Bidirectional Effect of IFN-γ on Th17 Responses in Experimental Autoimmune Uveitis

Pro- and ant-inflammatory effects of IFN-γ have been repeatedly found in various immune responses, including cancer and autoimmune diseases. In a previous study we showed that the timing of treatment determines the effect of adenosine-based immunotherapy. In this study we examined the role of IFN-γ in pathogenic Th17 responses in experimental autoimmune uveitis (EAU). We observed that IFN-γ has a bidirectional effect on Th17 responses, when tested both in vitro and in vivo. Anti-IFN-γ antibody inhibits Th17 responses when applied in the initial phase of the immune response; however, it enhances the Th17 response if administered in a later phase of EAU. In the current study we showed that IFN-γ is an important immunomodulatory molecule in γδ T cell activation, as well as in Th17 responses. These results should advance our understanding of the regulation of Th17 responses in autoimmunity.

IFN-g has either pro-or anti-inflammatory effects in various immune responses, including cancer and autoimmunity (18)(19)(20)(21)(22). The earlier finding that IFN-g inhibits Th17 responses (23,24) suggested that neutralization of IFN-g would elicit stronger Th17 responses (25,26). Because we previously found that adenosinebased immunotherapies are strictly "timing dependent" (27,28), we wished to investigate whether immunomodulation by treatments other than adenosine or adenosine receptors also might depend on the timing of treatment. In our previous study, we found that treatment of mice with induced experimental autoimmune uveitis (EAU) with adenosine deaminase (ADA), an enzyme that degrades adenosine, inhibited Th17 pathogenic T cell responses and suppressed EAU (27). The inhibitory effect of ADA was restricted to the active stage of disease; but ADA was ineffective if administered during the quiescent disease stage (27). Likewise, treatment of EAU-induced mice with an antagonist specific for adenosine receptor A2AR only inhibited EAU if given during the active phase of intraocular inflammation (28). We investigated the effect of adding anti-IFN-g antibody to in vitro responding T cells and EAU-induced mice at different time points to determine whether IFN-g has a regulatory effect on Th17 responses, and whether a similar timing effect might be identified. Our results demonstrated that anti-IFN-g antibody inhibited Th17 responses if provided during the initial phase of T cell activation (i.e., the early phase of EAU induction). However, the effect was reversed once T cell activation was initiated in vitro or when anti-IFN-g antibody was administered to EAU-induced mice in a later phase of EAU induction.
Our previous studies showed that activation of gd T cells closely correlated with augmented Th17 responses (29)(30)(31)(32). Therefore, we also examined the effect of anti-IFN-g treatment on gd T cell activation. Anti-IFN-g treatment during the early phase of EAU induction inhibited gd T cell activation, whereas treatment during a later phase of EAU induction enhanced gd T cell activation similar to the Th17 responses. Thus, in the current study we showed a "timing effect" for IFN-g similar to that of adenosine-based immunotherapy in EAU (27,33).

Animals and Reagents
Female C57BL/6 (B6), TCR-d -/-, and IFN-g -/mice were purchased from Jackson Laboratory (Bar Harbor, ME, USA); 12-to 16-week-old mice were used in all studies. All mice were housed and maintained in the animal facilities of the University of California Los Angeles. All protocols in this study were approved by the Committee on the Ethics of Animal Experiments of University of California, Los Angeles (IACUC permit number: ARC#2014-029-03A), in compliance with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health.

EAU Induction and Anti-IFN-g Treatment
EAU was induced in B6 mice by s.c. injection of 200 ml of emulsion containing 200 mg of human interphotoreceptor retinoid-binding protein (IRBP) 1-20 (Sigma-Aldrich, St. Louis, MO, USA) in CFA (Difco, Detroit, MI, USA) at six spots at the tail base and on the flank, and by i.p. injection with 300 ng of pertussis toxin. Mice were then examined three times a week until the end of the experiment (d 30 post-immunization).
For in vivo administration of IFN-g, immunized B6 mice were randomly divided into three groups (n=6), one of which received an i.p. injection of anti-IFN-g (100 mg/mouse) at d 0 (day of immunization and the second at d 8 post-immunization. The mice in the control group received PBS. At d 13 postimmunization (the time at which the highest T cell response is seen), responder T cells were purified from the spleen and draining lymph nodes stimulated in vitro with the immunizing peptide and APCs (irradiated spleen cells) under culture conditions that favor Th17 or Th1 autoreactive T cell expansion (medium containing 10 ng/ml of, respectively, IL-23 or IL-12) (24, 25) A schematic procedure of disease induction and examination of mice under investigation is shown in Scheme 1.

EAU Evaluation
The mice were examined three times a week for 30 d postimmunization. The clinical signs of EAU were evaluated using fundoscopic examination. Fundoscopic grading of disease was performed using the scoring system described previously (34).
At 30 d post-immunization, the mice were euthanized, and the eyes were collected for histological examination. For histology, whole eyes were collected at the end of the experiment and prepared for histopathological evaluation. The eyes were immersed for 1 h in 4% phosphate-buffered glutaraldehyde, then transferred to 10% phosphate-buffered formaldehyde until they were processed. Fixed and dehydrated tissues were embedded in methacrylate, and 5 mM sections were cut through the pupillary-optic nerve plane and stained with H&E. The eyes were fixed overnight at 40C in Davison's solution and then processed as paraffin-embedded blocks.

T Cell Preparation
Responder CD3 + T cells were purified from B6 mice immunized with the human IRBP 1-20 peptide (29,31,35). Nylon woolenriched splenic T cells from naive or immunized mice were incubated sequentially for 10 min at 4°C with FITC-conjugated anti-mouse gd TCR or ab TCR Abs and for 15 min at 4°C with anti-FITC Microbeads (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany). The cells were then separated into bound and non-bound fractions on an autoMACS ™ separator column (Miltenyi Biotec GmbH). To obtain a sufficient number of cells, we routinely pool the cells obtained from all six mice in the same group, before the T cells are further enriched using MACS column. The purity of the isolated cells, as determined by flow cytometric analysis using PE-conjugated Abs against ab or gd T cells, was >95%.

Assessment of Th1 and Th17 Polarized Responses
Responder CD3 + T cells (3 x 10 6 ) were co-cultured for 48 h with IRBP 1-20 (10 µg/ml) and with irradiated spleen cells (2 x 10 6 /well) as APCs in a 12-well plate under either Th17 polarized conditions (culture medium supplemented with 10 ng/ml of IL-23) or Th1 polarized conditions (culture medium supplemented with 10 ng/ ml of IL-12). The responder ab T cells were collected from IRBP 1-20 -immunized B6 mice, on d 13 post-immunization. Forty-eight hours after stimulation, IL-17 and IFN-g levels in the culture medium were measured using ELISA kits (R & D Systems) and the percentage of IFN-g + and IL-17 + T cells among the responder T cells was determined by intracellular staining after 5 d of culture, followed by FACS analysis, as described below (31,36).

Immunofluorescence Flow Cytometry for Surface and Cytoplasmic Antigens
In vivo primed T cells were stimulated with the immunizing antigen and APCs for 5 d. The T cells were then separated using Ficoll gradient centrifugation and stimulated in vitro for 4 h with 50 ng/ml of PMA, 1 mg/ml of ionomycin, and 1 mg/ml of brefeldin A (all from Sigma). Aliquots of cells (2 x 105 cells) were incubated for 30 min at 4°C with anti-ab or -gd TCR antibodies, then fixed, permeabilized overnight with Cytofix/ Cytoperm buffer (eBioscience, San Diego, CA, USA), and intracellularly stained with PE-conjugated anti-mouse IFN-g antibodies or FITC-labeled anti-mouse IL-17 antibodies. Data collection and analysis were performed on a FACScalibur flow cytometer using CellQuest software.

Cytokine Assays by ELISA
Cytokine (IL-17 and IFN-g) levels in the culture medium were measured using ELISA kits (R & D) following manufacturer's instructions.

Statistical Analysis
The results in the figures are representative of one experiment, which was repeated 3-5 times. The statistical significance of differences between groups in a single experiment was initially analyzed by 2-way Student's t-tests, and if statistical significance was detected, the Student-Newman-Keuls post-hoc test was subsequently used. A P value of less than 0.05 was considered a statistically significant difference and marked with one *; when P<0.01, two ** were used.

Anti-IFN-g Affects Th1 and Th17 Responses Differently In Vitro
To determine the effect of IFN-g on autoreactive antigen specific T cell responses in vitro, CD3 + T cells from the spleen and draining lymph nodes of B6 mice were isolated at the height of induced responses, 13 d after immunization with IRBP 1-20. They were stimulated with the immunizing antigen and irradiated splenic APCs under Th1-or Th17-polarized conditions, with/without anti-IFN-g (d 0). As demonstrated in Figures 1A, B, anti-IFN-g treated on day 0 (day of immunization) inhibited the response of Th17 cells to a much greater extent than it inhibited the response of Th1 cells; similarly, IL-17 production was decreased much more than production of IFN-g ( Figure 1E). The effect of anti-IFN-g was reversed when the antibody was added 8 d post-stimulation ( Figure 1C): with the Th17 cell and IL-17 responses enhanced. The effect on Th1 cells or IFN-g production was minimal.

Anti-IFN-g Affects Th1 and Th17 Responses Differently In Vivo
To determine whether anti-IFN-g antibody would have a similar effect in vivo, two groups of B6 mice (n=6) were immunized with IRBP 1-20 with or without an i.p. injection of anti-IFN-g (100mg/ mouse) on the day of immunization (d 0). Thirteen days postimmunization the mice were sacrificed and Th1 and Th17 SCHEME 1 | Schematic experimental procedure. responses were assessed. Figure 2 shows that anti-IFN-g antibody again significantly inhibited the Th17 response while inhibiting the Th1 response only slightly, similar to its effects seen in vitro.
The Effect of Anti-IFN-g on EAU Development and the Th17 Response Depends on Timing To determine whether the timing effect observed in vitro also applied in vivo, groups of B6 mice (n=6) were immunized with IRBP 1-20 /CFA. Because preliminary testing revealed that results differed when anti-IFN-g was injected between d 0 to 3 vs d 6 to 8 post immunization, two time points were selected for subsequent studies. Those mice injected on the day of immunization (d0) were designated as "early-treated", and those injected on day 8 after immunization were designated as "late-treated". As shown in Figures 3A, B, the number of IL-17 + cells among the responder T cells in early-treated mice was significantly decreased (from 9.5% in the control group to 3.3% in the treated group), whereas this number was significantly increased in late-treated mice (from 9.5% in the control group to 11.6% in the treated group) ( Figure 3A, upper panels). In contrast, numbers of IFN-g + T cells changed little between early-and late-treated mice ( Figure 3A, lower panels). Cytokine production ( Figure 3C) confirmed intracellular staining in that mice in the late group produced more IL-17, whereas those in the early group produced less IL-17. In vivo administration of anti-IFN-g had a stronger inhibitory effect on IFN-g production of responder T cells, when compared with in vitro tests (Figure 1), regardless of whether the antibody was injected early or late. The

IL-17 + T Cells Isolated From Late IFN-g Treated Recipients Are More Pathogenic
To determine whether pathogenic IRBP-specific IL-17 + T cell responses were enhanced in mice that received late treatment with anti-IFN-g we performed adoptive transfer tests. IL-17 + T cells from IRBP-immunized mice administered anti-IFN-g on d 8 were compared to PBS injected mice (i.e. the control) on d8. 2 x 10 6 IRBP-specific IL-17 + cells were adoptively transferred to naive B6 mice by i.p. injection after 2 d of in vitro stimulation with the immunizing antigen and splenic APCs, under Th17polarizing conditions. Adoptively transferred Th17 cells from late anti-IFN-g treated mice induced significantly more severe EAU in recipient mice when compared to controls ( Figure 3D). The total number of IL-17 + T cells was also significantly greater in anti-IFN-g treated mice (data not shown).

Similar Timing Effect Exists in gd T Cell Responses in Anti-IFN-g Treated Mice
We have previously shown that gd T cells regulate Th17 responses (29,31), whereby activation of gd T cells augmented Th17 reactivity (30)(31)(32). Therefore, we thought it important to examine the state of gd activation after anti-IFN-g treatment. In early anti-IFN-g treated mice, the frequency of gd T cells among total CD3 + T cells was significantly reduced (Figures 4A, B), suggesting that IFN-g is required for early-stage gd cell proliferation. Perhaps more importantly, the ratio between CD44 + and CD44gd T cells was greatly decreased (Figures 4C, D), indicating that neutralizing IFN-g at early stages disease had an inhibitory effect on gd activation in vivo. By contrast, the number of gd T cells were significantly increased among total CD3 + T cells in mice that received late anti-IFN-g treatment ( Figure 5A) compared to early anti-IFN-g treated or untreated B6 mice. Moreover, the number of IL-17 + among total gd T cells was significantly higher in mice that received late anti-IFN-g treatment ( Figure 5B).

gd T Cells Separated From Late Anti-IFN-g Treated Mice Enhance Th17 Activity In Vitro
To further investigate if the enhanced gd activation in mice with late anti-IFN-g treatment contributed to augmented Th17 responses, we compared the in vitro Th17 response of TCR-d -/responder T cells with or without the addition of gd T cells from early or late treated anti-IFN-g mice (30)(31)(32). Figure 5C shows that the Th17 response of the TCR-d -/responder T cells were significantly higher after adding gd T cells (5% of total CD3 + cells) isolated from late anti-IFN-g treated mice (right panel) compared to those without added gd T cells (left panel) or with added gd T cells isolated from early anti-IFN-g treated mice (middle panel).

Augmented Th17 Responses in IFN-g -/-Mice Are Contributed by Multiple Aberrant Immune Responses
Finally, we compared the immune response between IFN-gdeficient and wt-B6 mice, as a further means to better understand the role of IFN-g in pathogenic Th17 responses. Figure 6A shows that a greater number of responder T cells expressed IL-17 in IFN-g -/mice compared to B6 mice ( Figure 6A) when CD3 + T cells were stimulated with the immunizing antigen and APCs for 5 d; additionally, IFN-g -/-T cells produced significantly higher amounts of IL-17 ( Figure 6B). The number of gd T cells among the CD3 + splenic cells did not differ significantly between unimmunized IFN-g -/and B6 mice; however, the total number of gd T cells, as well as numbers of activated gd T cells (IL-17 + ), increased greatly among the IFN-g -/mice after immunization ( Figures 6C, D).
We also compared antigen-specific T cell responses between CD3 + T cells isolated from IFN-g -/and wt-B6 mice, respectively. The results of crisscross tests, in which the T cells derived from immunized IFN-g -/or B6 mice were stimulated either by splenic APCs of irradiated B6 mice or IFN-g -/mice ( Figure 6E) revealed that IFN-g -/-T cells produced significantly greater amounts of IL-17 when compared to B6 T cells. In addition, IFN-g -/splenic APCs had stronger stimulating activity compared to B6 splenic APCs, suggesting that APC function is also augmented in IFNg -/mice.
When Foxp3 + T cell responses of wt-B6 and IFN-g -/mice were compared, we found that the number of Foxp3 + T cells was significantly higher among CD3 + T cells in IFN-g -/mice ( Figure 7A). However, the EAU-inducing activity of IRBPspecific T cells isolated from IFN-g -/mice and from B6 mice was compared after a 2-d in vitro stimulation with the immunizing antigen and splenic APCs by adoptive transfer to naïve B6 mice. IRBP-specific T cells isolated from immunized IFN-g -/mice had significantly stronger EAU-inducing activity compared to those from immunized B6 mice ( Figures 7B, C). These results suggest that augmented Th17 responses in IFN-gdeficient mice are associated with multiple aberrant immune cell functions or that the increased number of Th17 cells in IFN-g -/exceeded the regulatory capacity of Tregs.

DISCUSSION
IFN-g production is a hallmark of T helper (Th)1 responses (1, 2) and exerts diverse effects on immune responsiveness. It strengthens innate immunity via induction of antimicrobial factors and degradative pathways in other immune cells, such as macrophages; it also augments the antigen-processing and antigen-presenting ability of APCs, stimulates antibody production by B cells, induces the expression of cytokines and chemokines required for the recruitment of myeloid cells to the site of inflammation, and increases the expression of TLRs, NOS, and phagocyte oxidase by macrophages (38). Previous studies demonstrated that the effect of IFN-g on immune responses could be either pro- (19,(39)(40)(41)(42) or anti-inflammatory (23,24,43,44). Elucidation of the bidirectional functions of IFN-g should increase our understanding of the regulated Th17 responses and their IFN-g mediated immunomodulation.
Early studies of IFN-g revealed it to be a key pathogenic molecule in human multiple sclerosis (MS) and in animal models, i.e., experimental autoimmune encephalitis (EAE) (39,40). Clinical studies showed that MS patients treated with IFN-g exhibited exacerbated symptoms (42), whereas MS patients treated with antibodies against IFN-g exhibited reduced clinical symptoms (19). However, a protective effect of IFN-g has also been demonstrated. Mice deficient in the IFN-g gene showed an increased incidence of EAE, with earlier disease onset and more severe symptoms (45,46). Injection of neutralizing antibodies to IFN-g exacerbated both actively and passively induced EAE (43,44). | Altered gd T cell responses in anti-IFN-g treated mice. gd T cell activation and expansion was inhibited in early anti-IFN-g treated mice. Freshly isolated CD3 + cells from naïve and immunized B6 mice, with or without anti-IFN-g administration (day 0), were stained with PE-anti-abTCR and FITC-anti-gdTCR antibodies before they were subjected to FACS analysis (A). They were also dually stained with PE-anti-mouse CD44 and FITC-anti-gdTCR antibodies (C). A summary of multiple assays is shown in (B, D). *P < 0.05, **P < 0.01. Injection of IFN-g to EAE-prone mice reduced the severity of disease symptoms and mortality (43,47). A similar protective effect of IFN-g was also found in other autoimmune diseases, including collagen-induced arthritis, EAU, autoimmune nephritis, and myocarditis (43,45,(48)(49)(50)(51)(52), underscoring the complex role of IFN-g in disease pathogenesis. Disease stage-specific effects of IFN-g were also observed. For example, administration of IFN-g to EAE mice during the inductive period led to disease exacerbation, while a similar treatment during the effector phase was protective (18)(19)(20)(21)(22)53). Although it has been well established that both Th1 and Th17 autoreactive T cells are pathogenic in various autoimmune diseases (24,36,54,55), including autoimmune uveitis (56), it remained largely undetermined whether therapeutic treatments directed at Th1 responses would also be effective in treating Th17 mediated disease. Our current results demonstrated that IFN-g is also an effective molecule modulating Th17 responses, although its effect is bidirectional.
The aims of our study were to determine the regulatory mechanisms of Th17 responses in autoimmune diseases. Based on our previous report that the protective effect of adenosinebased treatments is dependent on "timing" ─ ADA was protective when administered during the active phase of EAU but ineffective if administered prior to intraocular inflammation (27). Subsequently, we found that A2AR antagonist SCH58261 (SCH) effectively modulates aberrant Th17 responses in induced EAU. Likewise, timing of the treatment is important (28). We therefore questioned whether such a timing effect applied to A B C FIGURE 5 | gd T cells isolated from late anti-IFN-g treated mice have enhanced pro-Th17 activity.CD3 + responder T cells (1 5 x 10 6 /well) isolated from immunized TCR-d -/mice, with or without a prior anti-IFN-g administration were stimulated with the immunizing antigen and irradiated splenic APCs, under Th17 polarized condition. Total gd T cell numbers (A), as well as proportional number of IL-17 + versus total gd T cells (B) were compared. **P < 0.01. In vitro Th17 response of TCRd -/responder T cells were assessed with or without an addition of gd T cells isolated from early or late treated anti-IFN-g mice (C). The number of IL-17 + T cells among the gated CD4 + responder T cells was estimated after intracellular staining with anti-IL-17 antibody, 5 d after in vitro stimulation.  showing that an immunological manipulation will be more effective if a "timing" factor has been taken into consideration.
Comparing various immune responses of IFN-g -/and B6 mice showed that gd T cells are overly active in IFN-g -/mice, and that IRBP-specific T cells isolated from immunized IFN-g -/mice have stronger EAU-inducing activity. Moreover, splenic APCs separated from IFN-g -/mice have stronger T cell stimulating A B C FIGURE 7 | Multiple aberrant immune responses in IFN-g -/mice. Foxp3 + T cell expansion is increased in IFN-g -/mice. In a 24-well plate, the CD3 + responder T cells (1.5 x 10 6 /well) derived from B6 mice or IFN-g -/mice were cultured in medium containing a very low dose of IL-2 (1ng/ml) for 5 d, which preferentially favors Foxp3 + T cell expansion (37). The percentage of Foxp3 + T cells among abTCR + cells was determined by FACS analysis (A). Data are from one single experiment, which is representative of three independent experiments. IRBP-specific T cells isolated from immunized IFN-g -/mice have stronger pathogenic activity after adoptive transfer to naïve B6 recipients. IRBP-specific T cells were prepared from the responder T cells of IFN-g -/mice and B6 mice. After a 2-d in vitro stimulation with the immunizing antigen and splenic APCs. 2 x 10 6 cells were adoptively transferred to naïve B6 recipient mice via i.p. injection (B). Pathologic examination. H&E histologic sections from an eye in each group were obtained on day 30 post-immunization. Severe vitritis and chorioretinal folds occur in IFN-g -/mice compared to minimal vitreous inflammation and a normal retina in wt-B6 mice (C). activity, indicating that augmented Th17 responses seen in IFN-g -/mice are the result of altered immune modulators, leading to delayed onset of disease with later exacerbation. The enhancement or inhibition of EAU correlates with gd T cell activation. In previous studies we showed that a "timing effect" is a hallmark of adenosine-based immunotherapy. In the current study we showed that the immunomodulatory effect of IFN-g also involves both enhancement and inhibition of the Th17 response in a "timing effect". Continued efforts to elucidate the mechanisms underlying the influence of IFN-g on immune responsiveness eventually should enable us to achieve desired effects, and thus bring us closer to the therapeutic goal of IFN-gbased treatment of diseases.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The experimental protocols were approved by the Institutional Animal Care and Use Committee of the University of California Los Angeles (Protocol number: ARC#2014-029-03A).

AUTHOR CONTRIBUTIONS
DS, HK, and HS designed research. DS and HS performed the experiments and analyzed data. DS and HK wrote the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
This work was supported by NIH grants EY0022403 and EY018827 and grant from for Research to Prevent Blindness, NYC.