Ncf1-Associated Reduced Oxidative Burst Promotes IL-33R⁺ T Cell-Mediated Adjuvant-Free Arthritis in Mice¹²

A functional NADPH oxidase complex is essential for the response against invading pathogens, and therefore, an impaired reactive oxygen species (ROS)⁶ production results in high susceptibility to bacterial infections and can cause chronic granulomatous disease (CGD) (1). CGD patients have a higher frequency of autoimmune diseases and there are studies describing juvenile rheumatoid arthritis in CGD patients or an association between rheumatoid arthritis and a polymorphism in neutrophil cytosolic factor (Ncf)4 (p40^phox), another important cytosolic regulatory subunit of the NADPH oxidase complex (2, 3, 4). The involvement of a bacterial component in the induction of arthritis has been discussed for many years, and there are a number of bacteria that have the capacity to induce arthritis in animal models, such as Borrelia burgdorferi causing Lyme arthritis or Staphylococcus aureus inducing septic arthritis (5, 6, 7). In adjuvant-induced arthritis models, it is either a mix between bacterial fragments and a mineral oil or the oil only that can cause autoimmunity. The adjuvant itself is inducing an immune response, which is not tissue specific but rather causing general inflammatory infiltrates. Depending on the nature of the used adjuvant this immune response is biased toward a certain Th pathway, such as Th1, Th17, or Th2. Having this knowledge in mind, we were aware of the difficult situation studying autoimmunity phenotypes in a ROS-deficient mouse and using complete Freund’s adjuvant containing mycobacteria, as in the classical collagen-induced arthritis (CIA) model.

CIA is the standard model for RA but is likely to be dependent on several different pathways as it is induced by immunization with collagen type II (CII) emulsified in CFA. The use of an adjuvant such as CFA for the induction of arthritis is believed to be necessary to break the immune tolerance and allow a pathogenic autoreactive T cell response. CFA contains mycobacteria, which activates APC via pattern-recognition receptors that direct T cells toward a Th1 or Th17 pathway characterized by the production of IFN-γ, TNF-α, and IL-17 (8, 9, 10). However, this guidance of the T cell response by an adjuvant is also restricting our view on arthritis-regulating pathways. To circumvent these limitations, we have earlier established an arthritis model induced by injection of CII without any adjuvant. For this, the DBA.Vb12 transgenic (tg) mouse strain was used, which expresses a TCR specific for the immunodominant galactosylated CII epitope. In this adjuvant-free model, arthritis develops in a chronic progressive manner and is characterized by an infiltration of eosinophils and the dependency on IL-5 (11).

We have earlier demonstrated the importance of a functional NADPH oxidase complex and ROS production for the resistance to arthritis by the identification of Ncf1 as an arthritis susceptibility gene in rats and mice (12, 13). The Ncf1 protein (also denoted p47^phox) is an essential organizing subunit of the NADPH oxidase complex in phagocytic cells. The NADPH oxidase complex is a multicomponent electron carrier that is responsible for the reduction of oxygen, resulting in the production of ROS. These ROS molecules are of major importance in the inflammatory defense against invading pathogens, but the positional cloning of a natural-selected Ncf1 polymorphism revealed also a major function of this protein in regulating the immune response. Lack of radicals promotes inflammation because mice and rats that have a low oxidative burst, due to a dysfunctional Ncf1, are more susceptible to arthritis, an effect mediated via CD68⁺ cells, presumably monocytes/macrophages during the process of Ag presentation and by that affecting the level of T cell activation (12, 13, 14). Thus, the Ncf1 mutation is affecting a series of T cell-dependent arthritis models. It has proven effects in the mouse in CIA and occurs spontaneously postpartum (13). In rats, the mutation affects susceptibility to pristane-induced arthritis and CIA (15). Furthermore, it is regulating autoimmune encephalomyelitis and neuritis in rats (16, 17, 18) and mice (13) as well, a disease that is known to be mainly dependent on T cells. Instead, it has been shown that arthritis induced after transfer of sera from the K/BxN mouse is unaffected by deficiency in gp91, another component of the NADPH oxidase complex, supporting the idea of ROS regulating T cells because this model is not T cell dependent (19).

It has earlier been shown that mice lacking ROS production have decreased capacity to clear bacterial infections (20). This effect could also be influencing arthritis development in our Ncf1-mutated mice. To investigate the influence of oxidation in a clean arthritis model without any immunostimulatory bacterial adjuvant that could influence the arthritis susceptibility, the Ncf1 mutation was bred to the DBA.Vβ12tg background.

Furthermore, we have recently shown that Ncf1-regulated oxidation is particularly critical for the induction and maintenance of T cell tolerance toward tissue-specific CII in classical CIA (21). This was shown in a unique model where tolerance toward arthritis was induced by the cartilage-restricted expression of transgenic rat CII, which differs from mouse collagen by a single amino acid at position 266 (aspartic acid vs glutamic acid), making the immunodominant peptide identical to the rat and human CII (mutated mouse collagen (MMC)). The expression of MMC leads to an improved presentation of the peptide by APC and results in a natural T cell tolerance and protection to arthritis induced with rat CII (22). We have shown that reduced oxidative burst due to the addition of the Ncf1 mutation into the MMC mouse breaks tolerance in the classical CIA model (21). The introduction of the Ncf1 mutation onto the DBA.Vβ12 background and the combination with the MMC transgene provide the unique possibility to study T cell tolerance mechanisms in an adjuvant-free arthritis model and to investigate the role of ROS in this process. In this study, we show that an autoreactive Th2-associated pathway leads to more severe arthritis induced with CII without any adjuvant in mice that carry a mutation in Ncf1 and lack ROS.

Mice were bred and used for experiments in the animal facility of Medical Inflammation Research. DBA/1J mice originate from The Jackson Laboratory. The Vb12 tg mouse was generated by Dr. L. Mori (University Hospital Basel, Basel, Switzerland) on an SWR background (23) and has been backcrossed to DBA/1J for more than 15N generations, establishing the DBA.Vb12tg strain. The Ncf1 mutation (Ncf1*) was bred in from BQ.Ncf1* mice that have been described earlier (13) as well as the MMC transgene (22). The MMC mouse carries a mutated mouse CII gene, thereby expressing the rat CII_260–270 epitope in a cartilage-restricted manner, which was subsequently backcrossed to B10.Q (13N). The B10.Q.Ncf1*.MMC mice and the DBA.Vb12tg mice were intercrossed to generate the different variants used in these experiments: DBA.Vb12.Ncf1* and DBA.Vb12.Ncf1*.MMC littermates with corresponding controls DBA.Vb12.Ncf1^wt and DBA.Vb12.Ncf1^wt.MMC mice that have been backcrossed to the DBA.Vb12tg strain for 3–5N generations. The mice were sex and age-matched at the start of the experiment. All mice were kept and bred in a climate-controlled environment with a 12-h light/dark cycle, housed in polystyrene cages containing wood shavings, and fed standard rodent chow and water ad libitum. All experiments were approved by a local (Malmö/Lund, Sweden) ethical committee (license M70-04 and M107-07).

CIA was induced by an i.p. injection of 100 μg of pepsin-digested rat CII diluted in PBS in a total volume of 200 μl/mouse. The CII was purified from the Swarm rat chondrosarcoma as described previously (24, 25, 26). In the experiment of Fig. 1 and the MMC experiments, mice received a booster injection of 50 μg of CII in PBS i.p. Blood was collected in tubes containing heparin at days 5 and 20 for the blocking experiment and at days 14 and 77 for the MMC experiment. Plasma was taken after 10 min centrifugation at 4000 rpm and stored at −20°C until assayed. Arthritis development was monitored with a macroscopic scoring system of the four limbs ranging from 0 to 15 (1 point for midfoot digit or knuckle, 5 points for a swollen ankle). The scores of the four paws were added, yielding a maximum total score of 60 for each mouse. Other disease parameters, including incidence, mean cumulative score, and mean day of onset, were also evaluated.

IL-5 was blocked in vivo by injection of the protein G-purified neutralizing rat mAb (Ab) TRFK-5 as follows: days −1, 13, and 27 before/after CII immunization 50 μg of TRFK-5 plus 50 μg of normal rat serum (NRS)-IgG (control Ab). The control mice received the equivalent amount of control Ab only. Abs were injected i.p.

For the in vivo anti-IL-33R experiment, purified 3E10 Ab was used (27). DBA.Vb12.Ncf1* mice were injected with 100 μg of 3E10 plus 100 μg of anti-rat κ (60 min later; BD Biosciences) on days −1, 13, and 15 before/after CII immunization. The control mice of the same genotype received the equivalent amount of NRS-IgG (control Ab) plus 100 μg of anti-rat κ (60 min later; BD Biosciences) at the same time points. Abs were injected i.p.

For determination of relative frequency of eosinophils, blood was hemolyzed with ammonium chloride (0.84%, pH 7.4), washed with PBS, and stained with anti-CCR3-PE (83101; R&D Systems) and anti-CD4-APC (L3T4) for 20 min at room temperature (RT). Cells were washed again with PBS and acquired on a FACSort (BD Biosciences). CCR3⁺ cells were calculated as a percentage of CD4⁺ cells to determine the relative frequency of eosinophils.

For determination of relative frequencies of Th2 and B22a1⁺ T cells and their cytokine profile, splenocytes were restimulated with PMA/ionomycin (10 ng/ml/500 ng/ml) and in V-shaped 96-well plates for 6 h at 37°C. For the last 4 h of stimulation, 2 μg/ml brefeldin A was added to the cultures. After the 6-h stimulation, cells were washed, preincubated with anti-CD32/16 (2.4G2, homemade) for 10 min, and afterward stained for 15 min at RT with the following Abs: anti IL-33R-bio (3E10), clonotypic Ab B22a1-AL647 (P. Merky et al., manuscript in preparation), anti-CD4-PE-Cy5.5 (L3T4), anti-CD11b-FITC (M1/70), anti-CD45R-FITC (RA3-6B2), anti-MHC class II-FITC (7.16.17), and anti-CD49b-FITC (DX5), followed by washing and staining with streptavidin-QD655 for 10 min at RT. Then cells were washed, fixed, and permeabilized, according to the manufacturer’s protocol (BD Biosciences). For intracellular staining, the following Abs were used (20 min at RT): anti-IL-2-AF700 (JES6-5H4), IFN-γ-PE-Cy7 (GMX1.2) and anti-IL-17-PB (TC11-18H10.9), anti-IL-4-PE (BVD4-1D11), and anti-IL-5-PE (TRFK5). Data were acquired on a FACSAria (BD Biosciences) and analyzed using FloJo software (Tree Star).

The level of intracellular oxidative burst ex vivo was measured in blood from DBA.Vb12.Ncf1*.MMC mice and corresponding controls. Blood was collected in tubes containing heparin and was hemolyzed with ammonium chloride (0.84%, pH 7.4). Granulocytes were stained with allophycocyanin-conjugated anti-Gr-1 Ab (RB6-8C5; BD Biosciences) for 30 min (4°C). To determine the level of NADPH oxidase activity, we used a modified version of the oxidative burst activity flow cytometry assay described previously (28). Briefly, cells were resuspended in Dulbecco’s complete medium without FCS after staining and incubated for 10 min at 37°C with 3 μM dihydro-rhodamine 123 (Molecular Probes and Invitrogen Life Technologies), which, after oxidation by hydrogen peroxide (H₂O₂), peroxinitrite (ONOO⁻), and hydroxyl radicals (OH^Ÿ) to rhodamine 123, emits a bright fluorescent signal upon excitation by blue light. Stimulation with PMA (200 ng/ml; Sigma-Aldrich) for 20 min at 37°C was performed afterward. Cells were washed with PBS and then acquired on a FACSort (BD Biosciences). Cells were gated on cell type, R-123 fluorescence intensity was measured on FL-1, and results are expressed in relative fluorescence units.

Ab titers against rat CII in plasma were determined with ELISA in 96-well plates (Costar) coated overnight at 4°C with 50 μl/well of 10 μg/ml rat CII in 50 μl of PBS. All washes were performed with PBS (pH 7.4) containing 0.1% Tween 20. Plasma was diluted 1/500 in PBS and analyzed in duplicates. The amounts of bound IgG and IgM Abs were estimated after incubation with biotin-conjugated isotype-specific Abs (Southern Biotechnology Associates), followed by extravidin-peroxidase (Sigma-Aldrich) and developed with ABTS (Roche Diagnostics) as substrate, followed by detection in a Spectra Max at OD 405 nm (Molecular Devices).

Splenocytes were prepared from mice at day 6 after i.p. injection of CII in PBS and restimulated in vitro with 10 μg/ml galactosylated CII_259–273 peptide (with a β-d-galctopyranosyl-5-hydroxyl-l-lysine at position 264 that was synthesized as previously described (29) for determination of Ag-specific cytokine production. Supernatant of cultured cells were collected after 24 h (for IL-2) and 96 h. IL-2 levels were determined by coating microtiter plates with anti-IL-2 (clone Jes6-A12) in PBS overnight. After blocking with 1% BSA in PBS, supernatant was added to the plates. Biotinylated anti-IL-2 (Jes6–5H4) was used as secondary Ab. For detection of IFN-γ and IL-17, a similar protocol was used with the following set of Abs: anti-IFN-γ (AN18) and anti-IFN-γ-bio (R46-A2); and anti-IL-17 (TC11-18H10) and anti-IL-17-bio (TC11-8H4.1). For IL-5 detection, the OptEIA IL-5 ELISA set (BD Biosciences) was used. The ELISA results were quantified using the dissociation-enhanced lanthanide fluoroimmunoassay technique with europium-labeled streptavidin (Wallac and PerkinElmer Life and Analytical Sciences), according to the manufacturers’ recommendations. The plates were read on a fluorometer (VICTOR; PerkinElmer Life and Analytical Sciences).

Plasma concentration of COMP was determined by a competitive ELISA according to an earlier described method (30).

Quantitative data are expressed as mean ± SEM. Statistical analyses were performed using Mann-Whitney U test for all calculations, except frequency of arthritis that was analyzed using χ² test. Significance values are given for the difference between Ncf1^wt and Ncf1* mice of each MMC genotype, if nothing else is stated.

To investigate the influence of oxidation in an arthritis model without influence of immunostimulatory adjuvant and to exclude a possible effect of bacterial components, we have established an adjuvant-free arthritis model in mice. Since this model requires the highly arthritis-susceptible DBA/1 background and a tg TCR Vβ12 chain specific for CII, the Ncf1* was bred into the DBA.Vβ12tg mouse. In the resulting DBA.Vβ12.Ncf1* strain, it was possible to induce arthritis after a single i.p. injection of CII without any additional adjuvant. DBA.Vβ12tg mice with functional Ncf1 (Ncf1^wt) developed mild arthritis with an onset at day 17. Ncf1-mutated DBA.Vβ12tg mice, with a nonfunctional oxidative burst, developed a significantly more severe arthritis with earlier onset and increased incidence (Fig. 1 A).

To analyze the pathogenic pathway evoked in the dramatically enhanced arthritis in the Ncf1* mice, we addressed the role of eosinophils and IL-5, because this has earlier been shown to predominate in the adjuvant-free arthritis models (11). Injection of anti-IL-5 Ab (TRFK5) led to a significant reduction of the eosinophil population in blood tested on days 5 and 20 after disease induction (Fig. 1^,B). The treatment also had significant effects on arthritis development in groups that received neutralizing TRFK-5 Ab compared with controls. Arthritis was not only decreased in treated DBA.Vβ12 mice but also in treated mice with the Ncf1 mutation compared with the corresponding controls. By studying the effect of anti-IL-5 treatment in different phases of the disease, we could exclude the role of eosinophils acting as APC during priming. Rather, eosinophils are more likely to play a role as effector cells in adjuvant-free arthritis (see supplemental figure).⁷ Another evidence for this hypothesis is the fact that the treatment did not show an effect on the anti-CII Ab response (Fig. 1 C).

We have recently shown that the lack of ROS due to a mutation in Ncf1 mediates break of tolerance in the classical CIA model (21). To investigate the relevance of T cell tolerance toward CII in the adjuvant-free arthritis model and the impact of low ROS on the induction and maintenance of T cell tolerance, the MMC transgene, forcing the expression of the rat CII_260–270 epitope in a cartilage-restricted manner, was bred into the DBA.Vβ12.Ncf1* mouse. By that, we established a tool to study T cell tolerance toward CII in this unique adjuvant-free model.

First, it could be confirmed that the Ncf1 mutation lowers ROS on another genetic background and in the presence of the tg Vβ12 TCR and MMC (Fig. 2).

Next, arthritis was induced by i.p. injection of CII in PBS. These experiments revealed that T cell tolerance operates in adjuvant-free arthritis and that a functional oxidative burst is a prerequisite for the tolerance. The introduction of the Ncf1 mutation led to an early and severe development of arthritis comparable to that in littermates without the tolerizing MMC transgene and to an even earlier and more severe arthritis than in wild-type controls. Mice with functional Ncf1 and ROS production instead were significantly protected compared with littermate controls without MMC (Fig. 3,A and Table I). By investigating arthritis in more detail, we found that also the ongoing cartilage destruction measured by the level of free COMP in the chronic phase of arthritis was significantly elevated in mice with Ncf1 mutation (Fig. 3 B).

To investigate the effect on T cell-dependent autoimmunity, we analyzed the Ab response against CII. A decreased ROS production due to the presence of the Ncf1 mutation resulted in higher CII-reactive Ab levels measured in plasma on day 42 after immunization, in both MMC and non-MMC mice. By analyzing the Ab response in more detail, we found that the Ncf1 mutation elevated the levels of all measured IgG isotypes while only IgG1 reached significance, suggesting a possible polarization of the T cell response toward Th2 (Fig. 3 C).

Because the autoantibody response provided evidence for a possible regulation of T cell tolerance by the Ncf1 mutation in adjuvant-free arthritis, we wanted to investigate T cell phenotypes and responsiveness in more detail. Spleen lymphocytes from mice immunized 6 days earlier with CII in PBS i.p. were isolated and restimulated in vitro. Flow cytometry analysis showed that the number of CII-reactive T cells (B22a⁺) tended to be decreased in MMC mice, but the decrease was independent of the Ncf1 mutation (Fig. 4,A, upper left graph). Using intracellular flow cytometry, we investigated the number of IL-2-, IFN-γ-, and IL-17-producing CD4⁺ T cells. We observed a clear decrease in Th1 and Th17 populations in MMC mice, but the introduction of the Ncf1 mutation did not skew this cytokine pattern, and is therefore probably not the reason for the loss of tolerance towards arthritis (Fig. 4 A). Similar results were obtained when the level of IL-2, IFN-γ, and IL-17 were measured in supernatant of restimulated splenocytes by ELISA (data not shown).

Finally, because of the Ncf1-dependent increase of anti-CII IgG1 levels, we also wanted to investigate and compare the Th2 response in CII-tolerized mice with and without the Ncf1 mutation. Indeed, a significant increase in frequency of IL-4- and IL-5-producing CD4⁺ T cells was observed in MMC tg mice with the Ncf1 mutation, compared with MMC mice with a functional Ncf1. This particular cytokine profile is likely to be associated with T cells expressing IL-33R (31), a receptor for IL-33, which was observed to be expanded (Fig. 4,B). Analysis of the CII-specific CD4⁺ T cells coexpressing B22a and IL-33R or IL-4 and IL-5 revealed similar results, showing more CII-reactive CD4⁺ T cells with a Th2 phenotype in MMC tg mice with the mutation in Ncf1 (Fig. 4 C).

To investigate the role of IL-33R-expressing T cells, we induced arthritis in DBA.Vb12.Ncf1* mice with CII/PBS injection on day 0 and performed anti-IL-33R treatment at days −1, 13, and 15. Compared with controls that had been injected with a control Ab, the treated mice developed adjuvant-free arthritis with significant higher incidence (Fig. 5,A). In line with the possibility that the increase in arthritis incidence is due to an activation of the IL-33R⁺ T cells, we observed an increase of the number of IL-33R⁺ T cells and increased titers of IgG1 Abs to CII in the anti-IL-33R-treated group (Fig. 5, B and C).

The use of an arthritis model that can be induced by injection of CII protein only without any adjuvant enabled us to clarify that the autoimmune enhancing effect of a reduced oxidative burst is not dependent on adjuvants like mineral oil or bacterial components.

Importantly, the introduction of a mutation in the Ncf1 gene not only enhanced arthritis severity but also broke immune tolerance and enhanced an arthritogenic response associated with IL-5, eosinophils, and the activation of IL-33R-expressing T cells.

Taking these characteristics together, we suggest that the adjuvant-free arthritis model can be of importance in investigating new pathogenic pathways of arthritis. In this model, we have used a TCR tg mouse with a Vβ12 chain to increase the frequency of CII-reactive cells because the level of autoimmunity is depending on that and often the limiting factor in development of arthritis (23, 32). But as the tg T cells do still freely rearrange their endogenous TCRα-chain, only ∼5% of the T cell repertoire is recognizing CII (data not shown) and the conditions of T cell development and priming in this mouse are thus still relatively physiologic.

We have earlier observed that there was a correlation between arthritis development, IL-5 production, and eosinophil accumulation at the site of inflammation in the adjuvant-free arthritis model (11). In this study, we confirmed that finding by in vivo inhibition of IL-5 with neutralizing Abs that mediated a significant reduction on eosinophil numbers in the blood during disease progression and on the severity of arthritis. This was not only the case for Vb12tg mice but also for Vb12tg mice with the Ncf1 mutation and reduced ROS production, providing an interesting insight into the Ncf1-mediated pathomechanisms. A lower level of oxidation clearly increases arthritis susceptibility in adjuvant-free arthritis as it does in the classical CIA model, but this can only partially be explained by an effect on IL-5 and eosinophils. Because the reduction of eosinophil frequency is similar to that in treated mice with functional Ncf1, whereas the disease is still significantly higher in the Ncf1-deficient treated group, it is likely that oxidation is mediating arthritis also by other noneosinophil-dependent pathogenic mechanisms. This is supported by the fact that also in the untreated groups, arthritis is much more severe and frequent in mice with mutated Ncf1 than in Ncf1-sufficient controls, even if the eosinophil numbers in the blood are not different (day 5) or even higher in the Ncf1 wild-type group (day 20).

ROS are involved in multiple cellular processes, providing a number of possibilities of how a mutation in Ncf1 can affect arthritis susceptibility. We have suggested that the Ncf1 mutation affects ROS production in APC and, in turn, the APC–T cell interaction and the resulting T cell activation (12). Even if the exact mechanism is still unknown, this hypothesis is supported by the observation that a decreased ROS production or T cell membrane reduction enables arthritis transfer with CD4⁺ T cells (12, 15, 33). There are several possibilities of how ROS can influence the arthritogenicity of T cells. T cells are very much dependent on the oxidation status of their extracellular environment. Alterations of the membrane redox situation of T cells mediated by ROS determine T cell responsiveness and activation such as, for example, through the oxidation-sensitive T cell surface molecule linker for activation of T cells that can lower the TCR signaling after oxidation (34).

The arthritis induced after injection of CII without adjuvant seems to be eosinophil and IL-5 dependent, irrespective of whether the mice are Ncf1 mutated or not and irrespective of whether the mice are tolerant by MMC or not. Thus, the mice are affected by arthritis through a Th2-like pathway. It has been shown before that in conventional CIA the CII-specific CD4⁺ T cells found in the joint are mainly Tim-3⁺/IL-33R⁻, suggesting a Th1 phenotype (35). In this model, it is largely IL-33R⁺ CD4⁺ T cells that seem to mediate pathogenicity, which is clearly different from conventional, adjuvant-dependent CIA. That Th2-mediated arthritis could in certain cases be arthritogenic has been shown earlier. Svensson et al. (36) reported that IL-4-deficent mice develop a less acute arthritis if induced with s.c. immunization of CII in mineral oil only, i.e., without the mycobacterial component. In addition, it was shown that collagen Ab-induced arthritis is dependent on Th2 cytokines like IL-4 and IL-10 (37).

However, to explain the Ncf1-enhancing effect in this mouse model, we observed that the presence of the Ncf1 mutation in the MMC mouse background significantly increased the number of T cells expressing IL-33R. It has been described earlier that IL-33R is a surface marker for IL-4-independent Th2 cells and an important signaling molecule during Th2 activation as well as for the efficient conduction of their effector functions (27, 38, 39, 40, 41). Importantly IL-33R⁺ T cells produce IL-5, which is driving eosinophil expansion. It is a receptor for IL-33, and injection of this cytokine has been shown to enhance not only CIA but also an allergic bronchitis model, possibly through mast cell activation (41, 42).

Changes in cytokine production of T cells, influencing their arthritogenicity, have earlier been reported to be affected by ROS (14, 43). Several studies demonstrated that redox levels control the phenotypic response or polarization of T cells. Some groups reported that low intracellular glutathione levels in murine APC or oxidative stress in human PBMC, respectively, promote the differentiation toward a Th2 phenotype (44, 45). However, there are also reports suggesting that augmenting the intracellular thiol pool in mouse splenocytes with N-acetylcysteine blocked the induction of IFN-γ and instead promoted IL-4 production and Th2 differentiation (46). This is consistent with our results showing that a more reduced milieu in APC due to the Ncf1 mutation mediated lack of ROS induces a more pronounced Th2 environment. Nevertheless, it should be mentioned that we have also observed slight but not significant increases in Th1 and Th17 cytokine production and that after immunization with CII in CFA or IFA, a more pronounced Th1 and Th17 response is seen in the Ncf1-mutated mice (14). In a recent report by Liu et al. (47), it was demonstrated that ROS-deficient mice, with genetic ablation of Ncf1, develop proliferative macrophage lesions in an infection-free environment characterized by higher levels of Th1-, Th2-, and Th17-associated cytokines compared with wild-type mice. Furthermore, as we have already demonstrated earlier, their results implicate a role for Ncf1 in control of inflammation through the regulation of ROS response by macrophages (14, 47).

Liu et al. (47) also found that Ym1 expression by Ncf1^−/− macrophages was changed. Ym1/Ym2 protein, a chitinase-like lectin, is known to be induced by activated macrophages during Th2-biased immune response and implicated in the recruitment of eosinophils as a main chemotactic factor (48, 49). This is in line with our results showing that reduced eosinophil numbers correlate with less arthritis in Ncf1-mutated mice and that an activation of IL-33R⁺ cells elevates arthritis incidence and anti-CII IgG1 Ab response, suggesting a Th2 polarization.

Interestingly, in this study, the Ncf1- and ROS-mediated differences in cytokine production were primarily evident in the situation where the tolerizing MMC transgene was present, supporting the hypothesis that ROS molecules produced by APC affect the T cell activation threshold during T cell MHC-Ag complex interaction, either during central T cell tolerance induction in the thymus or in the periphery.

Taken together, our data underline the importance of a functional ROS production in maintaining resistance to autoimmunity, especially in the process of T cell tolerance induction and regulation of T cell response against tissue-specific CII in a unique clean arthritis model excluding the influence of contaminating adjuvant effects. Thus, our targeted investigations defining the role of NADPH oxidase-derived ROS may disclose novel molecular pathways for developing treatments for inflammatory diseases.

We thank the technicians at Medical Inflammation Research, Kristina and Carlos Palestro, Isabell Bohlin, and Johanna Ekelund, for taking excellent care of the animals.

The authors have no financial conflict of interest.