Patients with systemic lupus erythematosus (SLE) display reduced numbers and functions of invariant natural killer T (iNK T) cells, which are restored upon treatment with corticosteroids and rituximab. It is unclear whether the iNK T cell insufficiency is a consequence of disease or is a primary abnormality that precedes the onset of disease. To address this, we analysed iNK T cell function at different stages of disease development using the genetically lupus-susceptible NZB × NZW F1 (BWF1) model. We found that iNK T cell in-vivo cytokine responses to an iNK T cell ligand α-galactosylceramide (α-GalCer) were lower in BWF1 mice than in non-autoimmune BALB/c and major histocompatibility complex (MHC)-matched NZB × N/B10.PL F1 mice, although iNK T cell numbers in the periphery were unchanged in BWF1 mice compared to control mice. Such iNK T cell hyporesponsiveness in BWF1 mice was detected at a young age long before the animals exhibited any sign of autoimmunity. In-vivo activation of iNK T cells is known to transactivate other immune cells. Such transactivated T and B cell activation markers and/or cytokine responses were also lower in BWF1 mice than in BALB/c controls. Finally, we show that iNK T cell responses were markedly deficient in the NZB parent but not in NZW parent of BWF1 mice, suggesting that BWF1 might inherit the iNK T cell defect from NZB mice. Thus, iNK T cells are functionally insufficient in lupus-prone BWF1 mice. Such iNK T cell insufficiency precedes the onset of disease and may play a pathogenic role during early stages of disease development in SLE.
Invariant natural killer T (iNK T) cells are CD1d-restricted, lipid antigen-reactive, immunoregulatory T lymphocytes. iNK T cells recognize glycolipid antigens, such as α-galactosylceramide (α-GalCer), and rapidly produce a broad range of cytokines [1-3]. Activation of NK T cells in vivo leads to subsequent transactivation of other immune cells, such as conventional T, B, dendritic and NK cells [4-7]. It is therefore not surprising that iNK T cells are reported to modulate immunity in a broad spectrum of diseases, including allergy, atherosclerosis, autoimmunity, cancer and infections [8-10].
Several studies have shown that the numbers of NK T cells are significantly lower in the peripheral blood of patients with systemic lupus erythematosus (SLE) than in that of healthy controls [11-15]. This NK T cell deficiency correlates with disease activity and is restored to normal in patients treated with corticosteroids or rituximab [13-15]. NK T cell function, as measured by proliferation and cytokine production in response to iNK T cell ligand α-GalCer, is also impaired in patients with SLE [11, 13]. In family members of patients with SLE, the deficiency of iNK T cells is associated with the development of autoantibodies and clinical autoimmunity [16, 17]. These data suggest clearly that iNK T cells are impaired in patients with SLE. However, it is unclear whether the iNK T cell insufficiency is a consequence of disease or is a primary abnormality that precedes the onset of disease.
Animal studies have shown that CD1d-deficiency exacerbates lupus disease in the New Zealand black × New Zealand white (NZB × NZW) F1 (BWF1), MRL-lpr and hydrocarbon oil-induced models of lupus [6, 18, 19], and the deficiency of iNK T cells alone elicits disease in an otherwise normal mouse strain . These data suggest a protective role of CD1d-restricted iNK T cells in lupus. In fact, treatment with an iNK T cell ligand reduces lupus at least in the early stages of disease [21-24]. We have reported that a brief treatment with α-GalCer at a young age confers long-term protection against lupus . However, long-term repeated treatments with αGalCer have little or no effect on lupus  or can even exacerbate lupus nephritis, especially when given at later stages of disease in some animal models [21, 22, 25]. These data raise the possibility of using iNK T cell-based therapy during remission to prevent future disease flares in patients. Hence, understanding the dynamics of iNK T cell responses at different stages of lupus disease may facilitate the development and implementation of iNK T cell-based therapy.
Taken together, human and murine studies indicate that CD1d-restricted T cells may be an important therapeutic target, especially when considering the limited polymorphic nature of CD1 genes with ensuing therapeutic advantage over modulation of T cells restricted by highly polymorphic major histocompatibility complex (MHC) class I and class II systems. Hence, several studies have investigated the numbers and functions of iNK T cells in the peripheral blood of patients in relation to disease activity [11-15] and in lymphoid organs of animals with lupus (reviewed in ). However, analysis of iNK T cells in the peripheral blood of patients and in lymphoid organs of animals might not fully recapitulate the in-vivo status of iNK T cells, as iNK T cells might home to the sites of inflammation such as kidneys in the BWF1 mice . Hence, further studies are needed to understand the systemic in-vivo iNK T cell responses in lupus.
In this study, we assessed systemic in-vivo iNK T cell responses at different stages of disease in BWF1 (H2d/u) mice and in age-matched MHC-related control strains BALB/c (H2d) and NZB × N/B10.PL F1 (BPF1) (H2d/u), and parental strains NZB (H2d) and NZW (H2u). Our results show that iNK T cells display in-vivo hyporesponsiveness in BWF1 mice prior to the onset of disease, indicating a possible pathogenic role of iNK T cell insufficiency in lupus.
Materials and methods
NZB, NZW, BALB/c and B10.PL mice were purchased from the Jackson Laboratory and were bred locally to generate BWF1 and BPF1 mice. Female mice were used in all experiments. Experiments were performed according to the approved institutional research guidelines.
Flow cytometry analysis
Freshly prepared or cultured spleen cells were incubated with anti-CD16/32 (2·4G2; BD PharMingen, Franklin Lakes, NJ, USA) to block FcγRII/III, followed by staining with various conjugated monoclonal antibodies (mAbs) (all BD PharMingen), as indicated in the figure legends. iNK T cells were identified by CD1d/α-GalCer-tetramer staining . Stained cells were analysed using a Becton Dickinson (Mountain View, CA, USA) fluorescence activated cell sorter (FACS)Calibur flow cytometer and CellQuest software. Absolute cell numbers of each cell subset was calculated by the percentage of staining cells × times total cell number.
Mice were injected intravenously (i.v.) with α-GalCer (Kyowa Hakko Kirin Co. Ltd, Tokyo, Japan) dissolved in 0·1 ml vehicle or with vehicle [0·1% polysorbate-20 in phosphate-buffered saline (PBS)] alone. α-GalCer was injected at a 4 μg dose, which has been shown to reduce lupus manifestations in BWF1 mice . Mice were bled and killed 2 h after injection. Their spleen cells (106 cells in 1 ml) were cultured in RPMI-1640 medium supplemented with 10% fetal calf serum (FCS), 2 × 10−5 M 2-mecaptoethanol (ME), 20 mM HEPES, 1 mM sodium pyruvate and 100 μg/ml gentamicin with no stimulation for 2 h to collect the supernatants. The cytokine levels in the sera and the cultured supernatants were assayed by enzyme-linked immunosorbent assay (ELISA) .
In some experiments, spleen cells from α-GalCer- or vehicle-injected mice were cultured with 0∼100 ng/ml α-GalCer or 5 μg/ml concanavalin A (ConA) (Sigma, St Louis, MO, USA). Supernatants were collected after 2 days for cytokine assays by ELISA.
Cytokine secretion assay
To examine the cellular sources of cytokine production upon in-vivo or in-vitro α-GalCer-stimulation, a cytokine secretion assay was performed using the magnetic affinity cell sorting (MACS) cytokine secretion assay kit (Miltenyi Biotec, Auburn, CA, USA), following the manufacturer's protocol with some modifications . Briefly, stimulated or control spleen cells (1 × 107) were incubated at 37°C for 45 min with the cytokine catch reagent. After washing, cells were stained with phycoerythrin (PE)-conjugated cytokine detection antibody, followed by incubation with anti-PE microbeads. Cytokine-secreting cells were then positively selected using AutoMACS (Miltenyi Biotec). Cytokine-enriched cells were counterstained with NK1·1, CD1d/α-GalCer tetramer and T cell receptor (TCR)-β and analysed by flow cytometry. Dead cells and B cells, which can bind non-specifically to cytokine detection antibody via PE, were excluded by staining with propidium iodide and peridinin chlorophyll (PerCP)-conjugated B220 (BD PharMingen), respectively.
The statistical significance was analysed using Student's t- or Mann–Whitney U-tests. Differences were considered significant at P < 0·05.
BWF1 mice exhibit in-vivo iNK T cell hyporesponsiveness
Analysis of isolated iNK T cells from the peripheral blood of patients and spleen of animals may not fully recapitulate the in-vivo status of iNK T cells. Hence, we assessed systemic in-vivo iNK T cell responses in lupus mice by injecting an iNK T cell ligand α-GalCer. Two hours after an i.v. injection of α-GalCer, serum levels of all cytokines tested were significantly lower in BWF1 mice than in control mice in all age groups (Fig. 1a). To test further that this decrease in serum cytokine levels reflects the reduced cytokine production by α-GalCer-primed immune cells, spleen cells from α-GalCer-injected mice were cultured for 2 h with medium alone, and culture supernatants tested for cytokines. As shown in Fig. 1b, interferon (IFN)-γ and interleukin (IL)-4 were markedly lower in BWF1 mice than in control mice. This reduction in cytokine responses in BWF1 mice is particularly significant because iNK T cell numbers were not reduced in the lymphoid organs, except for thymus, of BWF1 mice compared to control mice, as also reported previously . These data indicate clearly a functional insufficiency of iNK T cells in lupus mice.
To evaluate the effect of disease status on iNK T cell responses, iNK T cell responses were analysed in BWF1 mice at pre-autoimmune (7-week), early autoimmune (14-week) and diseased (24-week) stages and compared with levels in age- and sex-matched non-autoimmune mice. We found that the iNK T cell hyporesponsiveness was seen at all ages, but the differences between BWF1 and control mice were generally more profound at a younger age (Fig. 1, and data not shown), suggesting that iNK T cell hyporesponsiveness in lupus is not a consequence of ongoing inflammation.
To determine if iNK T cell function can be restored in lupus mice by subsequent exposure to the glycolipid, we injected α-GalCer i.v., harvested spleen 2 h later, and cultured spleen cells with α-GalCer for 2 days. Although cytokine levels increased in BWF1 spleen cells upon recall with the same glycolipid, cytokine levels were still markedly lower in BWF1 mice than in BALB/c mice (Fig. 1b). Thus, recall iNK T cell responses are also impaired in BWF1 mice compared to control BALB/c mice.
Transactivation of other immune cells is an important function of iNK T cells [27-29]. To determine such transactivation of conventional T cells by in-vivo α-GalCer stimulation in BWF1 mice, we treated animals with α-GalCer or vehicle and cultured their spleen cells with ConA for 2 days. As shown in Fig. 2a, α-GalCer-primed mice had increased levels of most cytokines tested upon in-vitro ConA stimulation compared to vehicle-injected mice (P all < 0·01, except for IL-10). Such transactivation of T cells was always lower for all cytokines tested in BWF1 mice than in BALB/c mice at all ages tested (P all < 0·01, Fig. 2b). Expression of activation and memory markers on transactivated T cells (Fig. 2c) and of activation and co-stimulatory molecules on transactivated B cells (Fig. 2d) was also lower in BWF1 mice than in control BALB/c mice. Thus, in-vivo α-GalCer-induced transactivation of T and B cells is reduced in lupus mice compared to non-autoimmune mice.
Most cytokine-secreting cells immediately after in-vivo α-GalCer exposure are iNK T cells, whereas transactivated NK or T cells account for most cytokine-secreting cells after in-vitro stimulation
In contrast to markedly reduced in-vivo iNK T cell responses, cytokine responses of spleen cells cultured with α-GalCer for 48 h did not differ significantly between BWF1 and BALB/c mice (Fig. 3). Thus, lupus iNK T cells may retain the ability to respond in vitro in the absence of in-vivo milieu of BWF1 mice. In order to reconcile the discrepancy in our result showing reduced in vivo (Fig. 1), but largely intact in vitro (Fig. 3), iNK T cell responses in BWF1 mice, we performed a cytokine secretion assay using a cytokine catch reagent (Miltenyi Biotech) to identify cytokine-secreting cells. Two hours after a single i.v. injection of 4 μg α-GalCer, up to 2% of live B220− splenic lymphocytes in α-GalCer-primed BWF1 mice secreted IFN-γ, IL-2 or IL-4 (Fig. 4a). Most cytokine+ cells were CD1d/α-GalCer tetramer-positive (Fig. 4a). To confirm this further, spleen cells (1 × 107) were enriched for cytokine+ cells (Fig. 4b). Among gated cytokine-secreting cells, 81–92% cells were iNK T cells (TCR-β+tetramer+) (Fig. 4c). The remaining cytokine-secreting cells were TCR-β+ tetramer− conventional T cells (7–11%) and NK (TCR-β−NK1·1+) cells (6% for IFN-γ and < 1% for IL-2 or IL-4). Thus, IFN-γ, IL-2 or IL-4 secreting cells after brief in-vivo α-GalCer exposure are mainly iNK T cells.
Next, to identify which cells secrete cytokines upon in-vitro α-GalCer exposure in BWF1 mice, we cultured their spleen cells with α-GalCer and performed a cytokine secretion assay. As shown in Fig. 5a, percentages of IFN-γ-secreting cells were higher (3·2%) than IL-2-secreting (0·7%) or IL-4-secreting (1·2%) cells after 21 h of α-GalCer stimulation. However, 26% of IFN-γ secreting cells were NK cells (TCR-β−NK1·1+) and 19% were TCR-βhigh cells that are likely to be conventional T cells (Fig. 5b), as iNK T cells express intermediate levels of TCR-β . As expected, most (> 88%) cytokine-secreting cells in ConA-stimulated cultures were TCR-βhigh cells. Cytokine levels in culture supernatants essentially followed the same pattern as in the cytokine secretion assay (Fig. 5d). These data show that cytokine secretion shifted to a T helper type 1 (Th1) pattern 21 h after α-GalCer exposure, owing to increased IFN-γ-producing transactivated NK cells and T cells at this time-point.
Thus, a 2-h time-point appears to represent more accurately the direct in-vivo iNK T cell cytokine responses after α-GalCer injection, whereas transactivated conventional T and NK cells probably contribute to overall cytokine levels at later time-points after α-GalCer exposure in BWF1 mice.
BWF1 mice probably inherit iNK T cell hyporesponsiveness from their NZB parent
The iNK T cell defect in autoimmune-prone non-obese diabetic mice appears to be controlled genetically . Hence, we investigated whether one of the two parents, NZB and NZW, of BWF1 mice confer this defect. We found that 2 h after a single α-GalCer injection, NZB mice had the lowest serum levels of all cytokines tested among the three strains (Fig. 6a). Compared to the NZW parental strain, serum IL-4 and IL-13 levels were unchanged or reduced in BWF1 mice; however, serum IFN-γ levels were increased. The increased circulating IFN-γ in BWF1 mice compared to NZW mice may not necessarily represent an iNK T cell response to α-GalCer. Hence, we rested splenocytes harvested from α-GalCer injected mice for 2 h in cultures with no further stimulation (Fig. 6b), or stimulated splenocytes further with α-GalCer for 48 h (Fig. 6c). As shown in Fig. 6b, the levels of IFN-γ and IL-4 in 2 h culture supernatants were also the lowest in NZB mice. NZW mice had higher IFN-γ and similar IL-4 compared to BWF1 mice. In-vitro recall response to α-GalCer was also the lowest in NZB spleen cells among the three strains (Fig. 6c). Conventional T cell cytokine responses upon ConA stimulation in α-GalCer-injected animals were also lower in NZB mice than in NZW and BWF1 mice (Fig. 6d). Reduced iNK T cell responses in NZB mice compared to NZW and hybrid BWF1 mice suggests the possibility that BWF1 mice probably inherit iNK T cell hyporesponsiveness from their NZB parents.
In this study we demonstrate that, despite no reduction in iNK T cell numbers in organs other than thymus, in-vivo iNK T cell responses were lower in BWF1 mice than in non-autoimmune mice. BWF1 mice might inherit such functional insufficiency of iNK T cells from their NZB parent, as NZB mice had lower iNK T cell responses than all other strains tested. The iNK T cell hyporesponsiveness was seen at all ages, especially at a young age when animals have no evidence of autoimmunity. Although, circulating IL-4 and IFN-γ levels as detected by a highly sensitive in-vivo cytokine capture assay are slightly higher in BWF1 mice than in BALB/c mice , these minor differences do not explain the marked differences in cytokine responses to (Fig. 1). Furthermore, using a sandwich ELISA we detected no to minimal cytokines in the sera of BWF1 and BALB/c mice treated with vehicle alone. This suggests that iNK T cell hyporesponsiveness in lupus is not a consequence of ongoing inflammation or differences in baseline cytokine levels, but is a primary abnormality that may play a pathogenic role during the development of disease.
Previous studies in animal models of lupus (summarized in ) suggest that iNK T cell numbers are reduced in the thymus of all mouse models studied (MRL-lpr, MRL-fas+/+, BWF1 and pristane-induced model) at least at a young age [6, 18, 24, 26]. In the spleen, iNK T cells are also reduced in MRL-lpr mice compared to MHC-matched control strains , but their numbers remain unaffected in the BALB/c pristine-induced lupus model  and in BWF1 mice compared to MHC-matched CWF1, NZW and BALB/c control mice  or control B6 mice . Another study found increased numbers of iNK T cells in the spleen, liver and kidneys of BWF1 mice compared to B6 mice . Despite this increase in iNK T cell numbers, the in-vivo iNK T cell responses to α-GalCer are lower in BWF1 mice than in control strains. The reduced blood levels of cytokines upon α-GalCer injection may reflect the reduced iNK T cells in the thymus. However, older BWF1 mice that have no reduction in thymic iNK T cell numbers  still have reduced cytokine levels in response to α-GalCer injection. This suggests that iNK T cells must be severely hyporesponsive to glycolipid antigens in BWF1 mice.
We found that iNK T cells accounted for the bulk of cytokine-producing cells after a brief (2-h) α-GalCer exposure (Fig. 4), whereas transactivated NK and/or conventional T cells were major cytokine-producing cells after 21 h of α-GalCer exposure (Fig. 5). These observations may explain the results of previous studies that found increased IFN-γ levels after 6, 18 or 23 h of injecting α-GalCer in BWF1 mice compared to B6 mice [25, 26], which may be due to the increased IFN-γ production by transactivated NK or other immune cells in BWF1 mice. Thus, the timing of the analysis of iNK T cell responses after antigen exposure and robustness of transactivation of other immune cells may all dictate the eventual outcome of iNK T cell activation.
We found a few differences in cytokine levels between serum samples (Fig. 1a) and cultured splenocytes (Fig. 1b). For example, the initial response of IL-4 in serum of 7-week-old mice was not different, but the amount of IL-4 in cultured splenocytes was greatly different, between BWF1 and BALB/c mice. In-vivo production of cytokines in organs other than spleen and/or consumption of these cytokines might account for this difference.
Impairment of iNK T cell responses, which precedes the onset of disease (Fig. 1), is probably relevant in the pathogenesis of autoimmune diseases. Animal strains such as SJL mice, which have reduced iNK T cell numbers and responses, develop more severe autoimmune diseases, such as experimental autoimmune encephalomyelitis and induced lupus [22, 33, 34]. In autoimmune-prone MRL strains, the extent of reduction of iNK T cell numbers and responses appears to correlate with the severity of autoimmune disease that develops in these mice; MRL-lpr mice that develop severe disease in early life had a more profound reduction in NK T cell numbers and functions than MRL-fas+/+ mice that experience a milder disease course [19, 24]. A single injection of hydrocarbon oil pristane that induces lupus-like disease a few months later elicits dysfunction of iNK T cells within hours [6, 22]. A further reduction of iNK T cells elicits lupus-like disease in aged Jα18-deficient B6 mice  and accelerates lupus disease in CD1d-deficient pristane-injected BALB/c, BWF1 and MRL-lpr mice [6, 18, 19]. However, iNK T cell hyporesponsiveness alone might not be sufficient to elicit a classical lupus disease, as NZB mice that exhibited lower iNK T cell responses than BWF1 mice (Fig. 6) develop autoimmune haemolytic anaemia, but not a classical anti-nuclear autoantibody response and lupus nephritis [35, 36].
It is unclear how the decreased cytokine production by iNK T cells predisposes to the development of autoimmune disease. The reduced ability to make cytokines may be a manifestation of the reduced functionality of iNK T cells. Activated iNK T cells can normally inhibit autoantibody production [37, 38], in part by inducing activation-induced cell death of autoreactive marginal zone B cells . Hyporesponsive iNK T cells may be inefficient in their surveillance against autoreactive B cells, resulting in pathological autoantibody production.
As seen in the case of protein antigen-specific immune responses, secondary recall responses to α-GalCer (in-vivo injection followed by in-vitro restimulation) were higher than the primary in-vitro or in-vivo iNK T cell responses (Fig. 1b). Encouragingly, this increase in secondary iNK T cell responses occurred in both normal and lupus mice, although it was still lower in BWF1 mice than in control mice. These data suggest that iNK T cell hyporesponsiveness in BWF1 mice might be amenable, at least in part, to correction. In fact, a short-term treatment of young BWF1 mice with two injections of α-GalCer conferred long-term protection against lupus nephritis in BWF1 mice . In humans with SLE, NK T cell insufficiency was restored to normal in patients treated with corticosteroids or rituximab [13-15].
Mechanisms underlying iNK T cell insufficiency in lupus remain unclear. Previous studies have shown that iNK T cells expand initially upon α-GalCer stimulation [24, 30, 39], but may then switch to an anergic state in which the cells are unresponsive for a long time to subsequent doses of α-GalCer . We have reported that repeated administration of α-GalCer for several weeks worsens iNK T cell hyporesponsiveness in BWF1 mice . Thus, continuous in-vivo stimulation by self-glycolipid ligands might cause in-vivo iNK T cell hyporesponsiveness in lupus. In fact, dendritic cells isolated from BWF1 spleens elicit a more robust cytokine response by iNK T cells than dendritic cells isolated from an MHC-identical strain CWF1 (R. R. Singh et al., manuscript in preparation). A study in humans has shown that B cells are essential for iNK T cell expansion and activation in healthy donors, but fail to exert a similar effect in SLE patients . Other studies have implicated a role of site of antigen exposure in causing iNK T cell anergy; for example, a systemic administration of α-GalCer can result in iNK T cell anergy, whereas local administration of α-GalCer, such as by intranasal delivery, can avoid anergy induction . Thus, the duration and route of antigen exposure and status of antigen-presenting cells that play a role in iNK T cell anergy induction [42, 43] may be responsible for iNK T cell hyporesponsiveness in BWF1 mice. It also remains to be determined whether signalling molecules, such as programmed death (PD)-1 [44, 45] and Cbl-b , that are involved in iNK T cell anergy induction in normal mice are impaired in lupus.
In summary, reduced responsiveness of iNK T cells prior to the onset of disease in lupus-prone mice and the correlation of iNK T cell insufficiency with disease activity in SLE patients, as well as restoration of iNK T cell defect upon treatment in SLE patients, suggest an important role of iNK T cells in the pathogenesis of SLE. Improved understanding of mechanisms underlying iNK T cell defects in SLE and the development of strategies to overcome iNK T cell hyporesponsiveness will have important implications for iNK T cell-based therapies in SLE.
We thank Luc Van Kaer for helpful suggestions and Kyowa Hakko Kirin Co, Ltd (Tokyo, Japan) for providing synthetic αGalCer and β-GalCer. This work was supported in part by grants from the National Institutes of Health (R01 AI80778, R01 AR56465, R03 HD67413 and K08 AR62593), the American Heart Association (Western States Beginning-Grant-in-Aid), and Arthritis Foundation/American College of Rheumatology Research and Education Foundation Career Development Bridge Award.
The authors have no conflict of interest to disclose.