaction in vitro 31. Flow cytometry studies of spleen cells from the 3C11-treated mice confirmed that the CD1⁺ cells (biotin-conjugated anti-CD1 mAb [1B1] counterstained by streptavidin-PE) were neither depleted nor showed changes in the populations of T cells (FITC-conjugated anti-CD3 mAb), B cells (PE-conjugated anti-B22 mAb), NK and NKT cells (triple staining: FITC-conjugated anti-CD3 mAb, CyChrome 5–conjugated anti–TCR β chain mAb, and biotin-conjugated anti-NK1.1 mAb counterstained by streptavidin-PE), and macrophages (PE-conjugated Mac-1) (data not shown).

After RBC lysis, spleen cells were treated with FITC-conjugated anti-NK1.1 mAb, biotin-conjugated anti-Ly49C, and MicroBeads-conjugated anti–mouse pan-NK cells (DX5) (Miltenyi Biotec), and washed twice in PBS (pH 7.2) containing 0.5% BSA and 2 mM EDTA. Ab-labeled cells were treated with anti-FITC MicroBeads and streptavidin MicroBeads (Miltenyi Biotec) for 15 min, and washed twice. To harvest NK and NKT cell–enriched and depleted populations, cells were applied to Type MS⁺ positive selection column with MiniMACS (Miltenyi Biotec). Cells were stained with Cy-Chrome 5–conjugated anti–TCR β chain mAb, and depletion or enrichment was confirmed by flow cytometry. Cell numbers of depleted populations were adjusted to approximate the number used in the control studies.

For reconstitution experiments, CD1⁺ cells were depleted from the spleen cells. Following RBC lysis, column-enriched splenic T cells were incubated with biotin-conjugated anti-CD1 (1B1) and then treated with streptavidin MicroBeads before they were applied to Type MS⁺ positive selection column with MiniMACS. The CD1⁺ cell–depleted population in the effluent wash was counterstained by streptavidin-PE (Jackson ImmunoResearch Labs, Inc.) and analyzed by flow cytometry to confirm the quality of the depletion technique.

CD1 KO mice were γ-irradiated (cesium, 500 rad, Mark 1 irradiator; J.L. Shepherd and Associates) 1 d before receiving 2 × 10⁷/mouse whole spleen cells derived from F1 mice or spleen cells depleted of their NK1.1⁺ cells by magnetic beads. 7 d later, reconstituted CD1 KO mice were inoculated (ac) with OVA (50 μg/2 μl in HBSS). Spleens were removed 1 wk after the ac inoculation, dissociated cells were pooled, and splenic T cells were enriched as described above. Enriched splenic T cells were transferred to naive F1 mice as regulator cells with effector (derived from F1 mice) and stimulator cells (derived from F1 mice) prepared as described above and tested in a LAT assay. Any host versus graft disease that might have occurred was undetectable and did not interfere with the experimental outcome 15 d after reconstitution.

Data were subjected to analysis by analysis of variance and Scheffe's test. A value of P ≤ 0.05 was considered significant.

ACAID is a well-established experimental animal model for the study of immune deviation mediated through an immune-privileged site. Modulation of the model and analysis of the subsequent effects on the immune deviation–associated systemic tolerance can be assessed directly in the whole mouse, or in a LAT assay 14 for detection of antigen-specific regulatory T cells associated with ACAID. The postulate that NKT cells are important in ACAID was first assessed by looking for an ACAID-associated increase of NKT cells in the spleen.

B6 mice were inoculated (ac) with OVA, and 7 d later the spleens were extirpated, cells dissociated, and the numbers of NK and NKT cells were analyzed by flow cytometry after staining for the TCR β chain and the NK1.1 molecule. Analysis was performed on five individual mice per group. The flow cytometry data showed that the ratio of NKT cells to total gated lymphocytes from spleens (depleted of B cells and macrophages) was increased in all ACAID mice compared with naive mice (Fig. 1), subcutaneously or intravenously inoculated mice (data not shown) at 7 d after ac inoculation. In contrast to NKT cells, the number of NK cells did not change in the spleen during ACAID induction. Both the percent and the absolute number of NKT cells in the spleen began to increase as early as day 3 (data not shown), and peaked at day 7 (Fig. 1). These data show an association of splenic accumulation of NKT cells, not NK cells, with the induction of ACAID.

To directly test the involvement of the NKT cells in the development of ACAID, we examined the potential of CD1-deficient mice to develop immune deviation after ac inoculation of antigen. Since the CD1 molecule is essential for NKT cell development 30,31, CD1 KO mice do not have NKT cells, but do have NK cells and other lymphocyte subpopulations 32,33,34. CD1 KO mice and control WT mice were inoculated (ac) with OVA 7 d before subcutaneous sensitization with OVA and CFA, and challenged into the ear pinnae with OVA-pulsed PECs 7 d after subcutaneous sensitization (Fig. 2 A, protocol inset). When ear swelling was measured 24 and 48 h later, it was observed that ac-inoculated CD1 KO mice developed a positive DTH (ear swelling) response, but ac-inoculated WT mice exhibited a suppressed response (Fig. 2 A). While the level of the DTH response in CD1 KO mice was similar to that in WT mice, CD1 KO mice did not develop ACAID.

To further assess the NKT cell-dependent ACAID mechanism, spleen cells from ac-inoculated CD1 KO and WT mice were tested in a LAT assay (Fig. 2 B, black bars, protocol inset). In brief, CD1 KO mice were inoculated (ac) with OVA 7 d before harvesting, dissociating, and enriching T cells from the spleens for use as regulator cells. Regulator T cells were then cotransferred with OVA-primed T cells (effector cells from F1 mice) and OVA-pulsed PECs (stimulator cells from F1 mice) into the ear of F1 mice. In contrast to regulator cells from WT mice, regulator T cells from CD1 KO mice that received OVA were unable to suppress the DTH response (Fig. 2 B). An experiment was performed in B6 CD1 KO mice (hatched bars) to confirm the role of NKT cells in ACAID in a genetically homogeneous background. In this case, effector and stimulator cells were prepared from B6 mice, and then cells were transferred to the B6 recipient. These results show that NKT cells or other CD1-dependent populations were needed for the development of ACAID in general, and for the generation of the antigen-specific efferent regulatory T cells, in particular.

To confirm whether the defect in the CD1 KO mice that led to the failure of ACAID was actually the NKT cell deficiency, we reconstituted CD1 KO mice with whole spleen cells from WT mice (F1) (containing both NKT cells and CD1⁺ APCs), or spleen cells immunomagnetically depleted of NK1.1⁺ lymphocytes (that still contained CD1⁺ APCs). The successful depletion of NKT and NK cells was confirmed by flow cytometry analysis (Fig. 3 A). 7 d after reconstitution, the mice were inoculated (ac) with OVA, and after 8 d spleen cells were harvested for T cell enrichment. A typical profile of enriched splenic T cells from reconstituted CD1 KO mice (Fig. 3 B, bottom left panel) shows that ∼4.9% of T cells were donor derived (CD1⁺). (Total of CD1⁺ cells in the non-T splenic cells was 25%; data not shown.) To analyze the regulatory potential of the host CD1⁻ T cells, the spleen cells were further negatively selected against CD1 expression to yield populations that were CD1⁻ (Fig. 3 B, bottom right panel). These cells were then assessed as regulator cells in the LAT assay.

We observed that whole spleen cells were able to reconstitute the ACAID-inducing ability of the CD1 KO mice. In contrast, spleen cells depleted of NKT cells did not restore ACAID-inducing ability (Fig. 3 C). Because CD1 is expressed on all WT bone marrow–derived cells that were transferred, these results also show that antigen-specific-efferent regulatory cells that were generated did not express CD1, indicating that the transferred CD1⁺ NKT cells were not the effector cell in the regulation of the expression of DTH. The data support the postulate that NKT cells are essential in the induction of ACAID, and function by supporting the generation of antigen-specific efferent-regulatory T cells.

To confirm that NKT cells do not have a direct role in efferent regulation of DTH in another way, we depleted cells expressing NK1.1 antigens from the regulator cell population before testing in a LAT assay using specific Ab treatment and magnetic beads selection (Fig. 4 A). Treated cells were then assessed by flow cytometry before cotransferring with primed T cells and OVA-pulsed PECs to the ear pinnae. The results of the LAT assay showed that the NK and NKT cell–depleted populations retained their DTH-regulatory capability. Thus, the efferent-regulatory cell is a conventional NK1.1⁻ T cell. It is known that activated NKT cells can downregulate their NK1.1 molecules in vitro 35. Therefore, we were aware that the in vivo–activated NKT cell may not express NK1.1, and the Ab depletion of NK1.1⁺ cells treatment may not work. However, as also shown in Fig. 1, after OVA was inoculated in the ac NK1.1⁺ T cells were clearly present in the spleen. Therefore, together with our observations that CD1⁻ T cells (from CD1⁺ NKT cell–reconstituted CD1 KO mice) can function as regulators of DTH (Fig. 3 C), these data confirm that NKT cells are not direct efferent regulators of DTH in ACAID.

Another approach to assess the importance of NKT cells in induction of systemic tolerance used Abs specific for NK markers to create CD1⁺ mice with selective depletion of the NK1.1⁺ cell populations. B6 mice were depleted of NKT cells and NK cells, or NK cells only, and their ability to generate the efferent regulatory T cell in ACAID was compared to syngeneic mice treated with isotype control Ab. We reported previously that it is difficult to remove the NKT cell population in vivo with Abs to typical NK cell markers 36. Specifically, the inoculation of mice with anti-NK1.1 mAb or RαAsGM1 Ab depletes NK cells, but not NKT cells, although they both express the target antigens (36; and Fig. 5 A and B). Because the NKT cells that remained in the spleens after in vivo anti-NK1.1 mAb treatment expressed Ly49C 36, we mixed the anti-NK1.1 and anti-Ly49C Abs and effectively depleted NKT cells in the mice for these studies (Fig. 5 B). 24 h after the mixed Ab treatment, mice were inoculated (ac) with OVA, and the differently treated groups of mice were tested 1 wk later for their ability to generate efferent-regulatory T cells in a LAT assay. As expected, the NK-only depleted mice, previously treated with either RαAsGM₁ or anti-NK1.1 mAb, developed antigen-specific efferent-regulatory T cells (Fig. 5c and Fig. d), but the NKT and NK cell–depleted mice did not (Fig. 5 D). Therefore, together with studies in the CD1 KO mice (Fig. 3), these data show that the CD1-dependent cell responsible for the development of systemic tolerance and the generation of regulatory T cells in ACAID is the CD1-dependent NKT cell.

Clearly CD1 is needed for the development of NKT cells, but it is unknown if CD1 is required for NKT cell function in the generation of regulatory T cells in ACAID. We reasoned that if NKT cell interactions with CD1 were required for ACAID induction, we might be able to block ACAID by blocking the NKT cell interaction with CD1. Previously, anti-CD1 mAbs (3C11) were successfully used in vitro to block NKT cell–CD1 interactions 31. Therefore, mice were treated with anti-CD1 mAb 1 d before being inoculated (ac) with OVA and 8 d before harvesting the spleens, dissociating the cells, and column-enriching the splenic T cells for use as regulator cells in a LAT assay. 24 h after anti-CD1 treatment (3C11), dissociated spleen cells were stained with a different Ab for CD1 (1B1) to assess the presence of CD1⁺ cells in the spleen. Flow cytometry analyses showed that the Ab treatment did not alter the ratio or absolute number of CD1⁺ cells. In addition, anti-CD1 mAb treatment of mice did not alter the ratio or absolute number of NKT and NK cells (stained by anti–TCR-β and anti-NK1.1), T cells (stained by anti-CD3), B cells (stained by anti-B220), and macrophages (anti–Mac-1) in their spleens (data not shown). Groups of mice treated with control Ab developed antigen-specific efferent-regulatory T cells, but mice treated with anti-CD1 mAb did not (Fig. 6). Because the Ab treatment did not eliminate CD1⁺ cells, the most likely explanation is that an interaction between NKT cells and CD1 was blocked by the anti-CD1 mAb. Thus, we postulate that an interaction between the NKT cell and the CD1 molecule is required for the NKT cell to function in ACAID.

Since antigen inoculated into the ac is carried to the spleen, an argument could be made that there are few differences between the induction mechanisms of ACAID and the mechanisms involved in intravenously induced systemic tolerance. In fact, it could be argued that the leakage of antigen from the venules in the eye into the blood during ACAID induces intravenous tolerance. However, differences between the mechanisms involved in immune deviation via the ac or intravenous route have been reported 13. It has also been reported that intravenously induced tolerance generates CD8⁺ afferent-regulatory T cells in contrast to ACAID, which generates both CD4⁺ afferentregulatory T cells and CD8⁺ efferent-regulatory T cells 14.

CD1 KO mice and WT mice were inoculated ac or intravenously with OVA 7 d before subcutaneous sensitization with OVA and CFA and testing for DTH by challenging 14 d later with OVA-pulsed PECs into the ear. As before, ear swelling was measured 24 and 48 h later. As expected, WT mice developed immune deviation regardless of the route of inoculation (Fig. 7). Importantly, in contrast to the inability of CD1 KO mice to develop immune deviation when inoculated ac, intravenously treated CD1 KO mice were fully capable of developing immune deviation, and showed reduced ear-swelling responses (Fig. 7). Therefore, NKT cells do not participate in intravenously induced immune deviation, and ACAID is indeed a separate entity, with unique and locally maintained mechanisms of regulation.

This report shows that NKT cells (a) are required for the development of ACAID; (b) are specifically required for the generation of an efferent-regulatory T cell; (c) may interact with a CD1-expressing cell to mediate regulatory mechanisms for suppressing DTH responses in ACAID; and (d) are not required for intravenous tolerance. In addition, a relationship between ACAID and self-tolerance is suggested by reports that induction of ACAID in mice both prevented the onset of and also suppressed ongoing experimental autoimmune uveitis 37. Furthermore, while several previously published reports imply a role for NKT cells in preventing certain autoimmune disease in humans 38,39 and in mice 20,40,41, the mechanisms through which NKT cells might regulate autoimmunity remain unclear. Therefore, these results indicate that the autoimmunity associated with NKT cell defects (such as type 1 diabetes, scleroderma) may be mediated by disruption of organ-specific tolerance mechanisms that are similar to those mediating systemic tolerance to antigens introduced through immune-privileged sites.

The systemic tolerance to antigens introduced into the ac (ACAID) involves several steps. In the initial step, antigen introduced into the ac is carried to the spleen by specialized eye-derived F4/80⁺ APCs 42. A CD8⁺ efferent or effector stage regulatory T cell that can suppress a subsequent DTH response to the specific antigen is then generated. This study further shows that CD1-reactive NKT cells are required for the generation of the ACAID regulatory T cell. On the other hand, the LAT studies of CD1 KO mice reconstituted with CD1⁺ NKT cells show that CD1d expression by ACAID efferent-regulatory cells was not necessary (Fig. 3 B and C). Anti-CD1 blocking experiments nonetheless indicate a role for CD1, presumably functioning as a ligand for NKT cells in ACAID. Therefore, we suggest that the critical CD1-dependent interaction is between the NKT cell and the specialized ac-derived APC, and that NKT cell interactions with particular APCs may similarly be defective in some autoimmune diseases associated with loss of NKT cell function.

The ac contains aqueous humor that is filled with a mixture of immunosuppressive components including TGF-β 3,4,5,6,43. Moreover, we observed that addition of TGF β2 to thioglycolate-induced PECs in vitro induced an increased expression of their CD1d molecules (Sonoda, K.-H., and J. Stein–Streilein, unpublished observations). Thus, the eye-derived APCs may similarly express increased levels of CD1, and thereby stimulate NKT cells in the spleen. Alternative, and not mutually exclusive, hypotheses are that NKT cells are activated by a CD1-presented endogenous lipid antigen, accessory molecule, or cytokine produced by the APC.

A further alternative is that the NKT cells recognize CD1d expressed by another cell type in the spleen. Such a possible CD1⁺ NKT cell stimulator in the spleen is the marginal zone B cell. Niederkorn and colleagues showed that splenic B cells are needed for ACAID, and suggested that eye-derived APCs “hand over” their antigen and allow the splenic B cells to induce ACAID 44,45. Consistent with this hypothesis is the observation that the APCs in the spleen that express the highest density of CD1 are the marginal zone B cells 46.

Activated NKT cells may produce large amounts of a variety of cytokines 35 that, in the splenic microenvironment, likely contribute to the development of the efferent-regulatory T cell in ACAID. Cytokines, such as IL-4, may directly modulate activity of the efferent-regulatory cell. While IL-4 is most commonly thought to mediate NKT cell–dependent T cell regulation, we do not propose it to be the critical cytokine in NKT cell–dependent generation of the efferent-regulatory T cell since ACAID occurs in IL-4 KO mice 16,47,48. Another potentially important cytokine, already commonly associated with the development of ACAID, is TGF-β 43. A report has been published showing NKT cell production of TGF-β 49, and another shows that TGF-β may modulate APC function 50. Thus, the possibility arises that NKT cells respond to the CD1 or other signals by upregulating TGF-β production or its conversion from latent to active form. Strengthening this possibility is a recent report by Kosiewicz et al. showing that both CD4⁺ CD8⁻ and CD4⁻ CD8⁻ (double negative, DN) T cells from ACAID spleens produced TGF-β 16. NKT cells are notably either CD4⁺ or DN 22.

A significant technical observation in this study is that in our hands NKT cells are resistant to in vivo antibody treatments known to remove NK cells. While others report removal of NKT cells by anti-NK1.1 mAb 51,52, we could only eliminate NK cells and not NKT cells with either anti-NK1.1 mAb or RαAsGM1 Ab alone. However, when we used a mixture of Abs, we effectively eliminated NKT cells as well as NK cells. These results were not surprising, as we previously reported that while activating apoptosis in NK cells, the NK antigen–specific (anti-NK1.1) Ab treatment activated IL-4 synthesis in NKT cells 36. In our laboratory, when the same Ab was used to label cells for flow cytometry that was used for elimination, we could not find the cells because the antigen was either masked or downregulated. Thus, we consistently used different mAbs for depletion studies and the flow cytometric analyses.

In contrast to ACAID, intravenous tolerance could be induced in CD1 KO mice. The relationship between intravenous tolerance and ACAID is unclear, but intravenous tolerance does differ from ACAID in several respects. The intravenous administration of antigen cannot suppress the immune response in previously immunized hosts, whereas presentation of antigen via the ac does downregulate ongoing DTH responses in previously immunized hosts 53. Moreover, ACAID is mediated by both CD4⁺ afferent regulatory T cells and CD8⁺ efferent regulatory T cells 13, whereas intravenously induced tolerance only required CD8⁺ afferent-regulatory cells 14. (In fact, because intravenous tolerance mechanism does not require an efferent-regulatory cell, we could not test the intravenous tolerance capability of CD1 KO mice in a LAT assay.) It is proposed that intravenous tolerance reflects the intrinsic response of T cells to antigen in the absence of costimulatory molecules, and therefore may not require additional cells 54. In contrast, this report clearly shows that ACAID involves interactions between multiple cells, a process that may be necessary in order to suppress established immune responses. A recent paper by Zeng et al. suggests that NKT cells may be responsible for assisting the generation of antigen-specific regulatory T cells in the bone marrow, perhaps to regulate immune responses to self-antigens displayed on other bone marrow–developing cells 55. Thus, it seems likely that CD1-reactive NKT cells may not only be unique regulators of self-reactivity in immune privileged sites (ACAID), but also may contribute to self-tolerance through regional specialization in a variety of organs, tissues, and microenvironments in general.

We especially thank to Dr. J. Wayne Streilein for his many helpful discussions on ACAID and his generosity with ideas and historical facts. We also appreciate his critical reading of the manuscript. We appreciate our collegial interactions with Drs. Takeshi Kezuka, Jie Zhang-Hoover, and Douglas Faunce, as well as their ideas and technical suggestions. We thank Ms. Marie Ortega for management of the Schepens Eye Research Institute–Boston Biomedical Research Institute vivarium and breeding of the transgenic mice used in the experiments. Finally, we thank Mr. Vladimir Russakovsky for his dependable and outstanding technical support, and Ms. Anita Pearson for her preparation of the manuscript and her administrative support in general.

This work was supported in part by grants from the National Institutes of Health: (J. Stein-Streilein, RO1 EY11989-01; and S.P. Balk, RO1 AI42955 and RO1 AI33911), and by the Schepens Eye Research Institute.