Regulatory Role for NK Cells in a Mouse Model of Systemic Juvenile Idiopathic Arthritis

Systemic juvenile idiopathic arthritis (sJIA) is a chronic childhood disorder of unknown cause, characterized by arthritis and systemic inflammatory features, including rash and spiking fever (1). Clinically, patients with sJIA often have mild hepatosplenomegaly and lymphadenopathy associated with hematologic features such as anemia, granulocytosis, thrombocytosis, and increased levels of IL-6, IL-18, and S100 proteins (1–4). Approximately 10–15% of sJIA patients develop macrophage activation syndrome (MAS), an acute, potentially fatal complication of childhood disorders (2, 5, 6). MAS is categorized as a subtype of the hemophagocytic lymphohistiocytosis (HLH) diseases, a syndrome characterized by systemic hyperinflammation, hypercytokinemia, the presence of hemophagocytic macrophages in bone marrow aspirates of patients, and organ dysfunction (7). The excessive activation of T cells and macrophages in HLH and MAS has been linked to the defective cytotoxicity of CD8⁺ T cells and NK cells (8).

NK cells are innate immune cells best known for their ability to kill infected and malignant cells by the release of cytotoxic proteins (i.e., perforin and granzyme B) (9). In addition, NK cells have an important immune-modulating and regulatory functions by production of cytokines (i.e., IFN-γ and TNF-α) and by elimination of activated and autoreactive autologous immune cells to terminate inflammation (10, 11). NK cell activity is conditioned by the balance of activating and inhibitory receptors expressed on their cell surface and the expression of corresponding ligands on potential target cells (12). In sJIA patients, subtle changes of inhibitory and activating receptors have been shown, including decreased expression of inhibitory receptor KLRG1 and increased expression of activating receptor NKp44 (13).

In HLH, defects in the cytotoxic machinery of NK cells and cytotoxic T cells, including mutations in cytolytic effector molecules (e.g., PRF1) and in molecules involved in the cytotoxic granule release (e.g., UNC13D, STX11, and RAB27A, among others) underlie the pathogenesis (7, 14, 15). With respect to sJIA patients, some studies report decreased numbers or defective cytotoxic function of NK cells (16–18), which may explain why sJIA patients are predisposed to develop MAS (17, 19, 20). Other studies found no evidence for any NK cell defects in sJIA, and thus the role, if any, of NK cells in the pathogenesis of sJIA and/or MAS remains elusive (13, 16–18).

We previously described a mouse model that reflects clinical, hematological, and histopathological features of sJIA, relying on a single injection of CFA (containing heat-killed Mycobacterium butyricum) in IFN-γ knockout (KO) mice (21). Thus, CFA-challenged IFN-γ KO mice developed a chronic immune disorder characterized by arthritis, splenomegaly, lymphadenopathy, anemia, granulocytosis, thrombocytosis, and increased serum levels of IL-6. In contrast, wild-type (WT) mice challenged with CFA only developed mild and transient inflammatory disease. In a recent study, we demonstrated that NK cells of sJIA patients have a defective IFN-γ production (13). These findings emphasize the clinical relevance of the mouse model and provide a rationale for studying the role of NK cells in the pathogenesis of sJIA.

In the current study, we investigated the role of NK cells in the mouse model of sJIA. First, we comprehensively analyzed NK cell phenotype and function in IFN-γ KO mice and WT mice upon challenge with CFA. Second, by using different approaches we investigated the effect of NK cell targeting on the development of sJIA-like symptoms. Finally, to gain insight into the immunoregulatory mechanism of NK cells, we explored their capacity to eliminate activated autologous cells in vitro and ex vivo in the context of sJIA. Our data show that intact NK cell functioning is important to keep CFA-induced inflammation under control and to prevent development of sJIA-like disease.

Specific pathogen-free IFN-γ KO and WT BALB/c mice were bred in the Experimental Animal Centre of the Rega Institute, KU Leuven. The sJIA mouse model was induced in mice of 6–8 wk old, by s.c. injection of CFA (Difco) with added heat-killed Mycobacterium butyricum (1.5 mg/ml), as described by Avau et al. (21). Mice were euthanized between 19 and 28 d after CFA injection, when most fulminant clinical and hematological sJIA-like symptoms are present (21). The end-stage criteria for CFA-injected IFN-γ KO mice include significant weight loss and rash at the site of CFA injection, and as soon as the end point criteria are reached in one group of mice (usually in CFA-challenged IFN-γ KO mice), animals of all groups are euthanized. Of note, the onset of symptoms varies between experiments (usually between day 15 and 30). Age- and sex-matched noninjected or naive (NI) WT and IFN-γ KO littermates were included as controls. For NK depletion experiments, mice were treated i.p. with anti–IL-2Rβ Ab (clone TM-β1, 200 μg/mouse; Bioceros), anti–Asialo-GM1 Ab (anti-Asialo, polyclonal, 50 μl/mouse; Wako Laboratory Chemicals), or PBS (control group), starting 1 d prior to CFA immunization. Blocking of the NKG2D receptor was obtained with anti-NKG2D (clone CX5, 200 μg/mouse; Bioceros) (22). For depletion of CD4⁺ and CD8⁺ T cells, mice were treated with anti-CD4 Ab (clone GK1.5, 200 μg/mouse; Bioceros) or anti-CD8 Ab (clone YTS169, 200 μg/mouse; Bioceros), respectively, starting 1 d prior to CFA immunization. All experiments were approved by the Ethics Committee of KU Leuven (P182/2014).

Blood samples were obtained by heart puncture with heparin (LEO Pharma). Complete blood cell analysis was performed with a Cell-Dyn 3700 apparatus (Abbott Diagnostics). Plasma was isolated and stored at −80°C. IL-6 protein levels were measured by ELISA (IL-6 Quantikine; R&D Systems).

For histologic analysis, tissues were fixed in 10% formalin (VWR) and embedded in paraffin. Sections were stained with H&E staining (Merck). Spleens were fragmented and passed through a cell strainer to obtain single-cell suspensions. RBC lysis was performed with ACK lysing buffer (Life Technologies). Splenocytes (1 × 10⁵) were prepared for cytospin and stained by use of H&E staining (Merck). Cytospins were counted for the number of lymphocytes, monocytes, neutrophils, and immature neutrophils per 100 cells. Immature neutrophils were identified by their characteristic nuclear morphology, (i.e., ring-shaped nucleus). NK cells, CD4⁺ T cells, monocytes, and dendritic cells (DC) were enriched with EasySep mouse NK cell enrichment kit (negative selection), EasySep mouse CD4⁺ T cell isolation kit (positive selection), EasySep monocyte enrichment kit (negative selection), or EasySep CD11c⁺ isolation kit (positive selection), respectively, according to the manufacturer’s protocols (Stemcell technologies).

Splenocytes (5 × 10⁵) were cultured in α-MEM containing 10% FCS, penicillin (100 U/ml), and streptomycin (100 μg/ml) in chamber slides (Thermo Fisher Scientific). For differentiation to osteoclasts, the cells were incubated with receptor activator of NF-κB ligand (RANKL; 100 ng/ml; R&D Systems) and M-CSF (20 ng/ml; R&D Systems). After 3 d, the cells were restimulated with fresh medium and stimuli for 3–4 d, followed by staining of the cells for tartrate-resistant acid phosphatase (TRAP) as described (23, 24).

NK cytotoxic activity was measured by ⁵¹Cr release assay. NK cell–sensitive target cells included MHC class I–deficient YAC-1 and RMA-S target cells and NKG2D-ligand (Rae)–overexpressing RMA-Rae target cells (25–27). As control, the NK cell–irresponsive RMA cell line was included (28). ⁵¹Cr-labeled target cells were cultured with NK cells at the indicated E:T ratio for 4 h, and release of ⁵¹Cr was measured. The specific lysis was calculated as [(experimental release − spontaneous release)/(maximal release − spontaneous release)] × 100. Degranulation capacity of NK cells was measured by flow cytometric analysis of CD107a surface expression after 4 h incubation with target cells (E:T 1:1).

Cells were incubated with FcR block (Miltenyi Biotec) and stained with mAbs against mouse Asialo-GM1–Alexa Fluor 488 (polyclonal), CCR2-PECy7 (clone SA203G11), CD3-FITC, PerCPCy5.5, PECy7 or BUV737 (145-2C11), CD4-allophycocyanin eFluor 780 (RM4-5) or PECy5 (GK1.5), CD8-PECy5 or BV711 (53-6.7), CD11b-BUV395 (M1/70), CD11c-BV711 or PerCPCy5.5 (N418), CD19-allophycocyanin R700 (1D3), CD49b-BV421, PE or allophycocyanin (DX5), CD69-PECy7 (H1.2F3), CD80-BV421 (16-10A1), CD86-PE (GL1), CD107a-Alexa Fluor 647 (1D4B), CD122-BV421, PECy7 or BUV395 (TM-b1 or 5H4), Granzyme B-FITC (NGZB), KLRG1-PerCPCy5.5 (2F1), Ly49A-PE (A1), Ly49C/E-biotinylated (4D12) stained with streptavidin-allophycocyanin Cy7, Ly49C/I-PE (5E6), Ly49G2-FITC (eBio4D11), Ly6C-FITC, PerCPCy5.5 (HK1.4) or allophycocyanin Cy7 (AL-21), Ly6G-BV786 (1A8), MHC class II-PE (I-A/I-E; M5/114.15.2), NKG2A/C/E-PerCPCy5.5 (20d5), NKG2D-allophycocyanin (CX5 or C7), NKp46-PE (29A1.4), Perforin-PE (eBioOMAK-D), Rae-1-allophycocyanin (186107), TCRγδ-PE (eBioGL3). All Abs were from BD Biosciences, Thermo Fisher Scientific, BioLegend, or R&D Systems. Dead cells were excluded using Zombie Aqua 516 (BioLegend). Flow cytometry was performed on a BD LSR Fortessa X20 with DIVA software. Results were analyzed with FlowJo (V10). The flow cytometry plots showing the gating strategy for cellular identification are depicted in Supplemental Fig. 1.

Data were analyzed using GraphPad Prism 8 software. For comparison of two different groups, the two-tailed nonparametric Mann–Whitney U test was used for unpaired data. When three or more groups were included, data were first checked for statistical difference with the nonparametric Kruskal–Wallis test.

CFA challenge of IFN-γ KO mice results in the development of a sJIA-like disease, whereas WT mice only present a mild transient inflammation, as described by Avau et al. (21). In a first set of experiments, we investigated NK cell phenotype and cytotoxic capacity in the sJIA mouse model. Mice were euthanized around day 21 after CFA challenge, the time period where the animals showed most clinical and hematological sJIA symptoms (with exception of arthritis that manifests later) (21).

Flow cytometric analysis of spleen NK cells showed significantly reduced NK cell numbers in CFA-challenged IFN-γ KO mice (Fig. 1A). Furthermore, we found increased levels of inhibitory receptors NKG2A/C/E and Ly49A on NK cells of CFA-challenged IFN-γ KO mice as compared with their NI counterparts (Fig. 1B, 1C). The inhibitory receptors Ly49C/I and Ly49G2 showed a tendency of increased levels in CFA-challenged IFN-γ KO mice as compared with their NI counterparts, and Ly49C/E was significantly higher in CFA-challenged IFN-γ KO mice compared with their WT counterparts (Fig. 1C). In addition, the activating receptor NKG2D and maturation marker KLRG1 displayed decreased expression on NK cells of CFA-challenged IFN-γ KO mice in comparison with CFA-challenged WT mice (Fig. 1D).

NK cell lysis is regulated by the balance between inhibitory and activating receptors (29), indicating a potential change in the cytotoxic capacity downstream of these receptors. In line with this, we observed decreased levels of the NK cell cytotoxic proteins granzyme B and perforin in CFA-challenged IFN-γ KO mice compared with CFA-challenged WT mice (Fig. 1E), although in both genotypes, CFA induced a small increase of granzyme B and perforin as compared with NI controls. To measure the effective cytotoxicity of NK cells, their killing capacity was tested via ⁵¹Cr-release assay on three NK cell–sensitive tumor target cell lines (RMA-S, RMA-Rae, and YAC-1) and a control cell line irresponsive to NK cells (RMA). RMA-S and YAC-1 target cells have low to no MHC class I expression, whereas the RMA-Rae cell line has high expression of Rae, a ligand of the activating NKG2D receptor, rendering those cell lines more susceptible to NK cell–mediated killing (25–27). NK cell–mediated target cell lysis was comparable in all the tested NK cell–sensitive target cell lines. As can be seen in Fig. 1F, NI IFN-γ KO mice showed an inherently lower cytotoxic capacity than WT mice, which confirms the original data published by Dalton et al. (30). CFA failed to restore the defective cytotoxic killing in IFN-γ KO mice. In WT mice, injection with CFA caused a decreased target cell lysis in comparison with NI mice, but the killing was still significantly higher in all tested target cell lines as compared with CFA-challenged IFN-γ KO mice (Fig. 1F).

From these data, we can conclude that CFA-challenged IFN-γ KO mice present with a changed NK cell phenotype and a defective cytotoxic capacity against MHC class I–deficient and NKG2D-stimulating tumor target cells.

The differences in inflammatory and hematological manifestations between IFN-γ KO and WT mice upon challenge with CFA may be causally linked to their NK cell activity. To verify this and to examine the in vivo role of NK cells in CFA-induced inflammation, we depleted NK cells in WT mice and investigated whether this results in sJIA-like disease (as seen in CFA-challenged IFN-γ KO mice). NK cells were depleted by injection of Abs against IL-2R β-chain (anti–IL-2Rβ) or Asialo-GM1 (anti-Asialo), two generally accepted Abs to deplete NK cells in mice with a BALB/c background (31–33), in which the depleting anti-NK1.1 Abs cannot be used (34). Of note, IL-2Rβ and Asialo-GM1 are constitutively expressed by NK cells, but also by some other cells such as subpopulations of CD4⁺ T cells, CD8⁺ T cells, NKT cells, macrophages and eosinophils (35, 36). Given the potential cross-reactivity with other cell types, we analyzed the depletion capacity of both Abs in CFA-injected WT mice. In anti–IL-2Rβ- and anti-Asialo–treated mice, a significant reduction of NK cells was seen compared with CFA-injected control mice and was accompanied by a minimal decline in CD8⁺ T cells and CD4⁺ T cells. Macrophage numbers in anti–IL-2Rβ–treated mice were equal to those in CFA-treated control mice, whereas anti-Asialo–treated mice showed a reduction in macrophages compared with CFA-injected WT mice (Supplemental Fig. 2).

When compared with CFA-challenged IFN-γ KO mice, anti–IL-2Rβ and anti-Asialo-GM1 Abs both depleted NK cells in CFA-challenged WT mice to numbers comparable to those found in the diseased IFN-γ KO mice (Fig. 2A). Analysis of the development of sJIA-like symptoms showed that depletion of NK cells in CFA-injected WT mice resulted in an increased number and severity of sJIA-like features, comparable to CFA-challenged IFN-γ KO mice. Thus, NK-depleted mice developed significantly increased splenomegaly (Fig. 2B, left) and lymphadenopathy (Fig. 2B, right) compared with CFA-challenged WT mice. Blood analysis revealed granulocytosis in anti-Asialo– and anti–IL-2Rβ–treated mice compared with CFA-challenged WT mice (Fig. 2C, left). Anemia was not observed in NK-depleted CFA-challenged WT mice (Fig. 2C, center). Of note, in these experiments anemia was observed in only one CFA-challenged IFN-γ KO mouse, which is in line with our original publication in which ∼20% of diseased mice develop anemia (21). Thrombocytosis was observed in anti-Asialo–treated CFA-injected WT mice, and platelet counts were significantly increased compared with NI WT mice but not in the anti–IL-2Rβ–treated group (Fig. 2C, right). Other sJIA-like features, such as the pronounced increase of immature neutrophils in the spleen, were analyzed (21). NK cell depletion in WT mice resulted in an increase of immature neutrophils comparable to CFA-challenged IFN-γ KO mice (Fig. 2D). IL-6 plasma levels were found to be normal in anti-Asialo–injected mice but significantly increased upon depletion with anti–IL-2Rβ compared with CFA-challenged WT mice (Fig. 2E). Arthritis was not observed in the experiments, most likely because the mice were sacrificed before onset of arthritis [joint inflammation is a late feature seen in CFA-challenged IFN-γ KO mice (21)]. A parameter that may predict the potential development of arthritis is the formation of TRAP⁺-multinucleated osteoclasts ex vivo as described by Imbrechts et al. (24). Splenocytes were stimulated with osteoclast-differentiating factor RANKL and M-CSF, and after 7 d the osteoclasts were stained for TRAP. NK cell depletion by anti-Asialo and to a lesser extent by anti–IL-2Rβ increased the formation of osteoclasts. Large osteoclasts, containing more than 50 nuclei, were only seen in splenoctye cultures from mice injected with anti-Asialo (Fig. 2F). NK-depleted WT mice also showed a higher incidence of skin rash and massive inflammation at the site of CFA injection compared with nondepleted counterparts and were in this respect comparable to IFN-γ KO mice (data not shown). Histological analysis of spleen, liver, lymph nodes, and lungs did not reveal any additional abnormalities in NK cell–depleted mice as compared with WT and IFN-γ KO counterparts (data not shown). Of note, depletion of NK cell in CFA-challenged IFN-γ KO mice did not alter the disease course or symptoms (Supplemental Fig. 3).

Because we found diminished levels of CD4⁺ and CD8⁺ T cells upon injection of anti–IL-2Rβ and anti-Asialo (Supplemental Fig. 2), possibly as a consequence of cross-reactive depletion, an experiment was performed in which CD4⁺ T cells or CD8⁺ T cells in CFA-challenged WT mice were specifically depleted by anti-CD4 and anti-CD8 Abs. As can be seen in Supplemental Fig. 4, such a treatment had no effect on the development of sJIA symptoms, indicating that the reduction of CD4⁺ and CD8⁺ T cells by anti–IL-2Rβ and anti-Asialo may only have minimal effects. As a whole, these data indicate that the observed development of sJIA-like disease upon administration of anti–IL-2Rβ and anti-Asialo in CFA-challenged WT mice can be attributed to the depletion of NK cells, with only minimal effects by cross-reactivity.

Taken together, depletion of NK cells in CFA-challenged WT mice results in the emergence of a sJIA-like phenotype. This is in line with a regulatory role of NK cells on the development of inflammatory manifestations as seen in sJIA.

To provide insight into the role of NK cells at the molecular level, we opted to use a neutralizing Ab against NKG2D, a NK cell activating receptor. We used the clone CX5, which is a nondepleting anti-NKG2D Ab (22), and therefore this procedure also offers an alternative approach for studying the role of NK cell activity without depleting the cells. Because NKG2D expression was significantly reduced in CFA-challenged IFN-γ KO mice compared with CFA-injected WT mice (Fig. 1D), we hypothesized that blocking NKG2D in WT mice would result in sJIA-like disease as seen in IFN-γ KO mice.

Treatment of CFA-challenged WT mice with anti-NKG2D–blocking Ab induced characteristic symptoms of sJIA comparable to CFA-challenged IFN-γ KO mice such as increased splenomegaly and lymphadenopathy (Fig. 3A, 3B). Blood analysis showed increased granulocytosis and anemia in anti-NKG2D–treated mice as compared with CFA-challenged WT mice (Fig. 3C, 3D). Thrombocytosis was not observed in anti-NKG2D–treated CFA-challenged WT mice (Fig. 3E). Increased ex vivo formation of osteoclasts, as seen in NK-depleted mice, was not found in animals that received the anti-NKG2D–blocking Ab (data not shown).

Taken together, the neutralization of the NKG2D pathway seems to underlie some of the symptoms in this sJIA mouse model.

From the experiments described above, it can be concluded that sufficient numbers of functional NK cells are required to prevent immune triggering by CFA from escalating into a chronic inflammation as seen in sJIA. Several research reports have shown that NK cells are equipped to lyse activated immune cells, including CD4⁺ T cells, macrophages, and DC (37–42), making NK cells potentially important mediators for the clearance of activated immune cells in vivo. It has been suggested that impaired NK cytolytic activity in sJIA and MAS patients may interfere with lysis of activated immune cells (13, 43–45), allowing the survival of activated APC and further amplifying the proinflammatory nature of the disease (43).

To provide evidence for this concept, we analyzed the numbers of monocytes/macrophages and DC and evaluated the expression of various activation markers on these cells in CFA-challenged IFN-γ KO mice and NK cell–targeted WT mice (Fig. 4). Absolute numbers of monocytes/macrophages were increased in CFA-challenged IFN-γ KO mice and in CFA-challenged WT mice treated with anti-Asialo (Fig. 4A). Analysis of activation markers revealed increased expression of CD69, CD80, and/or CD86 in all CFA-challenged mice and was most pronounced in mice depleted of NK cells using anti-Asialo. Of special interest is the expression of CCR2, indicative for inflammatory monocytes (46), which was significantly increased in the three NK-targeted groups as compared with nondepleted CFA-challenged WT mice as well as in challenged IFN-γ KO mice (Fig. 4B). When the expression of Rae-1 (an important ligand for NKG2D) was analyzed, more Rae-1–expressing monocytes/macrophages were found in NK-targeted mice, and values reached statistical significance in mice receiving neutralizing anti-NKG2D Abs (Fig. 4C). Also, CFA-challenged IFN-γ KO mice showed more Rae-1–expressing monocytes/macrophages (Fig. 4C), indicating their potential capacity to trigger NK cell–mediated cytotoxicity, as it is known that Rae-1 expression potentiates killing via the activating NK cell receptor NKG2D (47). However, expression of NKG2D was significantly lower on NK cells from CFA-challenged IFN-γ KO mice (Fig. 1D).

Regarding DC, the data are less conclusive, although a similar trend of more mature cells as evident from expression of CD80, CD86, MHC class II, and CCR2, can be seen in anti-Asialo–treated mice (Fig. 4D, 4E). NKG2D blockade did not alter DC phenotype compared with CFA-challenged WT mice. Expression of Rae-1 was not different among the different experimental groups (Fig. 4F).

Taken together, targeting of NK cells in CFA-induced WT mice results in phenotypic changes of monocytes/macrophages which are associated with a more activated state and which are reminiscent to those seen in CFA-challenged IFN-γ KO mice.

To investigate whether the persistence of activated monocytes/macrophages is a direct result of the impaired NK cell function, we evaluated the cytotoxic potential of NK cells against autologous activated immune cells using a CD107a degranulation assay (Fig. 5) (48). In a first set of experiments, NK cells from CFA-challenged WT and IFN-γ KO mice were cocultured with CD4⁺ T cells, DC, and monocytes/macrophages from either CFA-challenged WT or IFN-γ KO mice. YAC-1 tumor target cells were included as positive controls for NK cell degranulation against tumor target cells. NK cells isolated from spleens of CFA-challenged IFN-γ KO mice showed a clear-cut lower expression of CD107a compared with CFA-injected WT mice after coculture with the tumor target cells YAC-1 (Fig. 5A, right). These data are in line with their defective killing of Cr⁵¹-labeled YAC-1 cells (Fig. 1F). Monocytes, and especially those from CFA-challenged IFN-γ KO mice, were the most important triggers for degranulation of NK cells. Furthermore, a significantly lower CD107a expression by NK cells of IFN-γ KO mice was seen after coculture with all tested autologous target cells (with the exception of DC from WT mice) (Fig. 5A).

To verify whether NKG2D is required for the cytolytic activity against autologous cells, the experiment was repeated and WT mice that received anti-NKG2D Abs were included. To obtain sufficient numbers of the different purified cell populations, pools of three to five mice were used in this experiment (Fig. 5B). The data confirmed that NK cell cytolytic activity was most pronounced against monocytes as target cells and was generally lower in NK cells from CFA-challenged IFN-γ KO mice as compared with their WT counterparts. Furthermore, in comparison with CFA-challenged WT mice, NK cells from anti-NKG2D–treated mice showed a reduced CD107a degranulation toward all tested autologous target cells. Remarkably, this was not observed against YAC-1 tumor target cells. This experiment further highlights that NKG2D is an important receptor in the regulatory activity of NK cells in a systemic disease as sJIA, most likely by killing activated immune cells.

The role of NK cells in the pathogenesis of sJIA remains a matter of debate. There are reports describing decreased numbers of NK cells (16, 20, 49–51), an aberrant NK cell phenotype (13, 52), or defective NK cell function (13, 16–18, 20, 50) in sJIA patients. In contrast, other reports state normal NK cell numbers (5, 13, 16, 53) and cytotoxic function (13). Also, it remains unclear whether the described NK cell dysfunction is rather a cause or a consequence of the altered immune pathways in sJIA. Discrepancies may be secondary to the heterogeneity of the condition or to differences in disease activity and therapy at the time of evaluation. Analysis of the role of NK cells in disease onset remains a challenge but can be addressed by the use of animal models. We recently developed a mouse model of sJIA in which CFA challenge of IFN-γ KO mice induces sJIA-like disease. In contrast to IFN-γ KO mice, WT mice only develop a mild transient inflammation (21).

In our study, we confirmed the cytotoxic NK cell defects in NI IFN-γ KO mice as reported by Dalton et al. (30). They described the development and phenotype of IFN-γ KO mice and reported that NK cell cytolytic activity in NI mice was restored upon treatment of mice with poly(I:C), from which it can be concluded that the NK cell defect in the absence of IFN-γ is not intrinsic (30). In our mouse model, CFA challenge did not restore the cytolytic NK cell defect. In addition, the expression of NK cell activating receptors, and of granzyme B and perforin, were all significantly lower in CFA-challenged IFN-γ KO mice as compared with WT counterparts. To investigate if the NK cell deficits in IFN-γ KO mice explain their increased susceptibility to the development of sJIA-like disease, NK cells in WT mice were depleted before and at the time of CFA challenge. In literature, NK cells are commonly depleted by means of Abs against NK1.1, Asialo-GM1, or IL-2Rβ. IL-2Rβ and Asialo-GM1 are constitutively expressed by NK cells, but also to some extent by additional cells of the immune system (35, 36), whereas NK1.1 is NK cell specific (54). Unfortunately, NK1.1-depleting Ab could not been used in our study because the protein is not expressed by mice of the BALB/c origin (34), and the sJIA-like mouse model cannot be induced in mice with a C57BL/6 background (21). Depleting NK cells by means of anti–IL-2Rβ and anti-Asialo Abs or blockade of NKG2D resulted in sJIA-like symptoms in CFA-challenged WT mice. Mice presented with splenomegaly, lymphadenopathy, granulocytosis, and thrombocytosis (the latter was not found in anti-NKG2D–treated mice). In addition, we detected high numbers of immature neutrophils in anti-Asialo–treated mice and increased IL-6 levels in anti–IL-2Rβ–treated mice. The two NK cell–depleting Abs did not lead to identical systemic manifestations in both mice, which might be explained by a difference in the depletion characteristics of the Abs (35, 36), affecting the number of CD8⁺ T cells, CD4⁺ T cells, and monocytes/macrophages besides depleting NK cells (Supplemental Fig. 2). Because monocytes/macrophages were highly activated in CFA-challenged IFN-γ KO mice, we would expect that their depletion by anti–IL-2Rβ and anti-Asialo would rather inhibit the overall inflammatory status. Depletion of CD4⁺ and CD8⁺ T cells had no effect on the disease course in CFA-challenged WT mice (Supplemental Fig. 4). Therefore, we conclude that the effects of anti–IL-2Rβ and anti-Asialo are predominantly due to depletion of NK cells. Considering the data obtained in CFA-challenged IFN-γ KO mice and in NK cell–depleted or NKG2D-blocked WT mice, we conclude that mice with defects in NK cells are more prone to develop a sJIA-like illness upon CFA challenge. In literature, there are contradictory reports on the role of NK cells or NKG2D in inflammatory disease models. For example, in collagen-induced arthritis, a model for rheumatoid arthritis that relies on the use of CFA, NK cells, and NKG2D were found to promote autoimmune joint inflammation (22, 55). NKG2D also contributed to the severity of experimental autoimmune encephalomyelitis (EAE), a mouse model for multiple sclerosis (MS), as was evident from the use of NKG2D-deficient mice (56). However, in the same EAE mouse model, a regulatory role of NK cells was described by using NK cell–depleting Abs (anti-NK1.1) (57). Galazka et al. described an NK cell (and NKG2D)–dependent way to induce EAE tolerance by administration of heat shock protein 70-peptides complexes. Further support for the regulatory role of NK cells in this autoimmune disease was shown by data obtained in MS patients (58). In line with our data presented in this study, it is worth to elaborate on the study of Bielekova et al. In an elegant way, these authors provided evidence for direct killing of autologous activated CD4⁺ T cells by NK cells from MS patients treated with daclizumab (59).

The fact that no full-blown sJIA disease was observed in NK cell–depleted CFA-challenged WT mice might indicate that NK cells are not the only regulatory factors involved in disease development (21, 24). Alternatively, IFN-γ deficiency may be required to develop a more complete sJIA-like disease in this mouse model, as other pathways may be affected by the absence of IFN-γ, including macrophage activation and induction of T cell responses (60). Nevertheless, eliciting symptoms in IFN-γ–competent mice might better reflect the situation of sJIA patients, because no mutations in the IFN-γ gene have been documented in sJIA patients and the exact role of IFN-γ in sJIA remains to be defined. Of note, IFN-γ levels in plasma of sJIA patients are not in line with the high plasma levels of IL-18, a known IFN-γ–stimulating cytokine, resulting in a decreased IFN-γ to IL-18 ratio in sJIA patients (13, 61). In addition, IL-18–induced IFN-γ production by NK cells was found to be defective in sJIA patients (13, 20).

In line with these data, NK cells are an important source of IFN-γ (62) which was also confirmed in our sJIA-like mouse model (A. Avau and P. Matthys, unpublished observations). Therefore, the phenotype of CFA-challenged WT mice after depletion of their NK cells may be partly due to reduced IFN-γ production by NK cells.

In addition to impaired NK cell cytotoxicity and cytokine production, IFN-γ KO mice present with other intrinsic defects. A reduced expression of MHC class I is seen in IFN-γ–deficient mice (63), which renders cells susceptible for NK cell–mediated killing via missing-self recognition (64). Also, recent studies have shown the induction of NK cell hyporesponsiveness and NK cell tolerance to missing-self in the context of MHC class I downregulation in WT mice (65). Indeed, in our study, the reduced levels of MHC class I in IFN-γ KO mice did not result in increased killing of activated cells by NK cells, as indicated by the decreased CD107a degranulation of NK cells against activated autologous immune cells in vitro (Fig. 5) and indirectly by the increased numbers of activated monocytes and macrophages in vivo in CFA-challenged IFN-γ KO mice (Fig. 4). These data suggest that NK cell cytotoxicity of CFA-challenged IFN-γ KO mice is not only defective in killing MHC class I–deficient tumor target cells but potentially also of autologous activated immune cells in vivo.

Monocytes/macrophages and DC have been demonstrated to be involved in the pathogenesis of sJIA and the development of MAS (43, 66–68). In our sJIA mouse model, CFA challenge of IFN-γ KO mice resulted in activation and increased numbers of monocytes/macrophages and to some extent of DC, indicating activation pathways independent from IFN-γ. CCR2-expressing monocytes were highly increased in NK cell-targeted CFA-challenged WT and IFN-γ KO mice, indicating an expansion of these cells in the absence of regulating NK cells. CCR2⁺ monocytes are described as a subset of inflammatory monocytes that are actively recruited to inflamed tissue (46). Our data are in line with those of Sepulveda et al. (69) who described, in a mouse model of primary HLH, an immunoregulatory role of NK cells by controlling tissue infiltration of macrophages and by limiting hyperactivation of T cells. Monocytes/macrophages from CFA-challenged IFN-γ KO mice and CFA-challenged WT mice with NKG2D blockade also displayed high expression of NKG2D ligand Rae-1 (Fig. 4), which may explain the more ready recognition and activation by NK cells from WT mice but not of their autologous NK cells (Fig. 5). This observation is in line with the significant lower expression of NKG2D in CFA-challenged IFN-γ KO mice and their diminished cytotoxic activity toward Rae-expressing tumor target cells (Fig. 1). Indeed, several research groups have reported the autologous killing of monocytes and macrophages by NK cells via the NKG2D pathway (40, 70–72). A disruption of this pathway, may lead to diminished control of inflammatory monocytes and macrophages. With respect to the analysis of DC, changes in their numbers and phenotype were comparable to those of monocytes but, overall, less pronounced and more restricted to anti-Asialo–treated mice. Our in vitro data show that DC are weak triggers of NK cell degranulation (as compared with monocytes) and are possibly also less targeted in vivo.

In conclusion, CFA-challenged IFN-γ KO mice presented with an altered NK cell phenotype and defective cytotoxic capacity. Disruption of NK cells in CFA-challenged WT mice increased the chronicity and severity of inflammation and mice developed sJIA-like manifestations, indicating a regulatory role for NK cells in the development of sJIA. Autologous monocytes were identified as major triggers for NK cell degranulation and analysis of their phenotype in NK cell-targeted mice identified a subset of activated and inflammatory (CCR2⁺) monocytes as target, with NKG2D as key receptor. From our mouse model, we conclude that intact NK cell function is required to prevent persistent activation of monocytes/macrophages in response to stimulation with CFA.

We thank Dr. Nele Berghmans for help with the experiments.

The authors have no financial conflicts of interest.