Superoxide Production by NADPH Oxidase Intensifies Macrophage Antiviral Responses during Diabetogenic Coxsackievirus Infection

This article is featured in In This Issue, p.1

Viral infections have long been suspected environmental instigators of type 1 diabetes (T1D) (1, 2). CMV, EBV, Rubella, Mumps, Rotavirus, retroviruses, and Coxsackieviruses have been associated with T1D (1, 3). Studies of Coxsackievirus infection in human cadaveric donors (4, 5) and the NOD mouse model of T1D have provided extensive knowledge of how viruses may trigger T1D (4–9). Specifically, these ssRNA viruses have a tropism for the pancreas, and all six serotypes (CB1-6) successfully infect islets (10, 11). Histological analysis of pancreatic islets from recent-onset T1D patients revealed the presence of a Coxsackievirus infection (4, 12), which in some cases, was β cell specific (5, 12). Prediabetic NOD mice infected with Coxsackievirus B3 (CB3) or B4 (CB4) displayed accelerated T1D progression (6, 11), which was partly due to the uptake of infected β cells by APCs and bystander activation of autoreactive T cells (8, 9). Furthermore, this acceleration required established insulitis (7, 13), suggesting that Coxsackievirus infections impact disease onset through exacerbating inflammation and breaking peripheral tolerance.

T1D is an organ-specific autoimmune disease consisting of chronic production of proinflammatory cytokines (TNF-α, IFN-γ, IL-1β), type I IFNs (IFN-α/β), and reactive oxygen species synthesis that drives autoimmune responses and β cell destruction (14). Classically activated M1 macrophages are a major source of reactive oxygen species and proinflammatory cytokines partly due to signal transducer and activator of transcription 1 (STAT1) activation (15). Importantly, macrophages are essential in the pathogenesis of T1D (16, 17). Within the NOD mouse, they display hyperactivation of the NF-κB signaling pathway (18), leading to aberrant production of inflammatory cytokines and chemokines to facilitate β cell cytotoxicity (19), but can also function as APCs to activate autoreactive T cells (20). Macrophages can sufficiently transfer the diabetogenicity of CB4 infection, as adoptive transfer of macrophages from CB4-infected NOD.scid mice was able to trigger diabetes onset in TCR-transgenic NOD.BDC-2.5 mice (9).

T1D is a relapsing and remitting disease exhibiting the hallmarks of chronic oxidative stress and inflammation that contribute to autoimmune dysregulation (21). Oxidative stress is intimately linked to inflammatory responses, and free radical production functions as a crucial third signal for efficient activation of T cells (22–25). Our laboratory has established the importance of oxidative stress in T1D, as the absence of NADPH oxidase (NOX)–derived superoxide, conferred by a mutation (Ncf1^m1J) on the essential p47^phox subunit of the NOX complex, delays T1D in NOD.Ncf1^m1J mice (16, 22). We have recently shown that loss of superoxide causes an inherent skewing of macrophages from an M1 to an M2 phenotype during spontaneous and adoptive transfer of T1D (26). Importantly, we also reported that NOX-deficient macrophages display diminished TLR3-dependent inflammatory responses following stimulation with the viral dsRNA mimic, poly(I:C) (27). Therefore, shifting macrophage differentiation away from an inflammatory M1 phenotype within the NOD mouse can significantly change disease outcome.

Because virus-induced T1D is a consequence of heightened innate immune antiviral responses, NOD proinflammatory M1 macrophage differentiation and immune responses to poly(I:C) are dependent on NOX-derived superoxide, and T1D-resistant NOD.Ncf1^m1J mice lack the ability to generate NOX-derived superoxide, we propose that the commonality of virus infections and induction of autoimmune destruction of pancreatic β cells is the generation of oxidative stress. However, the role for free radicals in antiviral responses in T1D remains unknown. Macrophages play critical roles in both T1D and in combating Coxsackievirus infection, and they express one of the highest levels of NOX2. Altogether, we hypothesized that NOX-derived superoxide production during Coxsackievirus infection would potentiate proinflammatory M1 macrophage antiviral responses and aid in triggering autoimmunity.

NOD/ShiLtJ and NOD.Ncf1^m1J mice were bred and housed under pathogen-free conditions at the Research Support Building animal facility at the University of Alabama at Birmingham. Female and male mice between 8 and 16 wk of age were used. Mice received standard chow and acidified water weekly except for NOD.Ncf1^m1J, which received 80 mg of trimethoprim/sulfamethoxazole (Hi-Tech Pharmacal) in drinking water. All animal studies were performed in accordance with the University of Alabama at Birmingham Institutional Animal Use and Care Committee in compliance with the laws of the United States of America.

NOD and NOD.Ncf1^m1J bone marrow–derived macrophages (BM-MΦ) were generated as previously described (27). CB3 (Woodruff variant) was graciously provided by Dr. M. Horwitz (University of British Columbia) and maintained as described previously described (6). To provide exogenous superoxide, culture media was supplemented with 1 mU/ml of xanthine oxidase (XO) (Sigma) as described previously (28). In vitro infections were performed at a multiplicity of infection (MOI) range of 10–50.

NOD and NOD.Ncf1^m1J mice were infected by i.p. injection of 100 PFU CB3 in HBSS, or HBSS alone as control. For diabetes incidence studies, female mice were infected or uninfected at 12 wk of age, and monitored several times weekly for diabetes by glucosuria (Diastix). Mice were considered diabetic following two consecutive blood glucose readings ≥300 mg/dl. To determine viral titers, whole pancreata were harvested, weighed, homogenized in HeLa growth media (DMEM supplemented with 10% heat-inactivated FBS, 4 mM l-glutamine, 1 mM sodium pyruvate, 100 U/ml penicillin-streptomycin), frozen and thawed three times, and used for plaque assays on HeLa cells (29).

For RNA analysis, BM-MΦ were harvested using TRIzol (Invitrogen) and pancreata were initially harvested with RNAlater (Ambion), then lysed in TRIzol using the TissueLyser II (Qiagen). mRNA was isolated by the RNeasy kit (Qiagen) and reverse transcribed into cDNA using Superscript III (Invitrogen). TaqMan gene expression was analyzed with the following primers (Applied Biosystems): Emr1 (Mm00802529_m1), Tlr3 (Mm00628112_m1), Ifih1 (Mm00459183), Ddx58 (Mm01216853_m1), Tnf (Mm00443258_m1), Ifnb1 (Mm00439552_s1), Isg15 (Mm01705338_s1), and Gapdh (Mm99999915_g1). Relative gene expression levels were calculated using the 2^{(−∆∆ cycle threshold)} method, normalized to Gapdh, and wild-type unstimulated levels set to 1. For Western blotting, 25 μg of whole cell lysates were separated, transferred to nitrocellulose membrane (Millipore), and probed with Abs against P-STAT1 (Y701), STAT1, melanoma differentiation–associated protein 5 (MDA5), retinoic acid–inducible protein I (RIG-I) (Cell Signaling), or β-actin (Sigma-Aldrich) as described (27), then visualized on an Odyssey CLx Imager with Image Studio v4.0 software using LI-COR secondary anti-rabbit or anti-mouse Abs conjugated to either IRDye 680RD or IRDye 800CW. TNF-α, CXCL10, CCL5 (R&D Systems), and IFN-α/IFN-β (PBL Assay Science) was quantified by ELISAs.

Pancreatic cells were isolated by collagenase digestion and macrophages were identified by surface staining with anti-F4/80 (clone BM8; eBioscience) and anti-CD11b (clone M1/70; BD Biosciences) (30). Intracellular staining was performed with anti–TNF-α (clone MP6-XT22; R&D Biosystems) and the appropriate isotype control (R&D Biosystems). Samples were collected (400,000 events) on an Attune NxT flow cytometer (Thermo Fisher Scientific) and analyzed with FlowJo version 10.0.7 Software (Tree Star) following the gating scheme (Supplemental Fig. 1).

NOD and NOD.Ncf1^m1J BM-MΦ were cultured on chamber slides (Lab-Tek) and infected with 10 MOI CB3. For immuno-spin trapping, macromolecular-centered free radicals were detected with 1 mmol/l 5,5-dimethyl-1-pyrroline-N-oxide (DMPO; Dojindo), then stained for DMPO adducts by immunofluorescence, as described (28). P-STAT1 (Y701) was detected with FITC-conjugated donkey anti-rabbit IgG (H+L) secondary Ab (1:500). Macrophages were stained with DAPI and anti-F4/80 surface staining. Images were obtained with an Olympus IX81 Inverted Microscope at a 40× objective and analyzed with cellSens Dimension imaging software version 1.12. To quantitate fluorescence intensity, three to six images were obtained for each treatment group, collected at the same exposure time, adjusted to the same intensity level for standardization, and measured using ImageJ Software (National Institutes of Health), as previously described (28).

Data were analyzed using GraphPad Prism Version 5.0 statistical software. The difference between mean values and SEM was assessed using the two-tailed Student t test, with p < 0.05 considered significant. Descriptive statistics that were calculated for serum IFN-α/IFN-β levels included means and standard deviations. The exact Wilcoxon rank-sum test was used; statistical tests were two sided and were performed using a significance level of 5% with SAS software, version 9.4 (SAS Institute).

Previous studies from our laboratory and others have underscored the importance of oxidative stress and free radical production during T1D progression in human patients and NOD mice (16, 22, 31, 32). Macrophages are major producers of NOX-derived superoxide, essential immune cells for T1D pathogenesis, and critical for the clearance of Coxsackievirus infection (33). To establish if diabetogenic CB3 infection (11) can induce NOX-derived superoxide production by macrophages, we performed in vitro CB3 infections of both NOD and NOD.Ncf1^m1J BM-MΦ and measured macromolecular-centered free radical adduct production by immuno spin-trapping (Fig. 1, Supplemental Fig. 2). As early as 4 h postinfection, NOD macrophages elicited an oxidative burst as shown by the appearance of green fluorescent puncta (Fig. 1A). Importantly, there was a decrease (p < 0.0001) in DMPO-adducts detected by immuno-spin trapping within CB3-infected NOD.Ncf1^m1J BM-MΦ cultures (Fig. 1B). These results demonstrate that upon Coxsackievirus infection, NOX-derived superoxide is induced by macrophages.

In the absence of NOX-derived superoxide production, NOD.Ncf1^m1J mice exhibit a delay in spontaneous and adoptive transfer of T1D (16, 22). Given that NOX-derived superoxide is produced by macrophages in response to CB3 infection (Fig. 1), we sought to determine if NOD.Ncf1^m1J mice exhibit a delay in CB3-induced T1D. Prediabetic female 12 wk old NOD and NOD.Ncf1^m1J mice were infected with 100 PFU of CB3, then observed for onset of T1D in comparison with HBSS-treated mice (Fig. 2A). NOD mice showed acceleration in diabetes onset as early as 2 wk and culminating at 5 wk postinfection (p = 0.0176) compared with HBSS controls (Fig. 2A). Acceleration of T1D was transient, as a cohort of CB3-infected NOD mice failed to succumb to autoimmune diabetes at 5 wk postinfection. NOD.Ncf1^m1J mice were significantly protected against spontaneous T1D (p < 0.0001), CB3-accelerated T1D at 5 wk postinfection (p = 0.0371), and at the end of the study (p < 0.05) compared with HBSS-treated NOD mice. These results show that superoxide production by the immune system plays a critical role in how viral infections trigger autoimmunity in NOD mice.

NOX-derived superoxide is essential for eradicating bacterial and fungal infections (34). Although the contribution of superoxide during viral clearance is not fully established, studies have suggested that a lack of dietary antioxidants can enhance Coxsackievirus infection (35). To determine if the loss of superoxide production impairs the clearance of CB3 infection, viral titers were measured from the pancreata of NOD and NOD.Ncf1^m1J at 3, 5, 7, 10, and 14 d postinfection (Fig. 2B). Surprisingly, even in the absence of NOX-derived superoxide, NOD.Ncf1^m1J pancreata harbored similar viral titers as compared with NOD pancreata and CB3 was efficiently cleared from NOD and NOD.Ncf1^m1J pancreata by 2 wk postinfection (Fig. 2B). Therefore, superoxide deficiency does not restrict the infectivity of CB3 within the pancreas, nor does it hinder immune clearance of CB3.

CXCL10, an inflammatory chemokine, is important in the acute immune response against Coxsackievirus infection (36) and expressed within the islets of T1D patients compared with healthy controls (37). Additionally, protective Ccl5 single nucleotide polymorphisms in T1D human studies are correlated with decreased serum-specific CCL5 production (38). To determine if superoxide deficiency caused diminutions in the chemokine response to CB3 infection, we compared levels of CXCL10 and CCL5 in the sera of NOD and NOD.Ncf1^m1J mice following infection (Fig. 3). CXCL10 was detected as early as day 3 and remained elevated in contrast to CB3-infected NOD.Ncf1^m1J mice through day 7 postinfection (Fig. 3A). There was a 4.3- (p = 0.0219) and 2-fold (p = 0.0483) reduction of CXCL10 in the sera of NOD.Ncf1^m1J mice at 5 and 7 d postinfection, respectively (Fig. 3A). CCL5 levels followed similar kinetics throughout infection, and in contrast to NOD, CCL5 levels were decreased 3.2- (p = 0.0268) and 3.7-fold (p = 0.0087) in the sera of CB3-infected NOD.Ncf1^m1J mice at 5 and 7 d postinfection, respectively (Fig. 3B). The marked decreases of serum CXCL10 and CCL5 during viral infection in NOD.Ncf1^m1J mice demonstrate an important role for NOX-erived superoxide in regulating chemokine production.

To determine if type I IFN synthesis is compromised in CB3-infected NOD and NOD.Ncf1^m1J mice, we assessed levels of IFN-α and IFN-β in the serum of mice at days 3 and 7 postinfection. At 3 d postinfection, NOD.Ncf1^m1J mice displayed similar levels of IFN-α and IFN-β production, as compared with NOD mice (Fig. 3C). However, by 7 d postinfection, IFN-α and IFN-β were still detectable in the NOD serum, but completely absent in the NOD.Ncf1^m1J serum (p = 0.018 and p = 0.007, respectively) (Fig. 3D). Thus, the synthesis of type I IFN at 3 d postinfection may facilitate viral clearance, whereas the extended superoxide-dependent type I IFN responses detected in NOD serum at day 7 may confer susceptibility to CB3-accelerated T1D.

With the confirmed loss of an oxidative burst by NOD.Ncf1^m1J macrophages upon CB3 infection (Fig. 1) and decreased circulating CXCL10, CCL5, and type I IFN levels (Fig. 3), we hypothesized that the delay in CB3-accelerated T1D in NOD.Ncf1^m1J mice (Fig. 2A) was partly due to the absence of synergistic effects of free radicals on proinflammatory macrophage antiviral responses. To address this, we analyzed innate immune antiviral responses within the pancreas of uninfected and CB3-infected NOD and NOD.Ncf1^m1J mice by quantitative RT-PCR (Fig. 4). By day 7 postinfection, an influx of macrophages was detected in the pancreata of both NOD and NOD.Ncf1^m1J mice as indicated by a 10.6- and 6.8-fold increase, respectively, in the transcript expression of Emr1, which encodes for the macrophage marker F4/80. Interestingly, analysis of the transcription factor, Stat1, involved in both type I IFN signaling and inflammatory M1 macrophage differentiation, was decreased 2.5-fold (p = 0.0012) in CB3-infected NOD.Ncf1^m1J pancreata compared with NOD. In addition to decreased Stat1, mRNA accumulation of inflammatory response gene Tnf was reduced by 4-fold (p = 0.0472). Antiviral mRNA accumulation was also severely attenuated within NOX-deficient pancreata, with an 8.3-fold (p = 0.0104) reduction in the expression of Ifnb1, and a 4.1-fold (p = 0.0041) decrease in expression of the IFN-stimulated gene, Isg15 (Fig. 4). Altogether, these results demonstrate that NOX-derived superoxide and the activation of reduction-oxidation (redox)-dependent signaling pathways enhance the innate immune antiviral response against CB3 infection.

To investigate the impact of superoxide deficiency on macrophage responses to viral infection, the pancreata from HBSS-treated or CB3-infected NOD and NOD.Ncf1^m1J mice were immunophenotyped for infiltrating macrophages (F4/80⁺ CD11b⁺) by flow cytometry (Fig. 5, Supplemental Fig. 1). Despite the dampened serum levels of CXCL10 and CCL5 in CB3-infected NOD.Ncf1^m1J mice (Fig. 3), and decreased (p = 0.03) total number of infiltrating cells within the NOD.Ncf1^m1J pancreas [22.5 × 10⁶ ± 3.4 × 10⁶ (NOD.Ncf1^m1J) versus 32.7 × 10⁶ ± 3.0 × 10⁶ (NOD)] (Fig. 5A), macrophage recruitment into the pancreas was similar between NOD and NOD.Ncf1^m1J mice by both percentage [22.9 ± 3.4 (NOD) versus 20.6 ± 4.6 (NOD.Ncf1^m1J)] and cell counts [8.6 × 10⁶ ± 1.9 × 10⁶ (NOD) versus 6.2 × 10⁶ ± 2.0 × 10⁶ (NOD.Ncf1^m1J)] at 7 d postinfection (Fig. 5A). To determine if the loss of NOX-derived superoxide hampers proinflammatory TNF-α synthesis of pancreas-infiltrating macrophages, intracellular cytokine staining was performed and revealed a decrease in the percentage (1.9-fold, p = 0.0293), count (2.3-fold, p = 0.0335), and geometric mean fluorescence intensity (gMFI) (2.4-fold, p = 0.0064) compared with CB3-infected NOD mice (Fig. 5B, 5D). Macrophages recovered from the pancreas at day 7 post-CB3 infection were in vitro challenged with LPS for 4 h. Interestingly, NOD.Ncf1^m1J macrophages still displayed decreased intracellular TNF-α synthesis compared with NOD, by percentage (2.1-fold, p = 0.0206), total number (2.9-fold, p = 0.0136), and gMFI (3.0-fold, p = 0.0255) (Fig. 5C, 5D). Therefore, our data demonstrate the importance of NOX-derived superoxide to mature and enhance proinflammatory M1 macrophage responses in the pancreas following CB3 infection.

To determine the redox-regulated mechanisms involved in macrophage antiviral responses, NOD and NOD.Ncf1^m1J BM-MΦ were infected with CB3 and induction of proinflammatory innate immune responses was assessed. First, to confirm the in vivo phenotype of a decreased antiviral signature within NOD.Ncf1^m1J pancreata (Fig. 4), mRNA accumulation of proinflammatory and type I IFN response genes was assessed (Fig. 6A). Whereas infection of NOD BM-MΦ with CB3 induced robust inflammatory responses, NOD.Ncf1^m1J BM-MΦ displayed reductions in Tnf (1.4-fold, p = 0.0001), Cxcl10 (2-fold, p < 0.0001), and Ccl5 (1.5-fold, p = 0.0023) transcripts, in addition to attenuated Ifnb1 (1.9-fold, p = 0.0106) and Isg15 (1.4-fold, p = 0.0059) type I IFN responses. Additionally, superoxide deficiency induced significant diminutions in the viral RNA sensors, Tlr3 (1.3-fold, p = 0.0026) and Ddx58 (1.5-fold, p = 0.0095), whereas Ifih1 remained unaffected (Fig. 6A). To corroborate the observed decrease in mRNA accumulation, ELISAs of cytokine and chemokine production were analyzed within culture supernatants. Compared to NOD, production of TNF-α (Fig. 6B), CXCL10 (Fig. 6C), CCL5 (Fig. 6D), and IFN-β (Fig. 6E) by NOD.Ncf1^m1J BM-MΦ were diminished 2-fold (p < 0.0001), 1.4-fold (p = 0.0001), 2-fold (p = 0.0009), and 11-fold (p < 0.0001), respectively. These results support our previous reports that inflammatory macrophage responses are diminished in the absence of superoxide production (26, 27) and, more importantly, provide evidence that redox status can influence macrophage antiviral responses.

To determine whether antiviral responses in macrophages were redox regulated, exogenous superoxide was added to CB3-infected NOD and NOD.Ncf1^m1J BM-MΦs via XO, an enzyme that produces superoxide as a byproduct during the production of uric acid from xanthine. Western blot analysis of cellular lysate at 24 h postinfection revealed that MDA5, a viral dsRNA sensor critical for antiviral responses, was dampened 2.6-fold (p = 0.0024) in CB3-infected NOD.Ncf1^m1J BM-MΦ compared with NOD (Fig. 7A, 7B). With the addition of XO to NOD.Ncf1^m1J BM-MΦ, expression of MDA5 increased by 1.6-fold, restoring to near NOD wild-type levels (Fig. 7A, 7B). Interestingly, even without CB3 challenge, XO addition enhanced MDA5 basal expression by more than 15-fold in both NOD (p = 0.0403) and NOD.Ncf1^m1J (p = 0.0495) BM-MΦ (Fig, 7A, 7B). RIG-I, another cytosolic viral RNA sensor, was upregulated by over 2-fold in NOD (p = 0.0188) and NOD.Ncf1^m1J (p = 0.0067) BM-MΦ with XO alone (Fig. 7C, 7D). Compared to NOD BM-MΦ, RIG-I expression in CB3-infected NOD.Ncf1^m1J BM-MΦ was decreased by 1.6-fold (p = 0.0011) and addition of exogenous superoxide restored RIG-I protein levels by 1.3-fold back to wild-type NOD levels (Fig. 7C, 7D). Together, these results suggest that NOX-derived superoxide enhances the sensitivity of macrophages to detect and respond to viral infections.

Because the absence of NOX-derived superoxide compromised M1 macrophage differentiation during spontaneous and adoptive transfer of T1D in NOD.Ncf1^m1J mice (26), we determined if CB3-infected NOD.Ncf1^m1J macrophages displayed decreases in STAT1 signaling, a transcription factor involved in inflammatory M1 differentiation and type I IFN signaling. By 6 h, the addition of XO alone induced expression of P-STAT1 (Y701) in NOD and NOD.Ncf1^m1J macrophages (Fig. 7E, 7F). Importantly, upon CB3 infection there was a 2.5-fold decrease (p = 0.0481) in P-STAT1 (Y701) in NOD.Ncf1^m1J macrophages, which was rescued with a 3.8-fold increase (p = 0.0108) upon addition of XO (Fig. 7F). Interestingly, STAT1 protein expression may also be redox regulated, as NOD.Ncf1^m1J macrophages displayed a 1.8-fold decrease (p = 0.0092) in total STAT1 levels upon CB3 infection (Fig. 7E, 7G).

To further assess STAT1 activation and M1 differentiation, nuclear translocation of P-STAT1 was analyzed by immunofluorescence staining (Fig. 8). Then 6 h postinfection, NOD and NOD.Ncf1^m1J BM-MΦ were costained for P-STAT1 (Y701), F4/80, and DAPI. Nuclear expression of P-STAT1 (Y701) was readily detected upon CB3 infection in F4/80⁺ NOD macrophages, and was significantly decreased (p < 0.0001) in CB3-infected NOD.Ncf1^m1J macrophages (Fig. 8). Similar to the immunoblot analysis, treatment with XO alone elicited nuclear translocation of P-STAT1 (Y701) for NOD and NOD.Ncf1^m1J macrophages (p < 0.0001 for both), and XO addition rescued the NOD.Ncf1^m1J phenotype upon CB3 infection (Fig. 8). Therefore, superoxide production can enhance M1 macrophage differentiation by increasing STAT1 phosphorylation and nuclear translocation following diabetogenic CB3 viral infections.

T1D is a chronic inflammatory autoimmune disease, and studies in both the NOD mouse model and human patients have shown that oxidative stress coincides with the inflammatory progression of T1D (32). Free radicals can influence innate and adaptive immune cell activation and effector responses in T1D (23–26). Recently, our laboratory demonstrated that NOX-derived superoxide within the islet microenvironment can influence macrophage differentiation and progression to T1D (26). Islet-resident NOD macrophages progressively displayed a proinflammatory M1 phenotype during spontaneous T1D, whereas NOD.Ncf1^m1J islet-resident macrophages were skewed toward a noninflammatory M2 phenotype and fostered a protective microenvironment (26). Given the prominent role of macrophages in driving autoimmunity (16, 17), we hypothesized that environmental cues, such as viral infections, may trigger diabetogenic proinflammatory macrophage responses that are regulated by the NOX complex.

Our results demonstrate that innate immune signaling by macrophages following CB3 infection is redox regulated and provides an essential link with diabetogenic viruses, oxidative stress, and the activation of autoimmunity in T1D. We provide evidence that superoxide-deficient NOD.Ncf1^m1J mice exhibited a significant delay in CB3-accelerated T1D in contrast to NOD mice. Interestingly, we also observed that not all CB3-infected NOD mice became diabetic and corroborated a previous report by Serreze et al. (13) using a CB4-accelerated T1D model. The discrepancy in acceleration and protection from Coxsackievirus-induced T1D may be due to the degree of insulitis within NOD mice prior to CB3 infection, as infections of NOD mice with low levels of insulitis exhibit a delay in T1D (39), whereas infections of NOD mice with established insulitis exhibit a break in tolerance and progression to T1D (11, 13). Given the range of virulence, persistence, and cellular specificity across variants within both CB3 (11) and CB4 strains (10), it is also plausible that the CB3 variant used in our studies causes an acute infection that is rapidly cleared prior to efficiently generating a sustained level of oxidative stress and inflammation necessary to activate all effector T cells in the pancreas, especially of those mice with lower insulitis at the time of infection. Our data support this possibility, as CB3 viral titer was not detected from the pancreata of NOD and NOD.Ncf1^m1J mice at 2 wk postinfection, suggesting that this strain confers an acute, rather than chronic infection. However, understanding how oxidative stress is involved in persistent Coxsackievirus infections could provide even further insight into how viruses can trigger T1D in humans.

Interestingly, even in the absence of superoxide, the antiviral response of NOD.Ncf1^m1J mice was sufficient for viral clearance, but unable to trigger T1D. The proinflammatory innate immune response by NOD.Ncf1^m1J macrophages following CB3 infection was not completely ablated, but was at sufficient levels to contribute toward viral clearance. This may be particularly true for NOD macrophages, as they are inherently hyperinflammatory and, in the absence of superoxide synthesis, NOD.Ncf1^m1J macrophages have an antiviral response that is still capable of CB3 clearance, but not sufficient to accelerate autoimmune diabetes. Our data assessing serum type I IFN responses support this conclusion, as the early day 3 type I IFN responses remained intact, and a prolonged response was seen at day 7 in T1D-susceptible NOD, but not NOD.Ncf1^m1J serum. Our results showing similar IFN-α levels between NOD and NOD.Ncf1^m1J serum also aligns with results from a study by Lincez et al. (40), showing that IFN-α responses are associated with Coxsackie B virus clearance. Additionally, the loss of NOX-derived superoxide may not compromise the antiviral response of other immune cells necessary for viral clearance including NK cells, CD8 T cells, and B cells. Future studies will determine if NOD.Ncf1^m1J CD8 T cells display increases in cytolytic function and/or if B cells produce an enhanced humoral response following CB3 infection.

One mechanism of protection against virus-induced T1D afforded by the Ncf1^m1J mutation was due to dampened inflammatory macrophage responses. Activated NOD macrophages exhibit a hyperinflammatory phenotype (18), which may also constitute an exacerbated antiviral response during diabetogenic viral infections. The role of innate immune responses has garnered great interest recently, as antiviral response signatures can be detected in T1D-susceptible individuals prior to autoantibody seroconversion (41, 42). TNF-α, IFN-β, CXCL10, and CCL5 were highly produced following CB3 infection of NOD macrophages, but were significantly reduced in NOD.Ncf1^m1J macrophages. Pancreas-infiltrating CD11b⁺ F4/80⁺ NOD.Ncf1^m1J macrophages were less inflammatory as shown by decreased TNF-α production in contrast to NOD mice. Further corroborating the dampened inflammatory response of CB3-infected NOD.Ncf1^m1J macrophages, a decrease in antiviral signaling pathway activation was also observed. CB3 infection of NOD macrophages resulted in the upregulation of the viral RNA sensors, MDA5 and RIG-I, which was dampened in NOD.Ncf1^m1J macrophages. Importantly, exogenous addition of superoxide enhanced the expression of STAT1, MDA5, and RIG-I in CB3-infected NOD.Ncf1^m1J macrophages to wild-type levels. Therefore, these results support the hypothesis that oxidative stress can enhance proinflammatory antiviral responses following diabetogenic viral infections in the NOD mouse and, ultimately, contribute to viral-induced T1D. Similar results were also observed with RIG-I and MDA5 expression in nasal epithelial cells following Influenza A Virus infection, as Duox2-derived hydrogen peroxide could induce RIG-I and MDA5 transcription (43). The findings in our study provide further mechanistic data of the connections with redox biology, dysregulated autoimmune responses in T1D, and how viral infections can exacerbate those interactions.

Our current study demonstrates that STAT1 expression and phosphorylation in macrophages are redox dependent upon CB3 infection, as exogenous superoxide restored STAT1 activation in NOD.Ncf1^m1J macrophages. The effect of free radicals on STAT1 may be direct and/or indirect via the upstream JAK1, JAK2, or TYK2 kinases. In support of the latter, Duox2-derived hydrogen peroxide can induce P-STAT1 (Y701) expression in glia by a JAK2-dependent mechanism (44). Additionally, oxidation of several protein tyrosine phosphatase family members exacerbated STAT1 signaling in insulitic prediabetic NOD pancreata (45). Therefore, these results warrant further investigation into the redox-dependent mechanism of STAT1 activation in proinflammatory M1 macrophage differentiation. Interestingly, whereas superoxide deficiency during the natural progression of T1D resulted in a skewed anti-inflammatory M2 macrophage phenotype (26), mRNA accumulation of M2-related markers including Cd206 and Stat6 were undetected in CB3-infected NOD.Ncf1^m1J pancreata or in vitro–challenged macrophages (data not shown). Altogether, blocking the redox-dependent cues necessary for M1 macrophage differentiation is sufficient to delay CB3-accelerated T1D.

We thank Jessie Barra, Dr. Jon Piganelli, Dr. Sasanka Ramanadham, and Dr. Ruth McDowell for critical reading of the manuscript. We are also appreciative of the biostatistical analyses performed by Dr. Robert Oster from the Division of Preventive Medicine at University of Alabama at Birmingham.

The authors have no financial conflicts of interest.