NCF1-dependent production of ROS protects against lupus by regulating plasmacytoid dendritic cell development and functions

Low capacity to produce ROS because of mutations in neutrophil cytosolic factor 1 (NCF1/p47phox), a component of NADPH oxidase 2 (NOX2) complex, is strongly associated with systemic lupus erythematosus in both humans and mouse models. Here, we aimed to identify the key immune cell type(s) and cellular mechanisms driving lupus pathogenesis under the condition of NCF1-dependent ROS deficiency. Using cell-specific Cre-deleter, human NCF1-339 variant knockin, and transgenic mouse strains, we show that low ROS production in plasmacytoid dendritic cells (pDCs) exacerbated both pristane-induced lupus and a potentially new Y-linked autoimmune accelerating locus–related spontaneous model by promoting pDC accumulation in multiple organs during lupus development, accompanied by elevated IFN-α levels and expression of IFN-stimulated genes. Mechanistic studies revealed that ROS deficiency enhanced pDC generation through the AKT/mTOR pathway and CCR2-mediated migration to tissues, which together with hyperactivation of the redox-sensitive stimulator of interferon genes/IFN-α/JAK1/STAT1 cascade further augmented type I IFN responses. More importantly, by suppressing these pathways, restoration of NOX2-derived ROS specifically in pDCs protected against lupus. These discoveries explain the causative effect of dysfunctional NCF1 in lupus and demonstrate the protective role of pDC-derived ROS in disease development driven by NCF1-dependent ROS deficiency.


Introduction
Systemic lupus erythematosus (SLE) is a chronic systemic autoimmune disease with a global incidence of around 0.1%, predominantly observed in women of child-bearing age (1). Patients with SLE present a broad range of clinical manifestations, including skin rash, arthritis, glomerulonephritis, neurological symptoms, hemolytic anemia, and vasculitis (2). The disease is characterized by loss of tolerance to ubiquitous self-antigens, high titers of autoantibodies directed against nuclear components such as dsDNA and nucleosomes, immune complex (IC) deposition, as well as involvement of multiple organs. Organ damage occurs as a consequence of the chronic, uncontrolled autoimmune responses, or secondarily because of vasculitis-and thrombosis-induced ischemia. The majority of patients with SLE display overactivation of the type I IFN system, a so-called type I IFN signature, indicated by increased serum level of IFN-α and expression of type I IFN-stimulated genes (ISGs), which correlates with SLE activity and severity (1). IFNs act as an immune adjuvant for diverse immune responses, including production of proinflammatory cytokines and autoantibodies by stimulating monocytes, T cells, and B cells (3). The exact cause of SLE has not been fully understood, yet it is believed that genetic predisposition together with exogenous factors such as viral infections, smoking, UV irradiation, and certain drugs lead to dysregulation of innate and adaptive immune systems, subsequently driving the clinical disease (4).
Low capacity to produce ROS because of mutations in neutrophil cytosolic factor 1 (NCF1/ p47phox), a component of NADPH oxidase 2 (NOX2) complex, is strongly associated with systemic lupus erythematosus in both humans and mouse models. Here, we aimed to identify the key immune cell type(s) and cellular mechanisms driving lupus pathogenesis under the condition of NCF1-dependent ROS deficiency. Using cell-specific Cre-deleter, human NCF1-339 variant knockin, and transgenic mouse strains, we show that low ROS production in plasmacytoid dendritic cells (pDCs) exacerbated both pristane-induced lupus and a potentially new Y-linked autoimmune accelerating locus-related spontaneous model by promoting pDC accumulation in multiple organs during lupus development, accompanied by elevated IFN-α levels and expression of IFN-stimulated genes. Mechanistic studies revealed that ROS deficiency enhanced pDC generation through the AKT/mTOR pathway and CCR2-mediated migration to tissues, which together with hyperactivation of the redoxsensitive stimulator of interferon genes/IFN-α/JAK1/STAT1 cascade further augmented type I IFN responses. More importantly, by suppressing these pathways, restoration of NOX2-derived ROS specifically in pDCs protected against lupus. These discoveries explain the causative effect of dysfunctional NCF1 in lupus and demonstrate the protective role of pDC-derived ROS in disease development driven by NCF1-dependent ROS deficiency.
Since type I IFNs are central and reflective of lupus severity and activity, we wondered if NOX2derived ROS ameliorate kidney damage in Ncf1 m1j/m1j .MN Tg mice by dampening IFN responses. We measured expression and phosphorylation of STAT1 (p-STAT1), one of the key transcriptional factors activating ISG expression. Higher STAT1 and p-STAT1 were found in the kidneys of Ncf1 m1j/m1j mice compared with WT mice (Figure 3, E and F). Of importance, we observed comparable levels of STAT1 and p-STAT1 in the kidneys of Ncf1 m1j/m1j .MN Tg and WT mice (Figure 3, E and F). Consistent with IFN signaling, we found more pDCs, the main producer of type I IFNs, infiltrating kidneys of Ncf1 m1j/m1j mice compared with WT mice and a similar number between Ncf1 m1j/m1j .MN Tg and WT mice ( Figure 3G). In addition to kidneys, we observed the same pattern in the spleen at 7 MPI (Supplemental Figure 2B). These results suggest that ROS produced by monocytes/macrophages and pDCs protect against PIL by controlling pDC accumulation at inflammatory sites and in lymphoid organs, thereby limiting IFN responses during disease development.
Restoration of ROS in CD11c + cells, predominantly DCs, alleviates PIL. Since restoration of ROS specifically in CD68 + cells limits pDC accumulation and IFN signature, fully protecting Ncf1 m1j/m1j .MN Tg mice from PIL exacerbation, we hypothesized that the protection is, at least partially, because of intrinsic ROS in pDCs. To further verify the role of NOX2-derived ROS produced by pDCs in lupus, we used the B6N.Q. CD11c-Cre + .TN3 (CD11c-Cre + .TN3) strain, which expresses functional NCF1 protein only in CD11c + cells. Unlike human pDCs, which are CD11c -, mouse pDCs express CD11c at a low level (21). As expected, the capacity to make an ROS response was partially restored in conventional DCs (cDCs) and pDCs of the CD11c-Cre + .TN3 strain. Neither CD11c-Cre + .TN3 nor CD11c-Cre -.TN3 mice displayed oxidative burst in neutrophils or macrophages/monocytes ( Figure 4A).
We then injected pristane to induce experimental lupus in CD11c-Cre + .TN3 mice. Mice with restored ROS in DCs developed lower levels of autoantibodies against nucleosomes and Smith/ribonucleoprotein (Sm/RNP) during the development of the disease, while there was no difference in the levels of anti-dsDNA IgG ( Figure 4B). Kidney injury was less severe in CD11c-Cre + .TN3 mice compared with CD11c-Cre -.TN3 mice, as indicated by lower urinary protein levels at 6 MPI ( Figure 4C). Furthermore, lower expression of STAT1 and other ISGs was found in kidneys from CD11c-Cre + .TN3 mice (Figure 4, D and E). Consistently, we observed milder renal infiltration of pDCs in CD11c-Cre + .TN3 male mice than in their sex-matched Crelittermates, which correlated with Stat1 expression (Figure 4, F and G). In contrast, kidney-infiltrating cDCs exhibited neither difference in cell counts nor a correlation with Stat1, further indicating that ROS-sufficient pDCs are the major regulatory cells ameliorating the disease (Figure 4, H and I).
In conclusion, our results show ROS produced by DCs ameliorate lupus and lupus nephritis through regulation of pDC accumulation and IFN responses.
ROS produced by DCs prevent spontaneous development of lupus in Ncf1-mutant mice with Yaa. Having found that ROS produced by DCs ameliorate PIL, we next asked if DC-derived ROS protect mice from ROS deficiency-induced disease exacerbation as seen in Ncf1 m1j/m1j .MN Tg . For this purpose, we established a mouse strain carrying both the Ncf1 m1j mutation and the Y-linked autoimmune accelerating locus (Yaa). Yaa is a region translocated from the X chromosome onto the Y chromosome, resulting in spontaneous lupus in the male mice with lupus-prone backgrounds (18). The duplicated Tlr7 gene in the Yaa and the resultant overexpression of TLR7 lead to lupus-like disease phenotypes, demonstrating the important role of TLR7-induced type I IFNs as the causative pathway for lupus (18). It is known that pDCs potently produce type I IFNs in response to ligation of TLR7, which can be activated by RNA-containing ICs and are associated with B cell activation and lupus development (22). Therefore, we considered Yaa mice as a useful tool to investigate the protective role of ROS in pDCs in lupus.
We earlier observed spontaneous development of signs of lupus in Balb/c.Ncf1 m1j mice; however, in mice with C57.black backgrounds such as B10.Q or C57BL/6, neither Ncf1 m1j mutation nor Yaa alone spontaneously induces lupus-like disease (5,23). We introduced Yaa into the CD11c-Cre.TN3 mouse line and observed the male mice over 5 months. BQ.CD11c-Cre -.TN3.Yaa males developed higher levels of autoantibodies against dsDNA and nucleosomes, as compared with BQ.Yaa ( Figure 5A). Comparable levels of anti-dsDNA JCI Insight 2023;8(7):e164875 https://doi.org/10.1172/jci.insight.164875 antibodies but not anti-nucleosome antibodies were found in the sera from BQ.CD11c-Cre + .TN3.Yaa and BQ.Yaa mice, suggesting that DC-specific ROS at least partially regulate autoantibody production ( Figure  5A). In addition, we found similar urinary protein levels in the 5-month-old BQ.CD11c-Cre + .TN3.Yaa and BQ.Yaa male mice, significantly lower than in BQ.CD11c-Cre -.TN3.Yaa mice ( Figure 5B). Consistently, we observed more total CD45 + cells, CD4 + T cells, and B cells in the kidneys of BQ.CD11c-Cre -.TN3.Yaa mice, while comparable numbers were found in BQ.CD11c-Cre + .TN3.Yaa and BQ.Yaa ( Figure 5C). Although there was only a trend in total CD4 + T cells, the percentage of Foxp3 + Tregs was significantly higher in Cre + mice than their Crelittermates (Supplemental Figure 3). These results indicate that DC-specific ROS protect ROSdeficient Yaa mice from kidney injuries and immune cell infiltration during lupus development. Restoration of ROS in DCs limited pDC accumulation in kidneys and spleens in BQ.CD11c-Cre + .TN3.Yaa mice, to the same level as in BQ.Yaa ( Figure 5, D and E). PDCs are also suggested to be involved in Sjögren's syndrome (SS), another chronic autoimmune disorder closely related to SLE. SLE and SS share possible underlying pathogenic pathways, such as TLR activation and type I IFN signaling (24). Since lymphocytic infiltration in salivary glands is characteristic of SS and usually concomitantly occurs with lupus activity in other organs, we immunophenotyped the infiltrates within salivary glands from the 3 strains. We found more CD45 + cells and predominantly B cells in the salivary glands in BQ.CD11c-Cre -.TN3.Yaa mice, as compared with BQ.Yaa ( Figure 5F). More importantly, comparable frequency of immune cells was found in BQ.CD11c-Cre + .TN3.Yaa and BQ.Yaa mice, indicating that DC-derived ROS fully prevent inflammation in salivary glands during lupus development in ROS-deficient Yaa mice.
Since lung involvement is common in SLE and has been reported in lupus-prone Yaa mice (25), we characterized leukocytes from bronchoalveolar lavage (BAL) fluids from BQ.Yaa and BQ.CD11c-Cre. TN3.Yaa males. Higher frequencies of B cells, CD4 + T cells, and neutrophils were found in BAL fluids from 5-month-old BQ.CD11c-Cre -.TN3.Yaa mice, as compared with the other 2 strains, while comparable frequency was seen between BQ.CD11c-Cre + .TN3.Yaa and BQ.Yaa mice ( Figure 5G). This indicates that ROS produced by DCs prevent lupus-associated lung inflammation.
In conclusion, NCF1-dependent ROS deficiency drives spontaneous lupus development in mice carrying Yaa, reflected by autoantibodies against dsDNA and nucleosomes, kidney damage, and involvement of lungs and salivary glands. More importantly, ROS restoration specifically in DCs protect against lupus by controlling autoantibody production and preventing immune infiltration, especially pDC accumulation, in multiple organs.
ROS regulate accumulation of pDCs and upregulation of ISGs at the initial stage of PIL. Using Ncf1 m1j/m1j , Ncf1 m1j/m1j .MN Tg , CD11c-Cre + .TN3, and CD11c-Cre + .TN3.Yaa mice, we demonstrated accumulation of pDCs in various target organs during PIL and spontaneous lupus development, which is regulated by NOX2derived ROS. Next, to investigate when the regulatory effects occur, we immunophenotyped the cells in the peritoneum, where initial activation of the immune system starts in the PIL model. By analyzing the pristaneinduced peritoneal pathology at the early stage (day 3 after pristane injection), we found that Balb/c.Ncf1 m1j/m1j mice exhibited dramatic accumulation of various immune cells, such as B cells, NK cells, macrophages, and most strikingly, pDCs within peritoneal exudates ( Figure 6A and Supplemental Figure 4A). In concert with the number of pDCs, higher expression of ISGs was found in the peritoneum of Balb/c.Ncf1 m1j/m1j mice than WT mice, suggesting a stronger ISG and inflammatory response ( Figure 6B). However, Ncf1 m1j/m1j .MN Tg and B10.Q mice displayed comparable frequency and numbers of pDCs in the peritoneum at day 3 after pristane injection ( Figure 6C), and CD11c-Cre + .TN3 mice showed significantly lower Stat1 expression than their Crelittermates ( Figure 6D). These results showed that CD68 + CD11c + cell-derived ROS regulation of pDC accumulation and type I IFN responses occur at the initial stage of lupus.
Insufficient ROS production in Ncf1 R90/90H mice carrying the human NCF1-339 SNP potentiates pDC responses. To verify that the alteration in pDCs and ISGs observed in Ncf1 m1j/m1j mice is relevant to humans with NCF1-339, we phenotyped the peritoneal cells from B10.Q.Ncf1 R90/90H mice at day 3 after pristane injection. This strain carries the human NCF1-339 SNP on an allele of Ncf1, leading to an amino acid exchange from arginine to histidine (from R90 to 90H), which results in reduced NOX2-derived ROS production (Supplemental Figure 4B). We found higher frequency and numbers of pDCs in the peritoneum of Ncf1 R90/90H mice compared with WT mice ( Figure 6E). Unlike in Ncf1 m1j/m1j mice, no difference in any other immune cell type was found in Ncf1 R90/90H , which further indicates the importance of pDCs (Supplemental Figure 4C). Moreover, Ncf1 R90/90H mice displayed higher levels of IFN-α in the peritoneal fluids, which was consistent with our observation in CD11c-Cre.TN3 strains (Supplemental Figure 5). Furthermore, we found higher expression of STAT1 in the peritoneal cells of Ncf1 R90/90H mice than their WT littermates (Figure 6, F and G). Our results demonstrate that moderate decrease in ROS production in Ncf1 R90/90H mice specifically promotes pDC accumulation and resultant type I IFN activation at an early stage of lupus.
Generation of pDCs is enhanced via AKT/mTOR pathway in ROS-deficient mice. Since ROS regulation of pDC accumulation occurs already at the initial stage of PIL, we hypothesized that ROS deficiency promotes generation of pDCs in BM. It is known that pDCs originate from common DC progenitors (CDPs) and common lymphoid progenitors (26,27). We found more CDP-derived pre-pDCs in Ncf1 R90/90H BM at day 3 after pristane injection, suggesting that the higher number of pDCs in ROS-deficient mice is at least partially due to enhanced pDC production ( Figure 7A). However, at naive status, the number of pre-pDCs was comparable between WT and Ncf1 R90/90H mice, indicating that the difference in pDC progenitors is regulated (Supplemental Figure 6). To further characterize how ROS affects pDC generation, we compared the proteomic landscape of WT and Ncf1 m1j/m1j BM-derived pDCs after in vitro stimulation by IFN-α, which autocrinally supports a broad range of pDC activities, including development, differentiation, maturation, and survival (28,29). BM cells were sorted by magnetic beads with anti-B220 mAb, and the resultant population was a mixture of mature pDCs and CCR9 -pDC precursors, making it suitable for studying pDC development (30). Besides deep expression profiling, to capture the differences in other aspects such as protein-protein interactions and posttranslational modifications, we also assessed the protein solubility and/or thermal stability with the Proteome Integral Solubility Alteration (PISA) assay (31). In fact, differentially expressed proteins in Ncf1 m1j/m1j pDCs were mainly enriched in processes related to hematopoiesis/myelopoiesis, immune responses, and redox activity ( Figure  7B). The PISA assay also revealed protein enrichment in hematopoietic development ( Figure 7C). To identify the underlying pathways, we analyzed the protein-protein interactions among the proteins clustered in the hematopoiesis process in the PISA assay with STRING database and found the PI3K/AKT/mTOR signaling pathway to be the major one involved in ROS-regulated pDC development ( Figure 7D). To validate the findings, we isolated BM SiglecH + cells, which consist of 50% pDC precursors and 20% mature pDCs (Supplemental Figure 7), and stimulated them with imiquimod, a TLR7 agonist, since TLR7 ligation has been shown to promote mTOR-dependent emergency myelopoiesis, leading to massive myeloid cell and DC expansion (32). We observed higher phosphorylation of mTOR and its upstream signal AKT in Ncf1 m1j/m1j cells, as compared with ROS-sufficient WT cells ( Figure 7E). To confirm that upregulation of AKT/mTOR pathway is ROS dependent, we abolished NOX2-derived ROS production in imiquimod-treated B10.Q splenocytes by a specific NOX2 inhibitor, GSK2795039 (GSK), and supplied Ncf1 R90/90H splenocytes with H 2 O 2 (33). At 20 μM GSK increased mTOR pathway in WT cells to a similar level as in Ncf1 R90/90H cells, while 20 μM H 2 O 2 reduced  U test (B and D-G), and 1-way ANOVA with Tukey's multiple comparison test (C). *P < 0.05, **P < 0.01, ***P < 0.001.  Figure 7F). Similar results were found in BM cells (Supplemental Figure 8A). These findings demonstrate that NCF1-dependent ROS control pDC generation through AKT/mTOR pathway.
Apart from myelopoiesis, enhanced cellular migration may also contribute to pDC accumulation during lupus. It is known that TLR7 activation induces pDC migration to the inflammatory sites via the CCL2/ CCR2 axis (34). CCR2 is an inducible inflammatory chemokine receptor involved in leukocyte traffic in both humans and murine lupus nephritis (35,36). It also promotes pDCs' egress from BM at steady state (37). We observed significantly higher CCR2 expression on mature pDCs in BM from Ncf1 R90/90H naive mice, as compared with WT mice, and a similar trend at day 3 after pristane injection ( Figure 7G). Decreased CCR2 expression after rapamycin treatment indicated its association with the mTOR pathway (Supplemental Figure  8B). We also measured other molecules known to be involved in pDC migration under certain conditions, including CXCR3, CXCR4, CCR5, CCR6, CCR7, β 1 /β 2 integrins, CD62L, and endothelial adhesion molecules ICAM-1 and VCAM-1 (38), on BM and peritoneal pDCs from naive and pristane-treated mice. None of these molecules showed different expression levels between WT and Ncf1 R90/90H pDCs (Supplemental Figure 9).
Taken together, low ROS production due to NCF1 deficiency enhances AKT/mTOR-dependent pDC generation and promotes pDC migration via CCR2, which may explain the accumulation of pDCs during lupus.
Hyperactive ROS-deficient pDCs exacerbate PIL. We next investigated whether intrinsic ROS regulate the functionality of pDCs. Since pDCs are rare cells, for subsequent experiments, we mostly used pDCs derived from BM.
It is known that pDCs potently produce type I IFNs in response to ligation of nucleic acid-sensing pattern recognition receptors (PRRs), such as TLR7 and TLR9, which have been implicated in lupus development (22). We stimulated BM-derived pDCs from WT and Ncf1 m1j/m1j mice with the TLR7 agonist imiquimod and TLR9 agonist CpG ODN2395. Both induced comparable level of IFN-α production by ROS-sufficient and -deficient pDCs ( Figure 8A). In addition, activation of the STING pathway has recently been shown to trigger IFN-α production in pDCs from SLE patients (39), although its involvement in experimental lupus models is under debate (40). We treated WT and Ncf1 m1j/m1j pDCs with the STING agonist 2′3′-cyclic guanosine monophosphate (2′3′-cGAMP) and found higher secretion of IFN-α by Ncf1 m1j/m1j cells ( Figure 8B). The autocrine/paracrine signaling where IFN-α further induces production of itself revealed no difference between the strains ( Figure 8C).
Next, we assessed the STAT1 pathway, which is downstream of IFN-α stimulation. As a member of ISGs, STAT1 has been reported to be upregulated at mRNA and protein levels in whole blood from Ncf1 m1j/m1j mice and NCF1-339-carrying patients (5, 7), yet its activity has not been clearly addressed in NCF1-deficient cells. We found higher phosphorylation of STAT1 in Ncf1 m1j/m1j pDCs than WT pDCs ( Figure 8D), and consequently, higher transcription of ISGs ( Figure 8E). These results indicate that ROS-deficient pDCs are hyperresponsive to IFN-α through STAT1 signaling, which forms a vicious cycle of type I IFN responses.
To verify that activation of STAT1 is regulated by ROS, we stimulated pDCs with or without H 2 O 2 supplement. Ncf1 m1j/m1j pDCs, in the presence of H 2 O 2 , displayed a comparable level of p-STAT1 to WT pDCs after IFN-α stimulation, while p-STAT1 in ROS-deficient pDCs cultured without H 2 O 2 was higher than in WT pDCs ( Figure 8F). We therefore conclude that ROS-deficient pDCs display hyperactivation of IFN-α/STAT1 signaling, which can be limited by ROS. JAK1, one of the tyrosine kinases giving upstream signals to STAT1, is known to be sensitive to ROS (41). We hypothesized that ROS regulate STAT1 by modifying JAK1 activity. To test the hypothesis, we measured JAK1 phosphorylation in pDCs upon stimulation with IFN-α and observed higher p-JAK1 in ROS-deficient Ncf1 m1j/m1j pDCs, as compared with WT pDCs ( Figure 8G). Blocking JAK1 activity with a JAK inhibitor, baricitinib, decreased the levels of p-STAT1 in ROS-deficient pDCs in vitro (Supplemental Figure  10A). Moreover, oral administration of baricitinib ameliorated spontaneous lupus in Ncf1 m1j/m1j .Yaa mice (Supplemental Figure 10, B-D). These results suggest that insufficient ROS regulation of JAK1 is at least partially responsible for stronger STAT1 activation in pDCs and exacerbation of lupus in Ncf1 m1j/m1j mice.
Having found that ROS-deficient pDCs displayed increased IFN-α production and type I IFN responses via the STING and JAK1/STAT1 pathways, we further investigated if these hyperreactive cells exacerbate lupus. We transferred Balb/c WT and Balb/c.Ncf1 m1j/m1j pDCs to pristane-treated Balb/c WT mice and followed them for 5 months. We found higher levels of autoantibodies against dsDNA, nucleosomes, Sm/RNP, and cardiolipin in the mice receiving ROS-deficient pDCs, as compared with those receiving ROS-sufficient pDCs ( Figure 8H). They also developed more severe lupus nephritis, as indicated by higher urinary protein concentration ( Figure 8I). Consistently, the mice receiving Ncf1 m1j/m1j pDCs exhibited more inflammatory cell infiltration, predominantly macrophages and B cells, in the kidneys ( Figure 8J and Supplemental Figure 11). Taken together, our results show that ROS-deficient pDCs are functionally hyperactive and promote lupus development.

Discussion
The association between NOX2-derived ROS deficiency and autoimmunity was first observed in CGD patients who developed a chronic granulomatous disease with symptoms of colitis, lupus, and arthritis (42). Originally based on positional cloning of a gene polymorphism underlying arthritis in animals (43), genetic studies later identified mutations in NOX2 subunits to be linked to various human autoimmune diseases, including SLE (4). The sequence of the human NCF1 locus is not yet fully clarified due to a large number of variable copies; however, exon sequencing revealed several SNPs (6). Recently, one of the amino acid-replacing SNPs, NCF1-339, leading to an NCF1 allele with lower ROS generation capacity, was discovered in 2 independent studies based on multi-ancestry cohorts to be strongly associated with SLE (7,8). To investigate the role of NOX2-derived ROS in lupus, Ncf1 m1j/m1j mice carrying a spontaneous mutation in the Ncf1 gene served as a scientific tool. Reduced formation of NETs was observed in pristane-injected Ncf1 m1j/m1j mice, which indicates that NOX2-derived ROS protect against lupus through NETs (11). Similar results have been obtained using samples from patients with SLE and NCF1-339 SNP carriers (14,44). Another mechanism was suggested by Hahn et al. that impaired phagocytosis of apoptotic cells by Ncf1 m1j/m1j monocytes and neutrophils may activate the immune system with undigested apoptotic materials in Ncf1 m1j/m1j mice (14). Consistent with this hypothesis, Geng and colleagues showed in NCF1-339 knockin mice that lower ROS production leads to reduced phagosomal acidification, thereby impairing proteolysis and efferocytosis of macrophages, which promotes follicular T helper 2 responses (9). Similarly, Meng et al. showed that ROS deficiency in mice carrying NCF1-339 leads to lower endosomal pH and enhanced TLR signaling in pDCs, which in turn results in increased type I IFN secretion and exacerbation of imiquimod-induced lupus (12).
In the current study, by using a set of cell-specific Cre-deleter and transgenic mouse strains, we systematically screened and were able to identify pDCs as the major immune cell type responsible for lupus exacerbation due to NCF1 deficiency. NCF1-dependent ROS specifically restored in these cells protected mice from lupus in both PIL and a Yaa-accelerating spontaneous model. Apart from previously proposed mechanisms contributing to lupus susceptibility in NCF1-deficient mice, we suggest another mechanism that intrinsic ROS produced by pDCs limit AKT/mTOR-dependent pDC generation, CCR2-mediated tissue accumulation, and pDC hyperactivation through type I IFN signaling, thereby protecting against lupus development.
Abnormally high serum level of IFN-α has long been documented in studies of SLE (45). Its correlation to lupus severity and activity makes it a marker of immune activation in the disease. PDCs, identified as major IFN-α-producing cells, were therefore carefully investigated in SLE patients (46). Increased frequency of pDCs was reported in PBMCs from lupus patients after stimulation by virus or combination of GM-CSF and IFN-α/γ in vitro (46). However, the number of circulating pDCs was lower in SLE patients (47). The finding of pDCs in tissues, such as skin and lymph nodes, from patients with SLE may explain the decreased number in PBMCs (48). Similarly, more pDCs were found in lymphoid tissues from lupusprone mice (49). Although ablation of pDCs in various models has proved their critical role in lupus development, few studies addressed mechanisms behind their accumulation in the disease (50)(51)(52)(53)(54).
In this study, by comparing the frequency and numbers of pDCs in ROS-deficient and -sufficient mice at different stages of lupus, we found that pDCs accumulated in multiple organs and correlated with local type I IFN signature as well as disease severity. This is regulated by pDC-intrinsic ROS and starts already at the initial stage of PIL, concomitant with enhanced pDC-poiesis. A recent study revealed transcriptional reprogramming in hematopoietic cells skewing toward the myeloid lineage in lupus mice and SLE patients (55). In line with this, we showed more pDCs developed from myeloid progenitors than those from lymphoid progenitors, which was, more importantly, promoted by ROS deficiency due to the Ncf1-339 SNP. Redox regulation of hematopoiesis is complex. A low level of ROS is required for the maintenance of quiescence status and the differentiation potential of hematopoietic cells, while an increase in intrinsic ROS production is necessary for activation of differentiation-related pathways, including AKT/mTOR (56). ROS have been reported to exert inhibitory or stimulatory effects in both hematopoietic process and the AKT pathway, depending on the sources of ROS, types of cells, and disease contexts (57)(58)(59). Here, we show that NCF1-dependent ROS at a physiological level limit development of pDCs in lupus through inhibition of the AKT/mTOR signaling. Consistently, the RNA-Seq analysis by Meng Furthermore, we observed CCR2 upregulation on NCF1-deficient pDCs, which may favor their migration and accumulation during lupus. Although the underlying mechanism is unclear, this may be mediated by the AKT pathway since it has been suggested that an AKT-dependent positive feedback loop between β-catenin and CCR2 sustains high expression of CCR2 (60). Nevertheless, very little is known regarding the role of ROS in expression of CCR2 and other chemokine receptors. On the other hand, the involvement of ROS in chemokine production has been demonstrated (61). RNA profiling data based on pDCs from mice carrying the human NCF1-339 SNP revealed upregulation of various chemokines including CCL3, CCL4, CXCL9, CXCL10, and CXCL11 via the NF-κB signaling pathway (12).
It is well established that pDCs produce IFN-α upon stimulation by nucleic acid-containing materials from various sources, including apoptotic/necrotic cells, pathogens, and NETs (62), effects most likely triggered through Fc-γ receptors (FcγRs) and PRRs. Apoptotic cell-derived materials, in the form of ICs with autoantibodies, have been shown to trigger IFN-α production in pDCs via FcγRIIa and contribute to lupus development (63). Upon FcγR-mediated uptake, nucleic acids are sensed by the endosomal PRRs, TLR7 and TLR9, leading to type I IFN production through IRF7 signal cascade (64). The pathogenic role of TLR7 and the regulatory role of TLR9 in pDCs have been demonstrated in spontaneous murine models of lupus and PIL (65,66). Recently, stimulation of another PRR, cyclic GMP-AMP synthase, by cytosolic DNA, was found to trigger IFN-α production in pDCs from patients with SLE and differentiation of pDCs in PIL, through activation of the STING pathway (39,67). Moreover, pDCs can produce IFN-α through an autocrine/paracrine loop (68). In the present study, the difference between IFN-α secretion in ROS-sufficient and -deficient pDCs was seen after treatment with STING agonists, not TLR7/9 agonists or IFN-α. On the contrary, Meng et al. found that pDCs from NCF1-339-carrying mice secreted higher levels of IFN-α upon TLR7/9 activation (12). As Meng's study demonstrated, NCF1-339 enhanced TLR signaling by lowering the pH values in late endosomes/lysosomes where TLRs locate. Since NOX2-derived ROS are known to be important in lysosome biogenesis (69), we suggest that the inconsistency between our and Meng's data can be due to molecular differences between Ncf1 m1j mutation and NCF1-339 SNP. Ncf1 m1j entirely blocks NOX2-derived ROS production because of aberrant NCF1 protein expression, which results in reduced lysosome formation, thereby counterbalancing the effects of endosomal acidification on TLR-dependent IFN-α secretion. In contrast, NCF1-339 only slightly impairs NCF1 translocation to endosomes, leading to low endosomal pH, which affects TLR cleavage and the TLR signaling (12). Nevertheless, both studies demonstrate high IFN-α production in ROS-deficient pDCs, consistent with the observation that pDCs with oxidative stress-related gene signatures were too exhausted to produce IFN-α in patients with SLE (47).
In the current study, we demonstrated that ROS deficiency enhances the IFN-α/JAK1/STAT1 pathway in Ncf1 m1j/m1j pDCs. This is in agreement with the previous finding that the genes related to the JAK/STAT pathway were enriched in the RNA-Seq analysis comparing R848-stimulated pDCs with and without the NCF1-339 SNP (12). Of importance, here we showed that hyperactivated IFN-α/ JAK1/STAT1 signaling, in concert with increased STING-dependent IFN-α production in Ncf1 m1j/m1j pDCs, acting as a positive feedback loop, augments type I IFN responses, which may be crucial to lupus exacerbation driven by ROS deficiency. This notion is strongly supported by the experiment showing that transfer of Ncf1 m1j/m1j pDCs aggravates PIL.
Taken together, in this study, for the first time to our knowledge, we demonstrated that NOX2-derived ROS in pDCs protect against lupus development. This opens possibilities for treatment of SLE at least in patients carrying relevant genetic variants.
PIL. Mice were injected intraperitoneally with a single dose of 500 μL pristane (MilliporeSigma, P2870) and followed up for 5-7 months. PIL arthritis was monitored using a macroscopic scoring system as described before (10). Briefly, 5 points were given to each visibly inflamed (erythema and swelling) ankle or wrist and 1 point to each inflamed toe. Histopathological evaluation on the ankle joints collected at the endpoint (5 MPI) by H&E staining was also performed (72).
Analysis of renal injury. Proteinuria was determined with semiquantitative urine testing strips (Uristix, 2857) using midstream urine. For immunofluorescence, cryosections of kidneys were examined by a Zeiss LSM800 Microscope or an Olympus IX73 Inverted Microscope after being stained with Alexa Fluor 488-conjugated goat anti-mouse IgG specific for Fcγ fragment (Jackson ImmunoResearch, 115-545-071). Deposits were semiquantitatively graded on a scale of 0-3 (0 = none; 1 = weak; 2 = moderate; 3 = strong) according to the intensity by 2 individual scorers in a masked manner. At least 10 glomeruli/section were analyzed.
Analysis of gene and protein expression. Frozen kidneys were homogenized in TRIzol Reagent (Thermo Fisher Scientific, 15596026). According to manufacturer's instructions, RNA from kidney homogenates and PECs was extracted with PureLink RNA Mini Kit (Thermo Fisher Scientific, 12183018A), and the concentration was determined by Thermo Fisher Scientific NanoDrop 2000 Spectrophotometers. CDNA was synthesized with iScript cDNA Synthesis Kit (Bio-Rad, 1708891). Gene expression was detected by Bio-Rad CFX96 Touch Real-Time PCR Detection System with iQ SYBR Green Supermix (Bio-Rad, 1708882). Results were expressed as Ct values normalized to β-Actin or Gapdh. Protein from kidney homogenates and cells was extracted and quantified by Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, 23225). Samples were subjected to electrophoresis and transferred to nitrocellulose membranes (Thermo Fisher Scientific, LC2001). Membranes were incubated with antibodies against STAT1 (clone: D1K9Y, Cell Signaling Technology), p-STAT1 Tyr701 (clone: D4A7, Cell Signaling Technology), p-AKT Thr308 (clone: 244F9, Cell Signaling Technology), p-mTOR Ser2448 (polyclonal, Cell Signaling Technology), and cyclophilin A (Thermo Fisher Scientific, PA1-025) at 1:1,000 dilution, followed by incubation with HRP-conjugated goat anti-rabbit IgG (Thermo Fisher Scientific, 31460) at 1:20,000 dilution. The Bio-Rad ChemiDoc XRS+ Gel imaging system was used for chemiluminescence detection. All the full Western blots are shown in the online supplemental material.
ROS measurement. Intracellular and extracellular oxidative burst were measured according to the protocols described before (7). Briefly, for intracellular ROS detection, cells were incubated with 3 μM DHR (Thermo Fisher Scientific, D23806) for 10 minutes after staining of cell surface markers, followed by stimulation with 200 ng/mL PMA (MilliporeSigma, P8139) for 20 minutes at 37°C. Fluorescence emitted from DHR after oxidation by ROS was detected by flow cytometry, and data were presented as relative geometric MFI of DHR in comparison with solvent control. For extracellular ROS measurement, neutrophils were magnetically isolated from BM with biotinylated anti-Ly6G mAb (BioLegend, clone: 1A8) according to manufacturer's instructions from Miltenyi Biotec. The isolated cells were then diluted in Hanks' balanced salt solution (HBSS; Thermo Fisher Scientific, 14025050) at a concentration of 10 × 10 6 cells/mL, and 50 μL/well was added to a white 96-well plate (Thermo Fisher Scientific, 136101). Cells were mixed with 50 μL/well of HBSS with 350 μg/mL isoluminol (MilliporeSigma, A8264), 3.5 U/mL HRP (MilliporeSigma, P8250), and 200 ng/mL PMA. Luminescence signal was detected immediately and followed at 37°C using BioTek Synergy 2.
In vitro pDC development. BM cells were flushed from 2 femurs and 2 tibias per mouse. After lysis of red blood cells with ammonium-chloride-potassium buffer for 1 minute at room temperature, cells were plated at 2 × 10 6 cells/mL in RPMI complete media containing 200 ng/mL FLT3L (BioLegend, 550704) for pDC differentiation. Cells were collected on day 7-9.
Transfer of pDCs. PDCs were derived from BM from Balb/c or Balb/c.Ncf1 m1j/m1j mice as described above. Balb/c mice were administrated intraperitoneally with the first injection of 5 × 10 5 pDCs at 24 hours after pristane injection. Afterward, 5 × 10 5 pDCs were injected every week intraperitoneally another 3 times. At 10 weeks after pristane injection, the mice were boosted with 5 × 10 5 pDCs intravenously.
PISA assay. PDCs were derived from BM and isolated with biotinylated anti-B220 mAb as described above. A total of 5 × 10 6 pDCs were plated in 6-well cell culture plates (Sarstedt, 83.3920) and stimulated with 500 U/mL IFN-α for 20 hours. For PISA assay, we followed our previously published protocol (31). After treatment, pDCs were collected by centrifugation at 350g and 4°C for 5 minutes, washed twice with PBS, and then resuspended in approximately 300 μL PBS. The cell suspension from each replicate was aliquoted into 10 in 96-well plates and heated in an Eppendorf gradient thermocycler (Mastercycler X50) in the temperature range of 48°C-59°C for 3 minutes. Samples were cooled for 3 minutes at room temperature and afterward snap-frozen with liquid nitrogen. The aliquots from each replicate were pooled and subjected to 4 more freeze-thaw cycles. The samples were then transferred into polycarbonate thick-wall tubes and centrifuged at 100,000g and 4°C for 20 minutes. The soluble protein fraction was then collected, protein concentration was measured using BCA assay, and 25 μg of each sample was taken. Dithiothreitol (DTT) was added to a final concentration of 10 mM, and samples were incubated for 45 minutes at room temperature. Subsequently, iodoacetamide was added to a final concentration of 50 mM, and samples were incubated at room temperature for 1 hour in the dark. The reaction was quenched by adding an additional 10 mM of DTT. Proteins were precipitated by methanol/chloroform and resuspended in 20 mM 4-(2-Hydroxyethyl)-1-piperazinepropanesulfonic acid (EPPS) buffer pH 8.5 with 8 M urea. Urea was diluted to 4 M by adding 20 mM EPPS. Lysyl endopeptidase (LysC; Wako) was added at a 1:75 w/w ratio at room temperature overnight. Samples were diluted with 20 mM EPPS to a final urea concentration of 1 M, and trypsin was added at a 1:75 w/w ratio, followed by 6-hour incubation at room temperature.
For profiling expression, we followed our previously published protocol (74). Cells were lysed with 1% SDS, 8 M urea, and Tris buffer pH 8.0 plus protease inhibitor. The cell lysates were subjected to 1 minute of sonication using Branson probe sonicator with 3-second on/off pulses and 30% amplitude. The concentration of proteins was measured using BCA assay, and the volume corresponding to 25 μg of protein was transferred from each sample to new tubes. The rest of the protocol is similar to above.
Acetonitrile was added to a final concentration of 20% and 200 μg Tandem Mass Tag (TMT) reagents was added to each PISA or expression sample, followed by incubation for 2 hours at room temperature. The reaction was quenched by addition of 0.5% hydroxylamine. PISA and expression samples were pooled, acidified by trifluoroacetic acid, cleaned using Sep-Pak cartridges (Waters), and dried using DNA 120 SpeedVac Concentrator (Thermo Fisher Scientific). The TMT multiplexing scheme is shown in Table 1.
Data processing. The raw liquid chromatography-tandem mass spectrometry data were analyzed by Pro-teomeDiscoverer 2.4. The Mascot search engine matched MS/MS data against the UniProt complete proteome database (human, version UP000005640_9606, 92,957 entries), unless otherwise specified. Ncf1 m1j encoding sequence was also added to the above fasta file. Trypsin/P was selected as enzyme specificity. No more than 2 missed cleavages were allowed. A 1% false discovery rate was used as a filter at both protein and peptide levels. For all other parameters, the default settings were used. TMTpro16 (channels 1-10 were used in the current study) was used for peptide quantification. Cysteine carbamidomethylation was set as a fixed modification, while methionine oxidation was selected as a variable modification. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD035964.
Statistics. For proteomics data, PISA experiments were performed in 2 replicates and expression experiments in 3 replicates. After removing the contaminants, protein intensities were normalized to the total protein intensity in each TMT channel. Subsequently, PISA data for each protein were normalized by corresponding expression channels (Ncf1 m1j/m1j or WT littermates), and 6 ratios were obtained for Ncf1 m1j/m1j and WT PISA samples. P values for the potential target proteins were calculated by t test based on fully normalized intensities between Ncf1 m1j/m1j and WT littermates. Two-sided Student's t test was applied to calculate P values. No peptides from the mutation site were quantified; therefore, for unbiased comparison, we removed the 2 peptides mapping to the sequence around this site in WT samples and recalculated the protein abundances in PISA and expression samples based on the other 12 peptides (there was only negligible change).
Statistical analysis was performed with GraphPad Prism version 8.0.1 by 2-tailed Mann-Whitney nonparametric test, Spearman's correlation, or 1-way or 2-way ANOVA as specified in figure legends. P < 0.05 was considered significant. Data points represent measurement values of biological replicates. All results are shown as mean ± SEM.
Study approval. Stockholm Animal Ethics Committee, Stockholm, Sweden, 4516-2017, approved all the animal experiments. The columns correspond to the TMT labels 1-10.